Key takeaways:
- The TriNetX dataset cannot determine whether predelivery GLP-1 RA prescriptions reflect inadvertent pregnancy exposure, planned washout periods, or continued first-trimester use — a key limitation for clinical interpretation.
- Prescriptions from cash-pay weight loss clinics and non-Epic-integrated systems are not captured, meaning the study likely underestimates total perinatal GLP-1 RA exposure.
- Investigators characterize the study as a high-level view of a now well-established phenomenon, with the next research priority being outcomes data on fetal and maternal effects in patients with documented periconceptional exposure.
The rapid rise in perinatal GLP-1 receptor agonist prescribing documented in a recent national EHR database study raises as many methodologic questions as clinical ones, according to a recent interview with study investigators.1
With the core prescribing trends covered in the first installment, Kevin Y. Xu, MD, MPH, and Jeannie C. Kelly, MD, MS, addressed the interpretive limitations that will shape how clinicians and researchers build on these findings.
A central question is whether the prescriptions captured in the predelivery window reflect inadvertent exposure during early pregnancy, planned discontinuation before conception, or continued use into the first trimester. The data cannot distinguish among those scenarios.
"From the data itself, we can't tell if these were inadvertent exposures, people purposely knowing that they're going to get pregnant coming off of these medications prior to pregnancy, or continuation of the medication through the first trimester," Kelly said.
Current guidelines across professional societies recommend discontinuing GLP-1 RAs 1 to 2 months before a planned conception and stopping immediately upon discovering an unplanned pregnancy. Kelly used that clinical context to inform her interpretation of the prescribing patterns.
“I don't think that there are that many physicians and clinicians out there prescribing GLP-1 medications purposefully throughout pregnancy," she said, suggesting the observed predelivery prescriptions likely reflect planned washout periods or inadvertent early-pregnancy exposure rather than intentional continuation.
Xu acknowledged the dataset's structural constraints. The TriNetX platform captures prescriptions integrated into the Epic EHR ecosystem, meaning prescriptions filled through cash-pay weight loss clinics or other non-integrated systems would not be reflected. The database also cannot confirm whether patients were actually taking the medication as prescribed, whether postpartum patients were breastfeeding, or at precisely what gestational age any given prescription was active.
"We have this balance between massive scope of the data and some payoff in terms of the data being more coarse," Xu said. "I like to see this as we have a thirty-thousand-foot view, and it's now time to dive quite a bit deeper."
On the other side of that tradeoff, Xu emphasized the precision of the delivery timing data. Because deliveries and procedures are entered directly into the medical record with date and time stamps, the study's exposure windows are anchored to reliable clinical events—an advantage over cohort studies relying on patient self-report. The result, he noted, is likely the largest GLP-1 prescribing trend study conducted in the US to date.
Both investigators framed the study as establishing the scale of a phenomenon that now demands more granular investigation—particularly outcomes data on fetal and maternal effects in patients with documented periconceptional or first-trimester exposure.
Reference:
1. Lessard C, Cary C, In A, et al. Prescribing Trends in glucagon-like peptide-1 medications among pregnant and postpartum persons. Obstet Gynecol. 2026;147(3):290-292. doi:10.1097/AOG.0000000000006161
