Main Article Content
Intention-to-treat analysis in the chronic suppurative otitis media trials
Abstract
Objectives: To determine the extent of application of the intention to treat principle in the chronic suppurative otitis media (CSOM) randomised controlled trials.
Design: Data were extracted from 28 CSOM randomised controlled trials.
Main outcome measures: Conceptual and methodological approaches of dealing with protocol deviations with respect to withdrawal, missing response and non-compliance.
Results: Of the 28 CSOM trials included in this study, only one (4%) trial mentioned intention-to-treat (ITT) analysis. However, 10(36%) other trials which did not mention ITT, had no protocol deviations and thus carried out an ITT analysis by default. It is highly likely that a biased treatment effect existed in the trial that mentioned ITT since the authors undertook a complete case analysis disregarding the 22% protocol deviators. There were no attempts in any of the trials to impute for missing responses and carrying out a sensitivity analysis. For trials with a big percentage of protocol deviations, the validity of their results are brought to question.
Conclusions: In practice, not all those entered into a randomised-controlled trial will complete the trial. Thus intention-to-treat analysis is an important aspect of randomized controlled trials of health care interventions which tries to bridge this gap. It is important for authors to explicitly state the protocol deviations, the methods used to handle them and the potential effect with reference to bias and study outcome.
(East African Medical Journal: 2003 80(6): 308-311)
Design: Data were extracted from 28 CSOM randomised controlled trials.
Main outcome measures: Conceptual and methodological approaches of dealing with protocol deviations with respect to withdrawal, missing response and non-compliance.
Results: Of the 28 CSOM trials included in this study, only one (4%) trial mentioned intention-to-treat (ITT) analysis. However, 10(36%) other trials which did not mention ITT, had no protocol deviations and thus carried out an ITT analysis by default. It is highly likely that a biased treatment effect existed in the trial that mentioned ITT since the authors undertook a complete case analysis disregarding the 22% protocol deviators. There were no attempts in any of the trials to impute for missing responses and carrying out a sensitivity analysis. For trials with a big percentage of protocol deviations, the validity of their results are brought to question.
Conclusions: In practice, not all those entered into a randomised-controlled trial will complete the trial. Thus intention-to-treat analysis is an important aspect of randomized controlled trials of health care interventions which tries to bridge this gap. It is important for authors to explicitly state the protocol deviations, the methods used to handle them and the potential effect with reference to bias and study outcome.
(East African Medical Journal: 2003 80(6): 308-311)
