Aims/hypothesis: We aimed to identify which surrogate index of insulin sensitivity has the strongest correlation with the reference measurement, the hyperinsulinaemic-euglycaemic clamp (HEC), to determine which surrogate measure should be recommended for use in large-scale studies.
Methods: A literature search (1979-2012) was conducted to retrieve all articles reporting bivariate correlations between the HEC and surrogate measures of insulin sensitivity (in fasting samples or during the OGTT). We performed a random effects meta-analysis for each surrogate measure to integrate the correlation coefficients of the different studies.
Results: The OGTT-based surrogate measures with the strongest pooled correlations (r) to the HEC were the Stumvoll metabolic clearance rate (Stumvoll MCR; r = 0.70 [95% CI 0.61, 0.77], n = 5), oral glucose insulin sensitivity (OGIS; r = 0.70 [0.57, 0.80], n = 6), the Matsuda index (r = 0.67 [0.61, 0.73], n = 19), the Stumvoll insulin sensitivity index (Stumvoll ISI; r = 0.67 [0.60, 0.72], n = 8) and the Gutt index (r = 0.65 [0.60, 0.69], n = 6). The fasting surrogate indices that correlated most strongly with the HEC and had narrow 95% CIs were the revised QUICKI (r = 0.68 [0.58, 0.77], n = 7), the QUICKI (r = 0.61 [0.55, 0.65], n = 35), the log HOMA-IR (r = -0.60 [-0.66, -0.53], n = 22) and the computer generated HOMA of insulin sensitivity (HOMA-%S; r = 0.57 [0.46, 0.67], n = 5).
Conclusions/interpretation: The revised QUICKI fasting surrogate measure appears to be as good as the OGTT-based Stumvoll MCR, OGIS, Matsuda, Stumvoll ISI and Gutt indices for estimating insulin sensitivity. It can therefore be recommended as the most appropriate index for use in large-scale clinical studies.