First of all I have to say thank you for taking a scientific approach in analysing Sellas. I mean this is really good to see given that there is so much shouting around on this topic and it is bloody hard to have a deeper discussion. I have been digging through Sellas for a long time and want to point out following problems that I see in your analysis:
The "Futility" issue (IDMC meetings)
The first analytical problem to me is the August 2025 IDMC (and also previous) recommendation. If the BAT arm and the GPS arm were performing similarly (or worse), the Hazard Ratio would approach 1.0. In such a scenario, the IDMC—which sees unblinded data—would have likely halted the trial for futility during the August 2025 (or earlier) review. The IDMC recommended the trial continue without modification. This implies that even with the "slow event accrual," there is still a statistically significant separation between the two curves that justifies reaching the 80th event.
Misapplication of Venetoclax "Super Responder" Data
That 21.6-month figure almost always includes patients who bridge to a Stem Cell Transplant (SCT). In AML, transplant is the only "reset button" for survival, whereas the REGAL trial specifically enrolls patients who are not candidates for SCT. For relapsed/refractory (R/R) AML patients who achieve CR2 but do not go to transplant, the mOS for HMA/Venetoclax is historically closer to 6.5–9.0 months, not 21.6.
The "Event Velocity" vs. "Salvage" Logic
Slow death rates because of better salvage therapies like Revumenib - salvage therapy extends life after a relapse. However, to reach a pooled mOS of 13.5+ months (which SELLAS reported as a blinded aggregate in early 2025), the patients must stay in remission or stay alive longer initially.
• Immunotherapy Signature: In clinical trials, a "slow-down" or "plateau" in deaths late in a trial is a classic hallmark of immunotherapy (vaccines), where responders simply stop relapsing. Control arms (BAT), even with modern salvage, typically show a steady, predictable decay in survival.
• The Math: If only 1.0 death per month occurred in 2025, it suggests the remaining survivors are concentrated in one arm that has reached a survival plateau. If both arms were simply "living longer due to salvage," the event accrual would be slower but still steady, not a near-total halt.
The "Lymphocyte Count"
Requiring >300 lymphocytes/µL filters out the sickest patients, padding the BAT arm. While 100% true, it most definitely gives a boost to GPS, because this is a vaccine; it requires a functioning immune system to work. By filtering for immune-competent patients, the trial is also ensuring that the GPS arm is at its maximum potential efficacy. If the trial were "diluted" with immunosuppressed patients, the vaccine would fail regardless of how the BAT arm performed.
Genetic Enrichment (NPM1/IDH)
If the trial is enriched for NPM1 (intermediate risk), the BAT arm does better. But if the trial is enriched for the typical CR2 population (high-risk/WT1+), GPS has a biological "home field advantage." You assume the selection bias only helps the control, ignoring that these same "healthy" fit patients are exactly the ones most likely to mount the robust T-cell response GPS requires to succeed.
I mean I agree with you 100% that we will see a totally different BAT arm mOS in REGAL study. Given the differences in salvage therapies what have come to market after phase 2 study, I have no doubt that BAT arm will produce a mOS in the range of 14-18 months. My view simply is that GPS arm will benefit even more from those new salvage therapies and will get to 40+ months mOS. It is a actually a statistical task – if you run statistical models with hard anchor data (enrollment pace, interim deaths, etc.), then you see that in no way can BAT arm mOS be lower than 13-14 months and at the same time you also need plateauing in GPS arm to get to this very low number of deaths. But most importantly, IDMC who sees unblinded data, has absolutely no point to say “continue without modification”, if GPS and BAT are not separating.
So great write up there is hope for humanity. 10% of the population would be expected to be eligible for revuforj and half of them should transplant! My napkin math with all of this factored in gives this trial a 1-8 chance of hr < .64 (bar of success) if expected bat mos is 13m (could be higher). A similar phenomena sunk glycomemtics, with control arm having a mos >12 and they didn't have revuforj or the filtering. This screams fail. I have puts.
Neutrophils over 1k are required to be considered in remission.
20k platelets is severe thrombopenia so these are the CR2p patients with incomplete platelet recovery, so these are patients with partial bone marrow failures indeed.
Same goes for lymphocyte count. ALC of 300 is pretty low and it's not a standardized ALC15 or ALC28 but rather a random snapshots, these patients can very well be in the ALC strata that do very poorly.
IDH is targetable so these patients are excluded from REGAL this leaves only the NPMmut patients reacting favourably to Venetoclax assuming they are in the 25% of BAT receiving VEN...
