DESIGN AND EXECUTION OF EARLY PHASE …...Group on Adaptive Designs in Clinical Drug Development. J...

12
DESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALS Maximising R&D Efficiency and Productivity An Aptiv Solutions White Paper

Transcript of DESIGN AND EXECUTION OF EARLY PHASE …...Group on Adaptive Designs in Clinical Drug Development. J...

Page 1: DESIGN AND EXECUTION OF EARLY PHASE …...Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs

DESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALSMaximising R&D Efficiency and Productivity

An Aptiv Solutions White Paper

Page 2: DESIGN AND EXECUTION OF EARLY PHASE …...Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs

DESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALSDESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALS

PAGE 1 PAGE 10

EXECUTIVE SUMMARY

�e adoption of an adaptive design strategy across the product development process brings a number of important bene�ts. �ese include enhanced R&D e�ciency, enhanced R&D productivity and, importantly, increased probability of success at phase III. �e opportunity to manage risk starts in the early phase of a development programme, with a diligent view to appropriate dose selection which impacts not only the success of the downstream clinical trial programme, but also manufacturing costs and value analyses.

Adaptive design trials enhance R&D e�ciency by reducing the need to repeat trials that just miss their clinical end-point or fail to identify the e�ective dose response at the �rst attempt. By avoiding the need to run these trials again, signi�cant cost and time savings are achieved. �is is possible through use of adaptive designs that enable additional patients to be added to achieve statistical signi�cance the �rst time around or by allowing a wider dose range to be studied and then picking doses early in the trial that are in the optimum part of the dose response curve. In addition, early stopping of development programmes because a product is ine�ective enables scarce resources to be redeployed in additional trials which may show more promise. All of these factors increase development e�ciency.

Adaptive designs increase R&D productivity by enabling more accurate de�nition of the e�ective dose response at phase II which enables better design of the pivotal phase III programme which in turn increases the probability of success of the overall development programme. A number of phase III trials fail because the dose is either too high and causes unwanted safety issues or is too low to show su�cient e�cacy. Adaptive design trials enable optimized dose selection before the pivotal trial phase is started.

Another opportunity provided by adaptive design is population enrichment where drug response can be

optimised to speci�c patient sub-populations that respond better to treatment. Many phase III studies fail because the overall e�cacy of treatment is diluted as a consequence of the drug being evaluated in the full trial population rather than in the speci�c subset where the drug works best. Adaptive design enables early selection of the appropriate patient population and increases the probability of success.

“Adaptive Design Enables Selection of the Right Dose for the Right Patient Before Commitment to Expensive Phase III Trials”

Phases I and II are critical steps in the clinical development process as this is where important information about the product has to be generated and assessed before the decision is taken to commit to expensive phase III pivotal studies. �is early phase of development is known as the “Learn Phase” and the data generated relates to the e�ective dose-response, the safety pro�le and therapeutic index, appropriate endpoints, and the patient population best treated by the product under evaluation.

Adaptive designs have an important role to play in both phase I and phase II, but it is in the latter phase where they add the greatest value in terms of risk management. Validated methodologies are available to design early phase adaptive trials and this is an area of signi�cant interest across the pharmaceutical industry.

“Getting it Right at Phase II is a Major Objective for Adaptive Design and is the Point at which Phase III Success is Defined”

Companies that adopt a comprehensive adaptive design strategy in the “Learn” phase will make better development decisions, will build development e�ciency, will increase the probability of success of the molecules that enter phase III, ultimately bringing e�ective products to the market more e�ciently. Competitor companies that persevere with conventional trial designs will remain slow, ine�cient

candidate optimisation to Proof of concept stages. Moira assists partner companies with both acquisition and due diligence of in licensed compounds and strategic support for the out-licensing of successful candidates. �is has included design, planning and management of, strategic and clinical development plans, due diligence, decision analysis, preclinical, POC and Phase II programmes in a range of, indications including respiratory, anti-infective, immunomodulation, CNS and cardiovascular. �is experience covers a large range of targets and mecha-nisms of action. In order to support the strategic and clinical development planning Moira has worked extensively with international Key Opinion Leaders in respiratory, anti-infective and transplant indications.

PHIL BIRCH,SVP, CORPORATE BRAND MANAGER

Dr. Phil Birch is a Senior Vice President at Aptiv Solutions and has global responsibility for co-ordinating market education and client awareness in the adaptive clinical trials �eld. Dr. Birch has worked in the pharmaceutical industry for over 27 years and has held executive positions in corporate development, business development and R&D in a number of consulting, biotechnology and top 10 pharmaceutical companies.

REFERENCES

1. Dragalin V. Adaptive Designs: Terminology and Classi�cation. Drug Information Journal, 2006; 40: 425-436. 2. Gallo P, Chuang-Stein C, Dragalin V, Gaydos B, Krams M, Pinheiro J. Executive Summary of the PhRMA Working Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs for Phase I Oncology Trials, Clinical Operations Summit, Brussels, 2012).4. Dragalin V, Hsuan F, and Padmanabhan SK. Adaptive Designs for Dose-Finding Studies Based on Sigmoid Emax Model. J Biopharmaceutical Statistics 2007; 17:1051-1070.5. Smith MK, Jones I, Morris MF, Grieve A and Tan K. Implementation of a Bayesian adaptive design in a proof of concept study. Pharmaceutical Statistics 2006; 5:39-50. 6. Bornkamp B, Bretz F, Dmitrienko A, Enas G, Gaydos B, Hsu CH, König F, Krams M, Liu Q, Neuenschwander B, Parke T, Pinheiro J, Roy A, Sax R, Shen F. Innovative approaches for designing and analyzing adaptive dose-ranging trials: White Paper from the PhRMA working group on “Adaptive Dose-Ranging Studies”. J of Biopharmaceutical Statistics 2007; 17: 965–995. 7. Dragalin V, Bornkamp B, Bretz F, Miller F, Padmanabhan SK, Patel N, Perevozskaya I, Pinheiro J, and Smith J. A Simulation Study to Compare New Adaptive Dose–Ranging Designs. Statistics in Biopharmaceutical Research 2010; 2:487-512.8. Orlo� J, Douglas F, Pinheiro J, Levinson S, Branson M, Chaturvedi P, Ette E, Gallo P, Hirsch G, Mehta C, Patel N, Sabir S, Springs S, Stanski D, Golub H, Evers M, Fleming E, Singh N, Tramontin T. �e future of drug development: advancing clinical trial design. Nature Review-Drug Discovery. 2009; 8:940-957.9. Padmanabhan SK and Dragalin V. Adaptive Dc-optimal designs for dose �nding based on a continuous e�cacy endpoint. Biometrical Journal 2010; 52:836-852. 10. Grieve AP and Krams M. ASTIN: a Bayesian adaptive dose-response trial in acute stroke. Clinical Trials 2005; 2:340-351. 11. Bornkamp B, Pinheiro J, Bretz F. MCPMod: An R Package for the Design and Analysis of Dose-Finding Studies. Journal of Statistical Software 2009; 29:1-23.12. Shen J, Preskorn S, Dragalin V, Slomkowski M, Padmanabhan SK, Fardipour P, Sharma A, and Krams M. How Adaptive Trial Designs Can Increase E�ciency in Psychiatric Drug Development: A Case Study. Innovations in Clinical Neuroscience 2011; 8: 26-34.13. Berry SM, Spinelli W, Littman GS, Liang JZ, Fardipour P, Berry DA, Lewis RL, and Krams M. A Bayesian dose-�nding trial with adaptive dose expansion to �exibly assess e�cacy and safety of an investigational drug. Clinical Trials 2010; 7: 121-135.

ABOUT THE AUTHORS

VLAD DRAGALIN, SVP, INNOVATION CENTRE

Dr. Vlad Dragalin is a well-known adaptive design expert with 25 years of experience in developing the statistical methodology of adaptive designs and with over 12 years experience in pharmaceutical industry. Vlad is a Senior Vice President, Software Develop-ment and Consulting, in the Innovation Center at Aptiv Solutions. Vlad works with clients to improve the drug development process through the use of adaptive designs and to assist them with the trial design, simulation and execution, developing adaptive trials software tools, and working with project teams in clinical operations, data management and statistics during study conduct.

SARAH ARBE-BARNES,SVP, TRANSLATIONAL SCIENCES

Dr. Sarah Arbe-Barnes has over 24 years of experience in drug development and project leadership in the Pharma and service sector industry to include in/out-licensing and due diligence, strategic development and business planning, fundraising for start-ups/SMEs( not-for-pro�t sector and VC), along with optimized market access strategies and marketing. Sarah has led recent product approval programmes, overseeing a broad range of clinical and nonclinical project design and management activities from the point of candidate selection. Sarah heads the Transla-tional Sciences division of Aptiv Solutions, which is a dedicated global consultancy group, focused on the provision of drug and device development advice and expertise for small molecules and biologics.

MOIRA THOMSON,VP, TRANSLATIONAL SCIENCES

Moira �omson has over 15 years of experience in the Pharmaceutical Industry, and is involved in the design, planning and management of global develop-ment projects for a mixture of small and medium sized Pharma, and Biotech companies covering lead

and will fail to make the commercial returns that investors and shareholders demand. “Adaptive Design Builds Competitive Advantage”

�is White Paper summarizes the important aspects of early phase adaptive design and provides a summary of selected case studies which demonstrate the value of this innovative approach. It has been written to inform senior R&D decision-makers about the critical role of adaptive design in early phase development and the signi�cant bene�ts that this approach can bring.

BACKGROUND TO EARLY PHASE ADAPTIVE TRIAL DESIGN

Adaptive Clinical Trials are viewed as supporting a change in the development process as focus moves from blockbuster products to more specialized medicines that require a �exible cost-e�cient approach to investigating their e�ectiveness. Adaptive designs are statistical methodologies applied to speci�c stages of drug development where real time learning from accumulating trial data is applied to optimize subsequent study execution.1 �e bene�t of an Adaptive Clinical Trial lies in the monitoring of data to make design modi�cation adjustments to an aspect of a trial leading to quicker development decisions around key go/no-go milestones that impact time to market and/or minimise costly R&D losses.

�e adaptations are prospectively de�ned prior to the start of the trial and can include stopping early either for futility or success, expanding the sample size due to greater than expected data variability, or allocating patients preferentially to treatment regimens with a better therapeutic index. Virtually any aspect of the trial may be the potential target of design modi�cations. Importantly, these modi�cations are a design feature aimed to enhance the trial, not a remedy for inadequate planning. �ey are not ad hoc study corrections made via protocol amendment.2

�e speci�c adaptations considered, and the basis of their implementation, are carefully de�ned based on strict rules, justi�ed statistically and scienti�cally, and are an integral part of the �nal, pre-recruitment trial protocol. Failing this would compromise the interpretation and acceptance of study results.

Adaptive designs often employ frequent interim analyses of all accumulated data (and, possibly, external trial data) to determine whether pre-planned design modi�cations will be ‘triggered’. Interim analyses partition the trial into multiple stages, each trial stage’s characteristics (number of arms, number of patients to be enrolled, their allocation between arms, stage duration, etc.) de�ned by the preceding interim analysis results. �e ability to periodically, or even continually, examine available data to determine whether trial modi�cations are necessary and implement pre-de�ned design changes when indicated, gives adaptive design its strength and �exibility.

�e use of adaptive design in the exploratory stage of clinical trials can increase the e�ciency of drug development by improving our ability to e�ciently learn about the dose-response and better determine whether to take a drug (and the right dose of the drug and in the right population) forward into later phase testing. �ese designs explicitly address multiple trial goals, adaptively allocate subjects according to ongoing information needs, and allow termination for both early success and futility considerations. �is approach can maximize the ability to test a larger number of doses in a single trial while simultaneously increasing the e�ciency of the trial in terms of making better go/no-go decisions about continuing the trial and/or the development of the drug for a speci�c indication or sub-population. During the “Learn” phase many aspects of the study design can be modi�ed: number of subjects, study duration, endpoint selection, treatment duration, number of treatments, patient population. �e adaptation of multiple trial features that are a legitimate concern in con�rmatory studies can become, in fact, an

appealing feature of adaptive designs in exploratory development, leading to enhanced learning to set the appropriate stage for con�rmatory trials.

“The Use of Adaptive Design in the Learn Phase Offers the Best Opportunity to Assess the Product Characteristics that Determine Success”

Judicious use of adaptive designs may increase the information value per resource unit invested by avoiding allocation of patients to non-e�cacious/unsafe therapies and allowing stopping decisions to be made at the earliest possible time point. Ultimately this will accelerate the development of promising therapies to bene�t patients.

Adaptive Clinical Trials can result in the more ethical treatment of patients from at least two perspectives. First, a larger number of patients can be randomized to more favorable, e�ective doses with fewer patients exposed to less e�ective doses. In addition, there is the potential for including fewer total patients in the early stages of the development programmes with attending lower risk of exposure to adverse events. A greater saving in patients may be obtained at the programme level rather than the individual study level as adaptive methodology applied across a programme will reduce the need to repeat ine�ective or inconclusive studies.

Adaptive designs represent an advanced methodology to support drug development. A prerequisite for their successful implementation is an understanding of the underlying methodology and their impact on the logistics of the trial, such as data management and monitoring procedures, electronic data capture/interactive web response services, drug supply management, and data-monitoring committee operating procedures. �ese designs have an impact on drug development strategies, trial protocols, clinical trial material doses and availability, informed consent forms, data analysis, and reporting plans. Adaptive design will also change the dynamics of

enrollment, randomization, and data capturing process, monitoring, and data cleaning systems. To be able to make informed decisions, the sponsor must have access to real time clinical data from all sources and be able to easily review and assess this data.

�is in turn requires �exibility through:

• Data management systems for rapid clinical data• access and data cleaning

• Drug supply systems for drug product planning • and management

• Optimal systems providing randomization, • medication kit management and emergency• unblinding

Planning of Adaptive Clinical Trials is a key feature in the success of such innovative approaches. Determining the optimal characteristics of the study design can be a complex yet critical decision that may require more upfront planning before the protocol can be �nalized. However, there are several commercial software packages that provide a powerful tool for planning, simulation and analysis of complex adaptive designs. FACTS™ is comprehensive software for adaptive designs in the “Learn” phase and includes early stage dose-escalation designs, adaptive dose-ranging studies, interim decision rules based on both e�cacy and safety responses, and population enrichment. ADDPLAN® is a validated software package for adaptive design in the “Con�rm” phase and incorporates sample size re-estimation, adaptive group sequential designs, multiple comparison procedures for multi-armed adaptive trials, including treatment selection designs, �exible combination of clinical research phases, and population enrichment designs.

TYPES OF ADAPTIVE DESIGNS IN “LEARN” STUDIES

Examples of adaptive designs in the Learn phase are shown in Figure 1. Phase I studies are already

FIGURE 5: Adaptive Design Creates Value at Level of the Single Trial, DevelopmentProgramme and Product Pipeline

�is aspect is discussed further in an accompanying white paper that explores the strategic utilisation of adaptive design at the development programme and product pipeline level. Importantly, recent data from Merck has indicated that adoption of an early phase adaptive trial strategy at this level saves at least $200m annually (Schindler J, Disruptive Innovations Conference, Boston, 2012).

“Adaptive Dose Ranging Studies Build Significant Efficiencies at Phase II and Optimize Identification of the Optimal Dose for Confirmatory Studies”

Several adaptive designs have been proposed and some already applied in practice using di�erent working models for the design engine: parsimonious monotonic sigmoid Emax9, Normal Dynamic Linear Model without monotonicity restrictions10, several simple models combined in a Multiple Comparison Procedure (MCP Mod).11

CASE STUDY

�e objective of this trial was to establish the correct dose(s) to take forward into a con�rmatory trial on a novel anti-psychotic drug in acute schizophrenia. �ere was data indicating that the dose-response might be non-monotonic. �erefore, the working model chosen for the design was the Normal Dynamic Linear Model, because it is �exible and robust, and also allows for capturing non-monotonic relationships. Patients were allocated to placebo, active comparator and seven doses of study drug. �e adaptive allocation was targeting both the minimum e�cacious dose (MED) and the dose achieving the maximum response (Dmax) on the primary endpoint, positive PANSS score at week 4. Weekly data updates were used to change the allocation and check early stopping rules based on predictive probabilities. A stopping rule for futility was based on a prespeci�ed threshold that no treatment dose achieves the clinically signi�cant di�erence (CSD) over placebo. A stopping rule for e�cacy required su�cient knowledge about the MED and Dmax and a prespeci�ed con�dence that the Dmax achieves CSD. For details, see Shen et al.12 �e adaptive design allowed earlier (5 months) decision making with a 99% con�dence in futility decision with $14m savings in direct grant cost. �e design, implementation, and outcome of a response adaptive dose-ranging trial of an investigational drug in patients with diabetes, incorporating a

dose expansion approach to �exibly address bothe�cacy and safety has also been reported recently by Berry et al.13 �e early (2 months) futility decision was achieved with a $3.6m saving in direct cost and additional cost savings in avoiding expensive production of phase III drug supplies.

FIGURE 4: Summary of Early Phase Case Studies

SUMMARY

Early phase adaptive design trials address a number of critical development questions that need to be answered before commitment is made to phase III con�rmatory trials. �e application of these innovative designs early in the product development process enhances clinical trial e�ciency, productivity and increases the probability of success at phase III.

�e deployment of these approaches across a pipeline of products brings substantial bene�ts to sponsor companies and enables more e�ective portfolio management and ultimately increases portfolio value (Figure 5).

adaptive by nature, but more recent adaptive designsprovide better estimate of drug safety (e.g. the maxi-mum tolerated dose) and better understanding of itsPK/PD characteristics. Combining the objectives of conventional Phase IIa and IIb studies in a single trial provides the obvious bene�t of reducing the timeline by running the two studies seamlessly under a single protocol with the same clinical team. Furthermore, such approaches achieve trial e�ciency by the prospective integration of data from both IIa and IIb in the �nal analysis.

FIRST-IN-HUMAN SINGLE ASCENDING DOSE ESCALATION DESIGNS

In contrast to traditional dose escalation designs, where for example, cohorts of subjects are allocated to ascending doses (and placebo) until a maximum tolerated dose (MTD) is empirically determined, the continual reassessment method (CRM) models the dose- and/or exposure-toxicity relationship to focus on the identi�cation of the dose-range of interest near the MTD and for this to be explored further. �ese adaptive dose escalation designs can be expanded to also e�ciently establish proof-of-mechanism or proof-of-target modulation, if validated biomarker endpoints are available.

FIGURE 1: Types of Adaptive Designs in the Learn Phase

�e traditional First in Human study objectives of safety and tolerability can be extended to include an

assessment of proof-of-mechanism in the very �rst study conducted with the investigational compound. �e ability to assess proof-of-mechanism in this earlydevelopmental setting provides an opportunity to greatly enhance the clinical development strategy for the compound. �e adaptation is intended to quickly hone in on the predicted dose-range of interest. �is avoids the risks associated with exposing patients in an ine�ective low dose range, whilst minimizing exposure to high doses with an unacceptable tolerability and safety pro�le.

“CRM Offers a Better Approach to Accurately Determining the MTD in Early Phase Oncology Studies”

CRM models are becoming routine for early phase oncology studies where accurate de�nition of the MTD is critical for selecting the optimum dose for e�cacy assessment. Conventional 3+3 designs have been shown to signi�cantly underestimate the MTD which increases the risk of poor outcome in e�cacy trials.3

SEAMLESS MULTIPLE ASCENDING DOSE AND PROOF-OF-CONCEPT STUDY DESIGNS

�e objective of seamless multiple ascending dose and proof-of-concept studies is to combine a multiple ascending dose (MAD) study in patients with the proof-of-concept of the investigational compound.

CASE STUDY

A good example is seen in the treatment of rheumatoid arthritis. �e study design involved a MAD escalation study, in which patients were randomly assigned only to the lowest treatment group or placebo in a 3:1 ratio (active treatment or placebo) which then converted into a parallel enrolling, proof of concept (POC) dose ranging study (see Figure 2). Consistent with the MAD paradigm, dose escalation (allowing subjects in the next higher dose level to

receive a second dose) occurred only after the safety review on 6 subjects assigned to active treatment in the preceding dose had been performed. �is safetyreview was undertaken by a data monitoring committee. Only after it was determined by the data monitoring committee that two consecutive doses of the experimental treatment were well tolerated at a given dose level, were subjects already enrolled in the next higher dose treatment group permitted to receive a second dose of the investigational product. In Stage 2, a utility combining biomarker observations with early readout of a clinical endpoint at week 4 was used to drop dose levels that did not ful�l prespeci�ed necessary conditions (Proof-of-Non-Viability). Viable cohorts and comparator were kept open until each of the surviving treatment arms reached a prede�ned number of patients. �e endpoint to assess proof-of-concept was a 12 week observation on the regulatory accepted endpoint for Rheumatoid Arthritis (ACR20).

�is adaptive design provides several advantages over more traditional clinical trial approaches, without compromising patient safety:

• The seamless design increased the utility of the• information obtained by allowing subjects in both• the MAD and POC stages to provide both • de�nitive safety and e�cacy data.

• Performing the MAD component in subjects with• rheumatoid arthritis receiving concomitant • methotrexate allowed for the earliest • characterization of safety and antigenicity of the• new compound in the clinically relevant • population.

• The adaptive design minimized the number of• subjects that were exposed to ine�ective doses of the• drug, while simultaneously focusing subjects to• doses that were most informative for accurate dose• selection for subsequent con�rmatory trials.

FIGURE 2: Trial Design for Seamless MAD/POC Case Study �erefore, the adaptive design optimizes the bene�t/risk balance for participating subjects via improved e�ciency of decision making in relation to the doses of the new drug studied. In this example the estimated cost saving through combining two trials into one was $1.2m and the development programme was accelerated by 9 months.

“The Development Programme was Accelerated by 9 Months and the Cost Saving of Combining Two Trials in One was Estimated at $1.2M”

PROOF OF CONCEPT AND DOSE-RANGING STUDY DESIGNS

An example of a proof of concept and dose-ranging study design is a dose-�nding study to evaluate the analgesic e�cacy and safety of a single dose of a new non-steroidal anti-in�ammatory drug in subjects experiencing pain after oral dental surgery.4 Six doses of the new drug, placebo and an active comparator are considered. �e adaptive design consists of three stages. At the �rst stage, subjects are equally randomized into three arms: placebo, active comparator, and the top dose of the new drug. At the �rst interim analysis, the study may be stopped either because of lack of assay sensitivity, i.e. inability to establish a prede�ned minimal bene�t of the comparator over and above placebo, or evidence that

the top dose of study drug could not achieve a minimal acceptable level of total pain relief at 8 hours. If the study continues to the second stage, additional subjects would be randomized to placebo, active control, and three doses of the new drug (low, medium, and high) in a 1:1:2:2:1 ratio. Again, the study could be stopped due to lack of assay sensitivity or futility. Otherwise, a D-optimal design will be used to determine both the doses of the new drug (among six possible) and subject allocation ratios for the third stage, see Figure 3. At the end of the study, a four parameter logistic model is used to �t the dose response and make recommendations about the dose to be taken in phase III.

FIGURE 3: Trial Design for POC and Adaptive Dose-Ranging Case Study

CASE STUDY

A full case study has been published in the pain �eld where a Bayesian adaptive dose ranging design was used to evaluate the e�cacy of a new analgesic drug in a proof-of-concept post-herpetic neuralgia trial.5 In total 7 doses of the drug were evaluated in an attempt to de�ne the e�ective dose response curve. �e maximum sample size was 280 patients, however, at the �rst interim analysis (80 patients) the study was stopped for futility as none of the doses showed any meaningful di�erence from placebo. �e study was formally stopped after 133 patients had been enrolled. Stopping of this trial saved an estimated direct grant cost of $0.76m with a further $1.3m savings in internal costs.

“Early Stopping of POC Trials Because of Lack of Efficacy Saves Significant Cost and Enables Valuable Resources to be Re-assigned to Alternative Programmes That May Have a Better Chance of Success”

A recent unpublished example shows that early stopping of a POC adaptive dose ranging trial in post-operative nausea and vomiting saved $0.5m in outsourced costs. In this example, 131 out of 200 patients had been enrolled when the decision was taken at the �rst interim analysis.

RESPONSE ADAPTIVE DOSE-RANGING STUDY DESIGNS

Recent simulation studies6,7 conducted by the PhRMA Working Group on Adaptive Dose-Ranging Studies have found that response-adaptive dose allocation designs are generally superior in performance to conventional pairwise comparisons approaches. Learning about the research question such as identifying a target dose, or estimating the dose-response, is optimized with these designs.

�e main objectives in an adaptive dose-ranging study are to detect the dose response, to determine if any doses(s) meets clinical relevance, to estimate the dose-response, and then to decide on the dose(s) (if any) to take into the con�rmatory Phase III. With adaptive designs for dose-ranging studies many more doses may be considered than in a conventional parallel group design without increasing sample size. �is is achieved mainly through an adaptive allocation rule that is changing the randomization ratio to di�erent treatment arms during the study and putting more subjects in the region that allow most accurate estimation of dose-response and better precision in selecting the target dose. For example, Orlo� et al.8 showed that “for half the sample size, the adaptive design is as powerful and e�cient as the standard approach”.

Page 3: DESIGN AND EXECUTION OF EARLY PHASE …...Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs

DESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALSDESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALS

PAGE 9 PAGE 2

EXECUTIVE SUMMARY

�e adoption of an adaptive design strategy across the product development process brings a number of important bene�ts. �ese include enhanced R&D e�ciency, enhanced R&D productivity and, importantly, increased probability of success at phase III. �e opportunity to manage risk starts in the early phase of a development programme, with a diligent view to appropriate dose selection which impacts not only the success of the downstream clinical trial programme, but also manufacturing costs and value analyses.

Adaptive design trials enhance R&D e�ciency by reducing the need to repeat trials that just miss their clinical end-point or fail to identify the e�ective dose response at the �rst attempt. By avoiding the need to run these trials again, signi�cant cost and time savings are achieved. �is is possible through use of adaptive designs that enable additional patients to be added to achieve statistical signi�cance the �rst time around or by allowing a wider dose range to be studied and then picking doses early in the trial that are in the optimum part of the dose response curve. In addition, early stopping of development programmes because a product is ine�ective enables scarce resources to be redeployed in additional trials which may show more promise. All of these factors increase development e�ciency.

Adaptive designs increase R&D productivity by enabling more accurate de�nition of the e�ective dose response at phase II which enables better design of the pivotal phase III programme which in turn increases the probability of success of the overall development programme. A number of phase III trials fail because the dose is either too high and causes unwanted safety issues or is too low to show su�cient e�cacy. Adaptive design trials enable optimized dose selection before the pivotal trial phase is started.

Another opportunity provided by adaptive design is population enrichment where drug response can be

optimised to speci�c patient sub-populations that respond better to treatment. Many phase III studies fail because the overall e�cacy of treatment is diluted as a consequence of the drug being evaluated in the full trial population rather than in the speci�c subset where the drug works best. Adaptive design enables early selection of the appropriate patient population and increases the probability of success.

“Adaptive Design Enables Selection of the Right Dose for the Right Patient Before Commitment to Expensive Phase III Trials”

Phases I and II are critical steps in the clinical development process as this is where important information about the product has to be generated and assessed before the decision is taken to commit to expensive phase III pivotal studies. �is early phase of development is known as the “Learn Phase” and the data generated relates to the e�ective dose-response, the safety pro�le and therapeutic index, appropriate endpoints, and the patient population best treated by the product under evaluation.

Adaptive designs have an important role to play in both phase I and phase II, but it is in the latter phase where they add the greatest value in terms of risk management. Validated methodologies are available to design early phase adaptive trials and this is an area of signi�cant interest across the pharmaceutical industry.

“Getting it Right at Phase II is a Major Objective for Adaptive Design and is the Point at which Phase III Success is Defined”

Companies that adopt a comprehensive adaptive design strategy in the “Learn” phase will make better development decisions, will build development e�ciency, will increase the probability of success of the molecules that enter phase III, ultimately bringing e�ective products to the market more e�ciently. Competitor companies that persevere with conventional trial designs will remain slow, ine�cient

candidate optimisation to Proof of concept stages. Moira assists partner companies with both acquisition and due diligence of in licensed compounds and strategic support for the out-licensing of successful candidates. �is has included design, planning and management of, strategic and clinical development plans, due diligence, decision analysis, preclinical, POC and Phase II programmes in a range of, indications including respiratory, anti-infective, immunomodulation, CNS and cardiovascular. �is experience covers a large range of targets and mecha-nisms of action. In order to support the strategic and clinical development planning Moira has worked extensively with international Key Opinion Leaders in respiratory, anti-infective and transplant indications.

PHIL BIRCH,SVP, CORPORATE BRAND MANAGER

Dr. Phil Birch is a Senior Vice President at Aptiv Solutions and has global responsibility for co-ordinating market education and client awareness in the adaptive clinical trials �eld. Dr. Birch has worked in the pharmaceutical industry for over 27 years and has held executive positions in corporate development, business development and R&D in a number of consulting, biotechnology and top 10 pharmaceutical companies.

REFERENCES

1. Dragalin V. Adaptive Designs: Terminology and Classi�cation. Drug Information Journal, 2006; 40: 425-436. 2. Gallo P, Chuang-Stein C, Dragalin V, Gaydos B, Krams M, Pinheiro J. Executive Summary of the PhRMA Working Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs for Phase I Oncology Trials, Clinical Operations Summit, Brussels, 2012).4. Dragalin V, Hsuan F, and Padmanabhan SK. Adaptive Designs for Dose-Finding Studies Based on Sigmoid Emax Model. J Biopharmaceutical Statistics 2007; 17:1051-1070.5. Smith MK, Jones I, Morris MF, Grieve A and Tan K. Implementation of a Bayesian adaptive design in a proof of concept study. Pharmaceutical Statistics 2006; 5:39-50. 6. Bornkamp B, Bretz F, Dmitrienko A, Enas G, Gaydos B, Hsu CH, König F, Krams M, Liu Q, Neuenschwander B, Parke T, Pinheiro J, Roy A, Sax R, Shen F. Innovative approaches for designing and analyzing adaptive dose-ranging trials: White Paper from the PhRMA working group on “Adaptive Dose-Ranging Studies”. J of Biopharmaceutical Statistics 2007; 17: 965–995. 7. Dragalin V, Bornkamp B, Bretz F, Miller F, Padmanabhan SK, Patel N, Perevozskaya I, Pinheiro J, and Smith J. A Simulation Study to Compare New Adaptive Dose–Ranging Designs. Statistics in Biopharmaceutical Research 2010; 2:487-512.8. Orlo� J, Douglas F, Pinheiro J, Levinson S, Branson M, Chaturvedi P, Ette E, Gallo P, Hirsch G, Mehta C, Patel N, Sabir S, Springs S, Stanski D, Golub H, Evers M, Fleming E, Singh N, Tramontin T. �e future of drug development: advancing clinical trial design. Nature Review-Drug Discovery. 2009; 8:940-957.9. Padmanabhan SK and Dragalin V. Adaptive Dc-optimal designs for dose �nding based on a continuous e�cacy endpoint. Biometrical Journal 2010; 52:836-852. 10. Grieve AP and Krams M. ASTIN: a Bayesian adaptive dose-response trial in acute stroke. Clinical Trials 2005; 2:340-351. 11. Bornkamp B, Pinheiro J, Bretz F. MCPMod: An R Package for the Design and Analysis of Dose-Finding Studies. Journal of Statistical Software 2009; 29:1-23.12. Shen J, Preskorn S, Dragalin V, Slomkowski M, Padmanabhan SK, Fardipour P, Sharma A, and Krams M. How Adaptive Trial Designs Can Increase E�ciency in Psychiatric Drug Development: A Case Study. Innovations in Clinical Neuroscience 2011; 8: 26-34.13. Berry SM, Spinelli W, Littman GS, Liang JZ, Fardipour P, Berry DA, Lewis RL, and Krams M. A Bayesian dose-�nding trial with adaptive dose expansion to �exibly assess e�cacy and safety of an investigational drug. Clinical Trials 2010; 7: 121-135.

