Truncated recurrent event survival models for methadone data.

Truncated multivariate survival models are proposed for the analysis of data from the South Australian Methadone Program. Injecting drug users may have several entries to and exists from the program. A marginal approach for recurrent events is found to provide a useful and simplified basis for modeling the retention of injecting drug users on methadone. The likelihoods and results of fitting two forms of truncated survival models are presented, together with recommendations for statistical analysis using S-PLUS. The relative merits of the marginal approach in this context and alternative ways of viewing the data are discussed.