Post-neoadjuvant imaging for prediction of pathologic complete response to KEYNOTE 522 in patients with early triple-negative breast cancer.

Abstract

e13153 Background: The KEYNOTE 522 trial established the role of neoadjuvant chemotherapy with pembrolizumab in patients with clinical stage II-III triple negative breast cancer (TNBC). We sought to assess real world KN522 pCR rates in a diverse cohort of patients with TNBC and investigate the ability of imaging to predict pathologic complete response (pCR). Methods: Retrospective analysis of women with non-metastatic TNBC who received neoadjuvant KN522 and had surgery at a single academic institution from 2021-2023 was conducted. KN522 was considered incomplete if chemotherapy or pembrolizumab were terminated early or cycles missed; dose reduction (DR) was assessed separately. Multivariable logistic regression examined factors associated with radiologic and pCR (ypT0/Tis N0). Standard 2x2 tables determined the positive and negative predictive values (PPV, NPV) of MRI and mammogram (MGM)/ultrasound (US) for treatment response. Results: Of 81 patients, 62% were NH White and 33% were NH Black. Median age was 55 (IQR 46-62), 30% lived in the highest Area Deprivation Index quintile, and nearly half had BMI ≥ 30. Only 21 (26%) completed all cycles of neoadjuvant KN522 without DR; 38 (47%) had early termination or missed cycles. 37% experienced any grade irAE. Median pre-treatment tumor size on diagnostic imaging was 3cm (IQR 2.2-4.4); for cN+ patients, 66% had > 1 abnormal lymph node (LN). Overall pCR rate was 59%, with highest pCRs in patients who completed pembrolizumab with incomplete chemotherapy +/- DR (100%), and those who completed both agents +/- DR (71%). On multivariate analysis, pCR was less likely with cN3 (OR 0.03, p = 0.026) and ECOG 1 (OR 0.14, p = 0.031), but not an incomplete or DR regimen. 58% (25/43) of the cN0 patients achieved pCR, and of these, 64% (16) had radiologic CR. 50% (19/38) of the cN+ patients had pCR, and 63% (12) of these had radiologic CR. Radiologic CR was less likely with increasing cT stage and early termination of pembrolizumab. BMI overweight (OR 19.1, p = 0.033) and irAE (OR 5.77, p = 0.036) were associated with incomplete pembrolizumab; NH Black race (OR 3.84, p = 0.045) was associated with any irAE. Overall, the PPV of post-KN522 imaging for pCR was 92.9%; the NPV was 58.5%. The PPV and NPV of MRI for pCR were both high at 83.3%. In cN+ patients, MRI had greater PPV for axillary pCR compared to MGM/US (88.9% vs. 75%). Conclusions: In the real-world application of KN522, our pCR rate aligns with clinical trial data and pCR is accurately predicted by post-neoadjuvant imaging. MRI should be considered in cases where this imaging will inform surgical decision-making, particularly breast conservation vs. mastectomy. Further, our finding that incomplete or DR regimens still result in pCR may allow patients to proceed to surgery if interval imaging suggests radiologic CR prior to KN522 completion.