Added Value of Female-Specific Factors Beyond Traditional Predictors for Future Cardiovascular Disease

In women, sex-specific factors related to hormonal and reproductive status are known to affect cardiovascular disease (CVD) risk (1). It is unknown whether female-specific factors have added value to the traditional risk factors for the purpose of predicting future CVD risk. Therefore, we aimed to evaluate the added value of female-specific factors in addition to traditional risk factors for the prediction of 10-year risk of CVD in women.

We used data collected between 1993 and 1997 in women 30 to 74 years of age from 2 Dutch population–based cohort studies (PROSPECT, MORGEN [Monitoring Project on Risk Factors for Chronic Diseases]) (2). Inclusion criteria for the present analysis were the following: known pregnancy status (never/ever), consent to linkage with disease registries, complete information on vital status and cardiovascular events, no previous CVD. In total, 15,922 women from the PROSPECT study, of whom 14,069 were ever pregnant, and 8,873 women from the MORGEN study, of whom 7,216 were ever pregnant, were included in the analyses. Missing values in candidate predictors were multiply imputed. Outcome was the 10-year risk of (non)fatal CVD. Women were classified into a risk category based on their predicted risk, using cutoff values according to the European Society of Cardiology 2007 treatment guidelines: <10% (low risk), ≥10% to <20% (intermediate risk), ≥20% (high risk). Female-specific factors (age at menarche, menopausal status/age, hormone use, gestational hypertension and diabetes, number of children, miscarriages/stillbirths) were added to the traditional predictors (age, diabetes, blood pressure, cholesterol, and smoking) using Cox proportional hazards models. We applied stepwise backward selection of the female-specific risk factors based on Akaike’s Information Criterion, forcing the traditional risk predictors to be retained in the final model. Reproductive status variables were investigated in ever-pregnant women only. Improvement by adding female-specific risk factors to traditional predictors was assessed in terms of discrimination (Harrell’s C-statistic), calibration (visualization of observed/predicted ratio in calibration plot), and net reclassification improvement (NRI).

In the PROSPECT study, the mean age was 58 ± 6 years and the mean systolic blood pressure was 133 ± 20 mm Hg, and 22% smoked. In the MORGEN study, the mean age was 46 ± 9 years, the mean systolic blood pressure was 119 ± 17 mm Hg, and 35% smoked. During a median follow-up of 11.7 years, 1,605 (PROSPECT, 10%) and 551 (MORGEN, 6%) CVD events occurred. On multivariable analysis, all traditional predictors were associated with an increased risk of CVD in both cohorts as well as early menopause and multiple miscarriages (Table 1). The C-statistics of the model with traditional risk factors (model 1) were 0.70 (95% confidence interval [CI]: 0.67 to 0.73) in the PROSPECT study, 0.70 (95% CI: 0.67 to 0.73) in the PROSPECT study ever pregnant women, 0.72 (95% CI: 0.67 to 0.73) in the MORGEN study, and 0.72 (95% CI: 0.67 to 0.77) in the MORGEN study ever pregnant women. The C-statistics with the retained female-specific factors (model 2) were 0.70 (95% CI: 0.68 to 0.73) in the PROSPECT study, 0.70 (95% CI: 0.67 to 0.73) in the PROSPECT study ever pregnant women, 0.72 (95% CI: 0.67 to 0.76) in the MORGEN study, and 0.73 (95% CI: 0.67 to 0.78) in the MORGEN study ever pregnant women. Including female-specific factors did not improve calibration of the models. Comparing models 1 and 2, the categorical NRI was −0.01 (95% CI: −0.02 to 0.00) in the PROSPECT study, −0.00 (95% CI: −0.02 to 0.01) in the PROSPECT study ever pregnant women, −0.01 (95% CI: −0.03 to 0.02) in the MORGEN study, and −0.01 (95% CI: −0.03 to 0.02) in the MORGEN study ever pregnant women.

Female-specific factors had no added value in additional to traditional risk factors for the prediction of the 10-year risk of CVD in women, nor did they improve calibration. However, some were associated with CVD risk on multivariable analysis and therefore certainly help to shed light on the biology and pathophysiology of CVD in women.

Female-specific factors may partly increase the risk of CVD through increasing the levels of the traditional cardiovascular risk factors (3) and partly act via independent mechanisms. However, the observed independent effect of early menopause and multiple miscarriages is too small to contribute to discrimination of women in whom CVD will and will not develop.

The strengths of our study include the size, the large number of endpoints, availability of a large number of female-specific factors, information on pregnancy complications, and the fact that it was population based. Furthermore, coefficients for the traditional risk factors were optimally fitted to our data instead of using published coefficients to avoid the risk of erroneously attributing a poor fit of traditional risk scores in these data to the added value of female-specific factors.

Several limitations of this study should be addressed. The presence of female-specific factors was assessed using self-administered questionnaires, which could have led to misclassification. No information about the presence of polycystic ovary syndrome, intrauterine growth restriction, or birth weight was collected.

Although female-specific factors are associated with CVD risk, they have no added value in addition to traditional predictors for the prediction of 10-year risk of CVD in women.

• American College of Cardiology Foundation