Kotani et al. reported results on the association between coffee consumption and serum adiponectin concentrations in Japanese men and women (1). They concluded that their study does not confirm our recent observation that frequent consumption of coffee is associated with higher adiponectin concentrations in a population of U.S. women (2) and suggest that ethnic differences may have contributed to this difference in results. However, one could offer several other potential explanations for the apparent discrepancy.
First, the study by Kotani et al. may have lacked statistical power to detect an association between coffee consumption and serum adiponectin because it included only 14 men and 23 women who consumed ≥4 cups/day. We observed 1.8 μg/ml higher adiponectin concentrations for nondiabetic women consuming ≥4 cups/day compared with <1 cup/week, but this association was not apparent for lower levels of coffee consumption (2). In the study by Kotani et al., the differences in adiponectin concentrations for ≥4 cups/day compared with <1 cup/day were 0.5 μg/ml (95% CI −2.3 to 3.3) for men and 0.0 μg/ml (−3.1 to 3.1) for women. These CIs show that the study by Kotani et al. does not exclude the difference in adiponectin concentrations that we observed. Thus, further studies including greater numbers of frequent coffee consumers are needed to confirm or refute our results on coffee consumption and adiponectin concentrations.
Second, residual confounding cannot be excluded as an explanation for results in observational studies. We adjusted for age and BMI in multivariate models, and additional adjustment for physical activity, energy intake, alcohol intake, and smoking strengthened the results. Additional adjustment for biological risk factors or medication use did not explain the observed association between coffee consumption and adiponectin in either the women with or without diabetes. Kotani et al. considered fewer potential confounders, and residual confounding may have obscured existing associations. As we concluded in our paper, however, randomized trials of the effects of coffee consumption on adiponectin concentrations will eventually be needed to establish beyond any doubt whether the observed association represents a causal effect.
