Climate change and cancer: An ecological evaluation of climate risks for cancer survivors.

Abstract

11093 Background: Changes in climate patterns will influence the frequency and severity of extreme weather events, leading to increased disasters (e.g., floods, fires), air pollution levels, and climate-related costs. Cancer survivors may suffer disproportionate effects of climate change-related events on mental health, care disruptions, morbidity, and mortality, yet evaluations of area-level climate risks and cancer survivorship are lacking. We examined future climate change risks faced by communities with high levels of cancer mortality. Methods: Data were from the Climate Vulnerability Index, a county-level measure of future climate change risks (45 variables) alongside a baseline vulnerability index (BVI) including health, socioeconomic, and environmental factors (139 variables). Variables ranging from 0 (least vulnerable) to 100 (most vulnerable) were split into high (rank ≥50) vs. low (rank <50) categories and linked to county-level cancer mortality data from the NCI’s SEER registry. We compared the top quartile of the U.S. population living in counties with the highest cancer mortality rates (54.7% of counties) to other areas. Data were randomly split using a 2:1 training/validation ratio. We identified candidate variables in bivariate analyses between climate risk factors (RFs) and area-level cancer mortality. Best-subset selection with logistic regression and K-fold cross-validation was used to derive a parsimonious model of adverse RFs. These were summed, creating a score, and split at the median to test in the validation set. All analyses were adjusted for the BVI. Results: Overall, N=3,139 counties were examined. In the training set (n=2,094), a model with 6 RFs was identified including weather-related events (increased hurricane, tornado, and precipitation exposure), air pollution factors (increased ozone/CO2 levels), increased climate-related costs, and future economic/productivity losses. Compared to low cancer mortality counties, high cancer mortality counties were more than threefold more likely to have high exposure to the 6 climate RFs (69.6% vs. 30.4% with 3-5 factors; OR=3.07, 95% CI, 2.52-3.74, p<.0001). In the test set (n=1,045), results were similar, validating the model. Overall, the BVI-adjusted model C-statistic was 0.80 (unadjusted, 0.73). The association of high cancer mortality areas and high exposure to climate RFs was evident for both high (OR=3.87, p<.0001) and low (OR=2.75, p<.0001) BVI counties. Conclusions: In this first-of-its-kind ecologic association study, we found that living in counties with high cancer mortality was associated with future climate change risks including extreme weather events, air pollution, climate-related costs, and economic factors. Future work is needed to identify potential mechanisms between climate events and worse cancer outcomes such as decreased treatment tolerability, treatment quality, or treatment access.