|Year : 2020 | Volume
| Issue : 3 | Page : 289-295
Study of Ki-67 index in the molecular subtypes of breast cancer: Inter-observer variability and automated scoring
Divya Meermira, Meenakshi Swain, Swarnalata Gowrishankar
Department of Histopathology, Apollo Hospitals, Jubilee Hills, Hyderabad, Telangana, India
|Date of Submission||03-Nov-2018|
|Date of Decision||28-Apr-2019|
|Date of Acceptance||17-May-2019|
|Date of Web Publication||22-Jul-2020|
Department of Histopathology, Apollo Hospitals, Jubilee Hills, Hyderabad, Telangana
Source of Support: None, Conflict of Interest: None
Background: Ki-67 index is an important prognostic marker in breast cancer and is also used to differentiate luminal A subtype from luminal B. Inter-observer variations in determining the index and the cut-off value to be considered in distinguishing the two subtypes remain problems in clinical practice.
Methods: MIB-1 immunohistochemistry was done on 200 cases of breast cancer with 50 cases in each molecular subtype. The Ki-67 scoring was done manually by two observers and automated method (using the software ImmunoRatio). The mean value of Ki-67 was calculated in each molecular group and in the entire estrogen receptor and progesterone receptor (ER/PR) positive group. The inter-observer variability between the two observers and the automated method was also assessed.
Results: The mean and median values of Ki-67 of all the 200 cases obtained by manual scoring was 31.13% and 29.65% by observer 1, 28.48% and 27.90% by observer 2, and 38.27% and 35.45% by the automated method. The mean Ki-67 value obtained by manual scoring, in luminal A, luminal B, HER2 enriched and triple negative was 21.07%, 37.19%, 33.72% and 27.27%, respectively. There was significant correlation between the two observers and with the automated scoring.. The mean value of the Ki-67 index in the ER/PR positive group was 29.1%.
Conclusion: The inter-observer correlation and the correlation with the automated scoring system of the Ki-67 index was good. 29.1% was the mean Ki-67 index in the ER/PR positive group and this value was within the acceptable range as per St Galen's recommendation.
Keywords: Automated scoring, breast cancer, inter-observer variability, Ki-67 labeling index
|How to cite this article:|
Meermira D, Swain M, Gowrishankar S. Study of Ki-67 index in the molecular subtypes of breast cancer: Inter-observer variability and automated scoring. Indian J Cancer 2020;57:289-95
|How to cite this URL:|
Meermira D, Swain M, Gowrishankar S. Study of Ki-67 index in the molecular subtypes of breast cancer: Inter-observer variability and automated scoring. Indian J Cancer [serial online] 2020 [cited 2020 Sep 20];57:289-95. Available from: http://www.indianjcancer.com/text.asp?2020/57/3/289/290552
See the related editorial Shet T. Ki.67 in breast cancer: Simulacra and simulation. Indian J Cancer 2020;57:231-3
| » Introduction|| |
Breast cancer is one of the most common malignancies in women and second most common cause of cancer death in the world. There is an increased breast cancer burden of 57% in India. Therefore, diagnostic and prognostic markers of breast cancer are gaining importance in clinical management.
Breast cancer is a genetically and clinically heterogeneous disease. It results due to alterations that take place in genes that govern cell growth and proliferation. The predominant form of breast cancer is sporadic, in which oncogenes that are mutated lead to uncontrolled cell proliferation.
Uncontrolled proliferation is the main feature of malignancy. One such marker of proliferation is Ki-67, or monoclonal antibody (MKI67). It is expressed in proliferating cells during all phases of cell cycle except G0 phase. It is expressed only in those cells with defective DNA repair process. As this defect leads to uncontrolled proliferation, Ki-67 is a marker of cell proliferation. It is assessed by immunohistochemical (IHC) staining using MIB-1 antibody.
There is a strong correlation between Ki-67 levels and other prognostic factors of breast cancer (e.g. Estrogen and Progesterone receptor (ER/PR, HER2)). In addition, these prognostic indicators are an important tool for selecting the treatment modalities for patients. Also, high Ki-67 levels are found in poorly differentiated tumors with a high mitotic rate and are associated with early recurrence of breast cancer.
