Abstract P188: The Pooled Cohort Equations Over-predict Risk in Women With Hypertensive Disorders of Pregnancy, Even After Refitting the Model
Introduction: The AHA guidelines for the Prevention of Cardiovascular Disease (CVD) in Women describe hypertensive disorders of pregnancy (HDP) as a failed stress test, which might unmask early CVD. An abundance of prediction models for CVD risk is available for the general population, but their validity in women with HDP is not established.
Hypothesis: The prognostic performance of the Pooled Cohort Equations (PCE) is lower in women with HDP compared to women without HDP and recalibrating and refitting the model will improve the prognostic performance.
Methods: Data were used from 27,339 women out of the MORGEN and PROSPECT cohorts; we excluded those who had never been pregnant. In total, 5,358 answered the question: ‘Did you suffer from high blood pressure during pregnancy?’ with ‘Yes’; and 15,266 with ‘No’. Outcome definition was equal to that in the original PCE model. MORGEN and PROSPECT were analyzed separately, because of differences in characteristics (e.g. MORGEN is younger and has more current smokers) and observed risks. First, we calculated the 10-year predicted risk and compared this with the observed risk. Subsequently, the model was updated in three steps: by recalibrating the mean linear predictor, by additionally updating the baseline hazard, and by refitting the full model. The performance of all models was quantified by calibration (calibration plot, expected:observed ratio) and discrimination (c-statistic).
Results: The Table shows that the original model over-predicts risk in all women, but more in women without HDP. Calibration plots improved most after refitting, which is confirmed by the expected:observed ratio, although the model still over-predicts. Refitting only improved discrimination in women with HDP, but not in women without HDP.
Conclusion: The PCE over-predicts risk in women with and without HDP, even after refitting the model. Discrimination is overall quite good, except for MORGEN women without HDP. Especially in women with HDP the model discrimination benefits from refitting.
Author Disclosures: V. Dam: None. N.C. Onland-Moret: None. W.M.M. Verschuren: None. J.M.A. Boer: None. K.G.M. Moons: None. E. Boersma: None. Y.T. van der Schouw: None.
- © 2017 by American Heart Association, Inc.