Article type
Year
Abstract
Background:
In randomized clinical trials (RCTs) exclusion of participants post-randomization from the final analysis population frequently happens. This leads to deviations from the intention-to-treat (ITT) principle, potentially leading to attrition bias. Modified ITT (mITT) is commonly used as an alternative to ITT; mITT is vaguely defined, but most often excludes patients who failed to receive treatment or patients without post-baseline assessment.
Objective:
To investigate potential bias of performing mITT compared to ITT.
Methods:
A systematic search in MEDLINE, EMBASE, and CENTRAL was performed. Inclusion criteria: blinded RCTs investigating approved biological agents for rheumatoid arthritis (RA); using the American College of Rheumatology 20 response (ACR20) as an outcome. Exclusion criteria: non-inferiority trials, as per protocol is the preferred analysis method. A meta-epidemiological random-effects analysis was used to calculate the odds ratio (OR) of analysis population (ITT and mITT). Statistical analyses were based on the GLIMMIX procedure in SAS.
Results:
Among 5,237 references, 72 RCTs were included and analysed (incl. 23,842 patients). Figure 1 illustrates the OR for each analysis population reflecting the pooled estimate of treatment effect; all in favour of the intervention. There was no statistically significant difference in treatment effect when applying mITT compared to ITT. Figure 2 illustrates the treatment effect corresponding to analysis population, when mITT is subdivided based on the number of modifications. The treatment effect increases gradually from ITT(0) to mITT(1) to mITT(2) and decreases again at mITT(3). mITT trials did not differ statistical significant from ITT trials, regardless of the number of modifications applied (4).
Conclusion: Both ITT and mITT analyses can be applied; however, more than one post-randomization modification may threaten a conservative estimate.
This work was supported by grants from the Oak and the Michaelsen foundation
In randomized clinical trials (RCTs) exclusion of participants post-randomization from the final analysis population frequently happens. This leads to deviations from the intention-to-treat (ITT) principle, potentially leading to attrition bias. Modified ITT (mITT) is commonly used as an alternative to ITT; mITT is vaguely defined, but most often excludes patients who failed to receive treatment or patients without post-baseline assessment.
Objective:
To investigate potential bias of performing mITT compared to ITT.
Methods:
A systematic search in MEDLINE, EMBASE, and CENTRAL was performed. Inclusion criteria: blinded RCTs investigating approved biological agents for rheumatoid arthritis (RA); using the American College of Rheumatology 20 response (ACR20) as an outcome. Exclusion criteria: non-inferiority trials, as per protocol is the preferred analysis method. A meta-epidemiological random-effects analysis was used to calculate the odds ratio (OR) of analysis population (ITT and mITT). Statistical analyses were based on the GLIMMIX procedure in SAS.
Results:
Among 5,237 references, 72 RCTs were included and analysed (incl. 23,842 patients). Figure 1 illustrates the OR for each analysis population reflecting the pooled estimate of treatment effect; all in favour of the intervention. There was no statistically significant difference in treatment effect when applying mITT compared to ITT. Figure 2 illustrates the treatment effect corresponding to analysis population, when mITT is subdivided based on the number of modifications. The treatment effect increases gradually from ITT(0) to mITT(1) to mITT(2) and decreases again at mITT(3). mITT trials did not differ statistical significant from ITT trials, regardless of the number of modifications applied (4).
Conclusion: Both ITT and mITT analyses can be applied; however, more than one post-randomization modification may threaten a conservative estimate.
This work was supported by grants from the Oak and the Michaelsen foundation