The importance of studying treatments in pregnant people

Illustration: several pregnant individuals stand in a line next to each other, showing their side profiles

December 21, 2023 – When the COVID-19 pandemic hit in 2020, it became clear early on that infection was associated with higher risks in pregnant people. They experienced worse outcomes in terms of severe illness, ICU admissions, and death, as well as preterm births and stillbirths. And yet, vaccine developers excluded them from clinical trials, and governments approved vaccines without pregnancy data—leaving those individuals in the lurch when trying to decide whether the vaccines were safe for them.

This issue, termed the “pregnancy paradox” by experts, was the focus of the 16th Kolokotrones Symposium held at Harvard T.H. Chan School of Public Health. During an afternoon of presentations and a panel discussion, experts discussed the importance of including pregnant people when testing vaccines and drugs, the challenges of collecting data for that population, and research methods that could help fill the data gap.

The December 15 event was organized by the CAUSALab at Harvard Chan School and the Harvard Program on Perinatal and Pediatric Pharmacoepidemiology (H4P).

Symposium organizer Sonia Hernandez-Diaz, professor of epidemiology, CAUSALab faculty member, and H4P co-director, opened the event with an overview of why research can be challenging during pregnancy. For example, some drugs can affect fetal development within the first 6–7 weeks of pregnancy, a time period when individuals may not realize yet that they are pregnant.

An ethical imperative for research

The COVID-19 vaccines are just one example of the longstanding issue of excluding pregnant people from clinical trials, according to Anne Lyerly, a professor of social medicine and research professor in obstetrics and gynecology at the University of North Carolina, Chapel Hill.

“In pregnancy, there is an effort to eliminate fetal risk. There’s the notion that no risk to the fetus is acceptable,” she said. As a result, potential drugs are not tested in pregnant people due to concerns that they could harm the fetus.

However, the lack of data poses risks of its own, after drugs are approved and start being used in the clinic. Pregnant bodies process drugs differently, so drug doses may end up too low to be effective or too high and cause toxicity, Lyerly noted. Or patients and physicians may take a “better safe than sorry” approach and avoid using the drugs altogether—even though it’s possible that a particular disease could lead to more harm than any potential risks of treatment, for both pregnant individuals and their fetuses.

“Research in pregnancy is an ethical imperative,” Lyerly said. “Pregnant people deserve protection, access, and respect.”

Lyerly highlighted efforts to improve the representation of pregnant people in clinical trials. For example, in 2018, the Department of Health and Human Services revised research regulations to reflect the fact that pregnant people should have autonomy to decide whether to participate in studies. The agency removed the classification of pregnant people as a vulnerable population—a designation used for groups susceptible to coercion including children, prisoners, and people with intellectual disabilities.

Filling the data gap

In the absence of clinical trial data, researchers can fill the gap by using evidence from large databases, according to the symposium experts. The observational data can be analyzed using causal inference methods that emulate a randomized trial.

Krista Huybrechts, associate professor of medicine at Harvard Medical School, associate professor in the Harvard Chan School Department of Epidemiology and H4P co-director, shared her research comparing the effectiveness of treatments for opioid use disorder.

“In any area of drug safety in pregnancy—the opioid epidemic is no exception—there is really an urgent need for data,” she said. “While randomized control trials are often the gold standard in terms of getting at a true causal effect, they’re often challenging to conduct in this population.”

Huybrechts mentioned a 2010 randomized trial that compared the effectiveness of using buprenorphine versus methadone during pregnancy in improving neonatal and maternal outcomes. The study results suggested that buprenorphine led to better outcomes, but the findings were uncertain given the small number of patients who participated.

To further investigate the question, Huybrechts and her colleagues analyzed a nationwide Medicaid database, which included information such as diagnoses and drug prescriptions. Because the database was large, the researchers were able to include data from thousands of patients in their study. They confirmed that buprenorphine treatment led to better neonatal outcomes, such as a reduced risk of neonatal abstinence syndrome, in which infants experience drug withdrawal symptoms. Furthermore, the researchers were able to calculate the risks with higher precision than the randomized trial.

Huybrechts discussed the challenges of using large databases to determine the effectiveness and safety of a particular treatment in pregnant people. Because treatments are not randomly assigned to patients, the results could be influenced by confounding factors such as demographics and people taking other medications at the same time. She emphasized that researchers need to be careful in their analyses.

“It’s not just the question of which data are you using. It’s a question of what are the methods that you are using,” she said.

Jay Lau

Illustration: iStock/Oksana Kalmykova