Alzheimer’s & Dementia 9 (2013) 438–444

Designing drug trials for Alzheimer’s disease: What we have learnedfrom the release of the phase III antibody trials: A report from the EU/US/

CTAD Task ForceBruno Vellasa,b,c,*, Maria C. Carrillox, Cristina Sampaioe, H. Robert Brashearf, Eric Siemersg,Harald Hampelh, Lon S. Schneideri, Michael Weinerj, Rachelle Doodyk, Zaven Khachaturianl,

Jesse Cedarbaumm, Michael Grundmann, Karl Broicho, Ezio Giacobinip, Bruno Duboiss,Reisa Sperlingq, Gordon K. Wilcockr, Nick Foxt, Philip Scheltensu, Jacques Touchonv,

Suzanne Hendrixw, Sandrine Andrieua,b,c, Paul Aisend;and the EU/US/CTAD Task Force Membersy

aUMR1027 Inserm, Toulouse, FrancebUniversity of Toulouse III, Toulouse, France

cGerontopole Toulouse, Toulouse University Hospital, Toulouse, FrancedDepartment of Neurosciences, University of California, San Diego, San Diego, CA, USA

eCHDI Foundation, Princeton, NJ, USAfJanssen AI, South San Francisco, CA, USA

gEli Lilly, Indianapolis, IN, USAhDepartment of Psychiatry, Goethe University, Frankfurt, Germany

iUniversity of Southern California Keck School of Medicine, Los Angeles, CA, USAjCenter for Imaging of Neurodegenerative Disease, VA Medical Center, University of California San Francisco, San Francisco, CA, USA

kAlzheimer’s Disease and Memory Disorders Center, Baylor College of Medicine, Houston, TX, USAlThe Campaign to Prevent Alzheimer’s Disease by 2020, Potomac, MD, USA

mBristol-Myers Squibb, Wallingford, CT, USAnGlobal R&D Partners, LLC, San Diego, CA, USA

oFederal Institute of Drugs and Medical Devices, Bonn, GermanypDepartment of Internal Medicine, Rehabilitation and Geriatrics, University of Geneva, Geneva, Switzerland

qBrigham and Women’s Hospital, Harvard Medical School, Boston, MA, USArUniversity of Oxford, Oxford, UK

sInstitute for Memory and Alzheimer’s Disease, UMR-S975, Salpetri�ere Hospital, Pierre & Marie Curie University, Paris, FrancetDementia Research Centre, UCL Institute of Neurology, London, UK

uDepartment of Neurology and Alzheimer Center, VU University Medical Centre, Amsterdam, NetherlandsvUniversity Hospital of Montpellier, Montepellier, France

wPentara Corporation, Salt Lake City, UT, USAxAlzheimer’s Association, Chicago, IL, USA

yTask Force Members: Paul Aisen, USA; Sandrine Andrieu, France;

Elizabeth Ashford, Switzerland; Wolfgang Barth, Switzerland; Robert Ber-

man, USA; H. Robert Brashear, USA; Karl Broich, Germany; Maria Car-

rillo, USA; Jesse Cedarbaum, USA; Francesca Cerreta, UK; William Cho,

USA; Caroline Cohen, France; Susan De Santi, USA; Catherine Debieuvre,

France; Rachelle Doody, USA; Ernest Dorflinger, Switzerland; Bruno Du-

bois, France; Judith Dunn, Switzerland; Michael Egan, USA; Nick Fox,

UK; Ezio Giacobini, Switzerland; Ana Graf, USA; Bjorn Aaris Gronning,

Denmark; Michael Grundman, USA; Harald Hampel, Ireland; Suzanne

Hendrix, USA; Patrick Kamphuis, Netherlands; Zaven Khachaturian,

USA; Stine Anna Larsen, Denmark; John Lawson, USA; Johan Lindborg,

USA; Enchi Liu, USA; Prisca Lucas, France; Valentina Mantua, Italy;

Glenn Morrison, Switzerland; Cristina Murat, France; Francois Nicolas,

UK; Susanne Ostrowitzki, Switzerland; Robert Paul, USA; Jana Podhorna,

France; Maria Pueyo, France; J. Michael Ryan, USA; Cristina Sampaio,

USA; Philips Scheltens, Netherlands; Rachel J. Schindler, USA;

Hans-Martin Schneble, France; Lon Schneider, USA; Adam Schwarz,

USA; Eric Siemers, USA; Reisa Sperling, USA; Andrew Stephens, Ger-

many; Aleksandra Stjepanovic-Agovic, France; Albrecht St€offler, Ger-

many; Joyce Suhy, USA; Patrick Therasse, Belgium; Jacques Touchon,

France; Bruno Vellas, France; Laura Vernoux, Belgium; Michael Weiner,

USA; Rico Wieggers, Netherlands; Gordon Wilcock, UK; Hans-Christian

Wirtz, USA; Sacha Wissink, Netherlands; Kristin Kahle Wrobleski, USA;

and Jean Zetlaoui, Switzerland.

