dodge_symposium mci 2014 v5

8/13/2019 Dodge_Symposium MCI 2014 V5

1/33

Biomarker Values Predictive of

Cognitive Decline:Baseline or Progression?Hiroko H. Dodge, PhD

As soc ia te Professo r o f Neuro logyDirecto r, Bio s ta t is t ics and Data Core, Layton Ag ing

and Alzheimers Disease Center Orego n Health and Science Univers i ty

Port land , ORAdjun c t Research As soc ia te Professo rMichigan Alzheimers Disease Center

An n A rbo r, MI


2/33

Nothing to disclose


3/33

Outline of my talkI. Discuss briefly several statistical

approaches useful for longitudinal

data analysisII. Introduce our most recent work in

which we aim to identify biomarker

values associated with declines incognitive functions, using ADNI 1data


4/33

Longitudinal Data AnalysisWhy important?

Knowledge of biomarkers predictive ofcognitive and functional decline canfacilitate selective recruitment ofparticipants for clinical trials.

Research efforts are moving towards earlyidentification of high risk subjects andprevention of progression. Identifyingbiomarkers associated with subtle declinesin cognitive functions among asymptomaticsubjects is especially critical.


5/33

I. (Some) Statistical Approaches

Useful in Longitudinal Data Analysis of Biomarkers andCognition

Change point modelLatent trajectory model


6/33

Change Point Model Developed by Dr. Charles Hall (Einstein

School of Medicine) Identify the location of changes

(acceleration points) in slopes in relation toa specific endpoint (incidence of dementiaonset, MCI, death, etc)

F u n c

t i o n


7/33

Silbert, Dodge et. al. Trajectory of white matter

hyperintensity burden preceding mild cognitiveimpairment. Neurology 2012;79:741-747(PMC3421153)

Buracchio, Dodge et. al. The trajectory of gaitspeed preceding MCI. Archives of Neurology.

Archives of Neurology, 2010; 67:980-986(PMC2921227)

Dodge, Ganguli, et al. Terminal decline andpractice effects in non-demented older adults.Neurology, 2011:77;722-730. (PMC3164394)

Change Point Models: Examples

14 yrs

6 yrs


8/33

A pp l ica tion to B iom arker da ta:Ident i fy A cc elerat ion Points

Jack et al., ( Lancet Neurol 2010), Sperling et al.,( Alzheimers and Dementia 2011)

Acceleration in changesstarts

t t t'

Time pointsOrder of

accelerations inbiomarkers


9/33

Latent Trajectory Model

Mixed Effects Model:each subject can havea different time slope

Slope Distribution

Average Trajectory


10/33

Latent Trajectory Model

Identifyhomogeneousgroups interms of

trajectoriesand estimatecharacteristicsassociatedwith eachgroupsimultaneously(mixturemodelling)


11/33

A PPL ICATION O F MIXEDEFFECTS MODEL Erten-Lyons, Dodge et. al.Neuropathological basis of age-associated brain atrophy. JAMA

Neurology 70:616-622, 2013.

Dodge, Ganguli et. al. CohortEffects in Age-AssociatedCognitive Trajectories. Journal ofGerontology: Medical Sciences.In press.

Mixed Effects Model: ExamplesVentricular Volume Expansion

By CERAD Neuritic Plaque CategoriesNone/Sparse Moderate/Freq


12/33

L ATENT TRA JECTORYANALYSIS Dodge, Kaye et. al. In-home walking speeds andvariability trajectoriesassociated with MildCognitive Impairment.Neurology, 2012:78(24):1946-1952.

(PMC3369505)

Latent Trajectory Model: Examples


13/33

II. Baseline or Progression?-

Biomarkers Predictive ofCognitive Declines

(using ADNI 1 data, applicationof mixed effects model)


14/33

Biomarkers vs. cognitiveoutcomes

C o g n

i t i v e

F u n c

t i o n s

time (age)

Individual Specific Deviations from populationmean trajectory (random components)

Variability explained by Age, education, (reservefactors )

Baseline biomarker values

Rate ofprogression inbiomarker values

Populationmean/averagetrajectory


15/33

Aim/Data

To examine which components ---baselinevalues or biomarker progressions--explains more variability in cognitive

declines in memory and executivefunctions. DATA: the Alzheimers Disease

Neuroimaging Initiative (ADNI 1). 526subjects with valid data in at least one ofour variables of interest were used in thisstudy.


16/33

Cognitive Outcomes

Trajectory (slope) of cognitive functions:1. ADNI-memory (ADNI-Mem) and2. ADNI-executive (ADNI-Exe).

(Crane et al., 2012; Gibbons et al., 2012)

The scores are psychometrically optimized

composite scores of memory and executivefunction, derived from items from ADNIneuropsychological tests


17/33

Approach: 2-stage

Stage 1. Individual-specific slope of thelongitudinal trajectory of each biomarker was estimated using mixed effects models.

Longitudinal trajectories of biomarkers model = + + + + ,

where ( , ) ~ 0, and

~ (0,2 ) .


18/33

Approach

Stage 2. Estimated individual-specific slope ofeach biomarker and observed baseline valueswere used as predictors of cognitive declinesusing mixed effects models. = + 1* < . + 2 1* . ( 0.5) +;

= + Age at baseline + 2 Sex + 3 Education + Apoe4 + + ;

2 =2 + 2 Sex + 22 Apoe4 + 23

Changes in diagnosis + 2 + 2 + 2 ;


19/33

Approach

Variability in cognitive declines explainedby subject-specific baseline biomarkervalues was compared with variabilityexplained by biomarker progressions.


