This retrospective study included patients with primary breast cancer who were treated with NAC from Jan 2014 to Nov 2020 at the Department of Breast Surgery of the First Affiliated Hospital of China Medical University and the Cancer Hospital of China Medical University.

All patients received a minimum of one cycle of NAC ahead of surgery. Patients with cancer in situ were excluded, as were patients with invasive breast cancer before NAC could be incorporated in cohort. Patients who received any kind of treatment prior to NAC or who presented with progressive or metastatic breast cancer were excluded. Patients with previous breast cancer, male patients, and those with synchronous invasion, bilateral, or inflammatory breast cancer were also excluded. 68 cases with incomplete or deficient IHC analysis were also excluded. In total, 828 patients met the above restriction standards and were included.

This retrospective study received permission from the institutional review board (IRB) of the First Affiliated Hospital and was in accordance with the Helsinki Declaration. All patients involved in the research gave informed consent in written agreements of specimens used for scientific research. The informed consent of retrospective research involvement could be waived based on the retrospective nature of the study. Pre-NAC core needle biopsy pathology and post-surgery regular pathology was extracted and saved in a database. Patient characteristics were collected including gender, age at diagnosis, body mass index (BMI), maximum tumor diameter, tumor grade and stage, axillary lymph node status, histologic type, NAC schedule and cycle number, histology grade, and clinical response to NAC. The local extent of breast cancer was measured via breast ultrasound, mammography, breast MRI, chest CT, bone scan and/or hepatobiliary and splenic ultrasound to verify distant metastasis. The final size of local breast cancer in our database was adopted following priorities: breast MRI > breast ultrasound > mammography. Every patient with suspicious lymph-node metastasis suggested by imaging examinations underwent ultrasound-guided core biopsy of ultrasound-graphically abnormal nodes for axillary node metastasis confirmation before starting NAC. The final histological assessments were all analyzed using hematoxylin and eosin staining and immunohistochemical (IHC). The histological type of specimens from incorporated patients was distinguished between two subgroups: general invasive breast carcinoma of no special type (IBC-NST) and Others. The latter group contained special subtypes such as lobular, mucinous and tubular carcinomas with ≥90% of the tumor as a pure special tumor type and mixed IBC-NST with special subtypes. Lymph nodes with micro-metastasis were considered positive, including the maximum diameter over 0.2 mm or the number of tumor cells over 200. Isolated tumor cells were considered negative.

Pathological complete response (pCR) was defined as no residual invasive tumor upon hematoxylin and eosin evaluation of the complete resected breast specimen and all sampled lymph nodes (noninvasive breast residuals) (ypT0/is, ypN0).

Histopathology is regarded as the gold standard for diagnosis. All breast tumor specimens were acquired from core needle biopsy or surgical resections, and every specimen was affixed into formalin-fixed, paraffin-embedded tissue sections for preservation. IHC staining was performed using Dako Autostainer Plus and EnVision Dual Link detection reagent (DAKO; Carpinteria, CA) with DAB (Dako). Biomarker status, including ER, PR and HER2, were defined by IHC in strict accordance with European Quality Assurance guidelines. ER and PR staining were assessed based on the American Society of Clinical Oncology/College of American Pathologists Guidelines [27,28]. Antibodies used in IHC include as anti-ER (Clone SP1, Dako), anti-PR (Clone PgR636, Dako), and Ki67 (Clone Mib-1, Dako). Hematoxylin II (Dako As Link 48) was used to counterstain specimens automatically. All tests incorporated external positive and negative controls. ER and PR stains were considered positive if immunostaining was seen in more than 1% of immunoreactive cells. HER2 status was ascertained via IHC using the Hercep Test ™kit (code K5204, Dako). HER2 expression was scored as 0, 1+, 2+, and 3 + according to ASCO guidelines. A score of 3+ was regarded as HER2+, with 0/1+ defined as HER2-. For cases scoring HER2 2+, a fluorescent in situ hybridization (FISH) test could be conducted. The measurement of Ki67 index was based on the spot with the highest intensity in a high-power field (400x) and 500–2000 cells were counted [29].

