Editor's note: On Wednesday night, Onconova Therapeutics (ONTX) announced negative results from a phase III study of its experimental drug rigosertib in patients with high risk myelodysplastic syndrome (MDS). Rigosertib plus best supportive care failed to prolong survival in these MDS patients compared to best supportive care alone. Onconova shares are down 39% to $8.40 in Thursday pre-market trading.
I have been writing about Onconova Therapeutics for quite some time and owned the stock. While acknowledging the outcome of the rigosertib high-risk MDS study would be a close call, I believed it would squeak by as a success. In biotech investing, you analyze the data and decide what you think is most likely to occur. Sometimes (a lot of the times) your best guess ends up being wrong. Unfortunately, that's what happened Wednesday night with Onconova.
Why did the rigosertib high-risk MDS trial fail?
In the first article I wrote on Onconova, I discussed the assumption used to predict survival for patients in the best supportive care arm. Cancer drug studies often fail not because the experimental therapy under-performs, but because the control arm out-performs expectations. For rigosertib to be successful, Onconova needed to make accurate predictions about how long patients with high-risk MDS would live if treated with best supportive care alone. For this study, the company assumed best supportive care patients (the control arm) would have a median overall survival of approximately 17 weeks. In reality, as we learned Wednesday night, these control arm patients reported a median overall survival of about 24 weeks. They lived longer than Onconova predicted.
As reported by Onconova Wednesday, high-risk MDS patients treated with rigosertib plus best supportive care showed a median overall survival of 8.2 months compared to 5.8 months in the best supportive care alone arm. Overall, rigosertib added to best supportive care reduced the risk of death by only 14% compared to best supportive care alone. This survival benefit favoring rigosertib was not large enough to be statistically significant.
Interestingly (at least to me), rigosertib actually outperformed survival expectations by about 3 weeks. The study failed because the control arm patients lived longer than expected, narrowing the difference in survival.
Why couldn't Onconova more accurately model the study with the proper survival assumptions for the control arm? In some ways, this was a difficult control arm to model because there are precious few studies of treatment in high-risk MDS patients. As I talked about previously, Prebet et al (2011) is one of the few studies that looks at high-risk MDS after Vidaza failures and found a median overall survival of 16 weeks for best supportive care. Of course, that is only one trial so the variance around that estimate is going to be quite large. Onconova was reasonable to rely on that study in estimating the effect of best supportive care but they left themselves with little room for error.
What is next for Onconova? The company talked about a survival benefit seen in a post-hoc subset of patients -- a common pivot for companies dealing with the aftermath of a failed study. The positive signal seen in this patient subset might be real, but I give it low probability because there is no clear reason why this subgroup of patients would respond any differently to rigosertib compared to all the patients enrolled in the study.
In fact, Onconova admits rigosertib survival in the subgroup of patients is similar to the survival reported overall. The difference is that patients in the subgroup treated with best supportive care alone have poorer survival, closer to the 17-month assumption used to design the study. Why is this happening? Onconova offered no explanation so unless it does, I am not optimistic that these positive subset results will be replicated in another study.
The failure of the rigosertib study in high-risk MDS patients is a setback but not the end for Onconova. The company noted Wednesday that a phase III study of rigosertib in low-risk MDS patients will be started as soon as possible. Some positive data in this patient population was presented last December at the American Society of Hematology annual meeting in which transfusion independence was reported for some patients. If Onconova can replicate this benefit in a large phase III trial, rigosertib is likely an approvable treatment for low-risk MDS. There is also the potential for rigosertib to be effective in head and neck cancer but I see that as a lower probability given the dearth of data in that setting.
At this point, an investment in Onconova is a bet on the potential that rigosertib can continue to demonstrate its ability to induce transfusion independence in low risk MDS.
Sobek is long Onconova.