The BAT Patient - In Detail

[mistergern]

I have worked directly with Claude to identify all of the known information about the IMPactMF BAT patients. There may be info we have missed but I have attempoted to include all the info that has been reported in publications as well as here and on Seeking Alpha. The following is the current exhaustive and to a degree definitive profile for IMPactMF BAT patients. Please feel free to correct anything that is wrong or misleading - we have attributed confidence levels to the info we are providing. This will continue for several chats:

How to Read This Document
This profile assembles, in one place, everything that can be responsibly said about the Best Available Therapy (BAT) control arm of the IMpactMF Phase 3 trial — the comparator against which imetelstat’s survival benefit will be judged. The behavior of this arm is the single most consequential unknown in the entire investment and clinical thesis, because the trial’s primary endpoint (overall survival) is decided by the gap between the two arms, not by either arm alone.
Three rules govern every claim in this document:
• Sourced vs. reconstructed vs. unknown. Every factual claim is tagged with a confidence level. Design facts from registries and the trial record are marked High. Estimates we have derived from assumptions are explicitly labeled as reconstructions. Items we cannot know are placed in the dedicated “What We Do Not Know” section and left empty rather than filled with speculation.
• Anchor discipline. Survival figures in myelofibrosis are meaningless without specifying the timepoint they are measured from. This document always distinguishes survival measured from ruxolitinib initiation versus from ruxolitinib failure/discontinuation — the latter being the trial-relevant anchor.
• No probability of success is asserted. This is a factual substrate, not a forecast. It is designed to let any reader build their own estimate on clean ground, and to retire claims that rest on confused or fabricated inputs.

A note on verification
Citations marked “verify” are drawn from secondary compilation and AI-assisted literature retrieval. They correspond to real, traceable papers but the specific figures should be confirmed against the primary source before being quoted as definitive. This caution exists because AI tools have been observed to state plausible-looking but inaccurate citations with false confidence.
 

[mistergern]

 
1. Executive Summary
The IMpactMF BAT arm is not “all patients who failed ruxolitinib.” It is a deliberately constructed, poor-prognosis subset: intermediate-2 or high-risk myelofibrosis, genuinely relapsed or refractory to a JAK inhibitor, not a candidate for further JAK inhibition, with measurable splenomegaly or symptoms, and limited to non-JAK-inhibitor therapies. Understanding this construction is the key to interpreting every survival figure in circulation.
The central findings of this synthesis:
• The best-matched estimate of imetelstat’s effect is HR 0.512 (imetelstat 30.7 months vs. matched real-world BAT 15.4 months; Kuykendall 2026). This is a cross-study, non-randomized comparison, and the effect has regressed from an earlier 0.35 estimate as the control cohort matured.
• Contemporary real-world post-ruxolitinib survival has risen sharply (to 34–40 months in recent cohorts), but this “era drift” is driven largely by access to newer JAK inhibitors and clinical trials — exactly the therapies IMpactMF excludes from its BAT arm.
• The era drift decomposes into two mechanisms with opposite implications: drug-access drift (design-excluded, does not enter the BAT arm) and ruxolitinib-regimen drift (earlier start, shorter duration — not excluded, but a both-arms effect that is largely neutral to the hazard ratio).
• The interim timing is itself informative. The interim fires at ~112 deaths, expected H2 2026. The fact that the trial has taken this long to accrue events constrains the feasible combinations of BAT and imetelstat survival, and is consistent with a durable imetelstat arm, elevated BAT survival, or both.
• The decisive requirement is separation, not reproduction. Imetelstat reproducing its Phase 2 absolute survival is not sufficient; the trial requires the survival curves to separate by enough to cross a stringent interim boundary (~p≤0.01). If BAT survival has risen, imetelstat must do proportionally better, not merely as well.

