Table 1.

Regression coefficients from logistic regression models for lung cancer using different smoking metrics and adjusting for other model covariates as shown^a. Models that include interactions (former^*metric) were fit separately. Residual deviances were calculated from models without interaction terms.

											Residual deviance
	smoking metric		former	current	age	age >70	yrsquit	start < 18	former* metric		Unadj	Adjusted
	Logcig-years								former * logcigyrs
Everyone	.02***		.36	.40	.02*	.10***	−.01†		−.001		2735	2663
Smokers	.02***		−.07		.002	.12***	−.004	−.40**	−.001		2266	2201
	Sqrt-pack								former*sqrt-pack
Everyone	.28***		.55*	.71**	.03***	.10***	−.03***		−.13*		2806	2648
Smokers	.29***		−.17		.04***	.12***	−.03***	−.21†	−.14*		2335	2186
	Log-cigp	Cig-time							former* logcigp	former*cigtime
Everyone	.83***	.03**	−1.26**	−1.13*	.03**	.10***	−.03**		−.63**	.004	2790	2648
Smokers	.90***	.02	−.16		.05**	.11***	−.04**	−.15	−.64**	.003	2265	2186

^aModels fit using everyone include never-smokers, and contain the following variables: smoking metric (as indicated), a former smoker indicator, a current smoker indicator, age, age > 70 (which is 0 if age < 70, and is age if age ≥ 70), years since quitting smoking, and gender. In a separate model, the interaction between the smoking metric and the former smoker indicator was also included. Models fit using smokers only include the smoking metric, a former smoker indicator, age, age > 70, years since quitting smoking, gender, and start < 18 (which is 1 if started smoking before age 18, and 0 otherwise). Gender was not significant in all models. In models with log (cigarettes smoked per day + 1) and years of smoking, the models include both the two smoking metrics. Significance of the coefficients is indicated by superscripts.

^†= (p > .05 but p < .10),

^*= (p < .05),

^**= (p < .01),

^***= (p < .0001)