NorthWest Biotherapeutics Inc (NWBO)

[Pyrrhonian]

Flipper found some contrary info to this (which states a very low alpha stopping boundary at first interim of 0.0004) from the book Analysis of Clinical Trials Using SAS. In it the following were noted:

-First interim analysis at 50% total events (77 events); stopping boundary 0.0021

-Second IA at 75% total events (115 events); stopping boundary 0.0097 [almost 0.01]

-Final analysis of designated events (153 events) 0.021

And this also was O'Brien-Flemming. The trial outlined here mirrors NW's similarly except in the final alpha of 0.02, which is actually one-sided (0.05 two-sided). NW's trial was initially designed to be higher powered at 0.01 one-sided and 0.02 two sided. Those same assumptions were used and carried over in the BSSR. Which is interesting, and the point is this: the 0.0004 stopping boundary for first interim I had previously quoted seems linked to using a two-sided formula. The one in this book is one-sided. However, NW's further powering suggests that the TWO-SIDED stopping boundary for their trial is higher (less stringent) than 0.0004, and is most likely closer to 0.002. This may be a better estimate of stopping boundaries for NW's trial:

•Check variance assumptions at 80% enrollment
•Interim for efficacy at 60% and 80% of 248 PFS events
•O’Brien-Fleming rule
•Stopping boundaries: a1= 0.002; a2= 0.01

Another interesting thing is alphapuppy's analysis which showed the following:

Hypothesis 1: The Control Patients in the DCVax-L GBM trial should relapse at a rate similar to what was shown in the Stupp trial. The Stupp trial is the best GBM data available and represents the standard of care for GBM patients. In order to compare the Stupp Trial to the current DCVax-L trial control patients, however, one must review the eligibility requirements for each trial. The DCVax-L Trial only allows patients to enter the trial, if "surgical resection with the intent of a gross total or near gross total resection" can be completed. The Stupp Trial contained 17% patient who just had a biopsy only and another 44% of the patients who only had a partial resection. Extent of surgical resection is a major prognostic factor for these tumors and historical reviews have shown the degree which the various surgical procedures can change survival rates. The Median progression free survival (NYSE:PFS) for all patients in the Stupp trial was 6.9 months. When one applies algebra proportions to correct for the differences in prognosis between the trials in terms of surgical extent, the likely median PFS for the control patients in the DCVax-L trial should be around 9 months.

Hypothesis 2: Given that we know the current number of events in the total DCVax-L trial at this time (first interim analysis) is 67, then one can model the control arm PFS to be similar to the corrected Stupp trial results and thus obtain an estimate for the probable number of events required in the control arm of the DCVax-L trial to give a 9 month median PFS. When one does this with a Kaplan Meier plot, one obtains a result of a minimum of 30 events in the control arm with the total number of patients in the control arm of 57. Then one can deduce that the number of events in the experimental arm for the DCVax-L trial would be 67-30 or 37 events.

Hypothesis 3:Using the derived 37 events in the experimental arm, one can model the experimental arm to be similar to the Stupp trial results but, in this case only allow for just 37 events out of 114 total patients in the DCVax-L experimental arm. When one does this with a Kaplan Meier plot, the median PFS for the experimental arm returns > 17.5 month median PFS in the DCVax-L experimental arm. When one applies log-rank and Wilcoxon significance testing to the two curves one obtains a significant result of p=<.0001 and p =.00015 respectively.

His analysis assumed 171 enrolled by 12/'13. Probably close, imo. But in his analysis he used an "at least" parameter for placebo group, meaning if more evented than his minimal assumption then obviously PFS for tx goes up.

In summary, if tx PFS is over 7 or 8 months higher than placebo (which nearly doubles it), then the first interim stopping boundary of 0.002 would be crossed.