ABOUT THE AUTHORS

VLAD DRAGALIN, SVP, INNOVATION CENTRE

Dr. Vlad Dragalin is a well-known adaptive design expert with 25 years of experience in developing the statistical methodology of adaptive designs and with over 12 years experience in pharmaceutical industry. Vlad is a Senior Vice President, Software Develop-ment and Consulting, in the Innovation Center at Aptiv Solutions. Vlad works with clients to improve the drug development process through the use of adaptive designs and to assist them with the trial design, simulation and execution, developing adaptive trials software tools, and working with project teams in clinical operations, data management and statistics during study conduct.

SARAH ARBE-BARNES,SVP, TRANSLATIONAL SCIENCES

Dr. Sarah Arbe-Barnes has over 24 years of experience in drug development and project leadership in the Pharma and service sector industry to include in/out-licensing and due diligence, strategic development and business planning, fundraising for start-ups/SMEs( not-for-pro�t sector and VC), along with optimized market access strategies and marketing. Sarah has led recent product approval programmes, overseeing a broad range of clinical and nonclinical project design and management activities from the point of candidate selection. Sarah heads the Transla-tional Sciences division of Aptiv Solutions, which is a dedicated global consultancy group, focused on the provision of drug and device development advice and expertise for small molecules and biologics.

MOIRA THOMSON,VP, TRANSLATIONAL SCIENCES

Moira �omson has over 15 years of experience in the Pharmaceutical Industry, and is involved in the design, planning and management of global develop-ment projects for a mixture of small and medium sized Pharma, and Biotech companies covering lead

and will fail to make the commercial returns that investors and shareholders demand. “Adaptive Design Builds Competitive Advantage”

�is White Paper summarizes the important aspects of early phase adaptive design and provides a summary of selected case studies which demonstrate the value of this innovative approach. It has been written to inform senior R&D decision-makers about the critical role of adaptive design in early phase development and the signi�cant bene�ts that this approach can bring.

BACKGROUND TO EARLY PHASE ADAPTIVE TRIAL DESIGN

Adaptive Clinical Trials are viewed as supporting a change in the development process as focus moves from blockbuster products to more specialized medicines that require a �exible cost-e�cient approach to investigating their e�ectiveness. Adaptive designs are statistical methodologies applied to speci�c stages of drug development where real time learning from accumulating trial data is applied to optimize subsequent study execution.1 �e bene�t of an Adaptive Clinical Trial lies in the monitoring of data to make design modi�cation adjustments to an aspect of a trial leading to quicker development decisions around key go/no-go milestones that impact time to market and/or minimise costly R&D losses.

�e adaptations are prospectively de�ned prior to the start of the trial and can include stopping early either for futility or success, expanding the sample size due to greater than expected data variability, or allocating patients preferentially to treatment regimens with a better therapeutic index. Virtually any aspect of the trial may be the potential target of design modi�cations. Importantly, these modi�cations are a design feature aimed to enhance the trial, not a remedy for inadequate planning. �ey are not ad hoc study corrections made via protocol amendment.2

�e speci�c adaptations considered, and the basis of their implementation, are carefully de�ned based on strict rules, justi�ed statistically and scienti�cally, and are an integral part of the �nal, pre-recruitment trial protocol. Failing this would compromise the interpretation and acceptance of study results.

Adaptive designs often employ frequent interim analyses of all accumulated data (and, possibly, external trial data) to determine whether pre-planned design modi�cations will be ‘triggered’. Interim analyses partition the trial into multiple stages, each trial stage’s characteristics (number of arms, number of patients to be enrolled, their allocation between arms, stage duration, etc.) de�ned by the preceding interim analysis results. �e ability to periodically, or even continually, examine available data to determine whether trial modi�cations are necessary and implement pre-de�ned design changes when indicated, gives adaptive design its strength and �exibility.

�e use of adaptive design in the exploratory stage of clinical trials can increase the e�ciency of drug development by improving our ability to e�ciently learn about the dose-response and better determine whether to take a drug (and the right dose of the drug and in the right population) forward into later phase testing. �ese designs explicitly address multiple trial goals, adaptively allocate subjects according to ongoing information needs, and allow termination for both early success and futility considerations. �is approach can maximize the ability to test a larger number of doses in a single trial while simultaneously increasing the e�ciency of the trial in terms of making better go/no-go decisions about continuing the trial and/or the development of the drug for a speci�c indication or sub-population. During the “Learn” phase many aspects of the study design can be modi�ed: number of subjects, study duration, endpoint selection, treatment duration, number of treatments, patient population. �e adaptation of multiple trial features that are a legitimate concern in con�rmatory studies can become, in fact, an

appealing feature of adaptive designs in exploratory development, leading to enhanced learning to set the appropriate stage for con�rmatory trials.

“The Use of Adaptive Design in the Learn Phase Offers the Best Opportunity to Assess the Product Characteristics that Determine Success”

Judicious use of adaptive designs may increase the information value per resource unit invested by avoiding allocation of patients to non-e�cacious/unsafe therapies and allowing stopping decisions to be made at the earliest possible time point. Ultimately this will accelerate the development of promising therapies to bene�t patients.

Adaptive Clinical Trials can result in the more ethical treatment of patients from at least two perspectives. First, a larger number of patients can be randomized to more favorable, e�ective doses with fewer patients exposed to less e�ective doses. In addition, there is the potential for including fewer total patients in the early stages of the development programmes with attending lower risk of exposure to adverse events. A greater saving in patients may be obtained at the programme level rather than the individual study level as adaptive methodology applied across a programme will reduce the need to repeat ine�ective or inconclusive studies.

Adaptive designs represent an advanced methodology to support drug development. A prerequisite for their successful implementation is an understanding of the underlying methodology and their impact on the logistics of the trial, such as data management and monitoring procedures, electronic data capture/interactive web response services, drug supply management, and data-monitoring committee operating procedures. �ese designs have an impact on drug development strategies, trial protocols, clinical trial material doses and availability, informed consent forms, data analysis, and reporting plans. Adaptive design will also change the dynamics of

enrollment, randomization, and data capturing process, monitoring, and data cleaning systems. To be able to make informed decisions, the sponsor must have access to real time clinical data from all sources and be able to easily review and assess this data.

�is in turn requires �exibility through:

• Data management systems for rapid clinical data• access and data cleaning

• Drug supply systems for drug product planning • and management

• Optimal systems providing randomization, • medication kit management and emergency• unblinding

Planning of Adaptive Clinical Trials is a key feature in the success of such innovative approaches. Determining the optimal characteristics of the study design can be a complex yet critical decision that may require more upfront planning before the protocol can be �nalized. However, there are several commercial software packages that provide a powerful tool for planning, simulation and analysis of complex adaptive designs. FACTS™ is comprehensive software for adaptive designs in the “Learn” phase and includes early stage dose-escalation designs, adaptive dose-ranging studies, interim decision rules based on both e�cacy and safety responses, and population enrichment. ADDPLAN® is a validated software package for adaptive design in the “Con�rm” phase and incorporates sample size re-estimation, adaptive group sequential designs, multiple comparison procedures for multi-armed adaptive trials, including treatment selection designs, �exible combination of clinical research phases, and population enrichment designs.

TYPES OF ADAPTIVE DESIGNS IN “LEARN” STUDIES

Examples of adaptive designs in the Learn phase are shown in Figure 1. Phase I studies are already

FIGURE 5: Adaptive Design Creates Value at Level of the Single Trial, DevelopmentProgramme and Product Pipeline

�is aspect is discussed further in an accompanying white paper that explores the strategic utilisation of adaptive design at the development programme and product pipeline level. Importantly, recent data from Merck has indicated that adoption of an early phase adaptive trial strategy at this level saves at least $200m annually (Schindler J, Disruptive Innovations Conference, Boston, 2012).

“Adaptive Dose Ranging Studies Build Significant Efficiencies at Phase II and Optimize Identification of the Optimal Dose for Confirmatory Studies”

Several adaptive designs have been proposed and some already applied in practice using di�erent working models for the design engine: parsimonious monotonic sigmoid Emax9, Normal Dynamic Linear Model without monotonicity restrictions10, several simple models combined in a Multiple Comparison Procedure (MCP Mod).11

CASE STUDY

�e objective of this trial was to establish the correct dose(s) to take forward into a con�rmatory trial on a novel anti-psychotic drug in acute schizophrenia. �ere was data indicating that the dose-response might be non-monotonic. �erefore, the working model chosen for the design was the Normal Dynamic Linear Model, because it is �exible and robust, and also allows for capturing non-monotonic relationships. Patients were allocated to placebo, active comparator and seven doses of study drug. �e adaptive allocation was targeting both the minimum e�cacious dose (MED) and the dose achieving the maximum response (Dmax) on the primary endpoint, positive PANSS score at week 4. Weekly data updates were used to change the allocation and check early stopping rules based on predictive probabilities. A stopping rule for futility was based on a prespeci�ed threshold that no treatment dose achieves the clinically signi�cant di�erence (CSD) over placebo. A stopping rule for e�cacy required su�cient knowledge about the MED and Dmax and a prespeci�ed con�dence that the Dmax achieves CSD. For details, see Shen et al.12 �e adaptive design allowed earlier (5 months) decision making with a 99% con�dence in futility decision with $14m savings in direct grant cost. �e design, implementation, and outcome of a response adaptive dose-ranging trial of an investigational drug in patients with diabetes, incorporating a

dose expansion approach to �exibly address bothe�cacy and safety has also been reported recently by Berry et al.13 �e early (2 months) futility decision was achieved with a $3.6m saving in direct cost and additional cost savings in avoiding expensive production of phase III drug supplies.

FIGURE 4: Summary of Early Phase Case Studies

SUMMARY

Early phase adaptive design trials address a number of critical development questions that need to be answered before commitment is made to phase III con�rmatory trials. �e application of these innovative designs early in the product development process enhances clinical trial e�ciency, productivity and increases the probability of success at phase III.

�e deployment of these approaches across a pipeline of products brings substantial bene�ts to sponsor companies and enables more e�ective portfolio management and ultimately increases portfolio value (Figure 5).

adaptive by nature, but more recent adaptive designsprovide better estimate of drug safety (e.g. the maxi-mum tolerated dose) and better understanding of itsPK/PD characteristics. Combining the objectives of conventional Phase IIa and IIb studies in a single trial provides the obvious bene�t of reducing the timeline by running the two studies seamlessly under a single protocol with the same clinical team. Furthermore, such approaches achieve trial e�ciency by the prospective integration of data from both IIa and IIb in the �nal analysis.

FIRST-IN-HUMAN SINGLE ASCENDING DOSE ESCALATION DESIGNS

In contrast to traditional dose escalation designs, where for example, cohorts of subjects are allocated to ascending doses (and placebo) until a maximum tolerated dose (MTD) is empirically determined, the continual reassessment method (CRM) models the dose- and/or exposure-toxicity relationship to focus on the identi�cation of the dose-range of interest near the MTD and for this to be explored further. �ese adaptive dose escalation designs can be expanded to also e�ciently establish proof-of-mechanism or proof-of-target modulation, if validated biomarker endpoints are available.

FIGURE 1: Types of Adaptive Designs in the Learn Phase

�e traditional First in Human study objectives of safety and tolerability can be extended to include an

assessment of proof-of-mechanism in the very �rst study conducted with the investigational compound. �e ability to assess proof-of-mechanism in this earlydevelopmental setting provides an opportunity to greatly enhance the clinical development strategy for the compound. �e adaptation is intended to quickly hone in on the predicted dose-range of interest. �is avoids the risks associated with exposing patients in an ine�ective low dose range, whilst minimizing exposure to high doses with an unacceptable tolerability and safety pro�le.

“CRM Offers a Better Approach to Accurately Determining the MTD in Early Phase Oncology Studies”

CRM models are becoming routine for early phase oncology studies where accurate de�nition of the MTD is critical for selecting the optimum dose for e�cacy assessment. Conventional 3+3 designs have been shown to signi�cantly underestimate the MTD which increases the risk of poor outcome in e�cacy trials.3

SEAMLESS MULTIPLE ASCENDING DOSE AND PROOF-OF-CONCEPT STUDY DESIGNS

�e objective of seamless multiple ascending dose and proof-of-concept studies is to combine a multiple ascending dose (MAD) study in patients with the proof-of-concept of the investigational compound.

CASE STUDY

A good example is seen in the treatment of rheumatoid arthritis. �e study design involved a MAD escalation study, in which patients were randomly assigned only to the lowest treatment group or placebo in a 3:1 ratio (active treatment or placebo) which then converted into a parallel enrolling, proof of concept (POC) dose ranging study (see Figure 2). Consistent with the MAD paradigm, dose escalation (allowing subjects in the next higher dose level to

receive a second dose) occurred only after the safety review on 6 subjects assigned to active treatment in the preceding dose had been performed. �is safetyreview was undertaken by a data monitoring committee. Only after it was determined by the data monitoring committee that two consecutive doses of the experimental treatment were well tolerated at a given dose level, were subjects already enrolled in the next higher dose treatment group permitted to receive a second dose of the investigational product. In Stage 2, a utility combining biomarker observations with early readout of a clinical endpoint at week 4 was used to drop dose levels that did not ful�l prespeci�ed necessary conditions (Proof-of-Non-Viability). Viable cohorts and comparator were kept open until each of the surviving treatment arms reached a prede�ned number of patients. �e endpoint to assess proof-of-concept was a 12 week observation on the regulatory accepted endpoint for Rheumatoid Arthritis (ACR20).

�is adaptive design provides several advantages over more traditional clinical trial approaches, without compromising patient safety:

• The seamless design increased the utility of the• information obtained by allowing subjects in both• the MAD and POC stages to provide both • de�nitive safety and e�cacy data.

• Performing the MAD component in subjects with• rheumatoid arthritis receiving concomitant • methotrexate allowed for the earliest • characterization of safety and antigenicity of the• new compound in the clinically relevant • population.

• The adaptive design minimized the number of• subjects that were exposed to ine�ective doses of the• drug, while simultaneously focusing subjects to• doses that were most informative for accurate dose• selection for subsequent con�rmatory trials.

FIGURE 2: Trial Design for Seamless MAD/POC Case Study �erefore, the adaptive design optimizes the bene�t/risk balance for participating subjects via improved e�ciency of decision making in relation to the doses of the new drug studied. In this example the estimated cost saving through combining two trials into one was $1.2m and the development programme was accelerated by 9 months.

“The Development Programme was Accelerated by 9 Months and the Cost Saving of Combining Two Trials in One was Estimated at $1.2M”

PROOF OF CONCEPT AND DOSE-RANGING STUDY DESIGNS

An example of a proof of concept and dose-ranging study design is a dose-�nding study to evaluate the analgesic e�cacy and safety of a single dose of a new non-steroidal anti-in�ammatory drug in subjects experiencing pain after oral dental surgery.4 Six doses of the new drug, placebo and an active comparator are considered. �e adaptive design consists of three stages. At the �rst stage, subjects are equally randomized into three arms: placebo, active comparator, and the top dose of the new drug. At the �rst interim analysis, the study may be stopped either because of lack of assay sensitivity, i.e. inability to establish a prede�ned minimal bene�t of the comparator over and above placebo, or evidence that

the top dose of study drug could not achieve a minimal acceptable level of total pain relief at 8 hours. If the study continues to the second stage, additional subjects would be randomized to placebo, active control, and three doses of the new drug (low, medium, and high) in a 1:1:2:2:1 ratio. Again, the study could be stopped due to lack of assay sensitivity or futility. Otherwise, a D-optimal design will be used to determine both the doses of the new drug (among six possible) and subject allocation ratios for the third stage, see Figure 3. At the end of the study, a four parameter logistic model is used to �t the dose response and make recommendations about the dose to be taken in phase III.

FIGURE 3: Trial Design for POC and Adaptive Dose-Ranging Case Study

CASE STUDY

A full case study has been published in the pain �eld where a Bayesian adaptive dose ranging design was used to evaluate the e�cacy of a new analgesic drug in a proof-of-concept post-herpetic neuralgia trial.5 In total 7 doses of the drug were evaluated in an attempt to de�ne the e�ective dose response curve. �e maximum sample size was 280 patients, however, at the �rst interim analysis (80 patients) the study was stopped for futility as none of the doses showed any meaningful di�erence from placebo. �e study was formally stopped after 133 patients had been enrolled. Stopping of this trial saved an estimated direct grant cost of $0.76m with a further $1.3m savings in internal costs.

“Early Stopping of POC Trials Because of Lack of Efficacy Saves Significant Cost and Enables Valuable Resources to be Re-assigned to Alternative Programmes That May Have a Better Chance of Success”

A recent unpublished example shows that early stopping of a POC adaptive dose ranging trial in post-operative nausea and vomiting saved $0.5m in outsourced costs. In this example, 131 out of 200 patients had been enrolled when the decision was taken at the �rst interim analysis.

RESPONSE ADAPTIVE DOSE-RANGING STUDY DESIGNS

Recent simulation studies6,7 conducted by the PhRMA Working Group on Adaptive Dose-Ranging Studies have found that response-adaptive dose allocation designs are generally superior in performance to conventional pairwise comparisons approaches. Learning about the research question such as identifying a target dose, or estimating the dose-response, is optimized with these designs.

�e main objectives in an adaptive dose-ranging study are to detect the dose response, to determine if any doses(s) meets clinical relevance, to estimate the dose-response, and then to decide on the dose(s) (if any) to take into the con�rmatory Phase III. With adaptive designs for dose-ranging studies many more doses may be considered than in a conventional parallel group design without increasing sample size. �is is achieved mainly through an adaptive allocation rule that is changing the randomization ratio to di�erent treatment arms during the study and putting more subjects in the region that allow most accurate estimation of dose-response and better precision in selecting the target dose. For example, Orlo� et al.8 showed that “for half the sample size, the adaptive design is as powerful and e�cient as the standard approach”.

Page 4: DESIGN AND EXECUTION OF EARLY PHASE …...Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs

DESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALSDESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALS

PAGE 3 PAGE 8

EXECUTIVE SUMMARY

�e adoption of an adaptive design strategy across the product development process brings a number of important bene�ts. �ese include enhanced R&D e�ciency, enhanced R&D productivity and, importantly, increased probability of success at phase III. �e opportunity to manage risk starts in the early phase of a development programme, with a diligent view to appropriate dose selection which impacts not only the success of the downstream clinical trial programme, but also manufacturing costs and value analyses.

Adaptive design trials enhance R&D e�ciency by reducing the need to repeat trials that just miss their clinical end-point or fail to identify the e�ective dose response at the �rst attempt. By avoiding the need to run these trials again, signi�cant cost and time savings are achieved. �is is possible through use of adaptive designs that enable additional patients to be added to achieve statistical signi�cance the �rst time around or by allowing a wider dose range to be studied and then picking doses early in the trial that are in the optimum part of the dose response curve. In addition, early stopping of development programmes because a product is ine�ective enables scarce resources to be redeployed in additional trials which may show more promise. All of these factors increase development e�ciency.

Adaptive designs increase R&D productivity by enabling more accurate de�nition of the e�ective dose response at phase II which enables better design of the pivotal phase III programme which in turn increases the probability of success of the overall development programme. A number of phase III trials fail because the dose is either too high and causes unwanted safety issues or is too low to show su�cient e�cacy. Adaptive design trials enable optimized dose selection before the pivotal trial phase is started.

Another opportunity provided by adaptive design is population enrichment where drug response can be

optimised to speci�c patient sub-populations that respond better to treatment. Many phase III studies fail because the overall e�cacy of treatment is diluted as a consequence of the drug being evaluated in the full trial population rather than in the speci�c subset where the drug works best. Adaptive design enables early selection of the appropriate patient population and increases the probability of success.

“Adaptive Design Enables Selection of the Right Dose for the Right Patient Before Commitment to Expensive Phase III Trials”

Phases I and II are critical steps in the clinical development process as this is where important information about the product has to be generated and assessed before the decision is taken to commit to expensive phase III pivotal studies. �is early phase of development is known as the “Learn Phase” and the data generated relates to the e�ective dose-response, the safety pro�le and therapeutic index, appropriate endpoints, and the patient population best treated by the product under evaluation.

Adaptive designs have an important role to play in both phase I and phase II, but it is in the latter phase where they add the greatest value in terms of risk management. Validated methodologies are available to design early phase adaptive trials and this is an area of signi�cant interest across the pharmaceutical industry.

“Getting it Right at Phase II is a Major Objective for Adaptive Design and is the Point at which Phase III Success is Defined”

Companies that adopt a comprehensive adaptive design strategy in the “Learn” phase will make better development decisions, will build development e�ciency, will increase the probability of success of the molecules that enter phase III, ultimately bringing e�ective products to the market more e�ciently. Competitor companies that persevere with conventional trial designs will remain slow, ine�cient

candidate optimisation to Proof of concept stages. Moira assists partner companies with both acquisition and due diligence of in licensed compounds and strategic support for the out-licensing of successful candidates. �is has included design, planning and management of, strategic and clinical development plans, due diligence, decision analysis, preclinical, POC and Phase II programmes in a range of, indications including respiratory, anti-infective, immunomodulation, CNS and cardiovascular. �is experience covers a large range of targets and mecha-nisms of action. In order to support the strategic and clinical development planning Moira has worked extensively with international Key Opinion Leaders in respiratory, anti-infective and transplant indications.

PHIL BIRCH,SVP, CORPORATE BRAND MANAGER

Dr. Phil Birch is a Senior Vice President at Aptiv Solutions and has global responsibility for co-ordinating market education and client awareness in the adaptive clinical trials �eld. Dr. Birch has worked in the pharmaceutical industry for over 27 years and has held executive positions in corporate development, business development and R&D in a number of consulting, biotechnology and top 10 pharmaceutical companies.

REFERENCES

1. Dragalin V. Adaptive Designs: Terminology and Classi�cation. Drug Information Journal, 2006; 40: 425-436. 2. Gallo P, Chuang-Stein C, Dragalin V, Gaydos B, Krams M, Pinheiro J. Executive Summary of the PhRMA Working Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs for Phase I Oncology Trials, Clinical Operations Summit, Brussels, 2012).4. Dragalin V, Hsuan F, and Padmanabhan SK. Adaptive Designs for Dose-Finding Studies Based on Sigmoid Emax Model. J Biopharmaceutical Statistics 2007; 17:1051-1070.5. Smith MK, Jones I, Morris MF, Grieve A and Tan K. Implementation of a Bayesian adaptive design in a proof of concept study. Pharmaceutical Statistics 2006; 5:39-50. 6. Bornkamp B, Bretz F, Dmitrienko A, Enas G, Gaydos B, Hsu CH, König F, Krams M, Liu Q, Neuenschwander B, Parke T, Pinheiro J, Roy A, Sax R, Shen F. Innovative approaches for designing and analyzing adaptive dose-ranging trials: White Paper from the PhRMA working group on “Adaptive Dose-Ranging Studies”. J of Biopharmaceutical Statistics 2007; 17: 965–995. 7. Dragalin V, Bornkamp B, Bretz F, Miller F, Padmanabhan SK, Patel N, Perevozskaya I, Pinheiro J, and Smith J. A Simulation Study to Compare New Adaptive Dose–Ranging Designs. Statistics in Biopharmaceutical Research 2010; 2:487-512.8. Orlo� J, Douglas F, Pinheiro J, Levinson S, Branson M, Chaturvedi P, Ette E, Gallo P, Hirsch G, Mehta C, Patel N, Sabir S, Springs S, Stanski D, Golub H, Evers M, Fleming E, Singh N, Tramontin T. �e future of drug development: advancing clinical trial design. Nature Review-Drug Discovery. 2009; 8:940-957.9. Padmanabhan SK and Dragalin V. Adaptive Dc-optimal designs for dose �nding based on a continuous e�cacy endpoint. Biometrical Journal 2010; 52:836-852. 10. Grieve AP and Krams M. ASTIN: a Bayesian adaptive dose-response trial in acute stroke. Clinical Trials 2005; 2:340-351. 11. Bornkamp B, Pinheiro J, Bretz F. MCPMod: An R Package for the Design and Analysis of Dose-Finding Studies. Journal of Statistical Software 2009; 29:1-23.12. Shen J, Preskorn S, Dragalin V, Slomkowski M, Padmanabhan SK, Fardipour P, Sharma A, and Krams M. How Adaptive Trial Designs Can Increase E�ciency in Psychiatric Drug Development: A Case Study. Innovations in Clinical Neuroscience 2011; 8: 26-34.13. Berry SM, Spinelli W, Littman GS, Liang JZ, Fardipour P, Berry DA, Lewis RL, and Krams M. A Bayesian dose-�nding trial with adaptive dose expansion to �exibly assess e�cacy and safety of an investigational drug. Clinical Trials 2010; 7: 121-135.

ABOUT THE AUTHORS

VLAD DRAGALIN, SVP, INNOVATION CENTRE

Dr. Vlad Dragalin is a well-known adaptive design expert with 25 years of experience in developing the statistical methodology of adaptive designs and with over 12 years experience in pharmaceutical industry. Vlad is a Senior Vice President, Software Develop-ment and Consulting, in the Innovation Center at Aptiv Solutions. Vlad works with clients to improve the drug development process through the use of adaptive designs and to assist them with the trial design, simulation and execution, developing adaptive trials software tools, and working with project teams in clinical operations, data management and statistics during study conduct.

SARAH ARBE-BARNES,SVP, TRANSLATIONAL SCIENCES

Dr. Sarah Arbe-Barnes has over 24 years of experience in drug development and project leadership in the Pharma and service sector industry to include in/out-licensing and due diligence, strategic development and business planning, fundraising for start-ups/SMEs( not-for-pro�t sector and VC), along with optimized market access strategies and marketing. Sarah has led recent product approval programmes, overseeing a broad range of clinical and nonclinical project design and management activities from the point of candidate selection. Sarah heads the Transla-tional Sciences division of Aptiv Solutions, which is a dedicated global consultancy group, focused on the provision of drug and device development advice and expertise for small molecules and biologics.

MOIRA THOMSON,VP, TRANSLATIONAL SCIENCES

Moira �omson has over 15 years of experience in the Pharmaceutical Industry, and is involved in the design, planning and management of global develop-ment projects for a mixture of small and medium sized Pharma, and Biotech companies covering lead

and will fail to make the commercial returns that investors and shareholders demand. “Adaptive Design Builds Competitive Advantage”

�is White Paper summarizes the important aspects of early phase adaptive design and provides a summary of selected case studies which demonstrate the value of this innovative approach. It has been written to inform senior R&D decision-makers about the critical role of adaptive design in early phase development and the signi�cant bene�ts that this approach can bring.

BACKGROUND TO EARLY PHASE ADAPTIVE TRIAL DESIGN

Adaptive Clinical Trials are viewed as supporting a change in the development process as focus moves from blockbuster products to more specialized medicines that require a �exible cost-e�cient approach to investigating their e�ectiveness. Adaptive designs are statistical methodologies applied to speci�c stages of drug development where real time learning from accumulating trial data is applied to optimize subsequent study execution.1 �e bene�t of an Adaptive Clinical Trial lies in the monitoring of data to make design modi�cation adjustments to an aspect of a trial leading to quicker development decisions around key go/no-go milestones that impact time to market and/or minimise costly R&D losses.

�e adaptations are prospectively de�ned prior to the start of the trial and can include stopping early either for futility or success, expanding the sample size due to greater than expected data variability, or allocating patients preferentially to treatment regimens with a better therapeutic index. Virtually any aspect of the trial may be the potential target of design modi�cations. Importantly, these modi�cations are a design feature aimed to enhance the trial, not a remedy for inadequate planning. �ey are not ad hoc study corrections made via protocol amendment.2

�e speci�c adaptations considered, and the basis of their implementation, are carefully de�ned based on strict rules, justi�ed statistically and scienti�cally, and are an integral part of the �nal, pre-recruitment trial protocol. Failing this would compromise the interpretation and acceptance of study results.

Adaptive designs often employ frequent interim analyses of all accumulated data (and, possibly, external trial data) to determine whether pre-planned design modi�cations will be ‘triggered’. Interim analyses partition the trial into multiple stages, each trial stage’s characteristics (number of arms, number of patients to be enrolled, their allocation between arms, stage duration, etc.) de�ned by the preceding interim analysis results. �e ability to periodically, or even continually, examine available data to determine whether trial modi�cations are necessary and implement pre-de�ned design changes when indicated, gives adaptive design its strength and �exibility.

�e use of adaptive design in the exploratory stage of clinical trials can increase the e�ciency of drug development by improving our ability to e�ciently learn about the dose-response and better determine whether to take a drug (and the right dose of the drug and in the right population) forward into later phase testing. �ese designs explicitly address multiple trial goals, adaptively allocate subjects according to ongoing information needs, and allow termination for both early success and futility considerations. �is approach can maximize the ability to test a larger number of doses in a single trial while simultaneously increasing the e�ciency of the trial in terms of making better go/no-go decisions about continuing the trial and/or the development of the drug for a speci�c indication or sub-population. During the “Learn” phase many aspects of the study design can be modi�ed: number of subjects, study duration, endpoint selection, treatment duration, number of treatments, patient population. �e adaptation of multiple trial features that are a legitimate concern in con�rmatory studies can become, in fact, an

appealing feature of adaptive designs in exploratory development, leading to enhanced learning to set the appropriate stage for con�rmatory trials.

“The Use of Adaptive Design in the Learn Phase Offers the Best Opportunity to Assess the Product Characteristics that Determine Success”

Judicious use of adaptive designs may increase the information value per resource unit invested by avoiding allocation of patients to non-e�cacious/unsafe therapies and allowing stopping decisions to be made at the earliest possible time point. Ultimately this will accelerate the development of promising therapies to bene�t patients.

Adaptive Clinical Trials can result in the more ethical treatment of patients from at least two perspectives. First, a larger number of patients can be randomized to more favorable, e�ective doses with fewer patients exposed to less e�ective doses. In addition, there is the potential for including fewer total patients in the early stages of the development programmes with attending lower risk of exposure to adverse events. A greater saving in patients may be obtained at the programme level rather than the individual study level as adaptive methodology applied across a programme will reduce the need to repeat ine�ective or inconclusive studies.

Adaptive designs represent an advanced methodology to support drug development. A prerequisite for their successful implementation is an understanding of the underlying methodology and their impact on the logistics of the trial, such as data management and monitoring procedures, electronic data capture/interactive web response services, drug supply management, and data-monitoring committee operating procedures. �ese designs have an impact on drug development strategies, trial protocols, clinical trial material doses and availability, informed consent forms, data analysis, and reporting plans. Adaptive design will also change the dynamics of

enrollment, randomization, and data capturing process, monitoring, and data cleaning systems. To be able to make informed decisions, the sponsor must have access to real time clinical data from all sources and be able to easily review and assess this data.

�is in turn requires �exibility through:

• Data management systems for rapid clinical data• access and data cleaning

• Drug supply systems for drug product planning • and management

• Optimal systems providing randomization, • medication kit management and emergency• unblinding

Planning of Adaptive Clinical Trials is a key feature in the success of such innovative approaches. Determining the optimal characteristics of the study design can be a complex yet critical decision that may require more upfront planning before the protocol can be �nalized. However, there are several commercial software packages that provide a powerful tool for planning, simulation and analysis of complex adaptive designs. FACTS™ is comprehensive software for adaptive designs in the “Learn” phase and includes early stage dose-escalation designs, adaptive dose-ranging studies, interim decision rules based on both e�cacy and safety responses, and population enrichment. ADDPLAN® is a validated software package for adaptive design in the “Con�rm” phase and incorporates sample size re-estimation, adaptive group sequential designs, multiple comparison procedures for multi-armed adaptive trials, including treatment selection designs, �exible combination of clinical research phases, and population enrichment designs.

