Background: Modified intention-to-treat (mITT) reporting in randomized controlled trials (RCTs) is characterized by varying and, sometimes, multiple types of deviation from intention-to-treat (ITT)[1]. However, it is unknown by how much the missing data affects the post-randomisation exclusions. Moreover, it is unknown whether all themissing data cases are considered in the post-randomization exclusions. Methods: 475 RCTs from a previous systematic search. From each trial mITT descriptions were retrieved and identified possible cases of missing data (e.g. descriptions covering lost to follow up, absence of post-baseline assessment, missing outcome) and possible protocol deviation (e.g. patients that did not take the intended treatment). Classification according to the type and number of mITT deviation was performed. Also, data on lost-to-follow up, missing data and withdrawals were recorded when not considered in the post-randomization exclusions. Multinomial regression model was used and odds ratios (OR) with confidence intervals (CI) were calculated. Results: The higher the number of types of mITT deviation, the higher the odds of the presence of a potential missing data in the mITT description (OR of trials with 1 type of mITT deviation 4.04 (CI, 2.64 6.22); OR of trials with more than 1 type of mITT deviation 9.47 (CI, 4.95 18.12)) . Of 18 trials with mITT descriptions reporting lost-to-follow up, 4 trials reported further lost-to-follow up not considered in the post-randomization exclusions. Of 144 trials that had mITT covering missing data descriptions, 57 had further missing data not considered in the post-randomization exclusions. No trials that reported patients withdrawals in the mITT description, had further withdrawals reported. Conclusion: Missing data appears to be an essential component in the mITT reporting. The mITT description is not a consistent and reliable method of reporting.
Reference: 1. Abraha and Montedori, Modified Intention to Treat Reporting in Randomised Controlled Trials: A Systematic Review. BMJ, 2010 in press.