BACKGROUND: Barbiturates, particularly phenobarbital, have been shown to be a tumour promoter in animal experiments and were found to be associated with increased risk of lung cancer in our cohort follow-up study to screen pharmaceuticals for possible carcinogenic effects. Sixteen more years of follow-up have accumulated permitting a more detailed evaluation of this association. METHODS: In all, 10,213 subscribers of the Kaiser Permanente Medical Care Program who received barbiturates between 1969 and 1973 from its San Francisco pharmacy were followed up through 1992 and their incidence of lung cancer at biennial intervals was compared with what was expected based on the experience of the entire pharmacy cohort (143,594). Smoking-habit data were available on about half of the barbiturate users and were used to adjust for cigarette smoking in both the observed/expected analysis and in Cox proportional hazards analysis. RESULTS: The initially elevated standard morbidity ratio of 1.55 (95% CI: 1.25-1.91) with 3-7 years of follow-up gradually decreased and stabilized at about 1.3 after 11-15 years of follow-up. This trend for diminishing relative risk over time was more pronounced among the never smokers but their initial excess risk was not statistically significant due to small numbers. A dose-response trend was observed, based on the number of prescriptions dispensed. Analytical control for cigarette smoking reduced but did not eliminate either the association or the dose-response trend. Most of the barbiturate-associated cases in never smokers were women and the predominant histological type was adenocarcinoma. CONCLUSIONS: These findings from up to 23 years of follow-up are not conclusive because of the continuing small number of never smokers who developed lung cancer. However, they strengthen and refine previous observations of a barbiturate-lung cancer association, which is probably not fully explained by confounding by cigarette smoking. The diminution of excess risk over time is consistent with a tumour promoter effect. Findings among the never smokers suggest that this possible effect may be greatest on adenocarcinomas in women.