Do we need Phase 3 trials in oncology?

Waste in cancer research has been disappointing to many whilst allowing significant capitalisation for pharmaceutical industry. Many improvements have been made and many more are yet to come. How should cancer research evolve? How can we decrease the waste and ensure better selection of pre-clinical models moving to clinic? Will we need phase 3 trials and what role RWE evidence will play? Will synthetic comparators give answers clinical practice needs? How can we overcome current incentives in the oncology market and stop over-production of expensive me-toos?
Speakers:
Chair: Dr. Leeza Osipenko
Dr. Marty Tanenbaum (Cancer Commons)
Dr. Julian Adams (Stand Up to Cancer)
Prof. Ian Tannock
Dr. Jonathan Kimmelman
Comments in chat:
Nathan Cherny: BEV in breast cancer was approved in a phase III study using a weak surrogate. It had no OS or QoL value.
Julian Adams: Nathan is absolutely correct.
Nathan Cherny: Many confirmatory demands are left hanging with data not forthcoming.
Eslam Maher: Are all these arguments against phase III trials? It doesn't seem like RWE can ever replace RCTs.
Nathan Cherny: There was a new low Grade glioblastoma approval last month

John Hickman: A recent survey by Tibau et al in the BMJ estimated that only 12% of molecularly targeted drugs recommended by the NCCN provided significant impacts on survival. Is this acceptable?
David Colquhoun: @Marty -you say IF they are safe and bioactive. But the problem is that you can't fulfil those conditions without phase 3
Nathan Cherny: Most single arm studies show ORR of only about 40%. There is no difference in the ORR of agents given accelerated approval or full approval. This leaves huge uncertainty if there is real clinical value
David Colquhoun: Exactly -thank you @Nathan
Michael: Fiendish problem also in psychoactive drugs. They are declared “safe enough” within a defined operational design domain (e.g. for depression, and for 6 weeks only). Then used way outside of their ODD in large populations.
But they are then sold for a panoply of symptoms and durations way outside of the ODD. Modal duration of serotonin prescriptions is YEARS e.g. in USA.
Pretty clear there are adverse effects including in tapering, and patients have invented techniques to taper themselves. Prescribers are clueless or are crying “help- I can’t help!” No relevant education/training.
Details of the safety cases that patients would dearly love to have, are held under lock and key.
Pressing practical matter to determine the fatness/length of the right tail of the adverse effect distribution.
For low probability/high severity, you have to test enough patients - 3-10x expected occurrence rate.
Can’t test your way out of it. Full stop. End of.
David Colquhoun: @Marty you ay be seeing nothing but regression to the mean!
David Colquhoun: And AI is only too often rubbish for scientific questions
David Colquhoun: LLMs are auto-complete on steroids. Totally untrustworthy
David Colquhoun: AI may or may not be much better in 5 years -don't believe the hype!