A case-control study identifying critical exposure windows in the association between ambient air pollution and spontaneous abortion.

To examine associations between early-pregnancy exposure to ambient air pollutants and spontaneous abortion (SA), identify lag windows using distributed lag non-linear models (DLNM), and explore the potential mechanisms via network toxicology. A hospital-based case-control study was conducted in Changzhi, Shanxi Province. Individual daily exposures were estimated using inverse distance weighting (IDW) based on residential geocoding and monitoring data. Multivariable logistic regression models were adjusted for BMI, gestational age, gravidity, and parity. DLNM were applied to pollutants showing statistical significance in the logistic model. Network toxicology was used to identify overlapping targets, construct PPI networks, and perform GO/KEGG enrichment analyses. O₃ Q3 (vs. Q1) was inversely associated with SA (OR = 0.42, 95% CI: 0.211–0.824). SO₂ showed positive gradient associations (Q3 vs. Q1: OR = 2.124, 95% CI: 1.068–4.291; Q4 vs. Q1: OR = 2.992, 95% CI: 1.536–5.946). DLNM suggested a lag-dependent SO₂ effect, peaking at lag 28 days (OR = 1.242, 95% CI: 1.0077–1.5363), with a significant association observed during lag 22–28 days, whereas O₃ displayed weak and inconsistent lag patterns. Network toxicology highlighted inflammation/immune-related pathways (e.g., Toll-like receptor and IL-17 signaling). SO₂ exposure was associated with increased SA risk with a more evident late-lag effect, while the inverse association for O₃ should be interpreted cautiously. Inflammation-related immune activation may be a plausible mechanistic pathway.