This meta-analysis discusses the consistency, strength, dose-response, independence, and generalizability of published cohort data on accident death relative risks in smokers. To locate data, three authors independently searched MEDLINE, and bibliographies of the pertinent studies found, for data which allowed estimation of an appropriate cigarette smoker accident death relative risk (and 95% confidence interval). Relative risks and dose-response were summarized by fixed effects and Poisson modeling, respectively. Four pertinent cohort studies including eight populations were located. Cigarette smoking predicted summary accident death relative risks of 1.51 (95% confidence interval 1.27-1.78) versus never smokers and 1.35 (1.17-1.57) versus ex-smokers. Summary dose-response trends were significant (P = 0.0000) versus never or least smoking referents. In individual studies, the smoking/accident death association persisted after adjustment or, in effect stratification, for age, race, sex, and occupation; occupation and time period; or numerous cardiac risk factors. This meta-analysis found significant, consistent, dose-response, often strong and independent (of age, race, and sex), prospective associations of smoking with accident death, internationally. Further studies and warnings of the smoking/accident death associations seem merited.