Core Needle Biopsy–Final Histopathology Discordance in Breast Phyllodes Tumors: Clinical, Radiological, and Ki-67-Related Factors

Background: Phyllodes tumors (PTs) are rare fibroepithelial breast neoplasms that can be difficult to distinguish from other breast lesions using core needle biopsy (CNB). This study aimed to evaluate discordance between preoperative CNB findings and final surgical histopathology, identify associated clinical and radiological factors, and explore differences in clinicopathological characteristics between benign and borderline/malignant PTs. Methods: This retrospective, single-center study reviewed 120 patients who underwent CNB followed by surgical excision between 2008 and 2024. After excluding 50 cases retrospectively classified as B3 lesions and eight cases with insufficient data, 62 patients were analyzed, including 34 with PT and 28 without PT on final histopathology. Among PTs, 27 were benign and seven were borderline/malignant. CNB diagnostic performance, clinical and radiological characteristics, and Ki-67 expression were evaluated. Multivariable logistic regression was used to explore factors associated with a final diagnosis of PT. Results: Within this selected surgical cohort, only 7 of 34 PTs (20.6%) were identified as suggestive of PT on CNB, yielding a sensitivity of 20.6% and specificity of 100%. Median tumor size was greater in the PT group than in the non-PT group (55 vs. 25 mm; p < 0.001) and remained associated with PT after adjustment (OR = 1.034; 95% CI, 1.004–1.065; p = 0.025). Within this selected surgical cohort, BI-RADS ≤3 was associated with final PT diagnosis compared with BI-RADS ≥4 (adjusted OR = 5.628; 95% CI, 1.575–20.109; p = 0.008). However, this counterintuitive association may reflect surgical selection and verification biases and should not be interpreted as evidence of clinical predictive utility. Continuous Ki-67 values did not differ significantly between PT and non-PT lesions (p = 0.307), whereas median Ki-67 was higher in borderline/malignant than benign PTs (10% vs. 0%; p = 0.004). Conclusions: Within this selected surgical cohort, preoperative CNB frequently failed to identify PTs, highlighting the limited generalizability of the observed diagnostic performance estimates to unselected breast biopsy populations. Tumor size was associated with final PT diagnosis, whereas the inverse association observed for BI-RADS category should be considered an exploratory, cohort-specific finding rather than evidence of clinical predictive utility. Ki-67 showed an exploratory, hypothesis-generating association with PT histological grade, requiring validation in larger cohorts. These findings support cautious interpretation of CNB results alongside clinical and radiological characteristics. The retrospective design, exclusion of B3 lesions, and potential selection biases limit generalizability.