*Corresponding author. Tel.: 1011 05 61 77 64 25.

E-mail address: vellas.bruno@gmail.com

1552-5260/$ - see front matter � 2013 The Alzheimer’s Association. All rights reserved.

http://dx.doi.org/10.1016/j.jalz.2013.03.007

B. Vellas et al. / Alzheimer’s & Dementia 9 (2013) 438–444 439

Abstract An international task force of investigators from academia, industry, nonprofit foundations, and

regulatory agencies met in Monte Carlo, Monaco, on October 31, 2012, to review lessons learnedfrom the recent bapineuzumab and solanezumab trials, and to incorporate insights gained from thesetrials into future clinical studies. Although there is broad consensus that Alzheimer’s disease (AD)should be treated during its earliest stages, the concept of secondary prevention has evolved to be de-scribed more accurately as treatment of preclinical, presymptomatic, or early AD. There continues tobe a strong emphasis on biomarkers and a need for new biomarkers; however, there has also been a re-alization, based on completed trials, that the most reliable indicator of clinical efficacy across the en-tire spectrum of disease from asymptomatic to AD dementia is likely a measure of cognition. The taskforce made many recommendations that should improve the likelihood of success in future trials, in-cluding larger phase 2 or combined phase 2/phase 3 studies, clear evidence of target engagement inthe central nervous system, evidence of downstream effects on biomarkers before initiating phase 3studies, consideration of adaptive and targeted trial designs, and use of sensitive measures of cogni-tion as the most robust indicator of treatment benefit.� 2013 The Alzheimer’s Association. All rights reserved.

Keywords: Alzheimer’s disease; Pharmacotherapy; Clinical trials

1. Introduction

In recent months, results have been reported from phase 3clinical trials of two antiamyloid immunotherapy agents—bapineuzumab and solanezumab—in participants with mildto moderate Alzheimer’s disease (AD). Neither trial achievedits primary end point; however, prespecified subgroup andcombined analyses of the solanezumab data showed a smallcognitive benefit that was more noticeable in the mild sub-group (not yet published). These results are consistent withthe emerging consensus that prevention ofADby treating peo-ple during the earliest stages of disease may be the best way toaddress the accelerating incidence of dementia as the world’spopulation ages. Moreover, although the results of the bapi-neuzumab trial were disappointing, there is much to learnfrom both the bapineuzumab and solanezumab research pro-grams that will inform the design of future trials. Againstthis backdrop, an international task force of investigatorsfrom academia, industry, nonprofit foundations, and regula-tory agencies met in Monte Carlo, Monaco, on October 31,2012, to review lessons learned from these trials and to docu-ment these insights for consideration in future clinical studies.

2. Lessons from antibody trials

Bapineuzumab is a humanized monoclonal antibody thatbinds to the N-terminus of the amyloid beta (Ab) peptide. Ina phase 2 study, bapineuzumab did not improve cognition orfunction significantly [1], but was shown to reduce the accu-mulation of fibrillar amyloid in the AD brain on amyloid pos-itron emission tomography (PET) [2]. Post hoc data analysissuggested a treatment effect and lower incidence ofamyloid-related imaging abnormalities–edema (ARIA-E) inapolipoprotein E (APOE) ε4 noncarriers; hence, two separatephase 3 studies were initiated in APOE ε4 carriers (study 302)and noncarriers (study 301). As a hedge against developmentof ARIA-E, carriers were randomized to receive only 0.5 mg/kg bapineuzumab or placebo, whereas noncarriers were ran-

domized to receive 0.5 mg/kg bapineuzumab, 1.0 mg/kg bapi-neuzumab, or placebo. Several issues have been identified thatmay have contributed to the disappointing results in phase 3.

