We examined the associations between physical activity (PA), fitness and all-cause mortality and compared their contributions, taking smoking status into consideration. A retrospective cohort study of 18,775 men was carded out between May 1995 and December 2003. Fitness was measured by maximum oxygen uptake and regular PA was defined as at least three times a week, for more than 30 min of leisure time PA. During the mean 6.4 yr of follow-up, 547 deaths were recorded. The hazard ratio (HR) (95% confidence interval [CI]) of regular PA for all-cause mortality was 0.63 (0.52-0.76). The HRs (95% CIs) for men with middle and highest tertile levels of fitness were decreased by 0.58 (0.47-0.70) and 0.58 (0.45-0.75) in comparison to men with lowest one. The inverse association between fitness and mortality was significant among the men who did not engage in regular PA, but not among those who did (p for interaction=0.031). Smoking status did not influence on the associations between regular PA, fitness and mortality. Our result suggested that regular PA and fitness predicted mortality in men. The influence of fitness on mortality was pronounced in the men who did not engage in regular PA.