CVD Risk Prediction: Basic Concepts, Current Status, and Future Directions Donald M. Lloyd-Jones MD ScM FACC FAHA Chair and Associate Professor Department.

CVD Risk Prediction: Basic Concepts, Current Status, and Future Directions

Donald M. Lloyd-Jones MD ScM FACC FAHAChair and Associate Professor

Department of Preventive Medicine Director, Program in Risk Estimation, Communication and

PreventionMedical Director, Clinical Trials Unit

Bluhm Cardiovascular Institute, Dept. of Medicine Northwestern University Feinberg School of Medicine

Disclosures

• None

Predicting the FuturePhilosophical Perspectives

• “Prediction is very difficult, especially about the future.” - Niels Bohr (Yogi Berra?)

• “There are many methods for predicting the future. For example, you can read horoscopes, tea leaves, tarot cards, or crystal balls. Collectively, these methods are known as ‘nutty methods.’ Or you can put well-researched facts into sophisticated computer models, more commonly referred to as ‘a complete waste of time.’” - Scott Adams (Dilbert)

Circulation 2010; 121; 1768-1777

April 20, 2010

Current Paradigm for Risk Estimation and Treatment: ATP-III

Estimate 10-year risk (FRS)

<10% >20% or DM10-20%

Further testing

Lifestyleand drug therapy

Lifestylemodification

“Intensity of prevention efforts should match the absolute risk of the patient”

Risk Factor Units Value Notes

Gender m

Age years 45

Total Cholesterol mg/dL 230

HDL mg/dL 40

Systolic Blood Pressure mmHg 135

Treatment for Hypertension {Only if SBP>120} yes (y) or no (n) n

Current Smoker yes (y) or no (n) n

Time Frame for Risk Estimate 10 years

Your Risk 5%

ATP-III Risk Estimator

Your Risk Estimate Comparative Risks for Lowest = Total Chol<160, HDL>60, Optimal SBP (<120), No Trt for Htn, Non-Smoker Same Age and Gender Low = Total Chol 160-199, HDL 50-59, Normal SBP (<130), No Trt for Htn, Non-Smoker

What doesWhat doesthis mean?this mean?

Hmmm….

• Seems unsatisfying

• But is it “wrong?”

• Can we do better?

• How do we know if we are doing better?

How Do We Measure the Performance of a Screening/Risk Prediction Test?

• NOT by anecdote

• NOR (solely) by independent RR

Circulation 2009

How Do We Measure the Performance of a Screening/Risk Prediction Test?

• We should use:– Sensitivity/specifity/predictive value– Discrimination

• Area under the ROC curve (AUC; C statistic)

– Calibration• Hosmer/Lemeshow and others

– Informativeness criteria• AIC, BIC

– Likelihood ratios (LR+ and LR-)– Reclassification (NRI, IDI)

NRI and IDI

• Net reclassification improvement (NRI)– Among those with events

No. moved up – No. moved down

Total number with events

– Among those without eventsNo. moved down – No. moved up

Total number without events

• Integrated Discrimination Index (IDI)– Represents difference in R2 between models– Also a good indicator of how far people move

}Sum = NRIRange -2 to +2

A Couple of Risk Scores to Help You Assess Risk in Your Patients

FRS 1991

FRS 1998

FRS/ATP-IIISCORE

QRISK

SCORE – S. Europe

SCORE – N. EuropeSCORE - Greece

ARIC Genetic RS

Reynolds for Women

Reynolds for Men

Cuore

FRS CVD 2008

PROCAM

Criticisms of Risk Scores

• “Risk scores don’t work, because they try to estimate risk for an individual based on average risk in a population”

• Uncertainty about performance

• Applicability to other race/ethnic groups

• Does not incorporate other/novel risk markers we might believe to be important

• Not useful for young adults, women

Area Under the Curve (AUC) or C Statistic for FRS

Study Design Sex AUC

Women’s Heath Prospective W 0.81

Rotterdam Study Nested C/C W/M 0.773

MONICA Germany Prospective M 0.735

Framingham Offspring Prospective W/M 0.74

Framingham Heart Prospective W/M 0.78

Cardiovascular Health Prospective W/M 0.73

Lloyd-Jones, Ann Intern Med 2006

Transportability of FRS

• D’Agostino et al, JAMA 2001– Well-calibrated and performed well in white and

black men and women– Ranked risk appropriately but consistently over-

estimated absolute risk in Puerto Rican, Japanese-American, Native American, and Chinese

