The risk-of-bias profile of neonatal randomised-controlled trials in Cochrane reviews: are we better over the years?

Lai NM1, Ong JM1, Chaiyakunapruk N2
1Taylor's University, Malaysia, 2Monash University Sunway Campus, Malaysia

Background:

The introduction of Neonatology as a subspecialty in 1960 has stimulated an enormous amount of neonatal research. A large proportion of neonatal randomised controlled trials (RCTs) are captured in the Cochrane Reviews. An integral part of a Cochrane Review is the risk-of-bias (RoB) assessment, in which study methodology is assessed under six major criteria. An evaluation of RoB in a representative sample of neonatal studies published over the last few decades has not been formally reported.

Objectives:

We determined the RoB profile of neonatal RCTs published since the 1950s.

Methods:

We conducted a secondary analysis from published Cochrane Reviews under the Neonatal Review Group. All reviews with RoB assessments were included. We extracted the review authors’ judgment on random sequence generation, allocation concealment, blinding, incomplete outcome data and selective reporting. We evaluated blinding by taking into account of its feasibility. We reported standard descriptive statistics (SPSS 20, Chicago, IL, USA).

Results:

A total of 1612 RCTs published between 1952 and 2012 were assessed in 258 reviews, with full RoB assessments performed in 551 trials (34%). Across the RoB criteria, the largest number of trials (72%) fared well (judged as ‘low-risk') in handling incomplete outcome data, while the fewest trials achieved blinding of outcome assessment (39%). Only 65 trials (12%) fared well across all criteria.

Over six decades, there has been progressive increase in the proportion of trials that fared well in random sequence generation, allocation concealment and selective reporting. However, such increase was not observed in blinding (30% to 55% since the 1980s) and incomplete outcome data (60% to 75% since the 1960s). The proportion of trials that fared well across all criteria did not increase until the 2000s (8% in 1980s, 6% in 1990s, 15% in 2000s and 32% in 2010s).

Conclusions:

The quality of neonatal RCTs has improved steadily over the last six decades, although the improvement was not consistent across all criteria, especially for blinding, which remained unsatisfactory in majority of the trials.