Amélioration des méthodes d’évaluation de la prévention en

population âgée

Nicola ColeyCERPOP

Equipe Vieillissement – Axe MAINTAIN

8,7 millions de cas de demence en Europe� La maladie d’Alzheimer (MA) représente 50 à 70% des cas

� Augmentation exponentielle de la prévalence avec l’âge

� 46,8 millions de cas dans le monde en 2015 → 131 millions en 2050

Winblad B et al. Lancet Neurol. 2016

(ADI World Alzheimer Report 2015)

� 7ème cause de mortalité mondiale en 2019 (OMS)

La maladie d’Alzheimer (MA)

Maladie neurodégénérative progressive

Atrophie cérébrale, plaques amyloïdes, dégénérescence neurofibrillaire

Perte progressive des fonctions cognitives (notamment mémoire)

Limitations dans les activités (instrumentales) de la vie quotidienne -> perte d’autonomie

Troubles d’humeur et comportement

Aucun traitement curatif

Longue période de latence (stade pré clinique): apparition des altérations physiopathologiques bien avant les symptômes

cliniques

Jack CR Jr. Lancet Neurol. 2013;12(2):207-16

Prévention

possible?

Timeline of AD research milestones

1906

Alois

Alzheimer

first

describes

the disease

1968

Development

of cognitive

measurement

scales

1984-6

Beta-

amyloid

& tau

identified

1987

First AD

drug trial

1993

First

symptomatic

treatment

approved

1998

First

prevention

trials

(inc.MCI)

2011

New AD

diagnostic

criteria

incorporating

biomarkers

2002

Last

symptomatic

treatment

approved

Many failed phase 2-3

trials

Vers la prévention

Revue systématique et méta analyses

Livingston G et al. Dementia prevention, intervention, and care. Lancet. 2020

• Niveau d’éducation (7%)

• Perte d’audition (8%)

• Lésions cérébrales traumatiques (3%)

• Hypertension (2%)

• Consomption élevée d’alcool (1%)

• Obésité (1%)

• Tabagisme (5%)

• Dépression (4%)

• Isolement sociale (4%)

• Sédentarité (2%)

• Pollution de l’air (2%)

• Diabète (1%)

40% des cas de démence attribuables à 12

facteurs de risque potentiellement modifiables :Enfance

Milieu de vie

Age avancée

Réflexion méthodologique:How to measure intervention effects in the new generation of 1° and 2° prevention trials?

Symtomatic AD trials

Co-primary outcome approach using gold-standard measures of cognition and function (lack of « hard » clinical endpoints)

Secondary prevention trials (MCI/prodromal AD)

Same co-primary outcome approach as in symptomatic trials?

Conversion to dementia?

Biomarkers?

Primary prevention trials

Biomarkers?

Incident dementia?

Cognitive decline?

• Power

issues

• Limited

functional

decline

• Lack of

standardisation/

validation

• Cost

• Invasiveness

• Sensitivity to

change?

• Clinical

relevance?

Towards validating new endpoints

Suitability of a single primary endpoint measuring both cognition and function in secondary prevention/early AD trials

Determining the clinical relevance of a cognitive composite score for primary prevention trials

Utility of dementia risk scores as primary outcome measures for early prevention trials

CDR-sum of boxes: scores range from 0-18

Cognitive

items

Functional

items

Tractenberg, J

Neuropsych Clinl

Neurosci 2005Hughes, Br J Psychiatry 1982; Morris, Neurology 1993

CDR-SB as a single primary endpoint

Analysis of 667 subjects with very early or mild-moderate AD from the REAL.FR study

CDR-SB measures

cognition & function

simultaneously

Similar rates of

change across

different disease

stages, with limited

floor/ceiling effects

Low variability in

change over time:

smaller trial sample

sizes

Little missing data

CDR-SB responsiveness & consequences for trial design

Coley N, Andrieu S, Jaros M, Weiner M, Cedarbaum J, Vellas B.

Alzheimers Dement. 2011;7(6):602-610

Impact

Replication of results in the US ADNI studyCedarbaum JM, Jaros M, Hernandez C, Coley N, Andrieu S, Grundman M, Vellas B. AlzheimersDement. 2013; 9(1 Suppl):S45-55

EMA and FDA guidance recommend single primary endpoint measuring cognition and function simultaneously for prodromal AD trials, citing both studies

Food and Drug Administration. Guidance for Industry. Alzheimer's disease: developing drugs for the treatment of early stage disease, 2013.European Medicines Agency, Committee for Medicinal Products for Human Use (CHMP). Guideline on the clinical investigation of medicines for the treatment of Alzheimer’s disease, 2018.

