SELLAS’s REGAL Trial Met Its Survival Goal: What the Topline Shows and What Is Still Missing

· 17 min read

SELLAS Life Sciences said on November 17, 2026 that its Phase 3 REGAL trial met its primary endpoint. Patients with acute myeloid leukemia who received galinpepimut-S (GPS) lived a median of 24.0 months, against 14.7 months for patients on the best available treatment chosen by their doctors. The hazard ratio for death was 0.59, with a two-sided p-value of 0.035, at the pre-specified analysis after 80 deaths among 126 patients (SELLAS press release, November 17, 2026). SELLAS plans to meet the FDA and to submit a Biologics License Application (BLA) in the first half of 2027. The confidence interval, the survival curves, the deaths in each arm and every secondary endpoint have not been published yet. The shares ended November 17 at $19.15, up 156% from Monday’s close of $7.47, after trading as high as $34.00 before the regular session opened (Yahoo Finance). After the close, SELLAS filed for a $150 million share offering.

The terms, briefly. Acute myeloid leukemia (AML) is a fast-growing blood cancer. REGAL enrolled patients whose leukemia had come back and who had reached a second remission (CR2, or CRp2 when platelet counts had not fully recovered) after more treatment, but who could not have a stem cell transplant from a donor, the main treatment that can cure AML once it has come back. GPS is a vaccine made of four peptides from the WT1 protein, which AML cells carry, meant to train T cells to attack the leukemia cells left after treatment. Overall survival (OS) measures how long patients live. A hazard ratio (HR) of 0.59 is an estimated 41% lower rate of death during follow-up than in the comparison arm, not a 41% gain in survival time. The comparison arm got the best available treatment (BAT): observation, a hypomethylating agent (azacitidine or decitabine), venetoclax and/or low-dose cytarabine. The trial was open-label: patients and doctors knew who got GPS.

Jump to the three scenarios ↓

Key points

  • The trial hit its primary endpoint. Median OS 24.0 months with GPS vs 14.7 months with BAT, HR 0.59, two-sided p=0.035, at the 80-death analysis the statistical plan set in advance. The figures are SELLAS’s topline and have not been peer-reviewed.
  • Both arms lived far longer than SELLAS’s 2022 design example. In 2022 SELLAS described significance as an HR of about 0.636, “for example” 12.6 vs 8.0 months. The control arm’s 14.7 months is close to double that example and far above the roughly 6 months SELLAS cited in January 2025 as the expected survival of similar patients. Part of the explanation may be the trial’s entry rules: patients needed an estimated life expectancy of more than 6 months. Parts of our earlier analyses, whose pooled-survival model came out too short, are contradicted (see below).
  • The evidence still to come decides how strong the result is. If the reported HR and p-value come from the same test, the 95% confidence interval is roughly 0.36 to 0.96 (our approximation), so the upper end is close to 1. The full data will show whether the curves separate clearly, how many patients were censored and how the control arm was treated.
  • The FDA path is not settled. SELLAS will meet the FDA to discuss “an accelerated filing strategy”, which it did not define, and plans a BLA in the first half of 2027. US law lets the FDA rely on one adequate and well-controlled trial plus confirmatory evidence, and an FDA draft guidance explains how; whether REGAL is enough is the FDA’s decision.
  • Cash, and the tools to raise more. SELLAS had $134.7 million on September 30 and used $22.4 million in operations in nine months; it says the cash covers at least the next twelve months. After the close on November 17 it filed for a $150 million underwritten share offering; its unused $150 million at-the-market program can be used after a 60-day lock-up. Five days after the January 2025 interim release, when it had $13.9 million in cash, it sold shares in a registered direct offering.

