Prognostic and Molecular Classication Value of Ki67 Index in Patients With Breast Cancer

Background: The Ki-67, which is a proliferative index, became more important in making treatment decisions of breast cancers. But a few factors limited its use in clinical practices, especially the assessment of positive cells percentage of Ki-67 and the cut-off value of Ki-67. In this study, we examined the expression of Ki-67 by immunohistochemical method, and evaluated the value of Ki-67 index in patients with breast cancers systematically. Methods: The expression of Ki-67 was detected by immunohistochemical method, the relationship between over-survive (OS), disease free survive (DFS) of patients with BCs and Ki-67 index (cid:0) and the value of Ki-67 in molecular classication of BC were analyzed with SPSS ver. 18.0 and GraphPad Prism software. Results: There were markedly difference between OS in patients with BC when the Ki-67 index arranged from 46% to 68% (χ2 = 6.87, P = 0.0088; χ2 = 7.69, P = 0.0056 respectively), and the same results were also found when the positive density was added to the Ki-67 index, but the positive density has limited value in assessment the value of Ki-67. There were markedly difference between DFS in BCs when the Ki67 index arranged from 50% to 58% (χ2 = 7.31, P = 0.0068; χ2 = 7.88, P = 0.005). When using 14% as a cut-off point to classify the molecular type of BC, the Luminal A-type patients had statistical differences with patients with HER-2 over-expression, with triple negative type BC in term of OS respectively (χ2 = 4.8, P = 0.028; χ2 = 3.89, P = 0.049). There’s statistical difference between OS of patients with HER-2 over-expression BC when the Ki-67 index fell in the range of 49%- 60% (χ2 =4.97, P = 0.0259; χ2 =5.61, P = 0.0178, respectively). There were statistical differences between the Luminal A-type BC and HER-2 over-expression Phosphatase; ER: estrogen receptor; PR: progesterone receptor.


Background
According to the latest national statistical data on cancer of China National Cancer Center in 2019, the incidence rate of breast cancer (BC) of Chinese was 45.29/100,000 in China. BC ranks the second among all cancers in term of incidence rate and the top position in term of incidence rate of cancers in women. Breast cancer is the fth cancer in view of death rate among all cancers in women, which amounts to about 70,000 cases / year. Ki-67 is a key marker in prognosis and molecular classi cation of BC. In 2011, a consensus was reached on the St Gallen Conference, which required to using 14% as the critical value of Ki-67 to distinguish Luminal A-type and Luminal B-type BCs [1,2]. Some guidelines also provided ambiguous expressions of critical value of Ki-67 as "high expression'' and "low expression". Some experts suggested to using ≥20% as the critical value of "high expression" state of Ki-67 [3,4]. The CSCO Guideline in 2019 suggested that the critical value of Ki-67 shall be determined according to practical situations of various laboratories. Most Chinese experts agreed that <15% is low expression and >30% is high expression. When Ki-67 index is 15% -30%, it is suggested to make clinical decisions according to secondary pathology consultation or other indexes [5].
In practical pathological diagnosis, the cut-off point of Ki-67 is di cult to be determined and has poor reproductivity. Different laboratories, technicians with different degree of familiarity and pathological diagnosis doctors with or without training might get different results. In particular, it is more di cult to have standard operation after the cutoff point of Ki-67 is solidi ed. Moreover, whether staining intensity of Ki-67 is meaningful to prognosis and molecular classi cation is a controversial di culty at present.
This study focuses on how to judge percentage of positive cells of Ki-67 simply and conveniently and whether there's a certain interval for the convenience of operation of clinical pathologists to improve repeatability.

Methods
Clinical data Para n samples of 288 patients with BC from January, 2005 to July, 2013 in NO. 989 th Hospital was collected for a retrospective study. Among these patients, there were 4 males (1.39%) and 284 females (98.61%). All respondents were aged between 25 and 84, averaging at 48 years old. Speci cally, there were 2 cases younger than 30 years old, 230 patients between 30-60 years old, and 56 patients over 60 years old. With respect to lymphatic metastasis, there were 154 cases without lymphatic metastasis (N0), 66 cases with 1-3 nodes of lymphatic metastasis (N1), 39 cases with 4-9 nodes of lymphatic metastasis (N2), and 20 cases with ≧10 nodes of lymphatic metastasis (N3). According to pathological stages of AJCC (pTNM), there were 8 cases in Stage 0, 42 cases in Stage IA, 104 cases in Stage IIA, 60 cases in Stage IIB, 43 cases in Stage IIIA, 1 case in Stage IIIB, 20 cases in Stage IIIC, and 1 case in Stage IV (Tab.1). The follow-up time lasted for 2 -157 months, averaging at 82 months. In addition, three cases had metastatic lesions and 6 cases had not undergone radical operation. None patient had received chemoradiotherapy before the operation. Samples were xed with 10% neutral formalin and embedded with para n. All patients had complete follow-up data. The inclusion standards of respondents were as follows: 1) all patients were diagnosed through breast puncture or postoperative examination after radical surgery. 2) None patients received chemotherapy or radiotherapy before operation. 3) All patients had complete follow-up data. 4) Patients died for breast cancer metastasis or relapse. Patients who met one of following conditions were excluded: 1) without complete follow-up data, 2) received thermotherapy or radiotherapy before operation, 3) patients died for reasons other than breast cancer metastasis or relapse, 4) patients with poor living quality caused by other reasons or patients died for other tumors. This study was approved by the ethics committee of NO. 989 th hospital and all included respondents signed the informed consent.
Immunohistochemical assay of Ki-67 in BC samples Tumor samples were xed in 10% paraformaldehyde solution buffered with neutral phosphate-buffered solution (PBS; pH 7.2), and embedded in para n for sectioning (2-5μm). For immunohistochemical analyses on Benchmark XT system (Ventana, AZ, USA), depara nized sections were immersed in 0.3% H 2 O 2 in methanol for 30 min to eliminate endogenous peroxidase activity, followed by incubation with PBS containing 1% normal serum corresponding to the secondary IgG and 1% bovine serum albumin (BSA) to reduce nonspeci c reactions. The sections were incubated with rabbit anti-human Ki-67 (MIB-1) (dilution 1:200; Maxim, Fuzhou, China). After incubation with biotinylated secondary Abs at room temperature (RT) for 1 hour, immune complexes were ampli ed with Streptavidin-Alkaline Phosphatase (SAP) immunohistochemical staining kits (Maxim, Fuzhou) following the manufacturer's instructions. Negative control was done by omitting the primary antibodies and substituting with corresponding normal isotype-matched immunoglobulin IgG.

Results analysis
Ki-67 protein positive lies on cell nucleus. It is judged positive for occurrence of yellow nuclei or brown particles. The whole section was previewed under a low power lens before interpretation to judge whether the staining is uniform. If there's even staining, samples were observed under a high power lens (40x) to count more than 500 cancer cells. If there's uneven staining, the hot spot region with concentrated positive cell number was chosen to count more than 500 cancer cells. Samples which approached to cutoff point were judged again by 2 another independent pathologists and the average Ki-67 index was calculated. Finally, the mean value was chosen as the nal Ki-67 index value. Samples with staining intensity unable to be determined and ambiguous cell pro les were excluded from counting. Besides, staining intensity of tumor cells was scored: light yellow (1 score), deep yellow (2 scores) and brown (3 scores). Staining index = percentage of positive cells× staining intensity. The minimum and maximum staining indexes were 1% and 285%, respectively.

Molecular classi cation
Molecular classi cation was performed by the consensus on the St Gallen Conference in 2011 (Luminal A-type and Luminal B-type BCs were distinguished by using 14% as the critical value of Ki-67) and suggestions of CSCO guideline in 2019 (If critical value of Ki-67 is smaller than <15%, it is low expression. If critical value of Ki-67 is higher than 30%, it is high expression) respectively. When expressions of ER (+), HER-2 (-) and Ki-67 are between 15% -30%, molecular classi cation of BC was performed according to percentage of positive expression cells of PR. Under this circumstance, PR=20% was used as the critical value according to suggestions of CSCO in 2019. If PR >20%, samples were determined as the Luminal Atype BC. If PR<20%, samples were determined as the Luminal B-type BC.

Statistical analysis
The statistical analyses were completed by using SPSS ver. 18.0 (SPSS Inc., Chicago, IL). Survival data (overall survival, OS and disease-free survival, DFS) were analyzed by using Kaplan-Meier survival curves and compared using log-rank tests among different groups. Statistical signi cance was 2-sided, and Pvalues < 0.05.

Results
Clinical-pathological characteristics of patients with BC Among the patients with BC enrolled in this study, invasive ductal carcinoma not otherwise speci ed (DCS, NOS) was found in 90.97% of patients, ductal carcinoma in situ was found in 2.78%, invasive lobular carcinoma (LCS) was found in 2.08%, invasive micropapillary carcinoma was found in 1.04%, mucinous carcinoma was found in 2.43%, medullary carcinoma was found in 0.35%, and apocrine  Table 1. The median follow-up time was 82 months (range from 2 to 157 months).

Relationship between Ki-67 index and overall survival (OS) of patients with BC
After staining and interpretation, there were 285 samples which could count more than 500 tumor cells.

Whether staining intensity of Ki-67 of tumor cells is signi cant to judging OS of patients with BC?
Staining intensity was added based on Ki-67 index. Results showed that there were statistical differences of OS between two groups when the critical value was 133% after accumulation of Ki-67 index and staining intensity and patients with high expression of Ki-67 index had relatively poorer prognosis (χ2 = 7.82, P = 0.0052. Tab.3 and Fig.1C). However, there's no statistical difference of OS between the group with staining intensity and the group without staining intensity (χ2 = 0.001, P = 0.9780, Tab.3). This re ected that staining intensity of Ki-67 was not valuable to OS prediction of patients with BC.
Relationship between expression of Ki-67 in BC and disease-free survival (DFS) of patients At the end of follow-up time, 36 patients among 288 patients with BC developed metastasis or in-situ relapse at different positions. According to statistical analysis, DFS of patients with BC declines gradually with the increase of Ki-67 index from 50% to 58% (χ2 = 7.31, P = 0.0068; χ2 = 7.88, P = 0.005 respectively). This revealed that Ki-67 index was conducive to judge DFS of patients with BC when it was in the range of 50% -58% (Tab. 4  Among 288 patients with BC, 258 patients with invasive ductal carcinoma adopted molecular classi cation according to standards in CSCO in 2019. If percentage of Ki-67 positive cells was lower than 15%, it is determined as low expression. If percentage of Ki-67 positive cells was higher than 30%, it was determined as high expression. When percentage of Ki-67 positive cells fell in the interval of 15% -30%, molecular classi cation of PC was performed according to state of HER-2 and percentage of PR positive cells. In other words, PR=20% was used as the critical value according to suggestions of the CSCO consensus in 2019 when expressions of HER-2 (-) and Ki-67 were in the range of 15% -30%. If PR > 20%, patients were determined as Luminal A-type BC; otherwise, patient were determined as Luminal Btype BC. According to molecular classi cation results, there were 91 cases of Luminal A-type BC, 79 Luminal B-type BC, 46 cases of HER-2 over-expression BC, and 42 cases of three-yin type BC. According to statistical analysis results, there were statistical differences between the Luminal A-type BC and HER-2 over-expression BC in term of OS (χ2 = 4.65, P = 0.0310). The patients with HER-2 over-expression BC presented higher death rate and poorer prognosis, but it presented no statistical difference with Luminal B-type BC and triple negative type BC in term of OS (Tab.6 and Fig.4A).
Relationship between Ki-67 index and prognosis of patients with speci c molecular classi cation based on CSCO in 2019 was analyzed. According to statistical analysis, there were statistical differences between two groups of Luminal B HER-2 (-) type when the critical value of Ki-67 index was 52% (χ2=6.61, P = 0.0101), ( Figure 4B). Moreover, there were statistical differences between two groups of HER-2 overexpression type when Ki-67 index fell in the critical interval of 47%-60% (χ2=4.97, P = 0.0259; χ2=5.61, P = 0.0178, respectively) (Fig. 4C, 4D.), but these two groups showed no statistical difference with other molecular subtypes. Otherwise, it is interested that all of 19 patients with Luminal A-type BC survived when Ki-67 index was ≤4% (within the follow-up visit period)

Discussion
Effects of Ki-67 in BC mainly could be divided into two parts: 1) molecular classi cation: this is particularly necessary when distinguishing Luminal A-type BC and Luminal B-type BC, because the former one only needs endocrine-therapy, while the later one needs systematic chemotherapy based on endocrine-therapy to achieve good clinical effects [6]. Hence, it is very necessary to distinguish these two types of BC in clinics. 2) judgment of prognosis of patients: many studies have demonstrated that patients with higher Ki-67 index had poor prognosis and they were easier to suffer relapse [7,8]. Patients with higher Ki-67 index will gain better pathological complete responses (pCRs), without considerations to HR state, HER-2 state and triple negative type of BC [9]. Some study even believed that Ki-67 index was unrelated with prognosis of patients with BC. Moreover, the prognosis was poorer when the Ki-67 index was lower [10]. Some study has demonstrated that other clinical pathological features of patients shall be considered during judging in uences of Ki-67 index on patients with BC, including histological grading, lymphatic metastasis, tumor volume, existence of vascular invasion, etc [11].
Therefore, Ki-67 has not been widely applied in clinics. In addition to differences among observers, some clinicians depend more on Ki-67 division between "high Ki-67 group" and "low Ki-67 group" in determining the treatment. However, the critical value of "low Ki-67 group" ranges between 0 and 28.6% [12,13], showing signi cant differences. Therefore, how to determine critical value of Ki-67 index and values of Ki-67 index in judgment of prognosis and molecular classi cation of patients with BC, and even how to make reasonable and e ciency use of Ki-67 in BC are of signi cant importance.
Immunohistochemical staining is the most common analytical method of Ki-67. This method is not only economic and simple, but also needs no complicated equipments. Immunohistochemical staining is easy to be promoted. However, Ki-67 has limited practical values in clinical practices, because there are many interference factors before, during and after experiment. Interferences before the experiment are mainly selection of stationary liquid, xed time and cold ischemia time. Interferences during the experiment generally include type of antibody, antigen repair method, staining development time and pro ciency of technicians. Interferences after the experiment mainly include counting technique of positive cells, professional training of observers, and accurate recognition of positive cells [14][15][16][17]. In uencing factors before and during the experiment can be avoided through standardized operation as much as possible, including choosing 10% neutral formalin as the xed liquid, controlling time of samples in xed liquid after separation within 30min, assuring at least 24hrs of xed time, and using the primary antibody of the same batch from the same factory, the same antigen repair method after optimization as well as the same technician. Counting of positive cells after experiment was the most di cult work. To optimize the counting method and decrease differences among observers as much as possible, the International Ki-67 in Breast Cancer Working Group offered some suggestions to promote applications of Ki-67 in BC: counting at least 3 high power elds (×40 objectives) or counting 500-1000 tumor cells, including invasion edges of tumor and hot spot region [18].
In this study, percentage of Ki-67 positive cells in 285 patients with BC was calculated and it found that there were statistical signi cant differences among patients with BC in term of OS when the critical value of Ki-67 index fell within the interval of 46% -68%. In other words, death rate of patients increased gradually and the prognosis deteriorated with the increase of Ki-67 index. Since the interval of Ki-67 index is large, it is easy to be promoted in pathologists and shows relatively high repeatability. Staining intensity of Ki-67 was included into the evaluation index, which also can in uence OS prediction of patients with BC. However, the group considering staining intensity and the group without consideration to staining intensity showed no statistical signi cant differences in OS prediction. This proved that staining intensity of Ki-67 has not signi cant values in OS prediction of patients with BC. Zhu pointed that that setting the critical value of Ki-67 index as 30% is an independent prognosis factor of OS and DFS of three-yin cancer [19]. Patients with triple negative cancer and Ki-67 > 30% had relatively poorer prognosis, especially for patients in Stage I. When the critical value of Ki-67 index was 10%, for both high expression (≥10%) and low expression (<10%), patients with BC show statistical differences in OS and recurrence-free survival (RFS) when there's low PR expression (<20%). In particular, patients with high Ki-67 expression and low PR expression had the poorest prognosis [20]. Similarly, Ki-67 index is signi cant to DFS judgment of patients with BC. According to this study, DFS of patients with BC decreased gradually when Ki-67 index value increased from 50% to 58%, but the interval of Ki-67 index to judgment of DFS was relatively small and special attentions shall be paid to it. However, some study show that when the critical value of Ki-67 index was determined 14%, patients with Luminal-type BC and high Ki-67 expression (≥14%) and low PR expression (<50%) achieved a relatively high clinical complete remission (CR) from neoadjuvant chemotherapy [21].
There's a "gray zone" during molecular classi cation to patients with BC based on Ki-67 index, especially when different methods were applied. During molecular classi cation of patients with BC based on the consensus of St. Gallen Conference in 2011, it is relatively di cult to set 14% as the critical value of Ki-67 index in practical applications. In particular, the repeatability is poorer when the primary counting is between 12% -16%. However, there are still counting errors by the same technician and the error is even larger by multiple technicians. In this study, the same observer was assigned to make three times of counting in 2 days for patients whose Ki-67 index ranges between 10% and 20%. Next, the mean value was used as the nal Ki-67 index. This can decrease errors among different observers as much as possible and the total cell counting is increased signi cantly. The biggest defect lies in high workloads and it is only applicable to the situation with small sample sizes. During molecular classi cation based on CSCO standards in 2019, it needs repeated counting or decision-making according to other evidences when the Ki-67 index ranges between 15% -30% [5], which absolutely increases workload and leads to poor repeatability as well as increased errors. How to decrease "grey zones" and differences in molecular classi cation of BC is the key and di culty of this study.
During molecular classi cation of patients with BC based on consensus on St. Gallen Conference in 2011, the OS of patients with Luminal A-type BC shows no statistical difference with those of patients with HER-2 over-expression and patients with triple negative type BC. OS of patients with HER-2 overexpression BC show statistical difference when Ki-67 index fell within 49% -60%, but there's no statistical difference among other subtypes. During molecular classi cation based on consensus of CSCO in 2019, OS of patients with Luminal A-type BC shows statistical difference with that of HER-2 over-expression BC in term of OS, but it shows no statistical difference with other subtypes. Differences caused by two different molecule classi cation methods might in uence prognosis judgment of patients. When Ki-67 index is 52%, there's statistical difference of OS among patients with Luminal B HER-2 (-) subtypes. Moreover, there's statistical difference of OS among patients with HER-2 over-expression BC when Ki-67 index is 47%. These further re ect differences of Ki-67 index among different molecular classi cation methods, which deserve special attentions.
In a word, a systematic study on application values of Ki-67 index to patients with BC is carried out.