Best reconstructed estimate of in-trial BAT median OS: approximately 18–22 months (plausible range 14–28), explicitly an estimate, not an observation. The single most important fact about this number is that it is blinded and will not be known to anyone outside the Independent Data Monitoring Committee until the interim readout.

[mistergern]

2. Trial Design Parameters Governing the BAT Arm
These are the structural facts of the trial that define who is in the BAT arm and what they may receive. Except where noted, these are design facts of high confidence, though some derive from trial-registry and secondary sources and are flagged accordingly.
2.1 Patient eligibility (entry into the trial)
Criterion Specification Confidence
Disease risk DIPSS intermediate-2 or high-risk myelofibrosis (primary MF, post-PV MF, or post-ET MF) High
Prior therapy Relapsed or refractory to a JAK inhibitor (ruxolitinib); genuine failure, not early elective switch High
JAKi candidacy Not a candidate for further JAK inhibitor therapy at screening High
Performance status ECOG 0–2 Med-High
Blast count Peripheral and marrow blasts < 10% (excludes accelerated/blast phase) Med-High
Disease burden Measurable splenomegaly and/or active symptoms High
Target enrollment ~320 patients High
Randomization 2:1, imetelstat : BAT (≈213 imetelstat, ≈107 BAT) High
Stratification By risk score (int-2 vs. high) and baseline platelet count Medium

2.2 What “BAT” is permitted to be
BAT is investigator-selected, and critically it is restricted to non-JAK-inhibitor therapy. Permitted options include:
• Hydroxyurea
• Danazol
• Thalidomide and analogs (e.g., lenalidomide)
• Interferon
• Hypomethylating agents (HMAs)
• Chemotherapy / radiotherapy
• Supportive care including transfusions

2.3 What is excluded from BAT — the design feature that matters most
The BAT arm explicitly excludes the newer JAK inhibitors (fedratinib, pacritinib, momelotinib) and concurrent enrollment in other clinical trials. [Confidence: Med-High] (Per trial registry and secondary compilation; the exact protocol language should be verified.) This singleexclusion is the mechanism by which the trial attempts to wall its control arm off from the modern survival improvements seen in real-world cohorts.

2.4 Crossover provision
BAT patients may cross over to imetelstat after protocol-defined disease progression (reported as a ≥25% increase in spleen volume, or palpable increase in splenomegaly, after ≥6 months of BAT), with sponsor approval. [Confidence: Medium]
Because the primary endpoint is overall survival analyzed by intent-to-treat (ITT), a patient who crosses over remains counted in the BAT arm. This preserves statistical integrity but means any effective post-progression therapy (including off-protocol access) can lengthen measured BAT-arm survival.

[mistergern]

3. What the BAT Arm Is NOT — Avoiding the Central Confusion
Most public confusion about IMpactMF stems from comparing the trial’s BAT arm to real-world myelofibrosis cohorts that are not comparable. The differences are decisive:
Dimension Real-world post-Rux cohorts IMpactMF BAT arm
Newer JAK inhibitors Freely used (fedratinib, pacritinib, momelotinib) — a major survival driver Excluded by protocol
Clinical trial access Patients often enroll in other trials post-failure Excluded by protocol
Population Heterogeneous; may include lower-risk, int-1, non-failed patients Strict int-2/high, genuine R/R only
Blast phase Often included (drags survival down) Excluded (blasts <10%)
Measurement Variable anchor (often from Rux initiation) From randomization (post-failure)

The most common error in circulation
The frightening real-world figures of 34–40 month survival do NOT describe the IMpactMF BAT arm.
Those numbers come from cohorts with access to newer JAK inhibitors and trials. IMpactMF’s BAT arm is built specifically to exclude those therapies. Applying the 34–40 month figure to the trial’s control arm is the single most frequent analytical mistake in public discussion of this trial.

