Diabetes
Per- and Polyfluoroalkyl Substance and Glycemia-Related Outcomes in Project Viva: A Causal Inference Approach Katlyn McGraw* Katlyn McGraw Arce Domingo-Relloso Linda Valeri Zilan Chai Feng Yan Eva Siegel Jorge Chavarro Marie-France Hivert Emily Oken Pi-I D. Lin Emma V. Preston Jordan Arvayo Burdeau2 Kathryn Tomsho Briana J.K. Stephenson Tamarra James-Todd Ami Zota
Objective: Exposure to per- and polyfluoroalkyl substances (PFAS) can adversely impact glycemic outcomes. Because some PFAS have significantly decreased over the last 20 years, we aimed to investigate whether an earlier decrease in PFAS exposure, during pregnancy, would improve later in life glycemic outcomes.
Methods: Project Viva, an eastern Massachusetts-based pre-birth cohort, enrolled pregnant individuals in 1999-2002. Among 442 participants, we measured plasma concentrations of six PFAS in 1) 1st trimester of pregnancy and 2) ~17-20 years post-pregnancy (i.e., midlife, 2017-2021) using HPLC-MS. Using the g-formula and g-formula extension of the Bayesian Kernel Machine Regression (GBKMR), we used a simulated intervention to estimate the association with glycemic outcomes (HbA1c, 2-hour postprandial oral glucose tolerance test – OGTT, fasting glucose, and HOMA-IR) when pregnancy PFAS were lowered to the 75th percentile of those measured at midlife, compared to the observed median pregnancy PFAS levels, accounting for time-varying confounding. We evaluated interventions for individual PFAS and joint PFAS using the g-formula and for the PFAS mixture using GBKMR to account for any multicollinearity. Models were adjusted for baseline age, race, education, income, pre-pregnancy BMI, smoking status, and time-varying confounders menopause, parity, and breastfeeding duration.
Results: Participants were of median age 33y at enrollment, primarily White (74%), and highly educated (77% ≥bachelor’s). Median levels of all PFAS decreased substantially from pregnancy to midlife, except for PFDA, with intraclass correlation values ranging from 23% for PFOS to 64% for PFDA. When pregnancy levels were decreased to the 75th percentile of observed midlife PFAS levels, there was a -24.2 (-25.3, -23.1), -9.09 (-10.4, -7.87), and -1.25 (-2.43, -0.13) mg/dL decrease in midlife postprandial OGTT for PFOS, PFOA, PFHxS, and MEFOSAA, respectively. Additionally, there was a -0.45 (-0.50, -0.40), -0.28 (-0.33, -0.23), and -0.05 (-0.09, -0.00) decrease in midlife HOMA-IR for PFOS, PFOA, and PFHxS, respectively. There were no significant decreases in outcomes when assessing the joint effect of decreased PFAS at pregnancy for the g-formula or GBKMR.
Conclusion: Earlier reductions in certain PFAS may improve glycemic outcomes. Actions to reduce PFAS now are more impactful long-term.