Epidemiology of the type of intention-to-treat reporting in meta-analyses

Article type
Authors
Montedori A1, Bonacini M2, Luchetta M3, Papini D4, Rocchi R1, Orso M1, Cozzolino F1, Abraha I1
1Regional Health Authority of Umbria, Italy
2Derriford Hospital, Plymouth, UK
3Azienda Sanitaria Locale3, Foligno, Italy
4Regional Agency for Health of Emilia Romagna, Italy
Abstract
Background: Despite the intention-to-treat (ITT) principle is the recommended method to perform unbiased analyses in randomised trials, it is inadequately applied and participants are often improperly excluded. Recently, a modified intention-to-treat (mITT) analysis with inconsistent descriptions is increasingly being used in randomised trials. The reporting of mITT in trials is significantly associated with post-randomisation exclusions. However, little is known about the impact of mITT reporting trials in trials included in meta-analyses.

Objectives: To estimate the prevalence of the type of ITT in meta-analyses and to assess the description of mITT reporting; and to determine differences in terms of reporting, methodological quality and effect size among the types of ITT.

Methods: We searched MEDLINE for systematic reviews indexed from 2005 to 2009. Montori’s search strategy was used (MEDLINE[Title/Abstract]) OR (systematic[Title/Abstract] AND review[Title/Abstract] OR meta- analysis[Publication Type]). From each year we randomly selected 10% of the records. Only therapeutic meta-analyses that included at least one randomised trial were considered. Diagnostic, prognostic, or epidemiological studies were excluded. Within each meta-analysis trials were classified according to the type of ITT analyses used as follows: ITT, trials reporting the use of standard ITT analyses; mITT, trials reporting the use of 'modified intention-to-treat’ analyses; or 'no ITT’ trials not reporting the use of any ITT analyses. Trials reporting the use of ITT with descriptions or conditions different from the standard ITT definition were classified as mITT. Items for quality measures such as allocation concealment, blinding, and early stopping were considered.

Results: We identified 200 meta-analyses out of 569 records published in 2005, 210 of 657 published in 2006, 190 of 804 published in 2007, 197 of 863 published in 2008, and 186 of 988 published in 2009.

Conclusions: Definitive results will be presented at the Colloquium.