Randomization, blinding, data handling and sample size estimation in papers published in Veterinary Anaesthesia and Analgesia in 2009 and 2019

Abstract

ObjectiveTo evaluate reporting of items indicative of bias and weak study design. Study designLiterature survey. PopulationPapers published in Veterinary Anaesthesia and Analgesia. MethodsReporting of randomization, blinding, sample size estimation and data exclusion were compared for papers published separated by a 10 year interval. A reporting rate of more than 95% was considered ideal. The availability of data supporting results in a publicly accessible repository was also assessed. Selected papers were randomized and identifiers removed for review, with data from 59 (57 in 2009, two in 2008) and 56 (52 in 2019, four in 2018) papers analyzed. Items were categorized for completeness of reporting using a previously published operationalized checklist. Two reviewers reviewed all papers independently. ResultsFull reporting of randomization increased over time from 13.6% to 85.7% [95% confidence interval (CI), 57.8–86.6%; p < 0.0001], as did sample size estimation (from 0% to 20%; 95% CI, 7.6–32.4%; p = 0.002). Reporting of blinding (49.2% and 50.0%; 95% CI, –18.3% to 20.0%; p = 1.0) and exclusions of samples/animals (39.0% and 50.0%; 95% CI, –8.8% to 30.8%; p = 0.3) did not change significantly. Data availability was low (2008/2009, zero papers; 2018/2019, two papers). None of the items studied exceeded the predetermined ideal reporting rate. Conclusions and clinical relevanceThese results indicate that reporting quality remains low, with a risk of bias.