©2012, Genentech

An integrated natural disease progression model of nine cognitive and biomarker endpoints in

patients with Alzheimer’s Disease

Angelica Quartino1*

Dan Polhamus2*, James Rogers2, Jin Jin1

*authors contributed equally to the work presented 1.  Genentech Inc. South San Francisco, CA 2.  Metrum Research Group, Tariffville, CT

Alzheimer’s Disease – a progressive neuro-degenerative disease 2

q  CSF biomarkers of disease (amyloid): ↓Aβ42, q  Brain amyloid load: ↑ amyloid PET imaging q  CSF biomarkers of disease (neuro-degeneration): ↑ t-tau and ↑ p-tau q  Brain atrophy: volumetric MRI (↓whole brain volume, ↓hippocampal volume, ↑ventricle size) q  Cortical activity: ↓ FDG-PET q  Cognitive and functional impairment scales: ↑ ADAS-Cog 12-item, ↑ CDR-SOB, ↓ MMSE

•  Trouble remembering recent events to inability to preform basic tasks and full time care •  Death within ~ 9 years from diagnosis

Jack et al, Lancet Neurology 12:207 (2013) Time

Background

Prog

ress

ion

of b

iom

arke

r abn

orm

ality

Min

Max

Mild to moderate AD

Objectives

To establish a natural disease progression model integrating multiple biomarkers and endpoints in patients with mild Alzheimer’s Disease1. è  An enhanced ability to identify and understand disease progression

and impact of covariates and drug treatment effect, rather than within endpoints

è  Ability to simulate realistic multivariate longitudinal data, to allow assessment of studies with co-primary endpoints

3

1) Polhamus D et al. AAIC 2013 CDR-SOB, vMRI in MCI

Alzheimer’s Disease Neuroimaging Initiative Database 4

MMSE: Mini-Mental State Examination * Data in the analysis was from ADNI database extracted on 22 January 2014 : www.loni.ucla.edu/ADNI

q  298 mild Alzheimer’s Disease subjects* q  Baseline MMSE 20-26 q  Baseline covariates e.g. age, gender q  Up to 3 years longitudinal changes in:

•  ADAS-Cog 12-item score •  CDR each of 6 items score •  volumetric MRI (hippocampal, ventricles)

Data and Methods

Natural history non-treatment study in USA/Canada

Software: OpenBUGS v. 3.2.2

Integrated Longitudinal Alzheimer’s Disease Model

Treatment Effect

Biomarkers Hippocampal volume Ventricular volume

Cognition/Function CDR 6-items scores

Cognition ADAScog12

Endpoints

Underlying AD “disease”

progression

Amyloid PET

CSF Aβ42, pTau, tTau

FDG-PET

Prognostic Factors

ApoE4 genotype

Baseline MMSE

Age and gender

Results 5

0

2

4

6

0 1 2 3 4 5Time

Und

erly

ing

dise

ase

stat

us

0 1 2 3 4 5 Time

6 4 2 0 U

nder

lyin

g D

isea

se S

tatu

s

20

40

60

80

0 2 4 6Underlying disease status

Endp

oint

raw

sco

re

80 60 40 20

0 2 4 6 Underlying Disease Status

End

poin

t Sco

re

20

40

60

80

0 1 2 3 4 5Time

Endp

oint

raw

sco

re

80 60 40 20

End

poin

t Sco

re

0 1 2 3 4 5 Time

Worse

Correlation between predicted disease status and observed endpoints 6 Results

Predicted underlying disease status

5

10

15

−2.5 0.0 2.5 5.0Underlying disease status

CDR6

DOM

abs

olut

e sc

ore

CDR SOB score

15

10 5

20

40

60

80

−2.5 0.0 2.5 5.0Underlying disease status

ADAS

12 a

bsol

ute

scor

e

ADAS-Cog12 score 80

60

40

20

4000

6000

8000

10000

−2.5 0.0 2.5 5.0Underlying disease status

Hipp

ocam

pus

abso

lute

sco

re

Hippocampus volume

-2.5 0.0 2.5 5.0

10000

8000

6000

4000 40000

80000

120000

160000

−2.5 0.0 2.5 5.0Underlying disease status

Vent

ricles

abs

olute

scor

e

Ventricular volume

-2.5 0.0 2.5 5.0

160000

120000

80000

40000

Obs

erve

d en

dpoi

nt s

core

●

●●

●

●

●

●●

●

●

●●

●

●

●

●●●

●

●

ADAS12 CDRSUM

Hippocampus Ventricles0

5

10

15

20

0

2

4

6

−1000

−750

−500

−250

0

0

10000

20000

30000

0 1 2 3 0 1 2 3Year

Cha

nge

from

bas

elin

e

7

observed data (red dots) and simulated data (black lines) with 95% credible interval (shaded area)

Results

The model accurately captures the central trend of observed data

0 1 2 3 0 1 2 3

Time (Years)

Cha

nge

scor

e fr

om b

asel

ine

ADAS-Cog12 score CDR SOB score

Hippocampus volume Ventricular volume

The model maintains observed correlation between endpoints Results

Observed correlation (red line) Simulated correlation (black distribution)

