Clinical Trials

Hanyan Yang yhy@zzu.edu.cn

2008.12

Treatment Once the nature of a patient’s illness has

been established and its expected, the next question is: what can be done about it? Is there a treatment that improves the outcome of disease?

This chapter describes ways of deciding whether a well-intentioned treatment dose in fact do more good than harms.

Studies of treatment effects

ideas about what might be useful treatment arise from virtually any activity within medicine.

Usually, the effects of treatments are considerably less dramatic. It is then nece- ssary to put ideas about treatment to a formal test, because a variety of conditions----coin- cidence, faulty comparisons, spontaneous changes in the course of disease, wishful thinking, etc----can obscure the true relation- ship between treatment and effect.

Clinical Trials

Experimental versus Observational Study Designs

Experimental Clinical Trials Community Trials

Observational Cohort Studies Case-Control Studies Cross-sectional Studies Ecologic Studies

Clinical Trial: Definition

A clinical trial is a prospective study comparing the effect and value of intervention(s) against a control in human beings.

A special kind of cohort study in which the conditions of study ---selection of treatment groups, nature of intervention, management during follow-up,etc.---are specified by the investigator for the purpose of making unbiased comparison.

Clinical Trials: Objective

To asses the efficacy and effectiveness of a new intervention or drug

To help establish the role of the new drug or intervention in clinical practice.

General Design of Clinical Trial

PAR S RStudy Group

Intervention

No Intervention

Outcome

NoOutcome

PAR = Population at RiskS = Sampling designR = Randomize interventionT = Elapsed time

T

Key Principles of a Clinical Trial

Randomization Controls Blinding

Key Elements of a Clinical Trial

Selection of subjects Allocation of exposure Blinding Data collection Statistical issues

Clinical Trials: Selection of Subjects

Population at risk, or target population, must be defined The intervention is intended to benefit this

population Criteria must be explicitly stated that capture

the population at risk Inclusion criteria – identifies the target

population Exclusion criteria – excludes people from

target population for safety reasons

Clinical Trials:Allocation of Treatment

Studies without controls - not a clinical trial Comparison groups

Historical controls Often advocated, but usually flawed Cannot exclude cohort effects Usually concludes that new treatment is better

than old Simultaneous non-randomized controls

Does not address bias Reduces chances of cohort effects

Randomization - true controls

Historical Controls

Comparison of current experience with the past

Problems Data collection differences introduce potential

bias Temporal effects not related to the intervention

or therapy

Clinical Trials:Simultaneous Non-Randomized Controls

A concurrent group of subjects is selected to act as a comparison group but are not offered the intervention or therapy

Problems The concurrent groups may be fundamentally

different for unrecognized reasons Systematic allocation may be abused

Clinical Trials: Randomization

Randomization will, on average, balance the known and unknown risk factors for the outcome under study.

Clinical Trials: Randomization

Allocation of treatment is randomized Every participant has an equal chance to

receive either the intervention or the control Stratified randomization may be used to

assign treatment within pre-defined strata Methods for randomization

Random number table Computer-generated list of assignments

Randomization often done at a coordinating center

Clinical Trials: Data Collection

Treatment What treatment did the subject actually take?

Assigned versus received Links directly with intent-to-treat analysis

Outcome Assessment of outcome must be comparable

between groups Avoid detection bias

Prognostic Profile at Entry Groups must have similar risk for outcome at

the start of the study

Clinical Trials: Blinding

Blinding refers to any attempt to make the various participants in a study unware of which treatment patients have been offered, so that the knowledge cannot cause them to act differently and thereby damage the internal validity of the study.

Clinical Trials: Blinding

Masking individuals participating in the trial from the treatment assignment

Levels of blinding Individual Data Collectors Investigators Analyst Data Safety Monitoring Board

Clinical Trials: Placebo

Placebo is an inert compound randomly allocated to subjects in a clinical trial.

Placebo arm is a true control for an intervention. Assess relative effect of intervention - relative

risk Assess risk for adverse events

Placebo arms are not ethical if there is an established intervention

Clinical Trials: Statistical Issues

Sample Size Determine for primary study outcome Must account for multiple comparisons Often requires statistical consultation

Monitoring of events Interim analyses are performed to determine

whether to stop the study early Sample size must be adjusted for the number of

interim analyses Data Safety Monitoring Board review

Analysis

Clinical Trials: Analysis

Fundamental measure of association Relative Risk

Efficacy

Efficacy = Incidenceplacebo

Incidencetreatment

Incidenceplacebo

-

Efficacy = 1 - Incidencetreatment

Incidenceplacebo

= 1 - Relative Risk

Summary

Promising ideas about what might be good treatment should be put to a rigorous test before being accepted.

The best test is a randomized controlled trial, a special case of a cohort study in which the intervention is allocated without bias.

Patients in clinical trials are usually highly selected, reducing generalizability. They are randomly allocated to receive either an experimental intervention or some comparison management: usual treatment, a placebo, or simple observation.

On the average, the compared groups have a similar prognosis just after randomization (and before the interventions), but differences not attributable to treatment can arise later, including dropouts and crossovers, co-interventions, and non-compliance.

Blinding all participants in the trial can help to minimize bias.