The Honolulu Heart Program continues to follow a cohort of Japanese-American men initially aged 45-68 years, of whom 4,000 had three acceptable measurements of forced expiratory volume in 1 second (FEV1) between 1965 and 1974 and were free of cardiovascular disease and cancer. The 6-year rate of change (slope) in FEV1 was calculated using a within-person linear regression method. Men were divided into tertiles based on the rate of change in FEV1. During 17 subsequent years of follow-up, 796 deaths occurred. The tertile with the greatest rate of decline in FEV1 (mean, -61 ml/year) had the highest age-adjusted total mortality rate (17.3/1,000 person-years), followed by rates of 13.2 for the middle tertile (mean, -25 ml/year) and 11.0 for men with the smallest change in FEV1 (mean, +9 ml/year) (test for trend, p < 0.0001). Using the Cox model, comparing the tertile with the smallest change in FEV1 as a reference group with the tertile with the greatest decline in FEV1, and after adjusting for age, hypertension, smoking, body mass index, alcohol intake, diabetes mellitus, and cholesterol, the authors found the relative risk (RR) for total mortality to be 1.48 (95% confidence interval (CI) 1.24-1.77). After stratification by smoking status, this association remained significant for past smokers (RR = 1.79, 95% CI 1.31-2.14), as well as for the low, < or = 42 (RR = 1.46, 95% CI 1.05-2.03), and high, > 42 (RR = 1.56, 95% CI 1.20-2.02), pack-year groups. An increased risk was also present for current smokers (RR = 1.29), but it was of borderline significance (p = 0.08). No association was found among never smokers. These data suggest that the rate of decline in FEV1 is a predictor of total mortality among smokers.