ARTICLE OPEN ACCESS CLASS OF EVIDENCE

Risk of natalizumab-associated PML in patientswith MS is reduced with extended interval dosingLana Zhovtis Ryerson MD John Foley MD Ih Chang PhD Ilya Kister MD Gary Cutter PhD

Ryan R Metzger PhD Judith D Goldberg ScD Xiaochun Li PhD Evan Riddle PhD Karen Smirnakis MPH

Rachna Kasliwal MPH Zheng Ren PhD Christophe Hotermans MD PhD Pei-Ran Ho MD and

Nolan Campbell PhD

Neurologyreg 201993e1452-e1462 doi101212WNL0000000000008243

Correspondence

Dr Campbell

nolancampbellbiogencom

AbstractObjectiveTo use the large dataset from the Tysabri Outreach Unified Commitment to Health(TOUCH) program to compare progressive multifocal leukoencephalopathy (PML) risk withnatalizumab extended interval dosing (EID) vs standard interval dosing (SID) in patients withmultiple sclerosis (MS)

MethodsThis retrospective cohort study included anti-JC virus antibody-positive patients (n = 35521)in the TOUCH database as of June 1 2017 The effect of EID on PML risk was evaluated with 3planned analyses using Kaplan-Meier methods stratified by prior immunosuppressant use Riskof PML was analyzed by Cox regression adjusted for age sex prior immunosuppressants timesince natalizumab initiation and cumulative number of infusions

ResultsThis study included 35521 patients (primary analysis 1988 EID 13132 SID secondaryanalysis 3331 EID 15424 SID tertiary analysis 815 EID 23168 SID) Mean average dosingintervals were 350 to 430 and 298 to 305 days for the EID and SID cohorts respectivelyHazard ratios (95 confidence intervals) of PML risk for EID vs SID were 006 (001ndash022 p lt0001) and 012 (005ndash029 p lt 0001) for the primary and secondary analyses respectivelyRelative risk reductions were 94 and 88 in favor of EID for the primary and secondaryanalyses respectively The tertiary analysis included no cases of PML with EID

ConclusionNatalizumab EID is associated with clinically and statistically significantly lower PML risk thanSID

Classification of evidenceThis study provides Class III evidence that for patients with MS natalizumab EID is associatedwith a lower PML risk than SID

RELATED ARTICLE

EditorialDosing interval ofnatalizumab in MS Dogood things come to thosewho wait

Page 655

MORE ONLINE

Class of EvidenceCriteria for ratingtherapeutic and diagnosticstudies

NPuborgcoe

These authors contributed equally to this work

From the Department of Neurology (LZR IK) NYU Langone Health New York University New York Rocky Mountain MS Clinic (JF RRM) Salt Lake City UT Biogen (IC ER KSRK ZR CH P-RH NC) Cambridge MA University of Alabama School of Public Health (GC) Birmingham and Division of Biostatistics (JDG XL) New York University School ofMedicine New York

Go to NeurologyorgN for full disclosures Funding information and disclosures deemed relevant by the authors if any are provided at the end of the article

The Article Processing Charge was funded by Biogen

This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License 40 (CC BY-NC-ND) which permits downloadingand sharing the work provided it is properly cited The work cannot be changed in any way or used commercially without permission from the journal

e1452 Copyright copy 2019 The Author(s) Published by Wolters Kluwer Health Inc on behalf of the American Academy of Neurology

Natalizumab a monoclonal antibody directed against theα4-integrin cell adhesion molecule is an efficacious treat-ment for relapsing forms of multiple sclerosis (MS) asdemonstrated by randomized clinical trials12 and real-worlddata34 The recommended treatment schedule (300 mg IVinfusion every 4 weeks) was selected to provide gt80 sat-uration of mononuclear cell α4β1-integrin receptors for asymp1month after administration56 For patients previouslyexposed to JC virus (JCV) natalizumab treatment isassociated with a risk of progressive multifocal leukoence-phalopathy (PML)7 Established risk factors for PML inanti-JCV antibody-positive patients include the level of anti-JCV antibodies in serum as assessed by anti-JCV antibodyindex the use of immunosuppressant therapy before nata-lizumab initiation and the duration of natalizumabtreatment89

In real-world practice treatment cessation treatmentinterruptions and deviations from recommended treatmentschedules are not unusual Several retrospective studies haveinvestigated the effect of extended interval dosing (EID)schedules (infusion intervals gt4 weeks) with the goal ofmaintaining natalizumab efficacy while reducing the risk ofPML1011 These studies which are limited by non-randomized designs small patient populations and variabledefinitions of EID nevertheless suggest that patientsswitching to natalizumab EID after a period of standard in-terval dosing (SID) continue to do well However becausePML is a rare event these studies did not have sufficientstatistical power to assess whether EID is associated with riskreduction of PML relative to SID Therefore the safety ofnatalizumab EID with respect to PML risk is not fullyknown

The Tysabri Outreach Unified Commitment to Health(TOUCHreg) program a risk evaluation and mitigationstrategy mandated by the US Food and DrugAdministration712 is designed to inform health care pro-viders and patients about PML and its known risk factors towarn against concurrent use of antineoplastic immunosup-pressant or immunomodulatory agents and to monitorpatients for the development of PML and other serious op-portunistic infections during treatment The TOUCH data-base captures all natalizumab infusion records patientdemographic information prior immunosuppressant ther-apy and anti-JCV antibody status data (since February 2012)It is the largest dataset in the world that can provide safetyinformation associated with alternative dosing intervals ofnatalizumab

