Temperature, fine particulate matter, and influenza: a time-stratified case-crossover study

Introduction: Evidence that fine particulate matter (PM 2.5 ) modifies the short-term association between ambient temperature and daily influenza cases remains limited, particularly in lower-pollution settings such as Japan. We assessed modification of the temperature–influenza association by PM 2.5 in Kawasaki City, Japan. Materials and methods: The influenza and meteorological source covered 4079 calendar days (1 March 2014–30 April 2025); the date-aligned PM 2.5 dataset covered 4049 days through 31 March 2025. We used a time-stratified case-crossover design with conditional quasi-Poisson regression and a distributed lag nonlinear model. The principal model retained 4028 days in 931 year–month–day-of-week strata. PM 2.5 was represented by the 8-day mean over lag 0–7; its low and high levels were 7.41 and 17.13 µg/m 3 (10th and 90th percentiles). Results: The source surveillance series recorded 309,759 influenza cases; 309,256 occurred during the 4049-day merged period. The overall temperature × PM 2.5 interaction was statistically significant ( p < 0.001). At 4.6 °C versus the median temperature (17.3 °C), cumulative relative risks (RRs) were 20.72 (95% CI: 16.26–26.42) at low PM 2.5 and 56.36 (44.87–70.80) at high PM 2.5 . The corresponding ratio of RRs (RRR) was 2.72 (2.01–3.69). In the pre-pandemic analysis, the cold RRR was 2.66 (1.78–3.96), whereas the post-pandemic estimate was in the opposite direction (0.20 [0.05–0.79]). Conclusions: Evidence for modification of the temperature–influenza association by PM 2.5 was observed in the full-period and pre-pandemic analyses, particularly at low temperatures. The opposite-direction post-pandemic estimate indicates substantial temporal heterogeneity and limits interpretation of the pooled result.