Background: The main strength of randomised controlled trials (RCTs) is that each group is balanced in all prognostic characteristics. The intention to treat (ITT) principle helps to preserve such balance, and it should be the main approach for analysis of RCTs. Trials using modified intention to treat (mITT) with arbitrary definitions and with substantial number of exclusions are increasingly appearing in medical literature. Objectives: The aim of this report is to search for any association between estimates of treatment effects among trials with respect to the approach used for analysis (ITT versus mITT). Methods: We performed a computerised search to retrieve articles of interest by identifying journals that were more likely to provide trials that used the mITT approach. Three major medical journals (JAMA, Lancet, NEJM) and three specialised journals (Antimicrobial Agents and Chemotherapy, American Heart Journal, and Journal of Clinical Oncology) were identified. All trials published in the 2006 were considered. To retrieve sufficient trials reporting mITT, the research was extended to other years. Data relating to reporting of an ITT approach, to allocation concealment and to blinding were extracted. Ratio of Odds Ratios (ROR) was calculated. A ROR less than 1 implies that mITT trials exaggerate intervention effect estimates. Phase I and II trials, equivalence and noninferiority studies and cluster RCTs were excluded. Results: Of 601 records, 156 trials reported an ITT approach, 67 trials reported modified ITT, and 56 studies did not cite any intention to treat. With respect to trials using ITT, trials reporting mITT were more likely to exclude patients from analysis reaching a mean proportion of exclusions of 14% (SD 24.8). Though adequate allocation concealment was similar in the groups reporting ITT and mITT analysis, compared to trials reporting ITT, mITT trials significantly exaggerated the effect estimate, on average by 15% [ROR 0.85 (0.81 to 0.88)]. The magnitude of the effect persisted even when considering only trials with adequate allocation concealment. Conclusions: This report provides evidence that the mITT approach in RCTs is associated with bias. Trial investigators must adhere to the true ITT analysis.