Cross-lag interaction between temperature and PM_2.5 reveals hidden windows of cardiovascular mortality risk in a...

Climate change and PM_2.5 co-occur in tropical megacities, yet their joint effects on cardiovascular mortality remain poorly resolved. Prior studies assess aligned-lag interactions, although PM_2.5 may prime susceptibility to later thermal stress. We mapped the temperature-PM_2.5 cross-lag surface to identify excess risk.

We conducted a time-series study in Bangkok regions (2014-2022), linking daily cardiovascular deaths to temperature and PM_2.5. Quasi-Poisson generalized additive models evaluated 8 × 8 cross-lag pairs (lags0-7) using a tensor-product interaction term, adjusted for humidity, time trends, and COVID-19. Interactions were predicted relative to minimum-mortality references, with subgroup analyses.

179,455 cardiovascular deaths were included, with mean temperature, PM_2.5, and relative humidity of 29.14 ± 1.66 °C, 26.49 ± 15.37 μg/m³, and 68.29 ± 8.96%, respectively. PM_2.5 itself showed a relative risk (RR) (95%CI) of lag1 = 1.029(1.018-1.040). The heat peaked at lag0 = 1.017(1.002-1.033); cold at lag4 = 1.044(1.031-1.058). Minimum-death references were 27.5 °C and 9.6 μg/m³. For visualization, the interaction-component surface at temperature lag3 × PM_2.5 lag0 showed a significant interaction surface of 74.79% and a maximum RR of 1.206(1.111-1.309) at 23.98 °C and 76.23 μg/m³. In the 64 lag-pair analysis, positive interaction patterns were concentrated around temperature lags 2-4 and PM_2.5 lags 0-2, rather than being restricted to a single lag pair. Peak RR was higher in females, 1.316(1.145-1.513), than in males, 1.173 (1.066-1.292), but smaller positive surface (23.44% vs. 73.38%). Greater in the younger age group (<65), 1.292(1.120-1.489), than in the older age group (≥65), 1.174(1.064-1.295), but a smaller surface (30.82% vs. 78.43%). Ischemic heart disease showed a stronger RR peak, 1.285(1.119-1.476), than cerebrovascular disease, 1.162(1.028-1.312), yet with a smaller surface (29.45% vs. 65.94%).

Co-exposure was asymmetric; PM_2.5 preceded temperature effects. Females and ischemic heart disease had higher risks; older age and cerebrovascular mortality showed broader vulnerability. Cross-lag evidence supports time-based warning for future research on early-warning interactions and targeted protection.