ARTICLE OPEN ACCESS

Refining eligibility criteria for amyotrophic lateralsclerosis clinical trialsRuben PA van Eijk MD MSc Henk-Jan Westeneng MD Stavros Nikolakopoulos PhD Iris E Verhagen MD

Michael A van Es MD PhD Marinus JC Eijkemans PhD and Leonard H van den Berg MD PhD

Neurologyreg 201992e451-e460 doi101212WNL0000000000006855

Correspondence

Dr van den Berg

lbergumcutrechtnl

AbstractObjectiveTo assess the effect of eligibility criteria on exclusion rates generalizability and outcomeheterogeneity in amyotrophic lateral sclerosis (ALS) clinical trials and to assess the value ofa risk-based inclusion criterion

MethodsA literature search was performed to summarize the eligibility criteria of clinical trials Theextracted criteria were applied to an incidence cohort of 2904 consecutive patients with ALS toquantify their effects on generalizability and outcome heterogeneity We evaluated the effect ofa risk-based selection approach on trial design using a personalized survival prediction model

ResultsWe identified 38 trials A large variability exists between trials in all patient characteristics forenrolled patients (p lt 0001) except for the proportion of men (p = 021) Exclusion ratesvaried widely (from 14 to 95 mean 598 95 confidence interval 526ndash667)Stratification of the eligible populations into prognostic subgroups showed that eligibilitycriteria lead to exclusion of patients in all prognostic groups Eligibility criteria neither reduceheterogeneity in survival time (from 220 to 205 months p = 009) nor affect between-patientvariability in functional decline (from 062 to 065 p = 025) In none of the 38 trials were theeligibility criteria found to bemore efficient than the predictionmodel in optimizing sample sizeand eligibility rate

ConclusionsThe majority of patients with ALS are excluded from trial participation which questions thegeneralizability of trial results Eligibility criteria only minimally improve homogeneity in trialendpoints An individualized risk-based criterion could be used to balance the gains in trialdesign and loss in generalizability

RELATED ARTICLE

EditorialHeterogeneity urgencygeneralizability andenrollment The HUGEbalance in ALS trials

Page 215

From the Department of Neurology (RPAvE H-JW IEV MAvE LHvdB) Brain Center Rudolf Magnus and Biostatistics amp Research Support (RPAvE SN MJCE) JuliusCenter for Health Sciences and Primary Care University Medical Center Utrecht the Netherlands

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

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

The Article Processing Charge was funded by The Netherlands ALS Foundation

Copyright copy 2019 The Author(s) Published by Wolters Kluwer Health Inc on behalf of the American Academy of Neurology e451

The clinical heterogeneity of amyotrophic lateral sclerosis(ALS) makes conducting clinical trials complex1 To increasethe potential to demonstrate therapeutic efficacy inves-tigators apply eligibility criteria to enroll a more homogeneouspopulation to improve protocol adherence or to excludepatients who are unlikely to benefit2ndash4 For ALS this oftenmeans excluding those patients with long disease durations orthose who are unlikely to survive the follow-up period Thusmany patients are excluded2 which could have importantconsequences for the generalizability of a trial (ie ldquoto whomdo the results of this trial applyrdquo)4 A low generalizabilitypotentially provides limited information about safety or effi-cacy for the general population245

Nevertheless a pragmatic approach (ie no selection) maynot be feasible in ALS because this increases endpoint vari-ability and may inflate sample size6 It is therefore importantto balance the generalizability of a trial (ie selecting a pop-ulation that represents the general population) and its end-point heterogeneity (ie selecting a sensitive population toshow efficacy) Prediction models such as the recently vali-dated ALS survival model7 might improve the selection ofpatients by using the predicted outcome as an inclusion cri-terion instead of a set of arbitrary criteria8

Virtually all ALS clinical trials have imposed various sets ofeligibility criteria but little is known about the consequences29

In this study we review the current practices of participantselection and assess the effects on trial populations efficacyendpoints and generalizability We then gauge the value of anindividualized risk-based selection criterion in balancing out-come heterogeneity and patient exclusion rates

MethodsThe effect of eligibility criteria on population characteristicsand trial endpoints was estimated in a 2-step approach Firstwe systematically reviewed the literature to compose a list ofcommonly used criteria and meta-analyzed the baselinecharacteristics of the included trials Subsequently we appliedthe extracted criteria to an incidence cohort to estimate theireffects on survival and functional decline

Search strategy and trial selectionTwo authors (RPAvE and IEV) individually searched thePubMed and Embase database for publications dating fromJanuary 1 2000 up to and including November 2017 usingthe following search terms amyotrophic lateral sclerosis ormotor neuron disease and clinical trial To harmonize the

comparison between clinical trials we included only ran-domized placebo-controlled clinical trials evaluating the ef-ficacy of a single pharmacologic agent We excluded clinicaltrials investigating multiple agents exclusively aiming to de-termine safety (phase I) having a nonclinical primary end-point or starting enrollment before the approval of riluzole(1996)