Ki-67 is used as a marker to define the molecular subtypes of breast cancer. The Ki-67 value is used to distinguish between the low risk, good prognostic group luminal A and high risk, bad prognostic luminal B subtypes of breast cancer. But the cut-off levels of this Ki-67 value used to make this distinction range from 5% to 34%.
Though a promising biomarker, the main limitation is the lack of inter-observer concordance. The variability of Ki-67 in different studies vary between 5% and 34%. The St. Gallen's 2015 meeting advocates that each laboratory determine their cut-offs based on the local laboratory values. This can be done by arriving at the mean value in the group of breast cancers with ER/PR positivity. This level could then be used to divide the luminal A and luminal B subtypes.
| » Materials and Methods|| |
The study was conducted over a period of 8 years (retrospective study from January 2009 to August 2015 and prospective study of 2 years from September 2015 to June 2017) in the Department of Histopathology, Apollo hospitals, Jubilee hills, Hyderabad. It included all the lumpectomy, simple mastectomy and modified radical mastectomy specimens of carcinoma breast with known ER, PR and HER2 status diagnosed at our department.
We analyzed 200 cases of breast carcinoma, which included 50 cases in each molecular subtype. Cases for evaluation were chosen from the histopathology records after application of the inclusion and exclusion criteria noted as above.
For retrospective cases, hematoxylin and eosin (H and E) stained section slides and paraffin blocks of representative tissue with known ER, PR and HER2 status were retrieved. For prospective cases, the specimens received were grossed after proper fixation and adequate sections were taken. Then for all the cases, H and E, along with IHC staining for ER, PR, HER2 and Ki-67 were performed [Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5].
|Figure 1: Luminal A molecular subtype; (a) ER positive, (b) PR positive, (c) HER2 negative, (d) Low MIB-1|
Click here to view
|Figure 2: Luminal B Her2 negative molecular subtype; (a) ER positive, (b) PR low positive, (c) HER2 negative, (d) High MIB-1|
Click here to view
|Figure 3: Luminal B HER 2 positive molecular subtype; (a) ER positive, (b) PR positive, (c) HER2 positive, (d) MIB-1|
Click here to view
|Figure 4: HER2/neu enriched molecular subtype; (a)ER negative, (b) PR negative, (c) Her2 positive (d) MIB-1|
Click here to view
|Figure 5: Triple negative; (a)ER negative, (b) PR negative, (c) Her2 negative (d) MIB-1|
Click here to view
Around 4-5 μm sections were cut from a paraffin block of tumor tissue and taken on glass slides. Coated glass slides with adhesive were taken for immunohistochemistry.
Clone used for ER was Ventana anti ER Rabbit monoclonal primary antibody SP1, for PR was Ventana anti-PR Rabbit monoclonal primary antibody 1E2, for HER2/neu was Ventana anti-HER2 Rabbit monoclonal primary antibody 4B5 and for Ki-67 was Dako monoclonal mouse anti-human antibody MIB-1. All of these were ready to use kits. No dilutions were performed. Ventana Benchmark XT-automated IHC slide staining system was used.
All the cases were then divided into the following four subgroups, based on ER, PR and HER2 status [Table 1].
|Table 1: The molecular subtypes were divided according to the St Gallen's recommendations as follows|
Click here to view
Two observers were selected to participate in the study.
Immunohistochemistry (IHC)-stained Ki-67 slides were submitted to two observers independently. Both of the observers were blinded regarding the patient outcome or other observer's scoring.
Scoring of Ki-67 was done independently, irrespective of fields of study and number of cells (preferably 1000, but minimum of 500 cells to be counted). After scoring, a category of high or low is designated based on a cut-off of 15%.
Then, the same slides were scored using the automated image analysis software IMMUNORATIO (automated determinator of Ki-67 index).,, The category of high or low was designated [Figure 6] and the results were then analyzed.
Method for interpreting and scoring of Ki-67,
Mouse anti-human Ki-67 monoclonal antibody, MIB-1 was used in IHC staining of Ki-67.
