An insight into the invasion of breast ductal carcinoma in situ based on clinical, pathological and hematological data

Inclusion and exclusion criteria

We adopted strict inclusion and exclusion criteria to ensure the rigor of this study. All enrolled patients met the following criteria: (a) Female gender with unilateral breast cancer; (b) initial diagnosis of breast cancer without evidence of distant metastasis; (c) pure DCIS or synchronous DCIS and IDC (referring to invasive carcinoma or microinvasive carcinoma) in breast lesions confirmed by puncture biopsy or excisional biopsy.

Patients with the following characteristics were excluded; (a) Concomitant lobular carcinoma; (b) concomitant invasive carcinoma in any other lesions and those who had already received systemic therapy; (c) failure to do immunohistochemical staining or those with pathology reports which did not indicate whether the immunohistochemistry was part of the DCIS component or the IDC component; (d) had severe blood disorders.

Enrolled patients

From 2012 to 2020, a total of 324 patients hospitalized in the Department of Breast Surgery at the First Hospital of China Medical University were enrolled to form the training cohort. From 2020 to 2021, an additional 65 patients were enrolled to form the validation cohort. All of them have received appropriate medical care including mastectomy/lumpectomy combined with sentinel lymph node biopsy (SLNB)/axillary lymph node dissection (ALND). Based on pathology, we first divided patients in the training cohort into invasion and non-invasion groups in order to investigate which subclones of tumor cells are more aggressive and what will occur in the immune and inflammatory system in the invasion process through comparing differences in DCIS immunohistochemical markers and hematological indicators between them. Second, we compared differences in immunohistochemical markers between DCIS and matched IDC components in the invasion group with an aim to explore changes in the heterogeneity of tumor cells during invasion. Conclusions are validated in the validation cohort (Fig. 1). All patients consented to their clinical information being used in this study and have signed a consent form. This study was approved by the Ethics Committee of China Medical University (Approval number: AF-SOP-07−1.1-01).

Pathological diagnosis

The handling of pathological specimens and immunohistochemical staining were performed as previously described (Liu et al., 2021). At least two experienced pathologists have independently reviewed and signed off on the report, and disputes between them were resolved by negotiation. According to American Society of Clinical Oncology and College of American Pathologists guidelines, expression of ER and PR was mainly based on the percentage of immunohistochemical (IHC)-stained positive nuclei and the intensity of staining with positive of them meaning that ≥ 1% of tumor cell nuclei was immunoreactive (Hammond et al., 2010). Expression of HER2 at the protein level was explored by IHC and was classified into 4 classes (0+, 1+, 2+ and 3+) based on the percentage of positively stained cells, morphology, and staining intensity of the cell membrane (Schaller et al., 2001). Detection of HER2 gene (ERBB2) amplification has also been used by pathologists as an adjunct in combination with IHC to make a more precise determination of HER2 expression (Wesoła & Jeleń, 2015). However, because anti-HER2 therapy has not been shown to be clearly beneficial in DCIS, gene expression testing is not routinely performed (Siziopikou et al., 2013). Therefore, we only discuss its expression on the protein level in this study.

Ki-67 is a biomarker responding to the proliferative activity of tumor cells (Chen & Wu, 2015). In routine pathological examinations, test for Ki-67 was performed and the percentage of positively stained cells was reported. In the present study, 30% was regarded as a cut-off value to distinguish between high and low expression.

Based on the modified Scarff-Bloom-Richardson grading system, pathologists scored specimens on adenoid formation, nuclear morphology, and nuclear schizograms respectively, then classified IDC into three histological categories (Grade 1, 2, and 3) on this basis (Elston & Ellis, 1991; Bansal et al., 2012). Similarly, DCIS was classified into three grades (Low, Medium, and High) based on a grading pattern containing assessment of nuclear grading, necrosis, nuclear splitting images and histology. In the process of data collection, we noted the coexistence of two or more histological grades in some DCIS lesions. We sought to explore its impact on invasion.

Collection of blood sample

Hematological data were obtained by searching the electronic medical records of patients for the period of their hospitalization. We reviewed the platelet count, lymphocyte count, and neutrophil count in peripheral blood samples and derived three hematological indexes: platelet lymphocyte ratio (PLR), neutrophil-lymphocyte ratio (NLR), and systemic inflammatory index (SII) on this basis. PLR was calculated as platelet count/lymphocyte count. NLR was calculated as neutrophil count/lymphocyte count. SII was calculated as neutrophil count × platelet count/lymphocyte count.

Statistical analysis

In the first comparison, age was the only continuous variable conforming to a normal distribution, and the student’s t-test was used for its statistical analysis. Mann–Whitney U test was used to analyze the statistical differences among the remaining non-normally distributed continuous variables and hierarchical variables. Pearson’s chi-square test was used for statistical analysis among categorical variables. Logistic regression models were used to further specify independent influencing factors. Variables with p < 0.1 in univariate analysis were adopted in subsequent multivariate analysis. In the second comparison, Wilcoxon’s test and McNamar’s test were used for statistical analysis of non-normally distributed continuous variables and categorical variables. Kendall’s tau-U rank correlation and linear by linear chi-square test were used to investigate the correlation between HER2 and pathological grade and their co-varying trends respectively. Continuous variables conforming to a normal distribution were presented as mean ±standard deviation (SD), otherwise, they are presented through median joint quartiles. All statistical analyses were two-tailed, and p < 0.05 was considered statistically significant. All statistical analyses were performed using SPSS software version 26.0 (IBM SPSS, Inc., Chicago, IL, USA).