MethodsStudy designThis retrospective cohort study included data collected in theTOUCH program as of June 1 2017 and included all patientswith a known positive anti-JCV antibody serostatus anda known status of prior immunosuppressant use PML data upto June 1 2017 from Biogenrsquos Tysabri Global Safety Databasewere also included in the study Patients with a history of anyinterval gt12 weeks (dosing gap) or lt3 weeks (overdose)between 2 consecutive infusions were excluded The 3 plan-ned analyses and their respective EID and SID inclusion cri-teria were developed and finalized under conditions blindedto PML events

Primary research questionThe objective of this study was to use the large real-worldTOUCH dataset to determine whether natalizumab EID wasassociated with a reduced PML risk compared with SIDBecause there is no precise understanding of the mechanismwhereby natalizumab causes PML or how dosing schedulesmight affect PML risk 3 planned analyses each with differentEID inclusion criteria were used to evaluate both the effect ofEID on PML risk and its potential mechanism

Classification of evidenceThis study provides Class III evidence that in patients withrelapsing-remitting MS natalizumab EID is associated withstatistically and clinically significant reductions in PML riskcompared with SID The evidence from this study is ratedClass III for the following reasons retrospective cohortstudy design imbalances between EID and SID groupsat baseline and limited available data on potentialconfounders

Data collectionPatient data collected in TOUCH include demographic in-formation the date and dose of each natalizumab infusion thedate and results of anti-JCV antibody testing (since 2012)performed in the previous 12 months and treatment withimmunomodulatoryimmunosuppressant therapies in theprevious 6 months The records of cases of PML are capturedand maintained in a separate pharmacovigilance database(Tysabri Global Safety Database)

Planned analyses and inclusion criteriaThe TOUCH dataset demonstrates considerable variability innatalizumab dosing whether intentional or unintentional inUS clinical practice Furthermore optimal EID infusion inter-vals and treatment duration are unknown Therefore 3 distinct

GlossaryADI = average dosing interval CI = confidence interval EID = extended interval dosing HR = hazard ratio JCV = JC virusMS = multiple sclerosis PML = progressive multifocal leukoencephalopathy SID = standard interval dosing TOUCH =Tysabri Outreach Unified Commitment to Health

NeurologyorgN Neurology | Volume 93 Number 15 | October 8 2019 e1453

analyses of EID vs SID were planned for this study Eachanalysis used different inclusion criteria (definitions) forpatients on EID and SID based on the number of dosesreceived during specified time periods to test different hy-potheses about the potential effect of EID on PML risk(figure 1) Patients could meet inclusion criteria for gt1analysis

The primary analysis assessed PML risk associated with thelast 18 months of recorded infusion history Patients who hadreceived le15 infusions in the last 18months of treatment wereincluded in the primary EID (EID-1deg) analysis group patientswho had received gt15 infusions in the last 18 months oftreatment were included in the primary SID (SID-1deg) analysisgroup

The secondary analysis assessed the effect of any prolongedperiod of EID in the patientrsquos infusion history on PML riskFor this analysis individual infusions were categorized asEID or SID An EID infusion was defined as any infusionpreceded by le10 infusions in the prior 365 days Patientsreceiving such EID infusions consecutively for ge6 monthswere included in the secondary EID (EID-2deg) analysis groupSimilarly an SID infusion was defined as any infusion

preceded by gt10 infusions in the prior 365 days and patientsreceiving such infusions consecutively for ge6 months wereincluded in the secondary SID (SID-2deg) analysis groupPatients with a history of both ge6 months of EID-2deg dosingand ge6 months of SID-2deg dosing were included in the EID-2degcohort only Patients with gt1 EID-2deg regimen were excludedincreasing the analytical rigor

The tertiary analysis assessed the effect of a dosing historyconsisting primarily of EID on PML risk Patients who hadreceived le10 infusions per year (annualized number of infu-sions) over their entire treatment history were included in thetertiary EID (EID-3deg) analysis group patients who had re-ceived gt10 infusions per year were included in the tertiary SID(SID-3deg) analysis group

Two prespecified sensitivity analyses were performed In thefirst cases of PML occurring before 2012 (before collectionof anti-JCV antibody test results in TOUCH) were assumedto be anti-JCV antibody positive and added to the 3 plannedanalyses described above In the second sensitivity analysisalternative EID definitions of le13 infusions in the last 18months and le9 infusions over any 12-month period wereused for inclusion in the primary and secondary EID analysis

Figure 1 Descriptions of the 3 planned analyses of PML risk and the definitions of EID and SID used in this study

Each hypothetical patient dosing diagram depicts 2 years of infusion history EID = extended interval dosing PML = progressive multifocal leukoencephal-opathy SID = standard interval dosing

e1454 Neurology | Volume 93 Number 15 | October 8 2019 NeurologyorgN

groups respectively Alternative inclusion criteria for thetertiary analysis were not tested

All analyses were performed on deidentified data collected inthe TOUCH program with patient consent and on PML datacollected via standard pharmacovigilance practices to monitornatalizumab safety as required by regulatory authorities Ad-ditional informed consent was not required

Statistical analysisDemographic and treatment history data for the overall studypopulation and for each EID analysis cohort were summarizedby descriptive statistics For the 3 planned analyses time-to-event (PML occurrence) analyses using Kaplan-Meier esti-mates of cumulative risk were performed for the EID and SIDcohorts Time to event was based on time since initiation ofnatalizumab treatment A log-rank test was performed tocompare the time to event between the EID and SID cohortsThe conditional probability of PML in each exposure epoch(defined as a series of 12 infusions) was derived for the EIDand SID cohorts with the life-table method stratified by priorimmunosuppressant use The PML hazard ratio (HR) in theEID and SID cohorts was estimated with a time-varyingcovariate Cox regression model adjusted for age sex calendaryear of the start of natalizumab treatment and prior immu-nosuppressant use (yesno) as covariates and the cumulativenumber of infusions as the time-varying covariate