TYPES OF ADAPTIVE DESIGNS IN “LEARN” STUDIES

Examples of adaptive designs in the Learn phase are shown in Figure 1. Phase I studies are already

FIGURE 5: Adaptive Design Creates Value at Level of the Single Trial, DevelopmentProgramme and Product Pipeline

�is aspect is discussed further in an accompanying white paper that explores the strategic utilisation of adaptive design at the development programme and product pipeline level. Importantly, recent data from Merck has indicated that adoption of an early phase adaptive trial strategy at this level saves at least $200m annually (Schindler J, Disruptive Innovations Conference, Boston, 2012).

“Adaptive Dose Ranging Studies Build Significant Efficiencies at Phase II and Optimize Identification of the Optimal Dose for Confirmatory Studies”

Several adaptive designs have been proposed and some already applied in practice using di�erent working models for the design engine: parsimonious monotonic sigmoid Emax9, Normal Dynamic Linear Model without monotonicity restrictions10, several simple models combined in a Multiple Comparison Procedure (MCP Mod).11

CASE STUDY

�e objective of this trial was to establish the correct dose(s) to take forward into a con�rmatory trial on a novel anti-psychotic drug in acute schizophrenia. �ere was data indicating that the dose-response might be non-monotonic. �erefore, the working model chosen for the design was the Normal Dynamic Linear Model, because it is �exible and robust, and also allows for capturing non-monotonic relationships. Patients were allocated to placebo, active comparator and seven doses of study drug. �e adaptive allocation was targeting both the minimum e�cacious dose (MED) and the dose achieving the maximum response (Dmax) on the primary endpoint, positive PANSS score at week 4. Weekly data updates were used to change the allocation and check early stopping rules based on predictive probabilities. A stopping rule for futility was based on a prespeci�ed threshold that no treatment dose achieves the clinically signi�cant di�erence (CSD) over placebo. A stopping rule for e�cacy required su�cient knowledge about the MED and Dmax and a prespeci�ed con�dence that the Dmax achieves CSD. For details, see Shen et al.12 �e adaptive design allowed earlier (5 months) decision making with a 99% con�dence in futility decision with $14m savings in direct grant cost. �e design, implementation, and outcome of a response adaptive dose-ranging trial of an investigational drug in patients with diabetes, incorporating a

dose expansion approach to �exibly address bothe�cacy and safety has also been reported recently by Berry et al.13 �e early (2 months) futility decision was achieved with a $3.6m saving in direct cost and additional cost savings in avoiding expensive production of phase III drug supplies.

FIGURE 4: Summary of Early Phase Case Studies

SUMMARY

Early phase adaptive design trials address a number of critical development questions that need to be answered before commitment is made to phase III con�rmatory trials. �e application of these innovative designs early in the product development process enhances clinical trial e�ciency, productivity and increases the probability of success at phase III.

�e deployment of these approaches across a pipeline of products brings substantial bene�ts to sponsor companies and enables more e�ective portfolio management and ultimately increases portfolio value (Figure 5).

adaptive by nature, but more recent adaptive designsprovide better estimate of drug safety (e.g. the maxi-mum tolerated dose) and better understanding of itsPK/PD characteristics. Combining the objectives of conventional Phase IIa and IIb studies in a single trial provides the obvious bene�t of reducing the timeline by running the two studies seamlessly under a single protocol with the same clinical team. Furthermore, such approaches achieve trial e�ciency by the prospective integration of data from both IIa and IIb in the �nal analysis.

FIRST-IN-HUMAN SINGLE ASCENDING DOSE ESCALATION DESIGNS

In contrast to traditional dose escalation designs, where for example, cohorts of subjects are allocated to ascending doses (and placebo) until a maximum tolerated dose (MTD) is empirically determined, the continual reassessment method (CRM) models the dose- and/or exposure-toxicity relationship to focus on the identi�cation of the dose-range of interest near the MTD and for this to be explored further. �ese adaptive dose escalation designs can be expanded to also e�ciently establish proof-of-mechanism or proof-of-target modulation, if validated biomarker endpoints are available.

FIGURE 1: Types of Adaptive Designs in the Learn Phase

�e traditional First in Human study objectives of safety and tolerability can be extended to include an

assessment of proof-of-mechanism in the very �rst study conducted with the investigational compound. �e ability to assess proof-of-mechanism in this earlydevelopmental setting provides an opportunity to greatly enhance the clinical development strategy for the compound. �e adaptation is intended to quickly hone in on the predicted dose-range of interest. �is avoids the risks associated with exposing patients in an ine�ective low dose range, whilst minimizing exposure to high doses with an unacceptable tolerability and safety pro�le.

“CRM Offers a Better Approach to Accurately Determining the MTD in Early Phase Oncology Studies”

CRM models are becoming routine for early phase oncology studies where accurate de�nition of the MTD is critical for selecting the optimum dose for e�cacy assessment. Conventional 3+3 designs have been shown to signi�cantly underestimate the MTD which increases the risk of poor outcome in e�cacy trials.3

SEAMLESS MULTIPLE ASCENDING DOSE AND PROOF-OF-CONCEPT STUDY DESIGNS

�e objective of seamless multiple ascending dose and proof-of-concept studies is to combine a multiple ascending dose (MAD) study in patients with the proof-of-concept of the investigational compound.

CASE STUDY

A good example is seen in the treatment of rheumatoid arthritis. �e study design involved a MAD escalation study, in which patients were randomly assigned only to the lowest treatment group or placebo in a 3:1 ratio (active treatment or placebo) which then converted into a parallel enrolling, proof of concept (POC) dose ranging study (see Figure 2). Consistent with the MAD paradigm, dose escalation (allowing subjects in the next higher dose level to

receive a second dose) occurred only after the safety review on 6 subjects assigned to active treatment in the preceding dose had been performed. �is safetyreview was undertaken by a data monitoring committee. Only after it was determined by the data monitoring committee that two consecutive doses of the experimental treatment were well tolerated at a given dose level, were subjects already enrolled in the next higher dose treatment group permitted to receive a second dose of the investigational product. In Stage 2, a utility combining biomarker observations with early readout of a clinical endpoint at week 4 was used to drop dose levels that did not ful�l prespeci�ed necessary conditions (Proof-of-Non-Viability). Viable cohorts and comparator were kept open until each of the surviving treatment arms reached a prede�ned number of patients. �e endpoint to assess proof-of-concept was a 12 week observation on the regulatory accepted endpoint for Rheumatoid Arthritis (ACR20).

�is adaptive design provides several advantages over more traditional clinical trial approaches, without compromising patient safety:

• The seamless design increased the utility of the• information obtained by allowing subjects in both• the MAD and POC stages to provide both • de�nitive safety and e�cacy data.

• Performing the MAD component in subjects with• rheumatoid arthritis receiving concomitant • methotrexate allowed for the earliest • characterization of safety and antigenicity of the• new compound in the clinically relevant • population.

• The adaptive design minimized the number of• subjects that were exposed to ine�ective doses of the• drug, while simultaneously focusing subjects to• doses that were most informative for accurate dose• selection for subsequent con�rmatory trials.

FIGURE 2: Trial Design for Seamless MAD/POC Case Study �erefore, the adaptive design optimizes the bene�t/risk balance for participating subjects via improved e�ciency of decision making in relation to the doses of the new drug studied. In this example the estimated cost saving through combining two trials into one was $1.2m and the development programme was accelerated by 9 months.

“The Development Programme was Accelerated by 9 Months and the Cost Saving of Combining Two Trials in One was Estimated at $1.2M”

PROOF OF CONCEPT AND DOSE-RANGING STUDY DESIGNS

An example of a proof of concept and dose-ranging study design is a dose-�nding study to evaluate the analgesic e�cacy and safety of a single dose of a new non-steroidal anti-in�ammatory drug in subjects experiencing pain after oral dental surgery.4 Six doses of the new drug, placebo and an active comparator are considered. �e adaptive design consists of three stages. At the �rst stage, subjects are equally randomized into three arms: placebo, active comparator, and the top dose of the new drug. At the �rst interim analysis, the study may be stopped either because of lack of assay sensitivity, i.e. inability to establish a prede�ned minimal bene�t of the comparator over and above placebo, or evidence that

the top dose of study drug could not achieve a minimal acceptable level of total pain relief at 8 hours. If the study continues to the second stage, additional subjects would be randomized to placebo, active control, and three doses of the new drug (low, medium, and high) in a 1:1:2:2:1 ratio. Again, the study could be stopped due to lack of assay sensitivity or futility. Otherwise, a D-optimal design will be used to determine both the doses of the new drug (among six possible) and subject allocation ratios for the third stage, see Figure 3. At the end of the study, a four parameter logistic model is used to �t the dose response and make recommendations about the dose to be taken in phase III.

FIGURE 3: Trial Design for POC and Adaptive Dose-Ranging Case Study

CASE STUDY

A full case study has been published in the pain �eld where a Bayesian adaptive dose ranging design was used to evaluate the e�cacy of a new analgesic drug in a proof-of-concept post-herpetic neuralgia trial.5 In total 7 doses of the drug were evaluated in an attempt to de�ne the e�ective dose response curve. �e maximum sample size was 280 patients, however, at the �rst interim analysis (80 patients) the study was stopped for futility as none of the doses showed any meaningful di�erence from placebo. �e study was formally stopped after 133 patients had been enrolled. Stopping of this trial saved an estimated direct grant cost of $0.76m with a further $1.3m savings in internal costs.

“Early Stopping of POC Trials Because of Lack of Efficacy Saves Significant Cost and Enables Valuable Resources to be Re-assigned to Alternative Programmes That May Have a Better Chance of Success”

A recent unpublished example shows that early stopping of a POC adaptive dose ranging trial in post-operative nausea and vomiting saved $0.5m in outsourced costs. In this example, 131 out of 200 patients had been enrolled when the decision was taken at the �rst interim analysis.

RESPONSE ADAPTIVE DOSE-RANGING STUDY DESIGNS

Recent simulation studies6,7 conducted by the PhRMA Working Group on Adaptive Dose-Ranging Studies have found that response-adaptive dose allocation designs are generally superior in performance to conventional pairwise comparisons approaches. Learning about the research question such as identifying a target dose, or estimating the dose-response, is optimized with these designs.

�e main objectives in an adaptive dose-ranging study are to detect the dose response, to determine if any doses(s) meets clinical relevance, to estimate the dose-response, and then to decide on the dose(s) (if any) to take into the con�rmatory Phase III. With adaptive designs for dose-ranging studies many more doses may be considered than in a conventional parallel group design without increasing sample size. �is is achieved mainly through an adaptive allocation rule that is changing the randomization ratio to di�erent treatment arms during the study and putting more subjects in the region that allow most accurate estimation of dose-response and better precision in selecting the target dose. For example, Orlo� et al.8 showed that “for half the sample size, the adaptive design is as powerful and e�cient as the standard approach”.

Clinical TrialTrial Efficiency

Better decision-makingMore information per $ invested

DevelopmentProgramme

Increased probability ofsuccess at Phase III

DevelopmentPipeline

Pipeline efficiency andproductivity

BenefitsStrategy

Value of a

dop

ting a

n adaptive stra

tegy

Page 5: DESIGN AND EXECUTION OF EARLY PHASE …...Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs

DESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALSDESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALS

PAGE 7 PAGE 4

EXECUTIVE SUMMARY

�e adoption of an adaptive design strategy across the product development process brings a number of important bene�ts. �ese include enhanced R&D e�ciency, enhanced R&D productivity and, importantly, increased probability of success at phase III. �e opportunity to manage risk starts in the early phase of a development programme, with a diligent view to appropriate dose selection which impacts not only the success of the downstream clinical trial programme, but also manufacturing costs and value analyses.

Adaptive design trials enhance R&D e�ciency by reducing the need to repeat trials that just miss their clinical end-point or fail to identify the e�ective dose response at the �rst attempt. By avoiding the need to run these trials again, signi�cant cost and time savings are achieved. �is is possible through use of adaptive designs that enable additional patients to be added to achieve statistical signi�cance the �rst time around or by allowing a wider dose range to be studied and then picking doses early in the trial that are in the optimum part of the dose response curve. In addition, early stopping of development programmes because a product is ine�ective enables scarce resources to be redeployed in additional trials which may show more promise. All of these factors increase development e�ciency.

Adaptive designs increase R&D productivity by enabling more accurate de�nition of the e�ective dose response at phase II which enables better design of the pivotal phase III programme which in turn increases the probability of success of the overall development programme. A number of phase III trials fail because the dose is either too high and causes unwanted safety issues or is too low to show su�cient e�cacy. Adaptive design trials enable optimized dose selection before the pivotal trial phase is started.

Another opportunity provided by adaptive design is population enrichment where drug response can be

optimised to speci�c patient sub-populations that respond better to treatment. Many phase III studies fail because the overall e�cacy of treatment is diluted as a consequence of the drug being evaluated in the full trial population rather than in the speci�c subset where the drug works best. Adaptive design enables early selection of the appropriate patient population and increases the probability of success.

“Adaptive Design Enables Selection of the Right Dose for the Right Patient Before Commitment to Expensive Phase III Trials”

Phases I and II are critical steps in the clinical development process as this is where important information about the product has to be generated and assessed before the decision is taken to commit to expensive phase III pivotal studies. �is early phase of development is known as the “Learn Phase” and the data generated relates to the e�ective dose-response, the safety pro�le and therapeutic index, appropriate endpoints, and the patient population best treated by the product under evaluation.

Adaptive designs have an important role to play in both phase I and phase II, but it is in the latter phase where they add the greatest value in terms of risk management. Validated methodologies are available to design early phase adaptive trials and this is an area of signi�cant interest across the pharmaceutical industry.

“Getting it Right at Phase II is a Major Objective for Adaptive Design and is the Point at which Phase III Success is Defined”

Companies that adopt a comprehensive adaptive design strategy in the “Learn” phase will make better development decisions, will build development e�ciency, will increase the probability of success of the molecules that enter phase III, ultimately bringing e�ective products to the market more e�ciently. Competitor companies that persevere with conventional trial designs will remain slow, ine�cient

candidate optimisation to Proof of concept stages. Moira assists partner companies with both acquisition and due diligence of in licensed compounds and strategic support for the out-licensing of successful candidates. �is has included design, planning and management of, strategic and clinical development plans, due diligence, decision analysis, preclinical, POC and Phase II programmes in a range of, indications including respiratory, anti-infective, immunomodulation, CNS and cardiovascular. �is experience covers a large range of targets and mecha-nisms of action. In order to support the strategic and clinical development planning Moira has worked extensively with international Key Opinion Leaders in respiratory, anti-infective and transplant indications.

PHIL BIRCH,SVP, CORPORATE BRAND MANAGER

Dr. Phil Birch is a Senior Vice President at Aptiv Solutions and has global responsibility for co-ordinating market education and client awareness in the adaptive clinical trials �eld. Dr. Birch has worked in the pharmaceutical industry for over 27 years and has held executive positions in corporate development, business development and R&D in a number of consulting, biotechnology and top 10 pharmaceutical companies.

REFERENCES

1. Dragalin V. Adaptive Designs: Terminology and Classi�cation. Drug Information Journal, 2006; 40: 425-436. 2. Gallo P, Chuang-Stein C, Dragalin V, Gaydos B, Krams M, Pinheiro J. Executive Summary of the PhRMA Working Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs for Phase I Oncology Trials, Clinical Operations Summit, Brussels, 2012).4. Dragalin V, Hsuan F, and Padmanabhan SK. Adaptive Designs for Dose-Finding Studies Based on Sigmoid Emax Model. J Biopharmaceutical Statistics 2007; 17:1051-1070.5. Smith MK, Jones I, Morris MF, Grieve A and Tan K. Implementation of a Bayesian adaptive design in a proof of concept study. Pharmaceutical Statistics 2006; 5:39-50. 6. Bornkamp B, Bretz F, Dmitrienko A, Enas G, Gaydos B, Hsu CH, König F, Krams M, Liu Q, Neuenschwander B, Parke T, Pinheiro J, Roy A, Sax R, Shen F. Innovative approaches for designing and analyzing adaptive dose-ranging trials: White Paper from the PhRMA working group on “Adaptive Dose-Ranging Studies”. J of Biopharmaceutical Statistics 2007; 17: 965–995. 7. Dragalin V, Bornkamp B, Bretz F, Miller F, Padmanabhan SK, Patel N, Perevozskaya I, Pinheiro J, and Smith J. A Simulation Study to Compare New Adaptive Dose–Ranging Designs. Statistics in Biopharmaceutical Research 2010; 2:487-512.8. Orlo� J, Douglas F, Pinheiro J, Levinson S, Branson M, Chaturvedi P, Ette E, Gallo P, Hirsch G, Mehta C, Patel N, Sabir S, Springs S, Stanski D, Golub H, Evers M, Fleming E, Singh N, Tramontin T. �e future of drug development: advancing clinical trial design. Nature Review-Drug Discovery. 2009; 8:940-957.9. Padmanabhan SK and Dragalin V. Adaptive Dc-optimal designs for dose �nding based on a continuous e�cacy endpoint. Biometrical Journal 2010; 52:836-852. 10. Grieve AP and Krams M. ASTIN: a Bayesian adaptive dose-response trial in acute stroke. Clinical Trials 2005; 2:340-351. 11. Bornkamp B, Pinheiro J, Bretz F. MCPMod: An R Package for the Design and Analysis of Dose-Finding Studies. Journal of Statistical Software 2009; 29:1-23.12. Shen J, Preskorn S, Dragalin V, Slomkowski M, Padmanabhan SK, Fardipour P, Sharma A, and Krams M. How Adaptive Trial Designs Can Increase E�ciency in Psychiatric Drug Development: A Case Study. Innovations in Clinical Neuroscience 2011; 8: 26-34.13. Berry SM, Spinelli W, Littman GS, Liang JZ, Fardipour P, Berry DA, Lewis RL, and Krams M. A Bayesian dose-�nding trial with adaptive dose expansion to �exibly assess e�cacy and safety of an investigational drug. Clinical Trials 2010; 7: 121-135.

ABOUT THE AUTHORS

VLAD DRAGALIN, SVP, INNOVATION CENTRE

Dr. Vlad Dragalin is a well-known adaptive design expert with 25 years of experience in developing the statistical methodology of adaptive designs and with over 12 years experience in pharmaceutical industry. Vlad is a Senior Vice President, Software Develop-ment and Consulting, in the Innovation Center at Aptiv Solutions. Vlad works with clients to improve the drug development process through the use of adaptive designs and to assist them with the trial design, simulation and execution, developing adaptive trials software tools, and working with project teams in clinical operations, data management and statistics during study conduct.

SARAH ARBE-BARNES,SVP, TRANSLATIONAL SCIENCES

Dr. Sarah Arbe-Barnes has over 24 years of experience in drug development and project leadership in the Pharma and service sector industry to include in/out-licensing and due diligence, strategic development and business planning, fundraising for start-ups/SMEs( not-for-pro�t sector and VC), along with optimized market access strategies and marketing. Sarah has led recent product approval programmes, overseeing a broad range of clinical and nonclinical project design and management activities from the point of candidate selection. Sarah heads the Transla-tional Sciences division of Aptiv Solutions, which is a dedicated global consultancy group, focused on the provision of drug and device development advice and expertise for small molecules and biologics.

MOIRA THOMSON,VP, TRANSLATIONAL SCIENCES

Moira �omson has over 15 years of experience in the Pharmaceutical Industry, and is involved in the design, planning and management of global develop-ment projects for a mixture of small and medium sized Pharma, and Biotech companies covering lead

and will fail to make the commercial returns that investors and shareholders demand. “Adaptive Design Builds Competitive Advantage”

�is White Paper summarizes the important aspects of early phase adaptive design and provides a summary of selected case studies which demonstrate the value of this innovative approach. It has been written to inform senior R&D decision-makers about the critical role of adaptive design in early phase development and the signi�cant bene�ts that this approach can bring.

BACKGROUND TO EARLY PHASE ADAPTIVE TRIAL DESIGN

Adaptive Clinical Trials are viewed as supporting a change in the development process as focus moves from blockbuster products to more specialized medicines that require a �exible cost-e�cient approach to investigating their e�ectiveness. Adaptive designs are statistical methodologies applied to speci�c stages of drug development where real time learning from accumulating trial data is applied to optimize subsequent study execution.1 �e bene�t of an Adaptive Clinical Trial lies in the monitoring of data to make design modi�cation adjustments to an aspect of a trial leading to quicker development decisions around key go/no-go milestones that impact time to market and/or minimise costly R&D losses.

�e adaptations are prospectively de�ned prior to the start of the trial and can include stopping early either for futility or success, expanding the sample size due to greater than expected data variability, or allocating patients preferentially to treatment regimens with a better therapeutic index. Virtually any aspect of the trial may be the potential target of design modi�cations. Importantly, these modi�cations are a design feature aimed to enhance the trial, not a remedy for inadequate planning. �ey are not ad hoc study corrections made via protocol amendment.2

�e speci�c adaptations considered, and the basis of their implementation, are carefully de�ned based on strict rules, justi�ed statistically and scienti�cally, and are an integral part of the �nal, pre-recruitment trial protocol. Failing this would compromise the interpretation and acceptance of study results.

Adaptive designs often employ frequent interim analyses of all accumulated data (and, possibly, external trial data) to determine whether pre-planned design modi�cations will be ‘triggered’. Interim analyses partition the trial into multiple stages, each trial stage’s characteristics (number of arms, number of patients to be enrolled, their allocation between arms, stage duration, etc.) de�ned by the preceding interim analysis results. �e ability to periodically, or even continually, examine available data to determine whether trial modi�cations are necessary and implement pre-de�ned design changes when indicated, gives adaptive design its strength and �exibility.

�e use of adaptive design in the exploratory stage of clinical trials can increase the e�ciency of drug development by improving our ability to e�ciently learn about the dose-response and better determine whether to take a drug (and the right dose of the drug and in the right population) forward into later phase testing. �ese designs explicitly address multiple trial goals, adaptively allocate subjects according to ongoing information needs, and allow termination for both early success and futility considerations. �is approach can maximize the ability to test a larger number of doses in a single trial while simultaneously increasing the e�ciency of the trial in terms of making better go/no-go decisions about continuing the trial and/or the development of the drug for a speci�c indication or sub-population. During the “Learn” phase many aspects of the study design can be modi�ed: number of subjects, study duration, endpoint selection, treatment duration, number of treatments, patient population. �e adaptation of multiple trial features that are a legitimate concern in con�rmatory studies can become, in fact, an

appealing feature of adaptive designs in exploratory development, leading to enhanced learning to set the appropriate stage for con�rmatory trials.

“The Use of Adaptive Design in the Learn Phase Offers the Best Opportunity to Assess the Product Characteristics that Determine Success”

Judicious use of adaptive designs may increase the information value per resource unit invested by avoiding allocation of patients to non-e�cacious/unsafe therapies and allowing stopping decisions to be made at the earliest possible time point. Ultimately this will accelerate the development of promising therapies to bene�t patients.

Adaptive Clinical Trials can result in the more ethical treatment of patients from at least two perspectives. First, a larger number of patients can be randomized to more favorable, e�ective doses with fewer patients exposed to less e�ective doses. In addition, there is the potential for including fewer total patients in the early stages of the development programmes with attending lower risk of exposure to adverse events. A greater saving in patients may be obtained at the programme level rather than the individual study level as adaptive methodology applied across a programme will reduce the need to repeat ine�ective or inconclusive studies.

Adaptive designs represent an advanced methodology to support drug development. A prerequisite for their successful implementation is an understanding of the underlying methodology and their impact on the logistics of the trial, such as data management and monitoring procedures, electronic data capture/interactive web response services, drug supply management, and data-monitoring committee operating procedures. �ese designs have an impact on drug development strategies, trial protocols, clinical trial material doses and availability, informed consent forms, data analysis, and reporting plans. Adaptive design will also change the dynamics of

enrollment, randomization, and data capturing process, monitoring, and data cleaning systems. To be able to make informed decisions, the sponsor must have access to real time clinical data from all sources and be able to easily review and assess this data.

�is in turn requires �exibility through:

• Data management systems for rapid clinical data• access and data cleaning

• Drug supply systems for drug product planning • and management

• Optimal systems providing randomization, • medication kit management and emergency• unblinding

Planning of Adaptive Clinical Trials is a key feature in the success of such innovative approaches. Determining the optimal characteristics of the study design can be a complex yet critical decision that may require more upfront planning before the protocol can be �nalized. However, there are several commercial software packages that provide a powerful tool for planning, simulation and analysis of complex adaptive designs. FACTS™ is comprehensive software for adaptive designs in the “Learn” phase and includes early stage dose-escalation designs, adaptive dose-ranging studies, interim decision rules based on both e�cacy and safety responses, and population enrichment. ADDPLAN® is a validated software package for adaptive design in the “Con�rm” phase and incorporates sample size re-estimation, adaptive group sequential designs, multiple comparison procedures for multi-armed adaptive trials, including treatment selection designs, �exible combination of clinical research phases, and population enrichment designs.

TYPES OF ADAPTIVE DESIGNS IN “LEARN” STUDIES

Examples of adaptive designs in the Learn phase are shown in Figure 1. Phase I studies are already

FIGURE 5: Adaptive Design Creates Value at Level of the Single Trial, DevelopmentProgramme and Product Pipeline

�is aspect is discussed further in an accompanying white paper that explores the strategic utilisation of adaptive design at the development programme and product pipeline level. Importantly, recent data from Merck has indicated that adoption of an early phase adaptive trial strategy at this level saves at least $200m annually (Schindler J, Disruptive Innovations Conference, Boston, 2012).

“Adaptive Dose Ranging Studies Build Significant Efficiencies at Phase II and Optimize Identification of the Optimal Dose for Confirmatory Studies”

Several adaptive designs have been proposed and some already applied in practice using di�erent working models for the design engine: parsimonious monotonic sigmoid Emax9, Normal Dynamic Linear Model without monotonicity restrictions10, several simple models combined in a Multiple Comparison Procedure (MCP Mod).11

CASE STUDY

�e objective of this trial was to establish the correct dose(s) to take forward into a con�rmatory trial on a novel anti-psychotic drug in acute schizophrenia. �ere was data indicating that the dose-response might be non-monotonic. �erefore, the working model chosen for the design was the Normal Dynamic Linear Model, because it is �exible and robust, and also allows for capturing non-monotonic relationships. Patients were allocated to placebo, active comparator and seven doses of study drug. �e adaptive allocation was targeting both the minimum e�cacious dose (MED) and the dose achieving the maximum response (Dmax) on the primary endpoint, positive PANSS score at week 4. Weekly data updates were used to change the allocation and check early stopping rules based on predictive probabilities. A stopping rule for futility was based on a prespeci�ed threshold that no treatment dose achieves the clinically signi�cant di�erence (CSD) over placebo. A stopping rule for e�cacy required su�cient knowledge about the MED and Dmax and a prespeci�ed con�dence that the Dmax achieves CSD. For details, see Shen et al.12 �e adaptive design allowed earlier (5 months) decision making with a 99% con�dence in futility decision with $14m savings in direct grant cost. �e design, implementation, and outcome of a response adaptive dose-ranging trial of an investigational drug in patients with diabetes, incorporating a

dose expansion approach to �exibly address bothe�cacy and safety has also been reported recently by Berry et al.13 �e early (2 months) futility decision was achieved with a $3.6m saving in direct cost and additional cost savings in avoiding expensive production of phase III drug supplies.

FIGURE 4: Summary of Early Phase Case Studies

SUMMARY

Early phase adaptive design trials address a number of critical development questions that need to be answered before commitment is made to phase III con�rmatory trials. �e application of these innovative designs early in the product development process enhances clinical trial e�ciency, productivity and increases the probability of success at phase III.

�e deployment of these approaches across a pipeline of products brings substantial bene�ts to sponsor companies and enables more e�ective portfolio management and ultimately increases portfolio value (Figure 5).

adaptive by nature, but more recent adaptive designsprovide better estimate of drug safety (e.g. the maxi-mum tolerated dose) and better understanding of itsPK/PD characteristics. Combining the objectives of conventional Phase IIa and IIb studies in a single trial provides the obvious bene�t of reducing the timeline by running the two studies seamlessly under a single protocol with the same clinical team. Furthermore, such approaches achieve trial e�ciency by the prospective integration of data from both IIa and IIb in the �nal analysis.

FIRST-IN-HUMAN SINGLE ASCENDING DOSE ESCALATION DESIGNS

In contrast to traditional dose escalation designs, where for example, cohorts of subjects are allocated to ascending doses (and placebo) until a maximum tolerated dose (MTD) is empirically determined, the continual reassessment method (CRM) models the dose- and/or exposure-toxicity relationship to focus on the identi�cation of the dose-range of interest near the MTD and for this to be explored further. �ese adaptive dose escalation designs can be expanded to also e�ciently establish proof-of-mechanism or proof-of-target modulation, if validated biomarker endpoints are available.

FIGURE 1: Types of Adaptive Designs in the Learn Phase

�e traditional First in Human study objectives of safety and tolerability can be extended to include an

assessment of proof-of-mechanism in the very �rst study conducted with the investigational compound. �e ability to assess proof-of-mechanism in this earlydevelopmental setting provides an opportunity to greatly enhance the clinical development strategy for the compound. �e adaptation is intended to quickly hone in on the predicted dose-range of interest. �is avoids the risks associated with exposing patients in an ine�ective low dose range, whilst minimizing exposure to high doses with an unacceptable tolerability and safety pro�le.

“CRM Offers a Better Approach to Accurately Determining the MTD in Early Phase Oncology Studies”

CRM models are becoming routine for early phase oncology studies where accurate de�nition of the MTD is critical for selecting the optimum dose for e�cacy assessment. Conventional 3+3 designs have been shown to signi�cantly underestimate the MTD which increases the risk of poor outcome in e�cacy trials.3

SEAMLESS MULTIPLE ASCENDING DOSE AND PROOF-OF-CONCEPT STUDY DESIGNS

�e objective of seamless multiple ascending dose and proof-of-concept studies is to combine a multiple ascending dose (MAD) study in patients with the proof-of-concept of the investigational compound.

CASE STUDY

A good example is seen in the treatment of rheumatoid arthritis. �e study design involved a MAD escalation study, in which patients were randomly assigned only to the lowest treatment group or placebo in a 3:1 ratio (active treatment or placebo) which then converted into a parallel enrolling, proof of concept (POC) dose ranging study (see Figure 2). Consistent with the MAD paradigm, dose escalation (allowing subjects in the next higher dose level to

receive a second dose) occurred only after the safety review on 6 subjects assigned to active treatment in the preceding dose had been performed. �is safetyreview was undertaken by a data monitoring committee. Only after it was determined by the data monitoring committee that two consecutive doses of the experimental treatment were well tolerated at a given dose level, were subjects already enrolled in the next higher dose treatment group permitted to receive a second dose of the investigational product. In Stage 2, a utility combining biomarker observations with early readout of a clinical endpoint at week 4 was used to drop dose levels that did not ful�l prespeci�ed necessary conditions (Proof-of-Non-Viability). Viable cohorts and comparator were kept open until each of the surviving treatment arms reached a prede�ned number of patients. �e endpoint to assess proof-of-concept was a 12 week observation on the regulatory accepted endpoint for Rheumatoid Arthritis (ACR20).

�is adaptive design provides several advantages over more traditional clinical trial approaches, without compromising patient safety:

• The seamless design increased the utility of the• information obtained by allowing subjects in both• the MAD and POC stages to provide both • de�nitive safety and e�cacy data.

• Performing the MAD component in subjects with• rheumatoid arthritis receiving concomitant • methotrexate allowed for the earliest • characterization of safety and antigenicity of the• new compound in the clinically relevant • population.