Differences between invasion and non-invasion groups in baseline characteristics

The mean age of all patients in the training cohort was 51.78 ± 11.40. One hundred and seventy-seven patients were divided into the invasion group and the other 147 patients were divided into the non-invasion group, with a ratio of 1.2:1. Detailed differences between them were shown in Table 1. The mean age of patients was 51.49 ± 10.97 in the invasion group and 52.14 ± 11.92 in the non-invasion group. Although patients in the invasion group were younger than those in the non-invasion group, the difference was negligible without statistical difference (p = 0.61). Interestingly, we noted a higher proportion of menopausal patients in the invasion group with a younger mean age compared with the non-invasion group (47.46% vs. 39.46%). This anomaly, although is not statistically significant (p = 0.10), does seem to suggest the important role of the endocrine factor in breast cancer. As for the surgical procedure, the proportion of patients in the non-invasion group who received lumpectomy and SLNB were both higher than those in the invasion group, although the differences were not statistically significant (36.05% vs. 29.38% p = 0.20 and 40.13% vs. 33.90% p = 0.25). In both groups, the proportion of left-sided breast cancers was higher than that of right-sided breast cancers (58.19% vs. 41.81% for the invasion group and 53.06% vs. 46.94% for the non-invasion group) but was not statistically significant (p = 0.33). We considered the two groups were balanced in baseline characteristics.

Differences between invasion and non-invasion groups in pathological parameters

For pathological parameters of DCIS including ER, PR, HER2, DCIS histological grade, and the mixture of histological grades, no statistically significant differences were noticed. Overall, the proportion of patients with positive hormone receptors was much higher than those with negative hormone receptors (68.83% vs. 31.17% for ER and 71.60% vs. 28.40% for PR). As for HER2, patients with a high expression on protein levels (3+) accounted for 46.67% of the whole, well above the theoretical average (25%). Both situations were in line with previous reports (Wolff et al., 2007; Chaudhuri et al., 1993; Wärnberg et al., 2001). Constrained by the cramped space and limited energy supply within the ductal system, the proliferative capacity of DCIS cells was weak, as demonstrated by the low Ki-67 expression in 83.95% of all patients.

Correlation of hematological indicators with invasion

For hematological indicators, statistically significant increases in platelet (p = 0.03), neutrophil (p = 0.02), and SII (p = 0.02) in the invasion group were noticed. The receiver operating characteristics curve (ROC) was then used to evaluate the discriminatory power of them and determine the cut-off value with the results shown in Fig. 2. The AUC for platelet, neutrophil, and SII were 0.567, 0.579, and 0.577, with cut-off values of 291.50, 3.49, and 347.20 respectively (corresponding to the maximum of the Youden index). After univariate and multivariate logistic regression analyses, high expression of platelet (>291.50) (odds ratio, 2.46; CI [1.35–4.46]; p = 0.003) and SII (>347.20) (odds ratio, 2.54; CI [1.56–4.12]; p < 0.001) were established as independent predictors for invasion (Table 2). On this basis, we speculated that invasion of tumor cells into mesenchyme might trigger changes in the body’s immune and inflammatory response.

The relationship between DCIS and IDC components

To explore changes in the tumor heterogeneity during the transition from DCIS to IDC, we compared immunohistochemical markers between DCIS and IDC components in the invasion group with results shown in Table 3. For ER and PR, similar positive rates in both components were noticed (72.39% vs. 74.63% for ER, p = 0.63 and 70.34% vs. 71.19% for PR, p = 1.00). The IDC component had a significantly higher expression of Ki-67 compared with the DCIS component (p < 0.001).

The relationship between DCIS and IDC regarding HER2 was shown in Table 4. We first performed statistical analysis using Wilcoxon’s test. The result showed p < 0.01, indicating that DCIS with different HER2 expression levels would develop into IDC with different HER2 expression levels. The results of Kendall’s tau-U rank correlation and linear by linear chi-square test suggested a positive correlation in HER2 expression between DCIS and IDC components. This correlation was not limited to the level of ’rank’ but was a linear relationship, implying that DCIS with higher HER2 expression tended to develop into IDC with higher HER2 expression. Subsequently, we noted that in the DCIS component, the groups with HER2 expression of 3+ and 2+ accounted for 41.17% and 34.56% respectively. However, in the IDC component, these percentages decreased to 28.68% and 33.09% respectively. In contrast, the groups with HER2 expression of 0+ and 1+ were differentially elevated, which was contrary to our statistical results (Figs. 3A–3C). In the statistical analysis for histological grade, we came to a similar conclusion as for HER2 (Table 5). Although the theory “DCIS with higher histological grades tend to develop into IDC with higher histological grades” holds true, we noticed that the majority of cells in all of the three grades of DCIS tended to develop into IDC grade 2 (Figs. 3D–3F). As a result, in DCIS and IDC components, the high-grade group and IDC grade 2 occupied the highest proportions respectively. We also made Sankey diagrams to show the relationship between DCIS and IDC (Fig. 4). Considering the change in hematological indicators in the invasion group, we speculated the immune and inflammatory response might be responsible for this phenomenon.

Verifications of the predictive power of SII and platelet

To verify the predictive power of hematological indicators, medical records of additional 65 patients were collected. Based on the expression of SII and platelet, they were divided into three groups: Group 1 (SII > 347.2 and platelet > 291.5), Group 3 (SII < 347.2 and platelet < 291.5), Group 2 (the rest of patients). After statistical analysis, the percentage of patients with invasion components was 75% in Group 1, which was significantly higher than 51.85% in Group 2 and 22.72% in Group 3 (p = 0.005) (Table 6).