For each analysis the PML HR estimate (EID vs SID) and its95 confidence interval (CI) from the Cox model were theprimary basis of inference Specifically if the HR upper 95CI limit was lt1 the EID cohort would be considered to havea lower risk of PML than the SID cohort If the HR pointestimate was ge09 and le11 the EID and SID cohorts wouldbe considered to have similar risks If the HR lower 95 CIlimit was gt1 the EID cohort would be considered to havegreater risk At the time of analysis plan specification theanticipated study population sizes and expected number ofPML events predicted asymp85 power to detect a risk reductionge50 (ie an HR le 05) as defined by the above rules ofinference

The statistical analysis plan was developed and finalized underconditions blinded to PML events PML data from theTysabri Global Safety Database were merged with TOUCHafter the analysis plan was finalized

Data availabilityData from the TOUCH program and pharmacovigilancedatabases used in the analyses described in this article are thesole property of Biogen and are not publicly available Theauthors and Biogen are fully supportive of allowing in-dependent assessment and verification of these resultsBiogen has established processes to share protocols clinicalstudy reports study-level data and patient-level data withqualified scientific researchers (supplementary file 1 linkslwwcomWNLA974)

ResultsPatientsOf the 90038 patients enrolled in TOUCH as of June 1 201735521 were anti-JCV antibody positive and eligible for thisstudy (figure 2) After application of the prespecified EID andSID inclusion criteria the study populations included 1988patients on EID and 13132 on SID in the primary analysis3331 patients on EID and 15424 on SID in the secondaryanalysis and 815 patients on EID and 23168 on SID in thetertiary analysis The most common reasons for patient ex-clusion were the presence of dosing gaps or overdoses intreatment history (primary secondary and tertiary analyses)and lt18 months of available dosing data (primary analysisonly)

The baseline demographics in the EID and SID groups werewell balanced across the 3 analyses (table 1) In all 3 analysespatients on EID had more natalizumab infusions and longertotal duration of natalizumab treatment than patients onSID Patients on EID included in the primary analysis hadreceived a median (range) of 37 (1ndash117) infusions beforestarting EID In the secondary analysis (in which each in-fusion was defined as either EID or SID) patients in theEID-2deg group had received a median (range) of 25 (1ndash121)infusions before starting EID For all 3 analyses the medianaverage dosing interval (ADI) over the entire treatmentduration was 335 to 414 days for patients on EID and 294to 298 days for patients on SID In the primary analysisduring the cohort-defining final 18 months of treatmentmedian ADI (interquartile range) was 422 (390ndash491) forpatients on EID patients and 297 (286ndash314) for thoseon SID

Risk assessmentThe Kaplan-Meierndashestimated cumulative risk of PML wassignificantly lower with EID than with SID (figure 3 AndashC) Inthe primary and secondary analyses cumulative risk appearedto separate after 24 to 36 months with separation increasing atlater time points Cox regression analysis also identified sig-nificant reductions in PML risk with EID treatment in theprimary and secondary analyses (both p lt 0001 table 2) Thecovariate-adjustedHR in the primary analysis was 006 (95CI001ndash022) corresponding to a relative risk reduction of 94 inpatients in the EID-1deg group vs those in the SID-1deg group Inthe secondary analysis the covariate-adjusted HR was 012(95 CI 005ndash029) corresponding to a relative risk reductionof 88 in patients in the EID-2deg group vs patients in the SID-2deggroup Because no cases of PMLwere observed with EID in thetertiary analysis the risk-reduction point estimate was 100and the Cox regression model 95 CI was nonestimable

Prior immunosuppressant use significantly increased PML riskCovariate-adjusted HRs were 292 (95 CI 167ndash511p lt 0001) in the primary analysis and 290 (95 CI 160ndash527p = 0001) in the secondary analysis (table 2) However thesignificance of this observation is limited by the small number of

NeurologyorgN Neurology | Volume 93 Number 15 | October 8 2019 e1455

patients with immunosuppressant use (95 for EID-1deg and 175for EID-2deg)

Sensitivity and post hoc analysesThe robustness of the 3 analyses was evaluated to determinethe effect of study design decisions on the results The firstsensitivity analysis examined the effect of excluding patientswithout known anti-JCV antibody status by including cases ofPML that occurred before 2012 under the assumption that allwere anti-JCV antibody positive This added 1 case of EID and67 cases of SID PML to the primary analysis 5 cases of EID and65 cases of SID PML to the secondary analysis and 0 cases ofEID and 71 cases of SID PML to the tertiary analysisWhen thesame post-2012 population denominators as the plannedanalyses were used (because anti-JCV antibody status is mostlyunknown for the pre-2012 population) HRs for EID vs SIDranged from lt001 to 009 in all 3 analyses (table 3)

The second sensitivity analysis investigated the effect of thenumber of EID doses required for inclusion in EID groups by

using alternative eligibility criteria The risk of PML was sig-nificantly lower for EID than for SID with the use ofthe alternative EID inclusion criteria of le13 infusions in theprevious 18 months (HR 010 95 CI 002ndash045) in theprimary analysis or le9 infusions over 12 months (HR 00195 CI lt 001ndash009) in the secondary analysis (table 3)Alternative EID inclusion criteria in the tertiary analysis werenot explored because no cases of EID-3deg PML were observed

Two post hoc analyses were carried out to address the in-fluence of potential selection biases on the composition of theEID analysis cohorts When the effect of excluding patientswith dosing gaps (intervals gt12 weeks between 2 infusions)was assessed by including patients with dosing gaps in the 3planned analyses of PML risk the resulting HRs ranged from008 to 016 (table 3)

Although all patients included in this study had tested positivefor anti-JCV antibodies at least once a second post hocanalysis was conducted to evaluate whether the duration of

