Tuesday, August 2, 2011

The perils of PFS as HEAT study primary endpoint

While PFS is the agreed upon primary endpoint for the HEAT study and reinforced via the trial’s SPA (Thermodox must demonstrate a 33% PFS improvement), I have and continue to express some concern over the use of this endpoint. To clarify, while PFS is a well-accepted endpoint by the FDA and very commonly used in clinical trials (in particular because it is thought to be strongly related to overall survival and other clinically significant endpoints, but PFS allows for quicker data generation/earlier NDA submission), I believe this might not be the best endpoint to measure the effect of Thermodox.

As it relates to the HEAT study, let’s clarify exactly what a PFS "event" could consist of:
  1. Local tumor progression – This would be progression relating to the originally ablated tumor.
  2. Intrahepatic distant progression – Emergence of an entirely new tumor in the liver, which may or may not be related to the originally ablated tumor.
  3. Extrahepatic disease progression – Tumor spread to other parts of the body
  4. Death – Death from any cause
You might already see where I am going with this. I strongly believe Thermodox will flex its muscle when it comes to reducing local tumor progression. Study after study has shown that local recurrence, when it occurs, happens just peripherally to the original tumor, most often due to an insufficient ablation margin, and much more likely in larger tumors.

When Dr. Borys, Celsion’s Chief Medical Officer was asked about the choice of PFS as an endpoint, he expressed that the company was “comfortable” with the endpoint because “the most common first site of recurrence is local recurrence.” Unfortunately, this is not universal in every study I have looked at (you can find these under RFA in Current Practice). For example, from the filename “WJGS-Current Status of RFA for HCC 2010.pdf”, we see the following in Table #1:

Keep in mind, these are from smaller studies than the HEAT trial, and are generally looking at smaller lesion sizes, so we do not have an apples to apples comparison (this cannot be stressed enough, since local progression rises exponentially with larger lesion sizes). Nevertheless, LTP was on the low side, while “new recurrences”, which included intrahepatic and extrahepatic distant spread, was relatively high.
This is a subtle, but important point for investors to keep in mind with regard to the HEAT study. I have seen some evidence that suggests LTP is related to some intrahepatic spread, so by controlling the tumor locally, Thermodox might be precluding the onset of new lesions elsewhere in the liver.

I, for one, am extremely excited to see what the HEAT study will reveal. 


  1. If I understand this correctly, per the table shown, in prior studies using RFA, patients have suffered new recurrences at 4-5x the rate of local tumor recurrence. So a therapy could be 100% successful in eradicating local tumors yet easily still fail the by PFS criterion. I can understand how a larger local lesion could result in higher odds of local progression, but still, based on this data, it seems to me the odds of CLSN's success are in fact poor. Can you explain why this is not the case?

  2. jdbear,

    Thanks for visiting my blog! I included that table as an illustration of what is in my opinion, the most significant challenge facing the HEAT trial, and that is, intrahepatic distant progression that will inevitably be captured by the PFS endpoint. If it was just local progression, which is a secondary endpoint I believe, I would not even be bringing up this concern.

    I stressed that the studies in that review article are by no means an apples to apples comparison. The HEAT trial is enrolling patients with multiple lesions (up to 4) as big as 7cm. And, local recurrence rates can get as high as 40% from what I have seen in the literature. The other issue is timing...for the most part, the most common FIRST recurrence is indeed local, but again, I have seen some variation here. So, even though patients in the Tdox arm might EVENTUALLY get intrahepatic spread, in the meanwhile, the control arm RFA only arm patients will get local recurrence with much greater probability than the Tdox + RFA arm.

    Overall, I remain very confident of the trial's success, precisely because I am expecting a very high rate of local recurrence in the HEAT trial.

    Appreciate you taking time to comment!