– After recalibration to population CHD rates, performed very well

• Chambless et al, AJE 2003– Similar AUC and calibration for AA and whites in

ARIC

A Couple of Biomarkers to “Help” You Assess Risk in Your Patients

CRP

BNP

sICAM-1PAI-1/tPA

Lp-PLA2

CAC score

D-dimer

HcyE-selectin

Fibrinogen

Leptin

Microalbuminuria

IL-6

IL-18

Factor VIII

MMP-1

Lp(a)CMV Ab

WBC countHSV-1 Ab

C-IMT

Biomarkers and CHD PredictionARIC Study

Folsom AR, Archives 2006

0

5

10

15

20

25

0-1% 2-4% 5-9% >=10%

Mu

ltiv

aria

ble

RR

fo

r C

VD

Framingham Estimated 10-Year CHD Risk

CRP<1.01.0-3.0>3.0

Does CRP Add to Risk Prediction?

Ridker, NEJM 2002

AUC – CRP: 0.81AUC + CRP: 0.81

Multimarkers and CHD Prediction Cardiovascular Health Study

Shlipak, M. G. et al. JAMA 2005;293:1737-1745.

Traditional: Smoking, DM, Sys HTN, LVH, Low PA, EtOHNovel: IL-6, CRP, Fn, Lp(a), Anemia, FVIII

AUC-: 0.73AUC+: 0.72

AUC-: 0.73AUC+: 0.74

No CKD CKD

Paynter et al. JAMA 2010; 303: 631

• 19,303 women from the WHS followed for median >12 years

• 101 SNPs from literature associated with CVD or intermediate phenotypes (12 with CVD)

• Created a genetic risk score based on number of risk alleles; Range: 73-125 (4-19)

Genetic Risk Score in WHS

101 SNP

Genetic Risk Score

12 SNP

Genetic Risk Score

HR (95% CI) Per 1 Allele

Δ in C Index

NRIHR (95% CI)

Per 1 AlleleΔ in C Index

NRI

Age-adjusted

1.02

(1.00-1.03)0.003 1.2

1.05

(1.01-1.09)0.004 0.6

FRS-adjusted

1.00

(0.99-1.01)0.000 0.5

1.04

(1.00-1.08)0.001 0.5

RRS-adjusted

1.00

(0.99-1.01)0.000 0.4

1.04

(1.00-1.07)0.000 0.8

Paynter et al. JAMA 2010; 303: 631

Screening for Coronary Calcium

RF-Adj HR

6.84

7.083.89

1.0 (ref)

MESA Study (Detrano, NEJM 2008)

C statisticsTraditional RFs: 0.79Plus CAC: 0.83

JAMA 2010; 303: 1610-1616

April 28, 2010

Reclassification Using CAC ScoreMESA

JAMA 2010; 303: 1610-1616

+23%

+13%

NRI = 0.25

C-IMT, Plaque and CHD Events

• 13,145 asx ppts in the ARIC Study– Mean age 54 y– 25% African-American

• Used ARIC-based traditional RF score to estimate 10-y risk for CHD (incl. revasc)

• Added C-IMT +/- plaque

• 1812 CHD events over 15 y

Nambi et al, JACC 2010; 55: 1600

C-statistics with Addition of C-IMT and Plaque

Nambi et al, JACC 2010; 55: 1600

Reclassification with Addition of C-IMT and Plaque

Nambi et al, JACC 2010; 55: 1600

Reclassified 23% (8.6%, 37.5%, 38.3% and 21.5%)NRI 0.099 [0.217 in intermediate]/ IDI 0.011

Summary• Existing biomarkers for clinical prediction

– Used in isolation are ineffective– Used in the context of 10 year risk estimation equations that

contain age and established, causal and treatable/modifiable risk factors, are very unlikely to lead us to the promised land of perfect discrimination or substantial reclassification

– Also misclassify people, leading to inappropriate down-staging of risk

– Might be useful as niche tests in sequential testing schemes for patients on the fence

• Disease screening appears promising, but requires more study for optimal implementation in a cost-effective and safe manner

Can longer-term risk estimatesprovide a useful adjunct to

10-year risk estimates?

