Translational mixed-effects PKPD modelling of recombinant human growth hormone - from hypophysectomized rat to patients

Abstract

Background and purpose: We aimed to develop a mechanistic mixed-effects pharmacokinetic (PK)-pharmacodynamic (PD) (PKPD) model for recombinant human growth hormone (rhGH) in hypophysectomized rats and to predict the human PKPD relationship.

Experimental approach: A non-linear mixed-effects model was developed from experimental PKPD studies of rhGH and effects of long-term treatment as measured by insulin-like growth factor 1 (IGF-1) and bodyweight gain in rats. Modelled parameter values were scaled to human values using the allometric approach with fixed exponents for PKs and unscaled for PDs and validated through simulations relative to patient data.

Key results: The final model described rhGH PK as a two compartmental model with parallel linear and non-linear elimination terms, parallel first-order absorption with a total s.c. bioavailability of 87% in rats. Induction of IGF-1 was described by an indirect response model with stimulation of kin and related to rhGH exposure through an Emax relationship. Increase in bodyweight was directly linked to individual concentrations of IGF-1 by a linear relation. The scaled model provided robust predictions of human systemic PK of rhGH, but exposure following s.c. administration was over predicted. After correction of the human s.c. absorption model, the induction model for IGF-1 well described the human PKPD data.

Conclusions: A translational mechanistic PKPD model for rhGH was successfully developed from experimental rat data. The model links a clinically relevant biomarker, IGF-1, to a primary clinical end-point, growth/bodyweight gain. Scaling of the model parameters provided robust predictions of the human PKPD in growth hormone-deficient patients including variability.

Figure 1

Plots of raw data from the included studies following i.v. (A) and s.c. (B) administration of rhGH, IGF‐1 induction following s.c. administration of rhGH (C) and one‐week IGF‐1 (D) and bodyweight (E) measurements following daily treatment, overlaid with a loess‐smoothed curve. Additionally, plots of raw data from the vehicle groups in the two long‐term studies (n = 20) indicating mean (±95% confidence interval) of bodyweight (F) and IGF‐1 (G) measurements over time. Colours correspond to dose levels of 0 μg (•), 44 μg (), 221 μg (), 1106 μg () and 3319 μg () rhGH absolute (PD studies) or per 100 g bodyweight (PKPD study).

Figure 2

Final model structure used in the mixed‐effects PKPD analysis of rhGH.

Figure 3

Prediction‐corrected VPCs to assess the models ability to describe the observed data and its variability. The time‐courses shown are: (A) rhGH following i.v. administration, (B) rhGH following s.c. administration, (C) IGF‐1 induction following s.c. administration of a single rhGH dose, (D) steady‐state levels of IGF‐1 and (E) bodyweight gain following daily s.c. doses of rhGH. The () represent observed data, the () represents the median of the observed data and the () represent the median of the 2.5 % and 97.5 % outer observations. The () represents a simulation‐based 95% confidence interval for the median; while the () represent a simulation‐based 95% confidence interval for the 2.5 % and 97.5 % model‐predicted percentiles.

Figure 4

Prediction‐corrected VPCs used to evaluate prediction of human PK following an i.v. bolus of rhGH. The graphs represent predictions with the distribution volumes scaled with an exponent of 0.8 (A), 0.9 (B) and 1.0 (C). The () represent observed data, the () represents the median of the observed data and the () represent the median of the 2.5 % and 97.5 % outer observations. The () represents a simulation‐based 95% confidence interval for the median; while the () represent a simulation‐based 95% confidence interval for the 2.5 % and 97.5 % model‐predicted percentiles.

Figure 5

Prediction‐corrected VPCs used to evaluate s.c. absorption (A) and corrected s.c. absorption (B) in order to assess prediction of IGF‐1 induction (C). The () represent observed data, the () represents the median of the observed data and the () represent the median of the 2.5 % and 97.5 % outer observations. The () represents a simulation‐based 95% confidence interval for the median; while the () represent a simulation‐based 95% confidence interval for the 2.5 % and 97.5 % model‐predicted percentiles.