Background: one of the most important aspects of randomized controlled trials (RCTs) is their risk of bias (RoB). Cochrane has a toolkit for the assessment of RoB. Many of the Cochrane Oral Health systematic reviews suffer from a relatively high risk of bias of the included studies. Therefore, there is a need to assess RoB to have a better understanding of the current situation.
Objectives: to investigate the frequency and risk of bias of randomized controlled trials published by the Cochrane Oral Health Review Group.
Methods: we used the “dentistry & oral health” filter in the Cochrane Library on 23 April 2019 to retrieve reviews in this field. Then, we collected the characteristics of the included RCTs in each review, including the first author, the title, the field of study, the country in which the study had been conducted, and the RoB assessments by review authors. We included the updated version of a review if there was more than one version. We analysed the data using Microsoft Excel.
Results: the Cochrane Oral Health Review Group has published 198 reviews so far (2.5% of all Cochrane Reviews), of which 27 have been withdrawn and 17 included no RCTs. Overall, 2884 RCTs were included in our analysis. The majority of these studies were in oral medicine, pediatric dentistry, and community oral health (27.2%, 20.0%, and 18.1%, respectively). More than 35% of the included studies were conducted in the USA (722 RCTs) and the UK (302 RCTs), followed by India, Italy, and Sweden (184, 182, and 163, respectively). RoB was assessed as low for each domain as follows: generating random sequence (52.9%), allocation concealment (27.9%), participants/personnel blinding (39.0%), outcome assessors blinding (55.0%), addressing incomplete outcome data (60.5%), free of selective reporting (58.0%), and free of other sources of bias (53.0%). In two domains - generation of the random sequence and allocation concealment - the rate of unclear RoB judgements was relatively high (41.5% and 59.7%, respectively).
Conclusions: RoB in five out of the seven domains of the Cochrane RoB tool was low in about half of the included RCTs in the reviews published by Cochrane Oral Health. However, these RCTs may suffer from selection bias.
Patient or healthcare consumer involvement: we did not involve any patients. However, we sought comments from some of the healthcare consumers in a dental school in Tehran, Iran.