Both studies used the same enrollment criteria, with theexpectation (based on the phase 2 data) that efficacy wouldbe higher in noncarriers. One potential reason this expecta-tion was not realized may be that the amyloid burden wasdifferent in the two groups. Amyloid imaging revealed that36.1% of subjects in the noncarrier study compared with6.5% in the carrier study had readings of a global cortical av-erage standard uptake value ratio in the region relative tostandardized gray matter below the preset threshold for am-yloid positivity, indicating that the amyloid burden waslower in carriers compared with noncarriers. Although therewas no a priori assumption that subjects in this trial would beamyloid positive even with a clinical diagnosis of AD, giventhat bapineuzumab targets amyloid, the inclusion of partici-pants without AD pathology may have contributed to thelack of efficacy. Other possible reasons for the lack of a treat-ment effect include insufficient dosing, duration of treat-ment, or the possibility that greater amyloid reduction maybe needed to affect the clinical course. It may also be thatsubjects were too far advanced for the drug to have an effect.Last, it is possible that amyloid removal alone may not besufficient to slow AD progression, and that treatment againsttau or other targets may be needed. All of these possible ex-planations for the lack of an apparent treatment effect arespeculative and require further investigation.

Logistical issues also introduced other challenges. Con-ducting a PET study across a network of PET centers in-creased the complexity of the study as investigators soughtto control variability in personnel, equipment, analysis, anddata submission. Monitoring for ARIA added further com-plexity [3]. During the 4-year trial, technology evolved andreporting procedures changed, introducing more variabilityinto the study. However, the bapineuzumab studies also re-sulted in improved procedures for identifying ARIA and

B. Vellas et al. / Alzheimer’s & Dementia 9 (2013) 438–444440

a better understanding of the natural history and risk of devel-oping ARIA-E (clearly greater for bapineuzumab than pla-cebo, increasing with dose and number of APOE ε4 alleles).

An important issue is why the phase 2 study signal wasnot replicated in phase 3. In the bapineuzumab phase 2 trial,there was, in fact, no treatment difference, with the only ap-parent difference being observed after multiple post hoccomparisons. Another issue was that the treatment differ-ence observed in a subset analysis of APOE ε4 noncarrierswas based on very small numbers and was driven primarilyby a greater than expected (nearly twice as great) decline inthe placebo group. Another factor that may have contributedto the phase 3 failure is that moving from a small study witha high level of operational control to a huge internationalstudy introduces additional variability. In addition, becausethe phase 2 trial was designed primarily for safety, it wasnot powered adequately to detect an efficacy signal. This ar-gues for securing a more definitive clinical efficacy signal inphase 2, for example, by conducting larger or combinedphase 2/phase 3 studies with greater power. Although thismay increase the probability of success in phase 3, such anapproach might also lengthen the drug development processfor disease-modifying treatments in AD.

Solanezumab, another humanized monoclonal antibody,targets an epitope in the middle of the Ab peptide. The pre-dominant message emerging from results of the two phase 3trials (EXPEDITION 1 and 2) is that targeting amyloid ap-pears to have a positive effect on cognition in mild AD, al-though there were no biomarker changes that indicateda treatment effect. Moreover, the trial suggested that clinicaldisease severity might be useful for stratification of subjectsto enhance an efficacy signal, although it was not useful inthe bapineuzumab and other trials.

The solanezumab trials also supported the need for largerstudies. In pooled analyses of the mild to moderate AD pop-ulations in the two studies, each with more than 1000 partic-ipants, there were positive signals in multiple cognitivemeasures. However, these signals were driven primarily bythe milder patients. When the mild subgroup was analyzedseparately, the number of subjects in each of the two studiesdid not provide sufficient power, but identical study param-eters allowed for pooling of the data. The pooled results re-vealed a 34% slowing on the Alzheimer’s DiseaseAssessment Scale-Cognitive Subscale (ADAS-cog14) andthe Mini-Mental State Examination (MMSE), and a slowingof functional decline on the Alzheimer’s Disease Coopera-tive Study-Activities of Daily Living (P5 .057) at 80 weeksin the mild subgroup. More important, the percent reductionin cognitive decline in participants with mild AD was rela-tively stable over 80 weeks, consistent with disease modifi-cation. The divergence between placebo and treatmentgroups continued in the mild group even after they advancedto moderate AD, suggesting that early intervention may sta-bilize progression to some extent.

Biomarker data in the solanezumab trials provided someevidence of target engagement as demonstrated by a substan-

tial increase in total cerebrospinal fluid (CSF) Ab1–40 andAb1–42, and lower CSF levels of free Ab1–40. In the mildsubgroup, there was a nonsignificant trend toward a treat-ment effect, reflected by reduced amyloid deposition onPET scans. However, there were no significant differencesin terms of brain volume or on CSF biomarkers of neural in-jury (i.e., CSF tau), raising concerns about whether the ther-apy modified the disease process or, alternatively, whetherthese markers have the necessary sensitivity to detect neuro-nal damage.