Data extraction and harmonizationFor each trial we extracted the eligibility criteria and baselinecharacteristics If a trial included multiple dosing groupsgroups receiving the experimental agent were collapsed bya fixed-effects within-study meta-analysis Predicted vital ca-pacity was recorded as either the forced vital capacity (FVC)or slow vital capacity because there is no true difference be-tween the 2 values10 When the mean Amyotrophic LateralSclerosis Functional Rating Scale (ALSFRS) score wasreported instead of the revised ALSFRS (ALSFRS-R) weadded 776 to the mean and 023 to the reported SD becausethis procedure provided the least systematic error duringsimulations The ALSFRS-R slope was calculated as the av-erage monthly change from randomization Studies reportingthe ALSFRS slope were transformed to the ALSFRS-R slopeby multiplying the mean ALSFRS rate of decline by 11 and itsstandard error by 114 If data were not reported as mean withSD (eg median and range) we transformed the reportedestimates according to the method of Hozo et al11 Finally ifonly the mean was reported without SD (6 of the studies)we imputed the natural logarithm of the SD from the naturallogarithm of the mean using linear regression12

Exclusion rate and incidence cohort of patientswith ALSOur primary goal was to estimate per set of eligibility criteriathe number of patients from the general population whowould be considered ineligible to participate (ie exclusionrate) The exclusion rate per trial was defined as the number ofineligible patients divided by the total number of patients(population size) An exclusion rate of 60 for exampleindicates that only 40 of the general population was po-tentially evaluated in a given trial the effectiveness and safetyare unknown for 60 of the population Because eligibilitycriteria are applied in a step-wise manner the exclusion rate isa cumulative buildup of several percentages An example ofhow this rate is calculated is presented in table 1

The denominator is crucial for the accuracy of the exclusionrate We therefore approximated the population size by in-cluding all consecutive patients with ALS in the Netherlandsdiagnosed between January 2006 and December 2016 in the

GlossaryALS = amyotrophic lateral sclerosis ALSFRS = Amyotrophic Lateral Sclerosis Functional Rating Scale ALSFRS-R =Amyotrophic Lateral Sclerosis Functional Rating Scalendashrevised CI = confidence interval ENCALS = European Network forthe Cure of ALS FVC = forced vital capacity

e452 Neurology | Volume 92 Number 5 | January 29 2019 NeurologyorgN

analysis A detailed description of the incidence cohort ofpatients with ALS in the Netherlands which has a coveragerate of 81 has been given elsewhere13 In short patientswith ALS in the Netherlands diagnosed with possible prob-able (laboratory supported) or definite ALS according to therevised El Escorial criteria are registered centrally at TheNetherlands ALS Center14 Patients are located by annualscreening of large medical center registries and by individuallycontacting Dutch neurologists13 For each patient we de-termined whether he or she participated in a clinical trial inthe same time period In addition clinical data were collectedto determine an individualrsquos eligibility for trial participation atthe day of diagnosis (figure e-1 available from Dryad doiorg105061dryad86f1m6g) Time-varying variables such asALSFRS-R total score and FVC were ideally collected atdiagnosis or within a 3-month time interval otherwise thescore was recorded as missing Complete survival data (dateof death or last follow-up) were obtained by checking theonline municipal population register at 3-month intervals Intotal 39 surviving patients (13) were followed up for lt6months and 139 (47) for lt12 months

Standard protocol approvals registrationsand patient consentsThe medical ethics committee and institutional review boardof the UniversityMedical Center Utrecht approved this study

Statistical analysisVariables necessary to determine eligibility contained onaverage 36 missing values kidney function was most oftenmissing (in 82 of the cases) Eligibility was based on allvariables provided (figure e-1 available from Dryad doiorg105061dryad86f1m6g) Missing data were accounted for bycreating multiple imputed datasets (n = 100) using predictivemean matching and bootstrapping discarding the first 100iterations (burn-in) The imputation model contained allcovariates survival time was modeled with Nelson-Aalenestimates15 Subsequently the extracted eligibility criteriawere applied to each imputed dataset Using the incidence

cohort as reference we estimated the proportion of patientseligible for each trial individually Results across imputationswere pooled with Rubin rules16 Baseline characteristics andeligibility percentages were summarized by meta-analysisContinuous variables were meta-analyzed using the rawreported means count variables were recoded to logit-transformed proportions Subsequently between-populationheterogeneity was evaluated with the CochraneQ test and theI2 statistic17 All meta-analytical models were based on a ran-dom-effects model fitted with restricted maximum-likelihoodestimation Standard errors of coefficients were adjustedaccording to themethodology given by Knapp andHartung18

