In this study, we used a multivariable negative binomial model with DLNM framework, to investigate the association between climate variables and PTB incidence in Brunei. We found significant results for minimum temperature, total rainfall, total sunshine hours and mean relative humidity, but with varying degrees of magnitude, direction and timing, in terms of the number of lagged weeks.
Firstly, we found that high minimum temperature was associated with low PTB incidence, and that this association was significant until the first 12 lagged weeks. This means that an increase in minimum temperature within the first 3 months prior could lower PTB incidence in Brunei. On the other hand, we found low total rainfall was also associated with low PTB incidence, with significant results observed from lag 12 onwards. Both findings were consistent with previous studies in China4 and Bangladesh23, which both used DLNM approach but with monthly and quarterly time series data, respectively. Our study findings were also consistent with another Chinese study that uses spatial panel modelling10. A plausible explanation for this finding relates to changes in human activities in both indoor and outdoor settings in response to temperature changes and rainfall, which in turn could lead to differences in the human-to-human transmission of Mycobacterium tuberculosis4. Such effects can be notably observed in Ethiopia, where the rainy season negatively affects healthcare-seeking behaviour for chronic illnesses6.
Secondly, we also found an inverse relationship between PTB incidence and low sunshine hours, and that this association was significant from 12 to 24 lagged weeks. This means that decrease in total sunshine hours 4 - 6 months prior could lower PTB incidence in Brunei. This finding contrasts to those found in countries with particularly temperate and subtropical climates, where TB incidence peak in spring/summer, and trough in autumn/winter1,7. Other studies suggest a relationship between vitamin D deficiency in winter months, a proxy for the lack of sunshine hours, with high TB notifications 3 - 6 months later24,25. Our finding could be due to the fact that Brunei lies near to the equator, where variations in daily sunlight exposure throughout the year are minimal. Indeed, multiple studies have shown stronger seasonality patterns in areas further from the Equator1. Further, although a Vietnamese study7 reported similar results to that in temperate and subtropical countries, the authors also noted that individuals in Vietnam often take measures to prevent sun exposure; a point that was not considered in their analysis, and is also a relevant limitation to the Bruneian setting. Hence, a plausible explanation for our finding could be again related to changes in human activities at both indoor and outdoor settings.
Lastly, we found that high mean RH was positively associated with high PTB incidence, with significant results observed from lag 8 onwards. This means that an increase in mean RH within the last 2 months could lead to increase PTB incidence. Although the reported cumulative RR at the 90th percentile for mean RH was very high (RR = 54.0 [95% CI: 5.67, 513.70], this result should be treated with caution due to the very wide 95% CI ranges. Despite this, the positive relationship of mean RH and PTB incidence remains robust, based on results from univariate and sensitivity analyses (S1 – S4 Table).
Our finding for mean RH could be explained by the unfavourable conditions for small droplets containing M. tuberculosis to evaporate slowly, thus allowing it to remain suspended in the air for a longer period of time26. Experiments also suggested that the quantity of small droplets or aerosolized particles containing M. tuberculosis produced by a TB patient is predictive of infection among household contacts27. However, it should be noted that this explanation could be more applicable to outdoor humidity conditions. Due to the daily hot and humid conditions in Brunei, the use of indoor air-conditioning is very common albeit for households who can afford it. Further studies that could incorporate data on air-conditioning usage in households or collecting socio-economic status data, or instead focusing on indoor climate conditions would help to determine the role played by mean RH in the human-to-human transmission of M. tuberculosis.
Understanding PTB seasonality is complex due to the ways it could manifest, that is either recently through exogenous infection, or remotely through endogenous reactivation of latent TB infection. It is still undetermined whether seasonal variation affects either or both types of infection, though recent infection is likely1. While TB cases in high and low TB burden areas tend to be driven, respectively, by recent and remote infection11,28, it is unclear where areas with intermediate TB burden, such as Brunei, stands. While our study findings and their possible explanations point more towards climate factors affecting human-to-human transmission, and thus leading to recent infections, it is still premature to rule out its contribution to remote infection. Future simulation studies on the driving factor in Brunei and/or equatorial countries could help to shed light on this.
Our study has several limitations. Firstly, our findings cannot be translated at the individual level as this is an ecological study. Secondly, climate data was only collected from a single meteorological station. While it is plausible that weather patterns could vary across Brunei (particularly rainfall), we chose to use data from this station alone as it was the only complete set of data available locally, and the station is based in the most populated district in the country. Thirdly, we did not incorporate non-climatic factors that are also known risk factors of PTB, such as age, presence of co-morbidities and socio-economic status, which could confound the study results. The main strength of our study is the availability of a 19-year long dataset with shorter weekly intervals. Analysing using time series with shorter time intervals could possibly lead to more accurate results. Also, we opted to report the multivariate model results for 5 crossbasis terms encompassing the different aspects of climate, which would theoretically control for possible confounding.
In conclusion, our study findings indicate that there is seasonal variation of PTB incidence in Brunei. PTB incidence is associated with minimum temperature, total sunshine hours, total rainfall and mean relative humidity; but with varying degrees of magnitude, direction and timing. These variations could be explained by environmental and social factors, mainly affecting human-to-human transmission. To better understand TB seasonality, future studies on the relative contribution of recent and remote TB infection in equatorial settings is warranted.