Among the lessons learned from the solanezumab and ba-pineuzumab trials is the need for caution in committing tolarge phase 3 trials in the absence of a clinical efficacy signalin phase 2. These trials also clearly point to the value of strat-ifying participants for amyloid-based disease modificationtrials based on clinical severity; however, the utility of strat-ifying by APOE ε4 status is not so clear. Stratification byAPOE ε4 status, however, may still be needed for safety rea-sons. Although amyloid positivity (imaging or CSF) may beuseful in selecting participants for trials of antiamyloid ther-apies, biomarker and structural neuroimaging studies inamyloid-negative subjects are also needed to understandmore completely the trajectory of clinical course and bio-marker changes in this heterogeneous group of participants.The Alzheimer’s Disease Cooperative Study will continue toexplore these data sets to understand more fully the meaningof negative amyloid scans, the relationship between bio-markers and cognitive measures, and patient characteristicsthat may relate to treatment response.

3. Evolving ideas about prevention

Although much has been written about AD prevention tri-als, new criteria proposed by the International WorkingGroup [4] and work groups convened by the National Insti-tute on Aging and Alzheimer’s Association [5], as well asgrowing biomarker evidence, suggest that secondary preven-tion is described more accurately as treatment of preclinical,presymptomatic, or early AD. However, current U.S. Foodand Drug Administration guidelines for AD treatment trials,which require demonstration of efficacy for dual primaryoutcomes—a broad cognitive measure and a functional orglobal measure to establish clinical relevance—cannot beapplied to preclinical AD trials in which there are no symp-toms to assess and no way to demonstrate clinically relevantbenefits other than assessing time to cognitive impairment orprogression of cognitive impairment. A surrogate marker forcognitive change or progression is needed, and regulatorshave suggested that a surrogate marker that is not fully val-idated may be sufficient if the surrogate is “reasonablylikely” to predict clinical benefits.

There are currently several trials that focus on this pre-clinical or early AD group. The Collaboration for Alz-heimer’s Prevention is an umbrella group, hosted byFidelity Biosciences Research, Inc., and the Alzheimer’s As-sociation, which encompasses three prevention initiatives:

B. Vellas et al. / Alzheimer’s & Dementia 9 (2013) 438–444 441

the Dominantly Inherited Alzheimer’s Network (DIAN), theAlzheimer’s Prevention Initiative, and the AntiamyloidTreatment of Asymptomatic Alzheimer’s (A4) trial. TheCollaboration for Alzheimer’s Prevention was developedto promote harmonization among the studies, prevent dupli-cation of effort, and share data. All three of these trials arebased on the view that the pathological process of AD beginslong before the development of dementia, and that amyloidplays a critical role in the pathogenesis of AD [6]. All threeinitiatives also include extensive biomarker studies for thepurposes of enrichment, evaluation of treatment effect, test-ing of the amyloid hypothesis, and validation of the bio-markers for their various uses.

TheA4 trial will test the hypothesis that lowering amyloidburden can slow the rate of decline in peoplewith evidence ofcerebral amyloid accumulation but no clear symptoms.Based on the findings of trial simulations, A4 will select cog-nitively normal subjects who are positive for cerebral amy-loid demonstrated by PET imaging and treat them witha biologically active compound (still to be selected) for3 years. An Ab-negative arm will also be included to studythe natural history and to map the trajectory of biomarkersand cognition in individuals with no measurable amyloid,and to understand more completely nonamyloid-relateddrivers of age-related cognitive decline. An ethics substudywill explore the effects of disclosure of amyloid status.

The bapineuzumab and solanezumab trials raise the con-cern that structural imaging does not reflect the effect oftreatment with antiamyloid monoclonals. These studies fur-ther suggest dissociation between the treatment effect oncognition and the treatment effect on brain volume. This re-alization has led to modification of the original A4 design,which would have used imaging or CSF biomarkers for in-terim adaptive decisions about treatment selection.

The results of trials to date, in which biomarkers have beenuninformative or have diverged fromcognitive results, suggestthat the most reliable indicator of clinical efficacy across theentire spectrumofdisease fromasymptomatic toADdementiais likely ameasure of cognition. Thus, the primaryoutcome forthe A4 trial will be a composite clinical measure that incorpo-ratesmeasures of episodicmemory (free and cued selective re-minding, delayed recall, and logical memory paragraphrecall), executive function (digit symbol span), and theMini-Mental State Examination (MMSE) [7]. As mentionedat the task force meeting by a study investigator, testing thiscomposite instrument in a variety of data sets suggests thata feasible trial size of 500 subjects per arm would have thepower to detect a treatment effect of 11% to 35%.

