Lung cancer incidence and PM _2.5 : Evaluating the association with varying spatial scales and durations of exposure estimates.

Abstract

10600 Background: Particulate matter with a diameter of 2.5 µm or less (PM 2.5 ) is classified as an IARC group 1 carcinogen based on decades of research. However, little is known about how the association between PM 2.5 and lung cancer incidence is impacted by varying spatial and temporal methods of estimating exposure. Methods: We performed a nested case-control analysis using the Department of Veterans Affairs (VA) Clinical Data Warehouse. Patients were eligible for inclusion if they had a recorded date of birth, smoking data, and quarterly location data for 40 or more quarters between 2009 and 2021. Cases were required to have at least 10 years of location data prior to diagnosis; diagnoses and diagnosis dates were identified using the VA Central Cancer Registry. Controls were matched in a 4:1 ratio on age (by birth quarter), smoking status (self-reported, defined as “Smoker” vs. “Never Smoker”), and the specific calendar quarters contributing to the PM 2.5 average. Modeled daily PM 2.5 estimates at a spatial resolution of 1 km 2 for the entire contiguous United States (US) were aggregated to quarterly estimates at three US Census geography levels (block group, census tract, and county). After case-control matching, we evaluated the impact of different spatial and temporal measures of PM 2.5 exposure on the association between PM 2.5 and lung cancer incidence. Specifically, we evaluated estimates at the block group, census tract, and county level with both 5-year and 10-year averages. Associations were evaluated using conditional logistic regression; model fit was assessed using the Akaike Information Criterion (AIC). Results: Approximately 4.7 million veterans met our inclusion criteria. Of these, 8451 veterans developed incident lung cancer at least 10 years after the observation period began; 33,150 matched controls were selected. All models showed a statistically significant increase in the risk of lung cancer with increasing PM 2.5 exposure, with odds ratios (ORs) ranging from 1.06 (95% CI 1.04-1.09) to 1.13 (95% CI 1.10-1.17) and p-values < 0.001 for all models. In both the 5-year and 10-year analyses, models with larger spatial units (county > census tract > block group) had higher ORs and superior model fit by AIC. All models with 10-year averages had higher ORs and lower AIC values compared to models with 5-year averages. Conclusions: Our findings add to the considerable body of research that demonstrates an epidemiologic association between PM 2.5 and lung cancer incidence. Furthermore, our results demonstrate that this association is robust to various methods of estimating patient PM 2.5 exposure. Finally, these data suggest that using county-level PM 2.5 estimates and an estimate duration of at least 10 years may be preferable to finer spatial scales and shorter estimate durations in future studies.