The invasive edge of the tumor was selected to score Ki-67. At least three fields were scored at the periphery of the tumor.
Only nuclear staining (plus mitotic figures which are stained by Ki-67) of any intensity were considered positive.
A minimum of 500 tumor cells were counted.
Ki-67 score was taken as the percentage of positively stained cells among the total number of malignant cells counted.
Internal positive controls of mitotic figures, normal ducts, lymphocytes or endothelial and stromal cells were taken.
The Ki-67 index value was calculated in all the 200 cases of breast cancer, the distribution range was observed and the median value was obtained. Mean Ki-67 index values of cases in each subgroup were calculated separately.
To study the inter-observer variability, the data collection sheets were given to two observers independently and the scores were obtained. Using the intra-class correlation coefficient value, the significance of variability was assessed.
To compare the manual scoring with automated scoring, a cut-off of 15% was considered to categorize high and low Ki-67. Therefore, cases with Ki-67 levels of ≥15% were taken as high and <15% were taken as low Ki-67., The average value of the two observers were taken for each case. The percentage of cases showing high/low Ki-67 by manual staining was compared with the percentage of cases showing high/low Ki-67 by automated scoring. Using Chi-square test, P value was calculated and significance was determined (</>0.05).
| » Results|| |
In this study, the mean values of Ki-67 of all the 200 cases and in each subgroup was studied, along with inter-observer variability and the variability among manual and automated score. The study was done on a total of 200 cases. The results were as follows:
Ki-67 index of all the 200 cases ranged from 0% to 89%; 10.8% cases were grade 1, 40.7% were grade 2, and 48.4% were grade 3 tumors histologically. Among luminal A cases, 11 cases were grade 1, 28 cases were grade 2, and 9 cases were grade 3. Among the triple-negative cases, 35 cases were grade 3 and 15 cases were grade 2.
Ki-67 index of all the 200 cases were calculated by manual (observer 1 and observer 2) and automated methods. The mean Ki-67 value of all the 200 cases of observer 1 was 31.13% and observer 2 was 28.48%. The average value was 29.81%. The mean value of Ki-67 of all the 200 cases obtained by automated method was 38.27%.
The median value of the Ki-67 of the 200 cases was 29.65 and 27.90 by observer 1 and observer 2, respectively. The median value of Ki-67 obtained by automated method was 35.45 [Table 2].
The mean value of Ki-67 was also calculated separately in the four subgroups. The values obtained by manual method were 21.07%, 37.19%, 33.72% and 27.27% in subgroup A, subgroup B, subgroup C and subgroup D, respectively.
The mean value of Ki-67 in ER/PR positive cases, which includes luminal A and luminal B subgroups is 29.1%.
Similarly, the mean values of Ki-67 in different subgroups obtained by automated method were 30%, 49%, 42% and 32% in subgroup A, subgroup B, subgroup C and subgroup D, respectively [Table 3].
The variability of manual scoring between the two observers was assessed. Inter-item correlation and intra-class correlation coefficient was calculated. Our study showed a good correlation of the Ki-67 values between the two observers. The inter-item correlation matrix was 0.895. The intra-class correlation coefficient (ICC) was 0.892. P value was observed to be 0.00 (i.e., <0.05), which is highly significant. Therefore, it shows that the variability between the two observers across the different subgroups by manual method is not significant.
Comparison of manual and automated scoring method
It was observed that there was significant correlation between the Ki-67 scores obtained by manual and automated scoring method; therefore, the variability was significantly less. The inter-item correlation coefficient was 0.741 and the intra-class coefficient value was 0.710. However, the variability was more than the inter-observer variability obtained by manual scoring method.
In our study, a cut-off of 15% was taken to distinguish high Ki-67 from low Ki-67. Ki-67 value of ≥15% was considered high and values<15% were considered as low Ki-67. Therefore, we observed that by manual scoring, out of all the 200 cases, 160 cases (80%) showed high Ki-67 and 40 cases (20%) showed low Ki-67 value.
Similarly, by the automated scoring method, 157 cases (78.5%) showed high Ki-67 value and 43 cases (21.5%) showed low Ki-67 value [Graph 1].
