Performance of the revised WHO cardiovascular disease risk prediction models for the Middle East and North Africa: a validation study in the Tehran Lipid and Glucose Study.

Abstract

Objective: We aimed to evaluate the performance of the revised WHO models in predicting the 10-year risk of cardiovascular disease (CVD) in Iran, as part of the Middle East and North Africa (MENA) region.

Study design and setting: We analyzed data from the Tehran Lipid and Glucose Study (TLGS), including 5162 participants (2241 men) aged 40-80 years without CVD at baseline (the 3^rd examination, 2006-2008), for the occurrence of CVD (myocardial infarction, coronary heart disease death, and stroke). We assessed the statistical performance of original and regionally recalibrated models-both laboratory- and non-laboratory-based- using discrimination (C-statistic) calibration (calibration plot and observed-to-expected[O:E] ratio), and clinical performance applying net benefit (NB), a measure of true positives penalized for a weight of false positives, a decimal value representing the expected proportion of true positive outcomes among total population.

Results: During the 10-year follow-up, 307 CVD events occurred. The cumulative incidence of CVD was 9.0% (95% CI: 8.0-10.0%) in men and 4.0% (3.0-5.0%) in women. For the laboratory-based model, the C-statistic was 0.72 (0.68-0.75) in men and 0.83 (0.80-0.86) in women; for the non-laboratory-based model, it was 0.70 (0.66-0.73) and 0.82 (0.79-0.86), respectively. The lab model underpredicted the risk (O:E=1.20[1.00-1.33] for men and 1.40 [1.13-1.60] for women). At the risk threshold of 10%, NB for the lab model was 0.03 (0.02-0.04) for men and 0.01 (0.004-0.01) for women; these values became zero or negative for thresholds over 20%. Regionally recalibrated models overestimated the risk (O:E <1) and showed lower NB.

Conclusion: The loss of specificity was not sufficiently offset by the increase in sensitivity provided by the regionally recalibrated models compared to the original models.