4. The Survival Evidence, Sourced and Anchor-Labeled
Every figure below is tagged by the timepoint it is measured from. The trial-relevant anchor is from ruxolitinib failure/discontinuation, because that is when IMpactMF patients are randomized.
4.1 The matched comparison (most trial-relevant)
Kuykendall et al., 2026 (ASCO/EHA), updated post hoc matched analysis:
Group Median OS (95% CI) Notes
Imetelstat (IMbark P2) 30.7 mo (25.5–36.9) 59 patients, 9.4 mg/kg, median f/u 46.2 mo
Matched RW BAT 15.4 mo (13.1–30.5) 54 closely-matched patients, median f/u 48.2 mo
Hazard ratio 0.512; p=.003 Unweighted; ATO & sIPTW concordant

Anchor: from JAK inhibitor discontinuation to death. [Confidence: High] (primary source).
Critical caveat on provenance: The numerator (imetelstat, 30.7 mo) is real Phase 2 trial data. The denominator (BAT, 15.4 mo) is a retrospective, single-center (Moffitt) real-world cohort matched to IMbark eligibility — not a randomized control. Matched real-world comparators systematically flatter the experimental arm, and this estimate already regressed from HR ~0.35 (Kuykendall 2021) to 0.512 (2026) as the control cohort grew. The true randomized HR is, on priors, likely somewhat higher (less favorable) than 0.512.

4.2 The historical post-failure literature (the strict-BAT analogs)
These cohorts measure survival from ruxolitinib failure/discontinuation and approximate the strict, poor-prognosis population. All figures should be verified against the primary source before being quoted as definitive.
Source (verify) N Median OS Post-failure context Conf.
Newberry 2017 (MD Anderson) 56 14 mo Pre-newer-JAK era High*
Kuykendall 2018 (Moffitt) 64 13 mo Salvage in 66%; some alloHSCT High*
Palandri 2020 (Europe) 218 13.2 mo (8.0–22.7) Incl. blast phase; chronic-phase subset 27.5 mo High
Mascarenhas 2020 (US claims) 290 11.1 mo (8.4–14.5) Claims/EHR/SEER-Medicare High*
Schain 2019 (Sweden/Norway) 71 16.0 mo (rel. surv.) Population registry High*
Kuykendall 2022 matched BAT 38 12 mo Matched to MYF2001 eligibility (closest analog) High*

* Median OS and anchor are high-confidence from accessible abstracts; 95% confidence intervals were not extractable for several and should be verified.

Synthesis of the historical cluster: correctly anchored post-failure survival in strict R/R populations clusters at 11–16 months. The cleanest single analog — the Moffitt cohort matched to imetelstat trial eligibility — is 12 months. [Confidence: High]

4.3 The chronic-phase wrinkle (Palandri)
Palandri 2020 found its overall 13.2-month figure was dragged down by blast-phase patients (3.9 mo). The chronic-phase discontinuers — which is what IMpactMF enrolls, since it excludes blasts ≥10% — had median OS of 27.5 months. [Confidence: High] This is the strongest single data-based argument that IMpactMF’s BAT arm could exceed the 12–16 month benchmark.
Important applicability limit: Palandri’s chronic-phase group still included ruxolitinib rechallenge and investigational agents — therapies IMpactMF’s BAT arm excludes. So the IMpactMF-applicable figure likely sits between the matched-Moffitt 12 and the Palandri chronic-phase 27.5.


 

[mistergern]

5. The Era-Drift Decomposition — The Analytical Crux
Contemporary real-world post-ruxolitinib survival has risen substantially. The Kuykendall 2026 registry data shows the magnitude and — crucially — the causes:
Cohort by diagnosis Dx→Rux (mean) Rux duration (mean) Median OS
Before 2016 (N=54) 46.7 mo 30.4 mo 15.4 mo
After 2016 (N=42) 10.5 mo 13.5 mo 34.2 mo
Before 2019 (N=69) 39.7 mo 26.1 mo 16.6 mo
After 2019 (N=27) 8.2 mo 15.0 mo 39.8 mo
The authors attribute the improvement to “shorter time from diagnosis to start of RUX and earlier switch to next-line JAKi or clinical trial.” [Confidence: High] This phrase contains two distinct mechanisms that must be separated:

