Ziopharm Palifosfamide Decision Delayed: What Does this Mean for the Likely Outcome?

Ziopharm (NASDAQ:ZIOP) recently announced that its Independent Data Monitoring Committee (IDMC) looked at the rate of events in the phase III trial and now expects to have enough events at the end of the 1st quarter 2013 instead of the end of 2012. The phase III trial comparing the addition of palifosfamide to a regime of doxorubicin in soft tissue sarcoma (STS) is event driven, which means that the trial is unblinded to the company once a certain number of events have occurred (in this case progression of disease). The major question is whether this delay is bullish or bearish for palifosfamide and the best way to figure this out is to model the trial to see what conditions were most likely to have caused the delay in events. In general, there are three major reasons why the trial endpoint would be pushed back. First, the trial could have enrolled more slowly than expected (moderately bearish). Second, the control arm is doing better than expected (certainly bearish). Third, the treatment arm is doing better than expected (certainly bullish). Investors often infer from their pre-existing beliefs as to which of the three is the main reason behind the delay, i.e. bears would focus on the first two while bulls would focus on the last. A better method, however, would be to actually model the enrollment and progressions to uncover the most plausible reason. Of course doing this involves a fair number of assumptions but as long as they are reasonable, it is a useful exercise.

The first aspect of the trial that has to be determined is enrollment, and there are some broad guidelines that the model has to follow. The company provided three key bits of information: Ziopharm started enrollment July 19th, 2010; the trial was about 90% enrolled by April 2nd 2012; and the trial completed enrollment by June 2nd, 2012. While we do not know exactly when patients entered the trial, table 1 provides a rough estimate as to the month that patients were enrolled. (Note the large number in the first month, and this relates to all patients in 2010, where I assume for simplicity that they all were treated January 1st 2011. If I modeled those 6 months separately it does not alter the conclusions.) Trials tend to start slowly and build up, so I assume an 11 patient/month enrollment for the 6 months between the start of the trial and the first of January 2011. In the beginning of 2011, the enrollment rate increases to 14/month and steadily rises to a peak of 28/month. These enrollment assumptions coincide with what we know of total enrollment as the model has 90% of patients having been treated by April 2012 and the trial is completed by June 2012. Outside of the enrollment, we must also assume a rate of PFS progression for the doxorubicin arm. To be conservative, I model a median PFS of 6 months and that 83% progress by the end of a year. This is similar to Maurel et al (2009) in which they found doxorubicin generated a median PFS of 6.5 months, 83% progressed by the end of 1 year, and 88% after 14 months. My assumed median PFS is rounded down to 6 months as that study had a relatively young population (median age of 49) and age is strongly correlated with response to treatment. In addition, that study did not have any ECOG 2 status patients, whereas the palifosfamide phase III trial allows in some sicker patients.



It is quite likely that the median PFS for the doxorubicin arm is actually closer to 5 months if not lower. The recent EORTC that looked at the addition of ifosfamide to a doxorubicin regime had a median PFS for the doxorubicin arm of 4.6 months (again with a younger population than the Ziopharm study). This is similar to the phase II PICASSO data that showed a median PFS for doxorubicin of 4.4 months (slide 9). Of course, STS is a diverse group of pathologies but even though they differ in their ability to be treated, the median PFS is unlikely to be over 6 as Van Glabbeke (2002) found that the progression free rate at 6-months ranged from a high of 56% (synovial sarcoma) to 38% (malignant fibrous histiocytoma). As such, while there is always a chance that the trial has an exceptionally rare outlier efficacy for the doxorubicin arm, the bulk of evidence from previous trials indicates that is highly unlikely. Modeling a median PFS of 6 months for doxorubicin is probably on the high side, but it is worth being conservative and use what would likely be a worst case scenario (and consistent with a bearish explanation for the delay).

Table 1 describes the number of patients dosed in each month as well as total number dose. The key, however, is the total PFS events, which is the estimated number of progressions given the patients dosed, when they are dosed, and the assumed rates of progression. As noted above for the doxorubicin arm it is assumed that 50% progressed by month 6. In addition, the model assumes that 55% have progressed by month 7, 60% by month 8, 65% by month 9, 75% by month 10, 83% by month 12, 88% by month 13, 95% by month 16, and 100% by month 19. These progressions are about a month or two later than what would be expected by most of the previous trials but again, I would rather err on the side of caution. The model then assumes that the palifosfamide arm progressed 60% slower than the doxorubicin arm (keep in mind that the trial was designed to find a PFS HR of 0.6). The final assumption that we need to make is how many PFS events are needed to stop the trial and I am using 80% (339). The company has not given specific guidance as to the precise number, so this number could ultimately be higher or lower.

What is interesting is that even assuming a relatively slow start to the trial (only 66 patients dosed in the first 6 months), and a doxorubicin efficacy on the high side of what is historical, the trial will generate enough events to end in March 2013. This triggering is also assuming a .60 PFS HR. In other words, a March 2013 data analysis with the above assumptions implies a PFS HR for palifosfamide of .60 (which would be statistically significant given the powering of the trial). We can use this model as a baseline to infer other outcomes. If, for instance, the doxorubicin arm performs more in line with the historical average, then the March 2013 data analysis implies a PFS HR for palifosfamide of well under 0.60. The same would be true if enrollment occurred quicker than I modeled. Of course, if enrollment was slower and a bolus occurred at the end (which the company noted was not the case) or if the doxorubicin effectiveness was higher than the 6 month median PFS, then the PFS HR of palifosfamide would be higher. That being said, if you only assume 40 patients were dose in 2010 and a larger bolus at the end, the estimate only changes to 337 PFS events in March 2013. While that would imply a slight lower PFS HR, this is unlikely as the company has stated that enrollment was relatively steady (and I would argue that only having 66 patients dosed in six months of 2010 is a relatively conservative estimate).

The biggest risk would be a doxorubicin effect higher than the 6 month median. Again, while it is not impossible, it would represent an almost historic performance of doxorubicin (and one done in an arguably sicker population than previous studies that showed a lower median PFS). In general, then, the delay in the data, while disappointing for having to wait longer, is a net positive for the odds of success. The bearish case on the delay simply does not seem plausible. First, trial enrollment has a minimal effect on the number of events, unless one assumes an exceptionally slow ramp, and even then the effect is not likely large enough to mean a statistically insignificant effect for palifosfamide. Second, the model indicates that a doxorubicin efficacy at an historically high level is consistent with a palifosfamide PFS HR of 0.60 or below. As such, this leaves the bullish explanation as the most plausible, and the news should lead us to upwardly revise our estimates that palifosfamide will demonstrate a robust and statistically significant effect.