Background: Randomized controlled trials (RCTs) represent the most valuable study design to evaluate the effectiveness of therapeutic interventions. However, flaws in design, conduct, analysis, and reporting of RCTs can cause biased results. Cochrane published a 'Risk of bias' (RoB) tool to standardize the assessment of RoB for authors of systematic reviews (SRs). RoB concerns eight items assessed as being at low, unclear or high RoB. Our objective was to provide an overview of the potential sources of bias in RCTs of the otorhinolaryngologic research field in the past literature (1950-2012), and to identify areas where improvement is still warranted.
Methods: We retrieved all otorhinolaryngologic Cochrane SRs published in 2012 and 2013 using a combination of search filters. From the included SRs, we adopted all RoB assessments by the SR authors of the included RCTs. Descriptive statistics of the RoB assessments of the included RCTs were computed. We plotted the development of the RoB per item (potential source of bias) per decade, and analyzed the development statistically with? a multinomial logistic regression analysis.
Results: We extracted data from 42 SRs and 402 included RCTs (median 7, range 1-40). In total 2356 RoB items were assessed (median per RCT 6, range 1-12). Thirty-six (9.0%) out of 402 RCTs were assessed with a low RoB on all items, and 208 (51.7%) RCTs were assessed with at least one item at a high RoB. The number of RCTs with high RoB assessments remained constant. On multinomial logistic regression, there appears to be an increase in recent decades in the number of RoB items judged as being at low RoB for random sequence generation, allocation concealment and blinding of outcome assessment. Most of the differences between decades, however, are not statistically significant.
Conclusion: Although there were some positive developments in the overall bias in RCTs in the otorhinolaryngologic literature, a further decrease in bias results is still warranted. Currently, biased RCTs are included in SRs and effects of interventions can be under- or overestimated, with implications for clinical patient care.