• The adaptive design minimized the number of• subjects that were exposed to ine�ective doses of the• drug, while simultaneously focusing subjects to• doses that were most informative for accurate dose• selection for subsequent con�rmatory trials.

FIGURE 2: Trial Design for Seamless MAD/POC Case Study �erefore, the adaptive design optimizes the bene�t/risk balance for participating subjects via improved e�ciency of decision making in relation to the doses of the new drug studied. In this example the estimated cost saving through combining two trials into one was $1.2m and the development programme was accelerated by 9 months.

“The Development Programme was Accelerated by 9 Months and the Cost Saving of Combining Two Trials in One was Estimated at $1.2M”

PROOF OF CONCEPT AND DOSE-RANGING STUDY DESIGNS

An example of a proof of concept and dose-ranging study design is a dose-�nding study to evaluate the analgesic e�cacy and safety of a single dose of a new non-steroidal anti-in�ammatory drug in subjects experiencing pain after oral dental surgery.4 Six doses of the new drug, placebo and an active comparator are considered. �e adaptive design consists of three stages. At the �rst stage, subjects are equally randomized into three arms: placebo, active comparator, and the top dose of the new drug. At the �rst interim analysis, the study may be stopped either because of lack of assay sensitivity, i.e. inability to establish a prede�ned minimal bene�t of the comparator over and above placebo, or evidence that

the top dose of study drug could not achieve a minimal acceptable level of total pain relief at 8 hours. If the study continues to the second stage, additional subjects would be randomized to placebo, active control, and three doses of the new drug (low, medium, and high) in a 1:1:2:2:1 ratio. Again, the study could be stopped due to lack of assay sensitivity or futility. Otherwise, a D-optimal design will be used to determine both the doses of the new drug (among six possible) and subject allocation ratios for the third stage, see Figure 3. At the end of the study, a four parameter logistic model is used to �t the dose response and make recommendations about the dose to be taken in phase III.

FIGURE 3: Trial Design for POC and Adaptive Dose-Ranging Case Study

CASE STUDY

A full case study has been published in the pain �eld where a Bayesian adaptive dose ranging design was used to evaluate the e�cacy of a new analgesic drug in a proof-of-concept post-herpetic neuralgia trial.5 In total 7 doses of the drug were evaluated in an attempt to de�ne the e�ective dose response curve. �e maximum sample size was 280 patients, however, at the �rst interim analysis (80 patients) the study was stopped for futility as none of the doses showed any meaningful di�erence from placebo. �e study was formally stopped after 133 patients had been enrolled. Stopping of this trial saved an estimated direct grant cost of $0.76m with a further $1.3m savings in internal costs.

“Early Stopping of POC Trials Because of Lack of Efficacy Saves Significant Cost and Enables Valuable Resources to be Re-assigned to Alternative Programmes That May Have a Better Chance of Success”

A recent unpublished example shows that early stopping of a POC adaptive dose ranging trial in post-operative nausea and vomiting saved $0.5m in outsourced costs. In this example, 131 out of 200 patients had been enrolled when the decision was taken at the �rst interim analysis.

RESPONSE ADAPTIVE DOSE-RANGING STUDY DESIGNS

Recent simulation studies6,7 conducted by the PhRMA Working Group on Adaptive Dose-Ranging Studies have found that response-adaptive dose allocation designs are generally superior in performance to conventional pairwise comparisons approaches. Learning about the research question such as identifying a target dose, or estimating the dose-response, is optimized with these designs.

�e main objectives in an adaptive dose-ranging study are to detect the dose response, to determine if any doses(s) meets clinical relevance, to estimate the dose-response, and then to decide on the dose(s) (if any) to take into the con�rmatory Phase III. With adaptive designs for dose-ranging studies many more doses may be considered than in a conventional parallel group design without increasing sample size. �is is achieved mainly through an adaptive allocation rule that is changing the randomization ratio to di�erent treatment arms during the study and putting more subjects in the region that allow most accurate estimation of dose-response and better precision in selecting the target dose. For example, Orlo� et al.8 showed that “for half the sample size, the adaptive design is as powerful and e�cient as the standard approach”.

• Single ascending dose escalation designs• Up-and-Down and CRM to find MTD• Establish Proof-of-Mechanism or Proof-of-Target Modulation

First-in Human

• Two-stage adaptive approach in patients• 1st stage – to identify MTD • 2nd stage – to select dose and exposure levels (necessary cond.)

MAD and PoC

• Start with the highest feasible tolerated dose and placebo• If a pre-specified futility condition is satisfied > stop • Otherwise, open enrollment to lower doses

PoC and ADRS

• SAD or MAD combined with Biomaker-based Efficacy• To identify the Optimal Safe Dose

Seamless PhaseI/II Design

• Finding a target dose (MED, EDp)• Response Adaptive Allocation• Covariate Adjusted Response Adaptive Allocation

Adaptive DoseRanging Design

CRM: Continual Reassessment Method; MTD: Maximum Tolerated Dose; MAD: Multiple Ascending Dose;SAD: Single Ascending Dose; MED: Minimum Effective Dose; EDp: Dose achieving 100p% of maximum effect

TherapeuticArea

Adaptive Design Trial Outcome Metrics

RheumatoidArthritis

Post-OperativeNausea &Vomiting

Post-HerpeticNeuralgia

Schizophrenia

Adaptive combinedMultiple Ascending Dose and Proof of Concept Study

Proof of concept andadaptive dose-ranging study design

Proof of concept andadaptive dose-ranging study design

Adaptive dose-ranging

Successful trial –criteria met for theachievement ofclinical POC inRheumatoid Arthritis

Early stopping forfutility

Early stopping forfutility

Early stopping forfutility

Programmeaccelerated by 9-month; $1.2m saving by combining two trials in one

$0.5m savings andredeployment of resource

$2m savings and re-deployment of resource

$14m savings and re-deployment of resource

Page 6: DESIGN AND EXECUTION OF EARLY PHASE …...Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs

DESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALSDESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALS

PAGE 5 PAGE 6

EXECUTIVE SUMMARY

�e adoption of an adaptive design strategy across the product development process brings a number of important bene�ts. �ese include enhanced R&D e�ciency, enhanced R&D productivity and, importantly, increased probability of success at phase III. �e opportunity to manage risk starts in the early phase of a development programme, with a diligent view to appropriate dose selection which impacts not only the success of the downstream clinical trial programme, but also manufacturing costs and value analyses.

Adaptive design trials enhance R&D e�ciency by reducing the need to repeat trials that just miss their clinical end-point or fail to identify the e�ective dose response at the �rst attempt. By avoiding the need to run these trials again, signi�cant cost and time savings are achieved. �is is possible through use of adaptive designs that enable additional patients to be added to achieve statistical signi�cance the �rst time around or by allowing a wider dose range to be studied and then picking doses early in the trial that are in the optimum part of the dose response curve. In addition, early stopping of development programmes because a product is ine�ective enables scarce resources to be redeployed in additional trials which may show more promise. All of these factors increase development e�ciency.

Adaptive designs increase R&D productivity by enabling more accurate de�nition of the e�ective dose response at phase II which enables better design of the pivotal phase III programme which in turn increases the probability of success of the overall development programme. A number of phase III trials fail because the dose is either too high and causes unwanted safety issues or is too low to show su�cient e�cacy. Adaptive design trials enable optimized dose selection before the pivotal trial phase is started.

Another opportunity provided by adaptive design is population enrichment where drug response can be

optimised to speci�c patient sub-populations that respond better to treatment. Many phase III studies fail because the overall e�cacy of treatment is diluted as a consequence of the drug being evaluated in the full trial population rather than in the speci�c subset where the drug works best. Adaptive design enables early selection of the appropriate patient population and increases the probability of success.

“Adaptive Design Enables Selection of the Right Dose for the Right Patient Before Commitment to Expensive Phase III Trials”

Phases I and II are critical steps in the clinical development process as this is where important information about the product has to be generated and assessed before the decision is taken to commit to expensive phase III pivotal studies. �is early phase of development is known as the “Learn Phase” and the data generated relates to the e�ective dose-response, the safety pro�le and therapeutic index, appropriate endpoints, and the patient population best treated by the product under evaluation.

Adaptive designs have an important role to play in both phase I and phase II, but it is in the latter phase where they add the greatest value in terms of risk management. Validated methodologies are available to design early phase adaptive trials and this is an area of signi�cant interest across the pharmaceutical industry.

“Getting it Right at Phase II is a Major Objective for Adaptive Design and is the Point at which Phase III Success is Defined”

Companies that adopt a comprehensive adaptive design strategy in the “Learn” phase will make better development decisions, will build development e�ciency, will increase the probability of success of the molecules that enter phase III, ultimately bringing e�ective products to the market more e�ciently. Competitor companies that persevere with conventional trial designs will remain slow, ine�cient

candidate optimisation to Proof of concept stages. Moira assists partner companies with both acquisition and due diligence of in licensed compounds and strategic support for the out-licensing of successful candidates. �is has included design, planning and management of, strategic and clinical development plans, due diligence, decision analysis, preclinical, POC and Phase II programmes in a range of, indications including respiratory, anti-infective, immunomodulation, CNS and cardiovascular. �is experience covers a large range of targets and mecha-nisms of action. In order to support the strategic and clinical development planning Moira has worked extensively with international Key Opinion Leaders in respiratory, anti-infective and transplant indications.

PHIL BIRCH,SVP, CORPORATE BRAND MANAGER

Dr. Phil Birch is a Senior Vice President at Aptiv Solutions and has global responsibility for co-ordinating market education and client awareness in the adaptive clinical trials �eld. Dr. Birch has worked in the pharmaceutical industry for over 27 years and has held executive positions in corporate development, business development and R&D in a number of consulting, biotechnology and top 10 pharmaceutical companies.

REFERENCES

1. Dragalin V. Adaptive Designs: Terminology and Classi�cation. Drug Information Journal, 2006; 40: 425-436. 2. Gallo P, Chuang-Stein C, Dragalin V, Gaydos B, Krams M, Pinheiro J. Executive Summary of the PhRMA Working Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs for Phase I Oncology Trials, Clinical Operations Summit, Brussels, 2012).4. Dragalin V, Hsuan F, and Padmanabhan SK. Adaptive Designs for Dose-Finding Studies Based on Sigmoid Emax Model. J Biopharmaceutical Statistics 2007; 17:1051-1070.5. Smith MK, Jones I, Morris MF, Grieve A and Tan K. Implementation of a Bayesian adaptive design in a proof of concept study. Pharmaceutical Statistics 2006; 5:39-50. 6. Bornkamp B, Bretz F, Dmitrienko A, Enas G, Gaydos B, Hsu CH, König F, Krams M, Liu Q, Neuenschwander B, Parke T, Pinheiro J, Roy A, Sax R, Shen F. Innovative approaches for designing and analyzing adaptive dose-ranging trials: White Paper from the PhRMA working group on “Adaptive Dose-Ranging Studies”. J of Biopharmaceutical Statistics 2007; 17: 965–995. 7. Dragalin V, Bornkamp B, Bretz F, Miller F, Padmanabhan SK, Patel N, Perevozskaya I, Pinheiro J, and Smith J. A Simulation Study to Compare New Adaptive Dose–Ranging Designs. Statistics in Biopharmaceutical Research 2010; 2:487-512.8. Orlo� J, Douglas F, Pinheiro J, Levinson S, Branson M, Chaturvedi P, Ette E, Gallo P, Hirsch G, Mehta C, Patel N, Sabir S, Springs S, Stanski D, Golub H, Evers M, Fleming E, Singh N, Tramontin T. �e future of drug development: advancing clinical trial design. Nature Review-Drug Discovery. 2009; 8:940-957.9. Padmanabhan SK and Dragalin V. Adaptive Dc-optimal designs for dose �nding based on a continuous e�cacy endpoint. Biometrical Journal 2010; 52:836-852. 10. Grieve AP and Krams M. ASTIN: a Bayesian adaptive dose-response trial in acute stroke. Clinical Trials 2005; 2:340-351. 11. Bornkamp B, Pinheiro J, Bretz F. MCPMod: An R Package for the Design and Analysis of Dose-Finding Studies. Journal of Statistical Software 2009; 29:1-23.12. Shen J, Preskorn S, Dragalin V, Slomkowski M, Padmanabhan SK, Fardipour P, Sharma A, and Krams M. How Adaptive Trial Designs Can Increase E�ciency in Psychiatric Drug Development: A Case Study. Innovations in Clinical Neuroscience 2011; 8: 26-34.13. Berry SM, Spinelli W, Littman GS, Liang JZ, Fardipour P, Berry DA, Lewis RL, and Krams M. A Bayesian dose-�nding trial with adaptive dose expansion to �exibly assess e�cacy and safety of an investigational drug. Clinical Trials 2010; 7: 121-135.

ABOUT THE AUTHORS

VLAD DRAGALIN, SVP, INNOVATION CENTRE

Dr. Vlad Dragalin is a well-known adaptive design expert with 25 years of experience in developing the statistical methodology of adaptive designs and with over 12 years experience in pharmaceutical industry. Vlad is a Senior Vice President, Software Develop-ment and Consulting, in the Innovation Center at Aptiv Solutions. Vlad works with clients to improve the drug development process through the use of adaptive designs and to assist them with the trial design, simulation and execution, developing adaptive trials software tools, and working with project teams in clinical operations, data management and statistics during study conduct.

SARAH ARBE-BARNES,SVP, TRANSLATIONAL SCIENCES

Dr. Sarah Arbe-Barnes has over 24 years of experience in drug development and project leadership in the Pharma and service sector industry to include in/out-licensing and due diligence, strategic development and business planning, fundraising for start-ups/SMEs( not-for-pro�t sector and VC), along with optimized market access strategies and marketing. Sarah has led recent product approval programmes, overseeing a broad range of clinical and nonclinical project design and management activities from the point of candidate selection. Sarah heads the Transla-tional Sciences division of Aptiv Solutions, which is a dedicated global consultancy group, focused on the provision of drug and device development advice and expertise for small molecules and biologics.

MOIRA THOMSON,VP, TRANSLATIONAL SCIENCES

Moira �omson has over 15 years of experience in the Pharmaceutical Industry, and is involved in the design, planning and management of global develop-ment projects for a mixture of small and medium sized Pharma, and Biotech companies covering lead

and will fail to make the commercial returns that investors and shareholders demand. “Adaptive Design Builds Competitive Advantage”

�is White Paper summarizes the important aspects of early phase adaptive design and provides a summary of selected case studies which demonstrate the value of this innovative approach. It has been written to inform senior R&D decision-makers about the critical role of adaptive design in early phase development and the signi�cant bene�ts that this approach can bring.

BACKGROUND TO EARLY PHASE ADAPTIVE TRIAL DESIGN

Adaptive Clinical Trials are viewed as supporting a change in the development process as focus moves from blockbuster products to more specialized medicines that require a �exible cost-e�cient approach to investigating their e�ectiveness. Adaptive designs are statistical methodologies applied to speci�c stages of drug development where real time learning from accumulating trial data is applied to optimize subsequent study execution.1 �e bene�t of an Adaptive Clinical Trial lies in the monitoring of data to make design modi�cation adjustments to an aspect of a trial leading to quicker development decisions around key go/no-go milestones that impact time to market and/or minimise costly R&D losses.

�e adaptations are prospectively de�ned prior to the start of the trial and can include stopping early either for futility or success, expanding the sample size due to greater than expected data variability, or allocating patients preferentially to treatment regimens with a better therapeutic index. Virtually any aspect of the trial may be the potential target of design modi�cations. Importantly, these modi�cations are a design feature aimed to enhance the trial, not a remedy for inadequate planning. �ey are not ad hoc study corrections made via protocol amendment.2

�e speci�c adaptations considered, and the basis of their implementation, are carefully de�ned based on strict rules, justi�ed statistically and scienti�cally, and are an integral part of the �nal, pre-recruitment trial protocol. Failing this would compromise the interpretation and acceptance of study results.

Adaptive designs often employ frequent interim analyses of all accumulated data (and, possibly, external trial data) to determine whether pre-planned design modi�cations will be ‘triggered’. Interim analyses partition the trial into multiple stages, each trial stage’s characteristics (number of arms, number of patients to be enrolled, their allocation between arms, stage duration, etc.) de�ned by the preceding interim analysis results. �e ability to periodically, or even continually, examine available data to determine whether trial modi�cations are necessary and implement pre-de�ned design changes when indicated, gives adaptive design its strength and �exibility.

�e use of adaptive design in the exploratory stage of clinical trials can increase the e�ciency of drug development by improving our ability to e�ciently learn about the dose-response and better determine whether to take a drug (and the right dose of the drug and in the right population) forward into later phase testing. �ese designs explicitly address multiple trial goals, adaptively allocate subjects according to ongoing information needs, and allow termination for both early success and futility considerations. �is approach can maximize the ability to test a larger number of doses in a single trial while simultaneously increasing the e�ciency of the trial in terms of making better go/no-go decisions about continuing the trial and/or the development of the drug for a speci�c indication or sub-population. During the “Learn” phase many aspects of the study design can be modi�ed: number of subjects, study duration, endpoint selection, treatment duration, number of treatments, patient population. �e adaptation of multiple trial features that are a legitimate concern in con�rmatory studies can become, in fact, an

appealing feature of adaptive designs in exploratory development, leading to enhanced learning to set the appropriate stage for con�rmatory trials.

“The Use of Adaptive Design in the Learn Phase Offers the Best Opportunity to Assess the Product Characteristics that Determine Success”

Judicious use of adaptive designs may increase the information value per resource unit invested by avoiding allocation of patients to non-e�cacious/unsafe therapies and allowing stopping decisions to be made at the earliest possible time point. Ultimately this will accelerate the development of promising therapies to bene�t patients.

Adaptive Clinical Trials can result in the more ethical treatment of patients from at least two perspectives. First, a larger number of patients can be randomized to more favorable, e�ective doses with fewer patients exposed to less e�ective doses. In addition, there is the potential for including fewer total patients in the early stages of the development programmes with attending lower risk of exposure to adverse events. A greater saving in patients may be obtained at the programme level rather than the individual study level as adaptive methodology applied across a programme will reduce the need to repeat ine�ective or inconclusive studies.

Adaptive designs represent an advanced methodology to support drug development. A prerequisite for their successful implementation is an understanding of the underlying methodology and their impact on the logistics of the trial, such as data management and monitoring procedures, electronic data capture/interactive web response services, drug supply management, and data-monitoring committee operating procedures. �ese designs have an impact on drug development strategies, trial protocols, clinical trial material doses and availability, informed consent forms, data analysis, and reporting plans. Adaptive design will also change the dynamics of

enrollment, randomization, and data capturing process, monitoring, and data cleaning systems. To be able to make informed decisions, the sponsor must have access to real time clinical data from all sources and be able to easily review and assess this data.

�is in turn requires �exibility through:

• Data management systems for rapid clinical data• access and data cleaning

• Drug supply systems for drug product planning • and management

• Optimal systems providing randomization, • medication kit management and emergency• unblinding

Planning of Adaptive Clinical Trials is a key feature in the success of such innovative approaches. Determining the optimal characteristics of the study design can be a complex yet critical decision that may require more upfront planning before the protocol can be �nalized. However, there are several commercial software packages that provide a powerful tool for planning, simulation and analysis of complex adaptive designs. FACTS™ is comprehensive software for adaptive designs in the “Learn” phase and includes early stage dose-escalation designs, adaptive dose-ranging studies, interim decision rules based on both e�cacy and safety responses, and population enrichment. ADDPLAN® is a validated software package for adaptive design in the “Con�rm” phase and incorporates sample size re-estimation, adaptive group sequential designs, multiple comparison procedures for multi-armed adaptive trials, including treatment selection designs, �exible combination of clinical research phases, and population enrichment designs.

TYPES OF ADAPTIVE DESIGNS IN “LEARN” STUDIES

Examples of adaptive designs in the Learn phase are shown in Figure 1. Phase I studies are already

FIGURE 5: Adaptive Design Creates Value at Level of the Single Trial, DevelopmentProgramme and Product Pipeline

�is aspect is discussed further in an accompanying white paper that explores the strategic utilisation of adaptive design at the development programme and product pipeline level. Importantly, recent data from Merck has indicated that adoption of an early phase adaptive trial strategy at this level saves at least $200m annually (Schindler J, Disruptive Innovations Conference, Boston, 2012).

“Adaptive Dose Ranging Studies Build Significant Efficiencies at Phase II and Optimize Identification of the Optimal Dose for Confirmatory Studies”

Several adaptive designs have been proposed and some already applied in practice using di�erent working models for the design engine: parsimonious monotonic sigmoid Emax9, Normal Dynamic Linear Model without monotonicity restrictions10, several simple models combined in a Multiple Comparison Procedure (MCP Mod).11

CASE STUDY

�e objective of this trial was to establish the correct dose(s) to take forward into a con�rmatory trial on a novel anti-psychotic drug in acute schizophrenia. �ere was data indicating that the dose-response might be non-monotonic. �erefore, the working model chosen for the design was the Normal Dynamic Linear Model, because it is �exible and robust, and also allows for capturing non-monotonic relationships. Patients were allocated to placebo, active comparator and seven doses of study drug. �e adaptive allocation was targeting both the minimum e�cacious dose (MED) and the dose achieving the maximum response (Dmax) on the primary endpoint, positive PANSS score at week 4. Weekly data updates were used to change the allocation and check early stopping rules based on predictive probabilities. A stopping rule for futility was based on a prespeci�ed threshold that no treatment dose achieves the clinically signi�cant di�erence (CSD) over placebo. A stopping rule for e�cacy required su�cient knowledge about the MED and Dmax and a prespeci�ed con�dence that the Dmax achieves CSD. For details, see Shen et al.12 �e adaptive design allowed earlier (5 months) decision making with a 99% con�dence in futility decision with $14m savings in direct grant cost. �e design, implementation, and outcome of a response adaptive dose-ranging trial of an investigational drug in patients with diabetes, incorporating a

dose expansion approach to �exibly address bothe�cacy and safety has also been reported recently by Berry et al.13 �e early (2 months) futility decision was achieved with a $3.6m saving in direct cost and additional cost savings in avoiding expensive production of phase III drug supplies.

FIGURE 4: Summary of Early Phase Case Studies

SUMMARY

Early phase adaptive design trials address a number of critical development questions that need to be answered before commitment is made to phase III con�rmatory trials. �e application of these innovative designs early in the product development process enhances clinical trial e�ciency, productivity and increases the probability of success at phase III.

�e deployment of these approaches across a pipeline of products brings substantial bene�ts to sponsor companies and enables more e�ective portfolio management and ultimately increases portfolio value (Figure 5).

adaptive by nature, but more recent adaptive designsprovide better estimate of drug safety (e.g. the maxi-mum tolerated dose) and better understanding of itsPK/PD characteristics. Combining the objectives of conventional Phase IIa and IIb studies in a single trial provides the obvious bene�t of reducing the timeline by running the two studies seamlessly under a single protocol with the same clinical team. Furthermore, such approaches achieve trial e�ciency by the prospective integration of data from both IIa and IIb in the �nal analysis.

FIRST-IN-HUMAN SINGLE ASCENDING DOSE ESCALATION DESIGNS

In contrast to traditional dose escalation designs, where for example, cohorts of subjects are allocated to ascending doses (and placebo) until a maximum tolerated dose (MTD) is empirically determined, the continual reassessment method (CRM) models the dose- and/or exposure-toxicity relationship to focus on the identi�cation of the dose-range of interest near the MTD and for this to be explored further. �ese adaptive dose escalation designs can be expanded to also e�ciently establish proof-of-mechanism or proof-of-target modulation, if validated biomarker endpoints are available.

FIGURE 1: Types of Adaptive Designs in the Learn Phase

�e traditional First in Human study objectives of safety and tolerability can be extended to include an

assessment of proof-of-mechanism in the very �rst study conducted with the investigational compound. �e ability to assess proof-of-mechanism in this earlydevelopmental setting provides an opportunity to greatly enhance the clinical development strategy for the compound. �e adaptation is intended to quickly hone in on the predicted dose-range of interest. �is avoids the risks associated with exposing patients in an ine�ective low dose range, whilst minimizing exposure to high doses with an unacceptable tolerability and safety pro�le.

“CRM Offers a Better Approach to Accurately Determining the MTD in Early Phase Oncology Studies”

CRM models are becoming routine for early phase oncology studies where accurate de�nition of the MTD is critical for selecting the optimum dose for e�cacy assessment. Conventional 3+3 designs have been shown to signi�cantly underestimate the MTD which increases the risk of poor outcome in e�cacy trials.3

SEAMLESS MULTIPLE ASCENDING DOSE AND PROOF-OF-CONCEPT STUDY DESIGNS

�e objective of seamless multiple ascending dose and proof-of-concept studies is to combine a multiple ascending dose (MAD) study in patients with the proof-of-concept of the investigational compound.

CASE STUDY

A good example is seen in the treatment of rheumatoid arthritis. �e study design involved a MAD escalation study, in which patients were randomly assigned only to the lowest treatment group or placebo in a 3:1 ratio (active treatment or placebo) which then converted into a parallel enrolling, proof of concept (POC) dose ranging study (see Figure 2). Consistent with the MAD paradigm, dose escalation (allowing subjects in the next higher dose level to

receive a second dose) occurred only after the safety review on 6 subjects assigned to active treatment in the preceding dose had been performed. �is safetyreview was undertaken by a data monitoring committee. Only after it was determined by the data monitoring committee that two consecutive doses of the experimental treatment were well tolerated at a given dose level, were subjects already enrolled in the next higher dose treatment group permitted to receive a second dose of the investigational product. In Stage 2, a utility combining biomarker observations with early readout of a clinical endpoint at week 4 was used to drop dose levels that did not ful�l prespeci�ed necessary conditions (Proof-of-Non-Viability). Viable cohorts and comparator were kept open until each of the surviving treatment arms reached a prede�ned number of patients. �e endpoint to assess proof-of-concept was a 12 week observation on the regulatory accepted endpoint for Rheumatoid Arthritis (ACR20).

�is adaptive design provides several advantages over more traditional clinical trial approaches, without compromising patient safety:

• The seamless design increased the utility of the• information obtained by allowing subjects in both• the MAD and POC stages to provide both • de�nitive safety and e�cacy data.

• Performing the MAD component in subjects with• rheumatoid arthritis receiving concomitant • methotrexate allowed for the earliest • characterization of safety and antigenicity of the• new compound in the clinically relevant • population.

• The adaptive design minimized the number of• subjects that were exposed to ine�ective doses of the• drug, while simultaneously focusing subjects to• doses that were most informative for accurate dose• selection for subsequent con�rmatory trials.

FIGURE 2: Trial Design for Seamless MAD/POC Case Study �erefore, the adaptive design optimizes the bene�t/risk balance for participating subjects via improved e�ciency of decision making in relation to the doses of the new drug studied. In this example the estimated cost saving through combining two trials into one was $1.2m and the development programme was accelerated by 9 months.

“The Development Programme was Accelerated by 9 Months and the Cost Saving of Combining Two Trials in One was Estimated at $1.2M”

PROOF OF CONCEPT AND DOSE-RANGING STUDY DESIGNS

An example of a proof of concept and dose-ranging study design is a dose-�nding study to evaluate the analgesic e�cacy and safety of a single dose of a new non-steroidal anti-in�ammatory drug in subjects experiencing pain after oral dental surgery.4 Six doses of the new drug, placebo and an active comparator are considered. �e adaptive design consists of three stages. At the �rst stage, subjects are equally randomized into three arms: placebo, active comparator, and the top dose of the new drug. At the �rst interim analysis, the study may be stopped either because of lack of assay sensitivity, i.e. inability to establish a prede�ned minimal bene�t of the comparator over and above placebo, or evidence that

the top dose of study drug could not achieve a minimal acceptable level of total pain relief at 8 hours. If the study continues to the second stage, additional subjects would be randomized to placebo, active control, and three doses of the new drug (low, medium, and high) in a 1:1:2:2:1 ratio. Again, the study could be stopped due to lack of assay sensitivity or futility. Otherwise, a D-optimal design will be used to determine both the doses of the new drug (among six possible) and subject allocation ratios for the third stage, see Figure 3. At the end of the study, a four parameter logistic model is used to �t the dose response and make recommendations about the dose to be taken in phase III.

FIGURE 3: Trial Design for POC and Adaptive Dose-Ranging Case Study

CASE STUDY

A full case study has been published in the pain �eld where a Bayesian adaptive dose ranging design was used to evaluate the e�cacy of a new analgesic drug in a proof-of-concept post-herpetic neuralgia trial.5 In total 7 doses of the drug were evaluated in an attempt to de�ne the e�ective dose response curve. �e maximum sample size was 280 patients, however, at the �rst interim analysis (80 patients) the study was stopped for futility as none of the doses showed any meaningful di�erence from placebo. �e study was formally stopped after 133 patients had been enrolled. Stopping of this trial saved an estimated direct grant cost of $0.76m with a further $1.3m savings in internal costs.

“Early Stopping of POC Trials Because of Lack of Efficacy Saves Significant Cost and Enables Valuable Resources to be Re-assigned to Alternative Programmes That May Have a Better Chance of Success”

A recent unpublished example shows that early stopping of a POC adaptive dose ranging trial in post-operative nausea and vomiting saved $0.5m in outsourced costs. In this example, 131 out of 200 patients had been enrolled when the decision was taken at the �rst interim analysis.

RESPONSE ADAPTIVE DOSE-RANGING STUDY DESIGNS

Recent simulation studies6,7 conducted by the PhRMA Working Group on Adaptive Dose-Ranging Studies have found that response-adaptive dose allocation designs are generally superior in performance to conventional pairwise comparisons approaches. Learning about the research question such as identifying a target dose, or estimating the dose-response, is optimized with these designs.

�e main objectives in an adaptive dose-ranging study are to detect the dose response, to determine if any doses(s) meets clinical relevance, to estimate the dose-response, and then to decide on the dose(s) (if any) to take into the con�rmatory Phase III. With adaptive designs for dose-ranging studies many more doses may be considered than in a conventional parallel group design without increasing sample size. �is is achieved mainly through an adaptive allocation rule that is changing the randomization ratio to di�erent treatment arms during the study and putting more subjects in the region that allow most accurate estimation of dose-response and better precision in selecting the target dose. For example, Orlo� et al.8 showed that “for half the sample size, the adaptive design is as powerful and e�cient as the standard approach”.

STA

GE

1ST

AG

E 2

Ascending MAD Until doses are open

Safety Decision: Subjects will receive a 2nddose only after a safety review of the 2nd dosein the preceding cohort.

The dose regimen is Q4 Weeks for 4 cycles.

2nd Stage begins after escalatingto the maximum tolerated dose

N=33 patients for each offive doses (10, 30, 50, 60,100 mg) and Placebo

Futility Decision: Based on ACR20and 25% reduction in CRP at 4 weeks

Internal DMC for safety & futility decisions:

Unblinded Medical Monitor Unblinded Biostatistion

Cohort 5: 100mg (n=3) : Placebo (n=1)

Cohort 4: 60mg (n=3) : Placebo (n=1)

Cohort 3: 50mg (n=3) : Placebo (n=1)

Cohort 2: 30mg (n=3) : Placebo (n=1)

Cohort 1: 10mg (n=3) : Placebo (n=1)

Stage I

DRG 900mg

DRG 750mg

DRG 600mg

DRG 450mg

DRG 300mg

DRG 150mg

Plbo

Ctrl

Total Ssize ~ 30:10 pats/arm

Stage II Stage III• Fit the Model• Find the D-Optimal Design• Determine Sample Size

DRG 900mg

DRG 750mg

DRG 600mg

DRG 450mg

DRG 300mg

DRG 150mg

Plbo

Ctrl

Total Ssize ~ 65:5:10 pats/arm

STOPFutility

STOPNAS

IA 1

STOPFutility

STOPNAS

IA II

Page 7: DESIGN AND EXECUTION OF EARLY PHASE …...Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs

DESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALSDESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALS

PAGE 5 PAGE 6

EXECUTIVE SUMMARY

�e adoption of an adaptive design strategy across the product development process brings a number of important bene�ts. �ese include enhanced R&D e�ciency, enhanced R&D productivity and, importantly, increased probability of success at phase III. �e opportunity to manage risk starts in the early phase of a development programme, with a diligent view to appropriate dose selection which impacts not only the success of the downstream clinical trial programme, but also manufacturing costs and value analyses.

