Study registries have improved the transparency of randomised controlled trials (RCTs) in some respects, but they lack sufficient information about study design and conduct to ensure a reliable critical appraisal of trial outcomes, a new analysis finds.
The number of clinical trial registries and registered trials has been growing since 2004, when the International Committee of Medical Journal Editors introduced a requirement for public registration at study inception, notes a research team led by Ludovic Reveiz from the Research Institute at Sanitas University Foundation in Bogotá, Colombia. There is also general agreement about the minimum protocol information that should be registered for a trial, as defined by the World Health Organization’s (WHO) 20-item Registration Data Set.
However, Reveiz et al point out in the open-access journal PLoS ONE, the limited methodological information included in the WHO Data Set does not allow for a full appraisal of trial quality. For example, it contains a specific item on ‘study type’ that should incorporate the type of study (interventional or observational) and details on the study design (method of allocation, masking, methods of randomisation and study phase where applicable).
Yet it is unclear, the researchers add, how much detail should be provided on methods of randomisation, allocation concealment and study blinding. This is “understandable because trial registries were initially conceived to identify the existence of a trial and not to provide all the required information concerning methodological issues”.
To explore the extent to which these concerns are borne out in practice, Reveiz et al assessed the quality of methodological information available for a random sample of ongoing RCTs. They used the WHO International Clinical Trials Registry Platform to identify 265 trials registered between 1 January and 12 December 2008 on ClinicalTrials.gov and six WHO Primary Registries, and open for recruitment on 12 December 2008.
The researchers then extracted data on key methodological items from each registry record, using an evidence-based source developed by the Cochrane Collaboration to define domains indicating the risk of bias and other important flaws in RCTs. The primary outcomes were the proportion of registry records with adequate reporting of random sequence generation, allocation concealment, blinding and trial outcomes.
The weighted overall proportions of RCTs with adequate reporting of key methodological items were 5.7% for random sequence generation; 1.4% for allocation concealment; 41% for blinding; and 66% for primary outcomes. There was adequate reporting of secondary outcomes in 46% of the RCTs samples, while harms outcomes scored just 5%, eligibility criteria 81%, follow-up duration 62% and study interventions 53%.
Reporting of details about sample size calculations was especially lax, with adequate descriptions in only 1% of records, Reveiz et al observed. Although the target sample size was reported for 97% of the RCTs, the number of participants in each study arm was declared in only 7%.
Details on methodology varied substantially across the trial registries used. There was also variation between registries in the characteristics of the data entry fields – for example, only three offered specific fields for random sequence generation and allocation concealment.
Reporting of methodological information in trial registries has not been a focus of early registration requirements, “and consequently the quality of reporting of trial methods in registry records is poor overall”, the authors concluded.
It is “imperative that disclosure of trial results in public databases or journal publications be accompanied by sufficient methodologic information to full appraise them,” they argued. “Full trial protocols remain a key source of information on trial methodologies, and should be made publicly available.”