Using Time-dependent Covariate Analysis to Elucidate the Relation of Smoking History to Warthin's Tumor Risk

Abstract

The authors aimed to elucidate the relation of the time-dependent smoking history parameters--age at smoking initiation and smoking intensity, duration, and latency--to the risk of Warthin's tumor, a benign tumor of the salivary gland for which cigarette smoking is a strong risk factor. They studied 117 cases of Warthin's tumor and 336 matched controls included in an Israeli nationwide case-control study of parotid gland tumors conducted from 2002 to 2003 by using the Cox regression model with time-dependent covariates, with age as the time axis. When current age and smoking duration were included in the statistical model, the authors show that the coefficient of a latency variable does not represent latency as such, but a balancing of the effects of age at initiation and time since cessation. They found a strong positive linear effect of duration of smoking, together with a positive nonlinear effect of intensity that levels off at higher intensities, and a negative effect of latency from 25 years onward. The latter finding implies that the effect of time since cessation dominates the effect of age at initiation, with risk decreasing sharply after smoking cessation. The relation of smoking variables to Warthin's tumor agrees with the patterns reported for lung cancer.