Figure 2 Patient flow diagram for primary secondary and tertiary PML risk analyses

For inclusion in any analysis patientsmust have had nodosing gaps (defined as an interval gt12weeks between 2 consecutive infusions) or overdoses (definedas an interval lt3 weeks between 2 consecutive infusions) EID = extended interval dosing JCV = JC virus PML = progressive multifocal leukoencephalopathySID = standard interval dosing TOUCH = Tysabri Outreach Unified Commitment to Health aEnrolled number as of June 1 2017 bAt least 1 occurrence ofdosing gap (interval gt12 weeks between 2 consecutive infusions) or overdose (interval lt3 weeks between 2 consecutive infusions) cPatients switchedbetween SID and EID more than once

e1456 Neurology | Volume 93 Number 15 | October 8 2019 NeurologyorgN

anti-JCV antibody seropositivity affected risk estimatesLongitudinal anti-JCV antibody status (ie antibody statusconversion from negative to positive at some point in time) asa time-varying covariate was incorporated into the Cox re-gression model The resulting HR (95 CI) estimates were005 (011ndash018) in the primary analysis and 011 (004ndash026)in the secondary analysis (table 3) This sensitivity analysiswas not performed for the tertiary analysis

EID was associated with a reduction in the conditional risk ofPML in each successive epoch of natalizumab treatment for all3 definitions of EID and SID (table 4) Over the first 4treatment epochs (le48 infusions) only 1 case of PML (in thesecondary analysis) was observed in EID groups no caseswere observed in the primary and tertiary analyses In the fifthand sixth epochs (49ndash72 infusions) PML risk was sub-stantially lower for EID than for SID across all 3 analyses(table 4)

Cases of EID PMLThirteen cases of PML were identified among patientsmeeting primary and secondary EID inclusion criteria Onecase met the primary analysis criteria only 10 cases met thesecondary analysis criteria only and 2 cases met criteria forboth analyses There were no cases of PML in the tertiaryanalysis At the time of PML diagnosis 8 of 13 patients all ofwhom were included in the secondary analysis had switchedback to SID from EID and had been on SID for ge28 weeksimmediately before PML diagnosis (data not shown) On thebasis of 3 different measures patients with PML with a historyof EID had greater natalizumab exposure than their respectiveoverall EID cohorts The patients with EID PML had longer

(median [quartile 1 3]) natalizumab treatment durations(primary analysis 74 [58 75] vs 59 [37 87] months sec-ondary analysis 75 [60 85] vs 56 [36 81] months) more(median [quartile 1 3]) natalizumab infusions before startingan EID regimen (primary analysis 54 [37 55] vs 37 [18 63]infusions secondary analysis 405 [19 565] vs 25 [13 44]infusions) and more (median [quartile 1 3]) total natalizu-mab infusions on average (primary analysis 68 [50 68] vs 50[31 755] secondary analysis 68 [58 83] vs 51 [31 75]infusions) Prior immunosuppressant use was also morecommon in cases of EID PML than in the overall EID cohorts(primary analysis 33 vs 5 secondary analysis 17 vs 5)Of the 7 cases of PML for whom pre-PML anti-JCV antibodyindex values were available 6 had index values gt15 (data notshown)

DiscussionTo address the question of PML risk with EID we conducteda retrospective cohort study using patient data collected bythe TOUCH program This is the largest study of PML riskassociated with natalizumab EID to date and provides anexample of how real-world data derived from a Risk Evalua-tion and Mitigation Strategy program can be rigorously ana-lyzed to address a clinically meaningful question of riskreduction Even though PML is an uncommon event the sizeof the TOUCH dataset provided sufficient power to producerobust and statistically significant results We evaluated PMLrisk in anti-JCV antibody-positive patients who met any of the3 inclusion criteria for natalizumab EID vs the risk in patientson SID using 3 different prespecified analyses to investigate

Table 1 Baseline characteristics natalizumab exposure and ADIs

Characteristic

Primary analysis Secondary analysis Tertiary analysis

EID-1deg group(n = 1988)

SID-1deg group(n = 13132)

EID-2deg group(n = 3331)

SID-2deg group(n = 15424)

EID-3deg group(n = 815)

SID-3deg group(n = 23168)

Women n ()a 1376 (69) 8846 (67) 2293 (69) 10239 (66) 539 (66) 15636 (67)

Age at first infusion mean (SD) y 429 (113) 440 (110) 430 (112) 439 (114) 420 (114) 439 (116)

Prior immunosuppressanttherapy n ()b

95 (5) 689 (5) 175 (5) 799 (5) 49 (6) 1310 (6)

No of natalizumab infusionsmedian (minimum maximum)

50 (11 132) 46 (17 142) 51 (6 137) 27 (7 142) 32 (2 103) 26 (1 142)

Duration of natalizumabtreatment median (minimummaximum) mo

59 (19 130) 44 (19 131) 56 (8 131) 26 (7 130) 43 (3 129) 25 (1 131)

ADI d

Mean (SD) 367 (49) 300 (16) 350 (49) 298 (17) 430 (54) 305 (26)

Median (Q1 Q3) 355 (333 388) 297 (288 308) 335 (317 369) 294 (287 305) 414 (392 448) 298 (288 314)

Abbreviations ADI = average dosing interval (over entire treatment history) EID = extended interval dosing Q1 = first quartile Q3 = third quartile SID =standard interval dosingFigure 1 provides definitions of EID and SID in the primary secondary and tertiary analysesa Information on patient sex was missing for lt1 of patients in each groupb Information on prior immunosuppressant therapy was missing for 4 to 5 of patients in each group