Comparison of cases with high and low Ki-67 values obtained from automated and manual methods [Table 4].
|Table 4: Comparison of high and low Ki-67 values between automated and manual methods |
Click here to view
From the above obtained values, using the Pearson's Chi-square test, P value was calculated. The P value obtained was 0.00, which is less than 0.05 (significant). Therefore, there is significant correlation between the manual and automated scoring, when a cut-off value of 15% is used.
| » Discussion|| |
Mean and median Ki-67 values
The main aim of our study was to assess the Ki-67 index in all the 200 cases of breast cancer that were included. The mean Ki-67 value of all the cases was calculated and the value obtained was 29.81%. The median value was 28.35% and the values ranges from 0% to 89%.
Our values were comparable to studies done by Thar Htet et al. and Jinzhong Sun et al. which showed mean Ki-67 values of 33.9%, 31.22% and 31.3%, respectively, and median Ki-67 values of 28%, 25% and 26%, respectively.
St Gallen's advocates that the mean/median value of Ki-67 of ER/PR positive cases, between 20% and 29% is acceptable and that value of the laboratory can be used as a cut-off to distinguish luminal A from luminal B.
Our study showed a mean value of 29.1% in ER/PR positive cases (luminal A and luminal B), which was within the acceptable range as per St Gallen's recommendations.
Mean Ki-67 value in the molecular subtypes
The mean Ki-67 value in each molecular subgroup was also studied which included 50 cases in each subgroup. Our study showed mean Ki-67 index value of 21.07%, 37.19%, 33.72% and 27.27% in subgroup A (luminal A), subgroup B (luminal B), subgroup C (HER2 enriched) and subgroup D (triple negative), respectively.
The mean values of Ki-67 obtained in subgroups A, B and C in our study were comparable with other studies (most closely with Rumiko Tashima et al. and Haroon et al.). But the mean value of Ki-67 of the subgroup D (triple negative) of our study was lesser than that seen in other studies. This was a retrospective study with archival material, in all of which the fixation and processing was not controlled, though internal and external controls were satisfactory in all. Loss of antigen due to suboptimal fixation and processing could be an explanation for this anomaly.
Luminal A subgroups have the best prognosis among all the molecular subtypes and show the best outcome. Luminal B have the second best prognosis after which is HER2 enriched. Triple negative cases have the worst prognosis and have poorest outcome. Therefore, the treatment modalities also change depending on subgroup. As Ki-67 is a marker of proliferation, its value increases from luminal A (low proliferative activity, low Ki-67) to triple negative (high proliferative activity, high Ki-67). Thus Ki-67 plays an important role in determining prognosis and also for the treatment modality.
A significant correlation was observed between the two observers (by manual scoring) in our study. Our findings were comparable with other studies such as Focke et al., Shui et al., Chung et al. and Yamamoto et al. which showed inter-observer concordance except for a study by Mikami et al. which showed high inter-observer variability.
As Ki-67 is an independent prognostic factor for disease-free survival, standardization of the staining techniques and scoring methods are necessary in order to incorporate this biomarker in routine practice.
Variability between manual and automated Ki-67 scoring
In our study, the Ki-67 values obtained by manual counting method were compared with the Ki-67 values obtained by automated scoring method using ImmunoRatio. It showed the reliability and accuracy of automated scoring method. But the variability was more than the inter-observer variability obtained by manual scoring.
Our study was comparable to studies done on ImmunoRatio by Min-Kyung Yeo et al., Fangfang Zong et al. and Tushar Mungle et al., all of which showed a good correlation between the Ki-67 scores. Therefore, ImmunoRatio could be used as a reliable tool to score Ki-67.
In our study, we also evaluated the percentage of cases with high and low levels of Ki-67 using the cut-off value as 15%. Using Pearson Chi-square test, P value obtained was less than 0.05. Therefore, there was a good correlation between manual and automated scoring when the cut-off level was 15%. A similar study by Mohammad et al. showed comparable results using 15% as the cut-off. In this study, the automated and manual determinations of Ki-67 were strongly correlated with correlation coefficient of 0.87 and r = 0.94.
