Etiologic heterogeneity across oral cavity cancers.

Authors

Yano Y, Tota JE, Graubard BI, Liao LM, Kahle L, Cheung LC, Abnet CC, Chaturvedi AK

Affiliations

• Division of Cancer Epidemiology and Genetics, National Cancer Institute, Rockville, Maryland, USA.
• Information Management Services Inc, Calverton, Maryland, USA.

Abstract

BACKGROUND: Oral cavity cancers are generally considered etiologically similar, with a majority (>75%) attributable to smoking/alcohol. Yet, incidence trends for oral cavity cancers have recently diverged in the United States and several European countries (with declines for most oral cavity cancers but increases for oral tongue cancers), suggesting etiologic differences. This study quantified etiologic heterogeneity across oral cavity cancers using data from the NIH-AARP Diet and Health Study.

METHODS: Cox regression models were used to quantify associations (hazard ratios [HRs] and population attributable fractions [PAFs]) and etiologic heterogeneity for risk factors across oral cavity anatomic sites.

RESULTS: Among 490,969 individuals (6,732,760 person-years), 765 incident oral cavity cancers occurred. There was substantial heterogeneity in the associations of smoking and alcohol within constituent oral cavity cancers (HRs for current smokers, >2 drinks/day vs. never-smokers/nondrinkers: floor of mouth = 14.7; hard palate and other mouth = 8.1; oral tongue = 4.9; gum = 2.6; lip = 1.5). PAFs for smoking/alcohol varied widely, ranging from 12.0% for the lip to 56% for the floor of mouth. We observed unexpected associations of increased oral tongue cancer risk with increased consumption of vegetables: total vegetables without legumes (HR per cup equivalent/day, 1.13; 95% CI, 1.01-1.27); Solanaceae (HR, 1.43; 95% CI, 1.20-1.70); Umbelliferae (HR, 2.31; 95% CI, 1.30-4.12); and dark green leafy vegetables (HR, 1.38; 95% CI, 1.07-1.78). Associations for total vegetables (HR, 1.12), Solanaceae (HR, 1.26), and dark green leafy vegetables (HR, 1.32) were remarkably similar in the PLCO trial, a prospective US cohort of ∼155,000 individuals.

CONCLUSIONS: The results highlight the need for disaggregation of oral cavity cancer sites in etiologic studies and reappraisal of widely accepted smoking/alcohol relative risks/attributable fractions.