NeurologyorgN Neurology | Volume 93 Number 15 | October 8 2019 e1457

Figure 3 Kaplan-Meier estimates of the cumulative probability of PML in EID vs SID groups in the (A) primary (B) secondaryand (C) tertiary analyses

CI = confidence interval EID = extended interval dosing HR = hazard ratio PML = progressivemultifocal leukoencephalopathy SID = standard interval dosingaEID vs SID Model includes age sex prior use of immunosuppressants EIDSID group and calendar year at the start of natalizumab treatment as covariatesThe Cox regression analysis could not be performed for the tertiary analysis because no cases of PML occurred in the EID-3deg group bNumber of patients whowere still in the study and did not have PML at the end of the specified time cCumulative number of cases of PML at the end of the specified time

e1458 Neurology | Volume 93 Number 15 | October 8 2019 NeurologyorgN

a wide range of dosing patterns used in real-world clinicalpractice For each of the 3 different analyses there wasa substantial reduction in PML risk with natalizumab EIDcompared with SID

The difference in dosing intervals between the overall EIDand SID groups was relatively modest (ADI 35ndash43 days forEID vs 30ndash31 days for SID) While these values combinedifferent treatment practices and dosing patterns the resultssuggest that extending dosing intervals by as little as 1 to 2weeks may produce a large reduction in PML risk It is un-likely that the main conclusions of this study were affected byoutliers at either end of the ADI range because patients withany dosing interval lt3 weeks or gt12 weeks in their historywere excluded from the planned analyses and third-quartile(75th percentile) ADI ranges for the EID cohort were 37 to45 days

Prespecified sensitivity analyses were performed to evaluatewhether and how the results were affected by the following 2study design elements (1) the inclusion only of patients withknown positive anti-JCV antibody status and (2) the numberof EID infusions required for inclusion in the EID-1deg or EID-2deg groups In addition a post hoc sensitivity analysis wasperformed to evaluate the effect of excluding patients witha history of dosing gaps (gt12 weeks between doses) Theresults of both the sensitivity and post hoc analyses werecomparable to those of the 3 planned analyses demonstratingthe robustness of the risk estimates and further strengtheningthe main conclusion that natalizumab EID is associated withlower PML risk than SID in at-risk patients

The possibility that physicians are more likely to switchpatients with longer durations of JCV seropositivity to EIDcreated a potential selection bias in the composition of theEID cohorts When anti-JCV antibody positivity status wasaccounted for as a time-varying covariate in a second post hoc

analysis the resulting HRs and 95 CIs were similar to thoseproduced in the original prespecified analyses indicating thatthis potential bias did not have a major bearing on the mainstudy conclusions

The identification of 13 cases of PML among the combined5249 patients in the EID-1deg andor EID-2deg groups indicatesthat while these EID regimens are associated with significantlylower PML risk than SID the risk is not completely elimi-nated No cases of PML were observed with the more strin-gent tertiary analysis although the EID-3deg group was relativelysmall (n = 815) Most of the cases of EID PML described herehad multiple risk factors for PML including longer overallnatalizumab treatment duration longer periods of SID beforeswitching to EID and a greater likelihood of prior immuno-suppressant use than patients in the corresponding overallEID cohorts In addition several of the cases of EID PML hadreturned to SID before PML diagnosis In a previously pub-lished case report of PML in a patient receiving natalizumabEID the affected patient also had elevated PML risk factorsincluding a prolonged period of SID preceding EID and ananti-JCV antibody index gt1513

The biological mechanisms underlying the observed PML riskreduction require additional research but partial reversal ofthe pharmacodynamic effects of natalizumab including de-creased receptor saturation increased soluble vascular celladhesion molecule expression and a reduced natalizumab-induced peripheral lymphocytosis has been reported to occur4 to 8 weeks after the last dose14 and may allow the rees-tablishment of some immune surveillance in the CNS Thepathogenesis of PML is complex and natalizumab may in-crease PML risk via mechanisms other than or in addition toreducing immune surveillance15 Therefore alternative hy-potheses for the observed risk reduction with EID that are notbased on an increase in immune surveillance should also beconsidered

Table 2 Effect of EID vs SID on PML risk in a Cox regression model in the primary and secondary analysesa

Risk factor

Primary analysis Secondary analysis

HR (95 CI) p Value HR (95 CI) p Value

Age 100 (098ndash102) 0999 099 (097ndash101) 0411

Sex (male female) 105 (058ndash163) 0828 099 (063ndash157) 0969

Prior immunosuppressant use (yes no) 292 (167ndash511) lt0001 290 (160ndash527) lt0001

Calendar year at the start of treatment 099 (088ndash112) 0881 094 (083ndash106) 0327

No of cumulative infusions 091 (087ndash095) lt0001 091 (087ndash094) lt0001

Dosing group (EID SID) 006 (001ndash022) lt0001 012 (005ndash029) lt0001

Abbreviations CI = confidence interval EID = extended interval dosing HR = hazard ratio PML = progressivemultifocal leukoencephalopathy SID = standardinterval dosingRefer to figure 1 for definitions of EID and SID in the primary secondary and tertiary analysesa Model includes age sex prior use of immunosuppressants EIDSID group and calendar year at the start of natalizumab treatment as covariates Modelingcould not be performed in the tertiary analysis because no PML events occurred in the tertiary analysis EID group