Adaptive design trials enhance R&D e�ciency by reducing the need to repeat trials that just miss their clinical end-point or fail to identify the e�ective dose response at the �rst attempt. By avoiding the need to run these trials again, signi�cant cost and time savings are achieved. �is is possible through use of adaptive designs that enable additional patients to be added to achieve statistical signi�cance the �rst time around or by allowing a wider dose range to be studied and then picking doses early in the trial that are in the optimum part of the dose response curve. In addition, early stopping of development programmes because a product is ine�ective enables scarce resources to be redeployed in additional trials which may show more promise. All of these factors increase development e�ciency.

Adaptive designs increase R&D productivity by enabling more accurate de�nition of the e�ective dose response at phase II which enables better design of the pivotal phase III programme which in turn increases the probability of success of the overall development programme. A number of phase III trials fail because the dose is either too high and causes unwanted safety issues or is too low to show su�cient e�cacy. Adaptive design trials enable optimized dose selection before the pivotal trial phase is started.

Another opportunity provided by adaptive design is population enrichment where drug response can be

optimised to speci�c patient sub-populations that respond better to treatment. Many phase III studies fail because the overall e�cacy of treatment is diluted as a consequence of the drug being evaluated in the full trial population rather than in the speci�c subset where the drug works best. Adaptive design enables early selection of the appropriate patient population and increases the probability of success.

“Adaptive Design Enables Selection of the Right Dose for the Right Patient Before Commitment to Expensive Phase III Trials”

Phases I and II are critical steps in the clinical development process as this is where important information about the product has to be generated and assessed before the decision is taken to commit to expensive phase III pivotal studies. �is early phase of development is known as the “Learn Phase” and the data generated relates to the e�ective dose-response, the safety pro�le and therapeutic index, appropriate endpoints, and the patient population best treated by the product under evaluation.

Adaptive designs have an important role to play in both phase I and phase II, but it is in the latter phase where they add the greatest value in terms of risk management. Validated methodologies are available to design early phase adaptive trials and this is an area of signi�cant interest across the pharmaceutical industry.

“Getting it Right at Phase II is a Major Objective for Adaptive Design and is the Point at which Phase III Success is Defined”

Companies that adopt a comprehensive adaptive design strategy in the “Learn” phase will make better development decisions, will build development e�ciency, will increase the probability of success of the molecules that enter phase III, ultimately bringing e�ective products to the market more e�ciently. Competitor companies that persevere with conventional trial designs will remain slow, ine�cient

candidate optimisation to Proof of concept stages. Moira assists partner companies with both acquisition and due diligence of in licensed compounds and strategic support for the out-licensing of successful candidates. �is has included design, planning and management of, strategic and clinical development plans, due diligence, decision analysis, preclinical, POC and Phase II programmes in a range of, indications including respiratory, anti-infective, immunomodulation, CNS and cardiovascular. �is experience covers a large range of targets and mecha-nisms of action. In order to support the strategic and clinical development planning Moira has worked extensively with international Key Opinion Leaders in respiratory, anti-infective and transplant indications.

PHIL BIRCH,SVP, CORPORATE BRAND MANAGER

Dr. Phil Birch is a Senior Vice President at Aptiv Solutions and has global responsibility for co-ordinating market education and client awareness in the adaptive clinical trials �eld. Dr. Birch has worked in the pharmaceutical industry for over 27 years and has held executive positions in corporate development, business development and R&D in a number of consulting, biotechnology and top 10 pharmaceutical companies.

REFERENCES

1. Dragalin V. Adaptive Designs: Terminology and Classi�cation. Drug Information Journal, 2006; 40: 425-436. 2. Gallo P, Chuang-Stein C, Dragalin V, Gaydos B, Krams M, Pinheiro J. Executive Summary of the PhRMA Working Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs for Phase I Oncology Trials, Clinical Operations Summit, Brussels, 2012).4. Dragalin V, Hsuan F, and Padmanabhan SK. Adaptive Designs for Dose-Finding Studies Based on Sigmoid Emax Model. J Biopharmaceutical Statistics 2007; 17:1051-1070.5. Smith MK, Jones I, Morris MF, Grieve A and Tan K. Implementation of a Bayesian adaptive design in a proof of concept study. Pharmaceutical Statistics 2006; 5:39-50. 6. Bornkamp B, Bretz F, Dmitrienko A, Enas G, Gaydos B, Hsu CH, König F, Krams M, Liu Q, Neuenschwander B, Parke T, Pinheiro J, Roy A, Sax R, Shen F. Innovative approaches for designing and analyzing adaptive dose-ranging trials: White Paper from the PhRMA working group on “Adaptive Dose-Ranging Studies”. J of Biopharmaceutical Statistics 2007; 17: 965–995. 7. Dragalin V, Bornkamp B, Bretz F, Miller F, Padmanabhan SK, Patel N, Perevozskaya I, Pinheiro J, and Smith J. A Simulation Study to Compare New Adaptive Dose–Ranging Designs. Statistics in Biopharmaceutical Research 2010; 2:487-512.8. Orlo� J, Douglas F, Pinheiro J, Levinson S, Branson M, Chaturvedi P, Ette E, Gallo P, Hirsch G, Mehta C, Patel N, Sabir S, Springs S, Stanski D, Golub H, Evers M, Fleming E, Singh N, Tramontin T. �e future of drug development: advancing clinical trial design. Nature Review-Drug Discovery. 2009; 8:940-957.9. Padmanabhan SK and Dragalin V. Adaptive Dc-optimal designs for dose �nding based on a continuous e�cacy endpoint. Biometrical Journal 2010; 52:836-852. 10. Grieve AP and Krams M. ASTIN: a Bayesian adaptive dose-response trial in acute stroke. Clinical Trials 2005; 2:340-351. 11. Bornkamp B, Pinheiro J, Bretz F. MCPMod: An R Package for the Design and Analysis of Dose-Finding Studies. Journal of Statistical Software 2009; 29:1-23.12. Shen J, Preskorn S, Dragalin V, Slomkowski M, Padmanabhan SK, Fardipour P, Sharma A, and Krams M. How Adaptive Trial Designs Can Increase E�ciency in Psychiatric Drug Development: A Case Study. Innovations in Clinical Neuroscience 2011; 8: 26-34.13. Berry SM, Spinelli W, Littman GS, Liang JZ, Fardipour P, Berry DA, Lewis RL, and Krams M. A Bayesian dose-�nding trial with adaptive dose expansion to �exibly assess e�cacy and safety of an investigational drug. Clinical Trials 2010; 7: 121-135.

ABOUT THE AUTHORS

VLAD DRAGALIN, SVP, INNOVATION CENTRE

Dr. Vlad Dragalin is a well-known adaptive design expert with 25 years of experience in developing the statistical methodology of adaptive designs and with over 12 years experience in pharmaceutical industry. Vlad is a Senior Vice President, Software Develop-ment and Consulting, in the Innovation Center at Aptiv Solutions. Vlad works with clients to improve the drug development process through the use of adaptive designs and to assist them with the trial design, simulation and execution, developing adaptive trials software tools, and working with project teams in clinical operations, data management and statistics during study conduct.

SARAH ARBE-BARNES,SVP, TRANSLATIONAL SCIENCES

Dr. Sarah Arbe-Barnes has over 24 years of experience in drug development and project leadership in the Pharma and service sector industry to include in/out-licensing and due diligence, strategic development and business planning, fundraising for start-ups/SMEs( not-for-pro�t sector and VC), along with optimized market access strategies and marketing. Sarah has led recent product approval programmes, overseeing a broad range of clinical and nonclinical project design and management activities from the point of candidate selection. Sarah heads the Transla-tional Sciences division of Aptiv Solutions, which is a dedicated global consultancy group, focused on the provision of drug and device development advice and expertise for small molecules and biologics.

MOIRA THOMSON,VP, TRANSLATIONAL SCIENCES

Moira �omson has over 15 years of experience in the Pharmaceutical Industry, and is involved in the design, planning and management of global develop-ment projects for a mixture of small and medium sized Pharma, and Biotech companies covering lead

and will fail to make the commercial returns that investors and shareholders demand. “Adaptive Design Builds Competitive Advantage”

�is White Paper summarizes the important aspects of early phase adaptive design and provides a summary of selected case studies which demonstrate the value of this innovative approach. It has been written to inform senior R&D decision-makers about the critical role of adaptive design in early phase development and the signi�cant bene�ts that this approach can bring.

BACKGROUND TO EARLY PHASE ADAPTIVE TRIAL DESIGN

Adaptive Clinical Trials are viewed as supporting a change in the development process as focus moves from blockbuster products to more specialized medicines that require a �exible cost-e�cient approach to investigating their e�ectiveness. Adaptive designs are statistical methodologies applied to speci�c stages of drug development where real time learning from accumulating trial data is applied to optimize subsequent study execution.1 �e bene�t of an Adaptive Clinical Trial lies in the monitoring of data to make design modi�cation adjustments to an aspect of a trial leading to quicker development decisions around key go/no-go milestones that impact time to market and/or minimise costly R&D losses.

�e adaptations are prospectively de�ned prior to the start of the trial and can include stopping early either for futility or success, expanding the sample size due to greater than expected data variability, or allocating patients preferentially to treatment regimens with a better therapeutic index. Virtually any aspect of the trial may be the potential target of design modi�cations. Importantly, these modi�cations are a design feature aimed to enhance the trial, not a remedy for inadequate planning. �ey are not ad hoc study corrections made via protocol amendment.2

�e speci�c adaptations considered, and the basis of their implementation, are carefully de�ned based on strict rules, justi�ed statistically and scienti�cally, and are an integral part of the �nal, pre-recruitment trial protocol. Failing this would compromise the interpretation and acceptance of study results.

Adaptive designs often employ frequent interim analyses of all accumulated data (and, possibly, external trial data) to determine whether pre-planned design modi�cations will be ‘triggered’. Interim analyses partition the trial into multiple stages, each trial stage’s characteristics (number of arms, number of patients to be enrolled, their allocation between arms, stage duration, etc.) de�ned by the preceding interim analysis results. �e ability to periodically, or even continually, examine available data to determine whether trial modi�cations are necessary and implement pre-de�ned design changes when indicated, gives adaptive design its strength and �exibility.

�e use of adaptive design in the exploratory stage of clinical trials can increase the e�ciency of drug development by improving our ability to e�ciently learn about the dose-response and better determine whether to take a drug (and the right dose of the drug and in the right population) forward into later phase testing. �ese designs explicitly address multiple trial goals, adaptively allocate subjects according to ongoing information needs, and allow termination for both early success and futility considerations. �is approach can maximize the ability to test a larger number of doses in a single trial while simultaneously increasing the e�ciency of the trial in terms of making better go/no-go decisions about continuing the trial and/or the development of the drug for a speci�c indication or sub-population. During the “Learn” phase many aspects of the study design can be modi�ed: number of subjects, study duration, endpoint selection, treatment duration, number of treatments, patient population. �e adaptation of multiple trial features that are a legitimate concern in con�rmatory studies can become, in fact, an

appealing feature of adaptive designs in exploratory development, leading to enhanced learning to set the appropriate stage for con�rmatory trials.

“The Use of Adaptive Design in the Learn Phase Offers the Best Opportunity to Assess the Product Characteristics that Determine Success”

Judicious use of adaptive designs may increase the information value per resource unit invested by avoiding allocation of patients to non-e�cacious/unsafe therapies and allowing stopping decisions to be made at the earliest possible time point. Ultimately this will accelerate the development of promising therapies to bene�t patients.

Adaptive Clinical Trials can result in the more ethical treatment of patients from at least two perspectives. First, a larger number of patients can be randomized to more favorable, e�ective doses with fewer patients exposed to less e�ective doses. In addition, there is the potential for including fewer total patients in the early stages of the development programmes with attending lower risk of exposure to adverse events. A greater saving in patients may be obtained at the programme level rather than the individual study level as adaptive methodology applied across a programme will reduce the need to repeat ine�ective or inconclusive studies.

Adaptive designs represent an advanced methodology to support drug development. A prerequisite for their successful implementation is an understanding of the underlying methodology and their impact on the logistics of the trial, such as data management and monitoring procedures, electronic data capture/interactive web response services, drug supply management, and data-monitoring committee operating procedures. �ese designs have an impact on drug development strategies, trial protocols, clinical trial material doses and availability, informed consent forms, data analysis, and reporting plans. Adaptive design will also change the dynamics of

enrollment, randomization, and data capturing process, monitoring, and data cleaning systems. To be able to make informed decisions, the sponsor must have access to real time clinical data from all sources and be able to easily review and assess this data.

�is in turn requires �exibility through:

• Data management systems for rapid clinical data• access and data cleaning

• Drug supply systems for drug product planning • and management

• Optimal systems providing randomization, • medication kit management and emergency• unblinding

Planning of Adaptive Clinical Trials is a key feature in the success of such innovative approaches. Determining the optimal characteristics of the study design can be a complex yet critical decision that may require more upfront planning before the protocol can be �nalized. However, there are several commercial software packages that provide a powerful tool for planning, simulation and analysis of complex adaptive designs. FACTS™ is comprehensive software for adaptive designs in the “Learn” phase and includes early stage dose-escalation designs, adaptive dose-ranging studies, interim decision rules based on both e�cacy and safety responses, and population enrichment. ADDPLAN® is a validated software package for adaptive design in the “Con�rm” phase and incorporates sample size re-estimation, adaptive group sequential designs, multiple comparison procedures for multi-armed adaptive trials, including treatment selection designs, �exible combination of clinical research phases, and population enrichment designs.

TYPES OF ADAPTIVE DESIGNS IN “LEARN” STUDIES

Examples of adaptive designs in the Learn phase are shown in Figure 1. Phase I studies are already

FIGURE 5: Adaptive Design Creates Value at Level of the Single Trial, DevelopmentProgramme and Product Pipeline

�is aspect is discussed further in an accompanying white paper that explores the strategic utilisation of adaptive design at the development programme and product pipeline level. Importantly, recent data from Merck has indicated that adoption of an early phase adaptive trial strategy at this level saves at least $200m annually (Schindler J, Disruptive Innovations Conference, Boston, 2012).

“Adaptive Dose Ranging Studies Build Significant Efficiencies at Phase II and Optimize Identification of the Optimal Dose for Confirmatory Studies”

Several adaptive designs have been proposed and some already applied in practice using di�erent working models for the design engine: parsimonious monotonic sigmoid Emax9, Normal Dynamic Linear Model without monotonicity restrictions10, several simple models combined in a Multiple Comparison Procedure (MCP Mod).11

CASE STUDY

�e objective of this trial was to establish the correct dose(s) to take forward into a con�rmatory trial on a novel anti-psychotic drug in acute schizophrenia. �ere was data indicating that the dose-response might be non-monotonic. �erefore, the working model chosen for the design was the Normal Dynamic Linear Model, because it is �exible and robust, and also allows for capturing non-monotonic relationships. Patients were allocated to placebo, active comparator and seven doses of study drug. �e adaptive allocation was targeting both the minimum e�cacious dose (MED) and the dose achieving the maximum response (Dmax) on the primary endpoint, positive PANSS score at week 4. Weekly data updates were used to change the allocation and check early stopping rules based on predictive probabilities. A stopping rule for futility was based on a prespeci�ed threshold that no treatment dose achieves the clinically signi�cant di�erence (CSD) over placebo. A stopping rule for e�cacy required su�cient knowledge about the MED and Dmax and a prespeci�ed con�dence that the Dmax achieves CSD. For details, see Shen et al.12 �e adaptive design allowed earlier (5 months) decision making with a 99% con�dence in futility decision with $14m savings in direct grant cost. �e design, implementation, and outcome of a response adaptive dose-ranging trial of an investigational drug in patients with diabetes, incorporating a

dose expansion approach to �exibly address bothe�cacy and safety has also been reported recently by Berry et al.13 �e early (2 months) futility decision was achieved with a $3.6m saving in direct cost and additional cost savings in avoiding expensive production of phase III drug supplies.

FIGURE 4: Summary of Early Phase Case Studies

SUMMARY

Early phase adaptive design trials address a number of critical development questions that need to be answered before commitment is made to phase III con�rmatory trials. �e application of these innovative designs early in the product development process enhances clinical trial e�ciency, productivity and increases the probability of success at phase III.

�e deployment of these approaches across a pipeline of products brings substantial bene�ts to sponsor companies and enables more e�ective portfolio management and ultimately increases portfolio value (Figure 5).

adaptive by nature, but more recent adaptive designsprovide better estimate of drug safety (e.g. the maxi-mum tolerated dose) and better understanding of itsPK/PD characteristics. Combining the objectives of conventional Phase IIa and IIb studies in a single trial provides the obvious bene�t of reducing the timeline by running the two studies seamlessly under a single protocol with the same clinical team. Furthermore, such approaches achieve trial e�ciency by the prospective integration of data from both IIa and IIb in the �nal analysis.

FIRST-IN-HUMAN SINGLE ASCENDING DOSE ESCALATION DESIGNS

In contrast to traditional dose escalation designs, where for example, cohorts of subjects are allocated to ascending doses (and placebo) until a maximum tolerated dose (MTD) is empirically determined, the continual reassessment method (CRM) models the dose- and/or exposure-toxicity relationship to focus on the identi�cation of the dose-range of interest near the MTD and for this to be explored further. �ese adaptive dose escalation designs can be expanded to also e�ciently establish proof-of-mechanism or proof-of-target modulation, if validated biomarker endpoints are available.

FIGURE 1: Types of Adaptive Designs in the Learn Phase

�e traditional First in Human study objectives of safety and tolerability can be extended to include an

assessment of proof-of-mechanism in the very �rst study conducted with the investigational compound. �e ability to assess proof-of-mechanism in this earlydevelopmental setting provides an opportunity to greatly enhance the clinical development strategy for the compound. �e adaptation is intended to quickly hone in on the predicted dose-range of interest. �is avoids the risks associated with exposing patients in an ine�ective low dose range, whilst minimizing exposure to high doses with an unacceptable tolerability and safety pro�le.

“CRM Offers a Better Approach to Accurately Determining the MTD in Early Phase Oncology Studies”

CRM models are becoming routine for early phase oncology studies where accurate de�nition of the MTD is critical for selecting the optimum dose for e�cacy assessment. Conventional 3+3 designs have been shown to signi�cantly underestimate the MTD which increases the risk of poor outcome in e�cacy trials.3

SEAMLESS MULTIPLE ASCENDING DOSE AND PROOF-OF-CONCEPT STUDY DESIGNS

�e objective of seamless multiple ascending dose and proof-of-concept studies is to combine a multiple ascending dose (MAD) study in patients with the proof-of-concept of the investigational compound.

CASE STUDY

A good example is seen in the treatment of rheumatoid arthritis. �e study design involved a MAD escalation study, in which patients were randomly assigned only to the lowest treatment group or placebo in a 3:1 ratio (active treatment or placebo) which then converted into a parallel enrolling, proof of concept (POC) dose ranging study (see Figure 2). Consistent with the MAD paradigm, dose escalation (allowing subjects in the next higher dose level to

receive a second dose) occurred only after the safety review on 6 subjects assigned to active treatment in the preceding dose had been performed. �is safetyreview was undertaken by a data monitoring committee. Only after it was determined by the data monitoring committee that two consecutive doses of the experimental treatment were well tolerated at a given dose level, were subjects already enrolled in the next higher dose treatment group permitted to receive a second dose of the investigational product. In Stage 2, a utility combining biomarker observations with early readout of a clinical endpoint at week 4 was used to drop dose levels that did not ful�l prespeci�ed necessary conditions (Proof-of-Non-Viability). Viable cohorts and comparator were kept open until each of the surviving treatment arms reached a prede�ned number of patients. �e endpoint to assess proof-of-concept was a 12 week observation on the regulatory accepted endpoint for Rheumatoid Arthritis (ACR20).

�is adaptive design provides several advantages over more traditional clinical trial approaches, without compromising patient safety:

• The seamless design increased the utility of the• information obtained by allowing subjects in both• the MAD and POC stages to provide both • de�nitive safety and e�cacy data.

• Performing the MAD component in subjects with• rheumatoid arthritis receiving concomitant • methotrexate allowed for the earliest • characterization of safety and antigenicity of the• new compound in the clinically relevant • population.

• The adaptive design minimized the number of• subjects that were exposed to ine�ective doses of the• drug, while simultaneously focusing subjects to• doses that were most informative for accurate dose• selection for subsequent con�rmatory trials.

FIGURE 2: Trial Design for Seamless MAD/POC Case Study �erefore, the adaptive design optimizes the bene�t/risk balance for participating subjects via improved e�ciency of decision making in relation to the doses of the new drug studied. In this example the estimated cost saving through combining two trials into one was $1.2m and the development programme was accelerated by 9 months.

“The Development Programme was Accelerated by 9 Months and the Cost Saving of Combining Two Trials in One was Estimated at $1.2M”

PROOF OF CONCEPT AND DOSE-RANGING STUDY DESIGNS

An example of a proof of concept and dose-ranging study design is a dose-�nding study to evaluate the analgesic e�cacy and safety of a single dose of a new non-steroidal anti-in�ammatory drug in subjects experiencing pain after oral dental surgery.4 Six doses of the new drug, placebo and an active comparator are considered. �e adaptive design consists of three stages. At the �rst stage, subjects are equally randomized into three arms: placebo, active comparator, and the top dose of the new drug. At the �rst interim analysis, the study may be stopped either because of lack of assay sensitivity, i.e. inability to establish a prede�ned minimal bene�t of the comparator over and above placebo, or evidence that

the top dose of study drug could not achieve a minimal acceptable level of total pain relief at 8 hours. If the study continues to the second stage, additional subjects would be randomized to placebo, active control, and three doses of the new drug (low, medium, and high) in a 1:1:2:2:1 ratio. Again, the study could be stopped due to lack of assay sensitivity or futility. Otherwise, a D-optimal design will be used to determine both the doses of the new drug (among six possible) and subject allocation ratios for the third stage, see Figure 3. At the end of the study, a four parameter logistic model is used to �t the dose response and make recommendations about the dose to be taken in phase III.

FIGURE 3: Trial Design for POC and Adaptive Dose-Ranging Case Study

CASE STUDY

A full case study has been published in the pain �eld where a Bayesian adaptive dose ranging design was used to evaluate the e�cacy of a new analgesic drug in a proof-of-concept post-herpetic neuralgia trial.5 In total 7 doses of the drug were evaluated in an attempt to de�ne the e�ective dose response curve. �e maximum sample size was 280 patients, however, at the �rst interim analysis (80 patients) the study was stopped for futility as none of the doses showed any meaningful di�erence from placebo. �e study was formally stopped after 133 patients had been enrolled. Stopping of this trial saved an estimated direct grant cost of $0.76m with a further $1.3m savings in internal costs.

“Early Stopping of POC Trials Because of Lack of Efficacy Saves Significant Cost and Enables Valuable Resources to be Re-assigned to Alternative Programmes That May Have a Better Chance of Success”

A recent unpublished example shows that early stopping of a POC adaptive dose ranging trial in post-operative nausea and vomiting saved $0.5m in outsourced costs. In this example, 131 out of 200 patients had been enrolled when the decision was taken at the �rst interim analysis.

RESPONSE ADAPTIVE DOSE-RANGING STUDY DESIGNS

Recent simulation studies6,7 conducted by the PhRMA Working Group on Adaptive Dose-Ranging Studies have found that response-adaptive dose allocation designs are generally superior in performance to conventional pairwise comparisons approaches. Learning about the research question such as identifying a target dose, or estimating the dose-response, is optimized with these designs.

�e main objectives in an adaptive dose-ranging study are to detect the dose response, to determine if any doses(s) meets clinical relevance, to estimate the dose-response, and then to decide on the dose(s) (if any) to take into the con�rmatory Phase III. With adaptive designs for dose-ranging studies many more doses may be considered than in a conventional parallel group design without increasing sample size. �is is achieved mainly through an adaptive allocation rule that is changing the randomization ratio to di�erent treatment arms during the study and putting more subjects in the region that allow most accurate estimation of dose-response and better precision in selecting the target dose. For example, Orlo� et al.8 showed that “for half the sample size, the adaptive design is as powerful and e�cient as the standard approach”.

STA

GE

1ST

AG

E 2

Ascending MAD Until doses are open

Safety Decision: Subjects will receive a 2nddose only after a safety review of the 2nd dosein the preceding cohort.

The dose regimen is Q4 Weeks for 4 cycles.

2nd Stage begins after escalatingto the maximum tolerated dose

N=33 patients for each offive doses (10, 30, 50, 60,100 mg) and Placebo

Futility Decision: Based on ACR20and 25% reduction in CRP at 4 weeks

Internal DMC for safety & futility decisions:

Unblinded Medical Monitor Unblinded Biostatistion

Cohort 5: 100mg (n=3) : Placebo (n=1)

Cohort 4: 60mg (n=3) : Placebo (n=1)

Cohort 3: 50mg (n=3) : Placebo (n=1)

Cohort 2: 30mg (n=3) : Placebo (n=1)

Cohort 1: 10mg (n=3) : Placebo (n=1)

Stage I

DRG 900mg

DRG 750mg

DRG 600mg

DRG 450mg

DRG 300mg

DRG 150mg

Plbo

Ctrl

Total Ssize ~ 30:10 pats/arm

Stage II Stage III• Fit the Model• Find the D-Optimal Design• Determine Sample Size

DRG 900mg

DRG 750mg

DRG 600mg

DRG 450mg

DRG 300mg

DRG 150mg

Plbo

Ctrl

Total Ssize ~ 65:5:10 pats/arm

STOPFutility

STOPNAS

IA 1

STOPFutility

STOPNAS

IA II

Page 8: DESIGN AND EXECUTION OF EARLY PHASE …...Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs

DESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALSDESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALS

PAGE 7 PAGE 4

EXECUTIVE SUMMARY

�e adoption of an adaptive design strategy across the product development process brings a number of important bene�ts. �ese include enhanced R&D e�ciency, enhanced R&D productivity and, importantly, increased probability of success at phase III. �e opportunity to manage risk starts in the early phase of a development programme, with a diligent view to appropriate dose selection which impacts not only the success of the downstream clinical trial programme, but also manufacturing costs and value analyses.

Adaptive design trials enhance R&D e�ciency by reducing the need to repeat trials that just miss their clinical end-point or fail to identify the e�ective dose response at the �rst attempt. By avoiding the need to run these trials again, signi�cant cost and time savings are achieved. �is is possible through use of adaptive designs that enable additional patients to be added to achieve statistical signi�cance the �rst time around or by allowing a wider dose range to be studied and then picking doses early in the trial that are in the optimum part of the dose response curve. In addition, early stopping of development programmes because a product is ine�ective enables scarce resources to be redeployed in additional trials which may show more promise. All of these factors increase development e�ciency.

Adaptive designs increase R&D productivity by enabling more accurate de�nition of the e�ective dose response at phase II which enables better design of the pivotal phase III programme which in turn increases the probability of success of the overall development programme. A number of phase III trials fail because the dose is either too high and causes unwanted safety issues or is too low to show su�cient e�cacy. Adaptive design trials enable optimized dose selection before the pivotal trial phase is started.

Another opportunity provided by adaptive design is population enrichment where drug response can be

optimised to speci�c patient sub-populations that respond better to treatment. Many phase III studies fail because the overall e�cacy of treatment is diluted as a consequence of the drug being evaluated in the full trial population rather than in the speci�c subset where the drug works best. Adaptive design enables early selection of the appropriate patient population and increases the probability of success.

“Adaptive Design Enables Selection of the Right Dose for the Right Patient Before Commitment to Expensive Phase III Trials”

Phases I and II are critical steps in the clinical development process as this is where important information about the product has to be generated and assessed before the decision is taken to commit to expensive phase III pivotal studies. �is early phase of development is known as the “Learn Phase” and the data generated relates to the e�ective dose-response, the safety pro�le and therapeutic index, appropriate endpoints, and the patient population best treated by the product under evaluation.

Adaptive designs have an important role to play in both phase I and phase II, but it is in the latter phase where they add the greatest value in terms of risk management. Validated methodologies are available to design early phase adaptive trials and this is an area of signi�cant interest across the pharmaceutical industry.

“Getting it Right at Phase II is a Major Objective for Adaptive Design and is the Point at which Phase III Success is Defined”

Companies that adopt a comprehensive adaptive design strategy in the “Learn” phase will make better development decisions, will build development e�ciency, will increase the probability of success of the molecules that enter phase III, ultimately bringing e�ective products to the market more e�ciently. Competitor companies that persevere with conventional trial designs will remain slow, ine�cient

candidate optimisation to Proof of concept stages. Moira assists partner companies with both acquisition and due diligence of in licensed compounds and strategic support for the out-licensing of successful candidates. �is has included design, planning and management of, strategic and clinical development plans, due diligence, decision analysis, preclinical, POC and Phase II programmes in a range of, indications including respiratory, anti-infective, immunomodulation, CNS and cardiovascular. �is experience covers a large range of targets and mecha-nisms of action. In order to support the strategic and clinical development planning Moira has worked extensively with international Key Opinion Leaders in respiratory, anti-infective and transplant indications.

PHIL BIRCH,SVP, CORPORATE BRAND MANAGER

Dr. Phil Birch is a Senior Vice President at Aptiv Solutions and has global responsibility for co-ordinating market education and client awareness in the adaptive clinical trials �eld. Dr. Birch has worked in the pharmaceutical industry for over 27 years and has held executive positions in corporate development, business development and R&D in a number of consulting, biotechnology and top 10 pharmaceutical companies.

REFERENCES

1. Dragalin V. Adaptive Designs: Terminology and Classi�cation. Drug Information Journal, 2006; 40: 425-436. 2. Gallo P, Chuang-Stein C, Dragalin V, Gaydos B, Krams M, Pinheiro J. Executive Summary of the PhRMA Working Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs for Phase I Oncology Trials, Clinical Operations Summit, Brussels, 2012).4. Dragalin V, Hsuan F, and Padmanabhan SK. Adaptive Designs for Dose-Finding Studies Based on Sigmoid Emax Model. J Biopharmaceutical Statistics 2007; 17:1051-1070.5. Smith MK, Jones I, Morris MF, Grieve A and Tan K. Implementation of a Bayesian adaptive design in a proof of concept study. Pharmaceutical Statistics 2006; 5:39-50. 6. Bornkamp B, Bretz F, Dmitrienko A, Enas G, Gaydos B, Hsu CH, König F, Krams M, Liu Q, Neuenschwander B, Parke T, Pinheiro J, Roy A, Sax R, Shen F. Innovative approaches for designing and analyzing adaptive dose-ranging trials: White Paper from the PhRMA working group on “Adaptive Dose-Ranging Studies”. J of Biopharmaceutical Statistics 2007; 17: 965–995. 7. Dragalin V, Bornkamp B, Bretz F, Miller F, Padmanabhan SK, Patel N, Perevozskaya I, Pinheiro J, and Smith J. A Simulation Study to Compare New Adaptive Dose–Ranging Designs. Statistics in Biopharmaceutical Research 2010; 2:487-512.8. Orlo� J, Douglas F, Pinheiro J, Levinson S, Branson M, Chaturvedi P, Ette E, Gallo P, Hirsch G, Mehta C, Patel N, Sabir S, Springs S, Stanski D, Golub H, Evers M, Fleming E, Singh N, Tramontin T. �e future of drug development: advancing clinical trial design. Nature Review-Drug Discovery. 2009; 8:940-957.9. Padmanabhan SK and Dragalin V. Adaptive Dc-optimal designs for dose �nding based on a continuous e�cacy endpoint. Biometrical Journal 2010; 52:836-852. 10. Grieve AP and Krams M. ASTIN: a Bayesian adaptive dose-response trial in acute stroke. Clinical Trials 2005; 2:340-351. 11. Bornkamp B, Pinheiro J, Bretz F. MCPMod: An R Package for the Design and Analysis of Dose-Finding Studies. Journal of Statistical Software 2009; 29:1-23.12. Shen J, Preskorn S, Dragalin V, Slomkowski M, Padmanabhan SK, Fardipour P, Sharma A, and Krams M. How Adaptive Trial Designs Can Increase E�ciency in Psychiatric Drug Development: A Case Study. Innovations in Clinical Neuroscience 2011; 8: 26-34.13. Berry SM, Spinelli W, Littman GS, Liang JZ, Fardipour P, Berry DA, Lewis RL, and Krams M. A Bayesian dose-�nding trial with adaptive dose expansion to �exibly assess e�cacy and safety of an investigational drug. Clinical Trials 2010; 7: 121-135.

