Use of Recursive Partitioning Analysis in Clinical Trials and Meta-Analysis of Randomized Clinical Trials, 1990-2016

BACKGROUND: Recursive Partitioning Analysis (RPA) is a very flexible non parametric algorithm that allows classification of individuals according to certain criteria, particularly in clinical trials, the method is used to predict response to treatment or classify individuals according to prognostic factors. OBJECTIVES: In this paper we examine how often RPA is used in clinical trials and in meta-analysis. METHODS: We reviewed abstracts published between 1990 and 2016, and extracted data regarding clinical trial phase, year of publication, type of treatment, medical indication and main evaluated endpoints. RESULTS: One hundred and eighty three studies were identified; of these 43 were meta-analyses and 23 were clinical trials. Most of the studies were published between 2011 and 2016, for both clinical trials and meta-analyses of randomized clinical trials. The prediction of overall survival and progression free survival were the outcomes most evaluated, at 43.5% and 51.2% respectively. Regarding the use of RPA in clinical trials, the brain was the most common site studied, while for meta-analytic studies, other cancer sites were also studied. The combination of chemotherapy and radiation was seen frequently in clinical trials. CONCLUSION: Recursive partitioning analysis is a very easy technique to use, and it could be a very powerful tool to predict response in different subgroups of patients, although it is not widely used in clinical trials.