Abstract 74: Relationship Between Preeclampsia and Subsequent Stroke: A Secondary Analysis of the Framingham Heart Study

Abstract

Introduction: Although prior studies have examined the later-life risk of stroke in women with a history of preeclampsia, they have relied on claims data or epidemiological studies with poorly detailed risk factor or outcome data. The Framingham Heart Study (FHS) is a landmark epidemiological study that began in 1948 and followed participants aged 30-59 at baseline until death. The study enrolled 5,209 residents of Framingham, Massachusetts, almost half of which were female. The adjudicated outcome of stroke was ascertained during follow-up visits until 2015, death, or loss to follow-up. Methods: We examined a cohort of 2,136 women from FHS for an association of preeclampsia with stroke occurrence. With each subject having their own unique history of exposures and treatments that vary over time, standard approaches of adjusting for confounding are biased when time-dependent confounders exist. Using a marginal structural model, the effects of possible time-dependent confounders were controlled for using inverse-probability-of-treatment weighted estimators. At each of the study visits, potential confounders were evaluated. Variables controlled for, across the study visits, included age, blood pressure, cholesterol, weight, glucose, treatment for hypertension, treatment for hyperlipidemia, and smoking. Results: Two hundred women (9.4%) had preeclampsia and 380 women (17.8%) had strokes. As compared with the 1,936 women who did not have a history with preeclampsia, the 200 women with a history of preeclampsia had a significantly higher risk of stroke (adjusted Hazard Ratio, 2.29; 95% CI, 1.13 - 4.63) when controlling for time-dependent confounders. Conclusion: A history of preeclampsia is associated with a significantly increased risk of a future stroke. Prior approaches to analyses of this nature have not fully emphasized the importance of controlling for time-dependent confounders with methods such as marginal structural modeling. We plan on analyzing the mediators of the observed association to determine which risk factors might prove beneficial for disease-modifying treatment in midlife.