What SELLAS reported

Item Topline, November 17, 2026
Primary endpoint Overall survival from study entry: met
Median OS 24.0 months with GPS vs 14.7 months with BAT
Hazard ratio 0.59 (41% lower risk of death); confidence interval not reported
p-value 0.035, two-sided; the test and the pre-specified threshold not reported
Analysis After the pre-specified 80th death; SELLAS announced the 80th event on October 28, 2026
Patients 126 randomized 1:1 (ClinicalTrials.gov lists 127 enrolled); patients per arm not reported
Design Global, randomized, open-label; GPS maintenance vs investigator’s choice of BAT
Safety “Generally well tolerated”; treatment-related adverse events mostly grade 1 or 2, most often injection site reactions; no rates given
Balance Arms “well balanced for prognostic stratification factors”; few patients went on to a transplant, “equally distributed” between arms
Next steps FDA meeting; BLA submission planned for the first half of 2027; detailed results at an unnamed “upcoming major medical conference” and in a journal

Swipe the table sideways on a phone.

The analysis is the trial’s final one: SELLAS’s filings describe the 80-event analysis as “the final analysis of the REGAL study” (10-Q for the third quarter of 2026). An independent data monitoring committee reviewed the 60-death interim analysis in January 2025 and recommended that the trial continue without changes (SELLAS, January 23, 2025). Deaths 60 to 80 took about 22 months, from December 2024 to October 2026, much longer than SELLAS had expected: in January 2025 it said the 80th event would come that year.

What the topline leaves out

SELLAS said “comprehensive statistical evaluation of all secondary endpoints and additional pre-specified analyses” is under way. Until the full results are out, these are unknown:

  • The confidence interval of the hazard ratio, the test behind p=0.035 and the significance threshold left after the interim analysis.
  • Deaths and patients in each arm, the number censored (patients whose survival is counted only until they leave the trial or the data are cut off) and the median follow-up.
  • The Kaplan-Meier curves, which show whether the arms separate early or late and whether the gap holds, and survival rates at fixed times such as 12 or 24 months.
  • How the control arm was treated: how many patients got venetoclax, a hypomethylating agent, low-dose cytarabine or observation, and the stratification factors behind “well balanced”.
  • The secondary endpoints: leukemia-free survival, landmark survival, T-cell responses to WT1 and measurable residual disease.
  • A safety table with grade 3 or higher events, serious adverse events and discontinuations.
  • Subgroups, such as CR2 vs CRp2, age and region, and whether patients from mainland China, where SELLAS’s licensee 3D Medicines could enroll, are included.

Reading the numbers

An approximate confidence interval. SELLAS did not report one. If the HR of 0.59 and the p-value of 0.035 came from the same test, the implied 95% interval is roughly 0.36 to 0.96 (our arithmetic: a two-sided p of 0.035 is a z-score of 2.11, which puts the standard error of the log hazard ratio near 0.25). The upper end close to 1 means the data are consistent with a survival benefit much smaller than the point estimate, as well as a larger one. That is typical for a trial of 126 patients and is not a sign of a problem by itself, but it is why the curves and the censoring matter.

The HR and the p-value do not quite match a simple formula. A common rule of thumb for a trial with two equal arms puts the standard error of the log hazard ratio at 2 divided by the square root of the number of deaths: about 0.22 for 80 deaths. With that, an HR of 0.59 would give a p-value near 0.018, and p=0.035 would fit an HR near 0.62. The gap usually means the two numbers come from different methods, for example a stratified log-rank test for the p-value and a Cox model for the hazard ratio. SELLAS has not said which methods it used; the full presentation should show it.

The control arm. REGAL’s 14.7-month control median is the biggest surprise in the release. SELLAS’s original statistical plan assumed about 8 months for all patients combined; the revised plan of 2022 assumed 12.5 months for all patients combined, with an example of 12.6 vs 8.0 months for GPS vs BAT (Q3 2022 call transcript). In January 2025 SELLAS cited an expected survival of about 6 months for similar patients, and in the earlier pilot study of GPS, 15 patients treated at the same center without GPS, who were not randomized, lived a median of 5.4 months, counted from the time they reached the second remission. REGAL counts from study entry and admitted only patients with an estimated life expectancy of more than 6 months, enrolled within 6 months of reaching the remission (ClinicalTrials.gov). Those rules select patients who were doing better than the historical figures describe. Newer treatment such as venetoclax and the treatment after leaving the trial may also play a part, but the baseline data must answer that; the release does not explain the difference.