CDR-SB now widely used as the sole clinical primary outcome measure in prodromal and mild dementia AD trials (approx. 20 trials registered in clinicaltrials.gov)

Outcomes in prevention trials

Type de critère de jugement principal:

Incidence de démence/MA

Déclin cognitif/performance cognitive

Biomarqueur

Chronologie des essais de prévention publiésType de critère de jugement et

résultat principal

Intervention pharmacologique non spécifique

Intervention pharmacologique spécifique

Intervention nutritionnelle

Entrainement cognitif

Exercice physique

Intervention multidomaine

Andrieu S, Coley N et al. Lancet Neurol. 2015

Revue de la littérature en 2015

Essais de prévention (randomisés et contrôlés)- de la MA- de la démence de toute

cause- du déclin cognitif

Analysis of the primary endpoints usedin the 51 prevention trials11 measured incident dementia

36 measured cognitive function/decline

Global cognition

Global test (e.g. MMSE)

Composite measure

Domain-specific cognitive function

4 measured biomarkers

• Most studies used their own individual

battery of tests

• Clinical relevance of most results unclear

Andrieu S, Coley N et al. Lancet Neurol. 2015

Composite cognitive scores• Increasingly popular primary outcome measures for

preclinical/prevention trials

• Combination of several sensitive cognitive tests into a single score

– More sensitive to change than individual tests/existing global cognitive scores

• Theory-driven and data-driven development suggest similar important domains

– episodic memory, orientation, executive function, language

15

Composite cognitive scores as primary outcome

measures• Supported by regulatory authorities

– “Alternatives to the ADAS like the NTB, the CDR-SB, the recently

proposed ADCS-PACC or others have been validated and may be used” EMA discussion paper on the clinical investigation of medicines for the treatment of Alzheimer´s disease and other dementias

• But not yet fully validated…

– Clinical relevance of changes on composite measures not well understood

• “The clinical relevance, the functional and long-term consequences of the subtle cognitive

changes measured by ADCS-PACC in earlier disease stages is currently unknown” (EMA)

– Long-term trajectories of cognitive decline not yet studied in treatment trials in preclinical/prevention settings 16

The GuidAge study• Multicentre, 5-year, randomised, double-blind, placebo-controlled trial of

Ginkgo biloba for AD dementia prevention

• 1414 placebo group participants included in this analysis– Age >70

– Spontaneously reported memory complaints

– MMSE>25

– CDR<1

– Free of dementia & major objective memory impairment (FCSRT score>10th percentile for age, sex &

sociocultural level)

– Free of major anxiety & depression (COVI<6, GDS<15)

• Cognitive function assessed annually in expert memory centres– Primary outcome: AD incidence (DSM-IV, NINCDS-ADRDA)

– Secondary outcome: cognitive decline (MMSE, FCSRT, TMT, verbal fluency)

Vellas B, Lancet Neurology 2012; Andrieu S, Curr Alz Res 200817

Calculation of the composite score

Spatial and temporal orientation: 10 MMSE orientation items

Episodic memory: FCSRT (free + total recall)

Language/semantic memory: Category fluency

Executive function: TMT-B

Composite score = (Zorientation + Zepidosic memory + Zverbal fluency + Zexecutive function)/4

18

Composite score change from baseline

-0.6

-0.4

-0.2

0.0

0.2

0 1 2 3 4 5

Co

mp

osi

te s

core

ch

an

ge

fro

m b

ase

lin

e

≤75 years

(N=657)

76-80 years

(N=504)

81-85 years

(N=218)

>85 years

(N=35)*

(76-80 & 81-85

vs. ≤75y)

By age (p<0.001)

*(all groups vs.

≤75y)

19

By ApoE genotype (p=0.001)

-0.6

-0.4

-0.2

0.0

0.2

0 1 2 3 4 5

Co

mp

osi

te s

core

ch

an

ge

fro

m b

ase

lin

e

ApoE Ɛ4-ve

(N=733)

ApoE Ɛ4+ve

(N=242)

*

*

By AD dementia status during follow-up (p<0.001)

-2.5

-2

-1.5

-1

-0.5

0

0.5

0 1 2 3 4 5

Time (years)

No

dementia

(N=1316)

AD

dementia

(N=73)

*

*

*

**

5-year composite score change from baseline

differed significantly by:

• Age

• ApoE genotype

• CDR progression status during follow-up

• AD dementia status during follow-up

2-year practice effects were observed in nearly all

subgroups

Coley N, Gallini A, Ousset PJ, Vellas B, Andrieu S. Alzheimers Dement. 2016;12(12):1216-1225