8

10

5

0

Hippocampus Ventricles ADAS12

0

5

10

0.0

2.5

5.0

7.5

0.0

2.5

5.0

7.5

10.0

CD

RSum

ADAS12Ventricles

−0.5 0.0 0.5 −0.5 0.0 0.5 −0.5 0.0 0.5Spearman's rho-0.5 0.0 0.5

Spearman’s rho

-0.5 0.0 0.5 -0.5 0.0 0.5

CD

R SO

B score

Hippocampus volume Ventricular volume ADAS-Cog12 score

AD

AS-C

og12 score Ventricular volum

e

10

5

0

7.5

5.0

2.5

0.0 10.0

7.5

5.0

2.5

0.0

Impact of prognostic factors on underlying disease progression rate 9 Results

Posterior estimated effect of normalized prognostic factors relative to reference subject (with uncertainty)

Faster disease progression Slower disease progression

ApoE4 positive

High baseline MMSE

Reference subject | ApoE4 negative | Baseline MMSE 23 | Age 75 |

Covariate effects on the slope of the latent score (per month)

Base

line

cova

riate

s

lambda_APOE

lambda_BMMSE

lambda_AGE

lambda_APOE.BMMSE

−0.010 −0.005 0.000 0.005

18 82

100

100

11 89

High age

Interaction high baseline MMSE-ApoE4 positive

MMSE: Mini-Mental State Examination

10 Results

Change in disease progression and endpoints for subpopulations

Baseline MMSE ApoE4 genotype

wor

se

0

4

8

12

0.0 2.5 5.0 7.5 10.0Year

Raw

scor

e

ApoE4 StatusNegativePositive

BMMSE2026

Theta progression of the average individual

0

4

8

12

0.0 2.5 5.0 7.5 10.0Year

Raw

sco

re

ApoE4 StatusNegativePositive

BMMSE2026

Theta progression of the average individual

Time (Years)

20

40

60

80

0.0 2.5 5.0 7.5 10.0Year

Raw

sco

re

ApoE4 StatusNegativePositive

BMMSE2026

ADAS−cog progression of the average individualADAS-Cog12 score

80

60

40

20 0

5

10

15

0.0 2.5 5.0 7.5 10.0Year

Raw

sco

re

ApoE4 StatusNegativePositive

BMMSE2026

CDR sum progression of the average individualCDR SOB score

15

10

5

0

3000

4000

5000

6000

0.0 2.5 5.0 7.5 10.0Year

Raw

scor

e

ApoE4 StatusNegativePositive

BMMSE2026

Hippocampus progression of the average individualHippocampus volume 6000

5000

4000

3000

0.0 2.5 5.0 7.5 10.0

40000

80000

120000

160000

0.0 2.5 5.0 7.5 10.0Year

Raw

scor

e

ApoE4 StatusNegativePositive

BMMSE2026

Ventricular progression of the average individualVentricular volume

160000

120000

80000

40000

0.0 2.5 5.0 7.5 10.0

0

4

8

12

0.0 2.5 5.0 7.5 10.0Year

Raw

sco

re

ApoE4 StatusNegativePositive

BMMSE2026

Theta progression of the average individual12

8

4

0

0.0 2.5 5.0 7.5 10.0

Scor

e Va

lue

Time (Years)

Underlying disease progression

●

●

●

●

−15

−10

−5

0

0.0 0.5 1.0 1.5 2.0Years

Cha

nge

from

bas

elin

e

DM factor●

●

50% reductionADNI AD progression

ADAS12ADAS-Cog12 score

0

-5

-10

-15

●

●

●

●

−5

−4

−3

−2

−1

0

0.0 0.5 1.0 1.5 2.0Years

Cha

nge

from

bas

elin

e

DM factor●

●

50% reductionADNI AD progression

CDRsumCDR SOB score

0

-1

-2

-3

-4

-5

Projected impact of a hypothetical treatment effect across endpoints 11 Application

Observed unadjusted mean of ADNI data (dots) Model predictions of a hypothetical treatment effect using baseline covariates (lines and shaded area)

●

●

●

●

−5

−4

−3

−2

−1

0

0.0 0.5 1.0 1.5 2.0Years

Chan

ge fr

om b

asel

ine

DM factor●

●

50% reductionADNI AD progression

CDRsum

●

●

●

●

−10.0

−7.5

−5.0

−2.5

0.0

0.0 0.5 1.0 1.5 2.0Years

Perc

ent c

hang

e fro

m b

asel

ine

DM factor●

●

50% reductionADNI AD progression

HippocampusHippocampus volume

0

-2.5

-5.0

-7.5

-10.0

0.0 0.5 1.0 1.5 2.0

●

●

●

●

−30

−20

−10

0

0.0 0.5 1.0 1.5 2.0Years

Perc

ent c

hang

e fro

m b

asel

ine

DM factor●

●

50% reductionADNI AD progression

VentriclesVentricular volume 0

-10

-20

-30

0.0 0.5 1.0 1.5 2.0

Cha

nge

scor

e %

Cha

nge

scor

e

Time (Years)

Summary 12

Alzheimer’s Disease Model Placebo Model Drug Effect Model

Longitudinal PKPD model for Alzheimer’s Disease

Model properties q  Greater insights of disease

progression, impact of patient covariates and drug treatment effect

q  Allows translation of information across endpoints and biomarkers

q  Ability to assess the sensitivity of the different endpoints in subpopulations

q  Ability to simulate realistic

multivariate longitudinal data

Application q  Comparison of novel-treatment

outcomes and placebo response to historical data across endpoints

q  Joint analysis of multiple endpoints (e.g. co-primary endpoints)

q  Trial design optimization for multiple endpoints

Alzheimer’s Disease Model

Co-authors: Dan, Jim and Jin

Colleagues and team members at Genentech

Alzheimer’s Disease Neuroimaging Initiative

Acknowledgements