5.1 Mechanism A — Drug-access drift (DESIGN-EXCLUDED)
Earlier switching to newer JAK inhibitors and clinical trials is the larger driver of the rise to 34–40 months. IMpactMF’s BAT arm excludes exactly these therapies. Therefore this component of the drift does not populate the trial’s control arm. [Confidence: Med-High]
5.2 Mechanism B — Ruxolitinib-regimen drift (NOT excluded, but HR-neutral)
Patients now start ruxolitinib earlier (46.7→8.2 months from diagnosis) and stay on it for less time (30.4→15.0 months) before failing. They therefore arrive at the point of failure healthier in disease-time. This modern regimen is standard practice and cannot be excluded — IMpactMF patients were treated this way. [Confidence: Medium]
However, this is a both-arms effect: the imetelstat arm patients also came through modern ruxolitinib practice and arrive equally healthier. Under proportional hazards, a lift applied to both arms cancels in the hazard ratio. It raises absolute survival in both arms while leaving the ratio — which decides the trial — approximately unchanged.

5.3 Mechanism C — The ITM/off-protocol crossover leak (the genuine residual)
Raised by the contributor “biopearl” and analytically sound: a BAT patient who leaves the study and obtains a newer JAK inhibitor off-protocol is still counted in the BAT arm under ITT (unlike a transplant patient, who may be censored). Such patients can form a longer tail on the BAT survival curve, partially importing the drug-access drift the design tried to exclude. [Confidence: Medium] This is the one channel by which Mechanism A can leak in, and it argues for holding the BAT estimate toward the higher end of its range.

The decomposition in one sentence
Of the era drift: the part that would compress the hazard ratio (drug access) is largely design-excluded; the part that leaks through (ruxolitinib timing) is a both-arms effect that is roughly HR-neutral; and the one genuine residual is off-protocol crossover counted under ITT.

6. The Disease-Modification Question and Proportional Hazards
Whether the both-arms RUX-timing lift is truly HR-neutral depends on whether imetelstat’s effect is proportional across patient health states. This is a genuine biological uncertainty.
The HR-neutral conclusion holds if and only if imetelstat applies a constant multiplicative effect on the hazard regardless of how healthy the population is (proportional hazards). Two scenarios diverge:
• If imetelstat is palliative-like (benefit concentrated in sicker patients): a healthier modern population gives it less room to outperform, and the both-arms lift would compress the HR. Bearish.
• If imetelstat is genuinely disease-modifying (the stated mechanism — fibrosis reversal, reduction in mutant allele burden): earlier intervention on less-advanced disease is where it has the most leverage. Healthier-at-failure patients have more intact marrow to preserve, so the effect is preserved or amplified. Neutral-to-favorable.
The documented mechanism — dose-dependent bone-marrow fibrosis reversal and VAF reduction correlating with survival — supports the disease-modifying interpretation. [Confidence: Medium] If correct, the modern ruxolitinib regimen delivers patients to imetelstat at the stage where it works best, which is consistent with the clinical view (attributed to Dr. Mascarenhas) that patients should be enrolled before their bone marrow “burns out.”
This remains a biological hypothesis, not a settled fact, and is one of the load-bearing assumptions of any bullish reading.
 

