BIOMARKERS IN BRAIN DISEASE

Biomarkers in Alzheimer’s Disease

Not Yet Surrogate Endpoints

N. Coley,a,b,c S. Andrieu,a,b,c J. Delrieu,c,d T. Voisin,a,c,d

and B. Vellasa,c,d

aINSERM U 558, Toulouse, FrancebUniversity of Toulouse III, Toulouse, France

cGerontopole, CHU Toulouse, FrancedAlzheimer’s Disease Clinical Research Center, CHU Toulouse, France

Clinical tests are currently used as endpoints in Alzheimer’s disease (AD) trials tomeasure disease progression based on cognitive, functional, or overall decline. Theseendpoints are not a perfect reflection of the underlying disease pathology and maybe insensitive to disease progression, especially in early AD. Furthermore, they aresubject to high variability, leading to large sample sizes and long trial durations. Abiomarker that could better reflect AD progression and also predict clinical benefits ofdrug treatments—a surrogate endpoint—would be of great use. Currently, no surrogateendpoints have been validated in AD. Structural imaging seems to be a better candidatethan plasma or CSF biomarkers, but is not yet validated as a surrogate endpoint. Moreprospective clinical trials are needed for the validation process. While AD biomarkerscannot currently be used as formal surrogate markers, they may nonetheless be usefulmeasures in clinical trials alongside clinical outcomes.

Key words: disease-modifying trials; biomarker; clinical trials; surrogate endpoint

Introduction

Biomarkers have several uses in clinical tri-als for Alzheimer’s disease (AD).1 For example,they may be used to select the target popula-tion, especially for trials aimed at patients withvery mild forms of AD. New disease-modifyingtreatments may be most effective in the earlierstages of the disease before neurodegenerationbecomes too widespread, therefore, it may beimportant to identify subjects at early stages ofthe disease. Biomarkers may also be used as pre-dictive markers of preclinical pathology, whichwould identify persons likely to develop clini-cally detectable disease in the future, or theycould be used for patient stratification, in or-

Address for correspondence: Prof. Bruno Vellas, Gerontopole/INSERM U 558, 170, avenue Casselardit, 31059 Toulouse Cedex 03,France. Voice: +33 5 61 77 76 49; fax: +33 5 61 49 71 09. vellas.b@chu-toulouse.fr

der to identify a subset of patients that a newdrug is expected to have the greatest effect.2

Additionally, non-AD-specific biomarkers mayalso be of use as safety measures, enablingthe detection of drug adverse events, such asinflammation.2

Perhaps more important, biomarkers maybe used as indicators of AD progression—inother words, as surrogate endpoints. In car-diovascular or cancer clinical trials, the pri-mary endpoint is typically the occurrence ofa distinct event, such as death, or recurrenceof a certain type of cancer or a cardiovascu-lar event. However, in AD, the use of mortal-ity endpoints is not feasible due to the verylarge sample sizes or follow-up periods thatwould be required, and there are no otherunequivocal events that could be measured.3

Therefore, current primary endpoints in ADtrials are based on clinical measures of cog-nitive or functional status as well as global

Biomarkers in Brain Disease: Ann. N.Y. Acad. Sci. 1180: 119–124 (2009).doi: 10.1111/j.1749-6632.2009.04947.x c© 2009 New York Academy of Sciences.

119

120 Annals of the New York Academy of Sciences

evaluations.4 These clinical tests are not a per-fect reflection of the underlying disease pathol-ogy, and, furthermore, they are subject to highvariability, leading to large sample sizes andmaking the detection of treatment effects dif-ficult, especially in disease modifying trials.5

Suitable surrogate biomarkers with less vari-ability would substantially increase the powerof a trial to detect disease-modifying therapeu-tic effects compared to standard clinical out-comes, thus leading to smaller sample sizes.Furthermore, short-term effects of a disease-modifying drug may be more easily detectablewith biomarkers than with clinical measuresalone, due to the slow progressive nature ofAD.2,6 Development of a biomarker surrogateendpoint should decrease the duration and sizeof future AD clinical trials.

