April 28, 2025-- Conflicting findings in environmental epidemiology have long stalled consensus on the health effects of toxic chemicals. A new study by Columbia University Mailman School of Public Health published in the American Journal of Epidemiology suggests that one major reason for these inconsistencies may be the limited exposure ranges in individual studies—leading to underpowered results and unclear conclusions.
Researchers used simulated data to examine how well individual and pooled studies can identify dose-response relationships between chemical exposure and health outcomes. Their findings point to a clear solution: pooling data across studies should be prioritized, even when confounding variables vary between cohorts.
"Underpowered studies—especially those with narrow exposure ranges—may produce misleading results about whether and how a chemical affects human health," said lead author Eva Siegel, PhD in the Department of Environmental Health Sciences. "Our simulations show that combining data across multiple cohorts is a natural and necessary step to strengthen conclusions in environmental health research."
The research focused on polychlorinated biphenyls (PCBs), a class of persistent organic pollutants (POPs). Specifically, the study explored the relationship between maternal exposure to PCB-153—the most commonly detected PCB congener in human blood—and birthweight, an association that has been inconsistently reported in previous studies.
"Some chemicals, like endocrine-disrupting POPs, may interfere with the body's systems even at very low doses," Siegel noted. "Understanding how health risks vary across the full exposure range is essential—but that requires broader data than most single studies can offer."
To address this gap, researchers created five hypothetical populations with different exposure distributions—from low to high—based on real data from three well-known birth cohorts: The Columbia Children's Center for Environmental Health (CCCEH) in New York City, The Environmental Health Fund (EHF) cohort in Israel, and The Child Health and Development Studies (CHDS) in California.
By simulating these distinct exposure environments and analyzing them both individually and collectively, the team assessed how well each approach could recover a "true" dose-response curve. Their results were clear: studies with limited exposure variability often failed to detect effects, while pooled data more accurately reflected the expected relationship.
"Our results show that despite potential differences in confounding factors across studies, the benefits of data pooling outweigh the challenges especially when every effort is made to fully harmonize data between studies.," said Pam Factor-Litvak, PhD, professor of Epidemiology at Columbia Mailman School, and senior author. "To emphasize, this approach is especially crucial in understanding low-dose chemical effects, where many individual studies lack sufficient range to detect patterns."
Other co-authors are Matt Lamb, Jeff Goldsmith, and Andrew Rundle, Columbia University Mailman School of Public Health; Andreas Neophytou, Colorado State University, Matitiahu Berkovitch, Tel Aviv University; and Barbara Cohn, Public Health Institute.
The study was supported by the National Institute of Environmental Health Sciences (grants F31ES032331 and T32ES023772.
Columbia University Mailman School of Public Health
