Forecasting PM2·5 induced lung cancer mortality and morbidity at county-level in China using satellite-derived PM2·5 data from 1998 to 2016: a modelling study

Abstract

BackgroundThe serious ambient fine particulate matter (PM2·5) pollution issue in China has received great attention in the past 10 years and it is one of key risk factors for developing lung cancer. Studies have investigated the effect of PM2·5 associated with regional transport and internal migration on lung cancer mortality and morbidity. However, evidence for a forecast of PM2·5-induced lung cancer across China is scarce. We aimed to predict the spatial distribution of lung cancer morbidity and mortality in China due to exposure to PM2·5 concentration. MethodsFor this modelling study, we used the lung cancer morbidity and mortality data (2006–14) from Chinese cancer registrations. The annual concentrations (1998–2016) of global surface PM2·5 (0·01° by 0·01° spatial resolution) were also collected. A spatial autocorrelation method was used to investigate spatial clusters of lung cancer on a county-level in China. Ridge regression (RR), partial least squares regression (PLSR), model tree (MT), regression tree (RT), and the combined forecasting model (CFM) were alternative methods used to predict the lung cancer morbidity and mortality. A kriging interpolation method was used to obtain the spatial distribution of lung cancer mortality and morbidity in 2015 and 2016. FindingsThe lung cancer morbidity increased from 49·70 per 100 000 population (2006) to 59·03 per 100 000 population (2014), and the lung cancer morbidity increased from 44·15 per 100 000 population to 47·23 per 100 000 population. Spatial autocorrelation analysis indicated that the morbidity and mortality of lung cancer and PM2·5 concentration were significantly spatially autocorrelated for each year in China (spatial correlation 0·26–0·42 for morbidity and 0·25–0·44 for mortality; p<0·0001). Results from kriging method suggested a similar spatial distribution in morbidity and mortality of lung cancer associated with PM2·5 in 2015 and 2016, and high morbidity and mortality of lung cancer areas were mainly located in central-east coast areas. InterpretationThe occurrence of severe PM2·5 pollution was not localised only in a county or city, but was a regional-level problem. Key findings from this study showed a large regional scale of lung cancer morbidity and mortality has occurred in China due to PM2·5 pollution. It is necessary to strengthen regional collaborative pollution management among the high PM2·5 concentration regions. FundingNone.