Breast cancer is the second most prevalent form of cancer among women, accounting for over 30% of all newly diagnosed cancer cases in women annually. Forecasts indicate 310,720 new cases and 42,250 deaths among American women in 2024 alone (1, 2). At some point in their lives, approximately 12.5% of women receive a breast cancer diagnosis (1). Currently, there are over four million breast cancer survivors in the United States (1). This includes women who are still being treated and those who have completed the treatment (1). However, this burden extends beyond mere statistics, encompassing a profound impact on patients, the healthcare system, and societal well-being.

The latest data obtained by the Centers for Disease Control and Prevention revealed that breast cancer commands the highest treatment expenditure among all cancers in the United States, with medical services amounting to $26.2 billion and prescription drugs totaling $3.5 billion (3).

Projections anticipate that roughly 56,500 new breast ductal carcinoma in situ (DCIS) diagnoses will occur in 2024 (1, 2). Currently, the extent and histological grade of lesions are the only factors that influence DCIS treatment (4). For patients with in-situ lesions only, guidelines recommend surgery alone or in combination with radiation for the treatment of patients with in-situ lesions only. Despite its prevalence, the precise subtypes of DCIS leading to aggressive cancers remain unidentified, necessitating urgent research. Women diagnosed with DCIS have a 1.8-fold increased risk of breast cancer-related mortality compared to women in general (5). This heightened risk was particularly pronounced among those diagnosed before the age of 35 years, where mortality rates were approximately 17 times higher than expected within nine years following diagnosis (5).

Since the introduction of mammography screening, breast DCIS has become a common diagnosis (6). About 20–25% of all malignant lesions found through national screening programs in industrialized nations are DCIS (6, 7).

In two previous studies, we found that human epidermal growth factor receptor 2 (HER2)-enriched high-grade DCIS was a high-risk subtype because it was strongly linked to having invasive components (8) and low progesterone receptor (PR) levels, correlated with HER2 overexpression and the presence of invasive components (9). The ultimate goal of our research is to improve the identification of DCIS lesions with a high potential to develop into invasive breast carcinoma (IBC). This information may pave the way for personalized treatment for women diagnosed with DCIS. We can potentially evaluate adjuvant immunotherapy (e.g., anti-programmed cell death protein 1/programmed death-ligand 1 (anti-PD-1/PD-L1) therapy) in a subset of patients at the highest risk of developing invasive breast cancer. In addition to enabling patients to avoid under- and over-treatment, changing the current standards can allow the health system to target individuals diagnosed with high-grade DCIS more effectively. In IBC, hormone receptors (HRs) for estrogen and progesterone, in addition to HER2 and Ki67 proliferation index, are all deciding factors according to a complex treatment algorithm (10). Conversely, we have not yet established the utility of immunohistochemistry (IHC) markers in the diagnosis of DCIS. Breast DCIS is a diverse entity, with different growth patterns and nuclear atypia ranging from subtle to severe. Certain DCIS subtypes may not pose an imminent risk, whereas others may suggest the possibility of progression to invasive breast cancer. If left untreated, around 20–25% of DCIS cases progress to invasive breast cancer within ten years (11). In other studies, this proportion has been reported to be as high as 50% (12). The precise DCIS subtypes that progress to aggressive cancers is currently unidentified. However, because almost all women diagnosed with DCIS undergo breast surgery, there is a lack of understanding of the possible consequences of not treating this disease. Despite this, a phase III randomized controlled clinical trial examined the risks and benefits of active monitoring versus standard DCIS therapy in patients with low-risk DCIS. The results of this study have yet to be published (13, 14). Routine IBC specimens frequently contain DCIS components. However, we do not devote much attention to the morphological features of the HR-, HER2-, and Ki67 status of this component, as these factors currently have little bearing on the clinical management of patients with IBC. Only a few studies have investigated the significance of these markers in DCIS according to their molecular subtypes (15–17).