Although the assessment of efficacy will not be driven bybiomarker results, everyone enrolled in the study will be re-quired to undergo PET amyloid imaging, volumetric mag-netic resonance imaging (MRI), and task-free functionalMRI to assess default network connectivity. At least halfthe participants will undergo lumbar puncture for CSF bio-marker assessment. Fluorodeoxyglucose (FDG)-PET willbe included in a subset if additional funding is obtained.

In the European Union, several initiatives are evaluatingmultidomain interventions for prevention of AD, includinglifestyle factors (e.g., nutrition, physical and cognitive exer-cise). Several trials have been funded, including the Multi-domain Alzheimer Prevention Trial (MAPT), which willtest the efficacy of omega 3 fatty acids and cognitive andphysical exercise in 1680 older adults with subjective mem-ory complaints. Neuroimaging biomarkers such as MRI,PET florbetapir, and FDG-PET are used in this trial. Othermultidomain trials in Europe include the Vitamin D3-Omega-3-Home Exercise-Healthy Ageing and LongevityTrial (DO HEALTH), which will test the effects of multido-main intervention on five primary end points, including therisk of cognitive decline; the Finnish Geriatric InterventionStudy to Prevent Cognitive Impairment and Disability (FIN-GER) trial; the Prevention of Dementia by Intensive Vascu-lar Care (PreDIVA) trial, which will test the efficacy ofinterventions aimed at cardiovascular risk factors in decreas-ing dementia risk; and the Healthy Aging Through InternetCounseling of the Elderly (HATICE) project, part of the Eu-ropean Commission’s Seventh Framework program, whichwill compare three ongoing clinical trials to identify the op-timal design for future dementia prevention studies.

4. Promise and limitations of biomarkers

Ten years ago, there was an expectation in the field thatbiomarkers would facilitate measurement of progressionand treatment response in AD. Thus far, however, bio-markers have proved most valuable for defining subject pop-ulations. Nonetheless, biomarkers have the potential to beextremely useful at all stages of drug development—first,to confirm that a test compound hits its target; second, to as-sess whether hitting the target alters the pathophysiology ofthe disease; and third, to determine whether altering thepathophysiology improves a person’s clinical status, or re-flects treatment response or side effects of new treatments[8]. Used in clinical trials, therefore, biomarkers offer thepotential for diagnosis, sample enrichment, characterizingthe mechanism of drug action, monitoring disease progres-sion, and assessing response to treatment. In addition, be-cause biomarker profiles reflect different stages of thepathogenic process, they can be used to select optimalsubjects for trials of different drugs and different forms ofdementia at different stages of disease [9].

Despite the massive amounts of data collected in the Alz-heimer’s Disease Neuroimaging Initiative (ADNI) and re-lated projects around the world, there is still much to learnabout how well biomarkers reflect pathology and how theyrelate to cognition and other clinical manifestations. Thereare also concerns that ADNI and other cohort data maynot be generalizable to the overall population because of se-lection bias. Standardization of biomarker tests and new bio-markers are also needed and underway [10–12].

Studies have begun to fit empiric data fromADNI and otherstudies onto the hypothetical model proposed by Jack and

B. Vellas et al. / Alzheimer’s & Dementia 9 (2013) 438–444442

colleagues [13]. For example, Bateman and colleagues [14]useddata from theDIANstudy tomap the trajectory of variousbiomarkers, beginning 25 years before expected symptom on-set, in subjectswith autosomal dominantmutations that ensuredevelopment of early-onset AD. Their study demonstrateda temporal ordering of biomarker changes as predicted, begin-ning with CSF Ab1–42 declining 25 years before onset. At15 years before onset, subjects showed amyloid depositionas measured by PET imaging along with increased CSF tauand hippocampal atrophy. This was followed by cerebral hy-pometabolism on FDG-PET and impaired episodic memory10 years before onset. Global cognitive impairment was seen5 years before onset, and individuals met diagnostic criteriaforADan average of 3 years after symptomonset.Whether in-dividuals with sporadic AD will follow similar patterns ofchange is not known. To round out knowledge of AD progres-sion, more data are needed in the oldest old, those older than80 years of age, who comprise the fastest growing subpopula-tion, and the subpopulationwith the highest prevalence ofAD.This population may represent a group with multiple patholo-gies and thus, different biomarker patterns.

