# Prediction of cardiovascular disease risk in women and individuals with polycystic ovary syndrome using the American Heart Association PREVENT model: A long-term population-based cohort study

## TL;DR

The study evaluated how well the PREVENT model predicts cardiovascular disease risk in Iranian women, including those with PCOS, finding it performs well but underestimates risk in older or high-risk individuals.

## Contribution

The study is the first to validate the PREVENT model's performance in predicting cardiovascular disease risk in Iranian women and those with PCOS.

## Key findings

- The PREVENT model showed good discrimination (C-statistic 0.84) and acceptable calibration for predicting 10-year CVD risk in Iranian women.
- Model performance was highest in women under 55 years, with some underestimation in older or high-risk individuals.
- The model maintained good discrimination in women with PCOS/isolated PCOS phenotypes (C-statistic 0.79) and non-PCOS controls (C-statistic 0.85).

## Abstract

This study aimed to assess the performance of the PREVENT risk model, in terms of discrimination and calibration in the overall female sample representative of the Iranian general population, across age groups, and among women with PCOS.

In this population-based prospective study, we used data from phases 3–7 of the Tehran Lipid and Glucose Study (2006–2021). A total of 3983 women aged 30–79 years without baseline CVD and with complete data for the PREVENT risk score were included, of whom 2117 had known PCOS or isolated PCOS phenotypes status. The PREVENT risk model was applied to estimate 10-year CVD risk. The primary outcome was incident CVD, which was defined as a composite of fatal and nonfatal atherosclerotic cardiovascular disease (ASCVD) and heart failure (HF), identified follow-ups. Model discrimination was evaluated using Harrell’s C-statistics and time-dependent AUC, and calibration was assessed by comparing predicted and observed 10-year CVD risks in general female population, by age group, and by PCOS status.

Among 3983 women (median follow-up 12.2 years; mean age 47.6 years; mean BMI 28.6 kg/m²), 911 had confirmed PCOS/Isolated-PCOS phenotypes. In the overall population, the PREVENT risk model demonstrated good discrimination (C-statistic 0.84, 95 % CI: 0.82–0.86; AUC 0.80) and satisfactory calibration in lower- and mid-risk deciles, with some underestimation at the highest risk deciles. Discrimination declined with increasing age, performing best in women aged 30–44 years (C-statistic 0.82, AUC 0.83). Stratified by PCOS status, the model maintained good discrimination in women with PCOS/isolated PCOS phenotypes (C-statistic 0.79, AUC 0.81) and non-PCOS/ non-isolated PCOS phenotypes controls (C-statistic 0.85, AUC 0.81), with calibration remaining satisfactory across subgroups, though high-risk quartiles in older women were slightly underestimated.

In this long-term, population-based cohort of Iranian women, the PREVENT risk model showed good discrimination and acceptable calibration for predicting 10-year CVD risk in both the general population and women with PCOS/Isolated-PCOS phenotypes. Performance was highest in women under 55 years, with some underestimation of risk in older or high-risk individuals.

Image, graphical abstract