Accepted Manuscript
Comparison of the methods for measuring the Ki-67 labeling index in adrenocortical carcinoma: manual versus digital image analysis
Human PATHOLOGY
W: B. Saunders
Yuto Yamazaki MD, Yasuhiro Nakamura MD, PhD, Yukiko Shibahara MD, PhD, Sachiko Konosu-Fukaya MD, Naomi Sato MD, Fumie Kubota MD, Yutaka Oki MD, PhD, Satoshi Baba MD, PhD, Sanae Midorikawa MD, PhD, Ryo Morimoto MD, PhD, Fumitoshi Satoh MD, PhD, Hironobu Sasano MD, PhD
| PII: | S0046-8177(16)00072-1 |
| DOI: | doi: 10.1016/j.humpath.2015.10.017 |
| Reference: | YHUPA 3826 |
| To appear in: | Human Pathology |
| Received date: | 15 July 2015 |
| Revised date: | 23 October 2015 |
| Accepted date: | 29 October 2015 |
Please cite this article as: Yamazaki Yuto, Nakamura Yasuhiro, Shibahara Yukiko, Konosu-Fukaya Sachiko, Sato Naomi, Kubota Fumie, Oki Yutaka, Baba Satoshi, Mi- dorikawa Sanae, Morimoto Ryo, Satoh Fumitoshi, Sasano Hironobu, Comparison of the methods for measuring the Ki-67 labeling index in adrenocortical carcinoma: manual ver- sus digital image analysis, Human Pathology (2016), doi: 10.1016/j.humpath.2015.10.017
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Yamazaki et al
Comparison of the Methods for Measuring the Ki-67 Labeling Index in Adrenocortical
Carcinoma: Manual versus Digital Image Analysis
Yuto Yamazaki MD$, Yasuhiro Nakamura MD, PhDª, Yukiko Shibahara MD, PhDª, Sachiko Konosu-Fukaya MDb, Naomi Sato MDb, Fumie Kubota MD“, Yutaka Oki MD, PhDª, Satoshi Baba MD, PhDe, Sanae Midorikawa MD, PhDf, Ryo Morimoto MD, PhDª, Fumitoshi Satoh MD, PhD8, and Hironobu Sasano MD, PhDª
ªDepartment of Pathology, Tohoku University Graduate School of Medicine, Sendai, Japan bDivision of Pathology, Tohoku University Hospital, Sendai “Division of Pathology, Tohoku Rosai Hospital, Sendai
Department of Family and Community Medicine, Hamamatsu University School of Medicine, Hamamatsu, Shizuoka, Japan
·Department of Diagnostic Pathology, University Hospital Hamamatsu University School of Medicine, Hamamatsu, Japan
Department of Radiation Health Management, Fukushima Medical University, Fukushima, Japan
ªDivision of Nephrology, Endocrinology, and Vascular Medicine, Department of Medicine,
ACCEPTED MANUSCRIPT
Yamazaki et al
Tohoku University Graduate School of Medicine, Sendai [AU: Postal codes?]
* Correspondence and reprint requests should be addressed to:
Yasuhiro Nakamura, MD, PhD
Department of Pathology
Tohoku University School of Medicine 2-1 Seiryo-machi, Aoba-ku, Sendai 980-8575 JAPAN
E-mail: yasu-naka@patholo2.med.tohoku.ac.jp
A® um dos sier PT Latok MANUSCRIPT
Disclosure: This study was supported by a Scholarship from the Takeda Science Foundation.
Running title: Ki-67 in adrenocortical carcinoma
ACCEPTED MANUSCRIPT
Yamazaki et al
Summary
Adrenocortical carcinoma (ACC) is a rare, highly malignant neoplasm harboring marked histologic heterogeneity. The Ki-67 labeling index (LI) is one of the most effective diagnostic and prognostic markers in ACC. However, its assessment has by no means been standardized. Therefore, in this study, we analyzed the Ki-67 LI in 18 ACC cases both by seven pathologists using microscopes (MA; manual analysis) and with digital image analysis (DIA) and also compared the Ki-67 LI obtained by selecting “hot spots” and formulating the “average” reading LAVENDEL of the whole tumor specimen. In addition, we performed statistical analysis of the association between Ki-67 LI and the clinical and pathologic features of individual cases. The DIA was significantly correlated with MA in hot spots but not in the average fields. The Ki-67 LI in hot spots was significantly and consistently higher than that in average areas by both MA and DIA, indicating intratumoral heterogeneity. The Ki-67 LI was significantly correlated with the Weiss criteria (eosinophilic cytoplasm, nuclear atypia, atypical mitoses, and sinusoidal invasion) by any mode of evaluation. The clinical outcome was significantly better in the patients with a Ki-67 < 10% than in those with a Ki-67 > 10% by MA in hot spots. The Ki-67 LI in hot spots measured by MA best reflected the clinical and pathologic features of ACC. Employment of DIA to obtain the Ki-67 LI in ACC requires further improvement, including correction of its overestimation of the value by counting non-tumorous cells and nuclear segmentation in areas of high cell density.
