Screening for Alzheimer's Disease Relation to Risk, Benefit Age and Genotype October 8, 2004 J. Wesson Ashford, M.D., Ph.D. Stanford / VA Alzheimer’s Center.

Screening for Alzheimer's DiseaseRelation to Risk, Benefit

Age and Genotype

October 8, 2004

J. Wesson Ashford, M.D., Ph.D.Stanford / VA Alzheimer’s Center

VAMC, Palo Alto, California

Slides at: www.medafile.com/ADben1.ppt

Why Diagnose AD Early?

• Safety (driving, compliance, cooking, etc.)• Family stress and misunderstanding (blame, denial) • Early education of caregivers of how to handle patient

(choices, getting started)• Advance planning while patient is competent (will, proxy,

power of attorney, advance directives)• Patient’s and Family’s right to know• Promotes advocacy for research and treatment development• Specific treatments now available

– May slow underlying disease process, the sooner the better– May delay nursing home placement longer if started earlier– May prevent conversion from Mild Cognitive Impairment to AD

BRIEF CLINICAL TOOLS FOR COGNITIVE ASSESSMENT

• MINI-MENTAL STATE EXAM• CLOCK DRAWING• ANIMAL NAMING (1 minute)• MATTIS DEMENTIA RATING SCALE• ALZHEIMER’S DISEASE

ASSESSEMENT SCALE (ADAS)• ACTIVITIES OF DAILY LIVING• GLOBAL CLINICAL SCALE• CLINICAL DEMENTIA RATING SCALE• GLOBAL DETERIORATION SCALE / FAST

Available Screening Tests• MMSE 10 -- 15 min

• Too long

• 7-Minute Screen 7 – 10 min• Too complex

• Clock Drawing Test 2 – 4 min• Not sensitive

• Mini-cog 3 – 5 min• Complex scoring, unclear adequacy

• Memory Impairment Screen 4 min• Need for slightly shorter, easier test

• (a suitably accurate test that takes less than 2 minutes is not available)

Need to Develop Better Screening and Early Assessment Tools

• Genetic vulnerability testing (trait risk)• Vulnerability factors (education, occupation, head

injury)• Early recognition (10 warning signs)• Screening tools (6th vital sign in elderly)• Positive diagnostic tests

– CSF – tau levels elevated, amyloid levels low– Brain scan – PET – DDNP, Congo-red derivatives

• Mild Dementia severity assessments• Detecting early change over time

– predicting progression, measuring rate

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Age

Perc

enta

ge AD

MCI

Non-Affected

Yesavavage et al., 2002Markov Chain model

Estimate MMSE as a function of time

0

5

10

15

20

25

30

-10 -8 -6 -4 -2 0 2 4 6 8 10

Estimated years into illness

MM

SE

scor

e

AAMI / MCI DEMENTIA

ALZHEIMER’S DISEASE

Ashford et al., 1995

AD all (easiest to hardest at p=.5)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

-4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10

DISABILITY ("time-index" year units)

PR

OB

AB

ILIT

Y C

OR

RE

CT

PENCILAPPL-REPWATCLOCATIONPENY-REPTABL-REPCLOS-ISRIT-HANDCITYFOLD-HLFSENTENCECOUNTYNO-IFSFLOORSEASONYEARPUT-LAPMONTHADDRESSDRAW-PNTDAYSPEL_ALLDATEAPPL-MEMPENY-MEMTABL-MEM

Mini-Mental State Exam items

Animals named in 1 min (mms>19) - CERAD data set

0

2

4

6

8

10

12

0 10 20 30 40

number of animals named

pe

rce

nt

of

tota

l

Normal Controls, CS = 1, n = 386

Alzheimer patients, CS = 0, n = 380

Animals named in 30 seconds (mms>19)

0

2

4

6

8

10

12

14

16

0 5 10 15 20 25

number of animals named

pe

rce

nt

of

tota

l

Normal Controls, n=386

Mild Alzheimer Patients, n=380JW Ashford, MD PhD, 2001

Animal naming ROCs (AUC 15: 0.74; 30: 0.83; 45: 0.86; 60: 0.87)

5

6

7

8

7

8

9

10

11

.

10

11

12

13

1415

16 17

0

10

20

30

40

50

60

70

80

90

100

0 10 20 30 40 50 60 70 80 90 100

False Positive Rate (%)

Sen

sitiv

ity (%

) animals in 15 secs

animals in 30 secs

animals in 45 secs

animals in 60 secs

Brief Alzheimer Screen (BAS)

• Repeat these three words: “apple, table, penny”.• So you will remember these words, repeat them again.• What is today’s date?

• D = 1 if within 2 days.

