External validation and comparison of four cardiovascular risk prediction models with data from the Australian Diabetes, Obesity and Lifestyle study

Loai Albarqouni, Jennifer A Doust, Dianna Magliano, Elizabeth Lm Barr, Jonathan E Shaw, Paul P Glasziou

Research output: Contribution to journalArticleResearchpeer-review

Abstract

OBJECTIVES: To evaluate the performance of the 2013 Pooled Cohort Risk Equation (PCE-ASCVD) for predicting cardiovascular disease (CVD) in an Australian population; to compare this performance with that of three frequently used Framingham-based CVD risk prediction models.

DESIGN: Prospective national population-based cohort study.

SETTING: 42 randomly selected urban and non-urban areas in six Australian states and the Northern Territory.

PARTICIPANTS: 5453 adults aged 40-74 years enrolled in the Australian Diabetes, Obesity and Lifestyle study and followed until November 2011. We excluded participants who had CVD at baseline or for whom data required for risk model calculations were missing.

MAIN OUTCOME MEASURES: Predicted and observed 10-year CVD risks (adjusted for treatment drop-in); performance (calibration and discrimination) of four CVD risk prediction models: 1991 Framingham, 2008 Framingham, 2008 office-based Framingham, 2013 PCE-ASCVD.

RESULTS: The performance of the 2013 PCE-ASCVD model was slightly better than 1991 Framingham, and each was better the two 2008 Framingham risk models, both in men and women. However, all four models overestimated 10-year CVD risk, particularly for patients in higher deciles of predicted risk. The 2013 PCE-ASCVD (7.5% high risk threshold) identified 46% of men and 18% of women as being at high risk; the 1991 Framingham model (20% threshold) identified 17% of men and 2% of women as being at high risk. Only 16% of men and 11% of women identified as being at high risk by the 2013 PCE-ASCVD experienced a CV event within 10 years.

CONCLUSIONS: The 2013 PCE-ASCVD or 1991 Framingham should be used as CVD risk models in Australian. However, the CVD high risk threshold for initiating CVD primary preventive therapy requires reconsideration.

LanguageEnglish
JournalThe Medical journal of Australia
DOIs
Publication statusE-pub ahead of print - 18 Jan 2019

Fingerprint

Life Style
Obesity
Cardiovascular Diseases
Northern Territory
Calibration
Population
Cohort Studies

Cite this

@article{71354d554ff2407fbc704b4e6a445835,
title = "External validation and comparison of four cardiovascular risk prediction models with data from the Australian Diabetes, Obesity and Lifestyle study",
abstract = "OBJECTIVES: To evaluate the performance of the 2013 Pooled Cohort Risk Equation (PCE-ASCVD) for predicting cardiovascular disease (CVD) in an Australian population; to compare this performance with that of three frequently used Framingham-based CVD risk prediction models.DESIGN: Prospective national population-based cohort study.SETTING: 42 randomly selected urban and non-urban areas in six Australian states and the Northern Territory.PARTICIPANTS: 5453 adults aged 40-74 years enrolled in the Australian Diabetes, Obesity and Lifestyle study and followed until November 2011. We excluded participants who had CVD at baseline or for whom data required for risk model calculations were missing.MAIN OUTCOME MEASURES: Predicted and observed 10-year CVD risks (adjusted for treatment drop-in); performance (calibration and discrimination) of four CVD risk prediction models: 1991 Framingham, 2008 Framingham, 2008 office-based Framingham, 2013 PCE-ASCVD.RESULTS: The performance of the 2013 PCE-ASCVD model was slightly better than 1991 Framingham, and each was better the two 2008 Framingham risk models, both in men and women. However, all four models overestimated 10-year CVD risk, particularly for patients in higher deciles of predicted risk. The 2013 PCE-ASCVD (7.5{\%} high risk threshold) identified 46{\%} of men and 18{\%} of women as being at high risk; the 1991 Framingham model (20{\%} threshold) identified 17{\%} of men and 2{\%} of women as being at high risk. Only 16{\%} of men and 11{\%} of women identified as being at high risk by the 2013 PCE-ASCVD experienced a CV event within 10 years.CONCLUSIONS: The 2013 PCE-ASCVD or 1991 Framingham should be used as CVD risk models in Australian. However, the CVD high risk threshold for initiating CVD primary preventive therapy requires reconsideration.",
author = "Loai Albarqouni and Doust, {Jennifer A} and Dianna Magliano and Barr, {Elizabeth Lm} and Shaw, {Jonathan E} and Glasziou, {Paul P}",
note = "{\circledC} 2019 AMPCo Pty Ltd.",
year = "2019",
month = "1",
day = "18",
doi = "10.5694/mja2.12061",
language = "English",
journal = "Medical Journal of Australia",
issn = "0025-729X",
publisher = "AUSTRALASIAN MED PUBL CO LTD",

}

External validation and comparison of four cardiovascular risk prediction models with data from the Australian Diabetes, Obesity and Lifestyle study. / Albarqouni, Loai; Doust, Jennifer A; Magliano, Dianna; Barr, Elizabeth Lm; Shaw, Jonathan E; Glasziou, Paul P.

