NorthWest Biotherapeutics Inc (NWBO)

[flipper44]

Googled: Crossover Arm Overall Survival


There are many recent articles and examples on crossover arm dynamics. In my reading thus far, it appears scientists include crossover arms for humanitarian reasons. Because of this, they accept Overall Survival can become much more complicated (if not impossible) to analyze for efficacy in some cases.

One reason patients do not enter a crossover arm is because the standard of care led to safety issues.
Example: Gemcitabine plus TH-cr-404 = treatment arm, Gemcitabine alone = control, but 16% could not crossover upon tumor progression because they had serious side effects from gemcitabine alone.
(Note: Of course, the Threshold Pharmaceutical example is not apples to apples with DCVAX-L, but it serves as an illustration).

http://www.thresholdpharm.com/pdf/supplemental_info_survival_pancreas_cancer_09-12.pdf


Inability to crossover should not be a problem (for patients) in the NWBO study (because the crossover arm does not require additional SOC after progression, and because DCVAX-l is essentially nontoxic), therefore it seems very likely almost all patients from the control arm who are alive at tumor progression will crossover.

The prior studies that have done this typically compare three different survival curves -- crossover arm, control arm and treatment arm. Because DCVAX-L is so safe, I don't think we will have a third OS control curve of any significance since maybe only 0-5% will not crossover.

Consequently, with OS, there will be crossover v. treatment arm. If we accept that PFS in the original control arm is likely to be somewhere around 9 months (this is higher than the 6.9 months or 8.1 months in prior trials because I am allowing for SOC improvement), then we can also assume the OS would have been (without crossover) as high as 17 months. This means at PFS = 9 months (control arm), there should be many alive in the crossover arm that might benefit from DCVAX-L. The issue that arises for patients and the review board is that it is likely the crossover arm will not do as well, because DCVAX-L does not work as well once tumor progression resumes active measurable progression. If this is born out in the curves, and it is statistically significant, the pressure to unblind the trial will be even stronger.

Going back to the Threshold Pharmaceutical example (which has three curves), the crossover arm still seemed to benefit from delayed treatment even though administration of TH-CR-404 was administered after progression resumed. (Again, this is an apples and oranges comparison, and it is not meant to compare the drugs...only to illustrate the crossover arm dilemma.)

Conclusion: Because the primary endpoint for DCVAX-L is PFS, OS will likely only be judged based upon two curves -- crossover arm v. treatment arm. Based upon past experience with DCVAX-L and what we can glean across 21 prior Dendritic therapy studies (some more alike than others), the curve is likely to demonstrate that delayed treatment is statistically more deadly. Consequently, this (delay) statistical significance may pressure the review board toward unblinding the trial earlier and moving everyone to the treatment arm.

Regardless, in my view, overall survival can not and will not be used to reflect negatively upon DCVAX-L efficacy. Why? Because OS will not include a meaningful control arm. Thus, the only meaningful OS comparison is S.O.C. treatment followed by immediate DCVAX treatment v. S.O.C. with delayed DCVAX treatment. Even if delay did not effect OS efficacy, it would not reflect negatively upon DCVAX-L treatment.

Thus, the only statistically significant endpoint that can reflect both.... either positively or negatively upon DCVAX-L is PFS. Because PFS is a much quicker and complete endpoint (compare against ICT-107 OS primary endpoint), the review board will have an easier time determining whether statistical significance is met. The review board will not be forced to extrapolate survival curves to determine efficacy. The only effect OS can really have on the review board is in the case if treatment delay with DCVAX-L proves deadly. If that is the case, it would likely place additional burden upon the review board to unblind the trial.