Incomplete reporting of experimental studies and items associated with risk of bias in veterinary research.

Abstract

In in vivo research, the reporting of core items of study design is persistently poor, limiting assessment of study quality and study reproducibility. This observational cohort study evaluated reporting levels in the veterinary literature across a range of species, journals and research fields. Four items (randomisation, sample size estimation, blinding and data exclusion) were assessed as well as availability of study data in publicly accessible repositories. From five general and five subject-specific journals, 120 consecutively published papers (12 per journal) describing in vivo experimental studies were selected. Item reporting was scored using a published scale (items ranked as fully, partially or not reported) according to completeness of reporting. Papers in subject-specific journals had higher median reporting levels (50.0 per cent vs 33.3 per cent, P=0.007). In subject-specific journals, randomisation (75.0 per cent vs 41.7 per cent, P=0.0002) and sample size estimation (35.0 per cent vs 16.7 per cent, P=0.025) reporting was approximately double that of general journals. Blinding (general 48.3 per cent, subject-specific 50.0 per cent, P=0.86) and data exclusion (general 53.3 per cent, subject-specific 63.3 per cent, P=0.27) were similarly reported. A single paper made study data readily accessible. Incomplete reporting remains prevalent in the veterinary literature irrespective of journal type, research subject or species. This impedes evaluation of study quality and reproducibility, raising concerns regarding wasted financial and animal resources.