RE: CC discusses AVA610318 Jan 2026 17:41
The fact 6103 designed using BOIN says they’re liaising with FDA and it’ll start there so MHRA sign off not hugely important. Credit to RAH for highlighting this.
https://x.com/blueberrymgmnt/status/2012896551639679104?s=46
#AVCT’s use of BOIN hasn’t really had much attention.
The old 3+3 design is very cautious. As we know from AVA6000, if you see 0/3 DLTs it escalates.
However, after 1/3 DLTs, it expands to 6 patients, and after 2/6 it declares the previous dose the MTD.
This frequently causes trials to stop escalation early or at low/subtherapeutic doses. Not a problem we had because of preCISION’s ability to remain “silent in the bloodstream”.
But for AVA6000 it meant moving desperately slowly.
Fast forward to AVA6103 and we will utilise BOIN, which uses optimal boundaries: an escalation boundary; and a deescalation boundary.
These are calibrated to minimise bad decisions. I.e. BOIN is designed to escalate more readily when the observed DLT rate is comfortably below the target. Thus allowing more patients at a promising dose without the rigid "6-patient cap" rule that 3+3 often imposes indirectly.
The effect of this is BOIN climbs the dose ladder MUCH faster and spends less time stuck at low doses when the true MTD is moderate or high.
When you are dealing with “normal” drugs this is important. When dealing with a platform which is designed to mask toxicity, it is essential.
When dealing with a highly toxic warhead (such as Exatecan in AVA6103) it will enable the full power of this warhead to be employed. Quickly.
By using BOIN, AVA6103’s trial will allocate patients closer to the current best estimate of the MTD.
BOIN is built on a ruled-based-interval-approach that treats the dose-finding “problem” as assigning patients to doses near the evolving estimate of the MTD. It is far more dynamic than 3+3 and should enable preCISION to shine much quicker.
It will naturally concentrate enrolment of patients at doses whose estimated toxicity is close to the target, whereas 3+3 often under-allocates to higher doses because of premature stopping or fear of toxicity.
Simulations consistently show BOIN assigns more patients (on average) to the true MTD across a wide range of dose-toxicity curves vs the 3+3 model.
It is also MSK’s preferred approach to trial design.
Not only will this ensure AVA6103 moves quickly, it will showcase the true potential of preCISION.
Roll on IND.
FDA is now open to Bayesian statistical approaches. A leap forward!
Bayesian statistics can help:
✅ Clinical trial design
✅ Finding the optimal dose
✅ Extrapolation to children
✅ Leveraging phase 2 results in phase 3