The size of the gap. The 9.3-month difference in medians is smaller than the 15.6 months (21.0 vs 5.4) of the pilot study that much of the investment case was built on, but it comes from a randomized trial. The pilot compared GPS-treated patients with 15 patients treated at the same center who were not randomized, and it counted survival from the second remission, not from study entry, so its medians are not directly comparable with REGAL’s.

The FDA path

SELLAS said it will meet the FDA “to discuss an accelerated filing strategy and advance preparations for a Biologics License Application (BLA) submission for GPS in the first half of 2027.” It did not say whether this means a rolling submission, priority review or something else. GPS has orphan drug and Fast Track designations in AML and a rare pediatric disease designation for pediatric AML (10-Q). Before the result, SELLAS wrote that it expected REGAL to be the basis of a BLA “subject to a statistically significant and clinically meaningful trial outcome and agreement with the U.S. Food and Drug Administration” (same source).

The legal standard is substantial evidence of effectiveness. US law allows it to be met with one adequate and well-controlled trial plus confirmatory evidence, and the FDA’s draft guidance on the standard, not yet final, explains how a sponsor can rely on “one adequate and well-controlled clinical investigation with confirmatory evidence” (FDA draft guidance, docket FDA-2019-D-4964). For REGAL, the questions are whether one open-label trial of 126 patients with p=0.035 is persuasive enough, and what could count as confirmatory evidence, such as the earlier pilot data or the T-cell and residual-disease results. Overall survival is an objective endpoint: knowing the treatment cannot change whether a patient is alive, unlike endpoints judged by investigators. The FDA will still look at dropouts, censoring and treatment after the trial in each arm, which an open-label design can affect.

A contrast from this week: on November 15, Summit said the FDA had rejected ivonescimab because its trial’s overall survival result was not statistically significant (our Summit analysis). REGAL’s primary endpoint was overall survival and, on SELLAS’s topline, it reached significance. That removes the issue that sank Summit’s application, but not the questions about the size and robustness of a small trial.

What our earlier analyses got right and wrong

BioRadar published three analyses of REGAL before the result: Why SELLAS Might Fail (July 2), What the Public Data Can and Cannot Tell Us (July 10) and the PETHEMA registry and the REGAL control arm (October 8). We compared their claims with the topline. Each article now carries a dated note.

Earlier claim Topline Status
The final analysis triggers at 80 deaths; an observed HR near 0.64 is roughly the significance boundary (July 10) Analysis after the 80th death; HR 0.59, significant Consistent
The sponsor’s 8-month control assumption may be pessimistic, but public data do not establish a 12 to 15 month control median (July 10) Control median 14.7 months The control lived longer, as the article allowed
Pooled median survival (both arms) about 15.5 months, working range 13 to 17, model estimates from about 6.5 to 17.7 (July 10, repeated October 8) No pooled median reported; with arm medians of 14.7 and 24.0 it lies between them, about 18.6 months under a simple exponential model (our arithmetic) Low end contradicted; central estimate likely too low
Scenario tables paired a long control median with a weak effect, for example a 14-month control with an HR near 0.92 and power of about 7% (July 10) 14.7-month control with HR 0.59 Contradicted: the pairing came from the pooled-survival fit, which underestimated survival
“Efficacy remains unproven” (July 10) Primary endpoint met Superseded by the topline, pending full data
Failure mode 1: REGAL misses statistical significance (July 2) p=0.035 Did not happen
Failure mode 3: the control arm lives far longer than expected and GPS loses significance (July 2) Control 14.7 months, GPS still significant The first half happened, the second did not
A good result may be followed by equity financing; $150 million ATM on file (July 2) $150 million underwritten offering filed after the close on November 17; ATM still unused that day Happened on the day of the result

Swipe the table sideways on a phone.

