OK, Finally, full part 2 of my post:
Part2:
In the new PR released today there’s a new important piece of information:
“Patients who received rexlemestrocel-L had a 68% reduction in the rate of recurrent hospitalizations from non-fatal heart attacks or strokes compared with controls, with a hospitalization rate of 1.90 per 100 patientyears of follow-up in the rexlemestrocel-L arm versus 5.95 per 100 patient-years of follow-up in the control arm (p=0.0002).”
What does this mean?? And what is the difference between this and primary endpoint?? It looks very similar, right?? So did the trial all of a sudden passed the primary???:
1. The original primary endpoint as stated in the trial NCT is: “Time to recurrent non-fatal decompensated heart failure major adverse cardiac events (HF-MACE).“ sounds very similar to “rate of recurrent hospitalizations from non-fatal heart attacks or strokes” isn’t it?... well not exactly…
2. The first difference is the omission of the word “decompensated”. Decompensated HF is different from compensated HF. Shortly: it is an HF that already causes significant symptoms such as fatigue, angina, palpitations, etc. stage II patients will be more compensated patients, whereas stage III patients will be more decompensated patients. During the course of the trial a compensated patient can become decompensated. So, the original endpoint measures the time to the first MI or Stroke, but one that occurs when the patient is already decompensated…. So, suppose there was a patient that had an MI while being compensated… this event will be skipped, and the analysis will keep looking for the patient to become decompensated before starting to look for such an event. This is how I understand this… Now I don’t know who designed this trial at the end, but this is a very weird decision. Why? Because, if that is what you are measuring, you should only recruit patients that are already decompensated. Why? Because it will make you trial more efficient and bigger numbers for the endpoint. I don’t know the breakdown, it may well be that almost all of them were decompensated patients, but if not… that’s a design flaw reducing the power of the trial imho. On the other hand – this decision resulted in recruiting the very important stage II patients. It will be very interesting to know how many were decompensated in the stage II , and III stages, and how many progressed to that.
3. The bigger difference is the time to event vs. rate. OK, this is interesting and smart and this is the MAJOR interesting point in this PR. The original analysis followed a patient and counted the time for the first non fatal MACE and then compared the KM curves. The analysis we get now calculates the average hospitalization rate PER patient year in control vs. treatment. And the difference is huge, with a 68% (!!) reduction. IMHO, this measure is much more important than the time to event analysis. Why? Because this analysis shows that the OVERALL number of MACE hospitalizations is significantly reduced. This is the real thing. This is the gold in the box. Think about it: if you use rex-L (in stage II or III btw), and simply count how many major (MACE event) hospitalizations a patient had, you see on average ~2 hospitalizations , while seeing ~6 for an untreated similar patient (!!). This is beyond huge. This with the MI/Stroke/CV deaths result, can change the way this trial is judged. Maybe even by the FDA.
4. So why didn’t the trial pass the time to event analysis?? This is weird indeed. Seeing such a huge reduction in hospitalizations overall, and not seeing it at all in the time to event…. Is it that many of the patients are compensated?? Is it that the trial was not powered enough to see a difference of ~4 months in time to event (12/6=2 months on average for controls, 12/2 = 6 months on average for treatment)? Is it because events come very much one after the other and correlated for untreated patients?... We need more information to understand this. BUT – really – the OVERALL expected number of such events is way more important than the time for the first one.
(20min delay)