Threshold Pharmaceuticals: Understanding The Recent Data

Threshold Pharmaceuticals (NASDAQ:THLD) recently announced positive phase IIb trial results of TH-302 in patients with pancreatic cancer. The median progression free survival (PFS) was 5.6 months for patients treated with gemcitabine in combination with TH-302 (combined analysis of both 240 mg/m2 and 340 mg/m2 doses), which compared quite favorably to a 3.6 month PFS for patients treated with gemcitabine alone. This difference was statistically significant with a p-value of 0.005. The results are basically in line with my expectation of a statistically significant PFS effect that I wrote about in a previous report. While the stock certainly moved in response to the results, I want to dissect it in more detail to figure out what exactly it means and does not mean.

The Path Forward

To understand the path forward, we need to understand that a phase IIb trial is not meant to provide definitive evidence of efficacy; rather, a successful phase IIb trial will provide an efficacy signal and important information about how to design a phase III trial with the greatest chance of success. From this perspective, the trial clearly showed an efficacy signal strong enough to move forward into phase III testing (at least in my opinion). A 2 month difference in PFS is not just statistically significant in the trial but clinically significant in locally advanced pancreatic cancer. Tarceva was approved by the FDA for locally advanced pancreatic cancer with a PFS benefit of 2 weeks- not 2 month but 2 weeks (look at table 7 in label for details). Locally advanced pancreatic cancer is a very difficult disease and the benefit seen with TH-302 treated patients is certainly something worth pursuing.

The company did not provide any data on the overall survival (OS) because that data takes more time to mature, so we cannot interpret that data. Of course, the Tarceva data also showed only a 2 week improvement in OS, which was statistically significant given the number of patients (521 treated with 261 in treatment group and 260 in control) in the trial. While we will likely get the OS data later, I would not read too much into it as this is a cross-over trial design. In other words, those who are in the gemcitabine arm are allowed to then take TH-302 after progression, which means that an OS analysis will not have an apples to apples comparison. There is still a chance that the trial shows a statistically significant OS effect, but with a cross-over design I would not expect one nor would I read that through to a prospective phase III trial results.

The company also provided the response rate, which was 22% in TH-302 arm compared to 12% in gemcitabine alone arm. This is a very large difference, as Tarceva only had an 8.6% response rate compared to its control of 7.9%. A couple of clarifying notes on this result. First, it is not obvious how Threshold calculated overall response rate, i.e. is it only complete response and partial response or did they include stable disease? Tarceva only included complete and partial responses, which might explain the absolute differences in the numbers. Second, the Tarceva results were not statistically significant (and it still got approved). Third, while the difference of 10% in response rate is large for TH-302, the company did not indicate whether or not it was statistically significant. This is not to say that it is not statistically significant, but it is not immediately clear from the information that we have.

Overall, however, the results are clearly a positive for Threshold and show significant signs of efficacy in a difficult to treat population. The results, from a clinical perspective, compare very favorably to a recently approved Tarceva. As such, it seems clear that the phase IIb was a success in that it provides enough of an efficacy signal to make a phase III program worth investigation. If these results maintain, then TH-302 would be an exceptionally effective compound in locally advanced pancreatic cancer. Of course, pancreatic cancer is often where good phase II data go away in phase III as such it is important to figure out what can go wrong.

What can we learn about the potential Phase III trial?

In all honesty, there is not enough information to make a reasonable accurate assessment as to how well this data will translate into phase III results. What I want to do in this section is highlight some issues that would need clarified before one can handicap a phase III trial.

Perhaps the biggest concern is that the trial was open labeled. While not a major issue when looking at OS, issues can arise when these trials use surrogate endpoints like PFS. While there are potentially many sources of bias, Bhattacharya , et. al. (2009) best highlight the major concern:

Simulation studies have demonstrated that differences in the timing of disease evaluations can significantly bias PFS analyses. Modeling by Friedlin et al demonstrated that performing more unscheduled tumor assessments in the control arm than in the experimental arm could result in an apparent improvement in outcome where none existed. This could occur in open-label studies when there is a greater concern regarding progressive disease in the control arm than in the experimental arm or, conversely, when toxicity is excessive in the experimental arm.

As such, the median PFS that were reported in the trial might not be the true measure of PFS. While this is certainly a possibility, a 2 month difference provides Threshold some significant wiggle room- so to speak. Ideally, we would like to see a blinded independent review of this data and other sensitivity analyses. These would create a greater sense of confidence in the reported PFS but there is no reason to expect that these would change the results. It is simply an open issue until we receive further clarity.

Another unknown is the difference between the two dosing schedules of TH-302. Previous studies showed a dosing effect, where the higher dose generated better responses. The company has not provided information about the median PFS of the different doses, so we do not know if the dosing effect maintained. Again, this is not a critique as the company cannot release everything in a first look but it will be interesting to see the split results later in the year. The dose response becomes most important when analyzing a potential phase III trial. If the higher dose had a better response and the company moves forward with that dosing schedule, then the reported data is likely underestimating the results as it is a blend of the two dosing schedules.

Finally, as noted above PFS is a surrogate endpoint, where the gold standard is OS. While there is a correlation between improvements in PFS and OS, it is certainly not a one-to-one relationship. Fortunately, the correlation is stronger when there is a shorter survival post progression (SPP) (Amir et. al. 2012). So while one needs to be careful about inferring potential OS effects from PFS data, there is likely a tighter correlation in locally advanced pancreatic cancer given a relatively short SPP.


In general, the trial was a success in that it generated a strong signal of efficacy. That being said no phase IIb trial will provide incontrovertible proof because that is not the point. In the second section, I highlighted a number of caveats but there is nothing to say that a bias definitely exists or that it would eliminate the effects; rather, these are areas that an investor or potential investor should follow to help you update your expectations about the drug. In biotech, we invest with incomplete information so you need to both analyze the new data but also figure out what you ideally need to know to form a more accurate assessment.

One final point is that TH-302 is part of a larger program and efficacy in this trial reads through to the entire program. The company is running a phase III trial in soft tissue sarcoma (STS) and this data will de-risk (to an admittedly small degree) that program as well. That being said TH-302 has consistently shown signs of efficacy across a number of diseases and each data point in and of itself does not prove it works but each successful trial increases the odds that it is an active compound. At this point, potential investors really should be concentrating more on the STS trial as that will be the next major value inflection point but is a discussion for another day.


Amir, E, b. Seruga, R. Kwong, IF Tanock, and A. Ocana. 2012. "Poor Correlation between Progression- Free Survival and Overall Survival in Modern Clinical Trials: Are Composite Endpoints the Answer?" European Journal of Cancer. 48(3) 385-8. Bhattacharya, Suman, Gwen Fyfe, Robert J. Gray, and Daniel J. Sargent. 2009. "Role of Sensitivity Analyses in Assessing Progression-Free Survival in Late-Stage Oncology Trials." Journal of Clinical Oncology. 27( 35): 5958-5964.

Freidlin B, Korn EL, Hunsberger S, and et al. 2007. "Proposal for the use of progression-free survival in unblinded randomized trials." Journal of Clinical Oncology. 25:2122-2126.

Disclosure: I am long THLD. I own shares of Threshold but that entire position is hedged with covered calls (short the April $7.50 calls).

About this article:

Author payment: $35 + $0.01/page view. Authors of PRO articles receive a minimum guaranteed payment of $150-500. Become a contributor »
Tagged: , , , Biotechnology
Problem with this article? Please tell us. Disagree with this article? .