ABOUT THE AUTHORS

VLAD DRAGALIN, SVP, INNOVATION CENTRE

Dr. Vlad Dragalin is a well-known adaptive design expert with 25 years of experience in developing the statistical methodology of adaptive designs and with over 12 years experience in pharmaceutical industry. Vlad is a Senior Vice President, Software Develop-ment and Consulting, in the Innovation Center at Aptiv Solutions. Vlad works with clients to improve the drug development process through the use of adaptive designs and to assist them with the trial design, simulation and execution, developing adaptive trials software tools, and working with project teams in clinical operations, data management and statistics during study conduct.

SARAH ARBE-BARNES,SVP, TRANSLATIONAL SCIENCES

Dr. Sarah Arbe-Barnes has over 24 years of experience in drug development and project leadership in the Pharma and service sector industry to include in/out-licensing and due diligence, strategic development and business planning, fundraising for start-ups/SMEs( not-for-pro�t sector and VC), along with optimized market access strategies and marketing. Sarah has led recent product approval programmes, overseeing a broad range of clinical and nonclinical project design and management activities from the point of candidate selection. Sarah heads the Transla-tional Sciences division of Aptiv Solutions, which is a dedicated global consultancy group, focused on the provision of drug and device development advice and expertise for small molecules and biologics.

MOIRA THOMSON,VP, TRANSLATIONAL SCIENCES

Moira �omson has over 15 years of experience in the Pharmaceutical Industry, and is involved in the design, planning and management of global develop-ment projects for a mixture of small and medium sized Pharma, and Biotech companies covering lead

and will fail to make the commercial returns that investors and shareholders demand. “Adaptive Design Builds Competitive Advantage”

�is White Paper summarizes the important aspects of early phase adaptive design and provides a summary of selected case studies which demonstrate the value of this innovative approach. It has been written to inform senior R&D decision-makers about the critical role of adaptive design in early phase development and the signi�cant bene�ts that this approach can bring.

BACKGROUND TO EARLY PHASE ADAPTIVE TRIAL DESIGN

Adaptive Clinical Trials are viewed as supporting a change in the development process as focus moves from blockbuster products to more specialized medicines that require a �exible cost-e�cient approach to investigating their e�ectiveness. Adaptive designs are statistical methodologies applied to speci�c stages of drug development where real time learning from accumulating trial data is applied to optimize subsequent study execution.1 �e bene�t of an Adaptive Clinical Trial lies in the monitoring of data to make design modi�cation adjustments to an aspect of a trial leading to quicker development decisions around key go/no-go milestones that impact time to market and/or minimise costly R&D losses.

�e adaptations are prospectively de�ned prior to the start of the trial and can include stopping early either for futility or success, expanding the sample size due to greater than expected data variability, or allocating patients preferentially to treatment regimens with a better therapeutic index. Virtually any aspect of the trial may be the potential target of design modi�cations. Importantly, these modi�cations are a design feature aimed to enhance the trial, not a remedy for inadequate planning. �ey are not ad hoc study corrections made via protocol amendment.2

�e speci�c adaptations considered, and the basis of their implementation, are carefully de�ned based on strict rules, justi�ed statistically and scienti�cally, and are an integral part of the �nal, pre-recruitment trial protocol. Failing this would compromise the interpretation and acceptance of study results.

Adaptive designs often employ frequent interim analyses of all accumulated data (and, possibly, external trial data) to determine whether pre-planned design modi�cations will be ‘triggered’. Interim analyses partition the trial into multiple stages, each trial stage’s characteristics (number of arms, number of patients to be enrolled, their allocation between arms, stage duration, etc.) de�ned by the preceding interim analysis results. �e ability to periodically, or even continually, examine available data to determine whether trial modi�cations are necessary and implement pre-de�ned design changes when indicated, gives adaptive design its strength and �exibility.

�e use of adaptive design in the exploratory stage of clinical trials can increase the e�ciency of drug development by improving our ability to e�ciently learn about the dose-response and better determine whether to take a drug (and the right dose of the drug and in the right population) forward into later phase testing. �ese designs explicitly address multiple trial goals, adaptively allocate subjects according to ongoing information needs, and allow termination for both early success and futility considerations. �is approach can maximize the ability to test a larger number of doses in a single trial while simultaneously increasing the e�ciency of the trial in terms of making better go/no-go decisions about continuing the trial and/or the development of the drug for a speci�c indication or sub-population. During the “Learn” phase many aspects of the study design can be modi�ed: number of subjects, study duration, endpoint selection, treatment duration, number of treatments, patient population. �e adaptation of multiple trial features that are a legitimate concern in con�rmatory studies can become, in fact, an

appealing feature of adaptive designs in exploratory development, leading to enhanced learning to set the appropriate stage for con�rmatory trials.

“The Use of Adaptive Design in the Learn Phase Offers the Best Opportunity to Assess the Product Characteristics that Determine Success”

Judicious use of adaptive designs may increase the information value per resource unit invested by avoiding allocation of patients to non-e�cacious/unsafe therapies and allowing stopping decisions to be made at the earliest possible time point. Ultimately this will accelerate the development of promising therapies to bene�t patients.

Adaptive Clinical Trials can result in the more ethical treatment of patients from at least two perspectives. First, a larger number of patients can be randomized to more favorable, e�ective doses with fewer patients exposed to less e�ective doses. In addition, there is the potential for including fewer total patients in the early stages of the development programmes with attending lower risk of exposure to adverse events. A greater saving in patients may be obtained at the programme level rather than the individual study level as adaptive methodology applied across a programme will reduce the need to repeat ine�ective or inconclusive studies.

Adaptive designs represent an advanced methodology to support drug development. A prerequisite for their successful implementation is an understanding of the underlying methodology and their impact on the logistics of the trial, such as data management and monitoring procedures, electronic data capture/interactive web response services, drug supply management, and data-monitoring committee operating procedures. �ese designs have an impact on drug development strategies, trial protocols, clinical trial material doses and availability, informed consent forms, data analysis, and reporting plans. Adaptive design will also change the dynamics of

enrollment, randomization, and data capturing process, monitoring, and data cleaning systems. To be able to make informed decisions, the sponsor must have access to real time clinical data from all sources and be able to easily review and assess this data.

�is in turn requires �exibility through:

• Data management systems for rapid clinical data• access and data cleaning

• Drug supply systems for drug product planning • and management

• Optimal systems providing randomization, • medication kit management and emergency• unblinding

Planning of Adaptive Clinical Trials is a key feature in the success of such innovative approaches. Determining the optimal characteristics of the study design can be a complex yet critical decision that may require more upfront planning before the protocol can be �nalized. However, there are several commercial software packages that provide a powerful tool for planning, simulation and analysis of complex adaptive designs. FACTS™ is comprehensive software for adaptive designs in the “Learn” phase and includes early stage dose-escalation designs, adaptive dose-ranging studies, interim decision rules based on both e�cacy and safety responses, and population enrichment. ADDPLAN® is a validated software package for adaptive design in the “Con�rm” phase and incorporates sample size re-estimation, adaptive group sequential designs, multiple comparison procedures for multi-armed adaptive trials, including treatment selection designs, �exible combination of clinical research phases, and population enrichment designs.

TYPES OF ADAPTIVE DESIGNS IN “LEARN” STUDIES

Examples of adaptive designs in the Learn phase are shown in Figure 1. Phase I studies are already

FIGURE 5: Adaptive Design Creates Value at Level of the Single Trial, DevelopmentProgramme and Product Pipeline

�is aspect is discussed further in an accompanying white paper that explores the strategic utilisation of adaptive design at the development programme and product pipeline level. Importantly, recent data from Merck has indicated that adoption of an early phase adaptive trial strategy at this level saves at least $200m annually (Schindler J, Disruptive Innovations Conference, Boston, 2012).

“Adaptive Dose Ranging Studies Build Significant Efficiencies at Phase II and Optimize Identification of the Optimal Dose for Confirmatory Studies”

Several adaptive designs have been proposed and some already applied in practice using di�erent working models for the design engine: parsimonious monotonic sigmoid Emax9, Normal Dynamic Linear Model without monotonicity restrictions10, several simple models combined in a Multiple Comparison Procedure (MCP Mod).11

CASE STUDY

�e objective of this trial was to establish the correct dose(s) to take forward into a con�rmatory trial on a novel anti-psychotic drug in acute schizophrenia. �ere was data indicating that the dose-response might be non-monotonic. �erefore, the working model chosen for the design was the Normal Dynamic Linear Model, because it is �exible and robust, and also allows for capturing non-monotonic relationships. Patients were allocated to placebo, active comparator and seven doses of study drug. �e adaptive allocation was targeting both the minimum e�cacious dose (MED) and the dose achieving the maximum response (Dmax) on the primary endpoint, positive PANSS score at week 4. Weekly data updates were used to change the allocation and check early stopping rules based on predictive probabilities. A stopping rule for futility was based on a prespeci�ed threshold that no treatment dose achieves the clinically signi�cant di�erence (CSD) over placebo. A stopping rule for e�cacy required su�cient knowledge about the MED and Dmax and a prespeci�ed con�dence that the Dmax achieves CSD. For details, see Shen et al.12 �e adaptive design allowed earlier (5 months) decision making with a 99% con�dence in futility decision with $14m savings in direct grant cost. �e design, implementation, and outcome of a response adaptive dose-ranging trial of an investigational drug in patients with diabetes, incorporating a

dose expansion approach to �exibly address bothe�cacy and safety has also been reported recently by Berry et al.13 �e early (2 months) futility decision was achieved with a $3.6m saving in direct cost and additional cost savings in avoiding expensive production of phase III drug supplies.

FIGURE 4: Summary of Early Phase Case Studies

SUMMARY

Early phase adaptive design trials address a number of critical development questions that need to be answered before commitment is made to phase III con�rmatory trials. �e application of these innovative designs early in the product development process enhances clinical trial e�ciency, productivity and increases the probability of success at phase III.

�e deployment of these approaches across a pipeline of products brings substantial bene�ts to sponsor companies and enables more e�ective portfolio management and ultimately increases portfolio value (Figure 5).

adaptive by nature, but more recent adaptive designsprovide better estimate of drug safety (e.g. the maxi-mum tolerated dose) and better understanding of itsPK/PD characteristics. Combining the objectives of conventional Phase IIa and IIb studies in a single trial provides the obvious bene�t of reducing the timeline by running the two studies seamlessly under a single protocol with the same clinical team. Furthermore, such approaches achieve trial e�ciency by the prospective integration of data from both IIa and IIb in the �nal analysis.

FIRST-IN-HUMAN SINGLE ASCENDING DOSE ESCALATION DESIGNS

In contrast to traditional dose escalation designs, where for example, cohorts of subjects are allocated to ascending doses (and placebo) until a maximum tolerated dose (MTD) is empirically determined, the continual reassessment method (CRM) models the dose- and/or exposure-toxicity relationship to focus on the identi�cation of the dose-range of interest near the MTD and for this to be explored further. �ese adaptive dose escalation designs can be expanded to also e�ciently establish proof-of-mechanism or proof-of-target modulation, if validated biomarker endpoints are available.

FIGURE 1: Types of Adaptive Designs in the Learn Phase

�e traditional First in Human study objectives of safety and tolerability can be extended to include an

assessment of proof-of-mechanism in the very �rst study conducted with the investigational compound. �e ability to assess proof-of-mechanism in this earlydevelopmental setting provides an opportunity to greatly enhance the clinical development strategy for the compound. �e adaptation is intended to quickly hone in on the predicted dose-range of interest. �is avoids the risks associated with exposing patients in an ine�ective low dose range, whilst minimizing exposure to high doses with an unacceptable tolerability and safety pro�le.

“CRM Offers a Better Approach to Accurately Determining the MTD in Early Phase Oncology Studies”

CRM models are becoming routine for early phase oncology studies where accurate de�nition of the MTD is critical for selecting the optimum dose for e�cacy assessment. Conventional 3+3 designs have been shown to signi�cantly underestimate the MTD which increases the risk of poor outcome in e�cacy trials.3

SEAMLESS MULTIPLE ASCENDING DOSE AND PROOF-OF-CONCEPT STUDY DESIGNS

�e objective of seamless multiple ascending dose and proof-of-concept studies is to combine a multiple ascending dose (MAD) study in patients with the proof-of-concept of the investigational compound.

CASE STUDY

A good example is seen in the treatment of rheumatoid arthritis. �e study design involved a MAD escalation study, in which patients were randomly assigned only to the lowest treatment group or placebo in a 3:1 ratio (active treatment or placebo) which then converted into a parallel enrolling, proof of concept (POC) dose ranging study (see Figure 2). Consistent with the MAD paradigm, dose escalation (allowing subjects in the next higher dose level to

receive a second dose) occurred only after the safety review on 6 subjects assigned to active treatment in the preceding dose had been performed. �is safetyreview was undertaken by a data monitoring committee. Only after it was determined by the data monitoring committee that two consecutive doses of the experimental treatment were well tolerated at a given dose level, were subjects already enrolled in the next higher dose treatment group permitted to receive a second dose of the investigational product. In Stage 2, a utility combining biomarker observations with early readout of a clinical endpoint at week 4 was used to drop dose levels that did not ful�l prespeci�ed necessary conditions (Proof-of-Non-Viability). Viable cohorts and comparator were kept open until each of the surviving treatment arms reached a prede�ned number of patients. �e endpoint to assess proof-of-concept was a 12 week observation on the regulatory accepted endpoint for Rheumatoid Arthritis (ACR20).

�is adaptive design provides several advantages over more traditional clinical trial approaches, without compromising patient safety:

• The seamless design increased the utility of the• information obtained by allowing subjects in both• the MAD and POC stages to provide both • de�nitive safety and e�cacy data.

• Performing the MAD component in subjects with• rheumatoid arthritis receiving concomitant • methotrexate allowed for the earliest • characterization of safety and antigenicity of the• new compound in the clinically relevant • population.

• The adaptive design minimized the number of• subjects that were exposed to ine�ective doses of the• drug, while simultaneously focusing subjects to• doses that were most informative for accurate dose• selection for subsequent con�rmatory trials.

FIGURE 2: Trial Design for Seamless MAD/POC Case Study �erefore, the adaptive design optimizes the bene�t/risk balance for participating subjects via improved e�ciency of decision making in relation to the doses of the new drug studied. In this example the estimated cost saving through combining two trials into one was $1.2m and the development programme was accelerated by 9 months.

“The Development Programme was Accelerated by 9 Months and the Cost Saving of Combining Two Trials in One was Estimated at $1.2M”

PROOF OF CONCEPT AND DOSE-RANGING STUDY DESIGNS

An example of a proof of concept and dose-ranging study design is a dose-�nding study to evaluate the analgesic e�cacy and safety of a single dose of a new non-steroidal anti-in�ammatory drug in subjects experiencing pain after oral dental surgery.4 Six doses of the new drug, placebo and an active comparator are considered. �e adaptive design consists of three stages. At the �rst stage, subjects are equally randomized into three arms: placebo, active comparator, and the top dose of the new drug. At the �rst interim analysis, the study may be stopped either because of lack of assay sensitivity, i.e. inability to establish a prede�ned minimal bene�t of the comparator over and above placebo, or evidence that

the top dose of study drug could not achieve a minimal acceptable level of total pain relief at 8 hours. If the study continues to the second stage, additional subjects would be randomized to placebo, active control, and three doses of the new drug (low, medium, and high) in a 1:1:2:2:1 ratio. Again, the study could be stopped due to lack of assay sensitivity or futility. Otherwise, a D-optimal design will be used to determine both the doses of the new drug (among six possible) and subject allocation ratios for the third stage, see Figure 3. At the end of the study, a four parameter logistic model is used to �t the dose response and make recommendations about the dose to be taken in phase III.

FIGURE 3: Trial Design for POC and Adaptive Dose-Ranging Case Study

CASE STUDY

A full case study has been published in the pain �eld where a Bayesian adaptive dose ranging design was used to evaluate the e�cacy of a new analgesic drug in a proof-of-concept post-herpetic neuralgia trial.5 In total 7 doses of the drug were evaluated in an attempt to de�ne the e�ective dose response curve. �e maximum sample size was 280 patients, however, at the �rst interim analysis (80 patients) the study was stopped for futility as none of the doses showed any meaningful di�erence from placebo. �e study was formally stopped after 133 patients had been enrolled. Stopping of this trial saved an estimated direct grant cost of $0.76m with a further $1.3m savings in internal costs.

“Early Stopping of POC Trials Because of Lack of Efficacy Saves Significant Cost and Enables Valuable Resources to be Re-assigned to Alternative Programmes That May Have a Better Chance of Success”

A recent unpublished example shows that early stopping of a POC adaptive dose ranging trial in post-operative nausea and vomiting saved $0.5m in outsourced costs. In this example, 131 out of 200 patients had been enrolled when the decision was taken at the �rst interim analysis.

RESPONSE ADAPTIVE DOSE-RANGING STUDY DESIGNS

Recent simulation studies6,7 conducted by the PhRMA Working Group on Adaptive Dose-Ranging Studies have found that response-adaptive dose allocation designs are generally superior in performance to conventional pairwise comparisons approaches. Learning about the research question such as identifying a target dose, or estimating the dose-response, is optimized with these designs.

�e main objectives in an adaptive dose-ranging study are to detect the dose response, to determine if any doses(s) meets clinical relevance, to estimate the dose-response, and then to decide on the dose(s) (if any) to take into the con�rmatory Phase III. With adaptive designs for dose-ranging studies many more doses may be considered than in a conventional parallel group design without increasing sample size. �is is achieved mainly through an adaptive allocation rule that is changing the randomization ratio to di�erent treatment arms during the study and putting more subjects in the region that allow most accurate estimation of dose-response and better precision in selecting the target dose. For example, Orlo� et al.8 showed that “for half the sample size, the adaptive design is as powerful and e�cient as the standard approach”.

• Single ascending dose escalation designs• Up-and-Down and CRM to find MTD• Establish Proof-of-Mechanism or Proof-of-Target Modulation

First-in Human

• Two-stage adaptive approach in patients• 1st stage – to identify MTD • 2nd stage – to select dose and exposure levels (necessary cond.)

MAD and PoC

• Start with the highest feasible tolerated dose and placebo• If a pre-specified futility condition is satisfied > stop • Otherwise, open enrollment to lower doses

PoC and ADRS

• SAD or MAD combined with Biomaker-based Efficacy• To identify the Optimal Safe Dose

Seamless PhaseI/II Design

• Finding a target dose (MED, EDp)• Response Adaptive Allocation• Covariate Adjusted Response Adaptive Allocation

Adaptive DoseRanging Design

CRM: Continual Reassessment Method; MTD: Maximum Tolerated Dose; MAD: Multiple Ascending Dose;SAD: Single Ascending Dose; MED: Minimum Effective Dose; EDp: Dose achieving 100p% of maximum effect

TherapeuticArea

Adaptive Design Trial Outcome Metrics

RheumatoidArthritis

Post-OperativeNausea &Vomiting

Post-HerpeticNeuralgia

Schizophrenia

Adaptive combinedMultiple Ascending Dose and Proof of Concept Study

Proof of concept andadaptive dose-ranging study design

Proof of concept andadaptive dose-ranging study design

Adaptive dose-ranging

Successful trial –criteria met for theachievement ofclinical POC inRheumatoid Arthritis

Early stopping forfutility

Early stopping forfutility

Early stopping forfutility

Programmeaccelerated by 9-month; $1.2m saving by combining two trials in one

$0.5m savings andredeployment of resource

$2m savings and re-deployment of resource

$14m savings and re-deployment of resource

Page 9: DESIGN AND EXECUTION OF EARLY PHASE …...Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs

DESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALSDESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALS

PAGE 3 PAGE 8

EXECUTIVE SUMMARY

�e adoption of an adaptive design strategy across the product development process brings a number of important bene�ts. �ese include enhanced R&D e�ciency, enhanced R&D productivity and, importantly, increased probability of success at phase III. �e opportunity to manage risk starts in the early phase of a development programme, with a diligent view to appropriate dose selection which impacts not only the success of the downstream clinical trial programme, but also manufacturing costs and value analyses.

Adaptive design trials enhance R&D e�ciency by reducing the need to repeat trials that just miss their clinical end-point or fail to identify the e�ective dose response at the �rst attempt. By avoiding the need to run these trials again, signi�cant cost and time savings are achieved. �is is possible through use of adaptive designs that enable additional patients to be added to achieve statistical signi�cance the �rst time around or by allowing a wider dose range to be studied and then picking doses early in the trial that are in the optimum part of the dose response curve. In addition, early stopping of development programmes because a product is ine�ective enables scarce resources to be redeployed in additional trials which may show more promise. All of these factors increase development e�ciency.

Adaptive designs increase R&D productivity by enabling more accurate de�nition of the e�ective dose response at phase II which enables better design of the pivotal phase III programme which in turn increases the probability of success of the overall development programme. A number of phase III trials fail because the dose is either too high and causes unwanted safety issues or is too low to show su�cient e�cacy. Adaptive design trials enable optimized dose selection before the pivotal trial phase is started.

Another opportunity provided by adaptive design is population enrichment where drug response can be

optimised to speci�c patient sub-populations that respond better to treatment. Many phase III studies fail because the overall e�cacy of treatment is diluted as a consequence of the drug being evaluated in the full trial population rather than in the speci�c subset where the drug works best. Adaptive design enables early selection of the appropriate patient population and increases the probability of success.

“Adaptive Design Enables Selection of the Right Dose for the Right Patient Before Commitment to Expensive Phase III Trials”

Phases I and II are critical steps in the clinical development process as this is where important information about the product has to be generated and assessed before the decision is taken to commit to expensive phase III pivotal studies. �is early phase of development is known as the “Learn Phase” and the data generated relates to the e�ective dose-response, the safety pro�le and therapeutic index, appropriate endpoints, and the patient population best treated by the product under evaluation.

Adaptive designs have an important role to play in both phase I and phase II, but it is in the latter phase where they add the greatest value in terms of risk management. Validated methodologies are available to design early phase adaptive trials and this is an area of signi�cant interest across the pharmaceutical industry.

“Getting it Right at Phase II is a Major Objective for Adaptive Design and is the Point at which Phase III Success is Defined”

Companies that adopt a comprehensive adaptive design strategy in the “Learn” phase will make better development decisions, will build development e�ciency, will increase the probability of success of the molecules that enter phase III, ultimately bringing e�ective products to the market more e�ciently. Competitor companies that persevere with conventional trial designs will remain slow, ine�cient

candidate optimisation to Proof of concept stages. Moira assists partner companies with both acquisition and due diligence of in licensed compounds and strategic support for the out-licensing of successful candidates. �is has included design, planning and management of, strategic and clinical development plans, due diligence, decision analysis, preclinical, POC and Phase II programmes in a range of, indications including respiratory, anti-infective, immunomodulation, CNS and cardiovascular. �is experience covers a large range of targets and mecha-nisms of action. In order to support the strategic and clinical development planning Moira has worked extensively with international Key Opinion Leaders in respiratory, anti-infective and transplant indications.

PHIL BIRCH,SVP, CORPORATE BRAND MANAGER

Dr. Phil Birch is a Senior Vice President at Aptiv Solutions and has global responsibility for co-ordinating market education and client awareness in the adaptive clinical trials �eld. Dr. Birch has worked in the pharmaceutical industry for over 27 years and has held executive positions in corporate development, business development and R&D in a number of consulting, biotechnology and top 10 pharmaceutical companies.

REFERENCES

1. Dragalin V. Adaptive Designs: Terminology and Classi�cation. Drug Information Journal, 2006; 40: 425-436. 2. Gallo P, Chuang-Stein C, Dragalin V, Gaydos B, Krams M, Pinheiro J. Executive Summary of the PhRMA Working Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs for Phase I Oncology Trials, Clinical Operations Summit, Brussels, 2012).4. Dragalin V, Hsuan F, and Padmanabhan SK. Adaptive Designs for Dose-Finding Studies Based on Sigmoid Emax Model. J Biopharmaceutical Statistics 2007; 17:1051-1070.5. Smith MK, Jones I, Morris MF, Grieve A and Tan K. Implementation of a Bayesian adaptive design in a proof of concept study. Pharmaceutical Statistics 2006; 5:39-50. 6. Bornkamp B, Bretz F, Dmitrienko A, Enas G, Gaydos B, Hsu CH, König F, Krams M, Liu Q, Neuenschwander B, Parke T, Pinheiro J, Roy A, Sax R, Shen F. Innovative approaches for designing and analyzing adaptive dose-ranging trials: White Paper from the PhRMA working group on “Adaptive Dose-Ranging Studies”. J of Biopharmaceutical Statistics 2007; 17: 965–995. 7. Dragalin V, Bornkamp B, Bretz F, Miller F, Padmanabhan SK, Patel N, Perevozskaya I, Pinheiro J, and Smith J. A Simulation Study to Compare New Adaptive Dose–Ranging Designs. Statistics in Biopharmaceutical Research 2010; 2:487-512.8. Orlo� J, Douglas F, Pinheiro J, Levinson S, Branson M, Chaturvedi P, Ette E, Gallo P, Hirsch G, Mehta C, Patel N, Sabir S, Springs S, Stanski D, Golub H, Evers M, Fleming E, Singh N, Tramontin T. �e future of drug development: advancing clinical trial design. Nature Review-Drug Discovery. 2009; 8:940-957.9. Padmanabhan SK and Dragalin V. Adaptive Dc-optimal designs for dose �nding based on a continuous e�cacy endpoint. Biometrical Journal 2010; 52:836-852. 10. Grieve AP and Krams M. ASTIN: a Bayesian adaptive dose-response trial in acute stroke. Clinical Trials 2005; 2:340-351. 11. Bornkamp B, Pinheiro J, Bretz F. MCPMod: An R Package for the Design and Analysis of Dose-Finding Studies. Journal of Statistical Software 2009; 29:1-23.12. Shen J, Preskorn S, Dragalin V, Slomkowski M, Padmanabhan SK, Fardipour P, Sharma A, and Krams M. How Adaptive Trial Designs Can Increase E�ciency in Psychiatric Drug Development: A Case Study. Innovations in Clinical Neuroscience 2011; 8: 26-34.13. Berry SM, Spinelli W, Littman GS, Liang JZ, Fardipour P, Berry DA, Lewis RL, and Krams M. A Bayesian dose-�nding trial with adaptive dose expansion to �exibly assess e�cacy and safety of an investigational drug. Clinical Trials 2010; 7: 121-135.

ABOUT THE AUTHORS

VLAD DRAGALIN, SVP, INNOVATION CENTRE

Dr. Vlad Dragalin is a well-known adaptive design expert with 25 years of experience in developing the statistical methodology of adaptive designs and with over 12 years experience in pharmaceutical industry. Vlad is a Senior Vice President, Software Develop-ment and Consulting, in the Innovation Center at Aptiv Solutions. Vlad works with clients to improve the drug development process through the use of adaptive designs and to assist them with the trial design, simulation and execution, developing adaptive trials software tools, and working with project teams in clinical operations, data management and statistics during study conduct.

SARAH ARBE-BARNES,SVP, TRANSLATIONAL SCIENCES

Dr. Sarah Arbe-Barnes has over 24 years of experience in drug development and project leadership in the Pharma and service sector industry to include in/out-licensing and due diligence, strategic development and business planning, fundraising for start-ups/SMEs( not-for-pro�t sector and VC), along with optimized market access strategies and marketing. Sarah has led recent product approval programmes, overseeing a broad range of clinical and nonclinical project design and management activities from the point of candidate selection. Sarah heads the Transla-tional Sciences division of Aptiv Solutions, which is a dedicated global consultancy group, focused on the provision of drug and device development advice and expertise for small molecules and biologics.

MOIRA THOMSON,VP, TRANSLATIONAL SCIENCES

Moira �omson has over 15 years of experience in the Pharmaceutical Industry, and is involved in the design, planning and management of global develop-ment projects for a mixture of small and medium sized Pharma, and Biotech companies covering lead

and will fail to make the commercial returns that investors and shareholders demand. “Adaptive Design Builds Competitive Advantage”

�is White Paper summarizes the important aspects of early phase adaptive design and provides a summary of selected case studies which demonstrate the value of this innovative approach. It has been written to inform senior R&D decision-makers about the critical role of adaptive design in early phase development and the signi�cant bene�ts that this approach can bring.

BACKGROUND TO EARLY PHASE ADAPTIVE TRIAL DESIGN

Adaptive Clinical Trials are viewed as supporting a change in the development process as focus moves from blockbuster products to more specialized medicines that require a �exible cost-e�cient approach to investigating their e�ectiveness. Adaptive designs are statistical methodologies applied to speci�c stages of drug development where real time learning from accumulating trial data is applied to optimize subsequent study execution.1 �e bene�t of an Adaptive Clinical Trial lies in the monitoring of data to make design modi�cation adjustments to an aspect of a trial leading to quicker development decisions around key go/no-go milestones that impact time to market and/or minimise costly R&D losses.

�e adaptations are prospectively de�ned prior to the start of the trial and can include stopping early either for futility or success, expanding the sample size due to greater than expected data variability, or allocating patients preferentially to treatment regimens with a better therapeutic index. Virtually any aspect of the trial may be the potential target of design modi�cations. Importantly, these modi�cations are a design feature aimed to enhance the trial, not a remedy for inadequate planning. �ey are not ad hoc study corrections made via protocol amendment.2

�e speci�c adaptations considered, and the basis of their implementation, are carefully de�ned based on strict rules, justi�ed statistically and scienti�cally, and are an integral part of the �nal, pre-recruitment trial protocol. Failing this would compromise the interpretation and acceptance of study results.

Adaptive designs often employ frequent interim analyses of all accumulated data (and, possibly, external trial data) to determine whether pre-planned design modi�cations will be ‘triggered’. Interim analyses partition the trial into multiple stages, each trial stage’s characteristics (number of arms, number of patients to be enrolled, their allocation between arms, stage duration, etc.) de�ned by the preceding interim analysis results. �e ability to periodically, or even continually, examine available data to determine whether trial modi�cations are necessary and implement pre-de�ned design changes when indicated, gives adaptive design its strength and �exibility.

