Guidelines often use the term expert opinion (EO) to qualify recommendations. We sought to identify the rationale and evidence type in EO recommendations. We searched multiple databases and websites for contemporary guidelines published in the last decade that used the term EO. We identified 1106 references, of which 69 guidelines were included (2390 recommendations, of which 907 were qualified as EO). A rationale for using EO designation was not provided in most (91%) recommendations. The most commonly cited evidence type was extrapolated from studies that did not answer guideline question (40% from randomised trials, 38% from observational studies and 2% from case reports or series). Evidence extrapolated from populations that were different from those addressed in the guideline was found in 2.5% of EO recommendations. We judged 5.6% of EO recommendations as ones that could have been potentially labelled as good practice statements. None of the EO recommendations were explicitly described as being solely dependent on the clinical experience of the panel. The use of EO as a level of evidence in guidelines remains common. A rationale for such use is not explicitly provided in most instances. Most of the time, evidence labelled as EO was indirect evidence and occasionally was very low-quality evidence derived from case series. We posit that the explicit description of evidence type, as opposed to using the label EO, may add clarity and transparency and may ultimately improve uptake of recommendations.
Background: The quality of reporting of harms data in randomised controlled trials (RCTs) has been reported to be suboptimal. Rheumatoid arthritis (RA) has seen a massive growth in novel pharmacotherapies in the last decade.
Objective: The aim of this study was to assess the quality of reporting of harms-related data in RCTs evaluating pharmacological interventions for RA according to the CONSORT (Consolidated Standards of Reporting Trials) statement on harms reporting extension.
Study selection: RCTs published between January 2011 and August 2016 in the five highest impact factor journals in general medicine and two in rheumatology subject categories as per 2015 Journal Citation Reports were included. Reports of secondary, supplementary or exploratory analyses of RCTs and non-inferiority trials were excluded. Two reviewers independently extracted data using a structured, pilot-tested, 18-item questionnaire developed based on CONSORT harms extension recommendations.
Findings: 68 RCTs were included in the review. Out of a maximum harms reporting score of 18, the mean (SD) score was 8.51 (3.5) (range=0-15). More than half (56.5%) of the RCTs reported ≤50% of items and only three (4.3%) RCTs reported more than 70% (score ≥14) of the items. Multilinear regression analyses found that region of trial origin (p=0.01), sample size (p=0.001) and whether the study was a long-term extension of a trial or not (p=0.04) were independent predictors associated with higher total harms reporting score.
Conclusions: The adherence to CONSORT harms extension was poor in recently published RCTs of pharmacological interventions for RA. There is a need to improve quality of harms reporting in RCTs to allow transparent and balanced assessment of the benefit-risk ratio in clinical decision making.