Patients’ characteristics
The characteristics of patients in the present cohort are summarized in Table 1. This study included 34 patients; all were female. The median age at the time of initial diagnosis was 53 (range: 31–77) years. All patients were diagnosed with TNBC based on the biopsy specimens; 31, 2, and 1 patients were diagnosed with invasive carcinoma of no special type, apocrine carcinoma, and invasive lobular carcinoma, respectively. Ten patients were initially diagnosed with clinical stage I, 10 with stage IIA, 9 with stage IIB, 2 with stage IIIA, 1 with stage IIIB, and 2 with stage IIIC, respectively, according to the 8th Union for International Cancer Control TNM Classification [20].
MRI findings before and after NAC
The most common enhancement pattern before NAC was solitary (20 patients, 58.8%), followed in order by separated (7 patients, 20.6%), grouped (5 patients, 14.7%), and replaced (2 patients, 5.9%) lesions (Fig 1A–D; Table 2). None of the patients in the present study showed mixed enhancement patterns.
Table 2 summarizes the tumor shrinkage patterns after NAC. The most common shrinkage pattern on MRI was type I (14 patients, 41.2%), followed in order by non-visualization (9 patients, 26.5%), type III (5 patients, 14.7%), type II (4 patients, 11.8%), and type IV (2 patients, 5.9%) (Fig 2A–E; Table 2). For the association between the initial findings and shrinkage patterns in MRI, the most common shrinkage pattern (type I) was observed in 60% of the patients with solitary patterns, followed by non-visualization, which was observed in 30% (Table 2). Patients with non-visualization patterns after NAC initially had either solitary or separated enhancement patterns. MRI shrinkage patterns did not significantly correlate with the clinicopathological parameters, including the menopausal status, clinical stage of the tumor, and Ki-67 labeling index.
Table 2. Correlation between initial contrast-enhancement patterns prior to neoadjuvant chemotherapy and shrinkage patterns
|
|
Initial enhancement patterns
|
Shrinkage patterns
|
Solitary
|
Grouped
|
Separated
|
Replaced
|
Non-visualization
|
6
|
0
|
3
|
0
|
Type Ⅰ
|
12
|
1
|
1
|
0
|
Type Ⅱ
|
1
|
2
|
1
|
0
|
Type Ⅲ
|
1
|
2
|
2
|
0
|
Type Ⅳ
|
0
|
0
|
0
|
2
|
|
|
|
|
|
Histopathological features following NAC
The most common histopathological regression grade was 1 (12 patients, 35.3%), followed by grades 5 (11 patients, 32.4%), 2 (6 patients, 17.6%), 4 (4 patients, 11.8%), and 3 (1 patient, 2.9%) (Fig 3A–D; Table 3). The most common shrinkage pattern on histopathology was type I, observed in 38.2% of the patients, followed in order by non-visualization (32.1%), type III (20.6%), type II (5.9%), and type IV (2.9%) (Table 3). The observed concordance between the shrinkage patterns on MRI and on histopathology was 41.2% (Kappa statistics: 0.181, p=0.07). Subsequently, we performed a direct comparison between MRI-based and histopathological shrinkage patterns in TNBC. Among 11 patients with no residual disease, six were diagnosed as having non-visualization pattern on MRI (sensitivity, 54.5%). In the other five patients, type I MRI shrinkage pattern was observed in 1 patient, type II in 2, and type III in 2 other patients (Table 3). Among the 23 patients who had residual disease, three did not show contrast-enhanced lesions on MRI (specificity 87.0%). Pathological shrinkage patterns did not significantly correlate with the clinicopathological parameters, including the menopausal status, clinical stage of the tumor, and Ki-67 labeling index.
Table 3. Correlation between MRI shrinkage patterns and pathological shrinkage patterns following neoadjuvant chemotherapy
|
|
|
Pathological shrinkage pattern
|
|
|
|
Non-
visualization
|
Type Ⅰ
|
Type Ⅱ
|
Type Ⅲ
|
Type Ⅳ
|
MRI shrinkage pattern
|
|
|
|
|
|
Non-visualization
|
6
|
2
|
0
|
1
|
0
|
Type Ⅰ
|
1
|
8
|
1
|
3
|
1
|
Type Ⅱ
|
2
|
0
|
0
|
2
|
0
|
Type Ⅲ
|
2
|
2
|
1
|
0
|
0
|
Type Ⅳ
|
0
|
1
|
0
|
1
|
0
|
|
|
|
|
|
|
Relationship between MRI and histopathological tumor sizes after NAC
Next, we assessed the residual tumor size on MRI and on histopathology after NAC in 25 patients with TNBC. There was a significant correlation between MRI tumor size and histopathological tumor size (Pearson’s correlation coefficient of 0.89, p <0.0001) (Fig 4). The mean difference between MRI-based and histopathology-based tumor diameter was 11.5 ± 12.7 mm. The following equation was developed based on the findings:
This equation can be applied to patients with MRI shrinkage patterns that are reflective of type I, II, III, and IV (excluding non-visualization). “Pathological residual tumor size” implies the total tumor size including invasive carcinoma and intraductal carcinoma components. If a resection size was estimated based on this equation before surgery, a safe margin of 10 mm beyond the calculated pathological size would lead to a positive margin rate of 4%. Moreover, a 20-mm safe margin would result in a positive margin rate of 0%.