(1207) Use of mean arterial blood pressure to predict superimposed preeclampsia with severe features

This was a secondary analysis of the Chronic Hypertension and Pregnancy (CHAP) trial. We excluded individuals with missing or invalid information regarding SIPE diagnosis. Our primary outcome was SIPE with severe features. We fitted Bayesian joint longitudinal and time-to-event models (linear and non-linear natural spline-based) using MAP from baseline and throughout pregnancy prior to the diagnosis of SIPE or censoring as repeated measures. A hazard ratio (HR) with a 95% credible interval was estimated. The fitted models were evaluated using the area under the receiver-operating curves (AUC) and Brier scores, estimated by 5-fold cross-validation 5 times. 


Results:

Of 2316 individuals with chronic hypertension, 600 (25.9%) developed SIPE with severe features. Compared to individuals without SIPE with severe features, those with SIPE with severe features had higher MAP, were more likely to be Black or smokers, and were more likely to have government insurance and diabetes (Table 1). Higher MAP was associated with an increased risk of SIPE with severe features (HR 1.15; 95% CI 1.12-1.19), indicating that each mmHg increase in MAP was associated with a 15% increase in the risk of SIPE with severe features. AUCs  (Table 2) were low-to-moderate, with better performance for intervals later in pregnancy and shorter in duration (e.g., 0.60 at 20-40 weeks to 0.71 at 34-37 weeks). Similarly, Brier scores were low-to-moderate indicating poor performance. Estimated HRs and prediction performance were nearly indistinguishable between linear and spline-based models.


Conclusion:

While repeated measures of MAP are associated with developing SIPE with severe features, the predictive ability of longitudinal MAP alone was fair to moderate. Future studies should consider longitudinal MAP in modeling the risk of SIPE with severe features, but additional predictors are necessary to improve predictive performances.