ACCEPTED MANUSCRIPT
Yamazaki et al
Keywords:
Adrenocortical carcinoma;
Digital image analysis;
Immunohistochemistry;
Ki-67 labeling index;
Weiss criteria
Ne Ven CORTE MOLTO MANUSCRIPT
1. Introduction
Adrenocortical carcinoma (ACC) is a rare, highly malignant neoplasm. The annual case number is 0.5 to 2 cases per million, amounting to 0.05%-2% of all malignant neoplasms [1,2]. The mean survival duration is approximately 3 years after the initial diagnosis [3-6]. The average diameter of an ACC is more than 6.5 cm in greatest dimension. An ACC weighs more than 50 g on average and has marked intratumoral heterogeneity [7].
The Ki-67 labeling index (Ki-67 LI) is considered one of the most pivotal diagnostic markers in differential diagnosis of adrenocortical adenomas versus carcinomas, and its cut-off value has been reported to be around 2.5%-5%, although a Ki-67 LI of 5% as the cut-off value was reported to yield high sensitivity (87.5%) and specificity (97.5%) [8]. In addition, Ki-67 LI is pivotal not only in differentiating adrenocortical carcinoma from adenoma, but also in
ACCEPTED MANUSCRIPT
predicting the clinical outcome of the patients with ACC [9-11]. The European Network for the Study of Adrenal Tumors (ENSAT) proposed that non-metastatic cases of ACC after complete resection could be classified into two groups by setting the cut-off value of Ki-67 LI at 10% [1]. E The cases with Ki-67 LI > 10% were considered a high-risk group requiring adjuvant chemotherapy with mitotane [1], and those with Ki-67 LI < 10% as low- or intermediate-risk group who may require mitotane but not other chemotherapy [1]. Fassnacht et al reported that, based on data from German ACC, non-metastatic ACC cases with vascular invasion and Ki-67 LI > 10% are candidates for adjuvant chemotherapy with mitotane plus additional cytotoxic agents such as streptozotocin [1]. Therefore, Ki-67 LI has a pivotal role in selecting the adjuvant chemotherapy after complete resection of non-metastatic ACC. The association of Ki-67 LI with clinical outcomes of ACC also has been reported [9-11]. In particular, McNicol and colleagues [10] and Morimoto et al. [11] independently reported that the disease-free survival (DFS) of the patients was significantly shorter when the cut-off value of Ki-67 LI was 3% and 7%, respectively. In addition, Stojadinovic and associates reported no significant difference in DFS or overall survival (OS) by setting the cut-off value of Ki-67 LI at 5% [9].
However, to the best of our knowledge, no one has reported comparison of the Ki-67 LI values in ACC determined by different methods. Standardization of Ki-67 LI measurement is still controversial in all human malignancies because of its low reproducibility and interobserver
ACCEPTED MANUSCRIPT
Yamazaki et al
differences [12]. For instance, breast cancer is one of the tumors in which Ki-67 LI critically influences the clinical outcome [13,14]. Mikami et al reported the concordance of Ki-67 LI in breast cancer analyzed by six pathologists and concluded that there was better concordance when the assessed field was predetermined, indicating that the selection of the evaluation area is pivotal for obtaining reproducible results [15].
Recently, digital image analysis (DIA) using image analyzing software in breast cancer was reported to be more reproducible and reliable than subjective counts by pathologists [16]. Therefore, in this study, we examined different methods of Ki-67 LI assessment in ACC and compared the results in order to confirm the value of Ki-67 LI in the clinical management of patients with this cancer.
