The present study examined the associations between a major adipokine, adiponectin, and adiposity indices as well as metabolic risk variables in a sample of 190 untreated asymptomatic men. Anthropometric measurements and a complete fasting plasma lipoprotein and lipid profile were obtained, and subjects underwent an oral glucose tolerance test. Fasting plasma adiponectin concentrations were determined by an ELISA. Although all adiposity and adipose tissue (AT) distribution indices were negatively correlated with plasma adiponectin levels (-0.14 </= r </= -0.32; P < 0.04), multiple regression analyses revealed that visceral AT accumulation was the only independent predictor of adiponectin levels, with 10% of its variance explained by visceral AT (P < 0.0001). Comparison of obese men with similar body mass index values (>/=30 kg/m(2)) but who markedly differed in their level of visceral AT (< vs. >/=130 cm(2); n = 15) revealed significant differences in adiponectin levels (7.0 +/- 3.0 vs. 11.1 +/- 4.9 microg/ml; P < 0.02 for men with high vs. low visceral AT, respectively). Finally, when men were stratified into tertiles of visceral AT and further classified on the basis of the 50th percentile of adiponectin levels (</= vs. >8.8 microg/ml), a 3 x 2 ANOVA revealed an independent contribution of adiponectin on the variation of high-density lipoprotein cholesterol levels (P < 0.002) and of the glucose area (P < 0.02). These results support the notion that adiponectin concentration is influenced to a greater extent by visceral than sc obesity. Furthermore, adiponectin predicts glucose tolerance and plasma high-density lipoprotein cholesterol levels in a manner that is partly independent from the contribution of visceral adiposity.