Celsion (CLSN) announced today that ThermoDox in combination with radiofrequency ablation (RFA) did not meet the primary endpoint of the Phase III HEAT Study in patients with hepatocellular carcinoma (HCC). The HEAT Study was designed to show a 33% improvement in PFS with 80% power. We will get more details from CLSN management as they analyze the data further. Meantime, I can only speculate on three realistic possibilities on why I was wrong, which I will detail below.
It seemed like such a sure thing. We had the enrollment data. We had the progression-free survival (PFS) data. We had what the placebo arm should do. We had the data to calculate how long the average patient was in the trial. We had the data to calculate how many patients should have had a PFS event. And most importantly, based on all this data, we had what seemed to be a huge amount of room for margin of error even if the control arm performed much better than expectations and guidance.
Management revealed on the conference call that the control arm performed around 20% better than expected, well within the margin of error. Yet the Phase III HEAT trial missed its primary endpoint, and in the words of the CEO: "it was not close."
So what went wrong?
1. Was management's expectations for the control arm better than they voiced publicly?
Management consistently guided for 12 months median control arm to PFS. This left room for up to 18 months in the control arm yet ThermoDox would still work. If management was internally expecting say 18 months yet telling the public 12 months, that could mean legal problems.
2. Did the treatment failure criteria throw everything off?
According to the trial protocol, "If a complete ablation is not achieved after these two study treatments, the subject will be considered a treatment failure and the patient will be discontinued and followed for survival only." Many people including myself used a 701 enrollee patient population when doing our calculations. Yet treatment failure is not one of the four criteria for a PFS event. These patients would should be excluded from the 701 enrollee figure for calculation purposes only. What's worse is if ThermoDox actually did reduce the chances of treatment failure, by default extra-sick patients from RFA-only would get excluded from the study and while other extra-sick ThermDox patients would be included, leaving the ThermoDox arm already starting off sicker than the RFA-only arm.
3. Was PFS was significantly more than 380?
It's conceivable that upon unblinding, the total PFS events were significantly higher than 380. If for example there were 500 PFS events instead of the expected 380 that I and others (and likely management) used in our calculations, then it would throw all calculations off because the calculations and statistical models absolutely depended on an accurate assumption of 380 events and not something greatly higher.
Seems like there's more unanswered questions now than before today. Management indicated that the data is still "fresh" and after further analysis will provide more details. We should learn more in the coming days and weeks. In the mean time, CLSN is an extremely risky stock now potentially on life support. Don't automatically assume an 80% decline means it's cheap especially with so few details so far. I look forward to further updates and possibly some hope in the future for shareholders and patients.