2. Materials and methods
2.1. Human adrenocortical carcinoma cases
The research protocols of this study were approved by the ethics committees of Tohoku University Graduate School of Medicine (Sendai, Japan), Fukushima Medical University (Fukushima, Japan), and Hamamatsu Medical University (Hamamatsu, Japan). We analyzed 18 ACC cases obtained from Tohoku University (n = 8), Fukushima Medical University (n = 4), and Hamamatsu Medical University (n = 6). Their clinicopathologic features are summarized in
ACCEPTED MANUSCRIPT
Yamazaki et al
Table 1. All the cases were confirmed to be ACC by application of the Weiss criteria after resection. The TNM class was determined according to the ENSAT classification of 2008 [17] (Supplementary Description).
AČound CA 30 Som Kan Kan Ka SCRIPT®
2.2. Immunohistochemical analysis
Hematoxylin and eosin-stained tissue sections were reviewed in all cases. After we reviewed all of the sections, we selected some for immunostaining in the greatest dimension of the tumor. The thickness of the sections was set at 3 um, and specimens were prepared in the Department of Pathology, Tohoku University School of Medicine. A mouse monoclonal antibody against Ki-67 (Dako, MIB-1 clone, dilution 1:100) was used. Antigen retrieval was performed by autoclaving the tissues at 121℃ for 5 min. We performed immunostaining manually and used sections of breast cancer as the control [15].
2.3. Analysis of Ki-67 labeling index
We analyzed Ki-67 LI in each tissue sections in both “hot spots” and average regions. We also performed both manual analysis (MA) and digital image analysis (DIA). Thus, four examinations (MA in hot and average spots; DIA in hot and average spots) were performed, and the results were compared with the clinicopathologic features of individual patients.
ACCEPTED MANUSCRIPT
2.4. Manual analysis
Seven pathologists in the two centers employing the same evaluation criteria independently assigned the Ki-67 LI to each section using standard light microscopy of both hot and average spots. They selected hot spots by estimating, through scanning of the whole immunostained slide preparation, which areas had Ki-67-positive cells that were the most
frequent or dense. These seven pathologists counted more than 500-1000 cells in one high-power field they had selected as the hot spot in each case. They also performed average field analysis, counting more than 1000 cells in 5-10 high-power fields chosen at random under light microscopy.
2.5. Digital image analysis
Digital image analysis was performed using HALO Cytonuclear ver. 1.4 (Leica, Buffalo
Grove, IL USA). All immunostained tissue sections were scanned by an Image Scope AT2 (Leica). The observer set the thresholds of the following parameters in each section (Supplementary Table): nuclear color, contrast, size, roundness, and segmentation. The pathologist first selected both positive nuclei (brown; stained by DAB) and negative nuclei (blue; stained by hematoxylin) and classified the results into the two groups. The pathologist then
ACCEPTED MANUSCRIPT
adjusted the thresholds of nuclear contrast to identify nuclei lightly stained by hematoxylin. The threshold of nuclear size was subsequently changed in an attempt to exclude lymphocytes and stromal cells, whose nuclei are smaller than those of tumor cells. Nuclear roundness also represents an important feature, as it can be used for putative exclusion of the spindle-shaped stromal cells. Finally, nuclear segmentation was altered to separate proximity nuclei in high cell-density areas. The DIA was performed by a single observer both in hot spots and in average areas. The hot spot was selected by the same pathologist and analyzed in one high-power field, counting more than 500 cells. In average areas, tumor areas of the whole sections were analyzed by DIA. All these parameters were set individually in each case. In the selected annotation area, the HALO software automatically calculated the number of positive cells among the total cells. Image examples of DIA are illustrated in Fig. 1.
Ma EDE,
2.6. Statistical analysis
Correlation of Ki-67 LI among methods was analyzed by Spearman’s technique (Fig. 2). The comparison of Ki-67 LI in hot spot and average areas was performed by the Wilcoxon t test (Fig. 3). Correlations of Ki-67 LI among methods with clinicopathologic factors were also analyzed by Spearman’s technique (Figs. 4 and 5). Survival periods were defined as the duration from the day of surgery to that of demise or the last day of disease-free status confirmed with
ACCEPTED MANUSCRIPT
censor code in Kaplan-Meier analysis (Fig. 6). Significant difference was set at P < . 05.
