Environmental tobacco smoke exposure and risk of stroke in nonsmokers: a review with meta-analysis.

Our objective was to provide a comprehensive review of the epidemiologic evidence relating stroke to exposure to environmental tobacco smoke (passive smoking) in lifelong nonsmokers. In all, 7 prospective, 6 case-control, and 3 cross-sectional studies were identified that provided relevant information. The 16 studies varied considerably in design, exposure indices used, and disease definition. Based on 24 sex-specific relative risk estimates, and using current spousal exposure (or nearest equivalent) as the exposure index, meta-analysis gave an overall estimate of 1.25 (95% confidence interval 1.16-1.36), with no significant heterogeneity. There was no significant heterogeneity by sex, year of publication, fatality, or disease end point. There was some indication that relative risks were less elevated in prospective studies and in US or European studies. No elevation was seen for subarachnoid hemorrhage. The estimate was similar using ever rather than current exposure, or total rather than spousal exposure. Adjustment for risk factors other than age had no material effect on the estimate. Eight studies provided dose-response estimates, the combined relative risk for the highest level of exposure being 1.56 (95% confidence interval 1.34-1.82). Many studies have evident weaknesses, and recall bias and particularly publication bias are major concerns. Currently, the association is only suggestive of a possible causal relationship. A clearer picture could be obtained using data from existing very large prospective studies that have already provided risk estimates for passive smoking and heart disease, and by conducting large, well-designed studies of incident stroke.