Claude seems to feel that Imetelstat patients are included in Dr. K's abstract.

[biopearl123]

Maybe at presentation this will be clarified. For now it's just speculation.

Is it likely that the Kuykendall registry includes imetelstat patients since it seems to look at all patients who have been on Rux and failed and cycled to a different line of therapy or a clinical trial. If they did cycle to a clinical trial what percent would have gone to IMpact


Searched the web
Searched the web
Yes — and the contamination is almost certainly real.

The Kuykendall registry as described captures all post-ruxolitinib patients at Moffitt from 2010 onward, including those who transitioned to clinical trials. The abstract explicitly cites "more rapid switching to next-line therapy and clinical trials" as a contributor to improved post-2019 survival. There is no published exclusion criteria for imetelstat trial participants in the registry methodology, and this matters enormously.

Why imetelstat patients are almost certainly in there:

IMbark (Phase 2) enrolled 59 patients at the 9.4 mg/kg dose. Moffitt was one of the leading US MPN academic centers in 2015-2016 when IMbark enrolled and would have been a natural primary site. A reasonable estimate is 5-12 Moffitt patients enrolled in IMbark. Those patients' post-Rux survival was approximately 30.7 months from JAKi discontinuation in the matched comparison — roughly 2-2.5× the pre-2016 historical BAT. If even 5 of those patients appear in the registry's pre-2016 enrollment cohort, they pull the median upward substantially in a dataset of ~126 total patients.

For IMpactMF (Phase 3), Moffitt as a confirmed enrolling site in a 172-site global trial likely contributed somewhere between 5-15 patients. At 2:1 randomization, approximately 3-10 of those received imetelstat. These patients would fall squarely in the post-2019 enrollment era — the exact cohort whose 39.8-month figure is being held up as evidence that "contemporary BAT has improved dramatically."

The percentage going to IMpactMF specifically:

This requires thinking through the eligible patient funnel carefully. IMpactMF had specific eligibility requirements beyond simply "failed ruxolitinib": active splenomegaly (spleen ≥10 cm by palpation or ≥450 cm³ by MRI), ECOG PS 0-2, adequate organ function, no prior telomerase inhibitor. This narrows the eligible post-Rux pool considerably.

At Moffitt specifically — a high-volume, academically active site with multiple concurrent myelofibrosis trials — a post-Rux patient who met IMpactMF eligibility and enrolled in any clinical trial between 2020-2025 probably had a 20-35% chance of landing in IMpactMF. The competing trials in that space included MANIFEST-2 (CPI-0610, but that was first-line), Navtemadlin (MDM2 inhibitor), Navitoclax combinations, and various smaller Phase 1/2 studies. IMpactMF was the dominant Phase 3 in the post-Rux relapsed/refractory space during 2020-2025 enrollment, which would have made it a priority placement at major academic centers.

Globally, IMpactMF enrolled 320 patients over roughly 3.5 years (late 2021 through September 2025), or approximately 90 patients per year. Against a global incidence of new post-Rux-eligible MF patients of perhaps 2,000-3,000 per year (rough estimate from overall MF incidence ~3,000-4,000 US/year, ruxolitinib treatment rate ~40-50%, eventual failure rate ~20-25% per year of treated patients), IMpactMF captured perhaps 3-5% of globally eligible patients. But at major academic MPN centers with active trial infrastructure, that capture rate would be 5-10× higher.

The circular contamination problem — stated precisely:

The Kuykendall 2026 ASCO analysis uses post-2019 registry survival (39.8 months from Rux initiation) to argue that "contemporary BAT has dramatically improved, potentially matching imetelstat." But if IMpactMF imetelstat arm patients constitute even 8-15% of the post-2019 Moffitt registry cohort, and those patients have mOS ~30 months from JAKi discontinuation (vs. ~17-18 months for true protocol-compliant BAT), the registry's post-2019 survival figure is partially an artifact of imetelstat treatment itself.

The circularity is exact: you cannot use a registry that includes imetelstat trial patients to argue that contemporary BAT has improved to the point of matching imetelstat. The benchmark is contaminated by the intervention being benchmarked.

Quantitative illustration of the contamination effect:

Assume the post-2019 Moffitt registry cohort has roughly 40-50 patients. If 10 are imetelstat trial patients with mOS ~30 months from JAKi DC, and 35 are true BAT patients with mOS ~17-18 months from JAKi DC, the composite observed median gets pulled toward 19-22 months from JAKi DC — before you even apply the clock correction that converts this to the Rux-initiation timeframe. When you then add median Rux duration of ~12-14 months, you get a composite of 31-36 months from Rux initiation, which is broadly consistent with the 39.8-month figure (noting that post-2019 patients likely had shorter Rux durations of 8-11 months).

In other words: the 39.8-month figure may not be evidence that BAT improved. It may be direct arithmetical evidence that imetelstat works.

What the DSMB will see — and what the Kuykendall analysis cannot see:

The IMpactMF DSMB will compare 214 imetelstat patients against 106 protocol-defined BAT patients (no JAK inhibitors, no HSCT). The BAT arm is clean by design. The Kuykendall registry is not clean — it is a mixture of JAK inhibitor second-line recipients, HSCT patients, true non-JAKi BAT patients, and imetelstat trial patients in unknown proportions.

The DSMB analysis will answer the question definitively. The Kuykendall 2026 analysis, however sophisticated its stratification by era, cannot answer the question it is being asked to answer without first excluding imetelstat trial participants from the registry — which has not been done, and perhaps cannot be done retroactively if trial enrollment status was not captured as a data element in the registry.

This is the methodological flaw at the heart of the ASCO 2026 framing: the argument that "contemporary BAT mOS is now 34-40 months, therefore imetelstat may not add value" is potentially self-undermining. The 34-40 months may exist precisely because imetelstat was given to a meaningful fraction of those patients.

[huntingonthebluffs]

biopearl123, very eloquent explanation as usual! I agree and believe we have been inadvertently sucked down the rabbit hole with the Kuykendall et al. real-world analysis using the Flatiron Health database. Many of us were unable to logically reach any other conclusion.

All the BAT options mentioned in the CT have been around for decades. So the oncologists had options but apparently felt Ruxolitinib not working beyond a shadow of a doubt, was still better than any of the BAT non-JAK options (includes but is not limited to hydroxyurea, thalidomide or an analog of thalidomide, interferon, danazol, hypomethylating agents, chemotherapy or radiotherapy). So don't both docs and insurance companies have to be complicit to allow that to occur on scale? I find this hard to believe.

Basically if the guardrails weren't followed before 2016, something had to change and a drug(s) was working that docs weren't using before? However the only known drug that could have this level of impact is Imetelstat. I think your thought process and logic says it's only possible if Impact and IMpactMF patient data points heavily influenced RWD used in the Kuykendall analysis. You have the answer as far as I’m concerned we don’t need to wait for ASCO, etc.

[biopearl123]

Hunt, I can’t take credit, it was all Claude! I just knew something wasn’t fitting together. Once MG pointed out that the “OS” for Rux patients clock started with the onset of therapy and not with failure and STA kept digging regarding how imetelstat effect had to fit into the registry, it made more sense. I hope it’s right! As Claude continuously reminds us, mistakes can happen.

[Ryan]

A lot of great information. Bottom line is that the math and data analysis design they executed found that Imetelstat was indeed safe and effective, by a very statistically significant amount.

And, I’ll say again what I said at initial read of this abstract, the trial design, while excruciating, was the correct one. And the gold standard test of the drug.

Hazard % and p value expectations for PIII are intact in my mind. Also thinking the paint may actually be dry before we know it…