NeurologyorgN Neurology | Volume 93 Number 15 | October 8 2019 e1459

The conclusions of this study are limited by several inherentbiases Patients on EID received more doses of natalizumabthan patients on SID which could have introduced a se-lection bias favoring more cases of PML in the EID cohortsbecause natalizumab exposure is a known PML risk factorConversely most patients had received gt2 years of SIDtreatment without developing PML before starting EID sothe EID groups may have included patients with inherentlyreduced risk of PML thereby introducing a selection bias infavor of fewer cases of PML in the EID groups In additionanti-JCV antibody index data are not available for allpatients in TOUCH therefore we do not know whetherindex values differ between the EID and SID cohorts and

whether any such difference plays a role in the risk reduc-tions observed in this study Because EID is used as an off-label strategy in clinical practice to reduce PML risk anti-JCV antibody index values might be higher in patients onEID than in those on SID as was observed in a retrospectivestudy of natalizumab EID11 If this is true the risk reduc-tions seen in the EID cohorts described here could po-tentially be even larger in EID and SID patient populationswith the same distributions of anti-JCV antibody indexvalues Lastly we note that the Kaplan-Meier risk estimatesare subject to increasing levels of uncertainty over time dueto reductions in the cohort population sizes especiallybeyond 5 years of treatment However the EID and SID

Table 3 PML HR (95 CI) for EID vs SID in the sensitivity and post hoc selection bias analyses

Planned analysis

Sensitivity analysis inclusion ofcases of PML without knownanti-JCV antibodyndashpositivestatusa

Sensitivityanalysisalternative EIDinclusion criteriab

Post hoc analysis inclusion ofcases of PML without knownanti-JCV antibodyndashpositivestatus and patients with dosinggapsc

Post hoc analysisduration of anti-JCVantibodyndashpositivestatusd

Primary analysis EID inthe last 18 mo oftreatment

le15 Infusions in thelast 18 mo

EID-1deg n 1989 998 7029 1988

SID-1deg n 13199 14122 17185 13132

PML HR (95 CI) 005 (002ndash016) 010 (002ndash045) 010 (004ndash020) 005 (001ndash018)

Secondary analysis EIDlasting ge6 mo at anytime in treatmenthistory

le10 Infusions over 12mo

EID-2deg n 3336 1870 9593 3331

SID-2deg n 15489 17902 16282 15424

PML HR (95 CI) 009 (004ndash018) 001 (lt001ndash009) 016 (010ndash024) 011 (004ndash026)

Tertiary analysismajority of treatmentreceived as EID

le10 Infusions per yearover the duration ofinfusion history

EID-3deg n 815 NA 6307 NA

SID-3deg n 23239 NA 27336 NA

PML HR (95 CI) lt001e NA 008 (003ndash017) NA

Abbreviations CI = confidence interval EID = extended interval dosing HR = hazard ratio JCV = JC virus NA = not analyzed PML = progressive multifocalleukoencephalopathy SID = standard interval dosingFigure 1 provides definitions of EID and SID in the primary secondary and tertiary analysesa Cases of PML assumed to be anti-JCV antibody positive and occurring before 2012 were added to the analysis populations This added 1 case of EID and 67cases of SID in the primary analysis 5 cases of EID and 65 cases of SID in the secondary analysis and 0 cases of EID and 71 cases of SID in the tertiary analysisb Alternative EID definitions were le13 infusions in the last 18 months in the primary analysis and le9 infusions over 12 months in the secondary analysis Analternative definition in the tertiary analysis was not exploredc Patients with dosing gaps gt12 weeks between infusions were added to the pre-2012 PML case sensitivity analysis cohortsd Cox regression modeling of JCV status as a time-varying covariate EID vs SID This model was not tested in the tertiary analysise The 95 CI was not estimable because no cases of PML occurred in the EID-3deg group

e1460 Neurology | Volume 93 Number 15 | October 8 2019 NeurologyorgN

curves are well separated even at 60 months (5 years) ofnatalizumab treatment when there are still substantialnumbers of patients in all 3 analysis groups The TOUCHprogram is ongoing so future analyses of the TOUCHdataset will add more patients and observation time to eachof the 3 groups and should provide greater certainty for therisk estimates

Finally and most important because the TOUCH programdoes not collect information about therapeutic efficacy wecould not assess the benefit-risk profile of EID compared withSID Several studies of patient outcomes after natalizumabdiscontinuation indicate that MS disease activity is suppressedfor at least 6 weeks and possibly as long as 12 weeks after thelast administration16ndash19 Furthermore 2 retrospective studieshave suggested that natalizumab efficacy is not compromisedby EID regimens1011 However the findings from thesestudies are limited by nonrandomized designs small studypopulations variable dosing practices and potential selectionbiases in the EID study populations In contrast to the clinicalresults model-based simulations of natalizumab exposurehave suggested that EID regimens (with 6- to 8-week inter-vals) may not confer adequate protection against MS diseaseactivity20

The information provided in this study is highly relevant tohealth care providers and patients considering initiating orcontinuing natalizumab therapy However the benefit-riskprofile of natalizumab EID has not been fully defined and it ispremature to suggest that SID should be replaced by EID orthat the established PML risk stratification measures shouldbe revised921 An ongoing randomized prospective trial ofEID vs SID (ClinicalTrialsgov identifier NCT03689972) willyield a more comprehensive understanding of both the ef-fectiveness and safety of natalizumab EID