For each patient we determined his or her risk profile bytaking the exponent of the centered linear predictor of theEuropean Network for the Cure of ALS (ENCALS) survivalmodel7 An estimated risk score of 05 indicates that the riskof dying during follow-up is half the risk of dying for theaverage patient with ALS whereas a score of 2 means thatthis risk is twice the average Using the quintiles of the in-dividual risk scores we defined 5 prognostic groups verylong long intermediate short and very short To estimatethe underlying survival time distribution we fitted a para-metric Royston-Parmar proportional hazard models with 3internal knots (based on Akaike information criterion) perprognostic subgroup1920 The interquartile range of thesurvival times was used to define survival time variabilityBootstrapping (n = 10000) was used to estimate 95confidence intervals (CIs) around the interquartile rangeVariation in ALSFRS-R rates of decline was estimated bylinear mixed-effects models as described previously6 Lon-gitudinal sample calculations by Eland (2009) determinedthe number of patients necessary to detect a 25 reductionin slope after 12 months (monthly follow-up) with 80power and a 2-sided α of 56 The sample size was used as anestimate of the sensitivity of the populations to detect a giventreatment effect All meta-analyses were performed with theR metafor package (version 20-0 Viechtbauer21) Linearmixed-effects models were fitted with the lmer function(lme4 version 11-12)22 Imputation of missing data wasperformed using the aregImpute function (Hmisc ver-sion 40ndash3)

Data availability statementAll protocols analyses and anonymized data will be shared byrequest from any qualified investigator

ResultsA total of 38 randomized placebo-controlled clinical trialswere included in this study (figure e-2 available from Dryaddoiorg105061dryad86f1m6g) their eligibility criteriawere applied to an incidence cohort of 2904 patients withALS Figure 1 summarizes the primary eligibility criteria andexclusion rates per trial On average 598 (95 CI526ndash667) of the patients would be excluded from trialparticipation on the day of diagnosis in the incidence cohort

Table 1 Example of the calculation of the exclusion ratefor the Pentoxifylline 2006 trial

Criteria LimitsExcludedn ()

Cumulativen ()

Age 18ndash80 y 205 (71) 205 (71)

Symptomduration

6ndash47 mo 773 (266) 934 (322)

Vital capacity le100 predicted 818 (282) 1500 (517)

El Escorial criteria Definite orprobable

1146(395)

2003 (690)

Using an incident cohort of patients with amyotrophic lateral sclerosis (ALSn = 2904) we calculated the individual and combined effects of 4 criteria onthe exclusion rate The cumulative multivariate exclusion rate in this ex-ample is 690 The cumulative exclusion rate is calculated by the sequentialapplication of all 4 criteria (the default method in current ALS trials)

NeurologyorgN Neurology | Volume 92 Number 5 | January 29 2019 e453

Exclusion rates varied widely between trials ranging from14 to 95 Between 2000 and 2010 533 of the patientswere excluded which increased to 655 between 2010 and2017 (p = 008) Not meeting a specific El Escorial category isthe most important reason for exclusion (23 95 CI18ndash28) followed by FVC (17 95 CI 14ndash20) anddisease duration (12 95 CI 8ndash15) (figure e-1 availablefrom Dryad doiorg105061dryad86f1m6g)

The enrolled populations of the 38 trials are presented infigure 2 and reveal a large variability between trials of all

patient characteristics (p lt 0001) except for the proportionof men (p = 021) If we select those patients who are eligiblefor gt50 of the trials from our general population (eligiblepopulation in table 2) and compare them with actual trialparticipants at our center (trial participants in table 2) largedifferences can be seen Actually enrolled patients (n = 26090 table 2 column 4) differ from their eligible populationin sex (more men p = 0019) age (younger p lt 0001)progression rate (slower p lt 0001) ENCALS risk profile(better p lt 0001) and survival (longer p lt 0001) Thissuggests that despite the already applied eligibility criteria

Figure 1 Overview of eligibility criteria and exclusion rates of ALS trials conducted between 2000 and 2017

Caterpillar plot of the exclusion rates at diagnosis per trial (n = 38) Exclusion rates ranged from 14 to 95 numerical results per trial are given in table e-1(available from Dryad doiorg105061dryad86f1m6g) ALS = amyotrophic lateral sclerosis ALSFRS (n) = number of selection items on Amyotrophic LateralSclerosis Functional Rating Scale Def = definite FALS = familial ALS FVC = forced vital capacity percent predicted G-CSF = granulocyte colony-stimulatingfactor IGF-1 = insulin-like growth factor-1 LI = lead-in ALSFRS-R slope LS = probable laboratory supported Prob = probable

e454 Neurology | Volume 92 Number 5 | January 29 2019 NeurologyorgN

only a selective subset of the eligible patients will participate intrials This finding is indicative of an additional latent selectionprocess

