In 70 healthy subjects with a large age range, the relationships between plasma tumor necrosis factor-alpha (TNF-alpha) and body composition, insulin action, and substrate oxidation were investigated. In the cross-sectional study (n = 70), advancing age correlated with plasma TNF-alpha concentration (r = 0.64, P < 0.001) and whole body glucose disposal (WBGD; r= -0.38, P < 0.01). The correlation between plasma TNF-alpha and age was independent of sex and body fat (BF; r = 0.31, P < 0.01). Independent of age and sex, a significant relationship between plasma TNF-alpha and leptin concentration (r = 0.29, P < 0.02) was also found. After control for age, sex, BF, and waist-to-hip ratio (WHR), plasma TNF-alpha was still correlated with WBGD (r = -0.33, P < 0.007). Further correction for plasma free fatty acid (FFA) concentration made the latter correlation no more significant. In a multivariate analysis, a model made by age, sex, BF, fat- free mass, WHR, and plasma TNF-alpha concentrations explained 69% of WBGD variability with age (P < 0.009), BF (P < 0.006), fat-free mass (P < 0.005), and plasma TNF-alpha (P < 0.05) significantly and independently associated with WBGD. In the longitudinal study, made with subjects at the highest tertiles of plasma TNF-alpha concentration (n = 50), plasma TNF-alpha concentration predicted a decline in WBGD independent of age, sex, BF, WHR [relative risk (RR) = 2.0; 95% confidence intervals (CI) = 1.2-2.4]. After further adjustment for plasma fasting FFA concentration, the predictive role of fasting plasma TNF-alpha concentration on WBGD (RR = 1.2; CI = 0.8-1.5) was no more significant. In conclusion, our study demonstrates that plasma TNF-alpha concentration is significantly associated with advancing age and that it predicts the impairment in insulin action with advancing age.