Author contributionsLana Zhovtis Ryerson draftingrevising the manuscript studyconcept or design analysis or interpretation of data acceptsresponsibility for conduct of research and will give final ap-proval study supervision John Foley draftingrevising themanuscript study concept or design analysis or interpretationof data accepts responsibility for conduct of research and willgive final approval Ih Chang study concept or design analysisor interpretation of data accepts responsibility for conduct ofresearch and will give final approval statistical analysis IlyaKister draftingrevising the manuscript study concept or de-sign analysis or interpretation of data accepts responsibility forconduct of research and will give final approval Gary Cutterdraftingrevising the manuscript analysis or interpretation ofdata accepts responsibility for conduct of research and will givefinal approval Ryan R Metzger draftingrevising the manu-script study concept or design accepts responsibility for con-duct of research and will give final approval Judith DGoldberg draftingrevising the manuscript analysis or in-terpretation of data accepts responsibility for conduct of re-search and will give final approval statistical analysis XiaochunLi draftingrevising the manuscript accepts responsibility forconduct of research and will give final approval statisticalanalysis Evan Riddle draftingrevising the manuscript studyconcept or design analysis or interpretation of data acceptsresponsibility for conduct of research and will give final ap-proval Karen Smirnakis data acquisition analysis or in-terpretation of data accepts responsibility for conduct ofresearch and will give final approval Rachna Kasliwal draftingrevising the manuscript accepts responsibility for conduct ofresearch and will give final approval acquisition of data ZhengRen analysis or interpretation of data accepts responsibility forconduct of research and will give final approval statisticalanalysis Christophe Hotermans draftingrevising the manu-script study concept or design analysis or interpretation of

Table 4 Life-table estimates of PML risk in patients included in the primary secondary and tertiary analyses

Natalizumab exposureepocha

No ofinfusions

Estimated risk of PML per 1000 patients (No of cases per adjusted no of patients)

Primary analysis Secondary analysis Tertiary analysis

EID-1deggroup SID-1deg group EID-2deg group SID-2deg group

EID-3deggroup SID-3deg group

1 1ndash12 0 (01806) 0 (011890) 0 (02980) 0 (013049) 0 (0662) 0 (018364)

2 13ndash24 0 (01659) 028 (310907) 0 (02722) 060 (69921) 0 (0510) 052 (713425)

3 25ndash36 0 (01366) 046 (48608) 044 (12292) 046 (36514) 0 (0371) 042 (49603)

4 37ndash48 0 (01080) 202 (136439) 0 (01841) 258 (124650) 0 (0265) 179 (137254)

5 49ndash60 123 (1810) 396 (194801) 145 (21380) 414 (143385) 0 (0169) 367 (205443)

6 61ndash72 170 (1589) 446 (153363) 204 (2980) 474 (112323) 0 (0104) 416 (163848)

Abbreviations EID = extended interval dosing PML = progressive multifocal leukoencephalopathy SID = standard interval dosingPML risk is shown as the incidence rate per 1000 patients (number of cases of PML per adjusted number of patients at risk) in anti-JC virus antibody-positivepatients without prior immunosuppressant use for the primary and secondary definitions Patientswith prior immunosuppressant use could not be analyzeddue to the insufficient number of patients The adjusted number of patients at risk was 95 in the EID-1deg group 689 in the SID-1deg group 171 in the EID-2deg groupand 747 in the SID-2deg group PML risk could not be calculated in the tertiary analysis of EID because no cases of PML occurred in this analysis Figure 1 providesdefinitions of EID and SID under the primary secondary and tertiary analysesa Data beyond 6 years are not shown

NeurologyorgN Neurology | Volume 93 Number 15 | October 8 2019 e1461

data accepts responsibility for conduct of research and will givefinal approval Pei-Ran Ho and Nolan Campbell draftingrevising the manuscript study concept or design analysis orinterpretation of data accepts responsibility for conduct ofresearch and will give final approval study supervision

AcknowledgmentThe authors acknowledge the many clinicians who pioneeredthe investigation of EID of natalizumab including JosephHerbert MD who was instrumental in organizing furtherstudy of this approach All named authors meet theInternational Committee of Medical Journal Editors criteriafor authorship for this manuscript and take responsibility forthe integrity of the work as a whole Biogen provided fundingfor medical writing support in the development of thismanuscript John Watson PhD from Ashfield HealthcareCommunications (Middletown CT) with input from theauthors wrote the first draft and revised subsequent drafts ofthe manuscript and Joshua Safran from Ashfield HealthcareCommunications copyedited and styled the manuscript perjournal requirements Biogen reviewed and provided feedbackon the manuscript to the authors

Study FundingFunded by Biogen

DisclosureL Zhovtis Ryerson reports personal compensation forspeaker bureau activities from Biogen Genentech and Tevaand advisory board activities from Biogen and Celgene andresearch support from Biogen J Foley reports personalcompensation and compensation for consulting activitiesfrom Biogen Genentech Genzyme and Teva I Chang is anemployee of and may hold stockstock options in Biogen IKister has served on advisory boards for Biogen and Gen-entech G Cutter has served on data and safety monitoringboards for AMO Pharmaceuticals Apotek Horizon Phar-maceuticals Merck MerckPfizer ModigenetechProlorNeurim Opko Biologics Reata ReceptosCelgene SanofiTeva NHLBI (Protocol Review Committee) and the Na-tional Institute of Child Health and Human Development(Obstetric-Fetal Pharmacology Research Unit oversightcommittee) has received compensation for consulting oradvisory board service from Argenix Atara BiotherapeuticsBioeq GmBH the Consortium of MS Centers (grant)Genentech Genzyme Innate Therapeutics Klein-BuendelMedDay Medimmune Novartis Opexa TherapeuticsRoche Savara Somahlution Teva TG Therapeutics andTransparency Life Sciences and is president of Pythagorasa private consulting company R Metzgerrsquos spouse is

employed by Sanofi and holds stock in Biogen J Goldbergand X Li report no disclosures relevant to the manuscript ERiddle K Smirnakis R Kasliwal Z Ren C Hotermans P-RHo and N Campbell are employees of and may hold stockstock options in Biogen Go to NeurologyorgN for fulldisclosures

Publication historyReceived by Neurology October 25 2018 Accepted in final formJune 11 2019

