Thanks Buddyboy for the short refresher on Bayesian methods and their application to clinical trials. The translation into plain English is quite helpful.
Integrating "prior knowledge" into the design and analysis of a clinical trial sounds great... But what if the drug you are testing breaks new ground and is at the cutting edge of innovation? In other words, if you are taking a truly new and creative approach to a problem, wouldn't prior knowledge tend to blunt innovation and steer the study towards conventional, existing practice?
Think about prior knowledge vis-a-vis CCR5 inhibitors not named leronlimab. Maraviroc, Vicriviroc, Cenicriviroc--there is a history of failed trials and underwhelming efficacy for the lot of them. Merck tested vicriviroc--combined with Keytruda, mind you--in a Phase 2 trial in mCRC MSS/pMMR (NCT03631407) that was completed in June 2021. Two partial responses in 42 patients, for an OOR of 5%. PFS of 2 months. OS of 9.8 months. How can this prior knowledge help inform our mCRC trial with leronlimab? A Bayesian protocol conceivably would have suggested scrapping leronlimab before our recent cancer trial even began, because those other attempts at CCR5 inhibition were inconclusive if not failures.
Specific to leronlimab, we have conducted two trials with confusing or poor results--the NASH trial, and the Severe Covid trial. I don't think we need Bayes to tell us that it sure would have been nice to give those poor suckers 4 weeks of leronlimab instead of 2... And what would Bayes, or AI for that matter, make of our NASH trial? Who can say... perhaps a Bayesian analysis would have convinced the FDA to allow four weeks in the covid trial. Or would have designed a better NASH trial, or made more sense out of the results. Will Bayesian methods do anything about outright corruption, if not incompetence, at the FDA? And more to the point--how will it deal with poorly designed historical trials and incomplete knowledge about CCR5 inhibitors as a class, in regards to a novel mab with a unique MOA?
It seems to me that the two biggest recent breakthroughs in cancer treatment--leronlimab paired with an ICI being one, and mRNA vaccines boosting the efficacy of ICIs, the other--were both accidental and retrospective. So I need to see how this Bayesian method will work in the real world, where a significant amount of "prior knowledge" turns out to be incomplete and contingent, if not outright wrong. It wasn't too far back that dietary fat was demonized, and carbs were cool. What would Bayes and the medical-industrial complex done with that? In my mind, truly innovative breakthroughs come about through collective knowledge, hard work, the occasional lucky accident, and the unparalleled ability of the human mind to bring disparate elements together in new, unique, and unconventional ways. Now, I'm an artist, so it figures I'd privilege the creative. But I've learned to value creativity in all walks of life, whether it happens in the darkroom, the clinic, or the laboratory. Creativity is absolutely necessary for innovation in the sciences, as odd as that may sound. Look at Dr Paul Maddon, a bio-chemical artist of the highest order, who managed to design an impeccably safe and effective drug, as compared to the other suboptimal CCR5 inhibitors out there.
Perhaps the Bayesian method will really do some good in streamlining clinical trials, with potential benefit to Cytodyn. Who knows! Adaptive trials, early stoppage of trials for success or failure, real-time modifications to trial parameters... That all sounds great, though of course it will depend on how these reforms play out. I was pleasantly surprised--almost shocked, really--at the flexibility shown by the FDA in our mCRC trial regarding dose escalation and providing for an ICI when appropriate. Perhaps we are an under-the-radar example of the FDA putting the Bayesian Methods to work? On the other hand, wouldn't a thorough review of prior knowledge have suggested the 350mg dose was useless in cancer if you want to boost PDL-1 numbers?
Bottom line--will this Bayesian reform of the clinical trial process reinforce conventional thinking at the expense of creativity in the clinic and the laboratory? And will this process compromise breakthrough innovation with a bias towards incremental advances and derivative, me-too drugs? More variations on existing symptom control, fewer cures because cures are infrequent in the historical literature?
And these days I can't help but wonder--will this be weaponized somehow, in some way, against small and innovative biotechs like Cytodyn, in favor of Big Pharma and their Wall Street masters? In this climate, is my investment safe from the finance bros and the Gilead goons? Will Trump feel entitled to a piece of leronlimab when the news breaks? That may sound outta line and conspiratorial... but hey, Prove Me Wrong! After the debacle of the severe covid trial, I admit to some paranoia and unease as we get closer and closer to the finish line. Time for a few deep breaths... and it's probably a good time to call it a night.
