SELLAS: Why The Bear Case Might Not Hold Up

Disclosure: The author holds a beneficial long position in SELLAS Life Sciences Group, Inc. (NASDAQ: SLS). This article is provided for informational and educational purposes only and is not financial advice. Although the author is a Medical Doctor, this content represents a personal analytical perspective and does not constitute medical advice, a diagnosis, or a treatment recommendation. The author receives no compensation for this article and has no business relationship with the company. Please see the full "Legal Information and Disclosures" section below.

In my article "Monte Carlo Simulation Of Elicio's Phase II AMPLIFY-7P Study Favors Vaccine Efficacy", I used Monte Carlo simulation to map the delayed readout of Elicio's AMPLIFY-7P trial onto combinations of control-arm survival and vaccine hazard ratio. After running similar simulations for SELLAS' Phase 3 REGAL trial of galinpepimut-S (GPS) in CR2 acute myeloid leukemia, I found that this approach adds rather little insight here. SELLAS (NASDAQ: SLS) has disclosed far more information than Elicio, to the point where the analysis largely collapses into a single question: what is the median overall survival in the control group?

If the best available therapy (BAT) arm achieves an mOS of around 8 months, consistent with historical benchmarks for non-transplant-eligible AML patients in second complete remission, you do not need any simulation at all. Trial success would be highly probable given that fewer than 50% of enrolled patients had died at a median follow-up of 13.5 months, meaning the pooled median survival had not yet been reached and, by SELLAS' own estimate, exceeds 12 months, and the positive IDMC recommendations in January and August 2025 to continue without modification.

Those benchmarks deserve a brief note on sourcing. The GPS Phase 1/2 study reported an mOS of 5.4 months in a matched historical cohort of CR2 patients receiving standard of care. SELLAS itself cited 6 months in the January 2025 interim analysis press release. Dr. M. Yair Levy, Director of Hematologic Malignancies Research at Texas Oncology Baylor University Medical Center and a REGAL Steering Committee member, put the figure at around eight months, though his estimate explicitly includes treatment with HMAs and/or a BCL-2 inhibitor. All three sources are company-adjacent rather than independent registry data, but they bracket a range of roughly 5 to 8 months that is consistent with the generally poor prognosis described in the broader relapsed AML literature. 

It is worth noting, however, that these historical benchmarks derive from heterogeneous patient populations that include patients who would never have been fit enough to enroll in a clinical trial. REGAL required patients to have achieved CR2 and to meet performance status criteria for enrollment, which enriches the study population for fitter, biologically more favorable patients. This selection effect alone could elevate both arms' survival above historical norms, independent of any treatment effect from venetoclax or GPS.

The real question is whether an mOS of 14, 16, or even more months is plausible in the BAT group. And that is something no Monte Carlo simulation can answer.

Consider the timeline. The final analysis requires 80 events. As of December 26, 2025, SELLAS reported 72. Today is March 11, 2026. More than ten weeks have passed and no announcement of the 80th event has been made. Enrollment completed in April 2024 with 126 patients randomized. The company itself noted that survival times appear longer than expected, which they framed as potentially positive. Given the historical benchmarks discussed above, the fact that 8 remaining events have not accumulated in over ten weeks among the 54 patients recorded as alive or censored as of that date tells you something about the tail of the survival curve. Some patients are living longer than expected. The question is whether it is the GPS arm, the BAT arm, or both.

It should be noted that some of the slowdown could also reflect operational factors: CRO event-verification lag, reduced site activity during the holiday period immediately surrounding the December 26 cutoff, or delays in death certificate processing across six or more countries. These explanations are not mutually exclusive with a genuine slowing of events, but they caution against overinterpreting the precise timing.

The IDMC data offer partial answers. The committee reviewed unblinded data at the interim analysis, triggered by 60 deaths, and confirmed that GPS exceeded the predetermined futility criteria. This means the data did not demonstrate that GPS is futile; it does not mean the IDMC found positive evidence of efficacy, and the distinction matters. No safety concerns were identified. They recommended continuation without modification. A subsequent review in August 2025 reached the same conclusion. The trial was neither stopped for futility nor for efficacy.

