OBJECTIVE: To evaluate the effects of moderate-intensity regular exercise on serum levels of tumor necrosis factor-a a (TNF-a a) and glucose and lipid metabolism parameters. DESIGN: Longitudinal intervention study of a 5 month exercise training program (30 ± 45 minaday, 4 ± 5 daysaweek). SUBJECTS: Forty-one healthy Japanese women aged 41 ± 69 y at baseline; 27 participants in the exercise program. MEASUREMENTS: Body mass index (BMI), waist-to-hip ratio (WHR), percentage body fat, and fasting levels for serum TNF-a a, serum soluble TNF receptor p55 (TNF-RI) and TNF receptor p75 (TNF-RII), serum lipids, HbA 1c , and serum insulin before and after exercise. RESULTS: In overweight to obese subjects, serum levels of TNF-a a, TNF-RI and TNF-RII were signi®cantly higher than those in lean subjects. There were signi®cant correlations between log serum TNF-a a and BMI, percentage body fat, WHR, HbA 1c and log insulin. TNF-RI was signi®cantly correlated with BMI, percentage body fat, WHR and HbA 1c . TNF-RII was also associated with BMI, percentage body fat and WHR. However, the correlation between TNF-RII and HbA 1c did not reach statistical signi®cance. Neither TNF-RI nor TNF-RII was correlated with log insulin. In contrast, TNF-a a, TNF-RI and TNF-RII were negatively correlated with HDL cholesterol. Regular exercise decreased BMI, percentage body fat, HbA 1c , serum TNF-a a, TNF-RI and TNF-RII and increased HDL cholesterol levels. In addition, exercise-induced change in serum TNF-a a was independently correlated with changes in HbA 1c and serum insulin, after being adjusted for the change in fat-free mass. CONCLUSION: Changes in serum TNF-a a that occur with exercise may play an important role in improving glucose metabolism parameters.