Eligible patients are compared to the general ALS population11 less likely to die during the first 24 months after diagnosis(pooled hazard ratio 089 95 CI 084ndash093 p lt 0001)When we stratify the eligible population into prognosticsubgroups by applying a personalized prediction model wecan seen that defining eligibility leads primarily to a temporarysurvival difference in the poorest prognostic group (6-monthsurvival increases from 655 to 816 figure 3) Largeproportions (45ndash66) of other prognostic subgroups arealso excluded while the effect on survival is minimal (anabsolute 23ndash46 increase at 12 months and 15ndash36increase at 18 months) Within the eligible population 127of patients still have a very poor prognosis and 163 are very

long survivors Results were similar for the individual trials(figure e-3 available from Dryad doiorg105061dryad86f1m6g) revealing a rather random exclusion process withinall prognostic subgroups Overall eligibility criteria reducedheterogeneity in survival time by only 69 (95 CI minus35 to163 p = 009) reducing survival time variability from 220months (95 CI 208ndash234) to 205 months (95 CI188ndash224) This further underscores that both very short- andlong-surviving patients remain in the trial population

Longitudinal ALSFRS-R scores up to 24 months after di-agnosis were available for 696 patients Patients with a poorprognosis are relatively underrepresented in the ALSFRS-Rdata (bar charts figure 4A) Between-patient ALSFRS-Rvariability does not differ between unselected and trial-eligiblepatients (slope variance 062 vs 065 p = 025) The averageALSFRS-R rate of decline in eligible patients however is

Figure 2 Between-trial variability in population characteristics of enrolled participants

Caterpillar plots of the reported sum-mary data for the experimental and pla-cebo groups (total number of patients10489) Solid black diamonds indicatethe meta-analyzed average with 95confidence interval Study heterogeneitywas present in all baseline characteristicsexcept for the proportion of men (I2 =11 p = 021) (A) Age at randomization(B) proportion of men (C) symptom du-ration (D) bulbar onset (E) vital capacity(VC) and (F) Amyotrophic Lateral Sclero-sis Functional Rating ScalendashRevised(ALSFRS-R)

NeurologyorgN Neurology | Volume 92 Number 5 | January 29 2019 e455

higher (091 vs 098 p = 014) which could reduce the samplesize by 53 (from 125 to 118 patients per arm) Figure 5shows the sample size reductions for the 38 trials relative totheir eligibility rate revealing a strong relationship betweenboth variables (Pearson r 058 95 CI 032ndash076 p lt 0001)The sample size estimates for trial populations defined by theENCALS risk score are shown in black and gray A sample sizereduction of 34 similar to Edaravone23 can be achieved byselecting only those patients with risk scores between 055 and33 This single selection criterion would mean that asymp48 ofthe patients remain eligible which would lead to an almost5-fold increase in eligibility rate compared to EdaravoneNone of the trials was more efficient than the predictionmodel in optimizing both the sensitivity and eligibility rate ofthe populations

DiscussionIn this study we show that on average 598 of the patientswith ALS are found to be ineligible to participate in clinicaltrials Although eligibility criteria reduce the number ofpatients with a poor prognosis there is an adverse exclusionprocess that leads to a substantial untargeted exclusion ofpatients from all prognostic subgroups Moreover currentlyapplied eligibility criteria select populations that may stillcontain a relatively high number of patients who show slow orfast progression rates and do not reduce between-patientvariability These findings raise questions regarding not onlythe value of currently applied eligibility criteria but also thegeneralizability of clinical trial results in ALS Using prediction

models could individualize participant selection and optimizethe balance between endpoint heterogeneity and the gener-alizability of trial results

The concept of generalizability plays a central role in thetranslation of trial results to medical decision making4 Clinicaltrials with highly selected subgroups are difficult to interpret inreal-world settings and the safety or effectiveness of a drug maybe unknown for the majority of the patients We show that598 of patients are excluded from participation at diagnosisThis percentage is however an underestimation because mostpatients will be enrolled a few months after diagnosis (6ndash9months) At that time 15 to 24 of our patients who couldtheoretically be prescribed the drug at diagnosis are deceasedand are thus never evaluated in clinical trials Moreover a largerproportion of the remaining patients will fail the criteria due todisease progression Together thesemay lead to exclusion ratesin real-world settings that approximate those reported in otherfields (80ndash96)452425 These high exclusion rates couldresult in the indirect removal of patients with specific drug-responsive pathways For example patients with ALSndashfrontotemporal dementia and familial ALS are often excludedhowever because these subtypes are related to the C9orf72repeat expansion26 they are indirectly related to C9orf72 dis-ease pathways27 As was shown recently28 it is possible that thetreatment effect is modified by pharmacogenetic interactionswhich could be missed by the indirect exclusion of specificsubgroups and may disguise important treatment clues Thesamemay hold true for the larger exclusion rates among bulbar-onset patients or women