  3. OK, thanks. Reviewing the slides in your subsequent posting also helped clarify this issue.

  4. I see your point, that Thermodox may be more effective for local tumor progression, but, how can you possibly expect that to have been a primary endpoint for the HEAT trial instead of PFS?

    Since there is no way to tell if a particular patient's disease progression will be local or not, there is little value in having a treatment that has only shown to be effective just for local progression.

    You need to have a primary endpoint that demonstrates efficacy for all patients that may get the treatment. Thus PFS, or OS (overall survival).

    P.S. since Thermodox does get doxorubicin in the bloodstream, I'd expect it will demonstrate a degree of efficacy over RFI (alone) for all forms of progression.

  5. Roxie,

    I agree, the FDA and the oncology community in general has basically said, "Look, local control is good, but for the most part, we want to know if local control translates into a clinically meaningful benefit to patients, with overall survival being the gold standard."

    PFS, as a reminder, remains a surrogate endpoint for overall survival, and an endpoint agreed upon for accelerated approval. Even if Thermodox shows an improvement in PFS and receives marketing approval, they WILL have to show an overall survival benefit to convert their approval to a "normal" approval. So, the accelerated approval is conditional upon post-marketing OS data.

    With all of that said, there are many studies in oncology that won't necessarily require PFS or OS as an endpoint, in other words, the community has agreed that other endpoints are equally valuable from a clinical benefit perspective. For example, in the Thermodox RCW trial, the original registrational protocol for their Phase 2 trial (which is likely to change since they are expanding the inclusion criteria and will have a control arm) called for a primary endpoint to determine durable complete local response. So, in the RCW setting, there is consensus that local control IN AND OF ITSELF is clinically meaningful.

    Back to your original point, in the HCC setting, local control is "not enough" from a regulatory perspective, hence PFS/OS. And by the way, I do expect that the predominant FIRST site of progression in the HEAT trial will be local, allowing Thermodox to flex its muscle.

    Thanks for the comment.

  6. Thanks for the thoughtful reply.

    The RCW trial is a Phase 2, meant not to establish efficacy, but the maximum tolerated dose; so I'm not sure how useful it is to compare Phase 3 endpoints to those on non-Phase 3 clinical trials.

    Can you explain your statement regarding how CLSN will still have to prove OS? How and when will that be done? Are you referring to the need for follow-on trials after HEAT? Will it be after an initial NDA approval and commercialization? Does this have something to do with the "rolling NDA" that they intend to file?

    (I have seen cases where improved PFS did not translate to OS, but that seem to be for less aggressive cancers.)

  7. Roxie,

    Actually, the originally planned RCW DIGNITY Phase II trial WAS registrational, meaning, enough for FDA approval. This was agreed upon with the agency, and it's in every one of their public presentations. This makes sense, since the originally defined patient population are very, very late stage patients. It's extremely difficult to recruit such patients, and this is precisely why they are thinking about expanding the protocol.
    I actually asked during the last conference call whether or not expanding the protocol would effectively render the trial "un-registrational", and the answer was yes.
    As for the HEAT study and OS, yes, indeed Celsion will eventually have to prove OS. So, assuming they hit PFS and gain eventual marketing approval, as a function of fast-track and the agreed upon PFS endpoint, this is called "accelerated approval" (PFS is not ALWAYS an accelerated endpoint, in this case, it is). Products that receive accelerated approval eventually need to be converted to "regular" approval, and this is where Celsion will need to provide and OS readout to confirm that an increase in PFS indeed DID translate into an OS benefit. For the record, Celsion needs top report 372 deaths for OS submission. OS takes a long time to mature, hence the rationale for this whole process. If you have any other questions on this, I highly, HIGHLY recommend you read this:
    Imagine if marketing approval could ONLY be granted in this case with OS data. We would probably have to wait another 2-3 years for that data to come out, THEN start the NDA process, etc. Needless to say, accelerated approval really saves manufacturers a lot of time.

  8. Should have probably spell-checked my last comment ;-)