Assessment of Phase 1 pharmacokinetic-pharmacodynamic predictions, based on Phase 3 clinical trials results.Depemokimab pharmacokinetics and response in blood eosinophil count reduction for asthma and chronic rhinosinusitis with nasal polyps patients
Lénaïg Tanneau1, Alexandra Lavalley-Morelle1, Anders Thorsted2, Anubha Gupta2, Stein Schalkwijk2, Jakob Ribbing1
1Pharmetheus, 2GlaxoSmithKline
Introduction: Depemokimab is the first ultra-long-acting biologic to have enhanced binding affinity, high potency, and an extended half-life. As a targeted ultra-long-acting anti-IL-5 (unlike other anti-IL-5 as mepolizumab dosed every 4 weeks), depemokimab has the potential to provide sustained inhibition of Type 2 inflammation, enabling dosing every 6 months. In previous analyses [1], model development and predictions of the depemokimab pharmacokinetic-pharmacodynamic (PKPD) relations were based on a first time in human (FTIH) single ascending dose study in adult patients with mild-moderate asthma (called Ph1 model in the following). This analysis integrated prior knowledge from mepolizumab, for which a PKPD model was established using data in multiple eosinophilic conditions, including asthma and chronic rhinosinusitis with nasal polyps (CRSwNP), and thereby allowed for predictions of the depemokimab PKPD in these patient populations, for which depemokimab was not yet investigated. Recently, depemokimab phase 3 studies in (severe) asthma and CRSwNP have been completed [2,3]. The PK and its relationship with blood eosinophil count (BEC) reduction have been analysed with pooled depemokimab data from three phase 1 studies (two in healthy volunteers [PK only] and the above FTIH) and four phase 3 studies (two studies in patients with severe asthma and two studies in patients with CRSwNP), called Ph3 model in the following. Objectives: To compare the predictions from the Phase 1 model to the Phase 3 results. Methods: The PKPD relationship between PK and BEC consisted of an indirect-response model, linked to individual predicted depemokimab concentration, with an inhibitory sigmoidal Emax model, supplemented by a placebo linear effect. Predictions of the exposure-response (ER) curves as well as model predictions of Week 52 a) absolute BEC and b) ratio of BEC to baseline change from placebo (RTBw52) were evaluated. First, a priori predictions used the Ph1 model, adjusted by patient characteristics from mepolizumab trials. Subsequently, Ph1 covariate-adjusted predictions were generated using the same Ph1 model but adjusted with patient characteristics (median covariate values) from the Phase 3 trials. Lastly, the Ph3 estimation represented predictions from the Ph3 model estimated on data including the Phase 3 trials and were compared to the values observed in these trials. The models were originally developed with NONMEM, and the predictions were made with mrgsolve and R. Results: Overall, the predictions that were made with the Ph1 model agreed with the Ph3 results. The ER curves were well predicted for patients with CRSwNP, and small changes in estimates were seen for the Ph3 asthma patients (higher EC50 (from 0.151 to 0.194 µg/mL) and lower Emax (from 0.892 to 0.848) than initially predicted with the Ph1 model). The differences in EC50 values between asthma indications in the mepolizumab trials were confirmed in the depemokimab trials, where FTIH patients had a 56% lower EC50 than patients belonging to Phase 3 studies (Ph1 prediction was 46% lower EC50). Moreover, the Hill factor was higher (steeper sigmoidal curve) in the Ph1 model (2.35) compared to the Ph3 model (1.64). Regarding the model predictions, a priori and Ph1 covariate-adjusted predictions showed good agreement due to the similarities in patients’ characteristics across the mepolizumab and depemokimab trials. However, when compared to the Ph3 estimation (as well as the actual Ph3 outcomes), the Ph1 model slightly underpredicted the response in patients with CRSwNP (median RTBw52 of 0.178 with the Ph1 covariate-adjusted predictions versus 0.158 estimated with the Ph3 model) and slightly overpredicted the response in patients with asthma (median RTBw52 of 0.143 with the Ph3 covariate-adjusted predictions versus 0.181 estimated with the Ph3 model). Potential causes for these discrepancies include higher than predicted trough concentrations and potentially other differences between the depemokimab Phase 1 and Phase 3 trials, including differences in population, study conduct, or bioanalytical assay. Conclusions: This analysis demonstrates that the use of model-informed drug development principles can be applied to generate prospective predictions that support dose selection. The Phase 1 model provided accurate predictions for the target absolute BEC (<100 cells/µL) at Week 52 and RTBw52 (<0.2, i.e. 80% reduction) in both asthma and CRSwNP patient populations.
[1] Barceló et al. Use of informative prior distributions from mepolizumab data to support depemokimab PKPD model analysis, Population Approach Group in Europe 31, 2023. [2] Jackson DJ, Wechsler ME, Jackson DJ, et al. Twice-Yearly Depemokimab in Severe Asthma with an Eosinophilic Phenotype. N Engl J Med. 2024;391(24):2337-2349. doi:10.1056/NEJMoa2406673 [3] Gevaert P, Desrosiers M, Cornet M, et al. Efficacy and safety of twice per year depemokimab in chronic rhinosinusitis with nasal polyps (ANCHOR-1 and ANCHOR-2): phase 3, randomised, double-blind, parallel trials. The Lancet, 2025.doi:10.1016/ S0140-6736(25)00197-7