I used to file “forever chemicals” under slow-motion harm, the kind of exposure that surfaces decades later as an off thyroid number or a liver enzyme your doctor frowns at. The idea that the same compounds could help tip a pregnancy into diabetes inside of six months felt like a stretch. Then I read the cohort data, and the stretch got shorter, with one strange catch I did not see coming: whether the chemicals mattered at all seemed to hinge on a vitamin.
The newest entry landed on September 23 in Environmental Health: a prospective birth cohort of 420 pregnant women in Ma’anshan, China, who had their per- and polyfluoroalkyl substances measured in early-pregnancy serum and were then followed to see who developed gestational diabetes. Seventy of them did, about 16.7 percent. Per doubling of serum concentration, the researchers reported higher gestational-diabetes risk across a whole cluster of these chemicals: PFNA, PFDA, PFOS, PFOA, and 6:2 Cl-PFESA, that last one a next-generation “replacement” PFAS that keeps turning up in Chinese blood samples anyway.
So far, so familiar. But the association was not spread evenly across the cohort, and that is what made me sit up. It concentrated almost entirely in the women who were low on vitamin D. In the vitamin-D-insufficient group, PFOA, PFNA, PFDA, and PFOS all showed significant positive links to gestational diabetes; in the women with adequate vitamin D, those same links faded to nonsignificant noise. The authors are honest that the formal interaction test held up for only two of the chemicals, so this is a signal, not a settled law. The shape of it is still hard to unsee.
That is where the biology gets strange, so that is what I want to chase. PFAS are endocrine disruptors, and one of their suspected tricks is latching onto the PPAR receptors that help run fat and glucose metabolism, nudging insulin signaling off its normal rhythm. Vitamin D is not just a bone nutrient. It sits on the pancreatic beta cells that pour out insulin and on the muscle and liver tissue that has to listen when insulin knocks; when it runs low, that listening gets sloppy, insulin resistance creeps up, and the pancreas has to shout to be heard. Wait, why would a fat-soluble vitamin change how an unrelated pollutant behaves in the body? The leading guess is that they are working the same machinery from opposite ends. The PFAS leans on the glucose-handling system, and enough vitamin D seems to keep slack in that system to absorb the shove. Take the vitamin away, and the chemical finally has something to push against.
If that sounds like an overreach from one modest cohort, I thought so too, until I saw the pattern repeat. A separate prospective cohort in Beijing, only 111 high-risk women, found that among the vitamin-D-deficient ones the estimated gestational-diabetes risk tied to PFAS was not subtle: risk ratios running from roughly 3.8 to 8.1 for each 1 ng/mL increase in several long-chain PFAS. In the vitamin-D-replete women that signal largely vanished, and the authors concluded that adequate vitamin D in early pregnancy could blunt the PFAS-linked risk. Small study, wide estimates, same dial turning the damage up or down.
And the American data? Here I have to be honest, because it cuts the other way, hard. Back in April, the NIH-funded ECHO Cohort Consortium pooled 5,229 pregnancies across 16 US cohorts and published in Diabetes Care one of the more deflating results in this field. The six-PFAS mixture was tied to lower odds of gestational diabetes, a prevalence ratio of 0.75 (95% CI, 0.58 to 0.96). Doubling PFOA nudged the odds down too, 0.92 (0.86 to 0.98). No individual PFAS raised risk. “The results, which were contrary to our initial hypothesis, do not imply any changes to gestational diabetes screening among women in the general population with elevated PFAS exposure prior to pregnancy,” lead author Anne Starling said. The largest, best-powered study of the bunch essentially found nothing to worry about.
| Measure | Value |
|---|---|
| Six-PFAS mixture | 0.75 (0.58–0.96) |
| PFOA per doubling | 0.92 (0.86–0.98) |
So which is it? A poison that helps set off diabetes, or a non-issue the big consortium already ruled out? I sat with that contradiction for a while, and I do not think it is the standoff it looks like. Look at what the null study did not do: it never stratified by vitamin D. It pooled everyone, the deficient and the replete, and asked for one average answer. If the harm lives almost entirely in the low-vitamin-D subgroup, the way both Chinese cohorts suggest, then folding that subgroup into a mostly vitamin-D-sufficient American population is exactly how you erase the signal, or even flip it, through the confounding and selection quirks that dog observational chemical studies. The ethnically diverse PETALS cohort shows the same messiness from another angle: PFNA at 1.40 (95% CI 1.24 to 1.58), PFOA at 1.15 (1.04 to 1.27), and PFDA at 1.23 (1.09 to 1.38) all pushed gestational-diabetes odds up, while PFHxS ran stubbornly inverse at 0.48 (0.38 to 0.60) and PFOS came out flat. These chemicals do not move as one bloc, and neither do the bodies carrying them.
| Measure | Value |
|---|---|
| PFNA | 1.40 (1.24–1.58) |
| PFOA | 1.15 (1.04–1.27) |
| PFDA | 1.23 (1.09–1.38) |
| PFHxS | 0.48 (0.38–0.60) |
The establishment read of the null pooled data is the reassuring one: nothing to change, carry on. But “no average effect” and “no effect in the women who are actually vulnerable” are not the same sentence, and treating them as identical is exactly how a subgroup harm disappears inside a comfortable average. Nobody consented to carrying industrial fluorochemicals in their blood during pregnancy in the first place. The least the science can do is ask who they hurt instead of flattening everyone into one number.
I am not going to pretend a 420-woman cohort settles anything, or that a pregnant woman should panic over a PFAS level she cannot easily measure or lower. What I can control is the modifier. Vitamin D deficiency in pregnancy is common, cheap to check, and cheap to fix, and the two vitamin-D cohorts both point to it as the difference between a chemical that sits quiet and one that bites. If I were pregnant, I would not wait for a consensus statement to reconcile these studies. I would get my vitamin D tested, and if it were low, correct it, because that is the one lever in this whole tangled story that is actually in my hands.
Sources
- Environmental Health – Early-pregnancy PFAS, vitamin status, and risk of gestational diabetes: a prospective birth cohort study (2026)
- Environment International – Modifying effects of vitamin D on PFAS and gestational diabetes: a prospective cohort in Beijing (2025)
- Healio – PFAS exposure not linked to increased gestational diabetes risk (ECHO Cohort, Diabetes Care, April 2026)
- PubMed – PFAS During Pregnancy and Gestational Diabetes: The ECHO Cohort
- BMC – PFAS and gestational diabetes in the ethnically diverse PETALS cohort
- Frontiers in Endocrinology – PFAS exposure and type 2 diabetes risk (mechanism review, 2022)