A meta-analysis was carried out to calculate a pooled estimate of relative risk of lung cancer following exposure to environmental tobacco smoke (ETS) and to determine whether there was any heterogeneity in the pooled estimates according to selected characteristics of the studies. A total of 35 case-control and five cohort studies providing quantitative estimates of the association between lung cancer and exposure to ETS published between January 1981 and March 1999 were identified. Using fixed- and random-effects models, we calculated pooled estimates of relative risk for exposure to ETS from subjects' parents (during childhood), spouses, and coworkers. As well, we investigated whether the pooled estimates of relative risk varied by study location, degree of control of potential confounding variables, proportion of cases confirmed histologically, proportion of surrogate respondents, nonresponse rates, and year of publication. The relative risk of lung cancer among non smoking women ever exposed to ETS from their husbands' smoking was 1.20 (95% confidence interval (CI): 1.12-1.29). The pooled relative risk was 1.19 (95% CI: 1.10-1.29) for case-control studies and 1.29 (95% CI: 1.04-1.62) for cohort studies. In various subgroup and meta-regression analyses, we found no statistically significant differences by selected characteristics of the studies. In addition, we found that the risk of lung cancer increased consistently with increasing levels of exposure. The 11 studies reporting relative risks among male non smokers yielded a pooled relative risk of 1.48 (95% CI: 1.13-1.92) for ever exposed to ETS, and the relative risk of lung cancer for ever being exposed to ETS at work was a 1.16 (95% CI: 1.05-1.28). These results are consistent with the hypothesis that exposure to ETS increases the risk of lung cancer. While there may be alternative explanations to the data, it is more likely that the observed association is not an artifact and that ETS causes lung cancer in non smokers.