Title:Use of Recursive Partitioning Analysis in Clinical Trials and Meta-Analysis of Randomized Clinical Trials, 1990-2016
Volume: 12
Issue: 1
Author(s): Martha Maria Fors, Carmen Elena Viada and Paloma Gonzalez
Affiliation:
Keywords:
Clinical trials, decision trees, meta-analyses, prediction, recursive partitioning analysis (RPA), recursive partitioning
method.
Abstract: 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.
