The purpose of the study is to compare surrogate estimates of insulin sensitivity with a directly measured insulin sensitivity index, steady-state plasma glucose (SSPG) from insulin suppression test (IST), in subjects with hypertension. Two hundred and twenty-eight hypertensive patients who received IST for SSPG were included for analysis. Estimates from fasting measurements alone, homeostasis model assessment for insulin resistance (HOMA-IR) and quantitative insulin sensitivity check index (QUICKI)), and indices from fasting and/or 2 h samples (ISI 0,120 and ISI TX ) were calculated. In addition to Pearson and partial correlations, variancecomponent models were used to test the relationship between surrogate estimates of insulin sensitivity and SSPG. A large proportion of variance owing to covariates in the variance-component models indicated the goodness of model fit, irrespective of the independence among variables. SSPG was positively correlated with logarithmic transformation (Log) (HOMA-IR) and negatively correlated with QUICKI, Log (ISI 0,120 ) and ISI TX (all Po0.0001). Log (ISI 0,120 ) seemed to have a better correlation with SSPG (r ¼ À0.72) than other measures in partial correlation. The proportion of variance owing to all covariates of Log (ISI 0,120 ) and ISI TX were larger than those of Log (HOMA-IR) and QUICKI in the variancecomponent models. After adjustments for demographic and obesity covariates, the proportion of variance explained by Log (ISI 0,120 ) were largest among the surrogate measures in the variance-component models. Our results showed that ISI 0,120 and ISI TX correlated better with SSPG than those used fasting measures alone (HOMA-IR and QUICKI). Log (ISI 0,120 ) currently showing the strongest association with SSPG than other estimates is adaptable for use in large studies of hypertension.