A Model-Informed Drug Development (MIDD) Approach for a Low Dose of Empagliflozin in Patients with Type 1 Diabetes

In clinical trials, sodium-glucose co-transporter (SGLT) inhibitor use as adjunct to insulin therapy in type 1 diabetes (T1D) provides glucometabolic benefits while diabetic ketoacidosis risk is increased. The SGLT2 inhibitor empagliflozin was evaluated in two phase III trials: EASE-2 and EASE-3. A low, 2.5-mg dose was included in EASE-3 only. As the efficacy of higher empagliflozin doses (i.e., 10 and 25 mg) in T1D has been established in EASE-2 and EASE-3, a modeling and simulation approach was used to generate additional supportive evidence on efficacy for the 2.5-mg dose. We present the methodology behind the development and validation of two modeling and simulation frameworks: M-EASE-1, a semi-mechanistic model integrating information on insulin, glucose, and glycated hemoglobin; and M-EASE-2, a descriptive model informed by prior information. Both models were developed independently of data from EASE-3. Simulations based on these models assessed efficacy in untested clinical trial scenarios. In this manner, the models provide supportive evidence for efficacy of low-dose empagliflozin 2.5 mg in patients with T1D, illustrating how pharmacometric analyses can support efficacy assessments in the context of limited data.

. Covariate forest plot on normalized AUCss for the final PK model. Point ranges represent the median (point) and 95% confidence interval (range) for the covariate effect based upon 500 simulations including parameter uncertainty. The shaded area marks covariate effect from 0.8 to 1.25. Reference subject: male, nonsmoker, total insulin dose = 0.6 IU/kg, AP = 73 IU/kg, TPRO = 68 g/L, eGFR = 99 mL/min/1.73 m 2 , and weight = 70 kg. AP, alkaline phosphatase; AUC, area under the curve; AUCss, area under the curve at steady-state; eGFR, estimated glomerular filtration rate; PK, pharmacokinetic; TPRO, total protein; WT, patient weight.
. Figure S2. M-EASE-1: External model evaluation for EASE-3 (out-of-sample) by longitudinal visual predictive check by dose for (a) HbA1c, (b) TDID, and (c) MDG. Red lines represent the 96.5 th , 50 th , and 2.5 th percentiles over 500 simulations. The red area is the 95% CI associated with these metrics. The interval between the 97.5 th and 2.5 th percentile is the 95% prediction interval. Clue lines represent the corresponding observed metrics. Whiskers on box plots represent 1.5× the IQR, with black dots representing observed data falling outside of 1.5× the IQR. CI, confidence interval; HbA1c, glycated hemoglobin; IQR, interquartile range; MDG, mean daily glucose; TDID, total daily insulin dose.
. Figure S4. M-EASE-2: Posterior predictive check for EASE-3 (out-of-sample) changes from baseline HbA1c by dose and week. Bar graphs are based on 500 simulations. The red line indicates the observed median delta value. The shaded interval indicates ±1.96 SE of the observed data. HbA1c, glycated hemoglobin; SE, standard error.
. Figure S6. Simulated empagliflozin AUCss by dose using final population pharmacokinetic model. Gray points represent individual observed steady-state AUC values. Box plots summarize simulated steady-state exposures. AUC, area under the curve; AUCss, area under the curve at steady-state.

Evaluation
The time course of the placebo effect could be sufficiently described for internal and external data.

M-EASE-1
Assumption Change in TDID can be described by empagliflozin drug effect

Justification
Due to a lack of information regarding the resolution in the time courses of changes in MDG and insulin and meal or exercise information, TDID was estimated independently from MDG. Therefore, the association of insulin reduction and changes in glucose levels was not considered mandatory to describe the impact of empagliflozin on the longer-term insulin dose changes.
Test Internal and external model evaluation.
Evaluation TDID data were appropriately described for internal and external data.

Assumption Change in HbA1c can be described by MDG levels
Justification MDG levels are affected by behavioral factors such as food intake and exercise, which are implicitly accounted for in the model.

Test
Internal and external model evaluation.

Evaluation
HbA1c change was appropriately described for internal and external data.

Justification Pretreatment optimization in EASE-2 caused a significant decrease in
HbA1c that could not be maintained throughout the study. An increase from Week 4 onward was observed in all randomization groups.
Test Nonlinear placebo models were tested as part of the indirect response model.

Evaluation
The time course of the placebo effect could be sufficiently described for internal and external data.
AUC50, AUCss at which half the maximal effect; AUCss, area under the curve at steady-state; Emax, maximal effect parameter for empagliflozin AUCss on TDID and MDG; Gmax, maximum serum glucose concentration; HbA1c, glycated hemoglobin; IC50, half maximal inhibitory concentration; Imax, maximum inhibition; MDG, mean daily glucose; T1D, type 1 diabetes; T2D, type 2 diabetes; TDID, total daily insulin dose. Full covariate PK model equations in Table S2 = Age, patient age; ALAG, oral absorption lag time; AP, alkaline phosphatase; CI, confidence interval; CL/F, apparent clearance after oral dosing; Cov, covariate; Cur, current; CV, coefficient of variation; eGFR, estimated glomerular filtration rate; Ka, absorption rate constant; PK, pharmacokinetic; Q/F, apparent (oral) intercompartmental clearance; Sex, patient gender; TDID, total daily insulin dose; TPRO, total protein; V2/F, apparent central volume of distribution after oral dosing; V3/F, apparent peripheral volume of distribution after oral dosing; WT, patient weight; ᵟ, residual variability; RSE, relative standard error; Ꞷ, inter-individual variance. a Estimated from placebo only data and fixed in the estimation of the impact of EMPA on TDID time course. b As EASE-2 included a pre-treatment insulin intensification phase and EASE-1 did not, study-specific effects were implemented on baseline insulin dose and the Emax parameter to allow for differences seen in observed data due to study design (see equations below). Although the data for the EASE-2 pre-treatment phase were not included in the analysis, the separate parameter effects were considered necessary for this study to account for the different relative starting point for these patients as affected by the pre-treatment difference. c Estimated from placebo only data and fixed in the estimation of the impact of EMPA on MDG time course. Table S3 (M-EASE  AUC50, AUCss leading to 50% of maximal effect; AUCss, area under the curve at steady-state; Base, baseline; CI, confidence interval; Cov, covariance; CSII, continuous subcutaneous insulin infusion; CV, coefficient of variance; EFF, power coefficient; eGFR, estimated glomerular filtration rate; Emax, maximal effect parameter for EMPA AUCss on HbA1c; EMPA, empagliflozin; HbA1c, glycated hemoglobin; INC, scale parameter reflecting the amplitude for insulin dose adjustment (applies only to EASE-1 during treatment week 1); INS, insulin; INSDT, insulin dose type (MDI vs. CSII); MDG, mean daily glucose; MDI, multiple daily injections; PBO, time-dependent MDG placebo effect; RSE, relative standard error; SEX, patient gender; SD, standard deviation; TDID, total daily insulin dose; WT, patient weight; ᵟ, residual variance; γ, insulin effect; Ꞷ, inter-individual variance.  A) Impact of prior variance on placebo−adjusted predicted median HbA1c change from baseline at 26 weeks.