Table 1 summarizes the analyses performed in this study. Table 1. Summary of Analysis Groups The first set of analyses sought to identify the occurrence or worsening of pain. The first analysis in this set examined smokers who reported having no pain at enrollment (Group A). The occurrence of pain in this group of smokers sellckchem was defined as reporting pain at any subsequent interview. The second analysis in this set included smokers who reported experiencing no pain or mild pain at enrollment (Group B). Worsening of pain in this group of smokers was defined as reporting moderate to severe pain at any subsequent interview. The second set of analyses sought to identify the resolution or improvement of pain. The first analysis in this set examined smokers who reported having any level of pain at enrollment (Group C).

The resolution of pain in this group of smokers was defined as reporting no pain at any subsequent interview. The second analysis in this set included smokers who reported moderate to severe pain at enrollment (Group D). The improvement of pain in this group of smokers was defined as reporting either no pain or mild pain at a subsequent interview. Generalized estimating equations (GEE) logistic regressions were used to evaluate the relationship between quitting smoking and the four different pain outcomes mentioned above at follow-up assessments, adjusting for other factors that we and others have shown in prior studies to be associated with pain (Mantyselka, Turunen, Ahonen, & Kumpusalo, 2003; Shi, Hooten, Roberts, & Warner, 2010).

The analyses included data from all follow-up assessments, exploring the relationship between smoking status and pain outcomes in each subject at each assessment. Because each subject could have multiple assessments, intra-individual correlations were taken into account by the GEE analysis. Age at enrollment (in 10-year intervals), sex, race, and education were not modeled as time dependent. Marital status, annual household income (in 10,000 U.S. dollars as a continuous variable), insurance coverage (any government or private plans), arthritis, quitting smoking, depressive symptoms, and BMI in four categories were entered into the models as time-dependent variables (i.e., were allowed to change at each assessment). In the multiple regression models, the independent contribution of covariates of interest was determined after adjustment for the effects of other covariates.

Adjusted odds ratios (ORs) were reported. Analyses was performed using Stata, version 10.0 (College Station, TX), and p < .05 was considered to be statistically significant. Results Table 2 shows the baseline characteristics of the study population. At the time of enrollment, 6,258 subjects were current smokers. The majority were aged 50�C60 years and were self-identified as non-Hispanic Batimastat Whites.