Possible effects of preenrollment “cycling” of Jaki

[biopearl123]

If we accept that JAKI drugs have not meaningfully improved OS and Imetelstat has, coming to the study later as a result of exhausting even the newer JAKI drugs, and the attendant delays in enrollment as a result, it does seem likely that the BAT arm will may have an accelerated death rate as a result, simply because these patients will be a a more advanced state in their disease with no hope of disease modification. Even with advanced disease there is some hope for disease modification in the Imetelstat arm, even after the delays imparted by trials with other JAK agents. This consideration could further separate the K-M curves. The trade off has been some likely delays in enrollment and a study that is taking forever.

On the other hand, I always thought patients in both arms were living longer because of general improvements in hematologic care and the benefits of being closely monitored in the study, and further thought that this benefit would apply to both arms fairly equally. Now with the recent comments regarding delays in enrollment in part being related to more and different JAKi drugs I am wondering about the selection of “healthier” patients at the time of enrollment. In other words, some percentage of patients die on JAKi therapy before they get to the study. (15-20%?) so the enrolled patients are “healthier” and will live longer than anticipated. Just looking for ways to understand the “study is taking longer because patients are living longer” vibe I have been getting lately. Another factor might be the availability (or not) of newer JAK agents in other countries and the effects on subsequent enrollment. Just trying to make sense of it all. bp

[Commsman]

I think part of the problem is that the math sets this up to be a long trial. As a thought experiment, pretend all imetelstat patients live forever. We would never reach an interim analysis or a conclusion. With the interim analysis being triggered by 35% deaths and only 33% patients on placebo, we would never reach the interim analysis let alone ever reach a final analysis. We would get close to reaching the interim analysis, but never quite get there. Overall, I think that the longer it takes, then the better it is looking for Imetelstat. Supposedly, if literally all patients in the placebo arm died, and all patients on imetelstat lived, they would call the trial for overwhelming efficacy. Of course, in the real world it's not likely to be so clean cut. So we wait....

[biopearl123]

Commsman, excellent! I sense a bit of Zeno’s dichotomy paradox there! I agree, the longer the better, unless the length is because the comparator arm is outliving the treatment arm…

[Ryan]

I know you try to be measured in your points but that counterpoint seems highly improbable…

Let’s check the greatest hits by our favorite CMOs:
In my opinion:::

“We believe that if IMpactMF confirms the clinical benefits of symptom response and overall survival observed in the Phase II IMbark study, that imetelstat could become a standard of care in relapsed/refractory myelofibrosis.”
This oldie but a goodie comes from Dr.Feller circa 2023 upon 50% enrollment

Then let’s connect it to my fave to date from Dr. Eid, fresh off the press on April 13 2026, paraphrased because I don’t have the direct quote handy:
Eid acknowledged that while the MF standard of care has expanded from one to multiple JAK inhibitors, he does not see evidence that any particular JAK inhibitor has improved survival, characterizing them instead as symptom-relief therapies.

A pointed comment.

Knowing the life expectancy based on studies Geron performed back in the 10’s (maybe came out ‘19, I’m spitballing here), is about 12-14 months on BAT. So there are new JAKs, the good Dr Eid is saying they are the same old same old, and we already know that Phase II showed clinical (not palliative) response.

So, yeah, no on your last counterpoint there BioPearl.

[biopearl123]

Ryan, I certainly was not suggesting that the comparator arm was outliving the study arm (or I would not be here), my point was only that cycling through JAKi does delay enrollment and some patients with advanced disease that could have been enrolled would have died before they could reach the study. (If no deaths occurred while this cycling occurred I would be surprised but I suppose is possible- I posited 15 to 20%)/ This would “enrich” those to did survive the “cycling” to a population that had patients prone to early death (a time when the K-M curves really diverge in the PII) removed by “natural selection”. I other words they had to get over yet another hurdle—time. While this may not be a strong point or be end up being completely irrelevant as I think you are suggesting, I am just trying to look for all the factors that have made this study take so long. I am just wondering if these patients enter the study at a different point on the K-M cure. There are clearly reasons for the delay, some good (as discussed, just being in a study and getting close care, some maybe not so good such maybe changing the population by delays brought about by additional unanticipated pre enrollment treatments with any or all of the four new JAKi. Could be way out in left field on this one but love the discussion and collective thought. And I do appreciate yours. Feel free to pick this apart further. Be interesting to see how this all pans out.