Much of the current research into drugs forAD is focused on so-called disease-modifyingtreatments. An AD treatment that is diseasemodifying should be able to reduce the diseaseprogression rate, have an effect on the phys-iopathological mechanism of the disease, andshould have a long-lasting effect on disability.7

The approval of symptomatic treatments forAD has traditionally been based on cognitiveand other clinical assessments, but this maynot be enough for a claim of disease modifi-cation. In order to gain regulatory approval asa disease-modifying drug, strong biomarker ev-idence will also be required.7–9

How Can Biomarkers Be Validatedas Surrogate Endpoints?

According to the Biomarkers DefinitionsWorking Group, a surrogate endpoint is “abiomarker that is intended to substitute for aclinical endpoint” (p. 91).10 Such endpoints aretherefore “expected to predict clinical benefit(or harm, or lack of benefit) based on epidemi-ologic, therapeutic, pathophysiologic, or otherscientific evidence” (p. 91).10 The size of thepharmacodynamic effect on biomarkers thatmust be achieved to predict a clinical benefit in

AD is unknown.11 For the European MedicinesAgency (EMEA), if a biomarker is to be ac-cepted as a surrogate endpoint, it should satisfycertain criteria, including, though not limitedto, responding to treatment, predicting clinicalresponse, and being compellingly related to thepathophysiological process of the dementingcondition.8 In addition, biomarkers should ide-ally show a “dose-effect response” and shouldbe able to predict response to treatment.6 Itwould also be advantageous for clinical trialsif biomarkers were easily obtainable (ideallyblood/urine) and objectively measurable andshowed less variation than clinical measures.

Which Are the Main Candidatesfor Surrogate Endpoints in AD?

Structural imaging can be used as a markerof disease severity and progression. For exam-ple, progressive cerebral atrophy can be trackedusing magnetic resonance imaging (MRI). Hip-pocampal and whole brain atrophy rates are thetwo most established markers of disease pro-gression on MRI.12 If a treatment could com-pletely stop disease progression, then atrophyshould slow to normal aging rates: approxi-mately 1–2% per year in normal subjects agedbetween 70–80 years, in contrast with 4–6%per year in AD for hippocampal atrophy13 and0.5% per year in normal subjects aged around70 years, in contrast with 2% per year in AD forwhole brain atrophy.14 Although a beneficial ef-fect on clinical outcomes remains paramount,MRI markers of disease progression might pro-vide evidence of disease slowing earlier or morecost-effectively than clinical outcomes.15 Suchmarkers of disease severity and progression maybe highly nonspecific (not relate to pathologi-cal progress) and yet still retain utility.1 Wholebrain atrophy rate seems to be the marker ofchoice at this time. It is a robust measure thatdiscriminates between groups and is related toclinical progression. Hippocampal atrophy per-forms better as a diagnostic marker and possi-bility as a marker of progression in the earliest

Coley et al.: Biomarkers as Surrogate Endpoints in AD 121

disease stage.16 For imaging biomarkers to beaccepted as surrogate outcomes in AD, multipleresults from disease-modifying trials supportingtheir validity will be required. Structural imag-ing also has the potential to greatly improve thepower of clinical trials.1

Specific markers of disease activity, such asplasma and cerebrospinal fluid (CSF) biomark-ers (concentrations of Aβ1–42, Aβ1–40, andtau) and molecular imaging markers (amyloidplaques), may also be important for diseasemodification trials. It would clearly be valu-able if a disease process could be sensitivelyand specifically measured in a trial of a drugdesigned to change this process.1 These mark-ers open new unique possibilities for evalua-tion of drug mechanisms and treatment efficacyin AD.17 However, these markers are not themarkers of choice for monitoring disease pro-gression. Indeed, CSF biomarkers cannot yettrack the naturalistic disease course, and amy-loid imaging reflects an on/off phenomenon. Itmay be difficult to demonstrate modification ofthe disease course with these markers, but theymay have some use in demonstrating a drugeffect on a relevant disease process.

