A parsimonious model of a mixed-meal test in patients with type 1 diabetes in insulin pump therapy

Background and aims: Currently available closed-loop insulin deliverysystems stem from sophisticated models of the glucose-insulin (G/I) systemmostly based on complex studies employing glucose tracer technology.We asked the question whether simpler studies based on the minimalmodeling approach (Bergman, 1981; Cobelli, 2007) could likewise reliablydescribe the G/I system during a mixed meal test (MMT) in patientswith type 1 diabetes (T1D). If effective, this approach would expedite thecreation of large bio-banks of virtual patients to develop robust models“to close the loop”. To this end, we tested the performance of a minimalmodel of the G/I system to characterize the glucose (G) and insulin (I)dynamics during MMT in T1D patients on insulin pump therapy (CSII).Materials and methods: In six T1D patients on CSII enrolled in theMMT-T1D Pilot Study we assessed on three separate days: (1) insulinsensitivity, by the hyperinsulinemic euglycemic clamp (HEC); (2) the G/Itime-courses during a standardized 5 h-MMT (MMT1: 292 Kcal; 38.9 gcomplex CHO, 8.9 g lipids, 14 g proteins); (3) theG/I time-courses duringa second identical (3 patients) or double-sized (3 patients) MMT(MMT2). The parameters estimated by modeling of the HEC were usedto cast a comprehensiveMMTmodel (GLUKINSLOOP.2), including theinsulin delivery system and the metabolic G/I system of each patient.GLUKINSLOOP.2 was implemented and run in the SAAM2.1 software.Results: The GLUKINSLOOP.2 model identified the G/I system parameters(among others: insulin sensitivity, SI; glucose effectiveness, SG;glucose distribution volume, Vd; time of oral carbohydrate appearancein the peripheral circulation, expressed as ICMTT, Intestinal CarbohydrateMean Transit Time) and provided a good fit of the G/I timecoursesin all studies, as proved by the analysis of mean weighted residuals(WR, mean ± SD; MMT1: WR(G)=-0.03±0.71; WR(I)=0.33±1.04; MMT2: WR(G)=0.09±0.86; WR(I)=0.01±1.37). Both the identicaland double-sized repeated MMT2s showed good reproducibility ofthe G/I system parameters (mean±SEM; MMT1: SI=0.57±0.12(mL˙min-1)/(pmol˙L-1); SG=15.6±10.1 mL˙min-1; Vd=11,482±1,131 mL; ICMTT=105±14 min; MMT2: SI=0.57±0.13; SG=38.2±13.6; Vd=11,609±1,031; ICMTT=101±11).Conclusion: The GLUKINSLOOP.2 model herein presented can providea fairly good and reproducible description of the G/I system in T1Dpatients on CSII and it may be used to build a bank of “virtual patients”.Our results might be relevant to strategies directed to improve the architectureof upcoming closed-loop CSII systems.Clinical Trial Registration Number: NCT01800734Supported by: EFSD/Novo Nordisk