• Spell the word “WORLD” backwards• S = 1 point for each word in correct order

• “Name as many animals as you can in 30 seconds, GO!”• A = number of animals

• “What were the 3 words I asked you to repeat?” (no prompts)• R = 1 point for each word recalled

BAS = 3 x R + 2/3 x A + 5 x D + 2 x S

Mendiondo et al., J Alz Dis 5:391, 2003

0

10

20

30

40

50

60

70

80

90Pe

rcen

t of V

alid

atio

n Sa

mpl

e

3-22 23 24 25 26 27-39

BAS Score

Mild AD

Control

JW Ashford, MD PhD, 2001 Mendiondo et al., 2004

BRIEF ALZHEIMER SCREEN (Normal vs Mild AD, MMS>19)

9

20

1413

1211

10

9

6

7

8

2627

25

0

10

20

30

40

50

60

70

80

90

100

0 10 20 30 40 50 60 70 80 90 100

False Positive Rate (%) (1-Specificity)

Tru

e P

osi

tiv

e R

ate

(%

) (

Se

nsi

tiv

ity)

animals 1 m AUC = 0.868

animals 30 s AUC = 0.828

MMSE AUC = 0.965

Date+3 Rec AUC = 0.875

BAS AUC = 0.983

JW Ashford, MD PhD, 2003

CONCLUSIONS on the BAS

• A single cut-off score provides reasonable sensitivity and specificity for the diagnosis of AD within 2 – 3 minutes

• Two cut-off points divide the population into 3 tiers– the first cut-off indicates a low likelihood of dementia– the second indicates a high likelihood of dementia– the remaining group falls into a ‘gray area’ in need of

closer scrutiny, follow-up, and more extensive testing

• A suitably short screen can be administered yearly to individuals over 60 y/o as a 6th vital sign

• Next direction – use of IRT to locate level of impairment

MEMTRAX - Memory Test(to detect AD onset)

• New test to screen patients for AD: – World-Wide Web – based testing, – CD-distribution– KIOSK administration

• Determine level of ability / impairment• Test takes about 1-minute• Test can be repeated often (e.g., quarterly)

• Any change over time can be detected• Free test is at: www.medafile.com

Issues in Screening• ROC analysis provides independent values for how the

screening test values affect the normal and patient populations, then plots their relationship with respect to each other (specificity vs sensitivity respectively)

• The value of the test must be calculated with respect to the risk of the disease– Risk is affected by age, genotype, many other factors– Accounting for a priori probability is Bayesian analysis

• The cost of the applying the test and the costs of false positive and false negative results as well as the benefits of correct positive and negative results must be assessed

• Alzheimer’s disease is not a dichotomous diagnosis in early stages, but a continuum, which would be better assessed with a probabilistic statement, that would be better calculated from Item Response Theory

• Item Response Theory allows combination of many items

RELATIVE RISK FACTORS FOR ALZHEIMER’S DISEASE

• Family history of dementia 3.5 (2.6 - 4.6)• Family history – Downs 2.7 (1.2 - 5.7)• Family history - Parkinson’s 2.4 (1.0 - 5.8)• Maternal age > 40 years 1.7 (1.0 - 2.9)• Head trauma (with LOC) 1.8 (1.3 - 2.7)• History of depression 1.8 (1.3 - 2.7)• History of hypothyroidism 2.3 (1.0 - 5.4)• History of severe headache 0.7 (0.5 - 1.0)• NSAID use or statin use 0.2 (0.05 – 0.83)

Roca, 1994, t’Veldt, 2002

JW Ashford, 2004, J. Molecular Neurosci. 23: 155-163

U.S. mortality rate by age1999 CDC / 2000 census

0.0001

0.0010

0.0100

0.1000

1.0000

0 10 20 30 40 50 60 70 80 90 100Age

Yea

rly

Haz

ard

Males, 2t = 8.2yrsFemales, 2t = 7.5 yrsAlzheimer incidence

Genes and Alzheimer’s disease(60% - 80 % of causation)

• Familial AD (onset < 60 y/o) (<5%)• all known autosomal dominant genes relate to amyloid

– Presenilin I, II (ch 14, 1)– APP (ch 21)

• Non-familial (late onset)– APOE

• Clinical studies suggest 40 – 50% due to 4• If is considered, may be 95% of causation• Population studies suggest 10 – 20% cause• Evolution over last 300,000 to 200,000 years

– At least 20 other genes suspected of relating to AD

Ashford & Mortimer, 2002, J. Alz. Dis. 4:1-9.

APO-E genotype and AD risk46 Million in US > 60 y/o //// 4 Million have AD(data from Saunders et al., 1993; Farrer et al., 1997)

GenT %pop %AD #pop #AD risk If all US

E2/2 1% 0.1% 0.5M .004M 0.8% .4 M

E2/3 12 % 4% 5.5M .18M 3.2% 1.5 M

E3/3 60% 35% 27.6M 1.4M 5.1% 2.3 M

E3/4 21% 42% 9.6M 1.7M 18% 8.2 M

E4/4 2% 16% .9M .6M 67% 30.7M

Ashford & Mortimer, 2002, J. Alz. Dis. 4:1-9.Raber, Huang, Ashford, 2004, Neurobiol Aging. 25:641-650

Mortality Equationst = age in years

Td = time for mortality rate to doubleRo = mortality rate at time zero

alpha = ln(2) / Td

• Mortality rate u(t)R = Ro x exp (alpha x t)

• Survival curve s(t)S = exp ( - Ro/alpha x (exp (alpha x t) –1 ))

