Accuracy of Post–Neoadjuvant Chemotherapy Image-Guided Breast Biopsy to Predict Residual Cancer

Importance: Image-guided breast biopsy of a residual imaging abnormality or tumor bed after neoadjuvant chemotherapy (NACT) is increasingly used to assess residual cancer, facilitate risk-adaptive surgery, and potentially identify exceptional responders in whom local therapy may be de-escalated. Objective: To further assess the accuracy of post-NACT image-guided biopsy to predict residual cancer in the breast. Design, Setting, and Participants: This diagnostic study analyzed multicenter patient-level data of patients with breast cancer who were treated with NACT and underwent image-guided biopsy before surgery at Royal Marsden Hospital in London, UK; Seoul National University Hospital in Seoul, South Korea; and MD Anderson Cancer Center in Houston, Texas. Data were analyzed from June to July 2019. Main Outcomes and Measures: Diagnostic accuracy of post-NACT image-guided biopsy. Final surgical pathology was used as reference standard. Results: Data from 166 women were analyzed. The median (range) age was 49 (25-76) years. The median (range) tumor size on pretreatment and posttreatment imaging was 33.5 (12-100) mm and 10 (0-100) mm, respectively. The overall pathologic complete response rate was 51.2% (n = 85) (16.1% [5 of 31] for hormone receptor-positive/ERBB2 (formerly HER2)-negative; 44.7% [21 of 47] for hormone receptor-positive/ERBB2-positive; 69% [20 of 29] for hormone receptor-negative/ERBB2-positive; and 66.1% [39 of 59] for triple negative). The majority (143 [86.1%]) underwent image-guided vacuum-assisted biopsy (VAB), and 23 had core-cut biopsy. The median (range) needle gauge was 10 (7-14), and the median (range) number of samples was 6 (2-18). When image-guided biopsy (VAB and core-cut biopsy) was representative (159 [95.8%]), the false-negative rate across the whole cohort was 18.7% (95% CI, 10.6%-29.3%). Subgroup analysis of patients with a complete/partial clinical response and residual imaging abnormality of 2 cm or smaller with at least 6 VABs taken (76 [45.8%]) demonstrated a false-negative rate of 3.2% (95% CI, 0.1%-16.7%), a negative predictive value of 97.4% (95% CI, 86.5%-99.9%), and an overall accuracy of 89.5% (95% CI, 80.3%-95.3%). Conclusions and Relevance: This large multicenter pooled data analysis suggests that a standardized protocol using image-guided VAB of a tumor bed measuring 2 cm or smaller with 6 or more representative samples allows reliable prediction of residual disease. These results could inform the design of de-escalation trials in NACT exceptional responders testing the safety of eliminating surgery.