ADNI data from noncognitively impaired participants whoconverted to mild cognitive impairment have revealed otherinteresting information about the use of biomarkers for pre-dicting conversion. The analyses of data from ADNI-GO(Grand Opportunities) and ADNI-2 have been particularly in-formative, because many of these participants had florbetapirPET scans at baseline. Because amyloid deposition is a rela-tively slow process, those who were amyloid positive in2012 were likely also amyloid positive in 2004. Thus, investi-gators can track the influence of amyloid status on the progres-sion of other biomarkers over time in many people withprodromal AD. These analyses have shown that amyloid pos-itivity is associatedwith a higher ADAS-cog score (i.e., worsecognition) in “normal” control participants, suggesting thatADAS-cog scores could be used as outcome measures in pro-dromal AD. In the MAPT trial, 36.7% were PET florbetapirpositive, and those individuals had more objective cognitiveimpairment than those without amyloid imaging.

More important, although amyloid PET scans or CSF am-yloid measurement may be useful for identifying subjectswho are highly likely to develop AD, the expense and lim-ited availability of PET scans and the difficulty of obtaininglumbar punctures limits their widespread application formost trial subjects in many multinational studies, and forgeneral clinical use. Thus, there have been efforts to developother tests that indicate amyloid positivity using techniquesother than PET. Hippocampal volume is, by itself, a poorpredictor of amyloid status, although a signature pattern ofregional brain atrophy including hippocampal atrophydoes associate with amyloid positivity, and this signal isstronger if APOE ε4 status is included in the model[15,16]. An even better indicator of amyloid status isobtained with arterial spin labeling MRI, which measuresregional cerebral blood flow; and the combination ofvolume and cerebral blood flow is even better yet.

Going forward, it will be important to classify each poten-tial biomarker with respect to its purpose, and to determine theprecision, replicability, sensitivity, specificity, and positive andnegative predictive values for each purpose. It is unlikely thata limited set of biomarkers reflecting the presence of AD pa-thology (e.g., Ab and tau) will be sufficient to detect and diag-nose the disease, determine disease severity and progression,and also serve as surrogate markers of treatment response.

5. Are targeted trials the answer?

The urgency to identify and develop effective treatmentsfor AD quickly and with fewer participants, and the poortrack record from recent randomized clinical trials, suggestthat trial designs must continue to evolve. Targeted trials,which have been used extensively in cancer clinical trials[17], may be especially useful in AD in which there is a greatdegree of heterogeneity and uncertainty. In AD, targeted tri-als might select the study population based on genotype (e.g.,APOE), cognitive status (e.g., MMSE), an AD biomarker(e.g., CSF Ab and tau, or amyloid PET), or another factorthat is thought to predict response to the new drug. A recentphase 3 clinical trial of bapineuzumab is an example of a tar-geted design that selected AD patients who were not carriersof the APOE ε4 allele, based on results of a phase 2 trial thatwere interpreted to suggest that noncarriers were responsiveto bapineuzumab whereas carriers were not. (The sponsorshedged their bet, however, by conducting a separate clinicaltrial that included only carriers of the APOE ε4 allele).

Two trial designs that have proved successful in cancertrials and may be useful in AD trials are a biomarker adap-tive threshold design [18] and an adaptive signature design[19]. The biomarker adaptive threshold design combinesa test for treatment effect in a heterogeneous group of partic-ipants with the establishment and validation of a threshold orcut point for a predictive (not prognostic) biomarker thatidentifies the sensitive subgroup in that population (e.g.,number of APOE ε4 alleles or MMSE score). In this design,a heterogeneous group is enrolled. If the null hypothesis isdisproved, the analysis moves on to test a particular bio-marker by looking only at those subjects who test positivefor that biomarker. By focusing on the biomarker of interest,this design lessens the likelihood of type 1 error.

The adaptive signature design combines the developmentof a classifier, such as a biomarker or biomarker signature,with an appropriately powered test for treatment effect. Thetreatment is compared with placebo in the unselected cohortand, if a significant effect is not seen, then results from the firsthalf of the participants are used to identify a classifier/bio-marker that predicts the response. Then, using this identifier,the treatment is compared with results from the placebo groupin the secondhalf of theparticipants.Both these designs enablethe evaluation of treatments in heterogeneous populations andallow for adaptation and improvement of the trial even as po-tential biomarkers for diagnosis, prognosis, and treatment pre-diction are being developed.

B. Vellas et al. / Alzheimer’s & Dementia 9 (2013) 438–444 443

Although targeted and adaptive designs have provedthemselves in other areas of therapy development, their effi-ciency depends on having effective drugs and accurate, in-formative, and precise biomarker assays. Moreover, ifbiomarkers are used to select subjects, the proposed treat-ment must be substantially more effective in the selectedgroup compared with the excluded group. However, testinglarge populations with drugs that only benefit small sub-groups is not a sustainable approach to drug development.