Table 2 Population characteristics of various populations

General population Eligible patients Trial participants

(n = 2904) (n = 1194)Trial participants alltrials (n = 10489)

Trial participantsUMCU (n = 260)

Age at onset y 638 (634ndash642) 614 (609ndash620) 559 (552ndash567) 558 (544ndash573)

Male (n) 578 (1678) 544 (649) 633 (6640) 623 (162)

Bulbar site of onset (n) 371 (1078) 351 (420) 212 (2220) 314 (82)

Diagnostic delay mo 100 (97ndash103) 94 (91ndash97) mdash 87 (81ndash93)

ALSFRS-R score at diagnosis 390 (388ndash393) 397 (394ndash400) mdash 415 (410ndash420)

DFRS mo 082 (079ndash085) 081 (078ndash085) mdash 068 (062ndash074)

FVC at diagnosis predicted 865 (855ndash874) 923 (911ndash935) mdash 994 (974ndash1014)

ENCALS risk profile 10 (094ndash107) 098 (091ndash105) mdash 062 (051ndash074)

Survival after diagnosis (median) mo 177 (170ndash184) 194 (186ndash204) mdash 278 (256ndash314)

ALSFRS-R slope after enrollment mdash mdash 097 (092ndash102) 111 (101ndash121)

Abbreviations ALSFRS-R = Amyotrophic Lateral Sclerosis Functional Rating ScalendashRevisedDFRS = (48minusALSFRS-Rscorediagnosis)symptomduration ENCALS=European Network for the Cure of ALS FVC = forced vital capacity UMCU = University Medical Center UtrechtENCALS risk profiles are relative risk estimated by the ENCALSmodel and can be interpreted as hazard ratio A relative risk of 062means that on average therisk of dying during follow-up is 38 lower compared to the general population General population includes all consecutive patients diagnosed withamyotrophic lateral sclerosis (ALS) in the Netherlands between 2006 and 2016 Eligible patients are patients in the Netherlands who are eligible for gt50 ofthe trials Trial participants all trials are meta-analyzed population characteristics of the 38 trials Trial participants UMCU are patients from the general ALSpopulation who participated in a clinical trial during the same time period at the UMCU Data are mean (95 confidence interval) or percent (n)

e456 Neurology | Volume 92 Number 5 | January 29 2019 NeurologyorgN

Because of the heterogeneous nature of ALS29 eligibilitycriteria aim primarily to reduce the amount of between-patient variation and to improve protocol adherence23 Weshow however that currently defined populations still con-tain fast- and slow-progressing patients and between-patientvariation is virtually unaffected A likely explanation is thatselection criteria are applied in a step-wise univariate manner(table 1) while progression rate is dictated by prognosis30

which is defined by a multivariate combination of predictors7

When a trial aims to exclude fast-progressing patients patientswho are on the lower limits of each criterion could still beenrolled The sum of the lower limits however means thatthese patients have a poor prognosis and fast progressionrates To exemplify a trial with only 2 criteria (age lt75 yearsand FVC gt60) would enroll a 74-year-old patient with anFVC of 61 but exclude a 76-year-old with an FVC of 104The first patient is likely to exhibit a faster rate of decline andis more likely to die during follow-up More important thereal-world effect is minimal The dexpramipexole study31 forexample excluded patients with a disease duration gt24months and an FVC lt65 However if none of the criteriahad been applied the statistical power would have been re-duced from 90 to 88 (assuming an inflation of the

reported SD of 3) It is doubtful whether a 2 gain in powerjustifies the exclusion of 25 to 40 of the patients

This questions the value of currently used eligibility criteriaand a revision would seem to be indicated There is a need tobalance endpoint heterogeneity (or sample size) and thegeneralizability of trials This balance could be achieved byusing individual risk scores rather than group-level criteriaThe risk score can be conceptualized as a summary of allavailable prognostic information per individual Replacingsets of several eligibility criteria by a single risk estimatewould allow investigators to select only those patients whoare the most likely to exhibit the investigator-preferred dis-ease pattern This could reduce between-patient heteroge-neity more effectively while balancing both generalizabilityand eligibility In our ALSFRS-R example for Edaravone23

instead of excluding 90 of the patients the same homo-geneity effect could be reached using a risk-based selectionwith 48 of the patients remaining eligible (a nearly 5-foldincrease)

Our study has several limitations that should be consideredFirst because populations may differ between countries our

