NorthWest Biotherapeutics Inc (NWBO)

[hyperopia]

CherryTree, yes there are several statistical methods to mitigate crossover effect, but none are really ideal for this trial. The most common method I’ve seen used in oncology trials is the rank-preserving structural failure time (RPSFT).

The RPSFT model
The RPSFT model allows a direct comparison of randomization groups by adjusting the OS of patients who cross over so that it reflects the OS had they not received the investigational treatment. The method is related to the accelerated failure time model in OS analysis in which prognostic variables measured on the individual level are assumed to act multiplicatively on the time scale, for example, affecting the rate of progression. . .

The RPSFT model is rank preserving because a constant factor is used for adjusting the time to event for each patient. Thus, if two patients are on the same treatment (either control or crossover), and patient i fails (dies) before patient j, before adjustment, patient i will also always fail before patient j after adjustment: the ranking in failure times is preserved. The model is structural (causal) in the sense that it assumes a defined relationship between the observed event time and the event time that would have been observed if crossover had not occurred.
A key assumption of the RPSFT model is that the investigational treatment causes a constant reduction in time to death, assumed equal for all patients before and after progression. This may be a reasonable assumption in some cases but not in others, which may restrict the use of the method to cases in which a constant proportional reduction in the time to event is biologically plausible

.


Analyzing Overall Survival in Randomized Controlled Trials with Crossover and Implications for Economic Evaluation

https://www.sciencedirect.com/science/article/pii/S1098301514018907#ab0005