Venetoclax in MYC-BCL2 double-hit lymphoma was as rational a combination as oncology produces. The disease is driven by BCL2 overexpression; the drug inhibits BCL2. The hypothesis was not speculative.
It killed six of 37 patients on treatment, four from sepsis, against one of 36 controls.
The mechanism was probably right. What it could not predict was what happens when a BCL2 inhibitor meets an already myelosuppressive backbone in patients with a median age of 65 — and febrile neutropenia rates, 43% against 37%, do not account for the gap. Something about the combination was worse than either part suggested.
The habit is to notice when you are reasoning from mechanism to safety, and to stop. Mechanism generates hypotheses about efficacy; it says almost nothing about toxicity in combination, because toxicity emerges from interactions nobody modelled. This is the argument for randomised safety data before adoption, and it is the argument that gets waived most often when the biology is elegant.
- Mechanism predicts efficacy hypotheses, not combination toxicity
- Additive myelosuppression is not additive — it compounds unpredictably
- Watch for on-treatment mortality separately from graded adverse events
- A trial stopped early for harm has told you something a completed one might not
- Elegant biology is where safety scepticism is most often suspended
The statistics, in plain English
Trials halted for harm are systematically imprecise: stopping early means fewer events and wider intervals, so the estimate of how much harm occurred is poor. That imprecision is a feature of acting responsibly, not a weakness of the evidence — the alternative is continuing until the interval narrows, which means continuing to randomise patients to a treatment already suspected of killing them.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for oncology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free