Over the years, Ki67 staining has increasingly permeated cancer research and diagnosis, with applications in numerous malignant tumors, including invasive breast cancer (IBC). The Ki67 proliferation index is a widely acknowledged marker of cellular proliferation, serving diagnostic, prognostic, and predictive roles with notable efficacy. In 1983, Gerdes et al. first discovered Ki67 protein (18). The name was derived from the city of origin (Kiel, Germany), and 67 was the number of original clones in a 96-well plate (19). During all stages of the cell cycle, except for the dormant phase (G0 phase), dividing cells show Ki67 expression in their nuclei (20). Initially, the main purpose of Ki67 was to examine cell cycle and cellular proliferation in normal and malignant tissues. It is now one of the most commonly used markers for determining tumor cell growth and behavior. Ki67 is a feasible marker for predicting the behavior of malignant tumors by quantifying the percentage of actively dividing tumor cells using IHC. Despite its extensive application, research continues to be conducted to improve Ki67 prognostic and predictive accuracy in various cancer types (21).

A large global phase III randomized trial included breast cancer patients with 1-3 positive axillary lymph nodes and a Ki67 index > 20% treated with adjuvant endocrine therapy and abemaciclib, a cyclin-dependent kinase 4 and 6 (CDK4/6) inhibitor. The primary endpoint was invasive disease-free survival. Results showed that while Ki67 was prognostic for recurrence, it was not predictive of response, as abemaciclib provided benefits regardless of Ki67 status (22, 23).

According to the International Ki67 in the Breast Cancer Working Group (IKWG), Ki67 ≤ 5% indicates low proliferation, and Ki67 > 30% indicates high proliferation (24).

Several changes have been made to the Ki67 cutoff values used to identify and differentiate Luminal A (LumA) and LumB subtypes in IBC at St. Gallen consensus meetings. This demonstrates the difficulty in defining low-grade tumors with a low proliferation index versus tumors exhibiting enhanced proliferation (25, 26) which are suitable for therapeutic decisions. In routine diagnosis, the vast majority of IBC cases have an intermediate (6–29%) Ki67 proliferation index (27). Only a limited number of studies (28, 29) have investigated the Ki67 proliferation index in breast DCIS. The purpose of this study was to identify a more consistent cutoff value for Ki67 and investigate whether the Ki67 proliferation index can contribute to identifying high-grade DCIS lesions with the highest potential to finally turn into IBC in humans. Furthermore, we aimed to study the possible relationships between HER2 overexpression, low PR levels, and high Ki67 expression.

The present study was a retrospective interventional investigation. Our entire study material included formalin-fixed paraffin-embedded (FFPE) histopathological specimens from a consecutive patient cohort stored in the diagnostic archive at Akershus University Hospital, Norway. Samples from 494 female patients diagnosed with ductal carcinoma in situ (DCIS) of the breast between 1996 and 2018 at Akershus University Hospital were collected, examined, and graded by experienced breast pathologists using the Van Nuys classification system (8, 30). The selected patients who met the inclusion criteria (8) received a prepaid return envelope that was enclosed in the letters they received, in addition to a sheet to sign and return if they objected. If they agreed, they would not need to undertake any action. We received reservations from ten patients; consequently, their cases were excluded from further examinations. To our knowledge, this is one of the largest DCIS biobanks in Europe that has been collected and approved for research purposes.

We chose to investigate high-grade (grade 3) DCIS cases because these lesions are believed to harbor the highest risk of recurrence and progression to IBC (30–33). We identified 357 patients with high-grade DCIS ( Figure 1 ) and stained their high-grade DCIS tissues using IHC for the estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2), and Ki67 proliferation index. ER and PR IHC positivity was defined as ≥ 1% positive DCIS cells, in accordance with the updated guidelines of the American Society of Clinical Oncology (ASCO) and the College of American Pathologists (CAP) (34) developed for IBC.