[mistergern]

7. The Timing Constraint — What the Interim Date Implies
The interim analysis is event-driven, not calendar-driven: it fires when approximately 112 deaths (35% of ~320 patients) have accrued, currently projected for H2 2026. The trial has been extended relative to earlier guidance (which projected an interim in early 2026 and final in early 2027). Both the trigger and the extension carry information.
7.1 The enrollment curve (backloaded)
Per the contributor “biopearl,” enrollment was backloaded. Observation periods must be measured to the interim (~Sep 2026), which is ~12 months after enrollment completion — a correction that several analyses initially missed.
Cohort (enrolled) Patients Follow-up by 9/26 Role at interim
Apr 2021–Dec 2023 164 (50%) ~33–65 mo Primary death engine
Dec 2023–Dec 2024 98 (30%) ~21–33 mo Meaningful contributor
Dec 2024–Jul 2025 49 (15%) ~14–21 mo Modest contributor
Jul 2025–Sep 2025 16 (5%) ~12–14 mo Minimal
Confidence on enrollment breakdown: [Confidence: Low-Med] (from a secondary contributor; should be verified against trial records). The K-M survival shape is front-loaded (steep early deaths, then a plateau), meaning late enrollees contribute deaths slowly and the interim is dominated by the mature early cohorts.

7.2 The constraint and its implication
For only ~112 deaths to have accrued by H2 2026, death accrual must be slow. With BAT largely spent by the interim, slow total accrual requires a durable imetelstat arm. This produces a feasibility relationship: any (BAT median, imetelstat median) pair must jointly reproduce ~112 deaths at the interim date.
Reconstructed feasible pairs (exponential survival, backloaded follow-up — illustrative, not precise):
BAT mOS BAT deaths by 9/26 Implied Imet mOS to fit Implied HR
14 mo ~87 ~140 mo ~0.06
18 mo ~77 ~110 mo ~0.16
22 mo ~71 ~95 mo ~0.21
26 mo ~64 ~84 mo ~0.27
30 mo ~58 ~73 mo ~0.32
Reconstruction. Two readings of this table are both honest: (1) every feasible pair with BAT ≤30 implies a low HR (halt-favorable); but (2) the required imetelstat medians (73–140 months) are biologically extreme, which signals either that BAT is running higher than these rows assume, or that the interim will slip past H2 2026, or that the exponential model overstates accrual. The tension is genuine and unresolved.

7.3 Separation, not reproduction
The decisive principle (articulated by “biopearl”): the trial requires the survival curves to separate enough to cross the boundary — not merely for imetelstat to reproduce its Phase 2 absolute survival. If BAT survival has risen, imetelstat must do proportionally better, not just as well. A drug that merely matches its Phase 2 number against a stronger BAT arm could fail to separate.

8. Reconstructed In-Trial BAT Estimate
Pulling the sourced evidence and the decomposition together yields a reasoned estimate — explicitly an estimate, not an observation.
Scenario BAT mOS Basis
Low / strict-historical 12–16 mo Matched analogs (Moffitt 12, Mascarenhas 11.1, Newberry 14, Schain 16)
Central / modern-adjusted 18–22 mo Historical + RUX-timing lift + chronic-phase eligibility
High / drift-and-leak 24–28 mo Strong RUX-timing effect + off-protocol crossover leak
Above 30 mo <15–25% likely Would require the design exclusions to fail substantially; the 34–40 mo figures require excluded therapies

Best reconstructed central estimate: 18–22 months, plausible range 14–28. [Confidence: Low-Med] Probability that in-trial BAT exceeds 30 months: estimated 15–25%, because the cohorts achieving 30+ months did so via the newer-JAKi/trial access that IMpactMF excludes. [Confidence: Low]
 

 

[mistergern]

9. What We Do Not Know
This section is the heart of the document. Each item below is a genuine unknown. None is filled with speculation. Where an item is unknowable until the readout, it is marked as such. This is the boundary that separates analysis from the fabrication that has circulated elsewhere (including AI-generated claims to “analyze the confidential Statistical Analysis Plan,” which no one outside the trial can do).

