The level of leptin increases with obesity, whereas that of adiponectin decreases with obesity. It is reported that the ratio of leptin to adiponectin (L/A) is associated with insulin resistance. It is difficult to evaluate insulin resistance in diabetic patients who have a dysfunction of insulin secretion. The aim of this study was to examine whether the L/A ratio is a useful marker for insulin resistance in diabetic patients. We examined L/A in the serum of a total of 139 Japanese patients with type 2 diabetes mellitus (66 women and 73 men) and 7 healthy individuals recruited in our hospital. Changes in the levels of leptin and adiponectin were observed using the oral glucose tolerance test and a hyper- and euglycemic clamp test. Twenty-one patients with type 2 diabetes mellitus were observed for more than 6 months after treatment with pioglitazone, and 31 patients with type 2 diabetes mellitus were observed for more than 6 months after the treatment with metformin. The mean value of L/A in 139 Japanese patients with type 2 diabetes mellitus was 1.22 +/- 1.41 (1.68 +/- 1.76 in women, 0.81 +/- 0.80 in men; P = .0002). In the clamp tests, L/A correlated with glucose infusion rate (GIR) (r(2) = 0.26, P = .0034). The correlation of L/A and GIR indicated a stronger correlation than either leptin (r(2) = 0.144, P = .03) or adiponectin alone (r(2) = 0.023, P = .41), or the homeostasis model assessment of insulin resistance (r(2) = 0.103, P = .08). The average hemoglobin A(1c) (HbA(1c)) improved from 10.2% +/- 1.2% to 9.2% +/- 1.6% (P = .0037) in 6 months after treatment with pioglitazone. Our results indicate pioglitazone to be effective for HbA(1c) improvement in subjects with high L/A and low L/A. The average HbA(1c) improved from 9.2% +/- 0.9% to 8.0% +/- 1.2% (P = .0002) in 6 months after treatment with metformin. Our results indicate metformin to be effective for HbA(1c) improvement in subjects with a low L/A. In conclusion, we demonstrate that L/A is different between male and female subjects. The correlation of L/A and GIR by the euglycemic hyperinsulinemic clamp test suggests that L/A is a useful indicator for the choice of drug to treat diabetes mellitus.