Validation of the Pooled - PowerPoint Presentation

Slide1

Validation of the Pooled Cohort 10-year Atherosclerotic Cardiovascular Disease Risk Equations

Paul Muntner, Lisandro D Colantonio, Mary Cushman, David C Goff Jr., George Howard, Virginia J Howard, Brett Kissela, Emily B Levitan, Donald M. Lloyd-Jones, Monika M Safford University of Alabama at Birmingham, University of Vermont, University of Colorado, University of Cincinnati, Northwestern University.

On behalf of REGARDS and REGARDS-MIThis study was supported by U01 NS041588 (NINDS) and R01 HL080477, K24 HL111154 (NHLBI)

Disclosures

Drs. Muntner, Howard, Levitan, and Safford have received grant funding from Amgen Inc. for work unrelated to this presentation.

Dr. Muntner has served on an advisory board for Amgen Inc.

Drs. Cushman and Safford have served as consultants for

DiaDexus

.

Background

In 2013, the American College of Cardiology / American Heart Association (ACC/AHA) published a guideline for the estimation of atherosclerotic cardiovascular disease (ASCVD) risk.This guideline included the development of the Pooled Cohort risk equations for estimating 10-year risk for incident ASCVD. These equations can be used to guide the decision to initiate statins for people 40-79 years without ASCVD or diabetes and with LDL-C of 70 to 189 mg/dL – “clinically relevant population”. Consideration of statin treatment is recommended for adults with a 10-year ASCVD risk ≥ 7.5%

Goff,

J Am

Coll

Cardiol

2013; Stone

, J Am

Coll

Cardiol

2013

Background

The Pooled Cohort risk equations were developed using data from several studies conducted before 2000. Marked declines in ASCVD incidence have occurred over the past 2 decades.These equations were reported to over-estimate risk in analyses of the Women’s Health Study, Women’s Health Initiative and the Physicians Health Study.Prior analyses: Did not focus on the population for whom the equations may inform a discussion to initiate statins.Did not have surveillance components

Rosamond, Circulation 2012;

Kleindorfer

, Stroke 2010;

Ridker

, Lancet 2013

Objective

To

evaluate the validity of

the Pooled Cohort risk equations in a contemporary US

population for whom the

equations are intended to inform discussions about initiating

statins.

We assessed

Calibration:

Do the Pooled Cohort risk equations

accurately

estimate

the

observed absolute risk level?

Discrimination:

Are

individuals with higher predicted risk more likely to have

events?

Methods

We used data from the REasons for Geographic And Racial Differences in Stroke (REGARDS) study: Population-based cohort of 30,239 blacks and whites ≥45 years of age residing in 48 contiguous US states and Washington, DCEnrolled between January 2003 and October 2007. All participants provided written informed consent.Primary analyses focused on the “clinically relevant population” Those for whom 10-year ASCVD risk can be used to guide decision-making We excluded participants taking statins, with ASCVD or diabetes, an LDL-C ≥ 190 mg/dL or < 70 mg/dL, and 80 years of age or older.

Howard,

Neuroepidemiology

2005; Safford, JAMA 2012

Methods – Statistical Methods

We calculated 10-year ASCVD

predicted

risk at

baseline using

the race-sex specific

Pooled Cohort risk

equations.

Participants were stratified by

decile

of 10-year predicted risk.

Calibration

was analyzed

by comparing

observed and predicted

number of ASCVD

events at 5

years:

We used

a modified

Hosmer-Lemeshow

test.

A chi-square >20 or p-value <0.05 indicates poor calibration.

Discrimination

was

analyzed:

We used the

C-index

.

A

C-index

between 0.70 and 0.80 is

good

and ≥0.80

is excellent.

Pooled Cohort risk equations

S0(t) at 5 years*S0(t) at 10 yearsMean scoreBlack women0.981940.953386.61White women0.988980.9665-29.18Black men0.957260.895419.54White men0.962540.914461.18

* Personal communication (Coady, S).

Goff, et al. Circulation 2013

Individual score calculation is based on:AgeTotal cholesterolHDL cholesterolSystolic blood pressureUse of antihypertensive medicationCurrent smokerDiabetes

HDL: high-density lipoprotein

Methods – Two sets of outcomes were evaluated

ASCVD outcomes – Non-fatal myocardial infarction, CHD death or non-fatal or fatal stroke.Adjudicated outcomes - Participant were contacted every 6 months and self-reported events were adjudicated. Surveillance outcomes - Medicare claims were searched for myocardial infarction and stroke events:Limited to participants 65 years of age and older.Medicare Part A coverage requiredOutcomes identified using validated algorithms.Outcomes were available through December 31, 2010.

