Methodological issues in the application of the minimal model: effects of glucose dose, basal glucose concentration,...

An intravenous glucose tolerance test (IVGTT) glucose dose of 0.3 g/kg has been adopted for measurement of insulin sensitivity using the minimal model. Traditionally, however, a dose of 0.5 g/kg has been used, which might be expected to improve the IVGTT insulin response and hence the effectiveness of minimal model analysis. In a preliminary study of 5 subjects given 0.3 and 0.5 g/kg IVGTTs, each lasting 120 minutes, we found a 53% increase in IVGTT insulin response at the higher dose, but there was a marked discrepancy in the difference between fasting and final glucose concentrations (-0.17 mmol.l-1 at 0.3 g/kg and -1.01 mmol.l-1 at 0.5 g/kg). Good agreement obtained between estimates of Si derived from 0.3 and 0.5 g/kg IVGTTs (mean Si: 0.3 g/kg test = 6.0 min-1.microU-1.ml, 0.5 g/kg test = 5.8 min-1.microU-1.ml: r = 0.97, p < 0.001) providing the final IVGTT glucose concentration rather than the fasting concentration was taken as the basal level for modelling analysis. These findings were confirmed and extended in two further studies; firstly in an analysis of the effects of choice of basal glucose concentration and the application of various modelling constraints in a cross-section of 66 subjects with a wide range of insulin sensitivities; then in a further study in which seven subjects were each given two 0.3 g/kg IVGTTs and one 0.5 g/kg IVGTT, with each test being prolonged to 300 minutes. Agreement between estimates of Si at the two different doses was again only achieved by taking the final IVGTT glucose concentration as basal (mean Si: 0.3 g/kg test = 4.7 min-1.microU-1.ml, 0.5 g/kg test = 3.8 min-1.microU-1.ml: r = 0.75, p < 0.05), although one anomalous test required that a constraint be applied to the modelling process for this agreement to obtain. Closest agreement between the 300 minute 0.3 and 0.5 g/kg IVGTTs was found when tests were modelled up to 180 minutes. An IVGTT duration of 180 minutes appears to be optimum for re-establishing the basal concentration necessary for effective modelling analysis. Application of constraints can markedly affect certain analyses and may introduce some bias; their use should be carefully monitored, although their effect on large datasets is likely to be small.