Oral cavity and oropharyngeal cancer incidence: 5-year survival vs. air pollution levels across New York State.

Abstract

e18111 Background: Smoking prevalence in the US has decreased but incidence of oral cavity and oropharyngeal cancers (OCC/OP) is increasing, particularly in never-smokers. Patterns of new diagnoses have been shown to vary geographically, but the relationship with air pollution has been inconsistent across studies. We examined the dispersion of OCC/OP cases in New York State (NYS) and created models to quantify associations between county-level ozone and PM2.5 with incident and 5-year survival rates of these diseases. Methods: Incidence and 5-year survival data were from NYS Cancer Registry; air pollution levels were from the CDC and SES characteristics were from the American Community Survey. Getis-Ord Gi was calculated to assess for hot/cold spots in incidence and 5-year survival. County levels of PM 2.5 and ozone were compared to county-level OCC/OP incidence and 5-year survival rates while controlling for SES county characteristics and percentage of adult smokers using negative binomial and fractional logit models respectively. Spatial autocorrelation was assessed with Moran’s I and when present, spatial lag models were used to test for spatial autocorrelation interference. PM 2.5 and ozone were tested individually. Results: A seven-county hot spot of OCC/OP incidence in western NYS was identified; with a corresponding cold spot comprised of New York City and two neighboring counties (p<.05). In upstate NYS, a two-county cold spot of 5-year survival rates was located (p<.05). Counties with higher levels of ozone had statistically significantly higher incidence of OCC/OP (IRR= 1.04; 95% confidence interval, 1.01 -1.07). Moran’s I showed no evidence of spatial autocorrelation in this model. PM2.5 was not statistically significantly correlated with OCC/OP incidence (IRR=.99; 95% confidence interval, .93 - 1.06). Negative binomial results are reported because while Moran’s I detected spatial autocorrelation, the spatial lag model showed no evidence of interference from this. Neither pollutant was statistically significantly associated with OCC/OP 5-year survival; however higher levels of PM 2.5 non-statistically significantly trended toward lower 5-year survival rates (.89; 95% confidence interval, .27 - 3.01). Conclusions: OCC/OP incidence and 5-year survival rates exhibited spatial clustering in NYS and incident rates were higher in counties with higher ozone levels. Neither of the pollutants was statistically significant with 5-year survival but higher levels of PM 2.5 trended toward lower 5-year survival rates. Results from this hypothesis generating analysis suggest that ozone might play a role in the development of oral and oropharyngeal cancers. Incidence Model Coefficient 95% Confidence Interval P-Value PM 2.5 .99 .93 to 1.06 .81 Ozone 1.04 1.01 to 1.07 <.01 5-Year Survival PM 2.5 .89 .27 to 3.01 .84 Ozone 1.02 .54 to 1.92 .95