Maternal ambient air pollution exposure and risk of stillbirth in Georgia, USA

Technical Summary: Maternal Ambient Air Pollution Exposure and Risk of Stillbirth in Georgia, USA

Problem Statement
Stillbirth, defined as pregnancy loss at or after 20 weeks of gestation, remains a significant public health burden in the United States, with approximately 23,000 cases occurring annually. While nearly one-third of stillbirths are etiologically unexplained, there is growing evidence linking maternal exposure to ambient air pollution with increased risk. However, existing literature on this association within the U.S. is limited by small sample sizes, restricted geographic coverage (often limited to California or Medicaid populations), and methodological constraints. Specifically, many prior studies treat air pollution exposure as time-invariant by averaging concentrations over fixed gestational windows, thereby ignoring temporal variability and short-term fluctuations. Furthermore, inconsistencies exist in the literature regarding which pollutants are harmful and during which gestational windows susceptibility is highest.

Methodology
This study utilized a matched case-control design using statewide vital records from Georgia spanning January 1, 2005, to February 27, 2014. The dataset included 8,384 stillbirth cases (defined as fetal death at ≥20 weeks) and 33,459 live birth controls. Cases and controls were matched on maternal residential county and conception month to account for residual unmeasured spatiotemporal confounding (e.g., healthcare access, reporting practices, seasonal variations).

Exposure Assessment:
Air pollution exposures were estimated using a random forest model fusing Community Multiscale Air Quality (CMAQ) model outputs with EPA Air Quality System (AQS) monitoring data. Daily average concentrations were provided on a 12 × 12 km grid and linked to census tracts via spatial area-weighted averaging. The study evaluated ten pollutants: CO, NO₂, NOₓ, O₃, PM₂.₅, PM₁₀, and PM₂.₅ constituents (NO₃, NH₄, EC, OC).

Statistical Analysis:
The authors employed stratified Cox proportional hazards models with time-varying covariates to estimate hazard ratios (HRs). The risk period began at 20 weeks of gestation. Five distinct exposure windows were analyzed:

1. First Month: Time-invariant average of the first four weeks.
2. Weekly: Time-varying weekly averages from week 20 to the end of follow-up.
3. First Trimester: Time-invariant average of weeks 1–13.
4. Second Trimester: Time-varying cumulative weekly average (weeks 14–27).
5. Third Trimester: Time-varying cumulative weekly average (weeks 28+).

The models were stratified by maternal county and conception month and adjusted for maternal age, race/ethnicity, marital status, and tract-level median household income. Sensitivity analyses included adjustments for temperature and subgroup analyses separating second-trimester (20–27 weeks) and third-trimester (≥28 weeks) stillbirths.

Key Results

• Demographics: The study population showed that stillbirth cases were associated with lower median household income, a higher proportion of Black residents, and lower marriage rates compared to controls.
• Second Trimester Susceptibility: The strongest and most consistent positive associations were observed for the second-trimester exposure window. For all stillbirths combined, significant positive associations were found for:
  • PM₁₀: HR = 1.07 (95% CI: 1.04, 1.11)
  • NO₂: HR = 1.06 (95% CI: 1.00, 1.13)
  • PM₂.₅: HR = 1.05 (95% CI: 1.01, 1.09)
  • NH₄: HR = 1.06 (95% CI: 1.02, 1.10)
• Subgroup Analysis: When analyzing second-trimester stillbirths separately, the magnitude of the HRs increased for most pollutants (e.g., CO HR rose to 1.15 for weekly exposure). Conversely, associations for third-trimester stillbirths were largely null or negative across most pollutants and exposure windows.
• Carbon Monoxide (CO): CO exhibited the largest magnitude of HRs across all pollutants, with positive associations observed in weekly and first-month windows, though confidence intervals were wider, suggesting higher uncertainty.
• Ozone (O₃): A notable finding was a high estimated HR for O₃ specifically during the third-trimester exposure window (HR = 1.12), contrasting with null or negative associations for other pollutants in that window.
• Sensitivity: Adjustments for temperature had minimal impact on the estimated HRs.

Significance and Claims
The authors claim this study is the first to evaluate maternal ambient air pollution and stillbirth in the U.S. using time-varying exposures on statewide vital records from Georgia. The study highlights several key contributions:

1. Temporal Resolution: By utilizing time-varying Cox models rather than static averages, the study captures fluctuations in exposure that may be critical for identifying risk windows.
2. Gestational Specificity: The research identifies a specific period of susceptibility—the second trimester—for PM₂.₅, PM₁₀, NO₂, and NH₄. This aligns with some international literature but contrasts with studies finding risks in the first or third trimesters, suggesting potential geographic or methodological heterogeneity.
3. Differentiation of Stillbirth Etiology: By separating second and third-trimester losses, the study reveals that risk factors may differ by the timing of the loss. The authors suggest that the null/negative findings in the third trimester could be due to "depletion-of-susceptibles" (where the most vulnerable pregnancies have already ended) or increased behavioral caution by mothers in late pregnancy.
4. Broad Pollutant Coverage: The inclusion of less frequently studied pollutants (NO₃, NH₄, EC, OC) expands the evidence base beyond the standard PM₂.₅ and NO₂ metrics.

The authors conclude that while air pollution is associated with increased stillbirth risk, the timing of exposure is a critical factor, with the second trimester emerging as a particularly vulnerable window for specific pollutants. They acknowledge limitations, including potential measurement error at the individual level, missing data on maternal behaviors (tobacco/alcohol), and the retrospective nature of the data, but maintain that their spatiotemporal matching and time-varying approach provide robust evidence for environmental risk factors in stillbirth.