Do Recent Trials Provide AnyEvidence for Claimsof Surrogacy Status?

In a 24-week Phase II trial of a tau ag-gregator inhibitor (TAI) in mild to moderateAD patients, using region of interest (ROI) andstatistical parametric mapping (SPM) analysesof single photon emission computed tomogra-phy (SPECT) data, it was demonstrated thatTAI therapy (remberTM; TauRx Therapeutics,Singapore) prevented the significant reductionin regional cerebral blood flow (rCBF) seen inthe placebo group; those treated with the drugshowed no clinical decline over 24 weeks (intotal, 135 subjects were imaged twice).18 Thegreatest treatment effect was demonstrated inthe brain regions characterized by the most se-vere tau aggregation pathology. It is not cur-

rently possible to directly measure the num-ber and extent of neurofibrillary tangles in thebrain in vivo, but rCBF changes demonstratedby SPECT have been shown to correlate withBraak pathological stage.19 Thus, using molec-ular brain imaging as an indirect measure ofneuronal function seems to indicate that TAImay have a disease- modifying effect. There issome evidence that this therapy affects clinicaloutcomes. 20

Both CSF and plasma biomarkers were as-sessed as secondary outcomes in a 12-weekPhase II trial in early AD patients of PBT2, ametal-protein attenuating compound thoughtto affect copper- and zinc-mediated oligomer-ization of Aβ.21 In the high-dose PBT2 group(250 mg) there was a significant reduction inCSF Aβ42 compared to placebo, with a sig-nificant dose effect. Treatment had no effecton CSF Aβ40, T-tau, or P-tau levels or on anyplasma biomarkers. While these results providesome evidence of the proposed mode of ac-tion of PBT2, there was no correlation betweenclinical response and CSF biomarker outcome.However, this was a short-term trial carried outin mild patients meaning that there may havebeen relatively little cognitive change duringthis period.

Other Phase II trials involving a gammasecretase inhibitor (semagacestat, LY450139)and an Aβ monoclonal antibody (solanezumab,LY2062430) have also demonstrated signifi-cant changes to plasma and CSF Aβ levels,but with no clear associated change in clinicaloutcomes.9,22 The lack of correlation betweenchanges in biomarkers and clinical scores inthese Phase 2 studies is not necessarily surpris-ing, given the time course of the biomarkerchanges and the intended clinical effects ofdrug.9

In the AN1792 trial of Aβ immunization inAD patients, changes in CSF biomarkers wereused as secondary outcomes in addition to neu-ropsychological tests and imaging. After 1 yearof follow-up, there was no effect of treatment onCSF Aβ42, but there was a significantly greaterreduction in CSF T-tau in antibody responders

122 Annals of the New York Academy of Sciences

compared to the placebo group.23 There wasno clear associated change in cognitive func-tion. However, only 21 patients were includedin the CSF biomarker analysis (11 antibody re-sponders and 10 placebo recipients), and themajority of patients in this trial had only re-ceived two injections before treatment was sus-pended, making it difficult to draw conclusions.