In this study, we identified that the interobserver agreement between Counting by Pathologist 1 (CP1) and CP2 varied with the selection of Ki67 cutoff values. Cohen’s kappa coefficient was higher, and the interobserver agreement between CP1 and CP2 was more reproducible in our assessed 50 cases when we used the 20% cutoff (moderate), compared to > 5–29% and ≥ 30% intervals (fair). The average mean and median Ki67 expression at the 20% cutoff value in the 50 assessed cases, scored by CP1 and CP2, were 20.5% and 23%, respectively, which could be a possible explanation for the higher reproducibility in favor of the 20% cutoff value. Nonetheless, these 50 cases were probably not large enough to cover the entire Ki67 expression spectrum in high-grade DCIS. For instance, the calculated Cohen’s kappa coefficient κ = 1.000 was unreliable because only one case was found to have a Ki67% of ≤ 5%. Poor, fair, and moderate Ki67 interobserver agreements have also been found in studies involving IBC cases (43–45).

In our 357 high-grade DCIS cohort, the majority of patients (64%) had a Ki67 value in a > 5–29% interval. This is in accordance with the findings in IBC (27). Harbeck et al. performed a large prospective phase III trial on endocrine therapy in IBC. They found that Ki67 levels ≥ 20%, along with other high-risk factors, could be used to identify patients who might benefit from certain treatments (46). An important issue in the analytical phase of Ki67 IHC assessment was the use of multiple counting methods (47–49).

We acknowledge that DCIS, classified as the LumA and LumB HER2^- subtypes, share several molecular features. The HRs are positive, and HER2 is negative in both subtypes. These two subtypes are defined and distinguished by an arbitrary and predefined Ki67 cutoff value, which has undergone several changes over the past decade. Thus, the Ki67 measurements, both mean, median, and maximum, should be interpreted cautiously in these two subtypes. We chose to combine these two subtypes to obtain a statistically large number of patients.

We demonstrated a statistically significant higher Ki67 distribution in LumB HER2⁺ and HER2-enriched subtypes (p-adjusted < 0.0001, Kruskal-Wallis test) when these subtypes were compared to the combined LumA and LumB HER2^- subtypes as an entire group. Remarkably, the comparison with the TPN subtype was not significant (p-adjusted = 0.5886, Kruskal-Wallis test) ( Figure 3 ).

The results of this study indicated that high-grade DCIS cases with low PR levels (< 20%) combined with high Ki67 expression (≥ 20%) were more likely to be of the HER2-enriched subtype (p < 0.0001, chi-squared test) ( Figure 5 ). By definition, the HER2-enriched subtype of invasive cancer is a non-dependent Ki67 entity, defined by the HER2 score 3+ and simultaneous negative HRs, irrespective of whether Ki67 is low or high. Despite this, DCIS cases with low PR levels (< 1%) combined with a high Ki67 expression constituted a strong majority of cases in this subtype ( Figure 5 ). Additionally, cases with low PR levels and simultaneously high Ki67 expression were significantly overrepresented among the combined group of LumB HER2⁺ and HER2-enriched subtypes, representing 79% (62 out of 78) of the cases (p = 0.0107, Fisher’s exact test) ( Figure 6 ). We suggest that, in a given high-grade DCIS lesion with a low PR level and high Ki67 expression, there is a 56% probability that the lesion is a HER2-enriched subtype and a 79% probability of HER2 overexpression (score 3+) (OR: 2.325; 95% CI: 1.229 to 4.518). In an earlier study, we found that the HER2-enriched subtype was significantly associated with invasive components (8). We believe that these findings can potentially identify high-risk DCIS patients.

Interobserver variations in Ki67 visual scoring are particularly evident in the intermediate interval (> 5% and < 30%) (24). As a result, in a recent guideline by the IKWG in 2021, Ki67 cutoff values of ≤ 5% or ≥ 30% were preferred for use in IBC. Ki67 clinical utility should be limited to the estimation of prognosis in anatomically favorable ER-positive and HER2^- patients and to determining which patients do not require adjuvant chemotherapy (24).

In 2008, the Breast International Group Trial 1-98 (BIG-1-98), set the cutoff value to define a high Ki67 proliferation index in IBC at 11% (50). The IKWG was established in 2009 to address preanalytical, analytical, and interlaboratory challenges regarding Ki67 (51, 52). Later, at the 2011 St. Gallen International Consensus Conference, a cutoff threshold of 14% was established (25), to distinguish between the LumA and LumB HER2^- subtypes of IBC. Two years later, at the 2013 St. Gallen meeting, the threshold for a high Ki67 proliferation index was increased to 20% following discussions on the poor repeatability of the previous threshold values (26).