3. Results
3.1. Correlation among four methods
The Ki-67 LI values in hot spot and average areas are summarized in Table 2. Significant correlation was detected among the results based on the seven pathologists’ semiquantitative evaluation in both hot spots (all pairs P ≤.0001) and average areas (all pairs P ≤.0005). However, there were interobserver differences. The MA was also significantly correlated with DIA in hot spots, but MA performed by pathologists 02-05 demonstrated no correlation with DIA in average fields (02 P =0746; 03 P = . 1736; O4 P = . 0794; 05 P = . 0746; see Fig. 2).
3.2. Hot spot versus average
The average Ki-67 LI in hot spots by MA and DIA were 23.3% and 18.2%, respectively, and the mean values in the average sites by MA and DIA were 10.1% and 6.1%, respectively. As illustrated in Fig. 3 above, the Ki-67 LI in hot spots was significantly higher than that in average fields by both MA and DIA (MA 10.1% vs. 23.3%, P = . 0030; DIA 6.0% vs. 18.2%, P = . 0025).
3.3. Association with clinicopathologic features of the cases
ACCEPTED MANUSCRIPT
The Ki-67 LI in hot spots was significantly correlated with the Weiss criteria by both MA and DIA (Fig. 5). However, Ki-67 LI was not significantly correlated with tumor size by any method (Fig. 4). When the Ki-67 LI cut-off value was set at 10% according to the ENSAT proposal, the results tended to be correlated with survival times of the ACC patients in both hot spot and average values by both MA and DIA. The patients with Ki-67 > 10% had a significantly poorer clinical outcome than those with Ki-67 < 10% by MA in hot spots by the Kaplan-Meier log-rank test (P = . 0179; Fig. 6). We tested other cut-off values of Ki-67 LI in hot spots and average regions by MA and DIA as 5% (MA hot spot P = . 17; MA average P = . 019; DIA hot spot P = . 81; DIA average P = . 53), 15% (MA hot spot P = . 019; MA average P = . 32; DIA hot spot P = . 82; DIA average P =. 996), and 20% (MA hot spot P = . 05; DIA hot spot P = . 35; MA average P = . 48; DIA average P = . 36).
HELEan
4. Discussion
4.1. Ki-67 LI in ACC
This is the first study comparing the different methods of Ki-67 LI analysis in ACC. This cancer has been known to harbor marked intratumoral heterogeneity [7]. Distribution patterns of Ki-67-positive cells varied enormously in our study: diffuse in the whole tumor areas, sporadic at random sites, and intensive in the marginal zone of tumors. However, hot spots tend to be
ACCEPTED MANUSCRIPT
Yamazaki et al
present in the marginal zones of tumors, indicating that tumor growth and cell proliferation could be active in these areas. In our cases with coagulation necrosis, Ki-67-positive cells were frequently detected around the necrotic area, which also suggested that high cell proliferation resulted in confluent necrosis with disturbance of the balance of tumor cell proliferation and cells tended to be localized in the marginal zones.
E
Tumor size also has been reported as one of the pivotal diagnostic factors to differentiate ACC from adenomas, but it is also true that 8%-13.5% of ACCs were <5 cm in greatest dimension. Therefore, tumor size is considered an important but by no means perfect diagnostic marker [16]. In addition, the tumor size was reported to be associated not only with the differential diagnosis, but also with clinical outcomes [18]. According to previously reported investigations, tumor size >12.5 cm in greatest dimension was significantly associated with poor survival after complete resection [18]. As illustrated in Fig. 4 above, the Ki-67 LI was not significantly correlated with tumor size, probably because of the small number of cases examined. Hormonal function also was not significantly correlated with Ki-67 LI in our study. However, the cases with high serum concentrations of dehydroepiandosterone (DHEAS) tended to have high Ki-67 LI in hot spots (MA P = . 108; DIA P = . 244).