Why neither? This is actually consistent with the mechanism of a therapeutic vaccine. GPS works by training the patient's T cells to recognize and kill WT1-expressing leukemic cells. Unlike a small molecule that acts immediately, a vaccine needs time to mount an immune response, and that response is inherently heterogeneous across patients. There will be non-responders whose outcome might even be worse than the control arm, since BAT patients might receive active treatment while non-responders get a vaccine that does not work for them. The responders, on the other hand, could show sustained, durable survival if the immune system successfully establishes long-term surveillance against WT1-positive residual disease. SELLAS disclosed that 80% of randomly selected GPS patients showed a specific T-cell immune response, surpassing the Phase 2 results. But that still leaves 20% who did not respond.

There is a scenario worth flagging here that sits between the bull and bear cases. What if GPS works for the 80% of immunologic responders, producing a genuine survival benefit, but the blended hazard ratio across all patients, diluted by the 20% non-responders, comes in at 0.70 or 0.75 rather than the more dramatic separation the trial may have been powered to detect? In that scenario, a real GPS effect exists but the trial fails to reach statistical significance on its primary endpoint.

The bimodal pattern of responders and non-responders would produce a rather late meaningful separation of the survival curves and an effect that is not dramatic in either direction at any single point in time. Early on, the curves would overlap because the vaccine has not yet primed the immune system. Then the non-responders would start relapsing and dying at rates similar to or worse than BAT, while the responders would diverge upward. At an interim look, this could easily look like a modest, non-significant trend. Only as follow-up lengthens and the responder tail stretches out would the full effect become visible. One could try to model this with a Monte Carlo simulation, but you would have to bake in assumptions about the responder fraction, the immune priming delay, and the differential hazard rates for responders versus non-responders. Each assumption biases the simulation considerably. I do not see how we gain information with an MC here that we cannot get from simply reasoning about the clinical reality of the control arm.

So instead of modeling event trajectories, let me reason about whether the bear case holds up.

To understand why this matters, a brief primer on the patient population. REGAL enrolled AML patients in second or later complete remission (CR2/CRp2) after second-line salvage therapy who are ineligible for or unable to undergo allogeneic stem cell transplantation. These patients have already relapsed once after initial treatment and then responded to a second round of aggressive chemotherapy. Their disease has demonstrated its ability to overcome prior therapy. The leukemic stem cells that survived induction and consolidation have proven themselves resistant. There is no approved maintenance therapy for these patients.

The permitted BAT options in REGAL include observation (with hydroxyurea if needed), a hypomethylating agent (azacitidine or decitabine), venetoclax, and low-dose cytarabine, or combinations thereof. Observation is the historical default. Low-dose cytarabine is an old, essentially palliative therapy with no meaningful survival impact in this setting. Injectable azacitidine and decitabine as maintenance agents have been studied but have not demonstrated significant survival benefit. The QUAZAR AML-001 trial did demonstrate an OS benefit for oral azacitidine (Onureg) in first-line maintenance, but that is a different formulation that is not bioequivalent to injectable azacitidine and was studied in CR1, not CR2. Oral azacitidine is not listed among the REGAL BAT options.

That leaves venetoclax as the only BAT component that could plausibly push the control arm's mOS far beyond historical norms. Venetoclax received accelerated FDA approval for newly diagnosed AML in November 2018 and full approval in October 2020, making it a relatively recent addition to the AML treatment landscape. This is exactly why the bear case exists: if enough BAT patients receive venetoclax-based regimens, the control arm might achieve an unprecedented mOS that narrows or eliminates the survival gap GPS needs to demonstrate.

But how realistic is this scenario?

