The Future of Molecular Tumour Boards

A decade ago, a molecular tumour board was unusual. Now many cancer centres run one. The question has shifted from whether to have one to what distinguishes a board that improves decisions from one that mainly documents them.

What a good board actually does

Assembles the right expertise. Molecular findings need a molecular scientist; pathology needs a pathologist; the treatment context needs the treating clinician. A board consisting mainly of one discipline reproduces that discipline's blind spots.

Reviews the case, not just the report. The most consequential contributions frequently concern something other than the variant list — that the pathology warrants review, that a prior scan was misread, that a test was performed on inadequate tissue.

States evidence levels explicitly. Not "consider drug X" but "drug X, supported by phase II data in this histology" or "supported by preclinical data only". The qualifier is the information.

Records what was not concluded. Cases arrive at a board precisely because they are uncertain. A record that captures only conclusions loses where the reasoning was thin, which is what a later reviewer most needs to know.

Where boards commonly underperform

Documenting rather than deciding. A meeting that records what was already going to happen consumes expert time without changing anything.

Volume over depth. Thirty cases in ninety minutes means three minutes each. Genuinely complex cases cannot be discussed in three minutes, and the format silently selects against depth.

No feedback loop. Most boards never learn what happened to the cases they discussed. Without that, calibration cannot improve — the board makes the same class of judgement indefinitely without ever finding out how it performed.

Unstructured output. A free-text note is difficult to audit and impossible to analyse in aggregate.

What is likely to change

Distribution. Remote participation means the relevant expert for a rare entity can join from elsewhere. For rare disease especially, this is the difference between a board with the necessary expertise and one without.

Machine preparation. Assembling the case — pulling the record together, retrieving relevant literature, screening the trial landscape — is preparatory work well suited to automation, freeing the meeting for judgement rather than assembly.

Structured records. Recording findings, evidence tiers, actionability categories and uncertainties in structured form makes boards auditable and comparable, and makes it possible to ask afterwards how often the board's judgements held up.

Longitudinal cases. Treating a case as a trajectory rather than a single discussion — returning to it as it evolves — fits the biology better than a one-off review does.

What should not change

The board exists to bring several kinds of expert judgement to a case that a single person would read less well. That is a human function.

Preparation can be automated. Retrieval can be automated. The judgement about what a thin evidence base supports, for this patient, in this situation — that is the meeting.