Integration of response, tolerability and dropout in flexible-dose trials: a case study in depression
A. Russu(1), E. Marostica(1), G. De Nicolao(1), A.C. Hooker(2), I. Poggesi(3,*), R. Gomeni(4), S. Zamuner(5)
(1) Department of Computer Engineering and Systems Science, University of Pavia, Pavia, Italy; (2) Department of Pharmaceutical Biosciences, Uppsala University, Uppsala, Sweden; (3) Clinical Pharmacology / Modelling & Simulation, GlaxoSmithKline, Verona, Italy; (4) Pharmacometrics, GlaxoSmithKline, King of Prussia, PA, USA; (5) Clinical Pharmacology / Modelling & Simulation, GlaxoSmithKline, Stockley Park, UK; * Current address: Advanced Modeling&Simulation, Janssen Pharmaceutical Companies of Johnson & Johnson, Milan, Italy
Objectives: The difficulties arising when analyzing depression trials are manifold, as a comprehensive model, in addition to the efficacy endpoints, should account for: (i) flexible dosing schemes, (ii) dropout events, and (iii) drug-related adverse effects. Simplified modelling approaches that neglect some of the above aspects may yield biased results. In this work we investigate an integrated approach based on the joint population modelling of response, tolerability and dropout. The proposed methodology is used to analyse data from a flexible-dose, placebo-controlled, Phase II depression trial. As an extension of previous work [1,2], in this study we account for flexible dosage regimen and adverse events as covariates in the dropout model.
Methods: The time course of the HAMD score was described as the sum of a Weibull and a linear function [3]. The dose escalation was included in the model as a covariate on two of the four structural parameters. We investigated three different dropout mechanisms: missing completely at random (MCAR), at random (MAR) and not at random (MNAR) [4]. The dropout probability was modulated using three covariates: the time course of the clinical outcome, dose escalation, and the occurrence of clinically relevant adverse events in the drug arm. The population model was implemented in WinBUGS 1.4.3 [5].
Results: With respect to previous approaches [1,2], which used only the HAMD score as a covariate in the hazard model, the proposed method achieved comparable goodness-of-fit to HAMD data. However, the inclusion of dose escalation and drug-related adverse events in the hazard function yielded a substantial benefit in the description of the dropout process, as witnessed by the Deviance Information Criterion [6], parameter estimates, and the modified Cox-Snell residuals [4]. Comparison of the dropout mechanisms suggested a MNAR dropout process in both treatment arms. The ability of the proposed model to reproduce realistic dropout patterns was assessed via Kaplan-Meier visual predictive checks [7].
Conclusions: Our results show the feasibility of a joint model accounting for the HAMD time course, discontinuities in the dosing schedule, dropouts and adverse events. Indeed, in the study here analyzed, the dropout process was influenced by all the above aspects. Thorough modelling approaches that integrate all the relevant information are necessary to provide a more comprehensive assessment of antidepressant drug response.
References:
[1] Russu A, Marostica E, De Nicolao G, Hooker AC, Poggesi I, Gomeni R, Zamuner S (2010), Integrated model for clinical response and dropout in depression trials: a state-space approach, Population Approach Group Europe (PAGE) 19th Meeting, Abstract 1852
[2] Hooker C, Gomeni R, Zamuner S (2009), Time to event modeling of dropout events in clinical trials, Population Approach Group Europe (PAGE) 18th Meeting, Abstract 1656
[3] Gomeni R, Lavergne A, Merlo-Pich E (2009), Modelling placebo response in depression trials using a longitudinal model with informative dropout, European Journal of Pharmaceutical Sciences 36, pp. 4-10
[4] Hu C, Sale ME (2003), A joint model for longitudinal data with informative dropout, Journal of Pharmacokinetics and Pharmacodynamics 30, pp. 83-103
[5] Lunn DJ, Thomas A, Best N, Spiegelhalter D (2000), WinBUGS - A Bayesian modelling framework: concepts, structure and extensibility. Statistics and Computing 10, pp. 325-337
[6] Spiegelhalter DJ, Best NG, Carlin BP, van der Linde A (2002), Bayesian measures of model complexity and fit (with discussion), Journal of the Royal Statistics Society 64, pp. 583-639
[7] Holford N (2005), The visual predictive check: superiority to standard diagnostic (Rorschach) plots, Population Approach Group Europe (PAGE) 14th Meeting, Abstract 738