Kiyota

, Am Heart J 2004,

Tirschwell

, Stroke 2002

Flowchart of participants included in the analysis

†

Defined

by use of

digoxin.

‡

Or non-HDL-C of 100 -

219 mg/

dL

for those without

a valid LDL-C

measurement.

Baseline characteristics of participants

Clinically relevant population

(n=10,997)

Medicare-linked

population

(n=3,333)

Age (years), mean (SD)

62

(

8)

71

(

4)

Blacks, n (%)

4,132 (

38)

1,095 (

33)

Men, n (%)

4,480 (

41)

1,476 (

44)

Current smoking, n (%)

1,626 (

15)

348 (10)

SBP (mmHg), mean (SD)

124.8 (

16)

128.3 (

16)

Antihypertensive

med,

n (%)

4,134 (

38)

1,491 (

45)

Total-C

(mg/

dL

), mean (SD)

203

(

31)

202

(

31)

HDL-C

(mg/

dL

), mean (SD)

54

(

17)

55

(

17)

The

clinically relevant population included those not taking statins, without ASCVD or diabetes, and with LDL-C 70 to 189 mg/

dL

.

SD: standard deviation; SBP

: systolic blood pressure;

HDL

: high density lipoprotein.

Calibration – Clinically relevant population

Results – Calibration and discrimination in the clinically relevant population

Events / person-years

Events in 5-years

5-year incidence

rate

*

Discrimination

Observed

†

Predicted

Observed

†

(

95% CI)

Predicted

C-index

(95% CI)

10-year predicted risk

Clinically relevant

population

0.72

(0.70-0.75)

<5%

28 / 14,816

32

33

1.9 (1.3-2.7)

1.9

5% to <7.5%

32 / 6,866

38

38

4.8 (3.4-6.7)

4.8

7.5% to <10%

34 / 5,853

41

46

6.1 (4.4-8.6)

6.9

≥10%

244 / 19,946

278

350

12.0 (10.6-13.6)

15.1

REGARDS participants without diabetes, with LDL-C 70 to 189 mg/

dL

‡ who were not taking statins are included in this table.

95% CI: 95% confidence interval. HDL-C: high density lipoprotein cholesterol; KM: Kaplan-Meier; LDL-C: low density lipoprotein cholesterol;

REGARDS:

REasons

for Geographic And Racial Differences in Stroke.

* Per 1,000 person-years.

† Kaplan-Meier adjusted.

‡ Or non-HDL-C of 100 - 219 mg/

dL

for those without a valid LDL-C measurement.

Calibration – REGARDS Medicare-linked population

Results - Calibration and discrimination for REGARDS Medicare-linked population

Events / person-years

Events in 5-years

5-year incidence

rate*

Discrimination

Observed

†

Predicted

Observed

†

(

95% CI)

Predicted

C-index

(95% CI)

10-year

predicted risk

Medicare linked

participants

0.67

(0.64-0.71)

5% to <7.5%

(Suppressed

)

9

7

5.3 (2.8-10.1)

4.0

7.5% to <10%

(

Suppressed)

15

12

7.9 (4.6-13.5)

6.4

≥10%

212 / 11,754

226

215

17.4 (15.3-19.8)

16.4

REGARDS participants without diabetes, with LDL-C 70 to 189 mg/

dL

‡ who were not taking statins are included in this table.

Suppressed – Medicare data are not presented in these cells due to a small sample size.

95% CI: 95% confidence interval. HDL-C: high density lipoprotein cholesterol; KM: Kaplan-Meier; LDL-C: low density lipoprotein cholesterol;

REGARDS:

REasons

for Geographic And Racial Differences in Stroke.

* Per 1,000 person-years.

† Kaplan-Meier adjusted.

‡ Or non-HDL-C of 100 - 219 mg/

dL

for those without a valid LDL-C measurement.

Conclusion

In this cohort of US adults for whom statin initiation

may be considered

based on

10-year predicted ASCVD risk:

Observed

and predicted 5-year ASCVD risks were similar indicating that these risk equations were well

calibrated.

Discrimination was moderate/good.

Previous results of over-estimation of ASCVD risk are likely due

to incomplete capture of ASCVD events and inclusion of participants taking statins

.

The current study supports the validity of the Pooled Cohort risk equations to inform clinical management decisions.

P Muntner and coauthors

Validation of the Atherosclerotic Cardiovascular Disease Pooled Cohort Risk Equations

Published online March 29, 2014

Available at www.jama.com and also at mobile.jamanetwork.com

jamanetwork.com