@Jdcramer has made a critical point below about KEYNOTE-689.
The EFS rules in KN-689 require careful review to interpret these results. A protocol-specific event was determined by blind independent central review (BICR), but if growth in the pre-surgical CT was perceived to be a “flare” (i.e. potential progression), surgery was supposed to proceed unless unresectable (protocol quote below). In order to be considered an event, repeat imaging was required 4-8 weeks later, and obviously if the tumor was still resectable the patient would go to surgery before then.
This definition means it was extremely difficult to have an event before surgery, even if the tumor grew and the surgery was more extensive than initially anticipated. Thus some patients may have been harmed by neoadjuvant treatment, but we would not see that in the event data.
Thus there is this important disconnect: 82 patients in the pembro arm stopped the drug due to progression (by the investigator, as shown in the CONSORT diagram), even though there were only 69 progression events (by BICR, used for endpoint analyses).
Is perioperative immunotherapy doing something favorable and important in a subset of these patients? Yes, and hopefully we can refine its use to the right population and with the right regimen (adding neoadjuvant chemo or RT?).
Is it distinctly possible/probable that some patients are progressing on neoadjuvant pembro, leading to a worse outcome in some domain, but we cannot see that in these data? Yes, unfortunately, also true.
Question for those who’ve dug into KEYNOTE-689:
If 43 pts had early ‘clinical progression’ (Table S2B), why isn’t there an early drop in EFS with neoadjuvant pembro?
How do you reconcile this?