Smoking history and incidence of tooth loss, comparing modeling methods

Objective: To evaluate the feasibility of a comprehensive smoking index (CSI) to accurately model the effects of smoking on tooth loss compared to conventional classifications.

Methods: We analyzed data of 989 dentate men in the VA Normative Aging and Dental Longitudinal Study who were not VA patients but received medical and dental care in the private sector and were followed with triennial medical/dental exams for up to 28 years (mean 17 years). Incident tooth loss was diagnosed at each dental exam by calibrated dental examiners. Subjects' smoking history was modeled using the CSI or various conventional approaches including categories of smoking history and packyears. The CSI accounts for intensity, recency and duration of smoking in a single continuous variable (range: 0 – 4.1). Conventional smoking categorizations included smoking status (never, former, current), and never, former (<10 vs. 10+ yrs) and current (<15,15-24, and 25+ cig/day). Tooth specific Cox proportional hazards models were fit for each exposure classification controlling for age, baseline socio-economic status and marital status and time varying BMI and alcohol use. Models were compared using Akaike's information criterion (AIC).

Results: 3041 incident tooth loss events were observed from a total of 399,538 tooth-years. There was a dose- and time-dependent association between smoking and tooth loss incidence. Per unit increase in CSI the hazard rate increased by 50% (95% CI: 38%, 63%). Smoking status alone provided the poorest model fit (AIC=56,976). Modeling smoking history with CSI provided the best model fit (AIC=56,849) compared to other categorizations of smoking.

Conclusion: These results provide further evidence to support the utility of a comprehensive smoking index to model smoking history parsimoniously. This index provides a novel approach to accounting for the complex interaction of smoking history on tooth loss while maintaining statistical efficiency.

Support: U.S. Dept. of Veterans Affairs, NIH R03DE016357, K24 DE00419.