Integrating "prior knowledge" into the design and analysis of a clinical trial sounds great... But what if the drug you are testing breaks new ground and is at the cutting edge of innovation? In other words, if you are taking a truly new and creative approach to a problem, wouldn't prior knowledge tend to blunt innovation and steer the study towards conventional, existing practice?
Think about prior knowledge vis-a-vis CCR5 inhibitors not named leronlimab. Maraviroc, Vicriviroc, Cenicriviroc--there is a history of failed trials and underwhelming efficacy for the lot of them. Merck tested vicriviroc--combined with Keytruda, mind you--in a Phase 2 trial in mCRC MSS/pMMR (NCT03631407) that was completed in June 2021. Two partial responses in 42 patients, for an OOR of 5%. PFS of 2 months. OS of 9.8 months. How can this prior knowledge help inform our mCRC trial with leronlimab? A Bayesian protocol conceivably would have suggested scrapping leronlimab before our recent cancer trial even began, because those other attempts at CCR5 inhibition were inconclusive if not failures.
Specific to leronlimab, we have conducted two trials with confusing or poor results--the NASH trial, and the Severe Covid trial. I don't think we need Bayes to tell us that it sure would have been nice to give those poor suckers 4 weeks of leronlimab instead of 2... And what would Bayes, or AI for that matter, make of our NASH trial? Who can say... perhaps a Bayesian analysis would have convinced the FDA to allow four weeks in the covid trial. Or would have designed a better NASH trial, or made more sense out of the results. Will Bayesian methods do anything about outright corruption, if not incompetence, at the FDA? And more to the point--how will it deal with poorly designed historical trials and incomplete knowledge about CCR5 inhibitors as a class, in regards to a novel mab with a unique MOA?
It seems to me that the two biggest recent breakthroughs in cancer treatment--leronlimab paired with an ICI being one, and mRNA vaccines boosting the efficacy of ICIs, the other--were both accidental and retrospective. So I need to see how this Bayesian method will work in the real world, where a significant amount of "prior knowledge" turns out to be incomplete and contingent, if not outright wrong. It wasn't too far back that dietary fat was demonized, and carbs were cool. What would Bayes and the medical-industrial complex done with that? In my mind, truly innovative breakthroughs come about through collective knowledge, hard work, the occasional lucky accident, and the unparalleled ability of the human mind to bring disparate elements together in new, unique, and unconventional ways. Now, I'm an artist, so it figures I'd privilege the creative. But I've learned to value creativity in all walks of life, whether it happens in the darkroom, the clinic, or the laboratory. Creativity is absolutely necessary for innovation in the sciences, as odd as that may sound. Look at Dr Paul Maddon, a bio-chemical artist of the highest order, who managed to design an impeccably safe and effective drug, as compared to the other suboptimal CCR5 inhibitors out there.
Perhaps the Bayesian method will really do some good in streamlining clinical trials, with potential benefit to Cytodyn. Who knows! Adaptive trials, early stoppage of trials for success or failure, real-time modifications to trial parameters... That all sounds great, though of course it will depend on how these reforms play out. I was pleasantly surprised--almost shocked, really--at the flexibility shown by the FDA in our mCRC trial regarding dose escalation and providing for an ICI when appropriate. Perhaps we are an under-the-radar example of the FDA putting the Bayesian Methods to work? On the other hand, wouldn't a thorough review of prior knowledge have suggested the 350mg dose was useless in cancer if you want to boost PDL-1 numbers?
Bottom line--will this Bayesian reform of the clinical trial process reinforce conventional thinking at the expense of creativity in the clinic and the laboratory? And will this process compromise breakthrough innovation with a bias towards incremental advances and derivative, me-too drugs? More variations on existing symptom control, fewer cures because cures are infrequent in the historical literature?
And these days I can't help but wonder--will this be weaponized somehow, in some way, against small and innovative biotechs like Cytodyn, in favor of Big Pharma and their Wall Street masters? In this climate, is my investment safe from the finance bros and the Gilead goons? Will Trump feel entitled to a piece of leronlimab when the news breaks? That may sound outta line and conspiratorial... but hey, Prove Me Wrong! After the debacle of the severe covid trial, I admit to some paranoia and unease as we get closer and closer to the finish line. Time for a few deep breaths... and it's probably a good time to call it a night.
12
