Thee Community Intervention Trial for Smoking Cessation (COMMIT) is a large-scale study involving 11 matched pairs of communities in North America. COMMIT is under the direction of the Public Health Applications Branch. The Biometry Branch has been responsible for all statistical aspects of the design, implementation, and analysis. Communities were randomized within pairs to active community-based intervention vs. comparison. Randomization occurred in 1988, with final outcome data collected in 1993 (finishing in early 1994); primary outcomes were published in February 1995, and were presented at the ASCO/NCI Symposium on Tobacco Addiction in June 1996. The objective of COMMIT was to establish and assess a cooperative program of smoking cessation strategies that would reach and aid smokers (with particular interest in heavy smokers) in achieving and maintaining cessation of cigarette smoking. Endpoint Cohorts totaling 10,019 heavy smokers and 10,328 light-to-moderate smokers were followed by telephone to determine quit rates. Baseline (1988) and final (1993) telephone surveys sampled households to determine changes in smoking prevalence. Methodological investigations at the beginning and end of the trial evaluated the gains from matching. Planning statistical analyses for the cohorts involved the exploration of randomization-based methods, incorporating techniques for adjusting observed differences using baseline prognostic factors, and for handling missing data using baseline covariates and patterns of intermediate outcome measures. Computer simulation studies were used to study the properties of various approaches. Other work involved approaches for weighting and adjusting results from the prevalence surveys, with consideration of probability of inclusion of each telephone number in the sampling frame, number of telephones in each household, nonresponse, and age-sex distribution for each community based on 1990 census data. Analysis efforts involved computer programming for permutation tests and associated test-based confidence intervals. The cohort analysis showed that the mean heavy smoker quit rate was 0.180 for intervention communities vs. 0.187 for comparison communities, a nonsignificant difference. Corresponding light-to-moderate smoker quit rates were 0.306 and 0.275; the difference was significant (p = 0.004), 90% confidence interval = [0.014, 0.047]. Smokers in intervention communities had greater perceived exposure to smoking control activities, which correlated with outcome only for light-to-moderate smokers. Overall smoking prevalence (ages 25-64) decreased 3.5% in intervention communities vs. 3.2% in comparison communities, a difference not statistically significant but consistent with the more powerful cohort analysis. The impact of this community-based intervention upon light-to-moderate smokers, although modest, has public health importance.
