ROBUST-RCT
Risk of Bias instrument for Use in Systematic reviews — for RCTs
CoreSR implements the ROBUST-RCT tool with GLASS structured assessment. The model evaluates each item using a two-step approach — study-level then outcome-level — flag disagreements, and provide transparent rationale for every judgement.
AI Assessment Is a Starting Point
AI-generated assessments should always be reviewed and verified by a human assessor with domain expertise. The structured AI assessment highlights the evidence used, but the final judgement is yours. For published systematic reviews, all RoB assessments should be conducted by at least two independent human reviewers.
How It Works
Five steps from PDF upload to a complete two-step risk of bias assessment for randomized controlled trials.
Step-by-Step Guide
Each step builds on the previous one. AI assessment runs automatically after you provide the study details.
Upload Study PDFs
Upload one or more full-text PDF files for the randomized controlled trial you want to assess. The AI extracts text from the PDFs to inform its judgements at both the study and outcome levels.
Enter Study Details
Provide the study title, authors, publication year, experimental intervention, and comparator. Specify the outcome(s) you want to assess at the outcome level.
Run AI Assessment
Click "Evaluate with AI" to run the GLASS structured assessment. GLASS evaluates each item at both the study level and outcome level.
Review and Override
Review each item's AI justifications and judgements at both levels. You can override any judgement. Pay attention to study-level items (randomization, allocation, blinding) as these apply across all outcomes.
Two-Step Assessment
ROBUST-RCT separates assessment into study-level items (assessed once) and outcome-level items (assessed per outcome).
Study-Level Items
These items are assessed once per study and apply to all outcomes. They address the fundamental design characteristics of the trial.
Randomization
Was the allocation sequence adequately generated? Was a random component used in the sequence generation process? Was the method of randomization appropriate for the study design?
Allocation concealment
Was the allocation adequately concealed? Were steps taken to ensure that the allocation sequence was not known to those enrolling participants prior to assignment?
Blinding
Were participants, personnel, and outcome assessors blinded to the intervention? If blinding was not possible, were adequate steps taken to reduce the impact of knowledge of allocation?
Outcome-Level Items
These items are assessed separately for each outcome. Risk of bias can differ across outcomes within the same study.
Incomplete outcome data
Were incomplete outcome data adequately addressed? Was attrition or exclusion from analysis reported and were reasons provided? Were methods used to handle missing data appropriate?
Selective reporting
Was the study free of selective outcome reporting? Were all pre-specified outcomes reported? Were there discrepancies between the protocol and the published results?
Other bias
Was the study apparently free of other problems that could put it at risk of bias? This includes baseline imbalances, early stopping, funding bias, or other methodological concerns.
Dual AI Review Process
GLASS assesses each item with structured step 1 and step 2 prompts.
When Reviewers Agree
If the AI output supports a clear item judgement, the suggested judgement is shown directly. A green "AI Agreed" badge is shown. The combined rationale is displayed for transparency.
When Reviewers Disagree
If the reviewers reach different judgements, the more conservative (higher risk) judgement is applied as a default. An orange "Disagreed" badge flags the item. Expanding the item shows each reviewer's individual judgement and a resolution rationale explaining why they differed.
Human Override
You can override any item judgement at both the study and outcome levels. Your decision always takes precedence over the AI. Overrides are tracked and clearly marked in exports for audit trails.
Judgement Scale
The study is judged to be at low risk of bias for this item.
Insufficient information to determine whether the item is at low or high risk of bias.
The study is judged to be at high risk of bias for this item.
Frequently Asked Questions
Common questions about ROBUST-RCT assessments in CoreSR.
What is ROBUST-RCT?
How does ROBUST-RCT differ from RoB 2.0?
What is the two-step approach?
When should I use ROBUST-RCT vs RoB 2.0?
How does the "Unclear" judgement work?
How does the AI review work?
How is the overall risk of bias determined?
Can I export the results?
Ready to assess?
Upload a study PDF and let AI provide a first-pass ROBUST-RCT risk of bias assessment.
Go to ROBUST-RCTReferences
1. Wang Y, Keitz S, Briel M, et al. Development of ROBUST-RCT: Risk Of Bias instrument for Use in SysTematic reviews — for Randomised Controlled Trials. BMJ. 2025;388:e081199. doi:10.1136/bmj-2024-081199
2. Wang Y, Parpia S, Couban R, et al. Compelling evidence from meta-epidemiological studies demonstrates overestimation of effects in randomized trials that fail to optimize randomization and blind patients and outcome assessors. J Clin Epidemiol. 2024;165:111211. doi:10.1016/j.jclinepi.2023.11.001
3. Guyatt GH, Oxman AD, Vist G, et al. GRADE guidelines: 4. Rating the quality of evidence — study limitations (risk of bias). J Clin Epidemiol. 2011;64(4):407-415. doi:10.1016/j.jclinepi.2010.07.017