ACCEPTED MANUSCRIPT
4.2. Ki-67 LI analysis by light microscopy
A significant correlation was detected among MA by conventional light microscopy performed by seven pathologists, both in hot spots and in average sites. As reported previously, VELIKONOCE E interobserver differences are unavoidable in manual analysis [15]. The best method to achieve the maximum interobserver concordance was said to be assessment of pre-chosen areas [12,15]. However, in practical settings, the area to be assessed is determined by each pathologist when evaluating the cases. Therefore, in this study, hot spots for MA were selected by each pathologist simulating the routine diagnostic settings, and even in this situation, significant correlation of Ki-67 LI in hot spots by MA was detected among seven pathologists (see Fig. 2). However, interobserver difference became much larger in the cases with high Ki-67 LI (Table 2). The threshold of immunopositivity also depends on the pathologists, because the results of Ki-67 LI tended to be low for one pathologist and high for another. The most influenced parameters for MA turned out to be “nuclear color” and “nuclear contrast.” While counting the cells,
pathologists could set the parameters simultaneously, recognizing tumor cells and differentiating them from non-tumor cells. These processes could become the same for well-trained pathologists without their being aware of it. However, the threshold “nuclear color” and “nuclear contrast” can be subjective. In MA, we employed the same evaluation methods used in routine clinical practice, but the precise thresholds of “weak” staining were dependent on each pathologist. In
ACCEPTED MANUSCRIPT
Yamazaki et al
DIA, when the brown colorimetric reaction of DAB was identified, the cells were counted as positive.
De Santa Mon aka la
In the average fields examined by MA, interobserver differences in Ki-67 LI became less, possibly because of the smaller size of the evaluated areas, and interobserver difference was thought to be less influenced by field selection. In our present study, the Ki-67 LI in hot spots was significantly higher than that in average fields, also indicating the marked intratumoral heterogeneity of ACC. Hot spots tended to be selected in marginal zones of the tumors, as described above. The Ki-67 LI was significantly correlated with the Weiss criteria in any measurement performed. Therefore, Ki-67 LI reflects the morphologic tumor features, including histologic atypia, mitotic activation, and invasive behavior. In this study, when setting the cut-off value of Ki-67 LI as 10% according to the ENSAT proposal, the patients with a Ki-67 LI of < 10% had a significantly better prognosis than those with a Ki-67 LI > 10% performed by MA in hot spots but not in average fields. These results could justify the threshold of Ki-67 LI of 10% despite the small number of cases examined in our present study (Fig. 6). In addition, the cut-off value of Ki-67 LI could be reliable as 15% in hot spots and 5% in average fields by MA. The Ki-67 LI in hot spots may therefore be considered the most important determinant of the clinical status of ACC patients.
ACCEPTED MANUSCRIPT
4.3. MA versus DIA
An MA analysis is simple, straightforward, and easy; but the interobserver differences can be marked [15], with low reproducibility and time-cost performance because of the a labor-intensive analysis. In contrast, DIA demonstrated high reproducibility because of the preservation of the parameters selected. In addition, the parameters can be applied in wide areas, which could result in high time-cost performance analysis. However, it is also true that the parameters are rather complicated, and it takes considerable time to set them. The DIA result is reported to be concordant with MA in several neoplasms such as neuroendocrine tumor (NET) and breast cancer [19,20]. The Ki-67 LI values in hot spots generally are regarded as standard practice in these tumors, counting more than 500-2000 cells in NET (by the ENETS/WHO proposal) [20] and more than 1000 cells in breast cancer [21]. In our study, the Ki-67 LI was indeed significantly correlated for MA and DIA in hot spots, but not in average areas. The DIA method overestimated the number of non-tumor or non-neoplastic cells (stromal, inflammatory, etc) regardless of the parameter settings.
A DIA analysis requires extremely complicated pattern recognition designed for machinery or instrumental parameters so that observers exclude non-tumorous cells with their own eyes, at least at this juncture. However, it is also true that pathologists are setting parameters simultaneously during their routine practice. The hot spot represents a rather narrow area, and
ACCEPTED MANUSCRIPT
Yamazaki et al
therefore, no discrepancy was noted between MA and DIA in these sites; but when analyzing a wider area, the discrepancy between these two modes became more pronounced, which could be influenced by overcounting of non-tumorous cells or undercounting proximity nuclei in areas of high cell density. According to a previous comparative analysis of pancreatic NETs, eyeball estimation yielded highly inaccurate and unreliable results and was not recommended for routine use [22]. Reid et al reported that DIA was not as accurate or cost-effective as camera-captured images with manual counting. Those investigators concluded that the latter was the most reliable and reproducible method [22]. In our study, selection of hot spots depended on each observer, in SIE accordance with routine practice. Despite the relatively small number of cases available for the study, MA in hot spots reflected clinicopathologic status in ACC. Further improvement, including more complicated histologic pattern recognition and proximity cell contour in high cell-density area, should be required in DIA.