Figure 3 Effect of selection criteria on overall mortality since diagnosis

Mortality since diagnosis for 5 prognostic subgroups defined by the European Network for the Cure of ALS personalized prediction model Colors representthe 5 prognostic subgroups very long (green) long (yellow) intermediate (orange) short (red) and very short (black) survival7 Solid lines in the Kaplan-Meierplots represent the survival patterns for all patients (n = 2904 A) whereas the dotted lines are the survival curves for the eligible population (n = 1194 B)defined by all patients who are eligible for gt 50 of the trials Boxplots provide the variability in survival time per prognostic subgroup (left all patients righteligible patients) Dotted lines in boxplot are the unstratified interquartile ranges (25th and 75th percentile) overall survival time variability was reduced by69 (95 confidence interval minus35 to 163 p = 009) Bar charts provide the number of patients in each subgroup

NeurologyorgN Neurology | Volume 92 Number 5 | January 29 2019 e457

results are limited to a specific geographic area In Italy andIreland for example the age at onset is slightly higher whichwould underestimate our current exclusion rate32 Second thiswork focused primarily on optimizing the balance betweenendpoint heterogeneity and generalizability of trial resultsHowever ALS remains a complex disorder and investigatorsmay apply eligibility criteria with imperatives other than simplyselecting the most responsive subgroup (eg safety pharma-codynamics or hypothesized mechanism or action) It maytherefore be insightful to distinguish between biological eligi-bility criteria (those that are set for the hypothesized drugmechanism) and design criteria (those that are set to op-timize the trial design) This is especially important whenconsidering the current developments toward a personal-ized genotype-oriented approach in ALS2829 Genotype- orbiomarker-oriented trials inherently have high exclusion ratesand potentially investigate (ultra) rare subgroups which com-promises their feasibility The antisense trial in SOD1-relatedALS33 for example had an enrollment rate of 1 patient per 4months per site which underscores the importance of opti-mizing the eligibility rate Therefore future clinical trials couldcombine both group-level biological (eg genetic marker ordisease pathway) and individualized risk-based criteria tooptimize both drug responsiveness and trial design

As final note our results supported by previous studies234

indicate a latent selection of patients in which young malepatients with a relatively mild disease are overrepresented intrial populations This latent selection process is most prob-ably the result of a multifactorial process one that cannot beestimated in the meta-data of published clinical trials Thereason for eligible patients with ALS declining study partici-pation is in 94 of cases the physical burden9 Given therelative underrepresentation of patients in advanced stages itis plausible that the physical burden is an important latentfactor one that becomes more apparent as the disease pro-gresses This is supported by the observation that dropout andnoncompliance are related to lower ALSFRS-R and FVCscores3 Similarly the physician may deem the patient unfit toundergo the trial and not offer the option of participating9 Sexdifferences may also play a role female patients are moreoften reluctant about medical testing or have an inability tocope with the protocol35 These latent factors may be over-expressed when 2 trials run simultaneously at the same siteThe patientrsquos choice for a certain trial could be related toa certain characteristic of the patient (eg sex or age) Ded-icated studies are needed to determine the relevance of eachfactor and to potentially develop strategies to reduce thislatent selection process (eg by reducing the physical burden

Figure 4 Longitudinal ALSFRS-R patterns in a subset of 696 patients

Colors of the boxplots represent the 5 prognostic subgroups very long (green) long (yellow) intermediate (orange) short (red) and very short (black)survival7 Purple dotted line is the average pattern of decline in the eligible population (n = 356) Solid purple line is the average pattern in the full dataset (n =696) Boxplots provide the variability in rates of decline over time (points permonth) estimated by the best unbiased linear predictors (BLUPs) from an linearmixed-effects model Dotted lines in boxplots represent the average rate of decline (A) All patients and (B) eligible patients ALSFRS-R = Amyotrophic LateralSclerosis Functional Rating ScalendashRevised

e458 Neurology | Volume 92 Number 5 | January 29 2019 NeurologyorgN

using home-based outcome measures or patient-reporteddata)36

Our results reveal that the majority of patients with ALS areexcluded from trial participation at diagnosis which raisesquestions regarding the generalizability of current trials Ex-clusion of ineligible patients only minimally improves homo-geneity in trial endpoints A risk-based selection criterion couldindividualize trial participant selection and may improve thebalance between endpoint heterogeneity and exclusion rates

Author contributionsDesign or conceptualization of the study RPAvE H-JWLHvdB Analysis or interpretation of the data RPAvEH-JW SN IEV MJCE Drafting or revising the manu-script for intellectual content all authors

Study fundingSupported by the Netherlands ALS Foundation (ProjectTryMe)

DisclosureR van Eijk H Westeneng S Nikolakopoulos I Verhagen andM van Es report no disclosures relevant to the manuscript MEijkemans received grants from The Netherlands Organizationfor Health Research and Development (Veni scheme) theThierry Latran Foundation and The Netherlands ALS Foun-dation (Stichting ALS Nederland) He received travel grantsfrom Baxalta and serves on the biomedical research advisorypanel of theUKMotorNeuroneDisease Association L van denBerg reports grants from The Netherlands ALS Foundation