References1 Miller DH Khan OA Sheremata WA et al A controlled trial of natalizumab for

relapsing multiple sclerosis N Engl J Med 200334815ndash232 Polman CHOrsquoConnor PW Havrdova E et al A randomized placebo-controlled trial

of natalizumab for relapsing multiple sclerosis N Engl J Med 2006354899ndash9103 Prosperini L Sacca F Cordioli C et al Real-world effectiveness of natalizumab and

fingolimod compared with self-injectable drugs in non-responders and in treatment-naive patients with multiple sclerosis J Neurol 2017264284ndash294

4 Butzkueven H Kappos L Pellegrini F et al Efficacy and safety of natalizumab inmultiple sclerosis interim observational programme results J Neurol NeurosurgPsychiatry 2014851190ndash1197

5 Rudick RA Sandrock A Natalizumab a4-integrin antagonist selective adhesionmolecule inhibitors for MS Expert Rev Neurother 20044571ndash580

6 Stuve O Bennett JL Pharmacological properties toxicology and scientific rationalefor the use of natalizumab (Tysabri) in inflammatory diseases CNS Drug Rev 20071379ndash95

7 Tysabri (natalizumab) [prescribing information] Cambridge Biogen 20188 Bloomgren G Richman S Hotermans C et al Risk of natalizumab-associated pro-

gressive multifocal leukoencephalopathy N Engl J Med 20123661870ndash18809 Ho PR Koendgen H Campbell N Haddock B Richman S Chang I Risk of

natalizumab-associated progressive multifocal leukoencephalopathy in patients withmultiple sclerosis a retrospective analysis of data from four clinical studies LancetNeurol 201716925ndash933

10 Bomprezzi R Pawate S Extended interval dosing of natalizumab a two-center 7-yearexperience Ther Adv Neurol Disord 20147227ndash231

11 Zhovtis Ryerson L Frohman TC Foley J et al Extended interval dosing of natali-zumab in multiple sclerosis J Neurol Neurosurg Psychiatry 201687885ndash889

12 Risk Evaluation and Mitigation Strategy (REMS) TYSABRI Outreach Unified Commit-ment toHealth (TOUCHreg) prescribing program [online] Available at fdagovdownloadsDrugsDrugSafetyPostmarketDrugSafetyInformationforPatientsandProvidersUCM288126pdf Accessed May 19 2018

13 Hervas-Garcıa JV Presas-Rodrıguez S Crespo-Cuevas AM et al Progressive multi-focal leukoencephalopathy associated to natalizumab extended dosing regimenNeurodegener Dis Manag 20155399ndash402

14 Plavina T Muralidharan KK Kuesters G et al Reversibility of the effects of natali-zumab on peripheral immune cell dynamics in MS patients Neurology 2017891584ndash1593

15 Major EO Yousry TA Clifford DB Pathogenesis of progressive multifocal leu-koencephalopathy and risks associated with treatments for multiple sclerosis a decadeof lessons learned Lancet Neurol 201817467ndash480

16 Fox RJ Cree BA De Seze J et al MS disease activity in RESTORE a randomized24-week natalizumab treatment interruption study Neurology 2014821491ndash1498

17 Kappos L Radue EW Comi G et al Switching from natalizumab to fingolimoda randomized placebo-controlled study in RRMS Neurology 20158529ndash39

18 Grimaldi LM Prosperini L Vitello G Borriello G Fubelli F Pozzilli C MRI-basedanalysis of the natalizumab therapeutic window in multiple sclerosis Mult Scler 2012181337ndash1339

19 Berkovich R Togasaki DM Cen SY Steinman L CD4 cell response to intervaltherapy with natalizumab Ann Clin Transl Neurol 20152570ndash574

20 Muralidharan KK Steiner D Amarante D et al Exposure-disease response analysis ofnatalizumab in subjects with multiple sclerosis J Pharmacokinet Pharmacodyn 201744263ndash275

21 Physician information and management guidelines for multiple sclerosis patients onTYSABRI therapy [online] Available at hpraiedocsdefault-source3rd-party-documentstysabri-physician-information-and-management-guidelines-(version-15-27th-april-2016)pdfsfvrsn=2 Accessed January 8 2019

e1462 Neurology | Volume 93 Number 15 | October 8 2019 NeurologyorgN

DOI 101212WNL0000000000008243201993e1452-e1462 Published Online before print September 12 2019Neurology

Lana Zhovtis Ryerson John Foley Ih Chang et al interval dosing

Risk of natalizumab-associated PML in patients with MS is reduced with extended

This information is current as of September 12 2019

ServicesUpdated Information amp

httpnneurologyorgcontent9315e1452fullincluding high resolution figures can be found at

References httpnneurologyorgcontent9315e1452fullref-list-1

This article cites 18 articles 5 of which you can access for free at

Citations httpnneurologyorgcontent9315e1452fullotherarticles

This article has been cited by 3 HighWire-hosted articles

Subspecialty Collections

httpnneurologyorgcgicollectionmultiple_sclerosisMultiple sclerosis

httpnneurologyorgcgicollectionincidence_studiesIncidence studies

httpnneurologyorgcgicollectionharm__risk_analysisHarm risk (analysis)

httpnneurologyorgcgicollectioncohort_studiesCohort studies

httpnneurologyorgcgicollectionclass_iiiClass IIIfollowing collection(s) This article along with others on similar topics appears in the

Permissions amp Licensing

httpwwwneurologyorgaboutabout_the_journalpermissionsits entirety can be found online atInformation about reproducing this article in parts (figurestables) or in

Reprints

httpnneurologyorgsubscribersadvertiseInformation about ordering reprints can be found online

ISSN 0028-3878 Online ISSN 1526-632XWolters Kluwer Health Inc on behalf of the American Academy of Neurology All rights reserved Print1951 it is now a weekly with 48 issues per year Copyright Copyright copy 2019 The Author(s) Published by

reg is the official journal of the American Academy of Neurology Published continuously sinceNeurology