Association of ambient temperature with tuberculosis incidence in Japan [TBN 057]

WHAT

This ecological study examined the association between ambient temperature and tuberculosis (TB) incidence across Japan from 2007 to 2019 using national surveillance data. A total of 335,060 newly confirmed TB cases were reported across all 47 prefectures during the study period. TB diagnosis was based on clinical symptoms (e.g., cough, sputum, fever, chest pain) combined with laboratory confirmation, including sputum smear, culture, nucleic acid testing, tuberculin skin tests, interferon-gamma release assays, radiography, or clinical judgment.

The analysis found a nonlinear relationship between temperature and TB incidence. The minimum risk temperature (MMT) was identified at 4.45°C, corresponding to the 10th percentile of mean temperature. Compared with this reference, high temperatures were associated with significantly increased TB risk. Specifically, at the 99th percentile (30.1°C), the relative risk (RR) of TB incidence was 1.52 (95% CI 1.04–2.23), indicating a 52% higher risk at extreme heat levels.

Cold temperature effects were smaller and short-term. At −1.2°C (1st percentile), the relative risk was 1.03 (95% CI 1.01–1.06), with the highest effect observed at lag 0 weeks, declining thereafter. In contrast, heat exposure demonstrated delayed effects, with increased TB risk observed at mid-term lag periods ranging from 2 to 26 weeks. These findings suggest that heat exposure may influence TB incidence over longer time frames compared with cold exposure.

Considerable regional variation was observed across prefectures. For example, the minimum risk temperature was 19.4°C (80th percentile) in Hokkaido, while in Okinawa it was 14.8°C (1st percentile), suggesting differences in population adaptation to local climates. These findings indicate that climate-related TB risk may depend on regional environmental conditions and population acclimatization.

Sensitivity analyses demonstrated that the association between temperature and TB incidence remained generally stable across different model specifications, including variations in degrees of freedom, lag periods, and autocorrelation adjustments. Heat-related effects were consistent across sensitivity analyses, whereas cold-related effects were less stable. Overall, results suggest that higher ambient temperatures are associated with increased TB incidence, with delayed and nonlinear effects.

---

HOW

This study used an ecological time-series design based on national surveillance data from Japan’s National Epidemiological Surveillance of Infectious Diseases system. Weekly counts of newly confirmed TB cases from 2007 to 2019 were collected for all 47 Japanese prefectures. Meteorological data—including mean temperature, minimum temperature, maximum temperature, and relative humidity—were obtained from monitoring stations located in each prefecture’s capital city. Diurnal temperature range (DTR) was calculated as the difference between daily maximum and minimum temperatures, and weekly averages were computed for analysis.

A two-stage meta-analysis framework was applied. In the first stage, generalized linear models with quasi-Poisson distribution and log link function were used to estimate prefecture-specific associations between temperature and TB incidence. To capture nonlinear and delayed effects, the study used distributed lag nonlinear models (DLNM) with natural cubic B-splines. Lag periods of up to 26 weeks were examined to evaluate delayed temperature effects on TB incidence.

Seasonality and long-term trends were controlled using natural cubic spline functions for week number and year. Additionally, autoregressive terms with 1- and 2-week lags were included to account for transmission dynamics and temporal autocorrelation. The minimum risk temperature (MMT) was identified from cumulative exposure-response curves, and relative risks for extreme cold and heat were calculated using the MMT as the reference.

In the second stage, mixed-effects meta-analysis was performed to pool prefecture-specific estimates and obtain national-level results. Meteorological variables such as mean temperature, diurnal temperature range, and relative humidity were included as meta-predictors to account for between-prefecture heterogeneity. Residual heterogeneity was assessed using Cochran’s Q-test and I-square statistics, and best linear unbiased predictions (BLUPs) were generated for prefecture-specific estimates.

Multiple sensitivity analyses were conducted to test robustness. These included varying degrees of freedom for seasonal adjustment, extending lag periods from 26 to 52 weeks, removing or adding autocorrelation terms, and replacing spline models with Fourier (trigonometric) terms. The findings remained largely consistent across models, supporting the robustness of the observed association between temperature and TB incidence.

Source: Wagatsuma, K., 2024. Association of ambient temperature with tuberculosis incidence in Japan: An ecological study. IJID regions, 12, p.100384.