Following the St. Gallen guidelines 2013 [2], high expression of PR was set as ≥20% and low expression of PR was defined as <20%. In accordance with the International Ki67 in Breast Cancer Working Group [30], Ki67 index was classified into two groups: low (<30%), and high (≥30%). To quantify the extent of Ki67 changes between pre- and post-NAC timepoints, we used the following equation: [define post-NAC Ki67 as A, define pre-NAC Ki67 as B, computational formula: ΔKi67%=(A−B)/B × 100%, maintaining sign]. If a patient achieved pCR after NAC, the post-NAC Ki67 index was defined as 0% and ΔKi67% was mathematically −100%. Representative IHC staining images of Ki67 subgroups are shown in Supplementary Figure S2.

We classified breast cancers into four subtypes based on HR status, HER2 status and Ki67 index according to the St. Gallen guidelines as follow [2,31]: Luminal A: (ER and PR positive, HER2 negative, “low” Ki-67, and a “low” recurrence risk based on multi-gene-expression assay results if available), Luminal B [“Luminal B-like (HER2 negative)”: ER positive, HER2 negative, and at least one of the following: “high” Ki-67, “negative or low” PR, or “high” recurrence risk based on multi-gene-expression assay if available. “Luminal B-like (HER2 positive)”: ER positive, HER2 over-expressed or amplified with any Ki-67, and any PR]. HER2-enriched (HER2 over-expressed or amplified, ER and PR absent) and TN (Negative ER, PR and HER2). The cut-point between “high” and “low” values for Ki67 varies and lacks conclusion. 14% cut-off value of Ki67 for subtype classification was adopted in St. Gallen guidelines 2013.

Multiple demographic features were analyzed using the Chi-square test. The survival-related indicators studied were DFS and OS. DFS was calculated from the date of initiation of the first regimen to the date of first event (locoregional relapse, distant relapse, or death) and OS was calculated from the date of surgery to the date of death or last follow-up. The Kaplan–Meier method was used to define the difference ratio, and survival curves were compared using the log-rank test [32]. Significance was assigned as p value < 0.05. Receiver operating characteristic curve (ROC) and area under the curve (AUC) were performed to calculate the optimal cut-off value determined by the Youden index with maximum sensitivity and specificity. Cox proportional hazards regression models were used to estimate relapse and survival risk between subgroups. The multivariate Cox proportional hazards model was implemented for Hazard Ratio (HR) and 95% confidence intervals (CI) to identify independent prognostic factors. Net reclassification improvement (NRI) was used to verify classification accuracy. All statistical data analysis was performed using SPSS 26.0 (SPSS Inc., Chicago, IL, United States) and R programming language (version 3.5.3; https://www.r-project.org/).

Patients with primary breast cancer who were treated with NAC were selected based on strict standards. 942 patients were initially included in the cohort, but 114 patients were eliminated for various reasons. A total of 828 female patients with primary breast cancer who received NAC were ultimately included in this retrospective study (Figure 1).

Basic demographic and pathologic features are shown in Table 1. The median age of entire cohort was 51 ± 9.65 years old (range: 23–76 years), of which 10.0% of patients were under 40 years old at diagnosis. Body mass index was used to distinguish subjects: overweight patients with index greater than or equal to 24.9 accounted for 48.2%, underweight patients with index under 18.9 accounted for 8.2 and 43.6% of patients had a healthy BMI between 18.9–24.9. Breast cancer pathological subtype was confirmed as invasive carcinoma of NST for 84.4% of patients. All other histological types represented 15.6% of the full cohort. A median of 4 NAC cycles were received (range: 1–9), with NAC classified into three groups: texane-based (10.7%), anthracycline-based (18.1%) and texane + anthracycline (71.1%). The average maximum tumor diameters before and after NAC were 3.41 ± 1.651 and 2.26 ± 1.648 cm, respectively. 81.9% patients presented with node-positive status at diagnosis. Finally, 138 subjects (16.7%) who received NAC achieved pCR, a commonly used measurement of NAC efficacy.