Adaptive or targeted trial designs can only go so far to-ward the goal of facilitating efficiency and improving thelikelihood of success. Other changes in trial design areneeded to address statistical and methodological issuesthat have contributed to recent preventive and treatment trialfailures. For example, in preventive trials such as Ginkgo Bi-loba Extract for the Prevention of AD (or GuidAge) [20], thefailure to demonstrate a protective effect may have resultedsimply from ineffective mechanisms or from methodologi-cal issues, including inadequate dose, insufficient durationof exposure, selection of an inappropriate population inwhich to test the drug, excessive number of dropouts, andthe use of inappropriate statistical tests to assess outcome.Biomarkers may enable more specific selection of appropri-ate populations, better targeting, identification of subjects atan earlier stage, and better assessment of treatment effects,but may also increase selection bias because of different per-ceptions of burden by different population groups. Similarly,innovative approaches that might reduce burden and mini-mize dropouts, such as increasing the use of Internet and dig-ital technologies, may be less acceptable in certainpopulation groups and thus could also affect selection bias.

6. Conclusions

Task force participants andothers in theADfield have beenencouraged by post hoc analyses in the solanezumab trial sug-gesting that targeting amyloidwith an antibody, at least duringthe early stages of AD, may provide some beneficial effect.The outcome of this trial and others provide lessons for futuretreatment trials, including DIAN, the Alzheimer’s PreventionInitiative, andA4, aswell as for futuremultidomain trials suchas MAPT, FINGER, DOHEALTH, Pre-DIVA, and HATICE.These lessons include the following:

� Secondary prevention studies as currently envisionedare conceptualized more appropriately as treatment tri-als of presymptomatic or subjects with mild cognitiveimpairment (i.e., when subjects are recruited within5 years of their expected onset of dementia). Largerphase 2 studies may be needed to predict phase 3 suc-cess more explicitly. Results must also be scrutinizedto ensure that the perceived benefit did not arise fromabnormal behavior of the placebo group. In the absenceof a clinical efficacy signal in phase 2, phase 3 studiescarry a considerable risk of failure. Thus, a single,combined phase 2/phase 3 study with initial efficacytesting at two or more promising doses in a moderately

sized study, and then moving directly into a pivotaltrial, may be worthy of consideration.

� Alternatively, clear evidence of target engagement inthe central nervous system and evidence of down-stream effects on relevant biomarkers may be estab-lished, as used in the solanezumab developmentprogram. This approach and combined phase 2/phase3 studies are not mutually exclusive, but may lead tomarkedly different durations of phase 2 development.

� New trial designs, including adaptive and targeted de-signs, should be explored.

� Disease severity may be a useful approach to stratifyparticipants who already have AD, or to focus the pa-tient population for trials.

� Stratification by APOE ε4 may be needed for safetyreasons (i.e., to reduce or evaluate of the risk ofARIA-E by APOE genotype).

� Amyloid positivity is a usefulmarker for identifying per-sons with amyloid pathology; however, both PET andlumbar puncture for CSF studies present problems thatlimit their widespread acceptability in global clinical tri-als. Thus, development of other measures as potentialsurrogates of amyloid positivity should be pursued.

� Although it may be useful to enrich trials with bio-markers or risk factors, it is important also to studysubjects who lack such markers to understand morefully the relationship among these biomarkers andrisk factors and cognitive decline.

� In the absence of reliable biomarker surrogates, verysensitive measures of cognition may be the most robustindicator of efficacy currently available for predictingtreatment benefit in phase 3 and for making clinical de-velopment decisions, particularly in presymptomaticpopulations.

� Harmonization is needed regarding the most appropri-ate cognitive end points for use in clinical trials.

� Improvements in trial design are needed to reduceselection bias and the number of dropouts, and toimprove validity.

Acknowledgments

Prof. Hampel was supported by the Katharina-Hardt-Found-ation, Bad Homburg, Germany.We thank Lisa J. Bain for ed-itorial assistance with this manuscript.

References

[1] Salloway S, Sperling R, Gilman S, Fox NC, Blennow K, Raskind M,

et al. A phase 2 multiple ascending dose trial of bapineuzumab in

mild to moderate Alzheimer disease. Neurology 2009;73:2061–70.

[2] Rinne JO, Brooks DJ, Rossor MN, Fox NC, Bullock R, Klunk WE,

et al. 11C-PiB PET assessment of change in fibrillar amyloid-beta

load in patients with Alzheimer’s disease treated with bapineuzumab:

a phase 2, double-blind, placebo-controlled, ascending-dose study.

Lancet Neurol 2010;9:363–72.

