MRI concentric partial response pattern as a predictor of pathological complete response in early triple-negative breast cancer treated with Keynote-522 neoadjuvant therapy.

Abstract

e12603 Background: In early triple-negative breast cancer (TNBC) current treatment involves neoadjuvant chemo-immunotherapy (NAC). Patients with residual disease after NAC are at the highest risk of recurrence. Previous studies have shown that tumor regression patterns in luminal breast cancer primarily exhibit concentric shrinkage, which is associated with pathological complete response (pCR). However, scarce studies in early TNBC have evaluated the correlation between imaging shrinkage patterns and pathological response. Therefore, further investigations to identify parameters that enable monitoring of response prior to surgery would add significant value to the strategy of personalized treatment. Methods: This single-institution study included patients with TNBC treated at the Alexander Fleming Institute who completed the Keynote-522 regimen. MRI was performed using a GE Voyager Air Edition 1.5 T system with a 16-Channel Breast XT Package and NeoCoil Breast Coil. Imaging evaluations were standardized and conducted by the same operator. Responses were categorized as no response, complete imaging response, or partial response. Partial responses were classified into concentric (uniform shrinkage) or fragmented/diffuse (non-uniform shrinkage). Pathological response was evaluated using the Residual Cancer Burden (RCB) scale. Statistical analyses were performed using IBM SPSS, with significance set at p < 0.05. Results: The study included 50 patients treated with the Keynote-522 regimen. The mean age was 46.38 years (SD 11.05). The mean tumor size was 35.98 mm (SD 18.4) by MRI. Lymph node involvement was present in 39.6% before treatment. A pathological complete response (pCR) was observed in 74% of patients. Regarding imaging responses, 44% of patients achieved a complete imaging response, and among these, 78% had a pathological complete response (pCR) with RCB 0 in the surgical specimen. A total of 52% of patients showed partial responses on MRI, categorized as 18 concentric and 8 fragmented; of these, 18 had a pathological RCB 0. In our cohort, larger tumor size was associated with a higher likelihood of fragmented partial responses (p = 0.03). Concentric partial imaging responses were significantly associated with higher pCR rates compared to fragmented responses. Among patients with concentric responses, the pCR rate was 83.3%, whereas for fragmented responses, it was 37.5% (p = 0.035). Conclusions: In our MRI assessments, concentric tumor shrinkage emerged as a strong predictor of complete pathological response. Identifying pre-surgical parameters to differentiate responders from non-responders could significantly enhance personalized treatment strategies. Further validation through larger cohorts is essential, and AI may support earlier and more precise recognition of these imaging patterns.