Premium
Dealing with heterogeneous populations in randomized wound trials: Challenges and potential solutions
Author(s) -
Vollenweider Daniela,
Ebneter Ingrid,
Mayer Dieter,
Hafner Jürg,
Steurer Johann,
Puhan Milo A.
Publication year - 2012
Publication title -
wound repair and regeneration
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.847
H-Index - 109
eISSN - 1524-475X
pISSN - 1067-1927
DOI - 10.1111/j.1524-475x.2012.00806.x
Subject(s) - clinical trial , medicine , randomization , comparability , randomized controlled trial , intensive care medicine , subgroup analysis , surgery , meta analysis , mathematics , combinatorics
Chronic wounds have a great variety of etiologies and manifestations that influence wound healing. Such heterogeneity potentially threatens the validity and clinical usefulness of trials if not considered appropriately. In 82 randomized wound trials retrieved from 10 C ochrane reviews, we assessed if and how authors considered wound and other prognostically important characteristics in the conduct and analysis of wound trials. We assessed whether these characteristics were discussed, reflected in the eligibility criteria, used for prestratification or for adjustments to ensure comparability of treatment groups, and whether subgroup analyses were conducted to identify heterogeneity of treatment effects. Nine percent of all trials explicitly discussed characteristics that influence wound healing in the introduction and 43% in the Discussion section. Ninety percent of trials had at least one prognostically important characteristic as eligibility criterion. Only 11% of trials used prestratification, and 6% adjusted the results for imbalances between treatment groups. Twenty‐seven percent performed subgroup analyses with prognostically important characteristics defining subgroups. Chronic wound trials use simple randomization, but rarely adapt the study design and analysis to take the heterogeneity of patients into consideration. Collaborative multicenter trials would overcome many of the limitations and provide statistical power to detect important treatment effects both overall and in subgroups.