COVID vaccines may be hard to find this season. Credit: Adobe Stock
COVID-19’s grip may have loosened, but its effects on pregnancies linger in data showing higher preterm birth risks for infected mothers. A new analysis from Panalgo, a Norstella company, presented at ICPE 2025, used linked U.S. claims data to match COVID-positive pregnant women against controls. The data revealed a statistically significant increase in preterm deliveries; lower rates of low birth weight were observed, with significance not specified.
More data will be needed to interpret the findings amid conflicting prior studies. “This was a starting point for looking at more longer-term impacts of a mother’s COVID-19 infection on the outcome of their children,” said Mike Munsell, Ph.D., director of real-world data and HEOR at Panalgo. He noted a rise in preterm births among infected mothers and, counterintuitively, lower rates of low birth weight, signals the team plans to track as the “children of the COVID era” move through early milestones.
Digging into the COVID-19 pregnancy data
Mike Munsell, Ph.D.
The ICPE 2025 poster, “The Effect of COVID-19 During Pregnancy on Fetal Outcomes: A Propensity Score-matched Analysis Using Mother-infant Linked Claims Data,” taps Norstella’s mother-infant linkage to examine how infection during pregnancy relates to delivery outcomes in the real world. The team matched mothers with a documented COVID-19 infection during pregnancy to contemporaneous mothers without recorded infection and controlled for the number of vaccinations and other relevant factors.
“At Norstella, we have very large datasets that we can link at the patient level,” said Munsell. “One of the ones that we have links mothers to their infants, and you can connect the characteristics of the mother with the outcomes of the infant.”
The core signal is straightforward on one dimension and puzzling on another. “Mothers who had COVID-19 during pregnancy were more likely to have a preterm birth, which was statistically significant compared to controls,” Munsell said.
At the same time, he noted lower rates of low birth weight among infants born to infected mothers. “That one’s a little bit of a puzzle,” he said. The literature varies, and the team plans to incorporate additional covariates. “There’s a lot of noise in this type of data,” he said, pointing to the need to broaden socio-demographic information to better understand fetal outcomes. “There’s a lot of value in being clear on the limitations of your data and the population you are studying. If you can be as clear and upfront about what you’re studying, that’s probably more practical and beneficial than being as general as possible.”
The children of the COVID era
Methodologically, the analysis relied on structured administrative claims and enrollment files to maximize completeness and establish clear timelines. “This one used structured claims data. We had enrollment files, which are particularly helpful for understanding the timeline for pregnancy,” Munsell said. “We identify the point in time when a child is born and link that with the mother’s record, and then approximate when the start of pregnancy was.” The same ecosystem allows deeper dives when needed. “We do have the ability within our dataset to tokenize more data and bring in a smaller sample with overlapping EMR and clinical notes and some of the text fields,” he said. He put the mother-infant linkage at “close to around 3 million mother-infant pairs,” drawn from “close to 170 million covered lives.”
Beyond the immediate results, Munsell presented the work as a starting point. “We’re at an interesting phase now. The pandemic has subsided, but a lot of work is happening on the ‘children of the COVID era.’ I have one as well; they’re entering kindergarten this year,” he said. “We would like to follow these infants further in the dataset” to clarify mechanisms behind the preterm-birth signal and to test whether the low-birth-weight finding persists after incorporating broader social and clinical factors.
The other ICPE poster: ES-SCLC in the real world
Munsell’s team also presented on extensive-stage small cell lung cancer (ES-SCLC ), analyzing U.S. treatment patterns, resource use, and outcomes. “Yes. That one also used administrative claims data,” he said. “Extensive stage is fairly aggressive, but there have been a lot of developments in treatment in the last five years, specifically immune checkpoint inhibitors becoming a first-line therapy along with chemotherapy.”
“Part of our work was looking at the rate for which, within this fairly recent dataset, you’re seeing patients receiving both of those as a first-line treatment,” Munsell said. “Close to 40% of our population had these newer treatments as a first-line therapy along with their chemotherapy. That was reassuring, to see that shortly after the approval and recommendation, patients were using those treatments.”
Despite uptake, burden remained high. “We’re still seeing close to two-thirds of patients having brain metastases at a certain point and over a third having a secondary primary malignancy, even with these new treatments,” he said. “So, this was mostly a study looking at the landscape of treatments available to these patients, but also still highlighting that the burden of illness is still pretty high.”
“One newer area that is just recently being looked at for these patients is radiation therapy happening further down the lines of treatment,” Munsell said. “We saw a smaller subset, about 15% of our patients, also receiving radiation therapy.” He added, “There was a recent sub-analysis from a clinical trial that referenced that, potentially, what they’re seeing in men is a potential increase in overall survival with radiation therapy. I don’t know the full mechanisms for why that is, and they’re still trying to confirm that. So we looked into whether we see a high proportion of men receiving radiation therapy. At this point in the data, it’s fairly equal among men and women receiving that treatment.”
“For this one, we also have the ability to link in mortality data,” Munsell said. “The hope with this cohort that we first highlighted was to confirm that the burden of illness is still relatively high, confirm that these new treatments are being used, and also confirm that there are still pretty severe outcomes. We can then follow these patients over time and do overall survival analysis, but the hope is also to pull in clinical notes to get a better description around disease staging at the time of index, which really impacts overall survival.” For now, he added, “We’re kind of going based on diagnosis codes right now, and treatment being indicative of the type of disease, which is a good starting point. But pulling in unstructured clinical notes, particularly for oncology, goes a really far way, because each of these patients has a very unique and individualized disease.”
Filed Under: clinical trials, Drug Discovery, Infectious Disease, machine learning and AI