The trial of tramiprosate (AlzheimedTM;Neurochem, Quebec, Canada) is the first com-pleted Phase III study of an anti-amyloid com-pound for the treatment of AD. As part of thistrial, an MRI substudy was carried out in ap-proximately half of the patients.11 It was there-fore possible to study the feasibility, sensitiv-ity, and clinical relevance of volumetric MRI(vMRI) as a disease modification endpoint in ananti-amyloid AD trial. In the MRI subset of pa-tients (n = 508), the mean change from baselinein the placebo group was 7.8 ± 9.2 pointsfor the Alzheimer’s Disease Assessment Scale,cognitive subscale (ADAS-cog) and 2.0 ± 2.6points for the Clinical Dementia Rating-Sumof Boxes (CDR-SB) at week 78. The meanchange from base line in hippocampal vol-ume was a decrease of 192.2 ± 223.0 mm3.At base line, there was evidence of a correla-tion between CDR-SB score and hippocam-pal volume (−0.25, P < 0.05), but there wasno significant correlation between ADAS-cogscore and hippocampal volume (−0.16, P =0.11). Correlations between the change frombase line ADAS-cog and CDR-SB scores andbaseline hippocampal volume were −0.17 (P =0.07) and 0.08 (P = 0.43), respectively. Thisfinding of a correlation approaching statisticalsignificance between the change in ADAS-cogscore and hippocampal volume change addssome support to the clinical validity of thevMRI biomarker and its usefulness to moni-tor disease progression. However, the magni-tude of the correlation was relatively low, per-haps suggesting that these measures may notreflect the same disease processes. For exam-ple, hippocampal volume may be involved inmemory and learning, and while the ADAS-cogdoes measure memory, it also measures multi-

ple other cognitive skills that are presumably af-fected by various regions of the brain. Further-more, hippocampal volume may affect mem-ory decline before any effects are detectable onthe ADAS-cog.11 This raises the point that wemay be searching for a correlation between twoevaluations that are not necessarily measuringthe same thing. Thus, vMRI may potentiallybe able to demonstrate treatment effects earlierin the disease relative to clinical cognitive end-points, and therefore correlations between thetwo may be weak at best. This study demon-strated the feasibility of using vMRI with a rel-atively large number of patients in an AD trials.However, it should be noted that there was asubstantial amount of measurement variability,perhaps due to the number of vMRI sites used.

Recent AD clinical trials have not providedany evidence that plasma or CSF biomarkersmight be useful as surrogate endpoints as therehave been no correlations between biomarkeroutcomes and clinical outcomes. On the otherhand, there is some evidence that certain imag-ing techniques may be of use, although furtherstudy is required.

Are Any Biomarkers Close to BeingValidated as Surrogate

Endpoints in AD?

A schema evaluating biomarkers as surro-gate endpoints across several domains was re-cently proposed by Lassere.24 The Quantita-tive Surrogate Validation Level of EvidenceSchema (QSVLES) is a method that could beused to evaluate the validity of biomarkers thatare currently being proposed as potential sur-rogate endpoints in AD clinical trials.6 Thisschema evaluates four different domains foreach proposed biomarker: (i) target, (ii) studydesign, (iii) statistical strength, and (iv) penal-ties. The “target” indicates the clinical endpointthat the biomarker is supposed to be substitut-ing, and the “study design” refers to the typesof study design that have been used to inves-tigate or validate the surrogacy status of the

Coley et al.: Biomarkers as Surrogate Endpoints in AD 123

TABLE 1. Scores of AD Biomarkers According to the QSVLES System for Validation of SurrogateEndpoints

Domain Hippocampal volume FDG-PET SPECT-rCBF Aβ Tau

Target 3 3 3 3 3Study design 2 2 0 0 0Statistical strength 3 3 0 0 0Penalties −4 −4 −1 −1 −1Total score 4 4 5 2 2LEVEL 4 4 5 5 5

Adapted from Ref. 6.FDG-PET: 18 flurodeoxyglucose-positron emission tomography.SPECT-rCBF: Single photon emission computed tomography-regional cerebral blood flow.Level = level of evidence supporting the validity of the biomarker as a surrogate. Levels 1 and 2 are considered

acceptable for true surrogates.