The limitation of the study was a lack of follow-up of our patients as well as the correlation of the molecular subtypes with the clinical stage. This study was an emphasis on Ki-67 index in the subtypes, the inter-observer variation and the use of an automated method.
| » Conclusion|| |
ER, PR, HER2 and Ki-67 immunohistochemistry should be done in all the cases with invasive breast cancer for molecular subtyping, because the prognosis and treatment modalities are different for each subgroup.
We recommend that scoring of Ki-67 should be done at the invasive edge of tumor and by counting a minimum of 1000 tumor cells to decrease the inter-observer variability.
The automated Ki-67 scorer, ImmunoRatio which was used in the study showed a fairly good correlation with manual scoring; hence, we recommend that the software ImmunoRatio could be used for scoring Ki-67.
The mean value of Ki-67 of ER/PR positive cases obtained was 29.1%. Therefore, we propose that a cut-off as 30% in our laboratory, which is in keeping with a high cut-off, according to the St Gallen's recommendation could be used in our laboratory to distinguish between luminal A and luminal B molecular subtype of breast cancer.
Department of histopathology, Apollo Hospitals, Jubilee Hills. Technical staff, Apollo Hospitals, Jubilee Hills.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
| » References|| |
Siegel RL, Miller KD, Jemal A. Cancer Statistics, 2017. CA Cancer J Clin 2017;67:7-30.
Ali I, Wani WA, Saleem K. Cancer scenario in India with future perspectives. Cancer Ther 2011;8:56-70.
Malhotra GK, Zhao X, Band H, Band V. Histological, molecular and functional subtypes of breast cancers. Cancer Biol Ther 2010;10:955-60.
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.
Jang MH, Kim HJ, Chung YR, Lee Y, Park SY. A comparison of Ki-67 counting methods in luminal Breast Cancer: The average method vs. the hot spot method. PLoS One 2017;12:1-15.
Jones RL, Salter J, A'Hern R, Nerurkar A, Parton M, Reis-Filho JS, et al
. The prognostic significance of Ki67 before and after neoadjuvant chemotherapy in breast cancer. Breast Cancer Res Treat 2009;116:53-68.
Dede DS, Gumuskaya B, Guler G, Onat D, Altundag K, Ozisik Y. Evaluation of changes in biologic markers ER, PR, HER 2 and Ki-67 index in breast cancer with administration of neoadjuvant dose dense doxorubicin, cyclophosphamide followed by paclitaxel chemotherary. J BUON 2013;18:366-71.
Mir R, Johnson H Jr, Mathur R, Wise L, Kahn LB. Ki-67 reactivity in breast carcinoma analyzed by a computer-assisted image system: Preliminary results. J Natl Med Assoc 1995;87:554-9.
Mannell A. The role of Ki-67 in breast cancer. S Afr J Surg. 2016;54:10-3.
Denkert C, Loibl S, Müller BM, Eidtmann H, Schmitt WD, Eiermann W, et al
. Ki67 levels as predictive and prognostic parameters in pretherapeutic breast cancer core biopsies: A translational investigation in the neoadjuvant GeparTrio trial. Ann Oncol 2013;24:2786-93.
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.
Mohammed ZMA, McMillan DC, Elsberger B, Going JJ, Orange C, Mallon E, et al
. Comparison of visual and automated assessment of Ki-67 proliferative activity and their impact on outcome in primary operable invasive ductal breast cancer. Br J Cancer 2012;106:383-8.
González-González R, Molina-Frechero N, Carreón-Burciaga RG, López-Verdín S, Robles-Bonilla C, Pereira-Prado V, et al
. Comparison between manual and automated methods for Ki-67 immunoexpression quantification in ameloblastomas. Anal Cell Pathol 2016;2016:1-8.
Tuominen VJ, Ruotoistenmäki S, Viitanen A, Jumppanen M, Isola J. ImmunoRatio: A publicly available web application for quantitative image analysis of estrogen receptor (ER), progesterone receptor (PR), and Ki-67. Breast Cancer Res 2010;12:R56.