First, venetoclax is not approved for the REGAL setting anywhere. The FDA granted approval for venetoclax in combination with azacitidine, decitabine, or low-dose cytarabine for newly diagnosed AML in adults who are age 75 years or older, or with comorbidities precluding intensive induction chemotherapy. The European Commission's approval is similar in scope: venetoclax in combination with a hypomethylating agent for newly diagnosed AML in adults ineligible for intensive chemotherapy. Neither covers CR2 maintenance, which is the REGAL population. Using venetoclax in this setting is off-label everywhere, including the United States. The VIALE-A trial demonstrated a median overall survival of 14.7 months for venetoclax plus azacitidine versus 9.6 months for azacitidine alone, but those patients had active, newly diagnosed disease where venetoclax plus azacitidine served as the primary antileukemic regimen. REGAL BAT patients are already in complete remission after salvage chemotherapy and would receive venetoclax as maintenance to prevent relapse in disease that has already demonstrated resistance to prior therapy. There is no controlled evidence that venetoclax-based maintenance in CR2 reproduces anything close to the frontline survival benefit.

Second, the geographic composition of the trial matters. REGAL enrolled 126 patients across approximately 110 sites. The trial launched in 2020 with sites in the U.S. and Europe only; Asian sites in India and China were added later, with China endorsed by the IDMC in April 2023, relatively late in the enrollment period. Among the European countries, the trial included sites in France, Germany, Greece, Hungary, Poland, and Serbia. SELLAS has stated that the U.S., Greece, and India were the top three enrolling countries, and Dr. Panagiotis Tsirigotis at the National and Kapodistrian University of Athens has been identified as the investigator who enrolled the highest number of patients in the study. Critically, study sites in the U.S. and Europe accounted for approximately 75% of patients enrolled, with the U.S. representing the single highest enrolling country. The remaining roughly 25% came from India and China.

Now consider venetoclax access across these geographies. In the United States, off-label use of oncology drugs is relatively common, and insurance coverage for off-label cancer treatments is more flexible than in most other countries. American investigators in the trial could plausibly prescribe venetoclax to their BAT patients.

But what about patients enrolled outside the U.S.? In India, venetoclax access remains limited and costly for indications beyond its approved label. Indian patients can only obtain it through cumbersome channels at substantial expense. For an off-label maintenance indication without guideline support, the probability of Indian BAT patients receiving venetoclax is very low. Greece, one of the trial's top three enrolling countries and home to the single highest-enrolling investigator, has well-documented constraints on oncology drug spending following years of austerity. Off-label venetoclax maintenance in CR2 is extremely unlikely to be reimbursed there. In Poland, venetoclax reimbursement is restricted even for certain approved indications. The bar for off-label AML maintenance reimbursement would be even higher. Hungary has some of the most stringent population restrictions for venetoclax access in Europe, even for approved indications. Serbia has limited oncology drug access overall. Germany and France offer the strongest healthcare systems among the European participants, and off-label venetoclax use is more plausible at their academic centers, but still not guaranteed for a CR2 maintenance indication that lacks any guideline recommendation.

The picture that emerges is this: the U.S. is the only geography where off-label venetoclax use in CR2 maintenance is realistically common. A significant share of the BAT patients come from the U.S. and could in theory receive venetoclax. But a substantial portion come from countries where venetoclax access ranges from restricted to nonexistent, even for approved indications. And as noted, Greece and India, two of the trial's three largest enrollment contributors, are among those countries. For those patients, "best available therapy" most likely means observation or HMA monotherapy, therapies that have not demonstrated meaningful survival improvements in this population.

Even within the U.S. contingent, it is not obvious that all or even most BAT patients would receive venetoclax. These are CR2 patients already in remission. Venetoclax combined with a hypomethylating agent carries real toxicity: severe cytopenias, risk of tumor lysis syndrome, and serious infections including fatal sepsis. For a fragile, already depleted patient currently in remission, an investigator might reasonably choose observation over an aggressive, toxic, off-label regimen. The decision to use venetoclax-based maintenance in a transplant-ineligible CR2 patient is not automatic, even at a major U.S. cancer center.

So what does the blended BAT arm likely look like? A mix of some venetoclax-treated U.S. patients and a larger number of patients on observation or HMA monotherapy from the U.S. and elsewhere. That blend might produce an mOS closer to historical benchmarks than the critics fear. Not 8 months, perhaps, because some venetoclax use, improved supportive care, and the selection of fitter patients inherent in trial enrollment will shift things somewhat. But 14 or 16 months? That would require a degree of venetoclax penetration and venetoclax efficacy in this specific setting for which there is no supporting evidence.