9.1 The blinded outcomes (unknowable until interim/readout)
Effective in-trial BAT median OS. The actual survival of the control arm. Known only to the IDMC.
The death split between arms at the interim. How many of the ~112 deaths are BAT vs. imetelstat — this is the observed hazard ratio. Blinded.
Imetelstat’s true randomized effect. The single load-bearing unknown. All estimates rest on a 59-patient Phase 2 post hoc vs. a non-randomized comparator. IMpactMF is the first randomized test.
Whether the survival curves have separated, and by how much. Blinded.

9.2 The trial machinery (unpublished, not knowable from outside)
The exact interim efficacy boundary. Estimated at ~p≤0.01 from design parameters, but the actual alpha-spending function in the Statistical Analysis Plan is not public.
The precise event count and timing of the interim. “~112 deaths, H2 2026” is a projection; the actual date depends on accrual.
Pre-specified subgroup analyses (e.g., by line of therapy). Whether these exist and what they would show is unknown. Claims to the contrary are fabrication.
IDMC composition and decision threshold behavior. The committee may continue even if the boundary is crossed; their disposition is unknown.

9.3 The population specifics (partially knowable when baseline data releases)
The actual baseline health of enrolled BAT patients vs. historical cohorts (ECOG, DIPSS mix, transfusion dependence).
The off-protocol crossover rate — how many BAT patients leave and access newer JAKis, the size of biopearl’s ITT leak.
The line-of-therapy composition — how many patients failed one vs. multiple prior JAK inhibitors.
The geographic enrollment split across ~172 global sites, which affects the blend of era-drift exposure in BAT.

9.4 The biological assumptions (cannot be confirmed pre-readout)
Whether proportional hazards holds — determines whether the both-arms RUX-timing lift is truly HR-neutral.
Whether imetelstat’s effect is preserved/amplified in healthier patients (disease-modification) or diminished (palliative-like).
How much the matched HR of 0.512 will regress under true randomization.

10. Research Legs — Where More Viable Information May Be Found
These are concrete, actionable avenues that could convert items from the “unknown” column into sourced estimates. Each is ranked by potential value.
10.1 Highest value
1. Matched global registry reconstruction. Pull non-US/non-European MF registries (Japanese, Korean, Chinese, Latin American) with post-2019 entry, matched to IMpactMF eligibility (DIPSS int-2/high, genuine R/R, no newer-JAKi access), measuring OS strictly from ruxolitinib failure. These low-era-drift geographies are the best proxy for the IMpactMF BAT floor. (Proposed by biopearl.)
2. Primary-source verification of the Kuykendall 2026 full publication. Confirm the matched-cohort treatment composition, the chronic-phase subset details, and whether any IMpactMF-applicable subgroup is reported.
3. Palandri 2020 chronic-phase treatment mix. Determine how much of the 27.5-month chronic-phase survival depended on rux rechallenge / investigational agents that IMpactMF excludes — this calibrates the upper bound of the in-trial BAT estimate.
10.2 Medium value
4. KOL commentary post-EHA 2026. Track whether Dr. Mascarenhas and other KOLs characterize the era drift as access-driven (excluded) or stage-driven (both-arms). Note: investigators are blinded to the trial’s outcome data; their value is clinical judgment on population condition, not survival numbers.
5. Geron disclosures on interim timing. Any tightening or slippage of the H2 2026 interim date is informative about accrual, and (via the constraint) about both-arm survival.
6. Trial registry / protocol amendments. Verify the exact BAT exclusion language, crossover criteria, and any enrollment-size or event-trigger changes against ClinicalTrials.gov / ANZCTR.
7. Fedratinib/momelotinib/pacritinib post-failure OS data. Quantify how much newer-JAKi access lengthens post-failure survival, to bound the size of the Mechanism-A drift that IMpactMF excludes.
10.3 Lower value / caution
8. Avoid: AI-generated reconstructions of the confidential SAP. These cannot be valid — the document is not public — and circulating them damages the credibility of legitimate analysis.
9. Avoid: applying real-world 34–40 month figures to the BAT arm without the access decomposition. This is the central error this document exists to correct.

