Karsten Kuritz 1, Petra Jauslin 1, Nikunjkumar Patel 2, Itay Perlstein 3, Avia Merenlender-Wagner 4, Tamar Bur-Nur 4, Anna Elgart 4, Rajendra Singh 2
1 IntiQuan AG (Basel, Switzerland), 2 Teva Branded Pharmaceutical Products R&D, Inc. (West Chester, USA), 3 Magic Wand Research LLC (Philadelphia, USA), 4 Teva Pharmaceutical Industries Ltd. (Netanya, Israel)
Introduction/Objectives:
Olanzapine is a second-generation antipsychotic drug for the treatment of schizophrenia. Oral administration requires daily dosing, often limiting long-term adherence. The currently available long-acting injectable (LAI) intramuscular formulation carries a risk of post-injection delirium/sedation syndrome (PDSS), thus requiring a risk evaluation and mitigation strategy. TV-44749 is a new LAI olanzapine formulation for once-monthly subcutaneous administration, designed to eliminate PDSS risk.
The Phase 3 study TV44749-CNS-30096 (SOLARIS) was conducted to evaluate efficacy and safety of TV-44749 in the treatment of schizophrenia. Patients with acute exacerbation of schizophrenia were randomized in a 1:1:1:1 ratio to an 8-week double-blind, placebo-controlled phase (Period 1), receiving 318, 425, or 531 mg TV-44749 or placebo once monthly, followed by an open label safety period of up to 48 weeks (Period 2). Participants on placebo in Period 1 were rerandomized in a 1:1:1 ratio to one of the active treatments in Period 2, and participants on active treatment retained their Period 1 treatment allocation.
The primary endpoint was change from baseline to Week 8 in Positive and Negative Syndrome Scale (PANSS) total score, a 30‑item clinician‑rated scale widely used to assess symptom severity of schizophrenia. Since high discontinuation rates in schizophrenia trials may bias longitudinal analyses [1,2,3], a joint PANSS–dropout model was evaluated in parallel with the exposure–response (ER) analysis.
The objective of this analysis was to characterize the ER relationship of TV-44749 for PANSS total score and to assess the impact of dropout on ER modeling through a joint PANSS-dropout model.
Methods:
Data from 665 participants and 7811 PANSS score measurements from the SOLARIS study were analyzed using nonlinear mixed effects modeling in NONMEM. PANSS total scores were described using a published multiplicative drug–disease model [1], where placebo and treatment effects jointly influence PANSS total score. The placebo response was represented by a mono-exponential function approaching a maximum (Pmax).
The drug effect was described by a maximum effect (Emax) model driven by TV-44749 exposure. A time-delay component was included to allow the treatment effect to emerge gradually. Two steady-state exposure metrics – average concentration (Cavg) and trough concentration (Ctrough) – were evaluated, with Cavg providing better predictive performance. Baseline PANSS total score was included as a covariate on both Pmax and Emax.
To evaluate informative dropout, a joint PANSS-dropout model with an exponential hazard function [1] was applied, allowing current PANSS severity to increase dropout risk. Model performance was assessed using visual predictive checks for both the original PANSS model and the joint PANSS-dropout model. Finally, simulations were performed in 2000 virtual patients using the SOLARIS sampling schedule to generate predicted PANSS score trajectories and corresponding dropout events.
Results:
Active treatment led to a marked reduction in PANSS total scores relative to placebo, with comparable efficacy observed across all three dose levels or exposure quartiles calculated across the doses. Dropout modeling revealed a weak influence of PANSS on the risk of treatment discontinuation (hazard coefficient 0.012), consistent with estimates from a Cox proportional hazards model (0.013) and lower than previously reported values (0.033–0.048) [1,2,3].
Diagnostic evaluations, including visual predictive checks for both the PANSS model and the joint PANSS-dropout model, revealed no appreciable time-dependent bias, supporting limited mutual influence between dropout and PANSS score trajectories in this study.
Simulation results further confirmed that incorporating a dropout component did not notably change predicted PANSS total score trajectories or estimated parameters compared to the PANSS ER model without dropout.
Conclusion:
TV-44749 demonstrated robust efficacy compared to placebo with a flat ER relationship across studied doses, indicating that the doses resulted in exposure levels close to the plateau of the ER curve.
Given a dropout rate of 28% in Period 1 (inpatients) and 67% in Period 2 (outpatients) in the SOLARIS study (consistent with expectations in schizophrenia clinical trials), a joint PANSS-dropout model was developed to rule out potential bias in ER characterization. The analysis showed that PANSS score had only a minor influence on dropout (hazard coefficient 0.012), and incorporating dropout did not substantially change model predictions and parameter estimates. These results confirm the robustness of the developed ER model, while underscoring the importance of evaluating dropout mechanisms to ensure unbiased longitudinal modeling.
References:
[1] Reddy VP, Kozielska M, Suleiman AA, Johnson M, Vermeulen A, Liu J, de Greef R, Groothuis GM M, Danhof M, Proost JH. Pharmacokinetic-pharmacodynamic modeling of antipsychotic drugs in patients with schizophrenia, part I: the use of PANSS total score and clinical utility. Schizophrenia Research. 2013;146(1-3):144-152.
[2] Kalaria SN, Zhu H, Farchione TR, Mathis MV, Gopalakrishnan M, Uppoor R, Mehta M, Younis I. A quantitative justification of similarity in placebo response between adults and adolescents with acute exacerbation of schizophrenia in clinical trials. Clinical Pharmacology & Therapeutics. 2019;106(5):1046-1055.
[3] Wang X, Gopalakrishnan M, Rich B, Gobburu JV, Larsen F, Raoufinia A. Exposure-Response Modeling in Adults and Adolescents With Schizophrenia to Support the Extrapolation of Brexpiprazole Efficacy to Adolescents. Journal of Clinical Pharmacology. 2024;60(7):848-859. doi:10.1002/jcph.2464.
[4] Collett D. Modelling Survival Data in Medical Research. 3rd ed. Boca Raton, FL: CRC Press/Taylor & Francis; 2015.
Reference: PAGE 34 (2026) Abstr 12280 [www.page-meeting.org/?abstract=12280]
Poster: Drug/Disease Modelling - CNS