20/33

Results


21/33

Notes for Table

Medial Temporal Cortical Thickness: summary variable by adding averagedmeans for left and right entorhinal (ERC), perirhinal (PRC) and posteriorparahipplocampal (PPHC) cortical region thicknesses.

Controlled variables: age at baseline, sex, education, apoe 4 allele (at leastone vs. none) and practice effects

N/A: Variability increased instead of decreased or had no changes, afterinclusion of the predictors in the model (i.e., including these variables didnot explain the variability of cognitive outcomes or caused more estimationerrors instead of explaining the variability).


22/33

OUTCOME: TRAJECTORIES OFMEMORY FUNCTION

Normal Group Among MCI* Among AD

Biomarker% of variability

explained bybiomarkers

% ofvariability

explained by

biomarkers

% of variabilityexplained bybiomarkers

CSF ttau baseline N/A N/A N/A CSF ttau progression N/A N/A N/A CSF Abeta 42 baseline N/A 5.10% N/A CSF Abeta 42 progression N/A 10.30% 6.60% FDG-PET baseline N/A 12.20% 30.00%

FDG-PET progression 1.50% 12.70% 84.00% Log_wmh/icv baseline N/A 0.50% N/A Log_wmh/icv progression N/A N/A 3.00% Hpcv/icv baseline N/A 9.00% 4.70% Hpcv/icv progression N/A 19.80% 26.00% Ventricles/icv baseline N/A 8.70% 4.20% Ventricles/icv progression N/A 39.40% 63.80% Total brain/icv baseline N/A 2.40% N/A Total brain/icv progression N/A 16.00% 26.00% Precuneus thickness baseline 4.52% N/A 12.12% Precuneus thickness progression 1.29% 4.87% 6.94% Medial Temporal cortical thickness baseline N/A 8.14% N/A Medial Temporal cortical thickness progression 4.44% 28.68% 34.67%


23/33

OUTCOME: TRAJECTORIES OFEXECU TIVE FUNCTION

Normal Group Among MCI* Among AD

Biomarker% of variability

explained bybiomarkers

% ofvariability

explained by

biomarkers

% of variabilityexplained bybiomarkers

CSF ttau baseline N/A N/A N/ACSF ttau progression N/A N/A N/ACSF Abeta 42 baseline N/A 23.70% N/ACSF Abeta 42 progression 2.50% 14.40% N/AFDG-PET baseline N/A 35.20% 18.20%

FDG-PET progression 6.20% 28.30% 39.50%Log_wmh/icv baseline N/A 2.40% N/ALog_wmh/icv progression N/A N/A N/AHpcv/icv baseline 9.20% N/A N/AHpcv/icv progression N/A 9.00% N/AVentricles/icv baseline N/A 13.00% 11.10%Ventricles/icv progression 16.90% 44.50% 65.10%Total brain/icv baseline N/A 9.50% 3.50%Total brain/icv progression N/A 32.20% 18.80%Precuneus thickness baseline 21.14% N/A 8.66%Precuneus thickness progression 9.78% 5.38% N/AMedial Temporal cortical thickness baseline N/A N/A N/AMedial Temporal cortical thickness progression N/A 21.55% 22.08%


24/33

Conclusions


25/33

A number of studies have shown support forthe hypothetical AD progression modeldeveloped by Jack et al (Jack et al., 2013;Jack et al., 2010).

Our study results also coincides with themodel to some extent; Across diagnosticgroups, the percentages of variability incognitive declines explained by functional

(FDG-PET) or structural (brain morphometric)biomarkers (either their baseline values orprogressions) increased significantly asdisease progressed from normal to MCI.


26/33

Among normal subjects in this dataset,precuneus thickness baseline values andmedial temporal cortical thinningprogressions explained variability indeclines of ADNI-Mem by over 4%.

Although this is small, it showed more

explanatory ability than CSF A42 and t -tau.


27/33

The relatively poor performance of CSF A42 and t -tau biomarker progression

values in explaining the cognitivetrajectories could be due to the fact that:The control group in ADNI is likely to be

heterogeneous and include significantnumbers of individuals who will notprogress to ADShorter duration of follow-up of thesemarkers might have reduced precisionof trajectory estimate in comparisonwith other markers in the current study.


28/33

Take Home MessageFor most biomarkers, biomarker progressions

were more associated with cognitive declinethan baseline values.This suggests that clinical trials which require

recruiting at-risk subjects could be improvedby using progression rather than baselinevalues in biomarkers to enrich the studysubjects.Future studies are warranted to estimate theincremental effectiveness of improving clinicaltrial statistical power by using biomarkerprogression criteria


29/33

Special Thanks to:Michigan A DCRoger Albin, MDRobert Koeppe, PhDJian Zhu, MSOrego n Health &

Science Univers i tyADC

Jeffrey Kaye, MDLisa Silbert, MDUC DavisDaniel Harvey, PhDUnivers i ty of Pi t tsb urg hMary Ganguli, MD

Fun ding Sources

Michigan ADC pi lo t

grant

R13 AG030995, Frid ayHarbor Ad vanced

Psychom et r ic WorkSh o p 2011

MCI Sym pos ium


30/33

CSF A42 (U2 series) in pg/mL


31/33

CSF T-tau (U2 series) in pg/mL


32/33

FDG-PET summary score


33/33

Medial Temporal CorticalThickness in Millimeters (mm)

dodge_symposium mci 2014 v5

Documents