0

5

10

15

20

25

30

Total cholesterol (mg/dL)160200240160200240

2535

45HDL-c

(mg/dL)Pre

dic

ted

10-

Yea

r R

isk

(%)

Smoker Non-smoker

SBP 130 mm Hg

0

5

10

15

20

25

30

Total cholesterol (mg/dL)160200240160200240

2535

45HDL-c

(mg/dL)Pre

dic

ted

10-

Yea

r R

isk

(%)

Smoker Non-smoker

SBP 150 mm Hg

10-Year Predicted Risks in ATP Risk Assessment Tool:Man, Age 45

Cavanaugh-Hussey, Berry,Lloyd-Jones, Prev Med 2008.

0

5

10

15

20

25

30

Total cholesterol (mg/dL)160200240160200240

3545

55HDL-c

(mg/dL)Pre

dic

ted

10-

Yea

r R

isk

(%)

Smoker Non-smoker

SBP 150 mm Hg

0

5

10

15

20

25

30

Total cholesterol (mg/dL)160200240160200240

3545

55HDL-c

(mg/dL)Pre

dic

ted

10-

Yea

r R

isk

(%)

Smoker Non-smoker

SBP 130 mm Hg10-Year Predicted Risks in ATP Risk Assessment Tool:Woman, Age 55

Cavanaugh-Hussey, Berry,Lloyd-Jones, Prev Med 2008.

Implications

• Vast majority of younger adults are considered to be at “low risk”

– Weight of age – 10-year risk window– Clinical treatment thresholds imposed

• BUT low risk ≠ “no risk”• Additional means for risk estimation and

communication needed to help men age <45 and women age <65

– Importance of addressing multiple moderate or single elevated risk factors for long-term CHD prevention

Rationale: Lifetime Risk Estimation

• Lifetime risk– The absolute cumulative risk of an individual

developing a given disease before death– Accounts for risk of disease of interest,

remaining life expectancy, and competing causes of death

– Reflects real-life risks better than Kaplan-Meier cumulative incidence

All Optimal

Not Optimal

Elevated

SBP/

DBP<120 and

<80120-139 or

80-89

140-159 or

90-99

≥160 or

≥100

TC <180 180-199 200-239 ≥240

DM No No No Yes

Smoking No No No Yes

Aggregate Risk Factor Burden

Lloyd-Jones, Circulation 2006; 113: 791-798

1 Major

≥2 Major

Lifetime Risk for ASCVD by RF Strata Framingham Heart Study, Age 50

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

50 60 70 80 900

0.1

0.2

0.3

0.4

0.5

0.6

0.7

50 60 70 80 90Attained Age

Ad

just

ed C

um

ula

tive

In

cid

ence

5%

36%

50%

69%

8%

27%

50%

Men Women

46%

39%

Lloyd-Jones, Circulation 2006

2 Major RFs1 Major RF1 Elevated RF1 Not Optimal RFOptimal RFs

0

2

4

6

8

10

12

14

16

18

20

20-29 30-39 40-49 50-59 60-79

Distributions of 10-Year and Lifetime Risk Strata by Age and Sex

US Adults, NHANES 2003-2006

0

2

4

6

8

10

12

14

16

18

20

20-29 30-39 40-49 50-59 60-79

Marma, Circ CQO 2010Age

Ind

ivid

ual

s (m

illio

ns)

High 10Low 10/High LifeLow 10/Low Life

Men Women

56% (87,000,000) have low short-term

but high lifetime predicted risk

A New Risk Estimator?Risk for Endpoint

Risk Factor Units Value Endpoint 10-Year Lifetime

Gender M or F M

Age years 45

Total Cholesterol mg/dL 230

HDL-Cholesterol mg/dL 40

Systolic Blood Pressure mm Hg 135

Treatment for Hypertension (if SBP >120) Y or N N

Current Smoker Y or N N

A New Risk Estimator?Risk for Endpoint

Risk Factor Units Value Endpoint 10-Year Lifetime

Gender M or F M Hard CHD 1.4% 42%

Age years 45

Total Cholesterol mg/dL 230 F/NF Stroke 0.2% 21%

HDL-Cholesterol mg/dL 40

Systolic Blood Pressure mm Hg 135

Total ASCVD

1.5% 46%

Treatment for Hypertension (if SBP >120) Y or N N

Vascular Age 54

Current Smoker Y or N N

Estimated Life-Years Lost

>10

The Future of Risk Prediction as a Strategy for CVD Prevention

• We need to understand – How well this approach works in clinical practice– How to improve adoption– How to optimize utilization

• Consider– Expanding endpoint to include all major CVD

endpoints (helps women)– Age-specific risk equations (but tough)– Long-term and lifetime risk estimation as an adjunct to

10-year risk