In: The Medical journal of Australia, 18.01.2019.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - External validation and comparison of four cardiovascular risk prediction models with data from the Australian Diabetes, Obesity and Lifestyle study

AU - Albarqouni, Loai

AU - Doust, Jennifer A

AU - Magliano, Dianna

AU - Barr, Elizabeth Lm

AU - Shaw, Jonathan E

AU - Glasziou, Paul P

N1 - © 2019 AMPCo Pty Ltd.

PY - 2019/1/18

Y1 - 2019/1/18

N2 - OBJECTIVES: To evaluate the performance of the 2013 Pooled Cohort Risk Equation (PCE-ASCVD) for predicting cardiovascular disease (CVD) in an Australian population; to compare this performance with that of three frequently used Framingham-based CVD risk prediction models.DESIGN: Prospective national population-based cohort study.SETTING: 42 randomly selected urban and non-urban areas in six Australian states and the Northern Territory.PARTICIPANTS: 5453 adults aged 40-74 years enrolled in the Australian Diabetes, Obesity and Lifestyle study and followed until November 2011. We excluded participants who had CVD at baseline or for whom data required for risk model calculations were missing.MAIN OUTCOME MEASURES: Predicted and observed 10-year CVD risks (adjusted for treatment drop-in); performance (calibration and discrimination) of four CVD risk prediction models: 1991 Framingham, 2008 Framingham, 2008 office-based Framingham, 2013 PCE-ASCVD.RESULTS: The performance of the 2013 PCE-ASCVD model was slightly better than 1991 Framingham, and each was better the two 2008 Framingham risk models, both in men and women. However, all four models overestimated 10-year CVD risk, particularly for patients in higher deciles of predicted risk. The 2013 PCE-ASCVD (7.5% high risk threshold) identified 46% of men and 18% of women as being at high risk; the 1991 Framingham model (20% threshold) identified 17% of men and 2% of women as being at high risk. Only 16% of men and 11% of women identified as being at high risk by the 2013 PCE-ASCVD experienced a CV event within 10 years.CONCLUSIONS: The 2013 PCE-ASCVD or 1991 Framingham should be used as CVD risk models in Australian. However, the CVD high risk threshold for initiating CVD primary preventive therapy requires reconsideration.

AB - OBJECTIVES: To evaluate the performance of the 2013 Pooled Cohort Risk Equation (PCE-ASCVD) for predicting cardiovascular disease (CVD) in an Australian population; to compare this performance with that of three frequently used Framingham-based CVD risk prediction models.DESIGN: Prospective national population-based cohort study.SETTING: 42 randomly selected urban and non-urban areas in six Australian states and the Northern Territory.PARTICIPANTS: 5453 adults aged 40-74 years enrolled in the Australian Diabetes, Obesity and Lifestyle study and followed until November 2011. We excluded participants who had CVD at baseline or for whom data required for risk model calculations were missing.MAIN OUTCOME MEASURES: Predicted and observed 10-year CVD risks (adjusted for treatment drop-in); performance (calibration and discrimination) of four CVD risk prediction models: 1991 Framingham, 2008 Framingham, 2008 office-based Framingham, 2013 PCE-ASCVD.RESULTS: The performance of the 2013 PCE-ASCVD model was slightly better than 1991 Framingham, and each was better the two 2008 Framingham risk models, both in men and women. However, all four models overestimated 10-year CVD risk, particularly for patients in higher deciles of predicted risk. The 2013 PCE-ASCVD (7.5% high risk threshold) identified 46% of men and 18% of women as being at high risk; the 1991 Framingham model (20% threshold) identified 17% of men and 2% of women as being at high risk. Only 16% of men and 11% of women identified as being at high risk by the 2013 PCE-ASCVD experienced a CV event within 10 years.CONCLUSIONS: The 2013 PCE-ASCVD or 1991 Framingham should be used as CVD risk models in Australian. However, the CVD high risk threshold for initiating CVD primary preventive therapy requires reconsideration.

U2 - 10.5694/mja2.12061

DO - 10.5694/mja2.12061

M3 - Article

JO - Medical Journal of Australia

T2 - Medical Journal of Australia

JF - Medical Journal of Australia

SN - 0025-729X

ER -