The Netherlands Organization for Health Research and De-velopment (Vici scheme) and The Netherlands Organizationfor Health Research and Development (Sampling and Bio-marker Optimization and Harmonization in ALS and OtherMotor Neuron Diseases [SOPHIA] Survival Trigger andRisk Epigenetic Environmental and Genetic Targets forMotor Neuron Health [STRENGTH] A Programme for ALSCare in Europe [ALS-CarE] project) funded through the EUJoint ProgrammemdashNeurodegenerativeDisease Research) andserving on the Scientific Advisory Board for Biogen Cytoki-netics Prinses Beatrix SpierFonds and the Thierry LatranFoundation Go to NeurologyorgN for full disclosures

Publication historyReceived by Neurology June 22 2018 Accepted in final form September19 2018

References1 Mitsumoto H Brooks BR Silani V Clinical trials in amyotrophic lateral sclerosis why

so many negative trials and how can trials be improved Lancet Neurol 2014131127ndash1138

2 Chio A Canosa A Gallo S et al ALS clinical trials do enrolled patients accuratelyrepresent the ALS population Neurology 2011771432ndash1437

3 Atassi N Yerramilli-Rao P Szymonifka J et al Analysis of start-up retention andadherence in ALS clinical trials Neurology 2013811350ndash1355

4 Rothwell PM External validity of randomised controlled trials ldquoto whom do theresults of this trial applyrdquo Lancet 200536582ndash93

5 Zimmerman M Mattia JI Posternak MA Are subjects in pharmacological treatmenttrials of depression representative of patients in routine clinical practice Am J Psy-chiatry 2002159469ndash473

6 van Eijk RPA Eijkemans MJC Ferguson TA Nikolakopoulos S Veldink JH van denBerg LH Monitoring disease progression with plasma creatinine in amyotrophiclateral sclerosis clinical trials J Neurol Neurosurg Psychiatry 201889156ndash161

7 Westeneng HJ Debray TPA Visser AE et al Prognosis for patients with amyotrophiclateral sclerosis development and validation of a personalised prediction modelLancet Neurol 201817423ndash433

8 Steyerberg EW Clinical Prediction Models A Practical Approach to DevelopmentValidation and Updating New York Springer 2009

Figure 5 Effect of trial eligibility criteria and the ENCALS risk scores on sample size

Sample size calculations were performed forvarious populations either selected by thetrial eligibility criteria (n = 38 orange) or de-fined by different cutoffs for the EuropeanNetwork for the Cure of ALS (ENCALS) riskscores (n = 1050 combinations) Results areexpressed as sample size inflation factor (IFx-axis) To exemplify an IF of 079means thatthe sample size is 21 smaller compared tothe sample size necessary when all patientsare included (green triangle 100 eligibility)Populations defined by the ENCALS riskscores for all possible combinations of cutoffvalues are shown in black and gray Blackdots are the selected populations thatresulted in the largest reduction in samplesize and highest eligibility rate To exemplifythe Edaravone 2017 trial resulted in thelargest reduction in sample size (minus34 IF066) with a 10 eligibility rate The ENCALSmodel could select a similarly sensitive pop-ulation with 48 of the patients remainingeligible Numerical effects of the differenttrials are provided in table e-1 (availablefrom Dryad doiorg105061dryad86f1m6g)

NeurologyorgN Neurology | Volume 92 Number 5 | January 29 2019 e459

9 Bedlack RS Pastula DMWelsh E Pulley D Cudkowicz ME Scrutinizing enrollment inALS clinical trials room for improvement Amyotroph Lateral Scler 20089257ndash265

10 Pinto S deCarvalhoMComparison of slow and forced vital capacities on ability to predictsurvival in ALS Amyotroph Lateral Scler Frontotemporal Degener 201718528ndash533

11 Hozo SP Djulbegovic B Hozo I Estimating the mean and variance from the medianrange and the size of a sample BMC Med Res Methodol 2005513

12 Fu R Vandermeer BW Shamliyan TA et al Handling Continuous Outcomes inQuantitative Synthesis Methods Guide for Effectiveness and Comparative Effec-tiveness Reviews Rockville Agency for Healthcare Research and Quality 2008

13 Huisman MH de Jong SW van Doormaal PT et al Population based epidemiologyof amyotrophic lateral sclerosis using capture-recapture methodology J NeurolNeurosurg Psychiatry 2011821165ndash1170

14 Brooks BR Miller RG Swash M Munsat TL World Federation of Neurology Re-search Group on Motor Neuron Disease El Escorial revisited revised criteria for thediagnosis of amyotrophic lateral sclerosis Amyotroph Lateral Scler Other MotNeuron Disord 20001293ndash299

15 White IR Royston P Imputing missing covariate values for the Cox model Stat Med2009281982ndash1998

16 Rubin DB Schenker NMultiple imputation in health-care databases an overview andsome applications Stat Med 199110585ndash598

