Reporting on blinding in trial protocols and corresponding publications was often inadequate but rarely contradictory Asbjørn Hro ´bjartsson a, * , Julie Pildal a , An-Wen Chan b , Mette T. Haahr a , Douglas G. Altman c , Peter C. Gøtzsche a a The Nordic Cochrane Centre, Rigshospitalet, Copenhagen, Denmark b Division of Dermatologic Surgery, Mayo Clinic, Rochester, MN, USA c Centre for Statistics in Medicine, University of Oxford, Oxford, UK Accepted 2 April 2009 Abstract Objective: To compare the reporting on blinding in protocols and articles describing randomized controlled trials. Study Design and Setting: We studied 73 protocols of trials approved by the scientific/ethical committees for Copenhagen and Fred- eriksberg, 1994 and 1995, and their corresponding publications. Results: Three out of 73 trials (4%) reported blinding in the protocol that contradicted that in the publication (e.g., ‘‘open’’vs. ‘‘double blind’’). The proportion of ‘‘double-blind’’trials with a clear description of the blinding of participants increased from 11 out of 58 (19%) when based on publications alone to 39 (67%) when adding the information in the protocol. The similar proportions for the blinding of health care providers were 2 (3%) and 22 (38%); and for the blinding of data collectors, they were 8 (14%) and 14 (24%). In 52 of 58 publications (90%), it was unclear whether all patients, health care providers, and data collectors had been blinded. In 4 of the 52 trials (7%), the protocols clarified that all three key trial persons had been blinded. Conclusions: The reporting on blinding in both trial protocols and publications is often inadequate. We suggest developing interna- tional guidelines for the reporting of trial protocols and public access to protocols. Ó 2009 Elsevier Inc. All rights reserved. Keywords: Blinding; Reporting; Protocol; Randomized clinical trials; Bias; Ethics committee 1. Background Blinding of key persons involved in clinical trials can reduce bias in effect estimates [1e7]. For example, an over- view of empirical studies of bias in meta-analyses found that beneficial effects on subjective outcomes were, on average, 25% lower in trials labeled ‘‘double blind’’ com- pared with similar trials that were not ‘‘double blind’’ [1]. Blinding of patients may result in more reliable report- ing of symptoms [2], and blinding of data collectors may result in more reliable reporting of outcomes that involve clinical judgment [3,5]. Furthermore, blinded patients may also differ from unblinded ones in their tendency for dropping out of the trial, or seeking additional treatment, and blinded health care providers may differ from un- blinded ones in how they attend to patients [4] or in their use of alternative forms of care. The reporting in a publication of who was intended to be blinded in a trial [8e11], and by which mechanisms [12,13], is often incomplete, making it difficult for readers to assess whom an intended blinding procedure involved, and whether it was likely to be successful. This is especially problematic for authors of systematic reviews assessing the risk of bias be- cause of lack of blinding. It is unclear how often lack of report- ing of blinding implies actual lack of blinding or undisclosed blinding (i.e., blinding that has been conducted but not reported). Previous studies [9,14] were based on contact with authors who might have given unreliable replies [15]. To our knowledge, the reporting of blinding in protocols, and its relation to the reporting of blinding in publications, have not previously been studied. Protocols focusing on the methodological aspects of trials are formulated before re- sults accumulate, and have no space restrictions. The infor- mation on blinding in protocols could, therefore, be more comprehensive and reliable than the information provided in publications and from contact to authors. Thus, our objective was to describe the information on blinding in trial protocols and in the corresponding * Corresponding author. The Nordic Cochrane Centre, Rigshospitalet, Dept 3343, Blegdamsvej 9 DK-2100 Copenhagen Ø, Denmark. E-mail address: ah@cochrane.dk (A. Hro ´bjartsson). 0895-4356/09/$ e see front matter Ó 2009 Elsevier Inc. All rights reserved. doi: 10.1016/j.jclinepi.2009.04.003 Journal of Clinical Epidemiology 62 (2009) 967e973