biomarker. “Statistical strength” refers to thestatistical significance and the strength of theassociation between the proposed biomarkerand the target, and “penalties” concern incon-sistent or conflicting evidence or a lack of evi-dence. Several criteria are assessed within eachdomain, leading to the assignment of a scoreranging from 0 to 5 for each domain. The scoresfor each domain are then combined to give anoverall level of evidence ranging from 1 to 5.Level 1 reflects the strongest evidence of valid-ity and level 5 the weakest evidence of validity.It is suggested that the term “surrogate” shouldonly be assigned to biomarkers obtaining levels1 and 2.24 In an evaluation of AD biomark-ers proposed as surrogate endpoints, MRI andpositron emission tomography (PET) measuresobtained level 4 status, and CBF (SPECT), Aβ,and tau were assigned to level 5 (Table 1).6

Other AD biomarkers assessed also fell intolevels 4 or 5. Thus, according to the QSVLEScriteria, none of these biomarkers can be con-sidered as surrogate endpoints as none of themwere ranked in levels 1 or 2. This was mainlyattributed to the lack of sufficient evidence thatwas derived from high-ranking studies; in theQSVLES system, high scores can only be ob-tained if evidence is gained from randomizedcontrol trials, and less emphasis is given toprospective observational studies.6

It is clear that the potential risks of relyingon invalid surrogate endpoints should not be

underestimated. Therefore, the development ofspecific guidance for the validation of biomark-ers as surrogate endpoints in AD should beencouraged. Despite being relatively new andas yet not widely used, the QSVLES may beproposed as a starting point for discussions.

Conclusion

At the current time, biomarkers may be use-ful for the diagnosis of AD and selection ofpatient populations for clinical trials, but thereare no validated biomarkers suitable for useas surrogate endpoint in AD trials. Whereasthere is no specific guidance for the validationof a SEP, it is recognized as an important re-quirement that changes of the surrogate markershould correspond to changes in clinical out-come and the magnitude of the change relatedto the intervention. Imaging markers, notablyMRI, seem to be more correlated with clini-cal outcomes than CSF or plasma markers andtherefore may be more easily developed as sur-rogate markers. Interpreting and validating theresults of biomarker studies in AD trials will bea challenge for the coming years and ultimatelywill require careful correlation with long-termclinical outcomes. The ideal setting to demon-strate this is the prospective randomized clinicaltrial. While AD biomarkers cannot currentlybe used as formal surrogate markers, they may

124 Annals of the New York Academy of Sciences

nonetheless be useful measures in clinical tri-als alongside, rather than as a replacement for,clinical outcomes.

Conflicts of Interest

The authors declare no conflicts of interest.

References

1. Lovestone, S. & M. Thambisetty. 2009. Biomarkersfor Alzheimer’s disease trials—biomarkers for what?A discussion paper. J. Nutr. Health Aging 13: 334–336.

2. Blennow, K. & H. Zetterberg. 2009. Use of CSFbiomarkers in Alzheimer’s disease clinical trials. J.

Nutr. Health Aging 13: 358–361.3. Thomas, R.G., J.D. Berg, M. Sano & L. Thal. 2000.

Analysis of longitudinal data in an Alzheimer’s dis-ease clinical trial. Stat Med. 19: 1433–1440.

4. Vellas, B., S. Andrieu, C. Sampaio, et al. 2008. End-points for trials in Alzheimer’s disease: a Europeantask force consensus. Lancet Neurol. 7: 436–450.

5. Aisen, P.S. 2009. Interpreting biomarker data in ther-apeutic trials. J. Nutr. Health Aging 13: 337–338.

6. Gispen-de Wied, C., M. Kritsidima & A. Elferink.2009. The validity of biomarkers as surrogate end-points in Alzheimer’s disease by means of theQuantitative Surrogate Validation Level of EvidenceScheme (QSVLES). J. Nutr. Health Aging 13: 376–387.

7. Vellas, B., S. Andrieu, C. Sampaio & G. Wilcock.2007. Disease-modifying trials in Alzheimer’s dis-ease: a European task force consensus. Lancet Neurol.

6: 56–62.8. EMEA. Committee for Medicinal Products for Hu-

man Use (CHMP). 2008. Guideline on medicinalproducts for the treatment of Alzheimer’s disease andother dementias. CPMP/EWP/553/95 95 Rev. 1.

