P. Girard, F. Varret
Service de Pharmacologie Clinique, Equipe Associée 643, Université Cl. Bernard, Lyon France.
Drug taking behavior is traditionally decomposed in 2 dimensions: accuracy (taking at scheduled times) and compliance (taking at the prescribed dose). Third dimension should be added when patients terminate their treatment prematurely for any reason. These drop-outs may not be independent from compliance: for example, under or over-compliant patients are more likely to experience inefficacy or severe side-effects, respectively, and so stop their treatments. We simultaneously model the 2 last components using a Markov model including fixed and random effects for compliance [1] and a parametric survival analysis where times to drop-out are assumed to follow a Weibull distribution. The model was applied to 66 depressed patients randomized between treatment A or B once daily. Compliance was assessed over 9 weeks using the MEMSä system. Maximum likelihood parameter estimates were obtained using NONMEM software. Probability of not-taking the treatment at day j was found to be highly variable between patients, influenced by not-taking at days j-1 and j-2 (Markov order 2), and by gender (better compliance for F than M). Probability of drop-out was 25% more for treatment A. Using the model for simulation purposes permit to reproduce drug holiday patterns as well as treatment duration in presence of drop-outs.
[1] Girard, P., Blaschke, T.F., Kastrissios, H., and Sheiner, L.B. A Markov mixed effect regression model for drug compliance. Stat.Med. 17(20):2313-2334, 1998.
Reference: PAGE 8 (1999) Abstr 153 [www.page-meeting.org/?abstract=153]
Poster: poster