We next analyzed IHC biomarkers. Patients with ER positivity made up 59.5% of the cohort. Patients with PR < 20%, or negative status, represented 62.8% of the cohort. Based on strict IHC staining, subjects positive for HER2 accounted for 31.5% of cases 87.6% of all cases had Ki67 expression ≥30% before NAC, while only 58.9% of the cohort had ≥30% Ki67 after NAC. 18.8% of HER2-positive patients received anti-HER2 therapy.

Based on biomarker status, patients were categorized into four subtypes: 43 subjects (5.2%) were categorized as Luminal A subtype breast cancer, 489 subjects (59.1%) had a Luminal B breast cancer, 148 subjects (17.9%) had HER2-enriched breast cancer, and 148 cases (17.9%) had TN breast cancer. IHC status and subtypes distribution are shown in Table 1.

We chose pCR as our evaluation criterion of pathological response to NAC. The median follow-up time was 62.00 ± 21.43 months. A associations between patient features and pathological response to NAC or survival were assessed via the Chi-square test (χ2), with results shown in Table 2. Age, body mass index, maximum tumor diameter before NAC, and HER2 status all had p values > 0.05, indicating no significant influence on prognosis. However, different carcinoma pathology subtypes, maximum diameter after NAC, and nodal status at diagnosis were all significantly associated with pCR, with all p values < 0.05. The IHC biomarkers ER, PR and Ki67 status before and after NAC, all associated with pathological response to NAC. The log-rank test was used for in analysis of different parameters with DFS and OS fully considering the follow-up time (Table 3). For DFS and OS, BMI and Ki67 after NAC both presented p value < 0.05.

Histological subtype further affected both pCR rates and survival status. Each subtype resulted in different pathological responses, as validated by the Chi-square test (p < 0.05), but there was no obvious change on patient prognosis (p = 0.244 > 0.05). Rates of pCR across subtypes are shown in Figure 2. The left side of the picture presents the relation between pCR and DFS. The right side of the picture presents the univariate analysis of the relationship between OS and pCR based on the Kaplan-Meier method. We calculated survival rates for patients with each subtype using the Kaplan–Meier method. pCR status had no impact on DFS outcome for patients with Luminal A (p = 0.784 > 0.05, Figure 2A) or Luminal B subtypes (p = 0.427 > 0.05, Figure 2B). However, both HER2-enriched and TN breast cancer subtype patients showed a significant association between pathological response to NAC and survival outcome (p = 0.043 < 0.05, and p = 0.042 < 0.05, respectively, Figures 2C,D). The corresponding Kaplan-Meier curves of pCR with OS outcome for four subtypes are presented in Figures 2E–H. The multivariate Cox analysis for the full cohort is available in Supplementary Table S1. The visual result of univariate analysis displayed via Kaplan-Meier curves is in Supplementary Figure S1.

The average Ki67 status before NAC was 39.66% ± 22.61%. In comparison, the average value after NAC was 25.00% ± 22.91%. The downward trend of Ki67 before and after NAC is displayed in Figure 3 for the whole cohort (A), Luminal subtype (B), HER2-enriched subtype (C), and TN subtype (D). As shown in the figure, Ki67 change before and after NAC presented a statistical difference (p < 0.001) in Luminal subtype. However, in HER2-enriched and TN breast cancer, the difference of Ki67 represented no statistical significance. To evaluate the degree of Ki67 decline, we compared post-NAC and pre-NAC proliferation indices [define post-NAC Ki67 as A, define pre-NAC Ki67 as B, the computational formula was Δ Ki 67 % = ( A − B ) ÷ B × 100 % , maintaining sign].

We next used ROC curve analysis to determine the optimal cut-off value for Δ Ki 67 % . Performing the calculations on SPSS 26.0, we found that Δ Ki 67 % had prognostic efficacy on survival outcomes among the full cohort. Based on the ROC calculation results in all patients, we defined an optimal cut-off of Δ Ki 67 % ≤ − 63 % (p < 0.05). The Δ Ki 67 % cut-off point in the Luminal B subtype was “−63%” (p = 0.047). Meanwhile the cut-off point in the TN breast cancer subtype was “−68%” (p = 0.009) (Table 4).

