Estimation of Population Insulin Sensitivity (S_I) from Sparsely Sampled Oral Glucose Tolerance Test (OGTT) Data

The oral glucose tolerance test (OGTT) minimal model analysis by Cobelli et al. obtains estimates of insulin sensitivity (S_I) based on mathematical models that are highly concordant with the euglycemic clamp, the gold standard for estimation S_I. This model requires a minimum of 7 blood samples during the 2-hour OGTT. However, in many OGTT studies, the number of samples is below the number that permits the estimation of S_I. One way to address (overcome) this issue is by pooling sparsely sampled data (<7 samples) from a specific cohort to obtain population estimates of parameters of S_I. Here we present a novel methodology of estimating population S_I using sparsely-sampled OGTT protocol. A cohort of 6 African American subjects (3 Female/ 3 Male) without a known history of T2D underwent 75-gm OGTTs with 6 samples performed at 0, 15, 30, 60, 90 and 120 minutes. The population S_I was estimated concurrently fitting the data from all subjects. We obtained population estimates of S_I using from 3 to 6 samples per subject. S_I estimates for each subject were also calculated. Sparse-sampling population estimated S_I varied from S_I=4.74±.75 10^-4.(mu/l)^-1.min^-1 for 4 samples to S_I= 6.40±2.40 10^-4.(mu/l)^-1.min^-1 for 3 samples. The individual estimates of S_I had a mean of S_I= 4.72±2.43 10^-4.(mu/l)^-1.min^-1. S_I obtained with the sparse-sampling methodology did not differ from the mean individual estimates of S_I. We conclude that Sparse-sampling estimation of S_I (<7 samples) is a useful tool for estimation of population parameters of S_I where there are not enough observations per subject to estimate individual insulin sensitivity. This methodology will provide better phenotyping of insulin sensitivity for studies where previously more crude measures were used.