Subgroup analysis in randomized controlled trials appeared to be dependent on whether relative or absolute effect measures were used

Roderick P Venekamp; Maroeska M Rovers; Arno W Hoes; Mirjam J Knol

doi:10.1016/j.jclinepi.2013.11.003

Subgroup analysis in randomized controlled trials appeared to be dependent on whether relative or absolute effect measures were used

J Clin Epidemiol. 2014 Apr;67(4):410-5. doi: 10.1016/j.jclinepi.2013.11.003. Epub 2014 Feb 5.

Authors

Roderick P Venekamp¹, Maroeska M Rovers², Arno W Hoes³, Mirjam J Knol³

Affiliations

¹ Department of Otorhinolaryngology, University Medical Centre Utrecht, PO Box 85500, 3508 GA Utrecht, The Netherlands; Julius Center for Health Sciences and Primary Care, University Medical Centre Utrecht, PO Box 85500, 3508 GA Utrecht, The Netherlands. Electronic address: R.P.Venekamp@umcutrecht.nl.
² Julius Center for Health Sciences and Primary Care, University Medical Centre Utrecht, PO Box 85500, 3508 GA Utrecht, The Netherlands; Department of Health Evidence, Radboud University Medical Center Nijmegen, PO Box 9101, 6500 HB Nijmegen, The Netherlands; Department of Operating Rooms, Radboud University Medical Center Nijmegen, PO Box 9101, 6500 HB Nijmegen, The Netherlands.
³ Julius Center for Health Sciences and Primary Care, University Medical Centre Utrecht, PO Box 85500, 3508 GA Utrecht, The Netherlands.

PMID: 24508145
DOI: 10.1016/j.jclinepi.2013.11.003

Abstract

Objectives: To assess whether relative or absolute effect measures were used in subgroup analyses of randomized controlled trials (RCTs) and study whether conclusions would change if subgroup effects were calculated on a different scale than reported.

Study design and setting: We studied all 327 RCTs published in 2010 in five major medical journals. For trials with a dichotomous primary outcome, we extracted reported main and subgroup effect measures. If crude subgrouping data were reported, we calculated the subgroup effects on both relative and absolute scales.

Results: Of the 229 RCTs with a dichotomous primary outcome, 120 (52%) performed subgroup analyses. In 106 of these 120 (88%) RCTs, relative effect measures were used for subgroup analyses, whereas an absolute scale was used in 9 (8%) trials. Two (2%) RCTs reported both relative and absolute subgroup effects. Crude data of the subgroups could be extracted in 41 of the 120 (34%) RCTs. Calculating subgroup effects on a different scale than reported lead to a change in conclusion in 17% of the 41 trials.

Conclusion: Almost all RCTs used relative effect measures for subgroup analyses. Interpretation of subgroup effects, however, appeared to be dependent on whether relative or absolute effect measures were used.

Keywords: Absolute risk reduction; Epidemiology; Randomized controlled trials; Relative risk reduction; Subgroup analysis; Treatment effects.

Publication types

Review

MeSH terms

Data Interpretation, Statistical*
Humans
Randomized Controlled Trials as Topic / methods*
Randomized Controlled Trials as Topic / standards
Risk Reduction Behavior
Treatment Outcome