Therefore, we defined a reduction of greater than 63% ( Δ Ki 67 % ≤ − 63 % ) as Δ Ki 67 % positive, and a reduction of less than 63% ( Δ Ki 67 > − 63 % ) as Δ Ki 67 % negative. Positive Δ Ki 67 % status presented a larger magnitude of change for Δ Ki 67 % between pre- and post- NAC, with negative status showing opposite. We used Chi-square tests to evaluate the relationship between demographic and pathological features and Δ Ki 67 % status (Table 5). Δ Ki 67 % -positive patients represented 37.1% of the full cohort. Histological type, number of chemotherapy cycles, type of chemotherapy regimen, clinical nodal status, molecular subtypes, pre- and post-NAC tumor size, and Ki67 all showed statistically significant relationship with Δ Ki 67 % status (p values < 0.05).

In summary, Δ K i 67 % -positive status related with better survival outcomes. We used the Kaplan-Meier method to affirm the correlation between Δ Ki 67 % status and survival outcomes in each molecular subtype. Survival curves of DFS and OS based on Δ K i 67 % status for the four subtypes are displayed in Figures 4A–H. Similar to Figure 2, the left side of the figure presents the Kaplan-Meier curves related with DFS and Δ K i 67 % , with the right side presenting the relationship between OS and Δ K i 67 % . As shown in the figure, the univariate log-rank test demonstrated Δ K i 67 % was statistically significantly related to DFS and OS in Luminal B and TN subtype, while it showed no definite effect in Luminal A and HER2-enriched subtypes (all p > 0.05).

Based on the multivariate Cox analysis, Δ K i 67 % status is a significant independent prognostic predictor of survival outcome regardless of DFS and OS, with DFS-HR = 3.495 (95% CI 1.723–7.088, p = 0.001) and OS-HR = 23.024 (95% CI 2.956–179.333, p = 0.003) for the Luminal B subtype (Table 6, corresponding forest plot in Figure 5.

Not only that, we tentatively continued to explore the subgroups of Luminal B patients based on the HER2-status. In Luminal B patients from our research, who with negative HER2 status were 256 (52.4%). The patients with positive HER2 status were 113 (23.3%). The patients with unknown HER2 status were 120 (24.5%). The Kaplan-Meier curves and multivariate-analysis results of relationship between ΔKi67% and survival outcome were attached in the Supplementary Presentation S1. Based on the statistical calculation, the analytical results of all subdivisions in Luminal B tumors fully supported the statistical significance of ΔKi67% (univariate and multivariate p < 0.05) except the DFS in HER2-positive Luminal B subtypes (univariate and multivariate p > 0.05). Combined, ΔKi67% was confirmed the statistically significant relationship with disease-free and overall survival outcome.

Among patients with TN breast cancer, Δ K i 67 % status also provided meaningful survival forecasts on DFS (p = 0.023 < 0.05) and OS (p = 0.019 < 0.05) presented in Figure 4. The relationship between Δ K i 67 % status and survival outcomes in the TN breast cancer subtype was confirmed by multivariate Cox analysis as well, with DFS-HR = 3.354 (95% CI 1.103–10.196, p = 0.033) and OS-HR = 30.774 (95% CI 3.552–266.644, p = 0.002) (Table 7, corresponding forest plot in Figure 6). Two forest plots represented that negative Δ K i 67 % status is a valid indicator for better prognostics. We further used the Kaplan-Meier method to affirm the correlation between Δ K i 67 % status and survival outcomes in each molecular subtype. Survival curves based on Δ K i 67 % status for the four subtypes are displayed in Figure 6. Δ K i 67 % status shows statistically significant differences in Luminal B and TN breast cancer patients. The NRI value comparing the prognostic capacity between Δ K i 67 % status and pCR in TN breast cancer subtype was 0.685 (95% CI 0.3336–1.0294, p < 0.001), also supporting our conclusions. In both Figures 5, 6 Δ K i 67 % had a statistically significant relationship with prognostic outcome.