Up front, I am not an expert. That said, I will make a few observations about your questions. As a starting point, below is the topline for the high FRα subgroup.
In the pre-specified high FRα subgroup (218/366, 60%)
- PFS was longer in patients who received mirvetuximab soravtansine compared with chemotherapy (HR 0.69, p-value 0.049). Given that the p-value in the entire study population exceeded 0.05, the statistical analysis plan for the study required the p-value in the high subset to be less than or equal to 0.025 to achieve statistical significance.
- Confirmed overall response rate was higher for mirvetuximab soravtansine than for chemotherapy (24% vs 10%, p-value 0.014).
- Overall survival was longer in patients who received mirvetuximab soravtansine compared with chemotherapy (HR 0.62, p-value 0.033).
"Will it be possible to extrapolate from the full data and additional exploratory analyses from the Phase 3 FORWARD I study, as to what type of trial should be conducted in a new phase 3 trial (number of treatments/ level of Fra) and the possibility of FDA approval if the results remain similar in the new trial?"
Short answer is yes, and that is exactly what IMGN will do. They will take the n=218 sample from F1 and use it as a guide to determine the parameters for F3. N=218 is a good sample size and larger companies commonly use data sets of a similar size from phase 2 testing to determine the design for a potentially pivotal trial. Assuming the 2:1 randomization held true, the MS arm was n=145 and the control n=73 or thereabouts. IMGN wants as inclusive as possible an indication and yet maintain a high probability of a positive outcome. I suspect the parameters of F3 will be very similar to the definition of the F1 high FRα subgroup. Using the F1 results, they may change the level of marker expression they used to define exactly what constitutes "high." This is very important as MS actually underperformed the control arm if one only looked at medium expression level and outperformed in the high subgroup, so where that demarcation point between medium and high is placed, could have a profound impact on the end result. On choice of endpoint, IMGN can use either PFS, with a subsequent or concurrent OS analysis for the label, or they can use OS itself. PFS has the advantage of an earlier result whereas the F1 data showed a better survival hazard ratio and lower p value than the PFS analysis. So PFS is cheaper and sooner, but OS may be more of a sure thing if the F1 trends hold true. FDA will almost certainly accept either of these choices. After the F1 embarrassment, I would expect a straightforward statistical action plan with little or no alpha splitting so as to retain as much p value as possible for the final analysis. The F1 subgroup had a very marginal p=.049 PFS outcome. Look for a trial with at least an additional 50% increase in sample size over the n=218 observed in the F1 subgroup to provide some margin of comfort.