Table 3.

Bias comparing Matsuda index and 1/HOMA-IR to insulin sensitivity derived from the hyperinsulinemic euglycemic clamp, adjusting for selected covariates

	Matsuda		1/HOMA-IR
	Model 1	Model 2	Model 1	Model 2
CKD
%Difference (95% CI)*	−15 (−30, 3)	−16 (−31, 1)	−23 (−37, −5)	−24 (−38, −6)
P value	0.09	0.06	0.02	0.01
BMI
%Difference (95% CI)*	−36 (−49, −19)	−34 (−47, −17)	−46 (−58, −31)	−44 (−56, −29)
P value	0.0002	0.0003	<0.0001	<0.0001

CKD, chronic kidney disease; CI, confidence interval.

^*Cell contents are %differences in surrogate insulin sensitivity indices (Matsuda index or 1/HOMA-IR, modeled separately as parallel log-transformed-dependent variables) associated with CKD (estimated GFR: <60 ml·min⁻¹·1.73m²) vs. no CKD or obesity (BMI ≥30 kg/m²) vs. no obesity.

Estimates were derived from multivariable regression models that included SI_clamp, age, sex, race, CKD, and obesity as independent variables (model 1) or these independent variables plus liver/spleen density and aspartate aminotransferase and alanine aminotransferase levels (model 2). Thus, in model 1 assessing the Matsuda index, participants with CKD were estimated to have a mean Matsuda index that was 15% lower than those without CKD at the same SI_clamp, age, sex, race, and BMI status.