B. Vellas et al. / Alzheimer’s & Dementia 9 (2013) 438–444444

[3] Sperling R, Salloway S, Brooks DJ, Tampieri D, Barakos J, Fox NC,

et al. Amyloid-related imaging abnormalities in patients with Alz-

heimer’s disease treated with bapineuzumab: a retrospective analysis.

Lancet Neurol 2012;11:241–9.

[4] Dubois B, Feldman HH, Jacova C, Dekosky ST, Barberger-Gateau P,

Cummings J, et al. Research criteria for the diagnosis of Alzheimer’s

disease: revising the NINCDS-ADRDA criteria. Lancet Neurol 2007;

6:734–46.

[5] Sperling RA, Aisen PS, Beckett LA, Bennett DA, Craft S, Fagan AM,

et al. Toward defining the preclinical stages of Alzheimer’s disease:

recommendations from the National Institute on Aging and the Alz-

heimer’sAssociationWorkgroup.Alzheimers Dement 2011;7:280–92.

[6] Rowe CC, Ellis KA, Rimajova M, Bourgeat P, Pike KE, Jones G, et al.

Amyloid imaging results from theAustralian Imaging, Biomarkers and

Lifestyle (AIBL) study of aging. Neurobiol Aging 2010;31:1275–83.

[7] FolsteinMF, Folstein SE,McHughPR. “Mini-mental state”: a practical

method for grading the cognitive state of patients for the clinician. J

Psychiatry Res 1975;12:189–98.

[8] Hampel H, Frank R, Broich K, Teipel SJ, Katz RG, Hardy J, et al. Bio-

markers for Alzheimer’s disease: academic, industry and regulatory

perspectives. Nat Rev Drug Discov 2010;9:560–74.

[9] Blennow K. Biomarkers in Alzheimer’s disease drug development.

Nat Med 2010;16:1218–22.

[10] Frisoni GB, Jack CR. Harmonization of magnetic resonance-based

manual hippocampal segmentation: a mandatory step for wide clinical

use. Alzheimers Dement 2011;7:171–4.

[11] Mattsson N, Andreasson U, Carrillo MC, Persson S, Shaw LM,

Zegers I, et al. Proficiency testing programs for Alzheimer’s disease

cerebrospinal fluid biomarkers. Biomark Med 2012;6:401–7.

[12] Mattsson N, Zegers I, Andreasson U, Bjerke M, Blankenstein MA,

Bowser R, et al. Reference measurement procedures for Alz-

heimer’s disease cerebrospinal fluid biomarkers: definitions and ap-

proaches with focus on amyloid beta42. Biomark Med 2012;

6:409–17.

[13] Jack CR Jr, Knopman DS, Jagust WJ, Shaw LM, Aisen PS,

Weiner MW, et al. Hypothetical model of dynamic biomarkers

of the Alzheimer’s pathological cascade. Lancet Neurol 2010;

9:119–28.

[14] Bateman RJ, Xiong C, Benzinger TLS, Fagan AM, Goate A,

Fox NC, et al. Clinical, cognitive, and biomarker changes in the

Dominantly Inherited Alzheimer Network. N Engl J Med 2012;

367:795–804.

[15] Tosun D, Joshi S, Weiner M. Alzheimer’s Disease Neuroimaging Ini-

tiative: neuroimaging predictors of brain amyloidosis in mild cognitive

impairment. Ann Neurol 2013 [in press].

[16] Tosun D, Weiner M. Alzheimer’s Disease Neuroimaging Initiative:

neuromorphometry predictors of amyloid-beta positivity in individ-

uals with MCI. Paper presented at: Alzheimer’s Association Interna-

tional Conference, Vancouver, British Columbia, Canada, 2012.

[17] Maitournam A, Simon R. On the efficiency of targeted clinical trials.

Stat Med 2005;24:329–39.

[18] Jiang W, Freidlin B, Simon R. Biomarker-adaptive threshold design:

a procedure for evaluating treatment with possible biomarker-

defined subset effect. J Natl Cancer Inst 2007;99:1036–43.

[19] Freidlin B, Simon R. Adaptive signature design: an adaptive clin-

ical trial design for generating and prospectively testing a gene ex-

pression signature for sensitive patients. Clin Cancer Res 2005;

11:7872–8.

[20] Vellas B, Coley N, Ousset PJ, Berrut G, Dartigues JF, Dubois B, et al.

Long-term use of standardised ginkgo biloba extract for the prevention

of Alzheimer’s disease (GuidAge): a randomised placebo-controlled

trial. Lancet Neurol 2012;11:851–9.