• Number of deaths per year d(t)D = -ds/dt = ( Ro + alpha x ln (1/s(t) )) x s(t)

= u(t) x s(t)

Dementia rate, for Td = 5 yrs

0.0001

0.001

0.01

0.1

1

10

100

1000

50 60 70 80 90 100

Age

Nu

mb

er

of

peo

ple

/yr

mean rateAPOE 4/4APOE 3/4APOE 3/3presenilin

J. W. Ashford, 2004

Probability Not Demented

00.10.20.30.40.50.60.70.80.9

1

50 60 70 80 90 100

Age

Pro

po

rtio

n o

f p

op

ula

tio

n

mean rateAPOE 4/4APOE 3/4APOE 3/3presenilin

J. W. Ashford, 2004

Probability of Dementia Onset

0

0.01

0.02

0.03

0.04

0.05

0.06

50 60 70 80 90 100

Age

pro

bab

ilit

y / y

r

mean rateAPOE 4/4APOE 3/4APOE 3/3Presenilin J. W. Ashford, 2004

Raber, Huang, Ashford, 2004, Neurobiol Aging. 25:641-650

U.S. AD Incidence by APOE(proportion of cases)

00.10.20.30.40.50.60.70.80.9

1

50 60 70 80 90 100Age

Pro

po

rtio

n /

Yea

r 4/4

3/43/3

Are we ready to do genetic testing to predict AD?

• The family members want it– They consider recommendations against genetic testing

to be “paternalistic”• Family members can make more powerful financial

decisions based on this knowledge than the relevance of insurance companies implementing changes in actuarial calculations

• Those at risk can seek more frequent testing– This is the best opportunity for early recognition

• Those at risk will be better advocates for research• Specific preventive treatments can be developed for each

genetic factor

Control: What happens without testing?

Total Population Risk=P

P

Have ADNo effective intervention

Do not have AD

P’

Helena Kraemer, 2003

Testing: What happens with testing?

Total Population

No ADAD

Unnecessary intervention OK No effective intervention Effective intervention

$ Testing $Testing $ Testing $ Testing$ Intervention $ Intervention

Iatragenic Damage? Clinical Wash Clinical Wash Clinical Gain

Major(?) Loss Minor (?) Loss Minor(?) Loss Major(?) Gain

PP’

SeSe’

SpSp’

Helena Kraemer, 2003

Cost – Benefit Calculation• I = incidence P = Total Population• X = cost of test, time to take (subject, ? Tester)• Sn = sensitivity of test = 1- False negative (Fn)• Sp = specificity of test = 1- False positive (Fp)• Cost:

– Total cost of test = P*X– Fp = $1000 * P * (1-I) * (1-Sp) (+false hope)– Tn = 0 (real peace of mind)– Fn = false peace of mind– Tp = (-$49000 * P * I * Sn) => $50,000 = NH cost/1year

• Avg Person Benefit= 49000*I*SN -X -1000*(1-I)*(1-Sp)

Sp 1 0.99 0.99 0.99 0.95 0.95 0.95 0.9 0.9 0.9 0.8

X I Sn 1 0.99 0.9 0.8 0.99 0.9 0.8 0.99 0.9 0.8 0.8

1 0.001 48 37.52 33.1 28.2 -2.44 -6.85 -11.8 -52.39 -56.8 -61.7 -161.6

1 0.005 244 231.6 210 185 191.8 169.8 145 142.1 120 95.5 -4

1 0.01 489 474.2 430 381 434.6 390.5 342 385.1 341 292 193

1 0.05 2449 2415 2195 1950 2377 2157 1912 2330 2109 1864 1769

10 0.001 39 28.52 24.1 19.2 -11.4 -15.9 -20.8 -61.39 -65.8 -70.7 -170.6

10 0.005 235 222.6 201 176 182.8 160.8 136 133.1 111 86.5 -13

10 0.01 480 465.2 421 372 425.6 381.5 333 376.1 332 283 184

10 0.05 2440 2406 2186 1941 2368 2148 1903 2321 2100 1855 1760

100 0.001 -51 -61.5 -65.9 -71 -101 -106 -111 -151.4 -156 -161 -260.6

100 0.005 145 132.6 111 86.1 92.8 70.75 46.3 43.05 21 -3.5 -103

100 0.01 390 375.2 331 282 335.6 291.5 243 286.1 242 193 94

100 0.05 2350 2316 2096 1851 2278 2058 1813 2231 2010 1765 1670

1000 0.001 -951 -961 -966 -971 -1001 -1006 -1011 -1051 -1056 -1061 -1161

1000 0.005 -755 -767 -789 -814 -807 -829 -854 -857 -879 -904 -1003

1000 0.01 -510 -525 -569 -618 -564 -609 -658 -613.9 -658 -707 -806

1000 0.05 1450 1416 1196 951 1378 1158 913 1331 1110 865 770

Future Directions for Alzheimer’s Disease

• Need more precise assessment of cognition from normal to severe dementia

• Need to use Item Response Theory and Computerized Adaptive Testing

• Screening tests need to be widely used, that implement computer administration

• Test analyses need to be risk/benefit-related• Better tests will lead to better treatments• Better treatments will lead to prevention