Nonetheless, the 2015 St. Gallen Conference did not attempt to determine a certain threshold for Ki67 but rather addressed that to be decided based on local laboratory guidelines and routines (53). During the preanalytical phase, multiple variables could have influenced the Ki67 evaluation. These included the type of specimen, type of fixative, length of fixation, and cold ischemic time (the interval from the time the specimen was removed during surgery to placement for tissue fixation) (54).

Compared with ER or HER2 IHC, Ki67 IHC seems to be more sensitive to fixation-related changes. When using fixatives other than neutral buffered formalin, delays of more than 3 hours (too short), 16 hours, or 14 days (too long) in fixation duration may result in a decline in the Ki67 index values. Prolonged fixation periods can potentially cause epitope degradation (55). Ki67 is also more vulnerable to antigen decay during long-term storage in FFPE blocks (56). The IKWG advises performing Ki67 IHC within five years of tissue placement in these blocks (24).

Overall, we deemed the quality of hematoxylin and eosin (HE) morphology and IHC staining in our materials to be technically adequate. In our previous study (8, 57), we experienced that suboptimal fixation was the likely cause of IHC failure in 12 of 422 samples. Nevertheless, we cannot rule out that some FFPE blocks could have poor quality for Ki67 IHC staining (cf. ratios < 1%), even though the ER, PR, and HER2 IHC markers showed acceptable features.

In our cohort (n = 357), we chose the hotspot method ( Figures 7A–F ) and calculated Ki67 ratios as a continuous variable rather than a categorical variable in accordance with the current local and national guidelines for the diagnosis, treatment, and follow-up of patients with IBC (58). The IKWG recently recommended global (average) counting instead of hotspot counting, which counts only the areas with the highest proliferative activity owing to its association with higher interobserver variability. In general, there are no established guidelines on how and where to count Ki67 in DCIS or where to expect to find hotspots.

In the assessment of IBC sections, Ki67-positive tumor cells are usually counted on the periphery of invasive lesions (invasion front). The central areas of IBC may be devoid of mitosis and Ki67-positive cells. In DCIS, there may exist a growth front with more pronounced proliferation on the periphery of the DCIS and possibly also a retrograde growth into the lobules (cancerization of lobules).

In addition, growth occurs in the basal layer of DCIS. We selected IHC sections from among those with the largest areas of DCIS, which mostly corresponded to the central areas of the lesions and hence may show poor mitotic figures. These included foci of lobular cancerization. Indeed, we found and counted hotspots in many lobules. Growth was observed in the basal cell layer of the central ducts (28), where the expression pattern varied from low to high.

Digital image analysis (DIA) platforms that can detect nuclear IHC biomarkers, such as Ki67, have made biomarker evaluations faster and easier to perform automatically (59). Thus, it is likely that future Ki67 counting precision, repeatability, and reproducibility will improve, as discussed by other researchers (21, 60, 61). Ki67 should not be widely used for clinical management without evaluation of quality assurance (EQA) because regular participation in such programs has been shown to substantially improve interlaboratory consistency (24, 62).

Our results emphasize the elusiveness of Ki67 in determining a robust and reproducible cutoff. Nevertheless, Ki67 is a valuable prognostic and predictive marker, although it has certain limitations (63–65).

The Ki67 interobserver agreement was moderate at a cutoff value of 20%. We demonstrated that patients with invasive components had considerably higher Ki67 values. Ki67 was substantially higher in the LumB HER2⁺ and HER2-enriched subtypes compared to the LumA and LumB HER2^- subtypes as a combined group. High Ki67 expression in high-grade DCIS can indicate potential invasive capabilities. Patients with HER2 overexpression (score 3+) were strongly overrepresented when the subtypes were subcategorized by combining a low PR level with a high Ki67 ratio. We believe that this combination has the potential to serve as a prognostic biomarker for invasive growth in high-grade DCIS. We also believe that this study contributes to the identification of DCIS lesions that ultimately progress to IBC and require tailored therapies.