�e use of adaptive design in the exploratory stage of clinical trials can increase the e�ciency of drug development by improving our ability to e�ciently learn about the dose-response and better determine whether to take a drug (and the right dose of the drug and in the right population) forward into later phase testing. �ese designs explicitly address multiple trial goals, adaptively allocate subjects according to ongoing information needs, and allow termination for both early success and futility considerations. �is approach can maximize the ability to test a larger number of doses in a single trial while simultaneously increasing the e�ciency of the trial in terms of making better go/no-go decisions about continuing the trial and/or the development of the drug for a speci�c indication or sub-population. During the “Learn” phase many aspects of the study design can be modi�ed: number of subjects, study duration, endpoint selection, treatment duration, number of treatments, patient population. �e adaptation of multiple trial features that are a legitimate concern in con�rmatory studies can become, in fact, an

appealing feature of adaptive designs in exploratory development, leading to enhanced learning to set the appropriate stage for con�rmatory trials.

“The Use of Adaptive Design in the Learn Phase Offers the Best Opportunity to Assess the Product Characteristics that Determine Success”

Judicious use of adaptive designs may increase the information value per resource unit invested by avoiding allocation of patients to non-e�cacious/unsafe therapies and allowing stopping decisions to be made at the earliest possible time point. Ultimately this will accelerate the development of promising therapies to bene�t patients.

Adaptive Clinical Trials can result in the more ethical treatment of patients from at least two perspectives. First, a larger number of patients can be randomized to more favorable, e�ective doses with fewer patients exposed to less e�ective doses. In addition, there is the potential for including fewer total patients in the early stages of the development programmes with attending lower risk of exposure to adverse events. A greater saving in patients may be obtained at the programme level rather than the individual study level as adaptive methodology applied across a programme will reduce the need to repeat ine�ective or inconclusive studies.

Adaptive designs represent an advanced methodology to support drug development. A prerequisite for their successful implementation is an understanding of the underlying methodology and their impact on the logistics of the trial, such as data management and monitoring procedures, electronic data capture/interactive web response services, drug supply management, and data-monitoring committee operating procedures. �ese designs have an impact on drug development strategies, trial protocols, clinical trial material doses and availability, informed consent forms, data analysis, and reporting plans. Adaptive design will also change the dynamics of

enrollment, randomization, and data capturing process, monitoring, and data cleaning systems. To be able to make informed decisions, the sponsor must have access to real time clinical data from all sources and be able to easily review and assess this data.

�is in turn requires �exibility through:

• Data management systems for rapid clinical data• access and data cleaning

• Drug supply systems for drug product planning • and management

• Optimal systems providing randomization, • medication kit management and emergency• unblinding

Planning of Adaptive Clinical Trials is a key feature in the success of such innovative approaches. Determining the optimal characteristics of the study design can be a complex yet critical decision that may require more upfront planning before the protocol can be �nalized. However, there are several commercial software packages that provide a powerful tool for planning, simulation and analysis of complex adaptive designs. FACTS™ is comprehensive software for adaptive designs in the “Learn” phase and includes early stage dose-escalation designs, adaptive dose-ranging studies, interim decision rules based on both e�cacy and safety responses, and population enrichment. ADDPLAN® is a validated software package for adaptive design in the “Con�rm” phase and incorporates sample size re-estimation, adaptive group sequential designs, multiple comparison procedures for multi-armed adaptive trials, including treatment selection designs, �exible combination of clinical research phases, and population enrichment designs.

TYPES OF ADAPTIVE DESIGNS IN “LEARN” STUDIES

Examples of adaptive designs in the Learn phase are shown in Figure 1. Phase I studies are already

FIGURE 5: Adaptive Design Creates Value at Level of the Single Trial, DevelopmentProgramme and Product Pipeline

�is aspect is discussed further in an accompanying white paper that explores the strategic utilisation of adaptive design at the development programme and product pipeline level. Importantly, recent data from Merck has indicated that adoption of an early phase adaptive trial strategy at this level saves at least $200m annually (Schindler J, Disruptive Innovations Conference, Boston, 2012).

“Adaptive Dose Ranging Studies Build Significant Efficiencies at Phase II and Optimize Identification of the Optimal Dose for Confirmatory Studies”

Several adaptive designs have been proposed and some already applied in practice using di�erent working models for the design engine: parsimonious monotonic sigmoid Emax9, Normal Dynamic Linear Model without monotonicity restrictions10, several simple models combined in a Multiple Comparison Procedure (MCP Mod).11

CASE STUDY

�e objective of this trial was to establish the correct dose(s) to take forward into a con�rmatory trial on a novel anti-psychotic drug in acute schizophrenia. �ere was data indicating that the dose-response might be non-monotonic. �erefore, the working model chosen for the design was the Normal Dynamic Linear Model, because it is �exible and robust, and also allows for capturing non-monotonic relationships. Patients were allocated to placebo, active comparator and seven doses of study drug. �e adaptive allocation was targeting both the minimum e�cacious dose (MED) and the dose achieving the maximum response (Dmax) on the primary endpoint, positive PANSS score at week 4. Weekly data updates were used to change the allocation and check early stopping rules based on predictive probabilities. A stopping rule for futility was based on a prespeci�ed threshold that no treatment dose achieves the clinically signi�cant di�erence (CSD) over placebo. A stopping rule for e�cacy required su�cient knowledge about the MED and Dmax and a prespeci�ed con�dence that the Dmax achieves CSD. For details, see Shen et al.12 �e adaptive design allowed earlier (5 months) decision making with a 99% con�dence in futility decision with $14m savings in direct grant cost. �e design, implementation, and outcome of a response adaptive dose-ranging trial of an investigational drug in patients with diabetes, incorporating a

dose expansion approach to �exibly address bothe�cacy and safety has also been reported recently by Berry et al.13 �e early (2 months) futility decision was achieved with a $3.6m saving in direct cost and additional cost savings in avoiding expensive production of phase III drug supplies.

FIGURE 4: Summary of Early Phase Case Studies

SUMMARY

Early phase adaptive design trials address a number of critical development questions that need to be answered before commitment is made to phase III con�rmatory trials. �e application of these innovative designs early in the product development process enhances clinical trial e�ciency, productivity and increases the probability of success at phase III.

�e deployment of these approaches across a pipeline of products brings substantial bene�ts to sponsor companies and enables more e�ective portfolio management and ultimately increases portfolio value (Figure 5).

adaptive by nature, but more recent adaptive designsprovide better estimate of drug safety (e.g. the maxi-mum tolerated dose) and better understanding of itsPK/PD characteristics. Combining the objectives of conventional Phase IIa and IIb studies in a single trial provides the obvious bene�t of reducing the timeline by running the two studies seamlessly under a single protocol with the same clinical team. Furthermore, such approaches achieve trial e�ciency by the prospective integration of data from both IIa and IIb in the �nal analysis.

FIRST-IN-HUMAN SINGLE ASCENDING DOSE ESCALATION DESIGNS

In contrast to traditional dose escalation designs, where for example, cohorts of subjects are allocated to ascending doses (and placebo) until a maximum tolerated dose (MTD) is empirically determined, the continual reassessment method (CRM) models the dose- and/or exposure-toxicity relationship to focus on the identi�cation of the dose-range of interest near the MTD and for this to be explored further. �ese adaptive dose escalation designs can be expanded to also e�ciently establish proof-of-mechanism or proof-of-target modulation, if validated biomarker endpoints are available.

FIGURE 1: Types of Adaptive Designs in the Learn Phase

�e traditional First in Human study objectives of safety and tolerability can be extended to include an

assessment of proof-of-mechanism in the very �rst study conducted with the investigational compound. �e ability to assess proof-of-mechanism in this earlydevelopmental setting provides an opportunity to greatly enhance the clinical development strategy for the compound. �e adaptation is intended to quickly hone in on the predicted dose-range of interest. �is avoids the risks associated with exposing patients in an ine�ective low dose range, whilst minimizing exposure to high doses with an unacceptable tolerability and safety pro�le.

“CRM Offers a Better Approach to Accurately Determining the MTD in Early Phase Oncology Studies”

CRM models are becoming routine for early phase oncology studies where accurate de�nition of the MTD is critical for selecting the optimum dose for e�cacy assessment. Conventional 3+3 designs have been shown to signi�cantly underestimate the MTD which increases the risk of poor outcome in e�cacy trials.3

SEAMLESS MULTIPLE ASCENDING DOSE AND PROOF-OF-CONCEPT STUDY DESIGNS

�e objective of seamless multiple ascending dose and proof-of-concept studies is to combine a multiple ascending dose (MAD) study in patients with the proof-of-concept of the investigational compound.

CASE STUDY

A good example is seen in the treatment of rheumatoid arthritis. �e study design involved a MAD escalation study, in which patients were randomly assigned only to the lowest treatment group or placebo in a 3:1 ratio (active treatment or placebo) which then converted into a parallel enrolling, proof of concept (POC) dose ranging study (see Figure 2). Consistent with the MAD paradigm, dose escalation (allowing subjects in the next higher dose level to

receive a second dose) occurred only after the safety review on 6 subjects assigned to active treatment in the preceding dose had been performed. �is safetyreview was undertaken by a data monitoring committee. Only after it was determined by the data monitoring committee that two consecutive doses of the experimental treatment were well tolerated at a given dose level, were subjects already enrolled in the next higher dose treatment group permitted to receive a second dose of the investigational product. In Stage 2, a utility combining biomarker observations with early readout of a clinical endpoint at week 4 was used to drop dose levels that did not ful�l prespeci�ed necessary conditions (Proof-of-Non-Viability). Viable cohorts and comparator were kept open until each of the surviving treatment arms reached a prede�ned number of patients. �e endpoint to assess proof-of-concept was a 12 week observation on the regulatory accepted endpoint for Rheumatoid Arthritis (ACR20).

�is adaptive design provides several advantages over more traditional clinical trial approaches, without compromising patient safety:

• The seamless design increased the utility of the• information obtained by allowing subjects in both• the MAD and POC stages to provide both • de�nitive safety and e�cacy data.

• Performing the MAD component in subjects with• rheumatoid arthritis receiving concomitant • methotrexate allowed for the earliest • characterization of safety and antigenicity of the• new compound in the clinically relevant • population.

• The adaptive design minimized the number of• subjects that were exposed to ine�ective doses of the• drug, while simultaneously focusing subjects to• doses that were most informative for accurate dose• selection for subsequent con�rmatory trials.

FIGURE 2: Trial Design for Seamless MAD/POC Case Study �erefore, the adaptive design optimizes the bene�t/risk balance for participating subjects via improved e�ciency of decision making in relation to the doses of the new drug studied. In this example the estimated cost saving through combining two trials into one was $1.2m and the development programme was accelerated by 9 months.

“The Development Programme was Accelerated by 9 Months and the Cost Saving of Combining Two Trials in One was Estimated at $1.2M”

PROOF OF CONCEPT AND DOSE-RANGING STUDY DESIGNS

An example of a proof of concept and dose-ranging study design is a dose-�nding study to evaluate the analgesic e�cacy and safety of a single dose of a new non-steroidal anti-in�ammatory drug in subjects experiencing pain after oral dental surgery.4 Six doses of the new drug, placebo and an active comparator are considered. �e adaptive design consists of three stages. At the �rst stage, subjects are equally randomized into three arms: placebo, active comparator, and the top dose of the new drug. At the �rst interim analysis, the study may be stopped either because of lack of assay sensitivity, i.e. inability to establish a prede�ned minimal bene�t of the comparator over and above placebo, or evidence that

the top dose of study drug could not achieve a minimal acceptable level of total pain relief at 8 hours. If the study continues to the second stage, additional subjects would be randomized to placebo, active control, and three doses of the new drug (low, medium, and high) in a 1:1:2:2:1 ratio. Again, the study could be stopped due to lack of assay sensitivity or futility. Otherwise, a D-optimal design will be used to determine both the doses of the new drug (among six possible) and subject allocation ratios for the third stage, see Figure 3. At the end of the study, a four parameter logistic model is used to �t the dose response and make recommendations about the dose to be taken in phase III.

FIGURE 3: Trial Design for POC and Adaptive Dose-Ranging Case Study

CASE STUDY

A full case study has been published in the pain �eld where a Bayesian adaptive dose ranging design was used to evaluate the e�cacy of a new analgesic drug in a proof-of-concept post-herpetic neuralgia trial.5 In total 7 doses of the drug were evaluated in an attempt to de�ne the e�ective dose response curve. �e maximum sample size was 280 patients, however, at the �rst interim analysis (80 patients) the study was stopped for futility as none of the doses showed any meaningful di�erence from placebo. �e study was formally stopped after 133 patients had been enrolled. Stopping of this trial saved an estimated direct grant cost of $0.76m with a further $1.3m savings in internal costs.

“Early Stopping of POC Trials Because of Lack of Efficacy Saves Significant Cost and Enables Valuable Resources to be Re-assigned to Alternative Programmes That May Have a Better Chance of Success”

A recent unpublished example shows that early stopping of a POC adaptive dose ranging trial in post-operative nausea and vomiting saved $0.5m in outsourced costs. In this example, 131 out of 200 patients had been enrolled when the decision was taken at the �rst interim analysis.

RESPONSE ADAPTIVE DOSE-RANGING STUDY DESIGNS

Recent simulation studies6,7 conducted by the PhRMA Working Group on Adaptive Dose-Ranging Studies have found that response-adaptive dose allocation designs are generally superior in performance to conventional pairwise comparisons approaches. Learning about the research question such as identifying a target dose, or estimating the dose-response, is optimized with these designs.

�e main objectives in an adaptive dose-ranging study are to detect the dose response, to determine if any doses(s) meets clinical relevance, to estimate the dose-response, and then to decide on the dose(s) (if any) to take into the con�rmatory Phase III. With adaptive designs for dose-ranging studies many more doses may be considered than in a conventional parallel group design without increasing sample size. �is is achieved mainly through an adaptive allocation rule that is changing the randomization ratio to di�erent treatment arms during the study and putting more subjects in the region that allow most accurate estimation of dose-response and better precision in selecting the target dose. For example, Orlo� et al.8 showed that “for half the sample size, the adaptive design is as powerful and e�cient as the standard approach”.

Clinical TrialTrial Efficiency

Better decision-makingMore information per $ invested

DevelopmentProgramme

Increased probability ofsuccess at Phase III

DevelopmentPipeline

Pipeline efficiency andproductivity

BenefitsStrategy

Value of a

dop

ting a

n adaptive stra

tegy

Page 10: DESIGN AND EXECUTION OF EARLY PHASE …...Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs

DESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALSDESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALS

PAGE 9 PAGE 2

EXECUTIVE SUMMARY

�e adoption of an adaptive design strategy across the product development process brings a number of important bene�ts. �ese include enhanced R&D e�ciency, enhanced R&D productivity and, importantly, increased probability of success at phase III. �e opportunity to manage risk starts in the early phase of a development programme, with a diligent view to appropriate dose selection which impacts not only the success of the downstream clinical trial programme, but also manufacturing costs and value analyses.

Adaptive design trials enhance R&D e�ciency by reducing the need to repeat trials that just miss their clinical end-point or fail to identify the e�ective dose response at the �rst attempt. By avoiding the need to run these trials again, signi�cant cost and time savings are achieved. �is is possible through use of adaptive designs that enable additional patients to be added to achieve statistical signi�cance the �rst time around or by allowing a wider dose range to be studied and then picking doses early in the trial that are in the optimum part of the dose response curve. In addition, early stopping of development programmes because a product is ine�ective enables scarce resources to be redeployed in additional trials which may show more promise. All of these factors increase development e�ciency.

Adaptive designs increase R&D productivity by enabling more accurate de�nition of the e�ective dose response at phase II which enables better design of the pivotal phase III programme which in turn increases the probability of success of the overall development programme. A number of phase III trials fail because the dose is either too high and causes unwanted safety issues or is too low to show su�cient e�cacy. Adaptive design trials enable optimized dose selection before the pivotal trial phase is started.

Another opportunity provided by adaptive design is population enrichment where drug response can be

optimised to speci�c patient sub-populations that respond better to treatment. Many phase III studies fail because the overall e�cacy of treatment is diluted as a consequence of the drug being evaluated in the full trial population rather than in the speci�c subset where the drug works best. Adaptive design enables early selection of the appropriate patient population and increases the probability of success.

“Adaptive Design Enables Selection of the Right Dose for the Right Patient Before Commitment to Expensive Phase III Trials”

Phases I and II are critical steps in the clinical development process as this is where important information about the product has to be generated and assessed before the decision is taken to commit to expensive phase III pivotal studies. �is early phase of development is known as the “Learn Phase” and the data generated relates to the e�ective dose-response, the safety pro�le and therapeutic index, appropriate endpoints, and the patient population best treated by the product under evaluation.

Adaptive designs have an important role to play in both phase I and phase II, but it is in the latter phase where they add the greatest value in terms of risk management. Validated methodologies are available to design early phase adaptive trials and this is an area of signi�cant interest across the pharmaceutical industry.

“Getting it Right at Phase II is a Major Objective for Adaptive Design and is the Point at which Phase III Success is Defined”

Companies that adopt a comprehensive adaptive design strategy in the “Learn” phase will make better development decisions, will build development e�ciency, will increase the probability of success of the molecules that enter phase III, ultimately bringing e�ective products to the market more e�ciently. Competitor companies that persevere with conventional trial designs will remain slow, ine�cient

candidate optimisation to Proof of concept stages. Moira assists partner companies with both acquisition and due diligence of in licensed compounds and strategic support for the out-licensing of successful candidates. �is has included design, planning and management of, strategic and clinical development plans, due diligence, decision analysis, preclinical, POC and Phase II programmes in a range of, indications including respiratory, anti-infective, immunomodulation, CNS and cardiovascular. �is experience covers a large range of targets and mecha-nisms of action. In order to support the strategic and clinical development planning Moira has worked extensively with international Key Opinion Leaders in respiratory, anti-infective and transplant indications.

PHIL BIRCH,SVP, CORPORATE BRAND MANAGER

Dr. Phil Birch is a Senior Vice President at Aptiv Solutions and has global responsibility for co-ordinating market education and client awareness in the adaptive clinical trials �eld. Dr. Birch has worked in the pharmaceutical industry for over 27 years and has held executive positions in corporate development, business development and R&D in a number of consulting, biotechnology and top 10 pharmaceutical companies.

REFERENCES

1. Dragalin V. Adaptive Designs: Terminology and Classi�cation. Drug Information Journal, 2006; 40: 425-436. 2. Gallo P, Chuang-Stein C, Dragalin V, Gaydos B, Krams M, Pinheiro J. Executive Summary of the PhRMA Working Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs for Phase I Oncology Trials, Clinical Operations Summit, Brussels, 2012).4. Dragalin V, Hsuan F, and Padmanabhan SK. Adaptive Designs for Dose-Finding Studies Based on Sigmoid Emax Model. J Biopharmaceutical Statistics 2007; 17:1051-1070.5. Smith MK, Jones I, Morris MF, Grieve A and Tan K. Implementation of a Bayesian adaptive design in a proof of concept study. Pharmaceutical Statistics 2006; 5:39-50. 6. Bornkamp B, Bretz F, Dmitrienko A, Enas G, Gaydos B, Hsu CH, König F, Krams M, Liu Q, Neuenschwander B, Parke T, Pinheiro J, Roy A, Sax R, Shen F. Innovative approaches for designing and analyzing adaptive dose-ranging trials: White Paper from the PhRMA working group on “Adaptive Dose-Ranging Studies”. J of Biopharmaceutical Statistics 2007; 17: 965–995. 7. Dragalin V, Bornkamp B, Bretz F, Miller F, Padmanabhan SK, Patel N, Perevozskaya I, Pinheiro J, and Smith J. A Simulation Study to Compare New Adaptive Dose–Ranging Designs. Statistics in Biopharmaceutical Research 2010; 2:487-512.8. Orlo� J, Douglas F, Pinheiro J, Levinson S, Branson M, Chaturvedi P, Ette E, Gallo P, Hirsch G, Mehta C, Patel N, Sabir S, Springs S, Stanski D, Golub H, Evers M, Fleming E, Singh N, Tramontin T. �e future of drug development: advancing clinical trial design. Nature Review-Drug Discovery. 2009; 8:940-957.9. Padmanabhan SK and Dragalin V. Adaptive Dc-optimal designs for dose �nding based on a continuous e�cacy endpoint. Biometrical Journal 2010; 52:836-852. 10. Grieve AP and Krams M. ASTIN: a Bayesian adaptive dose-response trial in acute stroke. Clinical Trials 2005; 2:340-351. 11. Bornkamp B, Pinheiro J, Bretz F. MCPMod: An R Package for the Design and Analysis of Dose-Finding Studies. Journal of Statistical Software 2009; 29:1-23.12. Shen J, Preskorn S, Dragalin V, Slomkowski M, Padmanabhan SK, Fardipour P, Sharma A, and Krams M. How Adaptive Trial Designs Can Increase E�ciency in Psychiatric Drug Development: A Case Study. Innovations in Clinical Neuroscience 2011; 8: 26-34.13. Berry SM, Spinelli W, Littman GS, Liang JZ, Fardipour P, Berry DA, Lewis RL, and Krams M. A Bayesian dose-�nding trial with adaptive dose expansion to �exibly assess e�cacy and safety of an investigational drug. Clinical Trials 2010; 7: 121-135.

ABOUT THE AUTHORS

VLAD DRAGALIN, SVP, INNOVATION CENTRE

Dr. Vlad Dragalin is a well-known adaptive design expert with 25 years of experience in developing the statistical methodology of adaptive designs and with over 12 years experience in pharmaceutical industry. Vlad is a Senior Vice President, Software Develop-ment and Consulting, in the Innovation Center at Aptiv Solutions. Vlad works with clients to improve the drug development process through the use of adaptive designs and to assist them with the trial design, simulation and execution, developing adaptive trials software tools, and working with project teams in clinical operations, data management and statistics during study conduct.

SARAH ARBE-BARNES,SVP, TRANSLATIONAL SCIENCES

Dr. Sarah Arbe-Barnes has over 24 years of experience in drug development and project leadership in the Pharma and service sector industry to include in/out-licensing and due diligence, strategic development and business planning, fundraising for start-ups/SMEs( not-for-pro�t sector and VC), along with optimized market access strategies and marketing. Sarah has led recent product approval programmes, overseeing a broad range of clinical and nonclinical project design and management activities from the point of candidate selection. Sarah heads the Transla-tional Sciences division of Aptiv Solutions, which is a dedicated global consultancy group, focused on the provision of drug and device development advice and expertise for small molecules and biologics.

MOIRA THOMSON,VP, TRANSLATIONAL SCIENCES

Moira �omson has over 15 years of experience in the Pharmaceutical Industry, and is involved in the design, planning and management of global develop-ment projects for a mixture of small and medium sized Pharma, and Biotech companies covering lead

and will fail to make the commercial returns that investors and shareholders demand. “Adaptive Design Builds Competitive Advantage”

�is White Paper summarizes the important aspects of early phase adaptive design and provides a summary of selected case studies which demonstrate the value of this innovative approach. It has been written to inform senior R&D decision-makers about the critical role of adaptive design in early phase development and the signi�cant bene�ts that this approach can bring.

BACKGROUND TO EARLY PHASE ADAPTIVE TRIAL DESIGN

Adaptive Clinical Trials are viewed as supporting a change in the development process as focus moves from blockbuster products to more specialized medicines that require a �exible cost-e�cient approach to investigating their e�ectiveness. Adaptive designs are statistical methodologies applied to speci�c stages of drug development where real time learning from accumulating trial data is applied to optimize subsequent study execution.1 �e bene�t of an Adaptive Clinical Trial lies in the monitoring of data to make design modi�cation adjustments to an aspect of a trial leading to quicker development decisions around key go/no-go milestones that impact time to market and/or minimise costly R&D losses.

�e adaptations are prospectively de�ned prior to the start of the trial and can include stopping early either for futility or success, expanding the sample size due to greater than expected data variability, or allocating patients preferentially to treatment regimens with a better therapeutic index. Virtually any aspect of the trial may be the potential target of design modi�cations. Importantly, these modi�cations are a design feature aimed to enhance the trial, not a remedy for inadequate planning. �ey are not ad hoc study corrections made via protocol amendment.2

�e speci�c adaptations considered, and the basis of their implementation, are carefully de�ned based on strict rules, justi�ed statistically and scienti�cally, and are an integral part of the �nal, pre-recruitment trial protocol. Failing this would compromise the interpretation and acceptance of study results.

Adaptive designs often employ frequent interim analyses of all accumulated data (and, possibly, external trial data) to determine whether pre-planned design modi�cations will be ‘triggered’. Interim analyses partition the trial into multiple stages, each trial stage’s characteristics (number of arms, number of patients to be enrolled, their allocation between arms, stage duration, etc.) de�ned by the preceding interim analysis results. �e ability to periodically, or even continually, examine available data to determine whether trial modi�cations are necessary and implement pre-de�ned design changes when indicated, gives adaptive design its strength and �exibility.

�e use of adaptive design in the exploratory stage of clinical trials can increase the e�ciency of drug development by improving our ability to e�ciently learn about the dose-response and better determine whether to take a drug (and the right dose of the drug and in the right population) forward into later phase testing. �ese designs explicitly address multiple trial goals, adaptively allocate subjects according to ongoing information needs, and allow termination for both early success and futility considerations. �is approach can maximize the ability to test a larger number of doses in a single trial while simultaneously increasing the e�ciency of the trial in terms of making better go/no-go decisions about continuing the trial and/or the development of the drug for a speci�c indication or sub-population. During the “Learn” phase many aspects of the study design can be modi�ed: number of subjects, study duration, endpoint selection, treatment duration, number of treatments, patient population. �e adaptation of multiple trial features that are a legitimate concern in con�rmatory studies can become, in fact, an

appealing feature of adaptive designs in exploratory development, leading to enhanced learning to set the appropriate stage for con�rmatory trials.

“The Use of Adaptive Design in the Learn Phase Offers the Best Opportunity to Assess the Product Characteristics that Determine Success”

Judicious use of adaptive designs may increase the information value per resource unit invested by avoiding allocation of patients to non-e�cacious/unsafe therapies and allowing stopping decisions to be made at the earliest possible time point. Ultimately this will accelerate the development of promising therapies to bene�t patients.

Adaptive Clinical Trials can result in the more ethical treatment of patients from at least two perspectives. First, a larger number of patients can be randomized to more favorable, e�ective doses with fewer patients exposed to less e�ective doses. In addition, there is the potential for including fewer total patients in the early stages of the development programmes with attending lower risk of exposure to adverse events. A greater saving in patients may be obtained at the programme level rather than the individual study level as adaptive methodology applied across a programme will reduce the need to repeat ine�ective or inconclusive studies.

Adaptive designs represent an advanced methodology to support drug development. A prerequisite for their successful implementation is an understanding of the underlying methodology and their impact on the logistics of the trial, such as data management and monitoring procedures, electronic data capture/interactive web response services, drug supply management, and data-monitoring committee operating procedures. �ese designs have an impact on drug development strategies, trial protocols, clinical trial material doses and availability, informed consent forms, data analysis, and reporting plans. Adaptive design will also change the dynamics of

enrollment, randomization, and data capturing process, monitoring, and data cleaning systems. To be able to make informed decisions, the sponsor must have access to real time clinical data from all sources and be able to easily review and assess this data.

�is in turn requires �exibility through:

• Data management systems for rapid clinical data• access and data cleaning

• Drug supply systems for drug product planning • and management

• Optimal systems providing randomization, • medication kit management and emergency• unblinding

Planning of Adaptive Clinical Trials is a key feature in the success of such innovative approaches. Determining the optimal characteristics of the study design can be a complex yet critical decision that may require more upfront planning before the protocol can be �nalized. However, there are several commercial software packages that provide a powerful tool for planning, simulation and analysis of complex adaptive designs. FACTS™ is comprehensive software for adaptive designs in the “Learn” phase and includes early stage dose-escalation designs, adaptive dose-ranging studies, interim decision rules based on both e�cacy and safety responses, and population enrichment. ADDPLAN® is a validated software package for adaptive design in the “Con�rm” phase and incorporates sample size re-estimation, adaptive group sequential designs, multiple comparison procedures for multi-armed adaptive trials, including treatment selection designs, �exible combination of clinical research phases, and population enrichment designs.

TYPES OF ADAPTIVE DESIGNS IN “LEARN” STUDIES

Examples of adaptive designs in the Learn phase are shown in Figure 1. Phase I studies are already

FIGURE 5: Adaptive Design Creates Value at Level of the Single Trial, DevelopmentProgramme and Product Pipeline

�is aspect is discussed further in an accompanying white paper that explores the strategic utilisation of adaptive design at the development programme and product pipeline level. Importantly, recent data from Merck has indicated that adoption of an early phase adaptive trial strategy at this level saves at least $200m annually (Schindler J, Disruptive Innovations Conference, Boston, 2012).

“Adaptive Dose Ranging Studies Build Significant Efficiencies at Phase II and Optimize Identification of the Optimal Dose for Confirmatory Studies”

Several adaptive designs have been proposed and some already applied in practice using di�erent working models for the design engine: parsimonious monotonic sigmoid Emax9, Normal Dynamic Linear Model without monotonicity restrictions10, several simple models combined in a Multiple Comparison Procedure (MCP Mod).11

CASE STUDY

�e objective of this trial was to establish the correct dose(s) to take forward into a con�rmatory trial on a novel anti-psychotic drug in acute schizophrenia. �ere was data indicating that the dose-response might be non-monotonic. �erefore, the working model chosen for the design was the Normal Dynamic Linear Model, because it is �exible and robust, and also allows for capturing non-monotonic relationships. Patients were allocated to placebo, active comparator and seven doses of study drug. �e adaptive allocation was targeting both the minimum e�cacious dose (MED) and the dose achieving the maximum response (Dmax) on the primary endpoint, positive PANSS score at week 4. Weekly data updates were used to change the allocation and check early stopping rules based on predictive probabilities. A stopping rule for futility was based on a prespeci�ed threshold that no treatment dose achieves the clinically signi�cant di�erence (CSD) over placebo. A stopping rule for e�cacy required su�cient knowledge about the MED and Dmax and a prespeci�ed con�dence that the Dmax achieves CSD. For details, see Shen et al.12 �e adaptive design allowed earlier (5 months) decision making with a 99% con�dence in futility decision with $14m savings in direct grant cost. �e design, implementation, and outcome of a response adaptive dose-ranging trial of an investigational drug in patients with diabetes, incorporating a

dose expansion approach to �exibly address bothe�cacy and safety has also been reported recently by Berry et al.13 �e early (2 months) futility decision was achieved with a $3.6m saving in direct cost and additional cost savings in avoiding expensive production of phase III drug supplies.

FIGURE 4: Summary of Early Phase Case Studies

SUMMARY

Early phase adaptive design trials address a number of critical development questions that need to be answered before commitment is made to phase III con�rmatory trials. �e application of these innovative designs early in the product development process enhances clinical trial e�ciency, productivity and increases the probability of success at phase III.

�e deployment of these approaches across a pipeline of products brings substantial bene�ts to sponsor companies and enables more e�ective portfolio management and ultimately increases portfolio value (Figure 5).

adaptive by nature, but more recent adaptive designsprovide better estimate of drug safety (e.g. the maxi-mum tolerated dose) and better understanding of itsPK/PD characteristics. Combining the objectives of conventional Phase IIa and IIb studies in a single trial provides the obvious bene�t of reducing the timeline by running the two studies seamlessly under a single protocol with the same clinical team. Furthermore, such approaches achieve trial e�ciency by the prospective integration of data from both IIa and IIb in the �nal analysis.

FIRST-IN-HUMAN SINGLE ASCENDING DOSE ESCALATION DESIGNS

In contrast to traditional dose escalation designs, where for example, cohorts of subjects are allocated to ascending doses (and placebo) until a maximum tolerated dose (MTD) is empirically determined, the continual reassessment method (CRM) models the dose- and/or exposure-toxicity relationship to focus on the identi�cation of the dose-range of interest near the MTD and for this to be explored further. �ese adaptive dose escalation designs can be expanded to also e�ciently establish proof-of-mechanism or proof-of-target modulation, if validated biomarker endpoints are available.

