Statistical Methods for Dealing with Selective Crossover in Randomised Controlled Trials

Balduzzi, Sara (2016) Statistical Methods for Dealing with Selective Crossover in Randomised Controlled Trials, [Dissertation thesis], Alma Mater Studiorum Università di Bologna. Dottorato di ricerca in Metodologia statistica per la ricerca scientifica, 27 Ciclo. DOI 10.6092/unibo/amsdottorato/7252.
Documenti full-text disponibili:
Documento PDF (English) - Richiede un lettore di PDF come Xpdf o Adobe Acrobat Reader
Download (972kB) | Anteprima


Background Selective crossover (SCO) occurs when patients in a randomised controlled trial (RCT) are given the opportunity to switch from the arm in which they were initially allocated to the other arm. Problems in data analysis and interpretation may arise. Objectives Main objectives were to assess: (i) the prevalence of SCO in RCTs assessing the efficacy of therapies for breast cancer (BC); (ii) whether different statistical methods provide different results, in particular when the outcome of interest is a time-to-event outcome. Methods RCTs assessing the efficacy of therapies for BC were searched in the medical literature. Different methods for dealing with SCO exist. Among naïve methods, the Intention-To-Treat analysis (ITT), the censored analysis and the analysis considering the treatment as a time-varying covariate were contemplated; among more complex methods, the inverse probability of censoring weighting analysis, the Loeys and Goetghebeur estimator, and the rank preserving structural failure time models were considered. The methods were evaluated trough simulations, considering twelve scenarios which differed by the proportion of patients crossing over, their underlying prognosis, and the magnitude of true treatment effect. Results Crossover occurred in the 23.5% of RCTs identified. Simulations highlighted that ITT analysis, always presented, tend to underestimate the actual treatment effect across all scenarios. Other naïve methods are able to better estimate the treatment effect, but when the probability of crossover is assumed to depend on prognosis (i.e. patients with a poor prognosis cross over more frequently than patients with a good prognosis), they failed. More complex models work better, but each of them make assumptions not always verifiable or likely to occur in the considered context. Conclusions The present work aims to highlight the need for attention being paid on the phenomenon of SCO in RCTs, and underlines the concerns related to it.

Tipologia del documento
Tesi di dottorato
Balduzzi, Sara
Dottorato di ricerca
Scuola di dottorato
Scienze economiche e statistiche
Settore disciplinare
Settore concorsuale
Parole chiave
Selective crossover Switch Intention-to-treat Inverse probability of censoring weighting Rank preserving structural failure time models
Data di discussione
4 Marzo 2016

Altri metadati

Statistica sui download

Gestione del documento: Visualizza la tesi