Fernando Carreño (1,2); Rashmi Mehta (3); Jon Collins (3); Daren Austin (4); Brandon Swift (3)
(1) UNC Eshelman School of Pharmacy, The University of North Carolina at Chapel Hill, Chapel Hill, NC, USA. (2) GlaxoSmithKline, Collegeville, PA, USA. (3) GlaxoSmithKline, Research Triangle Park, NC, USA. (4) GlaxoSmithKline, Great West Road, Brentford, UK
Objectives: The increase in systemic bile acids (BA) observed in primary biliary cholangitis (PBC) patients may play a causal role in cholestatic pruritus (itch) [1]. Linerixibat is a small molecule which selectively inhibits the ileal bile acid transporter blocking the reabsorption of BA in the gastrointestinal tract thereby lowering BA in the systemic circulation and reducing itch in PBC patients [2]. The pharmacological target of linerixibat is in GI lumen hence conventional concentration-response analysis has no utility. A k-PD model was implemented to characterize dose-response in the absence of plasma pharmacokinetic data. The advantage of this approach is the ability to model the pharmacodynamic endpoint over time driven by the kinetics of drug effect as a function of the dose [3].
The aim of this model-based analysis was to 1) develop a population k-PD model that characterizes the relationship between linerixibat dose following once and twice daily oral administration to changes in Worst Daily Itch (WDI) scores in PBC patients and 2) simulate expected changes in WDI scores following different dosing regimens of linerixibat to support Phase 3 dose selection.
Methods: Data from the Phase 2b dose-ranging study (NCT02966834) was included in the analysis [4]. After a placebo run-in period of 4 weeks, PBC patients (n=147) with itch were randomized to one of six dose regimens (placebo, 20 mg, 90 mg, 180 mg QD or 40 mg, 90 mg BID) as oral tablets or matching placebo for 12 weeks. Itch severity was recorded on an electronic diary twice daily (morning and evening) using 0 – 10 numerical rating scale, (NRS), with the worst of these two itch scores used as the WDI score. The WDI scores were then averaged over 7 days and defined as the Mean Worst Daily Itch (MWDI). The WDI was modelled on a continuous bounded scale with a logistic function used to transform the score in the effect compartment from [0, +∞] to the bounded NRS scale [0, 10] [5]. Stepwise covariate modelling (SCM) and parsimony were employed. The model was assessed by conventional goodness of fit methods. Simulations were performed to explore the effect of different linerixibat dose regimens on WDI scores.
Results: An inhibitory indirect response model with a linear placebo effect and an Emax linerixibat effect best described the WDI scores, with inter-individual variability on baseline itch score and placebo effect, and combined additive and proportional residual errors. Of the investigated covariates, only WDI score at randomization was statistically significant. Treatment group differences were not large, consistent with the primary statistical analysis at study Week 16, but a trend for decreased WDI scores with increasing dose was noted. The predicted mean change from baseline in MWDI scores at Week 16 were 1.65 (90% CI: 1.26, 2.06) and 1.84 (90% CI: 1.26, 2.32) for 40 mg and 90 mg BID compared to 0.67 (90% CI: 0.40, 0.96) for placebo. For the once daily regimens the mean change from baseline at week 16 is 1.15 (90% CI: 0.80, 1.58), 1.58 (90% CI: 1.28, 1.94) and 1.69 (90% CI: 1.32, 2.13) for 20 mg, 90 mg, and 180 mg QD. Simulations focused on BID regimens due to the marginally higher itch response and greater response with biomarkers of target engagement (data not shown) with BID regimens compared to QD. On average 50% (90% CI: 42%, 60%) of subjects are predicted to have a 2-point change from baseline in MWDI score with 90 mg BID, compared to 41% (90% CI: 33%, 49%) on 40 mg BID. Linerixibat also displays a dose-dependent increase in diarrhea, consistent with increased bile acid excretion. The rate of permanent discontinuation of study drug due to diarrhea was 4% and 14% for the 40 mg BID and 90 mg BID treatment arms, respectively Therefore, when factoring in the rate of permanent discontinuation for each treatment arm in the simulations, the average response rate at the higher dose of 90 mg BID (43%; 90% CI: 36%, 52%) is similar to the 40 mg BID (40%; 90% CI: 32%, 48%) treatment group.
Conclusions: A k-PD model of WDI scores accurately predicted placebo and linerixibat effects on itch for once and twice daily regimens over the course of 16 weeks. Whilst a higher dose of 90 mg BID predicts marginally higher clinically meaningful response, when treatment discontinuation due to diarrhea was incorporated, the percentage of responders was similar to the lower 40 mg BID regimen.
Disclosures:
This study was funded by GSK (201000, NCT02966834).
RM, JC, DA, and BS are employees of GSK and hold GSK shares.
F.C. is funded through a UNC/GlaxoSmithKline Pharmacokinetics/Pharmacodynamics Post-Doctoral Fellowship.
References:
[1] Lindor KD et al. Hepatology (2019) 69, 394-419.
[2] Hegade VS et al. Lancet (2017) 389, 1114-1123.
[3] Jacquimin P et al. J. Pharmacokinet Pharmacodyn (2007) 34, 57-85.
[4] https://clinicaltrials.gov/ct2/show/NCT02966834.
[5] Saito T et al. J Clin Pharmacol. (2017) 57, 1564-1572.
Reference: PAGE 29 (2021) Abstr 9706 [www.page-meeting.org/?abstract=9706]
Poster: Drug/Disease Modelling - Other Topics