FIGURE 1: Types of Adaptive Designs in the Learn Phase

�e traditional First in Human study objectives of safety and tolerability can be extended to include an

assessment of proof-of-mechanism in the very �rst study conducted with the investigational compound. �e ability to assess proof-of-mechanism in this earlydevelopmental setting provides an opportunity to greatly enhance the clinical development strategy for the compound. �e adaptation is intended to quickly hone in on the predicted dose-range of interest. �is avoids the risks associated with exposing patients in an ine�ective low dose range, whilst minimizing exposure to high doses with an unacceptable tolerability and safety pro�le.

“CRM Offers a Better Approach to Accurately Determining the MTD in Early Phase Oncology Studies”

CRM models are becoming routine for early phase oncology studies where accurate de�nition of the MTD is critical for selecting the optimum dose for e�cacy assessment. Conventional 3+3 designs have been shown to signi�cantly underestimate the MTD which increases the risk of poor outcome in e�cacy trials.3

SEAMLESS MULTIPLE ASCENDING DOSE AND PROOF-OF-CONCEPT STUDY DESIGNS

�e objective of seamless multiple ascending dose and proof-of-concept studies is to combine a multiple ascending dose (MAD) study in patients with the proof-of-concept of the investigational compound.

CASE STUDY

A good example is seen in the treatment of rheumatoid arthritis. �e study design involved a MAD escalation study, in which patients were randomly assigned only to the lowest treatment group or placebo in a 3:1 ratio (active treatment or placebo) which then converted into a parallel enrolling, proof of concept (POC) dose ranging study (see Figure 2). Consistent with the MAD paradigm, dose escalation (allowing subjects in the next higher dose level to

receive a second dose) occurred only after the safety review on 6 subjects assigned to active treatment in the preceding dose had been performed. �is safetyreview was undertaken by a data monitoring committee. Only after it was determined by the data monitoring committee that two consecutive doses of the experimental treatment were well tolerated at a given dose level, were subjects already enrolled in the next higher dose treatment group permitted to receive a second dose of the investigational product. In Stage 2, a utility combining biomarker observations with early readout of a clinical endpoint at week 4 was used to drop dose levels that did not ful�l prespeci�ed necessary conditions (Proof-of-Non-Viability). Viable cohorts and comparator were kept open until each of the surviving treatment arms reached a prede�ned number of patients. �e endpoint to assess proof-of-concept was a 12 week observation on the regulatory accepted endpoint for Rheumatoid Arthritis (ACR20).

�is adaptive design provides several advantages over more traditional clinical trial approaches, without compromising patient safety:

• The seamless design increased the utility of the• information obtained by allowing subjects in both• the MAD and POC stages to provide both • de�nitive safety and e�cacy data.

• Performing the MAD component in subjects with• rheumatoid arthritis receiving concomitant • methotrexate allowed for the earliest • characterization of safety and antigenicity of the• new compound in the clinically relevant • population.

• The adaptive design minimized the number of• subjects that were exposed to ine�ective doses of the• drug, while simultaneously focusing subjects to• doses that were most informative for accurate dose• selection for subsequent con�rmatory trials.

FIGURE 2: Trial Design for Seamless MAD/POC Case Study �erefore, the adaptive design optimizes the bene�t/risk balance for participating subjects via improved e�ciency of decision making in relation to the doses of the new drug studied. In this example the estimated cost saving through combining two trials into one was $1.2m and the development programme was accelerated by 9 months.

“The Development Programme was Accelerated by 9 Months and the Cost Saving of Combining Two Trials in One was Estimated at $1.2M”

PROOF OF CONCEPT AND DOSE-RANGING STUDY DESIGNS

An example of a proof of concept and dose-ranging study design is a dose-�nding study to evaluate the analgesic e�cacy and safety of a single dose of a new non-steroidal anti-in�ammatory drug in subjects experiencing pain after oral dental surgery.4 Six doses of the new drug, placebo and an active comparator are considered. �e adaptive design consists of three stages. At the �rst stage, subjects are equally randomized into three arms: placebo, active comparator, and the top dose of the new drug. At the �rst interim analysis, the study may be stopped either because of lack of assay sensitivity, i.e. inability to establish a prede�ned minimal bene�t of the comparator over and above placebo, or evidence that

the top dose of study drug could not achieve a minimal acceptable level of total pain relief at 8 hours. If the study continues to the second stage, additional subjects would be randomized to placebo, active control, and three doses of the new drug (low, medium, and high) in a 1:1:2:2:1 ratio. Again, the study could be stopped due to lack of assay sensitivity or futility. Otherwise, a D-optimal design will be used to determine both the doses of the new drug (among six possible) and subject allocation ratios for the third stage, see Figure 3. At the end of the study, a four parameter logistic model is used to �t the dose response and make recommendations about the dose to be taken in phase III.

FIGURE 3: Trial Design for POC and Adaptive Dose-Ranging Case Study

CASE STUDY

A full case study has been published in the pain �eld where a Bayesian adaptive dose ranging design was used to evaluate the e�cacy of a new analgesic drug in a proof-of-concept post-herpetic neuralgia trial.5 In total 7 doses of the drug were evaluated in an attempt to de�ne the e�ective dose response curve. �e maximum sample size was 280 patients, however, at the �rst interim analysis (80 patients) the study was stopped for futility as none of the doses showed any meaningful di�erence from placebo. �e study was formally stopped after 133 patients had been enrolled. Stopping of this trial saved an estimated direct grant cost of $0.76m with a further $1.3m savings in internal costs.

“Early Stopping of POC Trials Because of Lack of Efficacy Saves Significant Cost and Enables Valuable Resources to be Re-assigned to Alternative Programmes That May Have a Better Chance of Success”

A recent unpublished example shows that early stopping of a POC adaptive dose ranging trial in post-operative nausea and vomiting saved $0.5m in outsourced costs. In this example, 131 out of 200 patients had been enrolled when the decision was taken at the �rst interim analysis.

RESPONSE ADAPTIVE DOSE-RANGING STUDY DESIGNS

Recent simulation studies6,7 conducted by the PhRMA Working Group on Adaptive Dose-Ranging Studies have found that response-adaptive dose allocation designs are generally superior in performance to conventional pairwise comparisons approaches. Learning about the research question such as identifying a target dose, or estimating the dose-response, is optimized with these designs.

�e main objectives in an adaptive dose-ranging study are to detect the dose response, to determine if any doses(s) meets clinical relevance, to estimate the dose-response, and then to decide on the dose(s) (if any) to take into the con�rmatory Phase III. With adaptive designs for dose-ranging studies many more doses may be considered than in a conventional parallel group design without increasing sample size. �is is achieved mainly through an adaptive allocation rule that is changing the randomization ratio to di�erent treatment arms during the study and putting more subjects in the region that allow most accurate estimation of dose-response and better precision in selecting the target dose. For example, Orlo� et al.8 showed that “for half the sample size, the adaptive design is as powerful and e�cient as the standard approach”.

Page 11: DESIGN AND EXECUTION OF EARLY PHASE …...Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs

DESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALSDESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALS

PAGE 1 PAGE 10

EXECUTIVE SUMMARY

�e adoption of an adaptive design strategy across the product development process brings a number of important bene�ts. �ese include enhanced R&D e�ciency, enhanced R&D productivity and, importantly, increased probability of success at phase III. �e opportunity to manage risk starts in the early phase of a development programme, with a diligent view to appropriate dose selection which impacts not only the success of the downstream clinical trial programme, but also manufacturing costs and value analyses.

Adaptive design trials enhance R&D e�ciency by reducing the need to repeat trials that just miss their clinical end-point or fail to identify the e�ective dose response at the �rst attempt. By avoiding the need to run these trials again, signi�cant cost and time savings are achieved. �is is possible through use of adaptive designs that enable additional patients to be added to achieve statistical signi�cance the �rst time around or by allowing a wider dose range to be studied and then picking doses early in the trial that are in the optimum part of the dose response curve. In addition, early stopping of development programmes because a product is ine�ective enables scarce resources to be redeployed in additional trials which may show more promise. All of these factors increase development e�ciency.

Adaptive designs increase R&D productivity by enabling more accurate de�nition of the e�ective dose response at phase II which enables better design of the pivotal phase III programme which in turn increases the probability of success of the overall development programme. A number of phase III trials fail because the dose is either too high and causes unwanted safety issues or is too low to show su�cient e�cacy. Adaptive design trials enable optimized dose selection before the pivotal trial phase is started.

Another opportunity provided by adaptive design is population enrichment where drug response can be

optimised to speci�c patient sub-populations that respond better to treatment. Many phase III studies fail because the overall e�cacy of treatment is diluted as a consequence of the drug being evaluated in the full trial population rather than in the speci�c subset where the drug works best. Adaptive design enables early selection of the appropriate patient population and increases the probability of success.

“Adaptive Design Enables Selection of the Right Dose for the Right Patient Before Commitment to Expensive Phase III Trials”

Phases I and II are critical steps in the clinical development process as this is where important information about the product has to be generated and assessed before the decision is taken to commit to expensive phase III pivotal studies. �is early phase of development is known as the “Learn Phase” and the data generated relates to the e�ective dose-response, the safety pro�le and therapeutic index, appropriate endpoints, and the patient population best treated by the product under evaluation.

Adaptive designs have an important role to play in both phase I and phase II, but it is in the latter phase where they add the greatest value in terms of risk management. Validated methodologies are available to design early phase adaptive trials and this is an area of signi�cant interest across the pharmaceutical industry.

“Getting it Right at Phase II is a Major Objective for Adaptive Design and is the Point at which Phase III Success is Defined”

Companies that adopt a comprehensive adaptive design strategy in the “Learn” phase will make better development decisions, will build development e�ciency, will increase the probability of success of the molecules that enter phase III, ultimately bringing e�ective products to the market more e�ciently. Competitor companies that persevere with conventional trial designs will remain slow, ine�cient

candidate optimisation to Proof of concept stages. Moira assists partner companies with both acquisition and due diligence of in licensed compounds and strategic support for the out-licensing of successful candidates. �is has included design, planning and management of, strategic and clinical development plans, due diligence, decision analysis, preclinical, POC and Phase II programmes in a range of, indications including respiratory, anti-infective, immunomodulation, CNS and cardiovascular. �is experience covers a large range of targets and mecha-nisms of action. In order to support the strategic and clinical development planning Moira has worked extensively with international Key Opinion Leaders in respiratory, anti-infective and transplant indications.

PHIL BIRCH,SVP, CORPORATE BRAND MANAGER

Dr. Phil Birch is a Senior Vice President at Aptiv Solutions and has global responsibility for co-ordinating market education and client awareness in the adaptive clinical trials �eld. Dr. Birch has worked in the pharmaceutical industry for over 27 years and has held executive positions in corporate development, business development and R&D in a number of consulting, biotechnology and top 10 pharmaceutical companies.

REFERENCES

1. Dragalin V. Adaptive Designs: Terminology and Classi�cation. Drug Information Journal, 2006; 40: 425-436. 2. Gallo P, Chuang-Stein C, Dragalin V, Gaydos B, Krams M, Pinheiro J. Executive Summary of the PhRMA Working Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs for Phase I Oncology Trials, Clinical Operations Summit, Brussels, 2012).4. Dragalin V, Hsuan F, and Padmanabhan SK. Adaptive Designs for Dose-Finding Studies Based on Sigmoid Emax Model. J Biopharmaceutical Statistics 2007; 17:1051-1070.5. Smith MK, Jones I, Morris MF, Grieve A and Tan K. Implementation of a Bayesian adaptive design in a proof of concept study. Pharmaceutical Statistics 2006; 5:39-50. 6. Bornkamp B, Bretz F, Dmitrienko A, Enas G, Gaydos B, Hsu CH, König F, Krams M, Liu Q, Neuenschwander B, Parke T, Pinheiro J, Roy A, Sax R, Shen F. Innovative approaches for designing and analyzing adaptive dose-ranging trials: White Paper from the PhRMA working group on “Adaptive Dose-Ranging Studies”. J of Biopharmaceutical Statistics 2007; 17: 965–995. 7. Dragalin V, Bornkamp B, Bretz F, Miller F, Padmanabhan SK, Patel N, Perevozskaya I, Pinheiro J, and Smith J. A Simulation Study to Compare New Adaptive Dose–Ranging Designs. Statistics in Biopharmaceutical Research 2010; 2:487-512.8. Orlo� J, Douglas F, Pinheiro J, Levinson S, Branson M, Chaturvedi P, Ette E, Gallo P, Hirsch G, Mehta C, Patel N, Sabir S, Springs S, Stanski D, Golub H, Evers M, Fleming E, Singh N, Tramontin T. �e future of drug development: advancing clinical trial design. Nature Review-Drug Discovery. 2009; 8:940-957.9. Padmanabhan SK and Dragalin V. Adaptive Dc-optimal designs for dose �nding based on a continuous e�cacy endpoint. Biometrical Journal 2010; 52:836-852. 10. Grieve AP and Krams M. ASTIN: a Bayesian adaptive dose-response trial in acute stroke. Clinical Trials 2005; 2:340-351. 11. Bornkamp B, Pinheiro J, Bretz F. MCPMod: An R Package for the Design and Analysis of Dose-Finding Studies. Journal of Statistical Software 2009; 29:1-23.12. Shen J, Preskorn S, Dragalin V, Slomkowski M, Padmanabhan SK, Fardipour P, Sharma A, and Krams M. How Adaptive Trial Designs Can Increase E�ciency in Psychiatric Drug Development: A Case Study. Innovations in Clinical Neuroscience 2011; 8: 26-34.13. Berry SM, Spinelli W, Littman GS, Liang JZ, Fardipour P, Berry DA, Lewis RL, and Krams M. A Bayesian dose-�nding trial with adaptive dose expansion to �exibly assess e�cacy and safety of an investigational drug. Clinical Trials 2010; 7: 121-135.

ABOUT THE AUTHORS

VLAD DRAGALIN, SVP, INNOVATION CENTRE

Dr. Vlad Dragalin is a well-known adaptive design expert with 25 years of experience in developing the statistical methodology of adaptive designs and with over 12 years experience in pharmaceutical industry. Vlad is a Senior Vice President, Software Develop-ment and Consulting, in the Innovation Center at Aptiv Solutions. Vlad works with clients to improve the drug development process through the use of adaptive designs and to assist them with the trial design, simulation and execution, developing adaptive trials software tools, and working with project teams in clinical operations, data management and statistics during study conduct.

SARAH ARBE-BARNES,SVP, TRANSLATIONAL SCIENCES

Dr. Sarah Arbe-Barnes has over 24 years of experience in drug development and project leadership in the Pharma and service sector industry to include in/out-licensing and due diligence, strategic development and business planning, fundraising for start-ups/SMEs( not-for-pro�t sector and VC), along with optimized market access strategies and marketing. Sarah has led recent product approval programmes, overseeing a broad range of clinical and nonclinical project design and management activities from the point of candidate selection. Sarah heads the Transla-tional Sciences division of Aptiv Solutions, which is a dedicated global consultancy group, focused on the provision of drug and device development advice and expertise for small molecules and biologics.

MOIRA THOMSON,VP, TRANSLATIONAL SCIENCES

Moira �omson has over 15 years of experience in the Pharmaceutical Industry, and is involved in the design, planning and management of global develop-ment projects for a mixture of small and medium sized Pharma, and Biotech companies covering lead

and will fail to make the commercial returns that investors and shareholders demand. “Adaptive Design Builds Competitive Advantage”

�is White Paper summarizes the important aspects of early phase adaptive design and provides a summary of selected case studies which demonstrate the value of this innovative approach. It has been written to inform senior R&D decision-makers about the critical role of adaptive design in early phase development and the signi�cant bene�ts that this approach can bring.

BACKGROUND TO EARLY PHASE ADAPTIVE TRIAL DESIGN

Adaptive Clinical Trials are viewed as supporting a change in the development process as focus moves from blockbuster products to more specialized medicines that require a �exible cost-e�cient approach to investigating their e�ectiveness. Adaptive designs are statistical methodologies applied to speci�c stages of drug development where real time learning from accumulating trial data is applied to optimize subsequent study execution.1 �e bene�t of an Adaptive Clinical Trial lies in the monitoring of data to make design modi�cation adjustments to an aspect of a trial leading to quicker development decisions around key go/no-go milestones that impact time to market and/or minimise costly R&D losses.

�e adaptations are prospectively de�ned prior to the start of the trial and can include stopping early either for futility or success, expanding the sample size due to greater than expected data variability, or allocating patients preferentially to treatment regimens with a better therapeutic index. Virtually any aspect of the trial may be the potential target of design modi�cations. Importantly, these modi�cations are a design feature aimed to enhance the trial, not a remedy for inadequate planning. �ey are not ad hoc study corrections made via protocol amendment.2

�e speci�c adaptations considered, and the basis of their implementation, are carefully de�ned based on strict rules, justi�ed statistically and scienti�cally, and are an integral part of the �nal, pre-recruitment trial protocol. Failing this would compromise the interpretation and acceptance of study results.

Adaptive designs often employ frequent interim analyses of all accumulated data (and, possibly, external trial data) to determine whether pre-planned design modi�cations will be ‘triggered’. Interim analyses partition the trial into multiple stages, each trial stage’s characteristics (number of arms, number of patients to be enrolled, their allocation between arms, stage duration, etc.) de�ned by the preceding interim analysis results. �e ability to periodically, or even continually, examine available data to determine whether trial modi�cations are necessary and implement pre-de�ned design changes when indicated, gives adaptive design its strength and �exibility.

�e use of adaptive design in the exploratory stage of clinical trials can increase the e�ciency of drug development by improving our ability to e�ciently learn about the dose-response and better determine whether to take a drug (and the right dose of the drug and in the right population) forward into later phase testing. �ese designs explicitly address multiple trial goals, adaptively allocate subjects according to ongoing information needs, and allow termination for both early success and futility considerations. �is approach can maximize the ability to test a larger number of doses in a single trial while simultaneously increasing the e�ciency of the trial in terms of making better go/no-go decisions about continuing the trial and/or the development of the drug for a speci�c indication or sub-population. During the “Learn” phase many aspects of the study design can be modi�ed: number of subjects, study duration, endpoint selection, treatment duration, number of treatments, patient population. �e adaptation of multiple trial features that are a legitimate concern in con�rmatory studies can become, in fact, an

appealing feature of adaptive designs in exploratory development, leading to enhanced learning to set the appropriate stage for con�rmatory trials.

“The Use of Adaptive Design in the Learn Phase Offers the Best Opportunity to Assess the Product Characteristics that Determine Success”

Judicious use of adaptive designs may increase the information value per resource unit invested by avoiding allocation of patients to non-e�cacious/unsafe therapies and allowing stopping decisions to be made at the earliest possible time point. Ultimately this will accelerate the development of promising therapies to bene�t patients.

Adaptive Clinical Trials can result in the more ethical treatment of patients from at least two perspectives. First, a larger number of patients can be randomized to more favorable, e�ective doses with fewer patients exposed to less e�ective doses. In addition, there is the potential for including fewer total patients in the early stages of the development programmes with attending lower risk of exposure to adverse events. A greater saving in patients may be obtained at the programme level rather than the individual study level as adaptive methodology applied across a programme will reduce the need to repeat ine�ective or inconclusive studies.

Adaptive designs represent an advanced methodology to support drug development. A prerequisite for their successful implementation is an understanding of the underlying methodology and their impact on the logistics of the trial, such as data management and monitoring procedures, electronic data capture/interactive web response services, drug supply management, and data-monitoring committee operating procedures. �ese designs have an impact on drug development strategies, trial protocols, clinical trial material doses and availability, informed consent forms, data analysis, and reporting plans. Adaptive design will also change the dynamics of

enrollment, randomization, and data capturing process, monitoring, and data cleaning systems. To be able to make informed decisions, the sponsor must have access to real time clinical data from all sources and be able to easily review and assess this data.

�is in turn requires �exibility through:

• Data management systems for rapid clinical data• access and data cleaning

• Drug supply systems for drug product planning • and management

• Optimal systems providing randomization, • medication kit management and emergency• unblinding

Planning of Adaptive Clinical Trials is a key feature in the success of such innovative approaches. Determining the optimal characteristics of the study design can be a complex yet critical decision that may require more upfront planning before the protocol can be �nalized. However, there are several commercial software packages that provide a powerful tool for planning, simulation and analysis of complex adaptive designs. FACTS™ is comprehensive software for adaptive designs in the “Learn” phase and includes early stage dose-escalation designs, adaptive dose-ranging studies, interim decision rules based on both e�cacy and safety responses, and population enrichment. ADDPLAN® is a validated software package for adaptive design in the “Con�rm” phase and incorporates sample size re-estimation, adaptive group sequential designs, multiple comparison procedures for multi-armed adaptive trials, including treatment selection designs, �exible combination of clinical research phases, and population enrichment designs.

TYPES OF ADAPTIVE DESIGNS IN “LEARN” STUDIES

Examples of adaptive designs in the Learn phase are shown in Figure 1. Phase I studies are already

FIGURE 5: Adaptive Design Creates Value at Level of the Single Trial, DevelopmentProgramme and Product Pipeline

�is aspect is discussed further in an accompanying white paper that explores the strategic utilisation of adaptive design at the development programme and product pipeline level. Importantly, recent data from Merck has indicated that adoption of an early phase adaptive trial strategy at this level saves at least $200m annually (Schindler J, Disruptive Innovations Conference, Boston, 2012).

“Adaptive Dose Ranging Studies Build Significant Efficiencies at Phase II and Optimize Identification of the Optimal Dose for Confirmatory Studies”

Several adaptive designs have been proposed and some already applied in practice using di�erent working models for the design engine: parsimonious monotonic sigmoid Emax9, Normal Dynamic Linear Model without monotonicity restrictions10, several simple models combined in a Multiple Comparison Procedure (MCP Mod).11

CASE STUDY

�e objective of this trial was to establish the correct dose(s) to take forward into a con�rmatory trial on a novel anti-psychotic drug in acute schizophrenia. �ere was data indicating that the dose-response might be non-monotonic. �erefore, the working model chosen for the design was the Normal Dynamic Linear Model, because it is �exible and robust, and also allows for capturing non-monotonic relationships. Patients were allocated to placebo, active comparator and seven doses of study drug. �e adaptive allocation was targeting both the minimum e�cacious dose (MED) and the dose achieving the maximum response (Dmax) on the primary endpoint, positive PANSS score at week 4. Weekly data updates were used to change the allocation and check early stopping rules based on predictive probabilities. A stopping rule for futility was based on a prespeci�ed threshold that no treatment dose achieves the clinically signi�cant di�erence (CSD) over placebo. A stopping rule for e�cacy required su�cient knowledge about the MED and Dmax and a prespeci�ed con�dence that the Dmax achieves CSD. For details, see Shen et al.12 �e adaptive design allowed earlier (5 months) decision making with a 99% con�dence in futility decision with $14m savings in direct grant cost. �e design, implementation, and outcome of a response adaptive dose-ranging trial of an investigational drug in patients with diabetes, incorporating a

dose expansion approach to �exibly address bothe�cacy and safety has also been reported recently by Berry et al.13 �e early (2 months) futility decision was achieved with a $3.6m saving in direct cost and additional cost savings in avoiding expensive production of phase III drug supplies.

FIGURE 4: Summary of Early Phase Case Studies

SUMMARY

Early phase adaptive design trials address a number of critical development questions that need to be answered before commitment is made to phase III con�rmatory trials. �e application of these innovative designs early in the product development process enhances clinical trial e�ciency, productivity and increases the probability of success at phase III.

�e deployment of these approaches across a pipeline of products brings substantial bene�ts to sponsor companies and enables more e�ective portfolio management and ultimately increases portfolio value (Figure 5).

adaptive by nature, but more recent adaptive designsprovide better estimate of drug safety (e.g. the maxi-mum tolerated dose) and better understanding of itsPK/PD characteristics. Combining the objectives of conventional Phase IIa and IIb studies in a single trial provides the obvious bene�t of reducing the timeline by running the two studies seamlessly under a single protocol with the same clinical team. Furthermore, such approaches achieve trial e�ciency by the prospective integration of data from both IIa and IIb in the �nal analysis.

FIRST-IN-HUMAN SINGLE ASCENDING DOSE ESCALATION DESIGNS

In contrast to traditional dose escalation designs, where for example, cohorts of subjects are allocated to ascending doses (and placebo) until a maximum tolerated dose (MTD) is empirically determined, the continual reassessment method (CRM) models the dose- and/or exposure-toxicity relationship to focus on the identi�cation of the dose-range of interest near the MTD and for this to be explored further. �ese adaptive dose escalation designs can be expanded to also e�ciently establish proof-of-mechanism or proof-of-target modulation, if validated biomarker endpoints are available.

FIGURE 1: Types of Adaptive Designs in the Learn Phase

�e traditional First in Human study objectives of safety and tolerability can be extended to include an

assessment of proof-of-mechanism in the very �rst study conducted with the investigational compound. �e ability to assess proof-of-mechanism in this earlydevelopmental setting provides an opportunity to greatly enhance the clinical development strategy for the compound. �e adaptation is intended to quickly hone in on the predicted dose-range of interest. �is avoids the risks associated with exposing patients in an ine�ective low dose range, whilst minimizing exposure to high doses with an unacceptable tolerability and safety pro�le.

“CRM Offers a Better Approach to Accurately Determining the MTD in Early Phase Oncology Studies”

CRM models are becoming routine for early phase oncology studies where accurate de�nition of the MTD is critical for selecting the optimum dose for e�cacy assessment. Conventional 3+3 designs have been shown to signi�cantly underestimate the MTD which increases the risk of poor outcome in e�cacy trials.3

SEAMLESS MULTIPLE ASCENDING DOSE AND PROOF-OF-CONCEPT STUDY DESIGNS

�e objective of seamless multiple ascending dose and proof-of-concept studies is to combine a multiple ascending dose (MAD) study in patients with the proof-of-concept of the investigational compound.

CASE STUDY

A good example is seen in the treatment of rheumatoid arthritis. �e study design involved a MAD escalation study, in which patients were randomly assigned only to the lowest treatment group or placebo in a 3:1 ratio (active treatment or placebo) which then converted into a parallel enrolling, proof of concept (POC) dose ranging study (see Figure 2). Consistent with the MAD paradigm, dose escalation (allowing subjects in the next higher dose level to

receive a second dose) occurred only after the safety review on 6 subjects assigned to active treatment in the preceding dose had been performed. �is safetyreview was undertaken by a data monitoring committee. Only after it was determined by the data monitoring committee that two consecutive doses of the experimental treatment were well tolerated at a given dose level, were subjects already enrolled in the next higher dose treatment group permitted to receive a second dose of the investigational product. In Stage 2, a utility combining biomarker observations with early readout of a clinical endpoint at week 4 was used to drop dose levels that did not ful�l prespeci�ed necessary conditions (Proof-of-Non-Viability). Viable cohorts and comparator were kept open until each of the surviving treatment arms reached a prede�ned number of patients. �e endpoint to assess proof-of-concept was a 12 week observation on the regulatory accepted endpoint for Rheumatoid Arthritis (ACR20).

�is adaptive design provides several advantages over more traditional clinical trial approaches, without compromising patient safety:

• The seamless design increased the utility of the• information obtained by allowing subjects in both• the MAD and POC stages to provide both • de�nitive safety and e�cacy data.

• Performing the MAD component in subjects with• rheumatoid arthritis receiving concomitant • methotrexate allowed for the earliest • characterization of safety and antigenicity of the• new compound in the clinically relevant • population.

• The adaptive design minimized the number of• subjects that were exposed to ine�ective doses of the• drug, while simultaneously focusing subjects to• doses that were most informative for accurate dose• selection for subsequent con�rmatory trials.

FIGURE 2: Trial Design for Seamless MAD/POC Case Study �erefore, the adaptive design optimizes the bene�t/risk balance for participating subjects via improved e�ciency of decision making in relation to the doses of the new drug studied. In this example the estimated cost saving through combining two trials into one was $1.2m and the development programme was accelerated by 9 months.

“The Development Programme was Accelerated by 9 Months and the Cost Saving of Combining Two Trials in One was Estimated at $1.2M”

PROOF OF CONCEPT AND DOSE-RANGING STUDY DESIGNS

An example of a proof of concept and dose-ranging study design is a dose-�nding study to evaluate the analgesic e�cacy and safety of a single dose of a new non-steroidal anti-in�ammatory drug in subjects experiencing pain after oral dental surgery.4 Six doses of the new drug, placebo and an active comparator are considered. �e adaptive design consists of three stages. At the �rst stage, subjects are equally randomized into three arms: placebo, active comparator, and the top dose of the new drug. At the �rst interim analysis, the study may be stopped either because of lack of assay sensitivity, i.e. inability to establish a prede�ned minimal bene�t of the comparator over and above placebo, or evidence that

the top dose of study drug could not achieve a minimal acceptable level of total pain relief at 8 hours. If the study continues to the second stage, additional subjects would be randomized to placebo, active control, and three doses of the new drug (low, medium, and high) in a 1:1:2:2:1 ratio. Again, the study could be stopped due to lack of assay sensitivity or futility. Otherwise, a D-optimal design will be used to determine both the doses of the new drug (among six possible) and subject allocation ratios for the third stage, see Figure 3. At the end of the study, a four parameter logistic model is used to �t the dose response and make recommendations about the dose to be taken in phase III.

FIGURE 3: Trial Design for POC and Adaptive Dose-Ranging Case Study

CASE STUDY

A full case study has been published in the pain �eld where a Bayesian adaptive dose ranging design was used to evaluate the e�cacy of a new analgesic drug in a proof-of-concept post-herpetic neuralgia trial.5 In total 7 doses of the drug were evaluated in an attempt to de�ne the e�ective dose response curve. �e maximum sample size was 280 patients, however, at the �rst interim analysis (80 patients) the study was stopped for futility as none of the doses showed any meaningful di�erence from placebo. �e study was formally stopped after 133 patients had been enrolled. Stopping of this trial saved an estimated direct grant cost of $0.76m with a further $1.3m savings in internal costs.

“Early Stopping of POC Trials Because of Lack of Efficacy Saves Significant Cost and Enables Valuable Resources to be Re-assigned to Alternative Programmes That May Have a Better Chance of Success”

A recent unpublished example shows that early stopping of a POC adaptive dose ranging trial in post-operative nausea and vomiting saved $0.5m in outsourced costs. In this example, 131 out of 200 patients had been enrolled when the decision was taken at the �rst interim analysis.

RESPONSE ADAPTIVE DOSE-RANGING STUDY DESIGNS

Recent simulation studies6,7 conducted by the PhRMA Working Group on Adaptive Dose-Ranging Studies have found that response-adaptive dose allocation designs are generally superior in performance to conventional pairwise comparisons approaches. Learning about the research question such as identifying a target dose, or estimating the dose-response, is optimized with these designs.

�e main objectives in an adaptive dose-ranging study are to detect the dose response, to determine if any doses(s) meets clinical relevance, to estimate the dose-response, and then to decide on the dose(s) (if any) to take into the con�rmatory Phase III. With adaptive designs for dose-ranging studies many more doses may be considered than in a conventional parallel group design without increasing sample size. �is is achieved mainly through an adaptive allocation rule that is changing the randomization ratio to di�erent treatment arms during the study and putting more subjects in the region that allow most accurate estimation of dose-response and better precision in selecting the target dose. For example, Orlo� et al.8 showed that “for half the sample size, the adaptive design is as powerful and e�cient as the standard approach”.

Page 12: DESIGN AND EXECUTION OF EARLY PHASE …...Group on Adaptive Designs in Clinical Drug Development. J Biopharmaceutical Statistics 2006; 16:275-283. 3. Matano A, Bayesian Adaptive Designs

DESIGN AND EXECUTION OF EARLY PHASE ADAPTIVE CLINICAL TRIALSMaximising R&D Efficiency and Productivity

An Aptiv Solutions White Paper