17 Higgins JPT Green S Cochrane Collaboration Cochrane Handbook for SystematicReviews of Interventions Chichester Wiley-Blackwell 2008

18 Knapp G Hartung J Improved tests for a random effects meta-regression with a singlecovariate Stat Med 2003222693ndash2710

19 Royston P Parmar MK Flexible parametric proportional-hazards and proportional-odds models for censored survival data with application to prognostic modelling andestimation of treatment effects Stat Med 2002212175ndash2197

20 Rooney J Fogh I Westeneng HJ et al C9orf72 expansion differentially affects maleswith spinal onset amyotrophic lateral sclerosis J Neurol Neurosurg Psychiatry 201788281ndash287

21 Viechtbauer W Conducting meta-analyses in R with the metafor package J StatSoftware 20103648

22 Bates D Machler M Bolker B Walker S Fitting linear mixed-effects models usinglme4 J Stat Software 2015671ndash48

23 Edaravone Writing Group Safety and efficacy of edaravone in well defined patientswith amyotrophic lateral sclerosis a randomised double-blind placebo-controlledtrial Lancet Neurol 201716505ndash512

24 Zetin M Hoepner CT Relevance of exclusion criteria in antidepressant clinical trialsa replication study J Clin Psychopharmacol 200727295ndash301

25 Hoertel N Le Strat Y De Maricourt P Limosin F Dubertret C Are subjects intreatment trials of panic disorder representative of patients in routine clinical practiceResults from a national sample J Affect Disord 2013146383ndash389

26 Umoh ME Fournier C Li Y et al Comparative analysis of C9orf72 and sporadicdisease in an ALS clinic population Neurology 2016871024ndash1030

27 Wen X Westergard T Pasinelli P Trotti D Pathogenic determinants and mecha-nisms of ALSFTD linked to hexanucleotide repeat expansions in the C9orf72 geneNeurosci Lett 201763616ndash26

28 van Eijk RPA Jones AR Sproviero W et al Meta-analysis of pharmacogeneticinteractions in amyotrophic lateral sclerosis clinical trials Neurology 2017891915ndash1922

29 van Es MA Hardiman O Chio A et al Amyotrophic lateral sclerosis Lancet 20173902084ndash2098

30 van Eijk RPA Eijkemans MJC Rizopoulos D van den Berg LH Nikolakopoulos SComparing methods to combine functional loss and mortality in clinical trials foramyotrophic lateral sclerosis Clin Epidemiol 201810333ndash341

31 Cudkowicz ME van den Berg LH Shefner JM et al Dexpramipexole versus placebofor patients with amyotrophic lateral sclerosis (EMPOWER) a randomised double-blind phase 3 trial Lancet Neurol 2013121059ndash1067

32 DrsquoOvidio F Rooney JPK Visser AE et al Critical issues in ALS case-control studiesthe case of the Euro-MOTOR study Amyotroph Lateral Scler FrontotemporalDegener 201718411ndash418

33 Miller TM Pestronk A David W et al An antisense oligonucleotide againstSOD1 delivered intrathecally for patients with SOD1 familial amyotrophic lateralsclerosis a phase 1 randomised first-in-man study Lancet Neurol 201312435ndash442

34 Hardiman O Al-Chalabi A Brayne C et al The changing picture of amyotrophiclateral sclerosis lessons from European registers J Neurol Neurosurg Psychiatry201788557ndash563

35 Markanday S Brennan SL Gould H Pasco JA Sex-differences in reasons for non-participation at recruitment Geelong Osteoporosis Study BMC Res Notes 20136104

36 Wicks P Vaughan TE Massagli MP Heywood J Accelerated clinical discovery usingself-reported patient data collected online and a patient-matching algorithm NatBiotechnol 201129411ndash414

e460 Neurology | Volume 92 Number 5 | January 29 2019 NeurologyorgN

DOI 101212WNL0000000000006855201992e451-e460 Published Online before print January 9 2019Neurology

Ruben PA van Eijk Henk-Jan Westeneng Stavros Nikolakopoulos et al Refining eligibility criteria for amyotrophic lateral sclerosis clinical trials

This information is current as of January 9 2019

ServicesUpdated Information amp

httpnneurologyorgcontent925e451fullincluding high resolution figures can be found at

References httpnneurologyorgcontent925e451fullref-list-1

This article cites 33 articles 8 of which you can access for free at

Citations httpnneurologyorgcontent925e451fullotherarticles

This article has been cited by 1 HighWire-hosted articles

Subspecialty Collections

httpnneurologyorgcgicollectionrisk_factors_in_epidemiologyRisk factors in epidemiology

httpnneurologyorgcgicollectionparkinsons_disease_parkinsonismParkinsons diseaseParkinsonismfollowing 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