Tomohisa Saito (1), Nahoko Kasahara-Ito (1), Takaaki Ishida (1), Satofumi Iida (1), Kimio Terao (1)
(1) Chugai Pharmaceutical. Co., Ltd.
Objectives: Type 2 diabetes mellitus (T2DM) is a chronic metabolic disorder and has serious impacts on our lives. Most existing drugs for treating T2DM aim to regulate blood glucose levels by directly enhancing insulin action. On the other hand, sodium-glucose co-transporter 2 (SGLT2) inhibitors increase urinary excretion of excess glucose by inhibiting renal glucose reabsorption via SGLT2 and reduce blood glucose levels without risk of hypoglycemia. Tofogliflozin that was discovered and synthesized by Chugai Pharmaceutical Co., Ltd. (Tokyo, Japan) is one of the highly selective SGLT2 inhibitors administered orally for the treatment of T2DM patients [1] and was launched in Japan in 2014. We developed a population pharmacokinetic (PK) model to understand clinical pharmacology of tofogliflozin and to evaluate covariate effect on PK profile of tofogliflozin.
In addition, in order to evaluate the accuracy of parameters with real data, we used different types of evaluation methods to calculate prediction intervals of estimated parameters. A bootstrap method is a golden standard to evaluate the prediction intervals of the parameters and sampling importance resampling method has been in the rise to be used for the evaluation in recent years [2]. Therefore, we used these methods for the evaluation of prediction intervals for our actual concentration data.
Methods: Six clinical studies (CSG002JP, CSG003JP, CSG004JP, CSG006JP, CSG007JP, CSG010JP) were combined to prepare a dataset for the population analysis [3][4][5][6]. The dataset includes doses (range: 2.5 to 80 mg) and the following recruited subjects: 55 healthy volunteers, 9 patients with moderate hepatic impairment, 333 patients with T2DM, and 23 T2DM patients with moderate renal impairment. Full sampling has been done for the studies in healthy volunteers and some T2DM patients and trough sampling has been done for long term studies. A covariate search was performed using a Perl speaks NONMEM scm command [7] with stepwise forward and backward procedures.
After the model building, non-parametric bootstrap, parametric bootstrap and sampling importance resampling (SIR) methods were tested as the evaluation of parameter estimation using the tofogliflozin’s data [2].
All analyses were performed using NONMEM® version 7.4.1 (ICON Developed Solutions, Ellicott City, Maryland) with a fortran compiler, gFORTRAN 4.6.0 and first order conditional estimation method with INTERACTION option was applied to estimate the population parameters.
Results: A three-compartment model with transit absorption compartments assuming a log-normal distribution well described the PK profile of tofogliflozin. The estimated population mean of CL/F, V1/F, Q2/F, V2/F, V3/F and MTT were 10.9 (L/h), 38.8 (L), 0.255 (L/h), 4.96 (L), 13.4 (L), and 24.2 (L), respectively. The inter individual variability of each parameter was less than 60%. No clinically relevant covariates were identified in the covariate search though food intake at administration of tofogliflozin and body weight were incorporated in the model as structural covariates before the search. Body weight was included with allometry function using the power of 0.75 for CL/F and 1 for V/F. Post-meal condition increased MTT by 231%.
The accuracy of the estimated parameters was assessed with some methods. The SIR method and the parametric bootstrap method showed almost the same prediction intervals of each parameters whereas the non-parametric bootstrap method showed a slightly wider interval. The computing time of the SIR method was about 1 day while that of the parametric bootstrap method was about 20 days. However, it takes 2 to 3 days to find the sophisticated setting of SIR method.
Conclusions: The final model well fitted with the observed data. We tested different types of evaluation methods for parameters and the SIR method using tofogliflozin concentration data showed reasonable results with shorter computing time even though the setting of the calculation was relatively complex.
References:
[1] Suzuki M, Honda K, Fukazawa M, Ozawa K, Hagita H, Kawai T, Takeda M, Yata T, Kawai M, Fukuzawa T, Kobayashi T, Sato T, Kawabe Y, Ikeda S. J Pharmacol Exp Ther (2012) 341: 692–701.
[2] Dosne AG, Bergstrand M, Karlsson MO. J Pharmacokinet Pharmacodyn (2017) 44(6): 509-520.
[3] Kasahara-Ito N, Fukase H, Ogama Y, Saito T, Ohba Y, Shimada S, Takano Y, Ichihara T, Terao K, Nakamichi N, Kumagai Y, Ikeda S. Drug Res (Stuttg) (2017) 67(6): 349-357.
[4] Kaku K, Watada H, Iwamoto Y, Utsunomiya K, Terauchi Y, Tobe K, Tanizawa Y, Araki E, Ueda M, Suganami H, Watanabe D. Cardiovasc Diabetol (2014) 13: 65.
[5] Tanizawa Y, Kaku K, Araki E, Tobe K, Terauchi Y, Utsunomiya K, Iwamoto Y, Watada H, Ohtsuka W, Watanabe D, Suganami H. Expert Opin Pharmacother, (2014) 15(6): 749-766.
[6] Ikeda S, Takano Y, Schwab D, Portron A, Kasahara-Ito N, Saito T, Iida S. Drug Res (Stuttg) (2018) Aug 13. [Epub ahead of print].
[7] Khandelwal A, Harling K, Jonsson EN, Hooker AC, Karlsson MO. AAPS J. (2011) 13(3): 464-472.
Reference: PAGE 28 (2019) Abstr 8827 [www.page-meeting.org/?abstract=8827]
Poster: Drug/Disease Modelling - Endocrine