BACKGROUND: Substudy 02C of the phase I-II KEYMAKER-U02 study was designed to evaluate neoadjuvant pembrolizumab as monotherapy and in combination with investigational agents, followed by surgery and adjuvant pembrolizumab in participants with resectable stage IIIB-D melanoma. Here, we report a post hoc analysis of concordance of pathologic response assessment between local and central review and among central reviewers in the KEYMAKER-U02 substudy 02C. PATIENTS AND METHODS: Participants received neoadjuvant pembrolizumab plus vibostolimab, pembrolizumab plus gebasaxturev, or pembrolizumab monotherapy followed by resection and adjuvant pembrolizumab. Pathologic response was categorized by percent age of residual viable tumor (%RVT) in the treated tumor bed as follows: pathologic complete response (pCR; 0% RVT), near-pCR (RVT >0% but ≤10%), partial pathologic response (pPR; RVT >10% but ≤50%), and pathologic nonresponse (pNR; RVT >50%). RESULTS: By local review of 60 participants, 24 (40%) had pCR, 7 (12%) had near-pCR, 12 (20%) had pPR, and 17 (28%) had pNR. By central review of 60 participants, 23 (38%) had pCR, 7 (12%) had near-pCR, 14 (23%) had pPR, and 16 (27%) had pNR. The weighted kappa for concordance between local and central review for these categories was 0.79 (95% CI 0.70-0.89). The three central reviewers agreed on the response category for 54 of 61 participants (89%). For the seven participants (11%) without agreement, the maximum difference in response was one category; for four of these participants, the dissenting response assessment was close to a %RVT category cut point. The weighted Gwet's agreement coefficient was 0.94 (95% CI 0.90-0.98). CONCLUSION: These results confirm a high level of concordance in pathologic response assessment between local and central review and among central reviewers and provide confidence in the accuracy of pathologic response assessment following neoadjuvant therapy.
Tetzlaff et al. (Tue,) studied this question.