Simon Eckermann1, Michael Coory, Andrew R Willan. 1. Flinders University, Flinders Centre for Clinical Change & Health Care Research, Adelaide, South Australia, Australia. simon.eckermann@flinders.edu.au
Abstract
OBJECTIVE: When undertaking indirect comparisons, relative risk (RR) is often suggested as an appropriate indicator of treatment effect, particularly where baseline (common comparator) risks differ. In this article, we demonstrate that such use of RR in indirect comparisons is not necessarily stable with respect to framing of outcomes. STUDY DESIGN AND SETTING: Use of RR is shown to lead to inferential fallacies where, for example, a new therapy is suggested to reduce both mortality and survival risk. Conditions under which the inferential fallacy arises and an odds solution are illustrated in indirect comparison of natalizumab and interferon beta-1b for multiple sclerosis. RESULTS: Using RR, natiluzimab is suggested to be 30% more effective than interferon for progression (RR=0.70), but 16% less effective than interferon for no progression (RR=0.84). This inferential anomaly is avoided using odds ratios (ORs), with odds of progression (0.83) the reciprocal of that for no progression (1.21). CONCLUSION: Inferential fallacies with use of RR in indirect comparison provide scope for abuse with respect to choice in framing of outcomes, and confound decision making where both results are presented. The use of ORs overcomes this inferential fallacy, consistently informing inference with respect to direction of treatment effect in indirect comparisons.
OBJECTIVE: When undertaking indirect comparisons, relative risk (RR) is often suggested as an appropriate indicator of treatment effect, particularly where baseline (common comparator) risks differ. In this article, we demonstrate that such use of RR in indirect comparisons is not necessarily stable with respect to framing of outcomes. STUDY DESIGN AND SETTING: Use of RR is shown to lead to inferential fallacies where, for example, a new therapy is suggested to reduce both mortality and survival risk. Conditions under which the inferential fallacy arises and an odds solution are illustrated in indirect comparison of natalizumab and interferon beta-1b for multiple sclerosis. RESULTS: Using RR, natiluzimab is suggested to be 30% more effective than interferon for progression (RR=0.70), but 16% less effective than interferon for no progression (RR=0.84). This inferential anomaly is avoided using odds ratios (ORs), with odds of progression (0.83) the reciprocal of that for no progression (1.21). CONCLUSION: Inferential fallacies with use of RR in indirect comparison provide scope for abuse with respect to choice in framing of outcomes, and confound decision making where both results are presented. The use of ORs overcomes this inferential fallacy, consistently informing inference with respect to direction of treatment effect in indirect comparisons.
Authors: M Tadrous; L Wong; M M Mamdani; D N Juurlink; M D Krahn; L E Lévesque; S M Cadarette Journal: Osteoporos Int Date: 2013-11-28 Impact factor: 4.507