I want to be careful not to overstate this. Clinical trials produce surprises with remarkable regularity. The history of oncology is full of studies that failed or succeeded for reasons no one anticipated. There is also the Hawthorne effect to consider, amplified by the open-label design: investigators who know their BAT patients are not receiving an experimental vaccine might compensate by treating more aggressively, selecting therapies they would not have chosen outside a trial context. In an open-label trial, this effect operates in both directions. Investigators may work harder for their BAT patients precisely because they feel those patients are missing out on a potentially beneficial experimental therapy. And it is entirely possible that the BAT group performs better than the vaccine group. GPS is an immunotherapy, and immune responses are variable. The 20% non-responder rate alone could dilute the treatment effect enough to erase statistical significance if the BAT arm performs even moderately above expectations.

But return to the timeline. Seventy-two events by December 26, and now more than ten weeks of silence. In a population where 6- to 8-month mOS is the benchmark, the slowing of events is itself data. Combined with two positive IDMC reviews of unblinded data, the geographic reality of venetoclax access, and the limited efficacy of the remaining BAT options, the specific bear case that the control arm will produce an unprecedented mOS seems to rest on assumptions about drug access and efficacy that do not hold up well.

None of this guarantees that REGAL will succeed. The study could fail for any number of reasons, including a real but underpowered treatment effect diluted by non-responders, better-than-expected BAT performance driven by trial-enrichment for fitter patients, or the inherent variability of immune responses across a 126-patient study. The market started pricing in trial success in recent months, which reflects the fact that at some point any further delay of the readout is more likely signal than noise.

Follow me on X for frequent updates (@chaotropy).

Legal Information and Disclosures

General Disclaimer & No Financial Advice: The content of this article is for informational and educational purposes only. It represents the personal opinions of the author as of the date of publication and may change without notice. The author is not a registered investment advisor or financial analyst. This content is not intended to be, and shall not be construed as, financial, legal, tax, or investment advice. It does not constitute a personal recommendation or an assessment of suitability for any specific investor. This article does not constitute an offer to sell, a solicitation of an offer to buy, or a recommendation of any security. Readers should conduct their own independent due diligence and consult with a certified financial professional before making any investment decisions. This article is not a securities research report and has not been prepared in accordance with legal requirements designed to promote the independence of investment research.

Medical Disclaimer: Although the author possesses a medical background, the information presented here regarding clinical trials, survival endpoints, or pharmaceutical mechanisms is strictly for the purpose of educational discussion and general commentary regarding the underlying science. It does not constitute medical advice, a diagnosis, or a treatment recommendation, nor does it establish a physician-patient relationship. Readers should never disregard professional medical advice or delay in seeking it because of something read on this website. Always consult a qualified healthcare provider regarding any medical condition.

Accuracy and Third-Party Data: Clinical trial data, press releases, regulatory filings, and published literature referenced in this article are sourced from ClinicalTrials.gov, company press releases, peer-reviewed journals, and conference abstracts. While the author believes these sources to be reliable, the completeness, timeliness, or correctness of this data cannot be guaranteed. The author assumes no liability for errors, omissions, or the results obtained from the use of this information.

Disclosure of Interest: The author holds a beneficial long position in SELLAS Life Sciences Group, Inc. (NASDAQ: SLS). The author reserves the right to buy or sell these securities at any time without further notice. The author receives no direct compensation for the production of this content and maintains no business relationship with the company. The author has no access to non-public trial data, unblinded interim results, or internal company information. Readers should be aware that the author's financial interest in the company creates a potential for confirmation bias in the analysis and interpretation presented here.

Forward-Looking Statements & Risk: This article contains forward-looking statements regarding clinical trial outcomes and market potential. These statements are based on current publicly available information and are subject to significant risks and uncertainties. Actual results may differ materially. Investing in biotechnology and pharmaceutical securities involves a high degree of risk, including the potential for total loss of principal. Past performance is not indicative of future results. No reader should rely on this article as a basis for any investment decision.

Copyright: All content is the property of the author. This article may not be copied, reproduced, or published, in whole or in part, without the author's prior written consent.