Individuals with cytological atypia in sputum may be at increased risk for lung cancer. We conducted a longitudinal analysis of the association between lung cancer incidence and cytological atypia in sputum samples collected prospectively from an ongoing cohort of adults at high risk for lung cancer. Cohort members had a smoking history of > or = 30 pack-years and chronic obstructive pulmonary disease documented by pulmonary airflow testing. Sputum samples collected at baseline and periodically thereafter were examined by standard cytological methods. From the cohort of 2,006 people, there were 83 incident lung cancers over 4,469 person-years of observation. At baseline, the association between personal and behavioral characteristics, and sputum cytological atypia was assessed by multiple logistic regression. The association between sputum cytological atypia and incident lung cancer was then assessed by hazard ratios using proportional hazards regression analysis, adjusting for potential confounding factors. Cytological atypia graded as moderate or worse was associated with continuing cigarette smoking (adjusted odds ratio, 2.5; 95% confidence interval, 1.5-4.1), and with lower levels of intake of fruits and vegetables (P for trend = 0.04). Atypia was not associated with several other factors, including the degree of airflow obstruction, the use of vitamin supplements, nonsteroidal anti-inflammatory drugs, or metered-dose steroid inhalers. Incident lung cancer was increased among those with moderate or worse cytological atypia (adjusted hazards ratio, 2.8; 95% confidence interval, 1.4-5.5). This association was not confounded by other risk factors. We conclude that in this high-risk cohort, cytological atypia is associated with continuing smoking and low intake of fruits and vegetables, but that independent of these and other factors, the risk of incident lung cancer is increased among those with moderate or worse grades of cytological atypia in their sputum.