The Cochrane risk of bias tool is the standard instrument for judging how far a randomised controlled trial's result can be trusted within a systematic review. Its current version, RoB 2, works through five pre-defined domains, asks structured signalling questions in each, and arrives at a judgement of low risk, some concerns, or high risk of bias. Rather than scoring overall study quality, it targets the specific mechanisms, such as a broken randomisation sequence or selective reporting, through which a trial's estimate can be distorted.
This domain-based, mechanism-focused design is what distinguishes it from quality scales: it does not ask "is this a good study" but "in which specific ways might this result be wrong, and how serious is each".
Why bias is assessed by mechanism, not by a score
Older quality scales bundled many features into a single number, which obscured where a trial was actually weak: a study could score moderately overall while harbouring one fatal flaw. The Cochrane approach instead isolates distinct bias mechanisms and judges each separately, so a trial with impeccable randomisation but heavy unexplained dropout is flagged precisely on that dropout. This is also why the tool pairs with design-specific instruments. Randomised trials use RoB 2, while non-randomised observational studies need a different tool such as the Newcastle-Ottawa Scale or ROBINS-I, a matching covered in our overview of quality assessment tools.
The five domains of RoB 2
RoB 2 organises the assessment into five domains, each capturing a different point at which bias can enter a trial.
| Domain | Bias it captures |
|---|---|
| Randomisation process | Flaws in how the allocation sequence was generated and concealed |
| Deviations from intended interventions | Departures from the assigned treatment and how they were analysed |
| Missing outcome data | Bias from participants lost or excluded from analysis |
| Measurement of the outcome | Bias from how the outcome was assessed, including assessor awareness |
| Selection of the reported result | Cherry-picking among multiple analyses or outcomes |
Within each domain, a set of signalling questions guides the reviewer to an algorithm-driven judgement, and the five domain judgements roll up into an overall risk of bias rating for that outcome. Crucially, RoB 2 assesses bias per outcome, not once for the whole study, because a trial can be at low risk for one outcome and high risk for another.
RoB 2 versus the original RoB 1
The original tool, now called RoB 1, judged bias once per study across domains such as sequence generation, allocation concealment, and blinding, recorded as low, unclear, or high. RoB 2 reorganised this into the five domains above, replaced the vague "unclear" category with a clearer "some concerns" judgement, introduced structured signalling questions to make assessments more consistent between reviewers, and shifted the unit of assessment from the study to the individual outcome. The result is more reproducible but also more demanding, since you repeat the process for each outcome rather than once overall.
| Feature | RoB 1 | RoB 2 |
|---|---|---|
| Unit assessed | Whole study | Each outcome |
| Middle category | Unclear | Some concerns |
| Signalling questions | No | Yes |
| Judgement method | Reviewer assessment | Algorithm from signalling questions |
Known limitations
RoB 2 is not without criticism. It is time-intensive, especially across many trials and outcomes; the signalling-question algorithm can still produce judgements that experienced reviewers find counterintuitive; and inter-rater agreement, while improved, is not guaranteed without training and piloting. It also applies only to randomised trials, so a review mixing trials and observational studies must run two different tools and present them coherently. None of this argues against using it, but a careful review acknowledges these constraints rather than presenting the ratings as objective fact.
Where it sits in the review
Risk of bias assessment follows data extraction and informs the synthesis and the certainty of evidence. It is one stage of the full systematic review workflow, applied to the studies that passed your eligibility criteria. When you reach the reporting stage, document how those studies were identified and selected with our free PRISMA 2020 flow diagram tool.
Frequently Asked Questions
What is the Cochrane risk of bias tool?
The Cochrane risk of bias tool is the standard instrument for assessing how far a randomised controlled trial's results may be distorted by methodological flaws. Its current version, RoB 2, works through five domains using structured signalling questions and produces a judgement of low risk, some concerns, or high risk of bias for each outcome, rather than a single overall quality score.
What are the five domains of the Cochrane risk of bias tool?
RoB 2 has five domains: bias arising from the randomisation process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in measurement of the outcome, and bias in selection of the reported result. Each domain is judged with signalling questions, and the five domain judgements combine into an overall risk of bias rating for the outcome.
What is the difference between RoB 1 and RoB 2?
RoB 1 assessed bias once per study across domains such as sequence generation, allocation concealment, and blinding, using low, unclear, or high categories. RoB 2 reorganised the domains, replaced "unclear" with "some concerns", added structured signalling questions for consistency, and assesses bias per outcome rather than per study. RoB 2 is more reproducible but more time-consuming to apply.
What are the limitations of RoB 2?
RoB 2 is time-intensive, particularly when applied per outcome across many trials, and its signalling-question algorithm can occasionally yield judgements that reviewers find counterintuitive. Inter-rater agreement improves with training but is not automatic, and the tool applies only to randomised trials, so reviews including observational studies must use an additional tool alongside it.