Enrollment criteria for VIALE-A were so strict that meeting these criteria were so strict that meeting these criteria alone was associated with increased survival.
Quazar AML trial leading to oral Aza becoming standard of care was so flawed that Vinay Prasad wrote a whole paper using it as an example for poor methodology.
These criteria are the least consequential ones. Care to address the fact that >6 months life expectancy is required, and how patients are Veno super responders?
25% of BAT receiving veno. How on earth did you come up with that? Also, these are naive pts, so that is also a significant part of the argument.
Also, if veno is really 8 months mOS for these patients, then how does cutting off the left tail (life expectancy >6 mo) affect the median of an exponential distribution?
Ridiculous commentary, made to sound smart but is really stupid. So the doctors on 10/29/25 stating BAT AML CR2 life expectancy is short and you are smarter than they are? They actually treat patients and know exactly what arm they are in. IDMC multiple reviews, no halt for futility, no trial modifications. Try and model the 66 out of trial by March 1, 2024 and tell me again how BAT is superior? Out was dead, relapsed, suffering intol tox or have completed the then max 15 doses of GPS. Did you know the FDA allowed GPS to dose every 6 weeks to ad infinitum until relapse, then in year 2 extend dose period to once every other month, then in year 3, just 4 times per year. Yup, I really think it's BAT keeping the trial from hitting 80. What a joke you are.
Complete smear peice. Did Anson Funds - Moez Kassam pay you alot... just because he is about to lose his ass? CEO publicly told him to eat a big fat one. LOVEd that.
BAT is 8 to 9 months. Even at 16 month GPS still wins. You have a lot of assumptions for sucha strong opinion.
Aron - may I ask your background? Are you in the biomedical field or involved with running trials in any way? Serious question. Just trying to understand your background and qualifications to speak to the trial design and scientific and medical data please.
First of all I have to say thank you for taking a scientific approach in analysing Sellas. I mean this is really good to see given that there is so much shouting around on this topic and it is bloody hard to have a deeper discussion. I have been digging through Sellas for a long time and want to point out following problems that I see in your analysis:
The "Futility" issue (IDMC meetings)
The first analytical problem to me is the August 2025 IDMC (and also previous) recommendation. If the BAT arm and the GPS arm were performing similarly (or worse), the Hazard Ratio would approach 1.0. In such a scenario, the IDMC—which sees unblinded data—would have likely halted the trial for futility during the August 2025 (or earlier) review. The IDMC recommended the trial continue without modification. This implies that even with the "slow event accrual," there is still a statistically significant separation between the two curves that justifies reaching the 80th event.
Misapplication of Venetoclax "Super Responder" Data
That 21.6-month figure almost always includes patients who bridge to a Stem Cell Transplant (SCT). In AML, transplant is the only "reset button" for survival, whereas the REGAL trial specifically enrolls patients who are not candidates for SCT. For relapsed/refractory (R/R) AML patients who achieve CR2 but do not go to transplant, the mOS for HMA/Venetoclax is historically closer to 6.5–9.0 months, not 21.6.
The "Event Velocity" vs. "Salvage" Logic
Slow death rates because of better salvage therapies like Revumenib - salvage therapy extends life after a relapse. However, to reach a pooled mOS of 13.5+ months (which SELLAS reported as a blinded aggregate in early 2025), the patients must stay in remission or stay alive longer initially.
• Immunotherapy Signature: In clinical trials, a "slow-down" or "plateau" in deaths late in a trial is a classic hallmark of immunotherapy (vaccines), where responders simply stop relapsing. Control arms (BAT), even with modern salvage, typically show a steady, predictable decay in survival.
• The Math: If only 1.0 death per month occurred in 2025, it suggests the remaining survivors are concentrated in one arm that has reached a survival plateau. If both arms were simply "living longer due to salvage," the event accrual would be slower but still steady, not a near-total halt.
The "Lymphocyte Count"
Requiring >300 lymphocytes/µL filters out the sickest patients, padding the BAT arm. While 100% true, it most definitely gives a boost to GPS, because this is a vaccine; it requires a functioning immune system to work. By filtering for immune-competent patients, the trial is also ensuring that the GPS arm is at its maximum potential efficacy. If the trial were "diluted" with immunosuppressed patients, the vaccine would fail regardless of how the BAT arm performed.
