When I was at the recent BIO-CEO 2013 conference, I had the chance to meet with Ziopharm (ZIOP) CEO Jonathon Lewis. The discussion mainly focused on the upcoming data and how the company designed the trial to maximize the odds of success. In this article, I want to share both parts of that discussion and then additional work that I have done since the conference. In general, while it is always possible that the control arm is well outside of the bounds of expectations, it seems to be a highly unlikely scenario.
While it seems obvious to say, there are two main reason phase III trials fail after having a successful phase II: drug is not as active as expected or the control arm outperforms expectations. Given that palifosfamide is based off of a known active agent (ifosfamide), it seems unlikely that the drug is not active (see a previous report on the relationship between ifosfamide and palifosfamide). It seems more likely that if PICASSO III is going to miss on progression free survival (PFS), the culprit will be a control arm that outperforms expectations. While we do not know the exact median PFS of doxorubicin that would generate a miss, my previous report modeled the enrollment numbers in an attempt to estimate how high the control arm can go and still get a successful train. In general, it looks like even a median PFS of 6 months for the doxorubicin could generate a statistically significant hazard ratio (likely between 0.6 and 0.7). Given this previous modeling, the question then becomes whether the PICASSO III study will have a median PFS for doxorubicin greater than 6 months.
In general, once soft tissue sarcoma becomes metastatic, the prognosis is grim. In fact, metastatic status is by far the strongest predictor of progression. Of course, palifosfamide is looking at the metastatic group, so the question then is what differentiates the prognosis in this group and the answer is little. Glabbeke et al (1999: 154) found that "age is highly predictive for response to first-line chemotherapy, but performance status does not add predictive information to the model." In addition, Glabbeke et al (1999) noted that liver metastases are highly correlated with poor responses and that good performance status is related to longer survival but not necessarily response to chemotherapy. With this in mind, we can look at some data to judge its relevance for the PICASSO III trial.
Maurel et al (2009) has perhaps the highest median PFS for doxorubicin in recently published studies at 6.5 months. If PICASSO III has a median doxorubicin PFS of 6.5 months, then it would be problematic but how relevant is this study? The biggest difference between Maurel et al (2009) and the PICASSO II trial is the median age of 49 compared to 57 in PICASSO II. Glabbeke et al (1999) make clear that age matters in that the odds of surviving 1 year with an age between 40-50 are 52% but when 60 or older the 1 year survival is only 39%. As such, we would expect PICASSO III, which allows patients over 60 just as PICASSO II, to have a slightly lower median doxorubicin PFS. In addition, the Maeurel et al (2009) study had very few patients with liver metastases (a liver metastasis changes 1 year survival odds from 50% to 42%). It is unclear what portion of PICASSO II patients had liver metastases, so it is unknown how this would affect the expected doxorubicin median PFS. In general, however, we can use the Maurel et al (2009) as one comparator but given the younger population I think it is safe to put the "expected" median doxorubicin PFS in PICASSO at 6 months instead of 6.5 months. Of course, that is only one data point.
In some ways the closest apple to apple comparison would be the PICASSO II trial but there are some slight changes between PICASSO II and PICASSO III. The most significant difference is that PICASSO II allowed some second line patients (about 30% were second line- see slide 13), where PICASSO III is strictly first line. So the question is what is the median PFS for the first line doxorubicin patients in the PICASSO II trial? While there is no easy answer, I think we can take the control arm from the votrient (GSK drug recently approved in 2nd line STS based on PFS data) as a worst case scenario for what the 2nd line looked like in PICASSO II and then we can back out an estimate of the 1st line doxorubicin patients. The control group in the votrient phase III trial had a median PFS of 1.6 months. This is absolutely the worst case scenario and is completely unreasonable because this has 2nd, 3rd, and 4th line patients and the control group was treated with placebo and not an active drug like doxorubicin. Even if we assume a median PFS for 2nd line of 1.6 months, it backs out to a 5.62 PFS for the 1st line group in the PICASSO II trial (keeping in mind this is a rough estimate as we are working with medians and not means). If we use a more reasonable 3 months median PFS for the 2nd line patients, then we get a 5 month median PFS for the doxorubicin arm.
The third data point that we can use to triangulate an estimated median PFS for the doxorubicin arm in PICASSO III is the recent phase III trial EORTC 62012 trial of single agent doxorubicin versus doxorubicin plus ifosfamide in 1st line chemotherapy for metastatic STS. In this trial the median PFS for the doxorubicin arm was 4.6 months. This is in line with Ziopharm expectations and is actually likely high in that the study excluded patients older than 60, which we know means that they had easier to treat patients. One small problem with the trial is that it started in 2003 and it is likely that in the intervening time doctors have gotten better at using doxorubicin in treating STS. Of course, that is difficult to quantify, so I think we can use the 4.6 month as our data point.
So looking at three trials that are similar to PICASSO III we get estimated doxorubicin median PFS of 4.6 months, 5.62 months, and 6 months. If we equally weight these estimates, then it comes out to 5.41 months (keeping in mind that we used conservative assumptions when coming up with the values). In my previous timing estimates I used 6 months and found that the last progression event would take place in the middle of March 2013. Using the 5.41 months and assuming a hazard ratio of 0.6, it comes out to the last progression event taking place in early March/late February, which is about 2 weeks earlier than the ZIOP press release. What does this mean? First, remember that I used very conservative enrollment figures. Second, even if you want to continue to use the conservative enrollment figures, the press release in some ways implies a doxorubicin median PFS of 4.9 months (with a hazard ratio of .60), which is certainly not an unreasonable figure. Finally, you could lower the hazard ratio slightly to get an earlier last event. In this case, you could still see a statistically significant effect but the absolute benefit would be lower.
When all is said and done, it appears well within reason that the PICASSO III trial will hit its endpoints and will be a success. Even using very conservative estimates of the control arm and enrollment figures, we get very close to the timing of the last progression event. Is it an absolute slam dunk that the data works out? Of course not but taking all of the information we have about previous trial experience with doxorubicin, the enrollment timelines, and expected hazard ratio, it appears that the odds of successes are higher than that of failure. With the company trading at only a $370 million market capitalization and with 23% of the float short, it looks like a favorable risk/reward for ZIOP bulls.
Disclosure: I am long ZIOP.