Inspiration

Determining the impact of chronically used medications during early pregnancy on perinatal outcomes is a complex issue. Some individuals planning a pregnancy may choose to discontinue their long-term medications due to concerns about potential risks to the fetus, while others may continue treatment but switch to safer alternatives. Additionally, some individuals may initiate new medications to manage their chronic condition, either due to worsening symptoms during pregnancy or improved access to healthcare. The early stages of pregnancy are often underrepresented in clinical trials due to the difficulties in recruiting large, diverse patient populations for this period. To address these critical gaps, well-designed observational studies using real-world data are essential for generating evidence to guide clinical care and optimize outcomes for both mothers and their infants.

Cool methodological features

Group-based trajectory modeling is one of the unsupervised clustering methods to group individuals with similar longitudinal data patterns. It models the probability of being assigned to a given trajectory group and the trajectories as a polynomial function of time. This method can identify groups with similar patterns in medication trajectories in the pregnant population and provide more granular information than a dichotomized status, i.e., determined as fill or no fill of a prescription, to describe utilization patterns. For example, we may observe groups without a prescription fill in the first trimester showing different patterns—one group discontinuing medication before pregnancy and another discontinuing in the later first trimester. Determining treatment trajectories will help us better understand the discontinuation or switching patterns during pregnancy in persons with pre-existing HTN. 

To determine whether switching to preferred therapy by the first trimester reduces the risk of pregnancy loss, we will use the clone-censoring-weight method. This approach addresses the immortal time bias that arises from the post-eligibility assignment—two treatment strategies (switching vs. non-switching) are not distinguishable at the start of the follow-up. Additionally, persons who experience pregnancy loss earlier in the pregnancy have less opportunity to be assigned as switchers. The cloning process allows each person to contribute to both treatment arms until their treatment strategy becomes evident. At that point, the individual’s person-time is censored in the arm, inconsistent with their observed treatment. After that, the inverse probability of censoring weights is assigned to account for the selection bias introduced by censoring. 

Impact

Currently, labetalol and nifedipine are the preferred medications for managing chronic hypertension in pregnancy. Findings from our study will improve the understanding of their comparative effectiveness and safety, providing evidence to guide treatment decisions for individuals with pre-existing hypertension. Additionally, by evaluating real-world utilization patterns and outcomes associated with switching to preferred therapies in the first trimester, the research will clarify the role of pre-conception counseling and timely prenatal care in optimizing treatment strategies.

Next steps

After completing my dissertation, I plan to pursue an academic career. My broader research interests include drug benefit-risk assessment and the evaluation of drug safety during pregnancy, evaluating the intended and unintended consequences of drug policy and its effects on population health, understanding the inappropriate use of prescription drugs, and studying the long-term effectiveness and safety of emerging therapies, including biologics and small molecules, for autoimmune diseases. I am also passionate about leveraging data visualization techniques to enhance the communication and dissemination of complex findings to clinicians, policymakers, and the general public.