[Ryan]

Well I see there are almost 600 people on the board today - wonder if those are real people/genuine interest, or a bunch of bots.

Anyways, I found this fascinating when I asked about the much discussed crossover patients and how it relates to hazard ratio and the FDA panel calculations and decision making :::

Why Crossover Specifically Destroys Hazard Ratio Clarity
Here is the core problem, stated precisely:
The hazard ratio (HR) in a standard intent-to-treat (ITT) Cox regression compares the instantaneous risk of death between the two arms at any point in time. When BAT patients cross over to imetelstat and subsequently survive longer, those additional survival months are attributed to the BAT arm’s OS curve — even though the patient is now receiving imetelstat. This compresses the survival gap between the two arms and drives the HR toward 1.0 (no effect), making the treatment look weaker than it actually is.
Crossover leads to information loss and dilution of comparative clinical efficacy, and the choice of statistical method for adjusting for crossover greatly affects treatment-effect estimates, including hazard ratios.
The math works against Geron in a specific way here:
• The BAT arm has ~106 patients. Even if only 30–40% cross over (a conservative estimate for a dying MF population with few other options), that’s 30–40 patients whose post-crossover survival inflates the BAT arm’s Kaplan-Meier curve.
• The pre-specified alpha threshold for the interim analysis is demanding — because it’s being conducted at only 35% of events, the O’Brien-Fleming-style boundary typically requires a lower p-value than 0.05 (often around p<0.01 or below) to stop early for efficacy.
• If the unadjusted ITT HR is, say, 0.72 with p=0.04 — that’s statistically significant at the final analysis threshold but may not clear the more stringent interim bar, and the FDA will anchor to the ITT result.

[biopearl123]

Ryan, exactly. This is a partial explanation for why the study is taking longer but also requires us to understand just what the statisticians have in mind to “correct” for this and how the FDA will view the almost certain less dramatic separation of the OS curves and a “worse” HR and p value. There can be no purity in this study because it pays homage to compassion. That’s a good thing for patients but a bad thing for numbers like p values and hazard ratios. If the numbers are good, it is likely the FDA will give extra credit knowing what adjustments the statistical analysis required because of cross over. The statisticians will have to apply some “magic” to project what the number would have been if the patient hadn’t crossed over. This will require a statistical leap faith but will use acceptable methods. So there will be (at least) 3 Kaplan Meir curves, one for Imetelstat, one for crossover patients and one designed to reflect what the OS would have looked like if the patients who crossed over hadn’t.

[Secret Third Arm]

I’d like to add that the trial only allows crossover with the approval of the sponsor (Geron). That means that for every request since the trial started management knows how many patients have made the move from BAT to imetelstat. I believe that this knowledge accounted for the optimism of previous management. They may not know how long they are living after crossover, but they certainly know if the number is large enough to present a statistical problem.

[biopearl123]

Secret, good point. Management knows those things. Crossovers “enrich” the Imetelstat numbers and while as we discussed the OS numbers may get murkified the secondary endpoints may get strengthened as more data available re VAF, TSS, spleen, fibrosis etc. There may be some statistical spin coming our way.

[biopearl123]

Even though JAKi therapies don’t seem to be related to disease modification, they still can prolong life via effects on appetite, inflammation general sense of well being etc. Even fibrosis may rarely improve. According to Dr. Eid, patients are living longer in the study. I know we have beaten this one to death but previous JAKi therapies may affect enrollees, either because they are “healthier”, some have died “natural selection” or others had other benefits. The other usual suspect BAT therapies don’t seem to have changed much over the course of the study. The JAKi effects would have had to be manifest pre enrollment since JAKi drugs are not allowed in the BAT arm. I am sure the statisticians will try to weight this somehow.