The main lesson is about method. The July model fitted one survival curve to the pooled death counts and then split it into arms under assumptions. The pooled counts constrained the arms too tightly: when the control arm lives long, the model needed GPS patients to live only a little longer to match the slow pace of deaths. In reality both arms lived long and GPS patients longer still. A pooled median near 6.5 months was in fact never possible: in January 2025 SELLAS said fewer than half of the patients had died about 10 months after enrollment ended, which already put the pooled median above roughly 10 months unless many patients had left the trial early.

Money and dilution

  • Cash: $134.7 million on September 30, 2026. Operations used $22.4 million in the first nine months of 2026, about $7.5 million a quarter; at that pace the cash would last more than four years, but a BLA and a launch would raise spending (our arithmetic). The 10-Q says the cash covers planned operations “for at least the next twelve months” and that SELLAS “may pursue additional capital resources” (Q3 results; 10-Q).
  • Share count: 203.9 million shares on November 11, 2026, up from 153.1 million at the end of 2025. Most of the increase came from warrants: 50.4 million were exercised in the first nine months at an average of $1.69, bringing in $85.1 million. On September 30, about 8.1 million warrants were outstanding, at $1.41 to $2.00 (about 325,000 were exercised in October and November), plus 2.7 million options and 3.3 million restricted stock units.
  • Offering, November 17: at 4:14 p.m. Eastern Time, after the close, SELLAS filed a preliminary prospectus supplement for a $150 million underwritten offering of common stock, with J.P. Morgan and TD Cowen as joint book-running managers and a 30-day option for the underwriters to buy up to $22.5 million more. The price had not been set when we updated this article. At the $19.15 close, $150 million would mean about 7.8 million new shares, 3.8% of the 203.9 million outstanding, or about 9.0 million (4.4%) with the option; a lower price would mean more shares (our arithmetic). SELLAS and its executive officers and directors agreed, with limited exceptions, not to sell shares for 60 days without the underwriters’ consent (preliminary prospectus supplement).
  • Ready to sell: an automatic shelf registration filed in March 2026 and an at-the-market (ATM) program with TD Cowen for up to $150 million, all of it still available on November 17; the prospectus supplement says the ATM can be used after the 60-day lock-up ends or is waived. ATM sales do not need a new filing on the day; they show up in later reports. At the $19.15 close, the full $150 million would mean about 7.8 million more shares (our arithmetic).
  • Precedent: on January 28, 2025, five days after the interim analysis release, SELLAS agreed a registered direct offering with one institutional investor at a combined price of $1.27 per share and warrant (8-K, January 29, 2025). It then had $13.9 million in cash (end of 2024), against $134.7 million now.
  • Licenses: GPS is licensed from Memorial Sloan Kettering Cancer Center, which could receive up to $17.4 million in milestones per licensed product, if they are reached, and a tiered royalty in the mid-single digits on sales (10-K for 2025). In Greater China, SELLAS has licensed GPS to 3D Medicines, which could owe SELLAS up to $191.5 million in further milestones (10-Q); in July 2026 an arbitrator dismissed SELLAS’s claims against 3D Medicines in a dispute over milestone payments.

On the BioRadar Dilution Watch, SLS scores 39 of 100 (moderate) as of November 17: low cash burn against a large ATM program and 43% share growth in a year.

How the stock moved

SLS closed at $7.47 on Monday, November 16, the session before the release. That was 44% below the $13.27 close the day before our July 2 analysis and 28% below the $10.38 close on October 28, when SELLAS announced the 80th death. The shares had risen from $0.99 on January 23, 2025, the day of the interim analysis, to $15.46 on August 21, 2026, their highest close before the result (Yahoo Finance and BioRadar daily closes). At $7.47 and 203.9 million shares, the company was valued at about $1.5 billion (our arithmetic).