Prediction of future AD dementia

HR represents 5-year risk of AD dementia associated with a 1 SD decrease in

cognitive performance at baseline

– Estimated from Cox proportional hazards models

• Adjusted for age, sex, education

• Separate model for each cognitive test

Composite score MMSE CDR-SB

HR [95%CI] p HR [95%CI] p HR [95%CI] p

3.5 [2.6-4.7] <0.001 1.6 [1.3-2.0] <0.001 1.9 [1.6-2.2] <0.001

Study population = placebo group subjects, excluding those who developed non-AD dementia 20

Minimal Clinically Important Difference (MCID)

• To determine the smallest difference on the composite

measure that is clinically important

• Compare changes in the composite score across groups of

individuals known to have small but real differences in their

health status (anchor measures)

– What is a small but “clinically important” change in this

population?

• Progression from CDR 0 to 0.5

• Loss of independence in one IADL– Measured over 1 year (shortest measurable period in this study)

– “Baseline” for this analysis = 2 year visit (to avoid practice effects) 21

MCID: anchor-based 1-year mean change• 1st MCID estimate = difference in 1-year mean change on composite

in subjects who did and did not undergo change on anchor measures

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

2 3

Me

an

1-y

ea

r co

mp

osi

te s

core

ch

an

ge

Stable CDR 0

(N= 478)

CDR 0→0.5

(N=63)-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

2 3

No IADL decline

(N=749)

Loss of

independence in 1

IADL (N=35)

MCID = -0.28

(p<0.001)

MCID = -0.05

(p=0.484)

CDR anchor IADL anchor

22

Coley N, Gallini A, Ousset PJ, Vellas B, Andrieu S. Alzheimers Dement. 2016;12(12):1216-1225

MCID: anchor-based ROC curve analysis

• To describe the performance of composite measure change scores in discriminating between participants who do/do not undergo change on anchor measures– AUC = 0.7-0.8: acceptable discrimination

– AUC = 0.8-0.9: excellent discrimination

– AUC >0.9: outstanding discrimination

• Optimal cut-off determined using Youden Index (max. Sp & Se)

= 2nd MCID estimate

CDR anchor

(N=541, 63 decliners)

IADL anchor

(N=784; 35 decliners)

AUC

(95%CI)

0.7

(0.6-0.8)

0.5

(0.4-0.6)

Optimal cut-off ≤ -0.3 ≤ -0.3

Sensitivity 57.1% 37.1%

Specificity 80.3% 72.9%

23

Summary

• Baseline composite score was highly predictive of 5-year AD

dementia risk in a prevention trial setting

• 5-year composite score change from baseline differed

significantly by:

– Age

– ApoE genotype

– CDR progression status during follow-up

– AD dementia status during follow-up

• 2-year practice effects were observed in nearly all subgroups

• A first estimation of the MCID on this composite measure is -0.3

points (over 1 year)24

Dementia risk scores as surrogate outcomes –rationale, preliminary evidence and challenges

40% des cas de démences potentiellement attribuables aux facteurs modifiables

Comment mesurer l’effet d’une intervention qui débute dès la période « midlife » ?

Scores de risque de démence comme critères de substitution?

Sensibles au changement?

Capables de détecter l’effet d’une intervention?

Coley N, Hoevenaar-Blom M et al. Alzheimers Dement. 2020;16(12):1674-1685

CriteriaSubjective rating/5

CAIDE LIBRA ANU-ADRI BDSI DRS FDRS

Dementia prediction based on midlife data 5 4 4 0 1 1

External validation 5 2 3 2 1 0

Overall predictive accuracy 2 1 3 4 4 0

Importance of modifiable factors in total score 3 4 3 1 2 1

Validation as an RCT outcome measure 0 0 0 0 0 0

Use in RCTs as an outcome measure 3 2 4 0 0 0

Average rating 3 2 2 1 1 0

Candidate risk scores

ANU-ADRI: Australian National University AD Risk Index; BDSI: Brief Dementia Screening Indictor; CAIDE:

Cardiovascular Risk Factors, Aging and Incidence of Dementia; DRS: Dementia Risk Score; FDRS: Framingham

Dementia Risk score; LIBRA: Lifestyle for Brain Health

CAIDE LIBRA

Score MAPT PREDIVA HATICE Score MAPT PREDIVA HATICE

Age 0/3/4

Education 0/2/3

Sex 0/1

Blood pressure 0/2 0/1.6

Cholesterol 0/2 0/1.4

Diabetes 0/1.3 (I)

Renal dysfunction 0/1.1 -

CHD 0/1.0 -

Depression 0/2.1

BMI 0/2 0/1.6

Physical activity 0/1 0/1.1

Smoking 0/1.5 (I)