[mistergern]

11. Synthesis of Community Input
This document integrates analysis circulating in the investor/clinical community. Each contribution is assessed on merit:
Source Assessment
TrailBlazer7 Correct that the Phase 3 bar (HR target ~0.60) sits above the matched effect (0.51), and that the design excludes the drug-access component of era drift. The “hermetically sealed BAT” framing overstates it — RUX-timing drift and off-protocol crossover leak through — but the core point survives, because those leaks are largely HR-neutral or modest.
biopearl The most rigorous contributor. Key insights: (1) “separation, not reproduction” — the trial needs curve separation, not absolute Phase 2 replication; (2) the RUX-regimen drift happens pre-enrollment and is not design-excluded; (3) the ITT off-protocol crossover leak; (4) the front-loaded K-M shape. All sound.
mistergern Credited (by biopearl) for projection work using both historical and “new era” BAT times. The constraint-based timing analysis is the right method.
AI outputs (various) Mixed. Sound on bounded math and literature retrieval when forced to show work and tag confidence. Unreliable when producing fluent synthesis, advocacy prose, or — in one case — fabricated “analysis” of the confidential SAP. AI conclusions should be treated as fallible drafts requiring human verification, not authority.
 
12. The Honest Bottom Line
On the BAT arm specifically:
• The IMpactMF BAT arm is a deliberately poor-prognosis, non-JAK-inhibitor population. Its closest historical analog (matched Moffitt) is 12 months; its best reconstructed in-trial estimate is 18–22 months after accounting for modern ruxolitinib practice.
• The frightening 34–40 month real-world figures do not apply to it, because they are driven by newer-JAK-inhibitor and clinical-trial access that the protocol excludes.
• The era drift that DOES reach the BAT arm (ruxolitinib timing) is a both-arms effect and is largely neutral to the hazard ratio that decides the trial.
• The one genuine residual is off-protocol crossover counted under ITT, which argues for the higher end of the BAT range.

On what this means:
The trial will be decided by separation between the arms, measured as a hazard ratio against a stringent interim boundary. The best-matched estimate of that effect is HR ~0.51, an optimistic-but-credible anchor that is likely to regress somewhat under randomization. Whether the trial halts at interim depends on the one variable no external analysis can resolve: imetelstat’s true randomized effect, and whether the curves separate enough at ~112 events to cross the boundary.

This document deliberately states no probability of success. Its purpose is to provide the clean factual substrate on which such estimates should be built — and to retire the recurring errors (misapplied real-world survival, confused anchor timepoints, fabricated protocol claims, and “reproduction” mistaken for “separation”) that have driven unproductive speculation.


Sources and confidence levels are noted inline throughout. Figures marked “verify” should be confirmed against primary sources before publication or citation. This is a living reference; it should be updated as the EHA 2026 presentations, KOL commentary, and ultimately the interim readout provide new information.

[biopearl123]

MG, can I just say here that your analysis is elegant. I am blown away by your depth of understanding required to input the proper questions and issues to Claude. What has come out I think is a major contribution to this board. We all should acknowledge the time and effort you have put into this. Kudos. It’s so interesting that the BAT itself is so influenced by what happens before randomization (cycling through JAKi or other non Jaki) and after—a patient who leaves the study can still be given other newer therapies and they still if I have this right, still contribute to the ultimate in study BAT determination. Some of this may be “smoothed out” by the contributions from other enrolling centers that are non US/non European for reasons previously examined. Perhaps the statistical gurus will look at different subgroups to account for mOS effects of newer agent and changing practice patterns that could not have been totally clear at the time of study conceptualization. Anyway, thanks for all the work you have put into this. It could not have happened without a lot of self study and critical thinking on your part. Best Regards, bp

[Secret Third Arm]

Bravo MG.

[biopearl123]

Minor correction is that total enrollment is 327 (not 320), probably would not change much re analysis.