Yeo M, Kim HE, Kim SH, Chae BJ, Song BJ, Lee A. Clinical usefulness of the free web-based image analysis application ImmunoRatio for assessment of Ki-67 labelling index in breast cancer. J Clin Pathol 2017;10:1-5.
Polley M-YC, Leung SCY, McShane LM, Gao D, Hugh JC, Mastropasqua MG, et al
. An international Ki67 reproducibility study. JNCI J Natl Cancer Inst 2013;105:1897-906.
Penault-Llorca F, Radosevic-Robin N. Ki67 assessment in breast cancer: An update. Pathology 2017;49:166-71.
Soliman NA, Yussif SM. Ki-67 as a prognostic marker according to breast cancer molecular subtype. Cancer Biol Med 2016;13:496-504.
San TH, Fujisawa M, Fushimi S, Soe L, Wah N, Yoshimura T, et al
. Molecular subtypes of breast cancers from Myanmar women : A study of 91 cases at two pathology centers. Asian Pac J Cancer Prev 2017;18:1617-21.
Sun J, Chen C, Wei WEN, Zheng H, Yuan J, Tu YI, et al
. Associations and indications of Ki67 expression with clinicopathological parameters and molecular subtypes in invasive breast cancer : A population-based study. Oncol Lett 2015;10:1741-8.
Coates AS, Winer EP, Goldhirsch A, Gelber RD, Gnant M. Tailoring therapies-improving the management of early breast cancer : St Gallen international expert consensus on the primary therapy of early breast cancer 2015 Ann Oncol 2015;8:1533-46.
Tashima R, Nishimura R, Osako T, Nishiyama Y, Okumura Y, Nakano M, et al
. Evaluation of an optimal cut-off point for the Ki-67 index as a prognostic factor in primary breast cancer: A retrospective study. PLoS One 2015;10:1-10.
Haroon S, Hashmi AA, Khurshid A, Kanpurwala MA, Mujtuba S, Malik B, et al
. Ki67 index in breast cancer: Correlation with other prognostic markers and potential in Pakistani patients. Asian Pac J Cancer Prev 2013;14:4353-8.
Hennigs A, Riedel F, Gondos A, Sinn P, Schirmacher P, Marmé F, et al
. Prognosis of breast cancer molecular subtypes in routine clinical care: A large prospective cohort study. BMC Cancer 2016;16:734.
Shemin KMZ, Smitha N, Jojo A, Vijaykumar D. Molecular classification and prognostication of 300 node-negative breast cancer cases: A tertiary care experience. South Asian J Cancer 2015;4:160-2.
Focke CM, Bü Rger H, Van Diest PJ, Finsterbusch K, Gläser D, Korsching E, et al
. Interlaboratory variability of Ki67 staining in breast cancer. Eur J Cancer 2017;84:219-27.
Shui R, Yu B, Bi R, Yang F, Yang W. An interobserver reproducibility analysis of Ki67 visual assessment in breast cancer. PLoS One 2015;10:1-10.
Chung YR, Jang MH, Park SY, Gong G, Jung W-H. Interobserver variability of Ki-67 measurement in breast cancer. J Pathol Transl Med 2016;50:129-37.
Yamamoto S, Chishima T, Mastubara Y, Adachi S, Harada F, Toda Y, et al
. Variability in measuring the Ki-67 labeling index in patients with breast cancer. Clin Breast Cancer 2015;15:35-9.
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.
Fre JJ, Poulet B, Clough KB, Crouet H, Fourquet A. Ki-67 : Level of evidence and methodological considerations for its role in the clinical management of breast cancer : Analytical and critical review. Breast Cancer Res Treat 2012;132:895-915.
Zhong F, Bi R, Yu B, Yang F, Yang W, Shui R. A comparison of visual assessment and automated digital image analysis of Ki67 labeling index in breast cancer. PLoS One 2016;11:1-11.
Mungle T, Tewary S, Arun I, Basak B, Agarwal S, Ahmed R, et al
. Automated characterization and counting of Ki-67 protein for breast cancer prognosis: A quantitative immunohistochemistry approach. Comput Methods Programs Biomed 2017;139:149-61.
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6]
[Table 1], [Table 2], [Table 3], [Table 4]