Association Between Ambient PM2.5 Exposure and Leukocyte Telomere Length Among Rural Adults: Cross-Sectional Study in Northern Thailand

Ambient fine particulate matter (PM2.5) exposure has been associated with oxidative stress, inflammation, and biological aging; however, evidence regarding its relationship with leukocyte telomere length (LTL) remains inconsistent, particularly in rural Southeast Asian populations. This study aimed to investigate the association between ambient PM2.5 exposure and LTL across multiple exposure windows and examine whether oxidative stress and systemic inflammation were associated with PM2.5 exposure and LTL. This cross-sectional study included 179 adults from rural communities in Northern Thailand affected by biomass-burning pollution. Cumulative average ambient PM2.5 exposure was estimated for lag 0–7, 0–14, 0–30, 0–60, and 0–90 days, corresponding to 8, 15, 31, 61, and 91 days, respectively, including the blood-collection date (day 0). Relative LTL was quantified using quantitative polymerase chain reaction. Serum 8-hydroxy-2′-deoxyguanosine (8-OHdG), the Systemic Immune-Inflammation Index (SII), and the Systemic Inflammation Response Index (SIRI) were also assessed. Analyses included covariate-adjusted linear regression, exploratory wild cluster bootstrap inference, and sensitivity analyses. All p-values were unadjusted for multiple comparisons. Median relative LTL was 0.81 T/S units (Q1–Q3: 0.72–0.89). Adjusted models showed positive associations between PM2.5 and LTL at 0–30-, 0–60-, and 0–90-day windows, but not at 0–7- and 0–14-day windows. Per window-specific IQR increase in PM2.5 (8.13, 7.25, and 6.25 µg/m3, respectively), relative LTL was higher by 0.050 T/S units (95% CI: 0.012–0.089; nominal p = 0.011), 0.079 T/S units (95% CI: 0.035–0.123; nominal p < 0.001), and 0.094 T/S units (95% CI: 0.041–0.147; nominal p < 0.001). The reported confidence intervals and conventional p-values did not account for site–date clustering. In exploratory wild cluster bootstrap analyses using eight site-date groups, none of the five associations reached nominal significance (p = 0.247–0.787). The 0–30-day PM2.5 model explained an additional 3.8% of the variance in LTL (ΔR2 = 0.038; p = 0.011), while 8-OHdG contributed a further 2.4% (ΔR2 = 0.024; p = 0.041); SII and SIRI did not significantly improve model fit (p = 0.499 and 0.603, respectively). Sensitivity analyses showed positive longer-window associations with substantially varying magnitudes, whereas inverse shorter-window associations emerged in the NTAQHI-only analysis. Higher estimated ambient PM2.5 exposure was positively associated with higher measured LTL across the 0–30-, 0–60-, and 0–90-day windows in conventional adjusted models. However, sensitivity to sample composition and nonsignificant wild cluster bootstrap findings limit confidence in the robustness of these associations. Adding serum 8-OHdG, SII, or SIRI produced modest changes in model fit and did not establish mediation by oxidative stress or systemic inflammation. These exploratory findings should not be interpreted as evidence of within-person telomere change or a causal or protective effect of PM2.5.