On November 17 the first trades after the release came in pre-market trading from 8:30 a.m. Eastern Time, at prices as high as $34.00. In the regular session the shares traded between $16.72 and $24.89 and closed at $19.15, 156% above Monday’s close. About 113.9 million shares changed hands, a number equal to 56% of the shares outstanding (Yahoo Finance; our arithmetic). At $19.15 the company was valued at about $3.9 billion (our arithmetic). At 4:14 p.m., after the close, SELLAS filed for a $150 million share offering (see Money and dilution above).

SELLAS Life Sciences (SLS) daily closing prices, November 2024 to November 17, 2026The shares closed at $0.99 on January 23, 2025, when SELLAS said the REGAL trial would continue after its interim analysis, and five days later it sold shares and warrants at $1.27. They rose to $15.46 on August 21, 2026, their highest close before the result, closed at $10.38 on October 28, 2026, when SELLAS announced the 80th death that triggered the final analysis, and fell to $7.47 on November 16, 2026. On November 17, the day SELLAS reported that REGAL met its survival goal, they closed at $19.15.SELLAS (SLS): two years through the REGAL resultDaily closes, November 2024 to November 17, 2026↻ Replay$0$5$10$15$20Jan ’25Jul ’25Jan ’26Jul ’26Jan 23, 2025: interim analysis,trial continues; offering at $1.27Aug 21, 2026: highest closebefore the result, $15.46Oct 28, 2026: 80th death,final analysis triggeredNov 17, 2026: REGAL topline,closed at $19.15Source: Yahoo Finance closes; SELLAS SEC filings. Not investment advice.SELLAS Life Sciences (SLS) daily closing prices, November 2024 to November 17, 2026The shares closed at $0.99 on January 23, 2025, when SELLAS said the REGAL trial would continue after its interim analysis, and five days later it sold shares and warrants at $1.27. They rose to $15.46 on August 21, 2026, their highest close before the result, closed at $10.38 on October 28, 2026, when SELLAS announced the 80th death that triggered the final analysis, and fell to $7.47 on November 16, 2026. On November 17, the day SELLAS reported that REGAL met its survival goal, they closed at $19.15.SELLAS (SLS): two yearsthrough the REGAL resultDaily closes, November 2024 to November 17, 2026↻ Replay$0$5$10$15$20Jan ’25Jul ’25Jan ’26Jul ’26Jan 2025: interim,offering at $1.27Aug 2026: $15.46,high before resultNov 17 topline: $19.15Source: Yahoo Finance closes; SEC filings.Not investment advice.
SLS daily closes from November 18, 2024 to November 17, 2026, the day of the REGAL topline. Source: Yahoo Finance; events from SELLAS SEC filings.

Live tracker: SLS since this analysisSubscribers

The checkpoints this analysis set out, the price move since Nov 17, 2026 and every SEC filing since, updated as filings land.

  • 🔒 StatusFull REGAL results at a medical conference: confidence interval, Kaplan-Meier curves, censoring and control-arm treatment
  • 🔒 StatusSELLAS's account of the FDA meeting on an "accelerated filing strategy"
  • 🔒 StatusBLA submission, planned for the first half of 2027
  • 🔒 StatusFinancing: ATM sales or a new offering
  • 🔒 Status3D Medicines and patients from mainland China

The price move since the analysis and each checkpoint's status are in the tracker.

The analysis above stays free. The tracker is part of the trial and the subscription.

Three scenarios from here

This section is analysis, not a forecast. It describes paths the disclosed facts make possible and what each would require. We do not assign probabilities.

1. The full data hold up and the FDA accepts REGAL as the basis for a BLA

The detailed results show curves that separate and stay apart, censoring that is similar in both arms, a control arm treated as the protocol allowed and consistent results across subgroups. The FDA agrees that REGAL, with the earlier data as support, can carry an application, and SELLAS files in the first half of 2027. With priority review the FDA’s goal is eight months from submission for a new biologic (six months after a 60-day filing period), so a filing in the first half of 2027 would put the goal date between about September 2027 and February 2028 (our arithmetic from the PDUFA goals). What to look for: the conference presentation and its confidence interval, SELLAS’s account of the FDA meeting, the BLA submission and whether the FDA grants priority review.

