Article type
Year
Abstract
Background: The majority of systematic reviews focus on pair-wise, direct comparisons of treatments from randomised controlled trials (RCTs). However, this makes it difficult to determine which is the 'best' treatment when there are multiple treatment options. A mixed treatment comparison (MTC) provides a single combined analysis of all RCTs involving the treatments of interest.
Objectives: To illustrate the advantages of MTC analysis through an example of treatment post acute myocardial infarction, and to determine the extent to which MTC reduces uncertainty in relative treatment effects.
Method: Trials from two previous systematic reviews, one of 6 thrombolytic treatments and another on primary percutaneous transluminal coronary angioplasty compared to thrombolysis, were combined in a fixed effect MTC. For each of the 21 pairwise comparisons we computed a precision, P1, of the log odds ratio. A fixed effect 'head-to-head' standard meta-analysis was also run for each pairwise comparison and a precision, P0, computed. The difference in precision (P1 - P0) represents the additional contribution of MTC over standard meta-analysis in reducing uncertainty. We calculate the sample size required for a new RCT to achieve an equivalent reduction in uncertainty.
Results: Direct data exist on 10 of the potential 21 pairwise comparisons (4 are individual trials), with 136,345 patients. The MTC analysis provides a full set of odds ratios for all 21 comparisons. The results from the standard pairwise meta-analysis (where available) are consistent with the MTC analysis. The MTC analysis produces narrower confidence intervals. Three comparisons based on standard pairwise meta-analysis were statistically significant at the 5% level, compared to 11 of the MTC comparisons. To achieve an equivalent reduction of uncertainty across the 21 pairwise comparisons would require a total of 89,192 patients in additional RCTs.
Conclusions: Multiple treatment options are pervasive in health care. MTC analyses enable the ranking of treatments, reduction of uncertainty and an assessment of the need for further research. There is growing interest in developing 'umbrella reviews', bringing together all the Cochrane reviews on a specific condition. MTC analyses could contribute significantly to such an enterprise.
Objectives: To illustrate the advantages of MTC analysis through an example of treatment post acute myocardial infarction, and to determine the extent to which MTC reduces uncertainty in relative treatment effects.
Method: Trials from two previous systematic reviews, one of 6 thrombolytic treatments and another on primary percutaneous transluminal coronary angioplasty compared to thrombolysis, were combined in a fixed effect MTC. For each of the 21 pairwise comparisons we computed a precision, P1, of the log odds ratio. A fixed effect 'head-to-head' standard meta-analysis was also run for each pairwise comparison and a precision, P0, computed. The difference in precision (P1 - P0) represents the additional contribution of MTC over standard meta-analysis in reducing uncertainty. We calculate the sample size required for a new RCT to achieve an equivalent reduction in uncertainty.
Results: Direct data exist on 10 of the potential 21 pairwise comparisons (4 are individual trials), with 136,345 patients. The MTC analysis provides a full set of odds ratios for all 21 comparisons. The results from the standard pairwise meta-analysis (where available) are consistent with the MTC analysis. The MTC analysis produces narrower confidence intervals. Three comparisons based on standard pairwise meta-analysis were statistically significant at the 5% level, compared to 11 of the MTC comparisons. To achieve an equivalent reduction of uncertainty across the 21 pairwise comparisons would require a total of 89,192 patients in additional RCTs.
Conclusions: Multiple treatment options are pervasive in health care. MTC analyses enable the ranking of treatments, reduction of uncertainty and an assessment of the need for further research. There is growing interest in developing 'umbrella reviews', bringing together all the Cochrane reviews on a specific condition. MTC analyses could contribute significantly to such an enterprise.
PDF