Joint model for dropout in longitudinal trials in COPD patients
M.A. Boroujerdi (1), A. Facius (2), M. Danhof (1), O.E. Della Pasqua (1,3)
(1) Division of Pharmacology, Leiden/Amsterdam Center for Drug Research, Leiden, The Netherlands; (2) Dept. of Pharmacometrics, Nycomed GmbH, Konstanz, Germany; (3) Clinical Pharmacology Modelling & Simulation, GlaxoSmithKline, United Kingdom.
Objectives: Long term longitudinal studies often have large dropout rates. The dropout is termed informative if it is related to the primary variable used for the assessment of efficacy in clinical trials. Forced exhaled volume in one second (FEV1) is commonly used for the assessment of disease severity and progression in chronic obstructive pulmonary disease (COPD). The objective of this study was to assess the joint model methodology for the simulation of individual hazard of dropping out in clinical trials.
Results: Data from 1356 COPD patients from a 52-week study treated with placebo were evaluated. A two level mixed model was used to describe the FEV1 over time. The time to dropout was modelled with Cox regression with basal FEV1 as covariate. The Cox model provides a mean estimate for changing basal hazard with time. The random components of two-level model for FEV1 indicating individual differences in initial FEV1 (intercept) and the rate of change (slope) were used as driver for the proportional hazard model. The risk of dropping out was computed as the accumulation of hazard with the survival being the exponentiated risk. Simulations were then performed to describe predicted dropout rate. The mean simulation of survivals closely resembles the Kaplan-Meier survival probability estimates. An application of this parametric approach is shown for the purposes of clinical trial simulation.
Conclusions: The Cox proportional hazard model with the addition of random effects to account for between-subject variability provides the basis for the simulation of survival in placebo treated COPD patients. Our simulations suggest that not only differences in baseline, but also changes in FEV1 over time contribute to the hazard or survival during the trial.