ONCY At the end of 6 weeks, the CRO would have done a "data scrub" on all 80 patients. This would have covered off all of the demographics, the background clinical information and loose bits on the ITT patient group. The evaluation at 12 weeks should involve the collection of the interim clinical data, two blinded evaluations of the RECIST evaluations and delivery to the statistician(s). The statistician looks at the single (unblinded - all 80 patients) curve and compares these data to the data expected from a very well performing (upper percentile survival/progression) historical "control" group. From this statistical analysis comes the stats about clinical effect of Reolysin. The curve for all 80 patients would contain 40 "control" patients and 40 patients treated with Reolysin and, blinded, this 80 patient curve should look moderately better than a control curve. It is the job of the statistician to opine about the likelihood that unblinding the trial (80 patients) would produce a observable difference between treatment (40 patients) and control group (40 patients).
The language of the statistical analysis provided to the company is intentionally non-committal. For example, the company may hear many options......"there is an 80% chance that the data will resolve to (at least) a 6 week improvement in PFS"; "there is a 90% chance that the data will resolve to (at least) a 4 week improvement in PFS": "there is a 95% chance that the data will resolve to (at least) a 2 week improvement in PFS. In fact, the statistician would produce a pretty, curved dataset plotting an observable PFS difference vs. likelihood. There would be many iterations of the data provided to the company.......all before unblinding. The company looks at the data and decides about unblinding......or not. The CRO collects and scrubs the data; the stats person tortures the numbers and provides a "best guess" about the trends of the unblinded data; the company decides on the unblinding and subsequent statistical analysis.
The devil is in the SPA details. At what point is the company forced to unblind the trial? At what point will the company unblind the trial? We have heard 6 weeks, 12 weeks and, yet we are now past 18 weeks. To play the devil's advocate, if the stats person came to you after the 12 week scans and said that there is an 80% chance that the data are currently stat sig for a 6 week PFS difference (doubling of PFS), would you risk unblinding the trial? In a trial trending toward success, you are taking a 20% chance of negative stats (often spuriously negative) and a resultant futility decision that torpedoes the entire Ph III H+N program. What numbers convince you to unblind? How much risk to the entire Ph III program do you assume to see the data early?
As the data mature, so do the stats. At the 18 week scan, the statistician would have tighter confidence intervals and would be able to say that there is a 98% chance of a 6 week PFS improvement (maybe a 99.5% chance of a 4 week difference). Would you unblind now?? We do not know the constraints or flexibility built into the H+N SPA but this is the gist of the decision making process.
Once the company makes the decision to unblind, they will have to live with the results as they stand. They direct the CRO to break the blind; the unblinded data comes to the statistician and two curves are generated and compared. If the difference between the two curves is sufficient (ie: >70% likelihood of success with 400+ patients enrolled), the SPA allows for continuation of the trial. Anything less.....the trial is discontinued on the basis of futility.
I don't know how much flexibility the SPA allows for......no matter what, the survival data are rapidly maturing at this point. The PFS data determine the go vs. no-go decision but what happens if the raw survival data are ready to report? Not much interpretation on those numbers and, at this point, more than half of the ITT population has died. If the trial were unblinded today, there would be very, very strong survival data already. Survival trumps PFS but what would the FDA say if the PFS is weak but the right side of the survival curve is strong?
It is getting interesting.....the screws are turning tighter.
For most of the biotechs that I have followed over the years, delays in data release portends bad results. I hope that this trend does not hold true for the ONCY H+N data!! No matter what, the lifespan of this group of patients enforces the duration of the trial. In October (6 weeks from now), 6m survival stats are available on all 80 patients. Effectively, the statistical package would be complete. Not much ducking that milestone event!!
Dog days of August!!!
Cheers, Geneman