Genetic Enrichment (NPM1/IDH)
If the trial is enriched for NPM1 (intermediate risk), the BAT arm does better. But if the trial is enriched for the typical CR2 population (high-risk/WT1+), GPS has a biological "home field advantage." You assume the selection bias only helps the control, ignoring that these same "healthy" fit patients are exactly the ones most likely to mount the robust T-cell response GPS requires to succeed.
I mean I agree with you 100% that we will see a totally different BAT arm mOS in REGAL study. Given the differences in salvage therapies what have come to market after phase 2 study, I have no doubt that BAT arm will produce a mOS in the range of 14-18 months. My view simply is that GPS arm will benefit even more from those new salvage therapies and will get to 40+ months mOS. It is a actually a statistical task – if you run statistical models with hard anchor data (enrollment pace, interim deaths, etc.), then you see that in no way can BAT arm mOS be lower than 13-14 months and at the same time you also need plateauing in GPS arm to get to this very low number of deaths. But most importantly, IDMC who sees unblinded data, has absolutely no point to say “continue without modification”, if GPS and BAT are not separating.
So great write up there is hope for humanity. 10% of the population would be expected to be eligible for revuforj and half of them should transplant! My napkin math with all of this factored in gives this trial a 1-8 chance of hr < .64 (bar of success) if expected bat mos is 13m (could be higher). A similar phenomena sunk glycomemtics, with control arm having a mos >12 and they didn't have revuforj or the filtering. This screams fail. I have puts.
Some flaws in your cute little write up:
Neutrophils over 1k are required to be considered in remission.
20k platelets is severe thrombopenia so these are the CR2p patients with incomplete platelet recovery, so these are patients with partial bone marrow failures indeed.
Same goes for lymphocyte count. ALC of 300 is pretty low and it's not a standardized ALC15 or ALC28 but rather a random snapshots, these patients can very well be in the ALC strata that do very poorly.
IDH is targetable so these patients are excluded from REGAL this leaves only the NPMmut patients reacting favourably to Venetoclax assuming they are in the 25% of BAT receiving VEN...
Enrollment criteria for VIALE-A were so strict that meeting these criteria were so strict that meeting these criteria alone was associated with increased survival.
Quazar AML trial leading to oral Aza becoming standard of care was so flawed that Vinay Prasad wrote a whole paper using it as an example for poor methodology.
Do I need to continue?
These criteria are the least consequential ones. Care to address the fact that >6 months life expectancy is required, and how patients are Veno super responders?
I adressed the VEN "super responders" by adressing the IDH/NPM argument.
Sounds like you are blindly regurgitating something you read while lacking the understanding to connect these dots...
Life expectancy over 6 months benefits both arms equally so it's not really relevant but might in fact explain BAT mOS of 10-12 months.
GPS mOS is for sure exceeding the P2 results with higher immune response rate and extended dosing so that's not a concern.
25% of BAT receiving veno. How on earth did you come up with that? Also, these are naive pts, so that is also a significant part of the argument.
Also, if veno is really 8 months mOS for these patients, then how does cutting off the left tail (life expectancy >6 mo) affect the median of an exponential distribution?
Ridiculous commentary, made to sound smart but is really stupid. So the doctors on 10/29/25 stating BAT AML CR2 life expectancy is short and you are smarter than they are? They actually treat patients and know exactly what arm they are in. IDMC multiple reviews, no halt for futility, no trial modifications. Try and model the 66 out of trial by March 1, 2024 and tell me again how BAT is superior? Out was dead, relapsed, suffering intol tox or have completed the then max 15 doses of GPS. Did you know the FDA allowed GPS to dose every 6 weeks to ad infinitum until relapse, then in year 2 extend dose period to once every other month, then in year 3, just 4 times per year. Yup, I really think it's BAT keeping the trial from hitting 80. What a joke you are.
Complete smear peice. Did Anson Funds - Moez Kassam pay you alot... just because he is about to lose his ass? CEO publicly told him to eat a big fat one. LOVEd that.
BAT is 8 to 9 months. Even at 16 month GPS still wins. You have a lot of assumptions for sucha strong opinion.
BTW tell Moez I cant wait to drink his tears.
With full sincerity, I have never heard of Moez Kassam.
Go back and start your "research" over if you dont know who that is.... good god.
Then you have 0 credibility if you are admitting you dont know who that is. Absolutely 0 credibility.
Aron - may I ask your background? Are you in the biomedical field or involved with running trials in any way? Serious question. Just trying to understand your background and qualifications to speak to the trial design and scientific and medical data please.