Two California studies on wildfire smoke in pregnancy, and one result runs backwards
A 4.5-million-birth analysis found smoke exposure associated with higher birth weight, not lower. A separate cohort linked third-trimester smoke days to autism risk — but only when counting days, not concentration.
| Group | Value (value) |
|---|---|
| 1 to 5 days | 1.11 (1.01 to 1.22) |
| 6 to 10 days | 1.12 (0.96 to 1.31) |
| More than 10 days | 1.23 (1.04 to 1.44) |
Two studies published eight days apart in Environmental Science & Technology asked what wildfire smoke does to a pregnancy. Both used California data, both were large, and both produced results that are more complicated than the headline version of the question.
Taken together they are a useful case study in how exposure gets measured, because in each paper the choice of metric changed the answer.
Birth weight: the result nobody predicted
The first study, published 12 January, examined 4,537,418 singleton term births in California from 2007 to 2018 [s1]. Wildfire PM2.5 exposure was estimated using a bias-corrected Community Multiscale Air Quality model, linked to residence at delivery [s1]. Logistic and linear regression models adjusted for individual-level sociodemographic covariates and seasonality [s1].
The expectation, from the broader air pollution literature, is that particulate exposure in pregnancy reduces fetal growth. That is not what the analysis found.
A 1 µg/m³ increase in wildfire PM2.5 during pregnancy was associated with increased odds of large for gestational age and an increase in birth weight, and with moderately decreased odds of low birth weight and small for gestational age [s1]. In other words, the associations pointed toward larger babies, not smaller ones.
The associations were more pronounced among Hispanic individuals and those categorised as Other race [s1]. Among American Indian and Alaska Native births, the direction reversed: wildfire PM2.5 was associated with decreased odds of large for gestational age [s1].
The authors do not present this as a benefit of smoke. Their conclusion is that the results underscore the importance of understanding how wildfire PM2.5 impacts fetal growth, especially among marginalised groups [s1]. A result that runs against the prevailing hypothesis in one very large dataset is a reason to examine the exposure model, the confounding structure and the population — not a finding to act on.
Several explanations are available and none is settled by this paper. Exposure was assigned by residence at delivery, not by where a person actually spent the pregnancy. Wildfire seasons coincide with other seasonal patterns in behaviour, heat and diet. And term births only — excluding preterm deliveries — can introduce selection effects when the exposure also influences gestational length.
Autism: the metric decides the answer
The second study, published 20 January, used a southern California pregnancy cohort with exposures estimated from 2006 to 2014 at maternal addresses, across the whole pregnancy and by individual trimester [s2]. There were 3,356 autism diagnoses by age five [s2].
The design feature that makes this paper worth reading is that it used three different exposure metrics: mean wildfire PM2.5 concentration, the number of days of smoke exposure, and the number of waves of smoke exposure, with analysis restricted to days over PM2.5 thresholds of 3 and 5 µg/m³ [s2]. A wave was defined as exceeding 3 µg/m³ for two consecutive days [s2]. Sensitivity analyses examined the 75% of the cohort who did not move during pregnancy [s2].
Autism risk was associated with an increased number of wildfire-exposed days in the third trimester, and the association was strongest among non-movers [s2]. For non-movers, compared with no exposed days, hazard ratios were 1.108 (95% CI 1.010 to 1.215) for 1 to 5 days, 1.118 (0.957 to 1.307) for 6 to 10 days, and 1.225 (1.043 to 1.440) for more than 10 days [s2]. The middle category's interval crosses 1, so the dose-response is suggestive rather than clean.
Per wave increase in the third trimester, hazard ratios were 1.073 (1.009 to 1.140) for the entire cohort and 1.267 for non-movers [s2]. The confidence interval reported for that last figure in the paper's abstract reads 1.054 to 1.205, which does not contain the point estimate of 1.267 — an apparent error in the published text that readers should treat with caution rather than resolve in either direction.
And the finding that constrains all of the above: there was no association with mean wildfire PM2.5 concentration alone [s2].
The authors' conclusion is appropriately hedged — prenatal exposure to wildfire smoke may increase risk of autism among children [s2].
Why the metric result matters most
Air quality is almost always reported to the public as a concentration. Both of these papers suggest that concentration may be the wrong summary for wildfire smoke.
In the autism study, the day-count and wave metrics found associations that the mean concentration metric did not [s2]. That is consistent with a mechanism driven by episodes of exposure rather than by an averaged background level — and an average over nine months will flatten a week of dense smoke into near-invisibility.
The birth weight study used average daily wildfire PM2.5 during pregnancy [s1], the metric that found nothing in the autism analysis. Whether that contributed to its counterintuitive result is not something either paper can answer.
Limits, stated plainly
Both are observational studies of modelled exposure. Neither measured what any individual actually breathed. Both are confined to California, which has a distinctive fire regime, a distinctive population and distinctive housing stock. Autism diagnosis by age five is subject to ascertainment differences that track with the same demographic variables as smoke exposure. Neither study establishes causation, and neither should change any individual's decisions.
What to watch
Whether other regions replicate the day-count and wave findings, whether the birth weight direction reverses in cohorts that assign exposure by residential history rather than address at delivery, and whether the Environmental Science & Technology correction record clarifies the confidence interval discrepancy in the autism paper.
Sources
- [s1] Impact of Wildland Fire Smoke PM2.5 on Birth Weight in California. Environmental Science & Technology, 12 January 2026. https://doi.org/10.1021/acs.est.5c05040
- [s2] Prenatal Exposure to Wildfire and Autism in Children. Environmental Science & Technology, 20 January 2026. https://doi.org/10.1021/acs.est.5c08256
Sources
- Impact of Wildland Fire Smoke PM2.5 on Birth Weight in California — Environmental Science & Technology , January 12, 2026
- Prenatal Exposure to Wildfire and Autism in Children — Environmental Science & Technology , January 20, 2026
Three-quarters of US pregnancies since 2012 involved some wildfire smoke exposure
A 34-million-birth analysis maps how far smoke travels from the fires that make it. A companion review is blunt that what that exposure does, beyond the lungs, is still unsettled.
A year after the LA fires, the first hard emergency-department numbers are in
Cedars-Sinai counted a 118% rise in general-illness visits and a 46% rise in heart attack visits over the 90 days after the fires started. Abnormal blood results more than doubled.
Wildfire smoke keeps killing after the sky clears, a 1.25-million-person study finds
Berkeley researchers followed older Californians for a decade. Cumulative smoke exposure predicted death better than the number of smoke days did — which is not how air quality is usually reported.
Only five studies qualified for a review of wildfire smoke and children's lungs
A November systematic review screened 120 publications and found five eligible papers, all from the US and Canada. All pointed the same way; none could be pooled with the others.