Patients treated with
neoadjuvant chemotherapy (NAC) who achieve a
pathologic complete response (pCR) can be identified preoperatively and can potentially be spared the morbidity of surgery. The objective of this retrospective study was to estimate the diagnostic accuracy of preoperative magnetic resonance imaging (MRI) in predicting pCR in patients with different molecular subtypes of
breast cancer and to provide a basis for the selection of surgical methods.
METHODS: A total of 177 women with a primary
tumor fulfilled the study criteria; 18 of these patients (10.2%) achieved rCR, and 21 (11.9%) achieved a pCR. MRI diagnosis of rCR was significantly correlated with pCR with a Spearman's correlation coefficient of 0.686 in the entire population. The sensitivity, specificity, accuracy, pCR predictive value (PPV), and non-pCR predictive value (NPV) were estimated to be 66.67%, 97.44%, 93.79%, 77.78%, and 95.60%, respectively. Statistically significant correlations between rCR and pCR were found in
Luminal B high Ki67% (P<0.001), HER2-positive (P=0.0035), and triple-negative (P<0.001) subtypes, but not in
Luminal A and
Luminal B low Ki67% subtypes. On univariate analysis, the
tumor characteristics significantly associated with both rCR and pCR were small
tumor,
lymph node metastasis (LNM) negativity, early clinical stage, high grade, high Ki67% index, and different molecular subtype. On multivariate logistic regression analysis, grade 3
tumors (P=0.013), Ki67% ≥40% (P<0.000), and stage I
tumor (P=0.006) were independently associated with rCR. However, grade 3
tumors (P=0.001),
triple-negative breast cancer (TNBC), and clinical stages I and II
tumors (P=0.003; P=0.030) were independently associated with the likelihood of attaining a pCR.
CONCLUSIONS: The overall accuracy of MRI in predicting pCR in invasive
breast cancer patients who received NAC was 93.8%. The performance of MRI differed among molecular subtypes, and the highest PPV was found in TNBC (100%) and
Luminal B high Ki67% (75%) subtypes.