Alcohol -1/0 (I)

Medi. Diet 0/1.7 - -

Cognitive activity -3.2/0 - -

MAX SCORE 15 14.4 11.7(I)/ 8.9 12.7 13.3

Score components and data availability

No

n-

mo

dif

iab

leM

od

ifia

ble

Suitability of scoring systems for detecting

change?• CAIDE and LIBRA attribute points using categorical scoring

systems based on underlying prediction models

– E.g. CAIDE: SBP ≤140 = 0 points; SBP >140 = 2 points

• Large changes in individual risk factors (and dementia risk) may

not be registered if they do not cross the categorical cut off

points

• Proposal for alternative scoring systems:

– Z-scores

– Weighted z-scores

MAPT (FR) preDIVA (NL) HATICE (NL, FI, FR)

Sample size 1679 3526 2724

Intervention and

control conditions

• Multidomain intervention (cognitive training,

physical activity, nutrition counselling,

preventive consultation) + placebo

• Multidomain intervention + omega-3

supplement

• Omega-3 supplement

• Placebo

• Nurse-led

multidomain

cardiovascular care

• Usual care

• Coach-supported

interactive internet

platform to encourage self-

management of CV risk

factors & lifestyle change

• Static internet platform

Intervention duration 3 years 6-8 years 1.5 years

Primary outcome Cognitive function Dementia, disability Composite score of BMI, SBP

and cholesterol

Main secondary

outcomes

Functional status, physical status, depression,

dementia, health resource utilization

Cardiovascular events,

change in cognitive

function, depression

Cardiovascular events, lifestyle

parameters, cognitive

functioning

Main eligibility criteria - Age 70+

- Spontaneous memory complaint, and/or

limitations in one instrumental activity of daily

living, and/or slow gait speed

- Free of dementia

- Age 70-78

- Free of dementia

- Age 65+

- ≥2 CV risk factors and/or

history of CVD

- Computer literate

- Free of dementia

Datasets: main trial design features

Coley N, Hoevenaar-Blom M et al. Alzheimers Dement. 2020

Change over time

(up to 2 years FUP)

• 48 - 58% of intervention group participants

underwent change on the CAIDE score

• 30-31% decreased score (i.e. reduced dementia

risk)

• 18-27% increased score

• Up to 79% of participants changed over time on LIBRA

• Similar proportions increased (32-42%) and

decreased (29-43%) in score

Coley N, Hoevenaar-Blom M et al. Alzheimers Dement. 2020

Change in original score vs.

change in z-core version

• Z-score versions able to detect changes in risk factors

that were not registered by the original scores

(especially for CAIDE)

• Change of -1.25 to 1.25 points on z-score version

for subjects with zero change on original version

• Original versions changed by as much as +/- 5 points

for individuals showing no change on z-score versions

←amélioration du risque de démence

-0.8 -0.6 -0.4 -0.2 0.0 0.2

-0.3 -0.2 -0.1 0.0 0.1

-0.8 -0.6 -0.4 -0.2 0.0 0.2

-0.3 -0.2 -0.1 0.0 0.1

Différence d’évolution (IC95%) du score de risque entre groupe intervention et

contrôle

CAIDE score

original

CAIDE version z-

score

LIBRA score

original

LIBRA version

z-score

MAPT preDIVA HATICE Simulation effet « cliniquement important »

Conclusions

• Scores CAIDE et LIBRA (version d’origine ou version modifiée)

potentiellement utiles comme critère de substitution

• sensibles au changements

• capables de détecter un effet modeste, mais statistiquement

significatif, d’une intervention multidomaine

• Impact réel d’un effet modeste à court terme sur un score de

risque sur l’incidence de démence à long terme?

RemerciementsCDR-SB analysis

INSERM-Univ Toulouse

Sandrine Andrieu

Bruno Vellas

UCSF

Mike Weiner

Elan Pharmaceuticals

Jesse Cedarbaum

Mark Jaros

REAL.FR study investigators &

funders (PHRC)

Composite score analysis

INSERM-Univ Toulouse

Sandrine Andrieu

Adeline Gallini

Pierre-Jean Ousset

Bruno Vellas

GuidAge study group & funders

(Ipsen)

Risk scores analysis

INSERM-Univ Toulouse

Sandrine Andrieu

Amsterdam UMC

Marieke Hoevenaar-Blom

Jan-Willem van Dalen

Eric Moll van Charante

Edo Richard

Karolinska Institute

Miia Kivipelto

University of Eastern Finland

Hilkka Soininen

PRODEMOS consortium & funders

HATICE consortium & funders

preDIVA study group & funders

MAPT/DSA group & funders