2. The FDA wants more before or after approval

The FDA accepts that REGAL is positive but asks for more: additional analyses or longer follow-up before filing, a second trial or other confirmatory evidence, or answers on the open-label design, the control treatment or censoring. The timeline moves out and the spending goes up, which makes financing more likely. What to look for: words such as “confirmatory”, “additional trial” or “additional data” in SELLAS’s updates after the meeting, changes to the first-half 2027 filing target and new trial registrations.

3. The full data weaken the case

The detailed results show a confidence interval reaching close to 1, curves that cross or separate only late, more censoring in one arm, a control arm that received less effective treatment than expected, or a benefit limited to a subgroup. The FDA then sees one small open-label trial with a fragile result. A BLA could still be filed, but the outcome would be less certain, and an advisory committee meeting becomes more likely. What to look for: the same presentation as in the first scenario, the number of deaths and patients in each arm, landmark survival rates and the treatment patients received after the trial.

What to watch

  • The detailed results: the conference, the confidence interval, the Kaplan-Meier curves, deaths and censoring in each arm, the control-arm treatment mix and the secondary endpoints.
  • SELLAS’s account of the FDA meeting and what “accelerated filing strategy” turns out to mean.
  • The BLA submission, planned for the first half of 2027, and the FDA’s acceptance and review designation.
  • Financing: the price and final size of the offering filed on November 17, and any sales under the $150 million ATM after the 60-day lock-up.
  • Any update on 3D Medicines and patients from mainland China.

Bottom line

REGAL met the endpoint that matters most, overall survival, in a randomized trial, which is the evidence GPS lacked until now. The topline is still only a summary: the confidence interval, the curves and the treatment of the control arm, which lived much longer than SELLAS’s design example, will show how strong the result is. The next decisions are the FDA’s view of a single small open-label trial and the price of the $150 million share offering SELLAS filed for after the close; it had $134.7 million in cash on September 30.

FAQ

What were the REGAL trial results?

On November 17, 2026, SELLAS said REGAL met its primary endpoint: median overall survival was 24.0 months with galinpepimut-S vs 14.7 months with the best available treatment, a hazard ratio of 0.59 with a two-sided p-value of 0.035, at the pre-specified analysis after 80 deaths among 126 patients.

Is galinpepimut-S approved?

No. GPS is investigational. SELLAS plans to meet the FDA and to submit a Biologics License Application in the first half of 2027. Approval depends on the FDA’s review of the full data.

What is still unknown about REGAL?

The confidence interval of the hazard ratio, deaths and censoring in each arm, the Kaplan-Meier curves, the treatment the control arm received, the secondary endpoints, a detailed safety table and subgroup results. SELLAS said the details will be presented at an upcoming medical conference.

Why is the control arm’s 14.7 months notable?

It is close to double the 8 months SELLAS used as a design example in 2022 and far above the roughly 6 months SELLAS cited in January 2025 as the expected survival of similar patients. REGAL admitted only patients with an estimated life expectancy of more than 6 months and counts survival from study entry, which selects patients doing better than historical figures; the baseline characteristics and the control-arm treatment data should show how much this explains.

Sources

Sources read November 17, 2026. Prices are Yahoo Finance and BioRadar daily closes. The confidence interval, the pooled median and the cash-runway figures are our own approximations from the disclosed numbers, not company data. This article is for information only and is not investment advice. The author held SLS shares (a long position) when this was written.

BioRadar analyses take no bull or bear side, make no buy or sell calls and are never paid for. Read our editorial policy.

This data is for informational purposes only, not investment advice. BioRadar does not provide buy/sell recommendations. Past performance does not guarantee future results. Always do your own due diligence.