9. Siemers, E.R. 2009. How can we recognize “diseasemodification” effects? J. Nutr. Health Aging 13: 341–343.

10. Biomarkers Definitions Working Group. 2001.Biomarkers and surrogate endpoints: preferred def-initions and conceptual framework. Clin. Pharmacol.

Ther. 69: 89–95.11. Saumier, D., P.S. Aisen, S. Gauthier, et al. 2009.

Lessons learned in the use of volumetric MRIin therapeutic trials in Alzheimer’s disease: theAlzhemedTM (Tramiprosate) experience. J. Nutr.

Health Aging 13: 370–372.12. Jack, C.R., Jr., M.M. Shiung, J.L. Gunter, et al. 2004.

Comparison of different MRI brain atrophy ratemeasures with clinical disease progression in AD.Neurology 62: 591–600.

13. Schott, J.M., S.L. Price, C. Frost, et al. 2005. Measur-ing atrophy in Alzheimer disease: a serial MRI studyover 6 and 12 months. Neurology 65: 119–124.

14. Fox, N.C., R.S. Black, S. Gilman, et al. 2005. Effectsof Abeta immunization (AN1792) on MRI measuresof cerebral volume in Alzheimer disease. Neurology

64: 1563–1572.15. Fox, N.C. & J. Kenedy. 2009. Structural imaging

markers for therapeutic trials in Alzheimer’s disease.J. Nutr. Health Aging 13: 350–352.

16. Van Der Flier, W.M. & P. Scheltens. 2009. Knowingthe natural course of biomarkers in AD: longitudinalMRI, CSF and PET data. J. Nutr. Health Aging 13:353–355.

17. Nordberg, A. 2009. Molecular imaging markers inclinical trials in Alzheimer’s disease. J. Nutr. Health

Aging 13: 346–347.18. Wischik, C. & R. Staff. 2009. Challenges in the

conduct of disease-modifying trials in AD: practicalexperience from a Phase 2 trial of tau-aggregationinhibitor therapy. J. Nutr. Health Aging 13: 367–369.

19. Jobst, K.A., A.D. Smith, C.S. Barker, et al. 1992. As-sociation of atrophy of the medial temporal lobe withreduced blood flow in the posterior parietotempo-ral cortex in patients with a clinical and pathologicaldiagnosis of Alzheimer’s disease. J. Neurol. Neurosurg.

Psychiatry 55: 190–194.20. Wischik, C.M., P. Bentham, D.J. Wischik & K. Meng

Seng. 2008. Tau aggregation inhibitor (TAI) therapywith rember TM arrests disease progression in mildand moderate Alzheimer’s disease over 50 weeks.Alzheimer’s Dementia 4(Suppl 2): T167.

21. Lannfelt, L., K. Blennow, H. Zetterberg, et al.2008. Safety, efficacy, and biomarker findings ofPBT2 in targeting Abeta as a modifying therapyfor Alzheimer’s disease: a Phase IIa, double-blind,randomised, placebo-controlled trial. Lancet Neurol. 7:779–786.

22. Siemers, E.R., S. Friedrich, R.A. Dean, et al. 2008.Safety, tolerability and biomarker effects of an Aβ

monoclonal antibody administered to patients withAlzheimer’s disease. Alzheimer’s Dementia 4(Suppl 2):T774.

23. Gilman S, Koller M, Black RS, et al. 2005. Clinicaleffects of Abeta immunization (AN1792) in patientswith AD in an interrupted trial. Neurology 64: 1553–1562.

24. Lassere, M.N. 2008. The Biomarker-Surrogacy Eval-uation Schema: a review of the biomarker-surrogateliterature and a proposal for a criterion-based, quan-titative, multidimensional hierarchical levels of evi-dence schema for evaluating the status of biomarkersas surrogate endpoints. Stat. Methods Med. Res. 17:303–340.