Meal Tolerance Test (MTT): Nonlinear Mixed-Effects Modeling of Insulin Secretion
A. Largajolli (1), A. Bertoldo (1), C. Cobelli (1)
(1) Department of Information Engineering, University of Padova, Italy
Objectives: The C-peptide oral minimal model (COMM) allows to quantitatively assess the β-cell function (1). Up to now, the COMM has been identified at individual level (IND). This study aims at comparing the performance of the NLMEM to both the full and reduced MTT protocol (2). We also investigate the performance of the NLMEM techniques by randomly discarding the C-peptide data of the reduced meal protocol, firstly, by removing 25% and, then, by removing 50% of the original samples.
Methods:The analysis was performed on a dataset of 204 nondiabetic subjects that received a mixed meal (2) considering the full (21 samples in 7 hours) and the reduced (7 samples in 2 hours) protocol. Once COMM parameters are estimated, three β-cell responsivity indices can be derived: static (Φs, 10-9 min-1), dynamic (Φd, 10-9) and total sensitivity (Φt, 10-9 min-1). COMM was implemented and identified in NONMEM VI by using FOCE with interaction. The population parameter distributions were assumed lognormal, BSV was modeled with a full covariance matrix and the C-peptide (C) measurement error was modeled as Var(C)=2000+0.001∙(C)^2.
Results: The population description obtained from FOCE using the full MTT dataset in terms of fixed effects and BSV is comparable with the first and second-order moment obtained with IND. We detect an overestimation of the BSV with IND as reported in (3-4). The linear regression analysis between FOCE vs IND estimates results having a highest correlation of r=0.96 and a lowest correlation of r=0.66. The mean β-cell indexes obtained with the full MTT are the following (FOCE vs IND): Φs is 32.55±10.92 vs 33.9 ±13.85, Φd is 418.57±207.63 vs 374.32 ±290.21 and Φt is 143.58±104.44 vs 140.84±126.51. Regarding FOCE, β-cell values do not change with the reduced or the two discarded reduced protocols and the correlation values are: a)full vs reduced protocol: rs=0.87, rd=0.96, rt=0.98 b)full vs randomly 25% reduced protocol: rs=0.86, rd=0.94, rt=0.97 c)full vs randomly 50% reduced protocol: rs=0.81, rd=0.88, rt=0.95.
Conclusions: NLMEM can be successfully used to estimate the COMM parameters. The population description is comparable to that obtainable considering the IND approach. COMM parameter estimates with FOCE are not significantly different in a data poor context. This paves the way to other studies that aim to further narrow down the reduced protocol in order to better deal with the typical data poor epidemiological study condition.
References:
[1]Oral glucose tolerance test minimal model indexes of β-cell function and insulin sensitivity. Breda E, Cavaghan M K, Toffolo G, Polonsky K S, Cobelli C. 2001, Diabetes, Vol. 50, pp. 150-158.
[2] Two-hour Seven-sample oral glucose tolerance test and meal protocol. Minimal model assessment of beta-cell responsivity and insulin sensitivity in nondiabetic individuals. Dalla Man C, Campioni M, Polonsky K S, Basu R, Rizza R A, Toffolo G, Cobelli C. 2005, Diabetes, Vol. 54, pp. 3265-3273.
[3]Nonlinear models for repeated measurements data. M. Davidian, D.M. Giltinian. Boca Raton, Fla.: Chapman & Hall/CRC, 1998.
[4]Nonlinear mixed effects to improve glucose minimal model parameter estimation: a simulation study in intensive and sparse sampling. Denti P, Bertoldo A, Vicini P, Cobelli C. 2009, IEEE Trans Biomed Eng, Vol. 56, pp. 2156-66.