Insulin Sensitivity Indexes From a Single Sample in Nonobese Japanese Type 2 Diabetic Patients: Comparison with minimal model analysis

Insulin resistance is important not only in obesity and diabetes but also in essential hypertension, dyslipidemia, and atherosclerotic cardiovascular disease (1). We recently demonstrated that the homeostasis model assessment for insulin resistance (HOMA-IR) proposed by Matthews et al. (2) is validated against the minimal model–derived insulin sensitivity index (MINMOD-SI) in type 2 diabetic patients (3). Although the relationship between HOMA-IR and MINIMOD-SI in these patients was statistically significant, the correlation coefficient was low (r = 0.459, P = 0.021) (3). Therefore, a more highly correlated simple index is needed to measure insulin resistance in large type 2 diabetic populations.

The major drawback of HOMA-IR is that when the glucose or insulin concentration increases, the value is overestimated. Furthermore, HOMA-IR is not linear over wide ranges of insulin sensitivity in man (4). Katz et al. (5) recently proposed the quantitative insulin sensitivity check index (QUICKI) [1/(log glucose + log insulin)] as a novel index of insulin sensitivity. They demonstrated that the correlation between the euglycemic clamp and QUICKI is significantly better than the correlation between the euglycemic clamp and HOMA-IR in 56 subjects, including diabetic patients. We proposed another three insulin sensitivity indexes based on QUICKI. To the best of our knowledge, however, the relationships between MINMOD-SI and these formulas, including QUICKI, have not been fully investigated in type 2 diabetic patients.

Twenty-five nonobese Japanese type 2 diabetic patients underwent the minimal model approach to measure insulin sensitivity index as described previously (6). Their age (mean ± SD), was 45.4 ± 11.5 years (range 26–58), and their BMI was 20.2 ± 3.0 kg/m² (13.5–25.3). The HbA_1c level was 7.2 ± 1.5% (5.1–11.3). Fasting glucose and insulin levels were 6.7 ± 1.7 mmol/l (5.0–11.9) and 4.5 ± 1.5 μU/ml (1.7–8.4), respectively. Their insulin sensitivity index was 9.0 ± 7.1 [(μU/ml)⁻¹ min⁻¹] (range 3.2–31.3). Type 2 diabetes was diagnosed based on the criteria of the World Health Organization (7). Five patients were treated with sulfonylureas and the rest with diet alone. One might argue that the use of sulfonylureas in patients with diabetes might significantly affect the estimate of insulin resistance by HOMA, as these drugs are known to decrease fasting plasma glucose without substantially changing fasting plasma insulin (8). However, it seems unlikely because Bonora et al. (9) confirmed that in the validation studies of HOMA, the correlation of insulin sensitivity estimated by this method and that measured by the glucose clamp was not substantially different in diet- and sulfonylurea-treated type 2 diabetes.

The statistical analysis was performed with the Statview 5.0 system (Statview, Berkeley, CA). Spearman’s correlation coefficient by rank was used for the analysis. P < 0.05 was considered as significant.

The insulin sensitivity index obtained from the minimal model approach was best correlated with 1/(log glucose × log insulin) (r = 0.658, P < 0.001), followed by 1/(insulin × log glucose) (r = 0.615, P < 0.001), 1/(glucose × log insulin) (r = 0.596, P < 0.001), QUICKI (r = 0.521, P < 0.005), and HOMA-IR (r = 0.459, P = 0.021) in our diabetic patients.

In conclusion, although the current study was performed in a limited number of patients (n = 25) with a relatively narrow range of insulin sensitivity (3.2–31.3 [(μ U/ml)⁻¹ min⁻¹], it can be concluded that 1/(log glucose × log insulin) is highly correlated with the minimal model–derived insulin sensitivity in nonobese Japanese type 2 diabetic patients. Furthermore, the correlation coefficient between MINMOD-SI and 1/(log glucose × log insulin) (r = 0.658) was similar to that between MINMOD-SI and the insulin sensitivity index (ISI, composite) (r = 0.677) proposed by Matsuda and DeFronzo (3,10). Whereas the ISI (composite) requires an oral glucose tolerance test, 1/(log glucose × log insulin) can be calculated from the fasting state. Thus, 1/(log glucose × log insulin) is considered to be a simple and useful tool for the estimation of insulin resistance in nonobese Japanese type 2 diabetic patients.