The study by Wambua et al., Development and validation of a postpartum cardiovascular disease risk prediction model in women incorporating reproductive and pregnancy-related predictors [3], recently in BMC Medicine, set out to determine if the addition of pregnancy complications and other female-specific conditions to the QRISK®−3 risk equation would improve the 10-year risk prediction of CVD in postpartum women. Of the women followed in this large retrospective cohort study, 0.38% were found to have had a cardiovascular event within 10 years of delivery. And while a number of pregnancy, reproductive, and other female-specific conditions were found to be associated with cardiovascular events even when traditional risk factors were taken into consideration, the addition of these conditions to the QRISK®−3 risk equation only resulted in marginally improved discrimination and calibration. Others have tried adding female-specific risk factors to other risk scores with similar findings [4].
Postpartum CVR screening for those that had one or more pregnancy complication is now recommended around the world [5]. Recognizing that the CVD risk prediction scoring tools that are widely used did not include reproductive-age women in their creation and validation, they are of limited use [6], and especially for the 30-year and lifetime-risk scores, likely underestimate the true risk of future CVD [7]. The minimal benefit of adding reproductive- and pregnancy-related conditions to the QRISK®−3 model seen in the study by Wambua et al. may in part be explained by the inclusion of nulliparous and parous women in the sample population used to develop these models. More specifically, the relative importance of these predictors likely differs by parity, as seen in a prior simulation study [8], and may warrant developing separate models by parity or models that incorporate the dynamic nature of predictors over time. The sensitivity analysis performed in the study by Wambua et al., among a subgroup of women with preeclampsia, which showed improved calibration, may highlight this point since preeclampsia is more prevalent in nulliparous women. The minimal benefit of female-specific conditions may also in part be explained by the modeling of pregnancy complications as independent factors, ignoring the correlation across and between pregnancies. For example, women who experience pregnancy complications over multiple pregnancies or multiple complications within a single pregnancy are at higher risk of CVD. Additionally, traditional risk factors, like age, interact with many of the reproductive- and pregnancy-related conditions, placing these women at higher risk of future CVD. Finally, the relatively short follow-up in this cohort (median 3.7 years, IQR 1.4, 7.8) resulted in a low event rate, which may have also contributed to the marginal improvements in performance of the models.