# A randomised trial of AI-generated treatment recommendations versus tumour boards

Source: https://onco.cc/ideas/idea-data-ai-vs-tumour-board-rct/  
OnCo record `idea-data-ai-vs-tumour-board-rct` (Idea). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

Test head to head whether an AI that reads the record and the evidence recommends treatments as well as a panel of experts, and whether patients do as well.

## Summary

AI systems that propose treatment plans from the record and the literature have been compared with tumour boards only retrospectively, with concordance as the metric. The proposal is a prospective, randomised non-inferiority trial in a defined setting (for example, first-line metastatic NSCLC or colorectal cancer): patients are randomised to have their plan generated by the AI (with clinician sign-off and override) or by the standard board, with guideline concordance, time to treatment, trial enrolment and 12-month outcomes as endpoints, and full provenance logging for every AI recommendation.

## Fields

- Kind: Idea
- Last checked: 2026-09-08
- Hypothesis: AI-generated plans with clinician sign-off will be non-inferior to tumour boards on guideline concordance and clinical outcomes and superior on time to treatment and trial referral, in the studied settings.
- Rationale: Tumour board capacity is a bottleneck and boards show variable concordance; if AI can match them in defined settings, expert time can be redirected to complex cases. The claim is testable only prospectively.
- Proposed test: A 600-patient randomised non-inferiority trial in one indication across five centres; pre-registered with concordance, time to treatment and 12-month PFS endpoints.
- Maturity: speculative
- Actor: research

## Sources

- Bottleneck evidence (AI that is built but not validated or deployed): Wu et al., How medical AI devices are evaluated: limitations and recommendations from an analysis of FDA approvals (Nature Medicine 2021): https://doi.org/10.1038/s41591-021-01312-x

## Connected records

- ideas: [A tumour board assistant that cites its evidence and tracks outcomes](https://onco.cc/ideas/idea-data-tumour-board-evidence-assistant/), [Decision support that cites the exact trial and guideline line it relies on](https://onco.cc/ideas/idea-data-provenance-first-decision-support/)
- fronts: [AI & Computation](https://onco.cc/fronts/ai-computation/)
- bottlenecks: [AI that is built but not validated or deployed](https://onco.cc/bottlenecks/b-ai-validation/), [Knowledge reaches practice too slowly](https://onco.cc/bottlenecks/b-knowledge-diffusion/), [Not enough oncologists, nurses, pathologists, physicists](https://onco.cc/bottlenecks/b-workforce/)
- key papers: [How medical AI devices are evaluated: limitations and recommendations from an analysis of FDA approvals](https://onco.cc/key-papers/paper-wu-nat-med/)

---
JSON: https://onco.cc/api/v1/entities/idea-data-ai-vs-tumour-board-rct.json