Key result
Common 10-year risk scores substantially underpredict ASCVD risk by up to ~5-fold despite good discrimination.
Why the study?
Cardiovascular risk scores must be calibrated to their target populations to guide preventive treatment, yet Arab populations are under-represented in score development and validation.
Do the ADRS, PREVENT-10y, PREVENT-30y, and SCORE2 cardiovascular risk scores accurately predict incident ASCVD in two Emirati cohorts, and do recalibration parameters transfer between community and voluntary screening settings?
Cohort (n=10,677)
Yes
Do the ADRS, PREVENT-10y, PREVENT-30y, and SCORE2 cardiovascular risk scores accurately predict incident ASCVD in two Emirati cohorts, and do recalibration parameters transfer between community and voluntary screening settings?
While ADRS, PREVENT, and SCORE2 discriminated cardiovascular risk well in Emirati adults, calibration varied drastically by screening setting and baseline statin treatment, demonstrating that risk scores in the UAE must be locally recalibrated within their specific setting of intended use rather than relying on imported or transferred parameters.
Answers are grounded in this paper and cite their sources.
Underprediction in voluntary screening cautions against unadjusted use; leaves open need for prospective validation of recalibrated scores in Emiratis.
Background Cardiovascular risk scores guide preventive treatment through absolute-risk thresholds, so they must be calibrated to the population in which they are used. Arab populations are under-represented in score development and validation. We assessed the calibration and recalibration of 4 atherosclerotic cardiovascular disease (ASCVD) scores in 2 Emirati cohorts. Methods The locally derived Abu Dhabi Risk Score (ADRS), PREVENT-10y, PREVENT-30y and SCORE2 were applied, each within its published age range, to a community screening cohort (C1; n = 8,123 eligible, screened 2010–2013, mean follow-up 9.2 years, 282 ASCVD events) and a voluntary screening cohort (C2; n = 2,554, 2016–2017, 6.75 years, 352 ASCVD events). Incident coronary artery disease (CAD) was analyzed for ADRS-CAD. Discrimination was assessed by the area under the curve and calibration by the observed-to-expected ratio and calibration slope. Scores were recalibrated by closed testing, and each cohort’s parameters were applied to the other. Calibration was also examined by baseline lipid-lowering treatment. Results Discrimination was acceptable to excellent: the area under the curve was 0.898 for ADRS (as ADRS was derived in C1), 0.837 for PREVENT-10y, 0.783 for PREVENT-30y and 0.768 for SCORE2 in C1, and 0.825, 0.808, 0.766 and 0.738 in C2. ASCVD incidence was 0.38 per 100 person-years in C1 against 2.25 in C2, and calibration depended on the cohort more than on the score. In C1, ADRS and SCORE2 were well calibrated (observed-to-expected [O/E] ratios 1.01 and 0.97) and PREVENT-10y underpredicted (1.58). In C2, all three underpredicted (3.43, 4.94 and 3.54). Recalibration parameters did not transfer between cohorts: applied to the other cohort, they left 3.2- to 3.5-fold underprediction or 0.29- to 0.33-fold overprediction. Among untreated participants in C2, the scores as published accounted for only 41–58% of observed events (O/E 1.73 to 2.42). Cohort-wide correction overshot to about two-fold overprediction in this group (0.45 to 0.46), while treated participants remained underpredicted (1.27 to 1.35). Applied at guideline thresholds without recalibration, the scores placed below those thresholds participants who accounted for up to four-fifths of subsequent events in C2. Conclusions ADRS, PREVENT and SCORE2 ranked cardiovascular risk well in Emirati adults, but none predicted absolute risk correctly in both cohorts. In the community cohort, ADRS and SCORE2 were calibrated, and PREVENT-10y underpredicted. In the voluntary screening cohort, every 10-year score underpredicted three- to five-fold. A correction estimated in one cohort did not transfer to the other. In the voluntary screening cohort, a single correction did not fit lipid-lowering-treated and untreated participants alike. For untreated adults, the group for whom the decision to start treatment is made, the published scores missed about half of events, and the corrected scores overstated risk about two-fold. Any score adopted in the UAE should be recalibrated locally in the setting of intended use, taking baseline treatment into account, checked within subgroups defined by sex, age, diabetes and treatment, and re-estimated periodically.
No takes yet. Share an insight, caveat, or question.
AlKetbi et al. (2026) conducted a cohort in Atherosclerotic cardiovascular disease (ASCVD) risk prediction (n=10,677). Cardiovascular risk prediction scores (ADRS, PREVENT-10y, PREVENT-30y, SCORE2) vs. Observed incident ASCVD events across two Emirati cohorts was evaluated on Incident atherosclerotic cardiovascular disease (coronary artery disease or stroke) calibration and discrimination. ADRS, PREVENT, and SCORE2 showed good discrimination in Emirati adults, but 10-year scores underpredicted risk 3.4- to 4.9-fold in voluntary screening.
Synapse has enriched 10 related papers. Consider them for comparative context: