|Year : 2017 | Volume
| Issue : 2 | Page : 177-184
|Ki-67 evaluation in breast cancer: The daily diagnostic practice
Lukasz Fulawka1, Agnieszka Halon2
1 Department of Pathology, Lower Silesian Oncology Centre; Department of Pathomorphology and Oncological Cytology, Wroclaw Medical University, Wroclaw, Poland
2 Department of Pathomorphology and Oncological Cytology, Wroclaw Medical University, Wroclaw, Poland
Click here for correspondence address and email
|Date of Web Publication||19-Jun-2017|
| Abstract|| |
Context: Breast cancer is the most common malignancy in females. It is routinely classified according to the WHO histological typing. However, there is also a molecular classification of breast cancer which is routinely substituted with surrogate subtypes based on expression of estrogen, progesterone, and human epidermal growth factor receptor 2 receptors and proliferation index (PI). PI is defined as the percentage of Ki-67-positive cells among overall cell population. The method commonly applied by pathologists to determine PI is visual scoring of the sample. Strict recommendations for PI assessment do not exist. Thus, the mode of PI evaluation differs significantly between pathologists. Aims: The aim of our study was to evaluate the daily approach to defining the PI. Settings and Design: Four practicing nonscholar pathologists were asked to evaluate PI in cases of invasive breast carcinoma. Subjects and Methods: The study was performed on a group of 98 patients diagnosed with invasive breast carcinoma. Immunohistochemical reaction was performed with monoclonal antibody against human Ki-67 antigen using Ventana BenchMark XT. Statistical Analysis Used: Results were compared using Pearson's and Spearman's rank correlation coefficients and Fleiss and Cohen's kappa values. Results: Statistical analysis showed pairwise Pearson's coefficients ranging between 0.77 and 0.84 (P < 0.001) and Spearman's rank correlation coefficients ranging between 0.68 and 0.83 (P < 0.001). The Fleiss kappa value for the 14% cutoff point was 0.58 whereas for the 20% cutoff point was 0.60. The pairwise Cohen's kappa values ranged from 0.45 to 0.69 for the 14% cutoff point and 0.53 to 0.67 for the 20% cutoff point. Friedman's rank ANOVA test showed significant differences among the four pathologists (P < 0.001). Conclusions: Our study shows a significant difference in results and methods of evaluation of PI between pathologists.
Keywords: Breast cancer, Ki-67, proliferation index
|How to cite this article:|
Fulawka L, Halon A. Ki-67 evaluation in breast cancer: The daily diagnostic practice. Indian J Pathol Microbiol 2017;60:177-84
| Introduction|| |
Breast cancer is the most common malignancy in females with an age-standardized incidence rate of 43.1/100,000 women per year. Thus, it is of great interest to clinicians, pathologists, and basic research scientists. There are a few classifications of breast cancer. The most commonly used classifications are WHO histological classification and molecular subtypes. Research based on gene expression arrays identified four molecular subtypes of breast cancer, differing in natural history and response to therapy., However, it is impossible to use molecular techniques in routine diagnostics due to high cost and the need for trained personnel, specialized equipment, and sophisticated methods of tissue acquisition and fixation. Therefore, the surrogate definitions of subtypes were based on immunohistochemistry (IHC).,,, The subtypes include luminal A-like, luminal B-like, human epidermal growth factor receptor 2 (HER2)-positive, and triple-negative [Table 1]. To classify breast cancer cases into these types, the IHC assay of estrogen receptor (ER), progesterone receptor (PgR), HER2, and Ki-67 protein needs to be performed. Expression level of these factors is evaluated by pathologists during routine pathological examination.
|Table 1: Surrogate definitions of molecular subtypes of breast cancer according to St Gallen 2013 consensus|
Click here to view
Defining breast cancer subtype is of key importance because it determines treatment scheme. Tumors positive for ER belong to luminal-like types which are further divided into luminal A-like and luminal B-like types. Breast cancers negative for ER and PgR and positive for HER2 belong to HER2-positive type, whereas those negative for all of the above-mentioned markers are classified as triple negative., Both proliferation index (PI) and the level of PgR expression are currently used to distinguish between luminal A- and B-like subtypes. Based on the initial gene-expression research, the ≥14% cutoff point for PI was accepted to define the luminal B-like subtype. It was later adopted into routine diagnostics and recommended by St. Gallen consensus in 2011 [Table 2]. However, it was later modified at St. Gallen conference in 2013. The 20% cutoff point for PI has been accepted by most of experts. Moreover, tumors positive for ER and with low level of PgR expression (<20%) were included into luminal B-like group regardless of PI. There is also a third luminal subtype called luminal B-HER2-positive. It is defined as ER-positive tumor with overexpression or amplification of HER2, regardless of PgR and Ki-67.
|Table 2: Surrogate definitions of molecular subtypes of breast cancer according to St Gallen 2011 consensus|
Click here to view
The determination of ER and PgR and HER2 status is based on generally accepted guidelines  which are included in recommendations for pathological reporting. Evaluation of Ki-67 status is more challenging. Generally, Ki-67 is the most widely accepted marker of proliferation activity. It is so because Ki-67 is expressed in all phases of the cell cycle other than G0.
Proliferation activity is determined as the percentage of Ki-67-positive cells among overall cell population, the measurement commonly known as PI. The routine method of Ki-67 antigen detection is IHC with anti-Ki-67 antibody on slides obtained from formalin-fixed paraffin-embedded tissue blocks. Pathologists are now required to determine PI to 1% accuracy. The method commonly applied by pathologists to determine PI is visual scoring of the sample. In practice, the nuclei are rarely counted one-by-one in microscopic field to determine a precise quotient of Ki-67 positive and total sum of nuclei. Instead, PI is estimated roughly by eyeballing.
| Subjects and Methods|| |
The study was performed on a group of 98 patients diagnosed with invasive breast carcinoma between 2012 and 2013 at the Lower Silesian Oncology Centre. The material consisted of 50 mammotome biopsies, 19 trucut biopsies, 7 lumpectomy, 13 quadrantectomy, and 9 mastectomy specimens. Their surrogate biological subtypes according to St. Gallen 2011 recommendations are listed in [Table 3].
|Table 3: Surrogate biological subtypes of study group according to St Gallen 2011 consensus|
Click here to view
Ethics, consent, and permissions
The study was approved by the Bioethical Committee of Wroclaw Medical University and the Director of Lower Silesian Oncology Centre.
IHC reaction was performed with monoclonal antibody against human Ki-67 (Clone 30-9, Ventana Medical Systems, Inc., Tucson, AZ, USA) and ultraView Universal DAB Detection Kit (Ventana Medical Systems, Inc., Tucson, AZ, USA). The procedure was performed automatically using Ventana BenchMark XT, in line with the producer's manual.
PI was scored visually by four pathologists blinded to each other's scores. The mode of scoring was significantly different among specialists. The pathologists were asked to fill in the survey regarding the mode of PI assessment [Figure 1]. The questions were as follows:
- Do you take hotspots or randomly chosen areas into account?
- Having assessed the first area, do you examine an adjacent or a remote field?
- Do you count cells one-by-one or by eyeballing?
- How many areas do you examine?
- What magnification do you use?
- What is the brown color (DAB, Diaminobenzidine) intensity cutoff value?
Statistica 12 (StatSoft, Tulsa, OA, USA) and Microsoft Excel Software were used for statistical analysis. Pearson's coefficients and Spearman's rank correlation were calculated for each pair of observers. Interobserver agreements were assessed with Cohen's kappa for each pair of observers and Fleiss kappa for all sets of observers. PI variable was categorized into binary variable. The division was made twice, for both 14% and 20% cutoff points. Nonparametric Friedman's ANOVA was also implemented to compare results between all observers.
| Results|| |
The mode of PI evaluation differed significantly between pathologists as shown by results in [Figure 1]. Hotspots were chosen by all but one pathologist (pathologist 4). Pathologist 1 was the only one to have taken into account the areas adjacent to each other. Pathologists 1–3 did their count by eyeballing, whereas Pathologist 4 counted the nuclei one-by-one. The number of areas examined differed significantly between the pathologists and was as follows: 3, at least 5, 4, and 20. All but one observer did their count using ×400 magnification. Pathologist 3 used ×200 magnification. Pathologist 1 counted all brown-colored nuclei regardless of intensity, whereas the other observers took into account nuclei with 2+ or higher intensity. The results of PI assessment for each case and observer are shown in [Figure 2]. The pairwise Pearson's coefficients were as follows [Figure 3]: 0.82, 0.84, 0.77, 0.83, 0.75, 0.76 (P < 0.001). The Spearman's rank correlation coefficients were the following [Figure 3]: 0.83, 0.77, 0.68, 0.83, 0.72, and 0.67 (P < 0.001). The Fleiss kappa value for the 14% cutoff point was 0.58 whereas for the 20% cut-off point was 0.60. After excluding Pathologist 4 from the dataset, the Fleiss kappa was 0.63 for the 14% cutoff point and 0.66 for the 20% cutoff point. The pairwise Cohen's kappa values for the 14% cutoff point were as follows [Table 4]: 0.69, 0.54, 0.56, 0.68, 0.62, 0.45. The respective values for the 20% cutoff point were as follows [Table 4]: 0.66, 0.67, 0.53, 0.67, 0.53, and 0.56. Friedman's rank ANOVA test showed significant differences among the four pathologists (P < 0.001). However, post hoc tests showed statistically significant differences only between PI values assessed by Pathologist 4 compared to the other observers [Table 5]. The differences between Pathologists 1, 2, and 3 were not statistically significant (P value of Friedman's test excluding Pathologist 4 was 0.16).
|Figure 3: Matrix scatterplot for four pathologists with correlation coefficients. RP: Pearson's linear correlation coefficient, RS: Spearman's rank correlation coefficient|
Click here to view
|Table 4: Pairwise Cohen's kappa values for the 14% and 20% cutoff points|
Click here to view
|Table 5: P values of the pairwise Friedman's ANOVA with post hoc Bonferroni adjustment|
Click here to view
| Discussion|| |
PI determined as a percentage of Ki-67 antigen-positive cancer cells is an important marker in tumor diagnostics. It is of great value in breast cancer because it is used to distinguish between luminal A-like and luminal B-like subtypes., The ≥14% cutoff point for PI best correlated with the initial gene-expression definition of luminal B type. It was later accepted by St. Gallen consensus in 2011 to be used in routine diagnostics  and modified at St. Gallen conference in 2013. A level of ≥20% has been accepted by most of experts to define luminal B-like subtype. The current St. Gallen 2015 recommendations that were announced to the public while this paper was being written maintain the role of the Ki-67 marker in differentiating between luminal A-like and luminal B-like HER2-negative breast cancer. It is also stated that “Ki-67 use requires knowledge of local laboratory values” and “the minimum value of Ki-67 required for 'luminal B-like' was for the majority of the panel 20%–29%. About one-fifth (of voters) stated that Ki-67 should not be used for this distinction.”
Our fellow pathologists and we call into question the St. Gallen recommendations formulated by clinical oncologists without taking into account the opinion of pathologists. It is noteworthy that European pathologists and clinical oncologists rely more strongly on Ki-67 index than their American colleagues. It is reflected in the College of American Pathologists Cancer protocol template for breast biomarker reporting which states that “routine testing of breast cancers for Ki-67 expression is not currently recommended by either American Society of Clinical Oncology or the National Comprehensive Cancer Network.”
The aim of our study was to evaluate the daily approach to defining the PI. Ki-67 assessment has been discussed in several papers. However, they are based on their respective authors' work only. In the present study, we asked our colleagues who are practicing nonscholar pathologists to define PI by the method they use in their daily practice. They were unaware of the scientific approach that was used to evaluate their results. That was to ensure that PI was evaluated without “science-induced bias.” Moreover, due to anonymity, the observers were unrestricted and were not afraid that their mode of assessment would be criticized by others.
The pairwise Pearson's coefficients ranged from 0.75 to 0.84 (P < 0.001), whereas Spearman's rank correlation coefficients were 0.67–0.83 (P < 0.001). These results were similar to other authors' findings (Pearson's coefficients 0.73–0.91, P≤ 0.0001; Spearman's rank correlation 0.74–0.91, P≤ 0.0001; three observers, thirty cases). The Fleiss kappa value for the 14% cutoff point was 0.58 whereas for the 20% cutoff point was 0.60. Other authors reported Fleiss kappa as 0.65 for the 14% cutoff point (four observers, 99 cases). The pairwise Cohen's kappa values ranged from 0.45 to 0.69 for the 14% cutoff point (0.54–0.69 when Pathologist 4 was excluded) and 0.53–0.67 for the 20% cutoff point (0.66–0.67 when Pathologist 4 was excluded). Other authors divided their group into three categories depending on PI values: ≤8%, 8%–30% and >30%. The Cohen's kappa values were 0.56–0.72 (five observers, ten cases). In yet another study, there were three groups of patients: ≤15%, 16%–30%, and >30%. The Cohen's kappa values differed significantly from ours (0.13–0.61; three observers, thirty cases). Friedman's ANOVA with post hoc Bonferroni adjustment showed a statistically significant difference between Pathologist 4 and other observers (P = 0.0002).
What are the causes of significant differences between Pathologist 4 and other observers? There are a few factors that affect PI evaluation results. One of them is the selection of areas. It goes without question that giving the precise number of Ki-67-positive nuclei is impossible. Even if we spared no effort to count them, a bias would still exist resulting from tumor sampling on gross dissection and slice cutting. Thus, evaluation of PI on the basis of a few areas selected by the observer is all that can be done. The International Ki-67 in Breast Cancer Working Group recommends that PI “should be scored in fields that are seen to be representative on an initial overview of the whole section.” The authors state that heterogeneity “can occur” and hotspots “may occur.” In our opinion, Ki-67 staining heterogeneity occurs in almost all cases of breast cancer [Figure 4] and [Figure 5]. Hence, using the words “may” or “can” seems inappropriate. Even in cases that seem to be otherwise homogeneously stained, the heterogeneity is apparent on precise counting. The difference in PI between the fields that appear to be equal can reach as much as 20% for one-by-one counting in the image displayed on the computer screen (unpublished results). In the present study, Pathologist 4 chose a random representative area in the scanning view fields, whereas other observers evaluated PI in hotspot areas. In similar studies, hotspot areas were preferred.,
|Figure 4: Ki-67 staining heterogeneity. (a) Scanning view. The area on the left consists of apparently lower number of positive nuclei compared to the area on the right, (b) higher magnification of the same tumor from less proliferative area, (c) higher magnification of the same case from more proliferative area|
Click here to view
|Figure 5: Ki-67 staining heterogeneity. (a) Apparent hotspot area, (b) the same tumor from randomly chosen area. Please note brisk lymphocytic infiltrate around the group of tumor cells which is also the factor affecting the precision of proliferation index evaluation|
Click here to view
In November 2012, an informal breast pathology group of the Polish Society of Pathologists decided to choose one hotspot and count nuclei in areas adjacent to it [Figure 1], which was an unprecedented approach. In the present study, Pathologist 1 chose this way of counting.
The International Ki-67 in Breast Cancer Working Group recommended to count at least 1000 cells and suggested that 500 cells should be accepted as the absolute minimum. However, our fellow colleagues and we agree that it is impracticable in laboratories like ours, where a significant number of breast cancer specimens need to be examined. We also maintain that in countries with substantial shortage of pathologists, precise one-by-one counting of tumor nuclei may delay the diagnosis. Thus, in daily routine diagnostics, PI is estimated roughly by eyeballing. In our survey, three out of four persons semi-quantitatively estimated PI by eyeballing, whereas Pathologist 4 stated that he counted the nuclei precisely. The mode of counting applied by Pathologist 4 was to precisely count the nuclei in twenty cells squared area for each twenty fields evaluated. In their study, Vörös et al. observed that all three persons did their count using their individual approach. Pathologist 1 counted 100 cells (positive and negative) and then determined the number of positive nuclei in the same 100 cells. Observer 2 counted ten tumor cells to estimate the area encompassed by these cells. Then, the positive nuclei were counted in the area 10 times greater. Observer 3 determined the number of the stained nuclei in twenty or thirty cells and then multiplied the proportion by four or three. In the study by Mikami et al., each pathologist applied two different modes to assess PI. First, the percentage of positive nuclei was estimated semi-quantitatively and then subdivided into ten groups every 10%. In the second method, 1000 nuclei in total were counted including the stained ones.
We agree with other authors that ×400 magnification should be applied.,, In our study, one person (Pathologist 3) used ×200 magnification. The Ki-67 staining intensity can be different between nuclei in the tumor [Figure 6]. Several authors recommend to count all the positive nuclei regardless of intensity.,, In our study, only one person, Pathologist 1, counted all brown-colored nuclei regardless of intensity. The other observers took into account nuclei with 2+ or higher intensity. Other authors observed that pathologists do not take “very weakly, faintly staining nuclei” into account. Varga et al. investigated the relevance of difference in staining intensity taken into account by different observers. Their results suggest that this is not the factor that significantly influences the interobserver variability.
PI scoring in breast cancer is of key importance because it plays critical role in distinguishing between luminal A-like and luminal B-like subtypes of breast cancer. In turn, cancer subtype determines treatment scheme. The 20% cutoff point for PI has been accepted by most of experts according to St. Gallen 2013 and 2015 recommendations.
| Conclusion|| |
Our study revealed significant differences between pathologists asked to evaluate PI. Considering 20% cutoff point, the significant number of patients would be differently classified into low and high proliferative activity group [Table 6], asterisks (*)]. These patients would have been administered different therapeutic strategies. On account of this, the issue of significant interobserver variability has been often raised during pathology conferences. However, there is still lack of precise, generally accepted, guidelines on how PI should be measured.
|Table 6: Pairwise classifification into low and high proliferative activity group based on the 20% cutoff point|
Click here to view
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
| References|| |
Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser S, Mathers C, et al
. GLOBOCAN 2012 v1.0, Cancer Incidence and Mortality Worldwide: IARC Cancer Base No. 11; 2013. Available from: http://www.globocan.iarc.fr
. [Last accessed on 2015 Apr 13].
Perou CM, Sørlie T, Eisen MB, van de Rijn M, Jeffrey SS, Rees CA, et al.
Molecular portraits of human breast tumours. Nature 2000;406:747-52.
Goldhirsch A, Winer EP, Coates AS, Gelber RD, Piccart-Gebhart M, Thürlimann B, et al.
Personalizing the treatment of women with early breast cancer: Highlights of the St. Gallen International Expert Consensus on the Primary Therapy of Early Breast Cancer 2013. Ann Oncol 2013;24:2206-23.
Goldhirsch A, Wood WC, Coates AS, Gelber RD, Thürlimann B, Senn HJ; Panel Members. Strategies for subtypes – Dealing with the diversity of breast cancer: Highlights of the St. Gallen International Expert Consensus on the Primary Therapy of Early Breast Cancer 2011. Ann Oncol 2011;22:1736-47.
Cheang MC, Chia SK, Voduc D, Gao D, Leung S, Snider J, et al.
Ki67 index, HER2 status, and prognosis of patients with luminal B breast cancer. J Natl Cancer Inst 2009;101:736-50.
Nielsen TO, Hsu FD, Jensen K, Cheang M, Karaca G, Hu Z, et al.
Immunohistochemical and clinical characterization of the basal-like subtype of invasive breast carcinoma. Clin Cancer Res 2004;10:5367-74.
Lester SC, Bose S, Chen Y, Connolly JL, de Baca M, Fitzgibbons PL, et al
. Protocol for the Examination of Specimens from Patients with Invasive Carcinoma of the Breast; 2012. Available from: http://www.cap.org
. [Last accessed on 2015 Aug 07].
Dowsett M, Nielsen TO, A'Hern R, Bartlett J, Coombes RC, Cuzick J, et al.
Assessment of Ki67 in breast cancer: Recommendations from the International Ki67 in Breast Cancer working group. J Natl Cancer Inst 2011;103:1656-64.
Gnant M, Thomssen C, Harbeck N. St. Gallen/Vienna 2015: A brief summary of the consensus discussion. Breast Care (Basel) 2015;10:124-30.
Fitzgibbons PL, Dillon DA, Alsabeh R, Berman MA, Hayes DF, Hicks DG, et al.
Template for reporting results of biomarker testing of specimens from patients with carcinoma of the breast. Arch Pathol Lab Med 2014;138:595-601.
Vörös A, Csörgo E, Nyári T, Cserni G. An intra- and interobserver reproducibility analysis of the Ki-67 proliferation marker assessment on core biopsies of breast cancer patients and its potential clinical implications. Pathobiology 2013;80:111-8.
Pavlakis K, Vrekoussis T, Tsipoura A, Tsenga A, Arapantoni-Dadioti P, Gavresea T, et al.
Her2 negative luminal breast carcinoma and Ki-67 evaluation. Breast 2012;21:529-33.
Varga Z, Diebold J, Dommann-Scherrer C, Frick H, Kaup D, Noske A, et al.
How reliable is Ki-67 immunohistochemistry in grade 2 breast carcinomas? A QA study of the Swiss Working Group of Breast- and Gynecopathologists. PLoS One 2012;7:e37379.
Mikami Y, Ueno T, Yoshimura K, Tsuda H, Kurosumi M, Masuda S, et al.
Interobserver concordance of Ki67 labeling index in breast cancer: Japan Breast Cancer Research Group Ki67 ring study. Cancer Sci 2013;104:1539-43.
Vörös A, Csörgo E, Kovári B, Lázár P, Kelemen G, Rusz O, et al.
Different methods of pretreatment Ki-67 labeling index evaluation in core biopsies of breast cancer patients treated with neoadjuvant chemotherapy and their relation to response to therapy. Pathol Oncol Res 2015;21:147-55.
Varga Z, Cassoly E, Li Q, Oehlschlegel C, Tapia C, Lehr HA, et al.
Standardization for Ki-67 assessment in moderately differentiated breast cancer. A retrospective analysis of the SAKK 28/12 study. PLoS One 2015;10:e0123435.
Viale G, Giobbie-Hurder A, Regan MM, Coates AS, Mastropasqua MG, Dell'Orto P, et al.
Prognostic and predictive value of centrally reviewed Ki-67 labeling index in postmenopausal women with endocrine-responsive breast cancer: results from Breast International Group Trial 1-98 comparing adjuvant tamoxifen with letrozole. J Clin Oncol 2008;26:5569-75.
Offersen BV, Sørensen FB, Knoop A, Overgaard J; Danish Breast Cancer Cooperative Tumour Biology Committee. The prognostic relevance of estimates of proliferative activity in early breast cancer. Histopathology 2003;43:573-82.
Fulawka L, Halon A. Proliferation Index Evaluation in Breast Cancer Using ImageJ and ImmunoRatio Applications. Anticancer Res 2016;36:3965-72.
Department of Pathology, Lower Silesian Oncology Centre, Pl. HIrszfelda 12, 53-413 Wroclaw, Poland
Source of Support: None, Conflict of Interest: None
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6]
[Table 1], [Table 2], [Table 3], [Table 4], [Table 5], [Table 6]
|This article has been cited by|
||Practical approaches to automated digital image analysis of Ki-67 labeling index in 997 breast carcinomas and causes of discordance with visual assessment
| ||Ah-Young Kwon,Ha Young Park,Jiyeon Hyeon,Seok Jin Nam,Seok Won Kim,Jeong Eon Lee,Jong-Han Yu,Se Kyung Lee,Soo Youn Cho,Eun Yoon Cho,Irina V. Lebedeva |
| ||PLOS ONE. 2019; 14(2): e0212309 |
|[Pubmed] | [DOI]|
||Double-positive in triple-negative? How significant is basal cytokeratin expression in breast cancer?
| ||Anusree Majumder,Rajat Jagani,Atoshi Basu |
| ||Medical Journal Armed Forces India. 2019; |
|[Pubmed] | [DOI]|
| Article Access Statistics|
| Viewed||8530 |
| Printed||102 |
| Emailed||0 |
| PDF Downloaded||436 |
| Comments ||[Add] |
| Cited by others ||2 |