Up To it!   Home > Italian biomedical research highlights

Highlights of Italian biomedical research: recent publications

New  Increasing use of "modified intention to treat" analysis: a worrisome trend of poor quality reporting of RCTs

When a randomized clinical trial is conducted, it is hoped that all patients receive the treatment to which they were initially assigned and that all planned measurements are actually made. In this ideal situation, the benefits of randomization are preserved and bias is not introduced by manipulations to make up for missing data. However, in real-life situations, the planned protocol may be violated (for example, when patients do not receive the allocated treatment) or data may be missing (if researchers fail to perform all tests or patients withdraw from the study). When patients' baseline and outcome data are assessed within the context of the group to which they were originally assigned, irrespective of the treatment they actually received, the analysis is said to be done according to the "intention to treat" (ITT) principle. This way of evaluating a study avoids introducing bias from the exclusion of patients after randomization. In the case that data are missing, it is impossible to conduct a strict ITT analysis and, when the problem is severe, the RCT is compromised.

A clear summary of the ITT principle is given in the latest version of the CONSORT statement. Considering the difficulties in achieving the conditions necessary for a strict ITT analysis, the statement now recommends that authors specify how patients were included in the analysis and why any were excluded, rather than resorting to equivocal expressions such as "modified intention to treat". In fact, this term is seen in some published reports of RCTs, making these papers difficult to evaluate and raising doubts as to their quality. To determine the extent to which this ambiguous analytical approach is used—and to understand how it is defined by authors—Italian researchers Abraha and Montedori did a systematic review of RCTs in which this term was reported.

The study, recently published in BMJ (PMID: 20547685), evaluated 475 RCTs in which a modified ITT analysis was reported. The number of papers using this term increased linearly over the years from 1997 to 2006. Although this set of papers was not compared with a control set that used a strict ITT analysis, the study did document a high proportion (88%) of trials with industry funding as well as signs of poor reporting such as the absence of a flow chart (in 49%). The term "modified" was used to describe an analysis that had one or more different types of deviations from a true ITT, with patients having been eliminated because of changes in treatment, improper inclusion, or missing baseline or outcome data. These variations indicate that a single modified ITT approach does not exist, confirming the term's ambiguous nature. As the authors pointed out, use of the modified ITT analysis is "erratic" and "unpredictable", and is evidence of a growing "shortcoming in the quality of reporting". Readers of the clinical literature should be alert to the possibility that studies that used this approach may suffer from bias and may be selectively reported. Journal editors should insist that accepted papers adhere to the most recent version of the CONSORT statement. Posted 30 July 2010.

© UpTo infotechnologies. All rights reserved.