You are right Antihama and since I wrote my post I had thoughts similar to yours except that I will go even further than you and because I now believe that the survival results of the 35 patients who were never treated with L, should be combined only with the results of the 64 member crossover group and not at all with the Treatment group.
Now let me explain my rationale. NWBO had no problem in carrying out the early L treatment of the 232 because all of those patients were on trial only about 3 months after their chemoradiation and were in adequate shape to receive their L treatment. Witness the fact that they ended up with all 232 Treatment patients which is exactly 2/3 of the original 348 patients who were originally planned to be in the Treatment group.
The situation with the 99 intend to crossover group may have been quite different. Many of those patients were perhaps not in good shape because once they progressed, they progressed rapidly and it was too late for them to be treated with L. Such patients may have been subject to salvage treatment. Also of those 35 patients there may have been some that were in adequate shape but for one reason or another received a different treatment. Finally, perhaps a few of those 35 never progressed.
I believe that those 99 should be included in one "intend to crossover" group.
and should be compared to external controls that have progressed. This intend to crossover group may have an mOS which may not differ greatly from that of the matched external control, but I bet that after 36 months on trial we will see a hell of a difference which favors our trial's 99 patient intend to crossover group.
I base the expected superiority of the 99 patients in the intend to crossover group's on the longer and fatter tail which this group will display. From the overall survival curve in the NYAS presentation, I estimated that the number of Treatment survivors at 33 months past randomization (36 months past surgery) will come to about 58 patients. The 2018 SNO presentation estimated that 36 months past surgery (33 months past randomization), 87 patients were still at risk. I am estimating that at least 80 such patients still existed 3 months later (at 36 months past randomization). The November 2018 SNO estimate was perhaps made a few months earlier than November but at that time probably all the trial's living patients were at least 30 months on trial and that estimate was therefore pretty accurate.
So of at least 80 post 36 months survivors, I have estimated a total of about 58 post 36 months survivors in the Treatment group. Therefore at least about 22 would be found in the group of 99 intend to cross over group.
This is just an estimate and we may perhaps have fewer 36 month survivors in the group of 99 but that could still be a robust number to be found among the 64 crossovers.
The results for the crossover group may turn out to be excellent and I hope that the JA validates my estimate.
Market Data 
Markets 
AI
