Joint effects of air pollution and meteorological factors on drug-resistant tuberculosis incidence in Eastern China: A time-series analysis of lagged and interactive associations
Аннотация
Environmental exposures have been associated with tuberculosis, but evidence specific to drug-resistant tuberculosis (DR-TB) remains limited. We examined lagged associations, relative predictor importance, and joint and interaction associations of meteorological factors and air pollutants with monthly DR-TB incidence rates in Pudong New Area, Shanghai. Monthly DR-TB incidence rates and 13 environmental exposures from January 2019 to December 2023 comprised 60 observations. Quasi-Poisson log-link working mean models estimated single-lag associations at 0–3 months and linear-exposure distributed-lag associations over 0–6 months. Cumulative, joint, and formal interaction models used 54 common months. Temporal adjustment included a time spline and annual sine and cosine terms. Bootstrap random forests assessed predictor importance; additive joint models and generalized linear and additive models assessed joint associations and four exposure-pair interactions. Benjamini–Hochberg correction was applied within each testing family. Heteroskedasticity and autocorrelation consistent covariance estimates assessed sensitivity to serial correlation. During the 60-month study period, the mean monthly incidence rate of DR-TB was 0.053 cases per 100,000 population per month. Lagged analyses identified heterogeneous temporal patterns across environmental exposures. Relative-humidity estimates were sensitive to temporal model specification, with a P90-versus-P50 cumulative RR of 0.894 (95% CI: 0.362–2.209) after full temporal adjustment; however, no lag-specific or cumulative associations remained statistically significant after Benjamini–Hochberg correction. Random forest analysis identified PM 10 , relative humidity, precipitation, O 3 , and SO 2 as the five exposures with the highest predictive importance, accounting for 52.2% of total importance. Exploratory joint-exposure analyses showed adjusted contrasts with pointwise 95% CIs that included 1. No interaction between environmental exposure pairs remained significant after correction. This ecological analysis provided limited evidence for environmental associations with DR-TB after temporal adjustment and multiple-testing correction. Predictor importance identifies candidates for further investigation, while imprecise estimates and specification sensitivity warrant longer studies with individual exposure and clinical data.