5. Conclusion
A Ki-67 LI performed by MA or a pathologist in hot spots is considered the most valuable measurement in ACC harboring marked intratumoral heterogeneity, because it is reflective of outcome of clinicopathologic status and its simplicity. However, interobserver difference could be marked in hot spots, depending on the selection of the field and subjective
ACCEPTED MANUSCRIPT
Yamazaki et al
thresholds of nuclear immunointensity. A DIA can be utilized for analysis of Ki-67 LI in hot spots in terms of its high reproducibility and high time-cost performance, while it should be limited to the cases in which non-neoplastic cells are difficult to differentiate from neoplastic cells in ACC. In routine pathology practice, hot spots analyzed by surgical pathologists is the most reliable method in ACC because of the small number of cases. A DIA is useful for tumors of other organs, including NETs or breast cancer. In order to use DIA for practical diagnosis, all of the observers should be well trained in setting the complicated parameters based on pathologist views. Although the value of Ki-67 LI might be influenced by other unevaluable factors, including section thickness, dyeing time, and state of preservation, standardization of Ki-67 LI measurement is necessary to make a pathologic diagnosis.
References
[1] Fassnacht M, Kroiss M, Allolio B. Update in adrenocortical carcinoma. J Clin Endocrinol Metab 2013;98:4551-64.
[2] Allolio B, Fassnacht M. Clinical review: adrenocortical carcinoma: clinical update. J Clin Endocrinol Metab 2006;91:2027-37.
[3] Volante M, Buttigliero C, Greco E, Berruti A, Papotti M. Pathological and molecular features of adrenocortical carcinoma: an update. J Clin Pathol 2008;61:787-93.
ACCEPTED MANUSCRIPT
Yamazaki et al
[4] Paton BL, Novitsky YW, Zerey M, et al. Outcomes of adrenal cortical carcinoma in the United States. Surgery 2006; 140:914-20.
[5] Fulmer BR. Diagnosis and management of adrenal cortical carcinoma. Curr Urol Rep 2007;8:77-82.
[6] Kirschner LS. Emerging treatment strategies for adrenocortical carcinoma: a new hope. J Clin Endocrinol Metab 2006;91:14-21.
[7] Aubert S, Wacrenier A, Leroy X, et al. Weiss system revisited: a clinicopathologic and immunohistochemical study of 49 adrenocortical tumors. Am J Surg Pathol 2002;26:1612-9.
[8] Schmitt A, Saremaslani P, Schmid S, et al. IGFII and MIB1 immunohistochemistry is helpful for the differentiation of benign from malignant adrenocortical tumours. Histopathology 2006;49:298-307.
[9] Stojadinovic A, Ghossein RA, Hoos A, et al. Adrenocortical carcinoma: clinical, morphologic, and molecular characterization. J Clin Oncol 2002;20:941-50.
[10] McNicol AM, Struthers AL, Nolan CE, Hermans J, Haak HR. Proliferation in adrenocortical tumors: correlation with clinical outcome and p53 status. Endocr Pathol 1997;8:29-36.
[11] Morimoto R, Satoh F, Murakami O, et al. Immunohistochemistry of a proliferation
ACCEPTED MANUSCRIPT
Yamazaki et al
marker Ki-67/MIB1 in adrenocortical carcinomas: Ki-67/MIB1 labeling index is a predictor for recurrence of adrenocortical carcinomas. Endocrine J 2008;55:49-55.
[12] Polley M-YC, Leung SCY, Gao D et al, An international study to increase concordance in Ki-67 scoring. Mod Pathol 2015;28:778-86.
13] Yerushalmi R, Woods R, Ravdin PM, Hayes MM, Gelmon KA. Ki-67 in breast cancer: prognostic and predictive potential. Lancet Oncol 2010;11:174-83.
[14] Sheri A, Dowsett M. Developments in Ki-67 and other biomarkers for treatment decision making in breast cancer. Ann Oncol 2012;23;219-27.
HOLA JEVALENTINE
[15] Mikami Y, Ueno T, Yoshimura K, et al. Interobserver concordance of Ki-67 labeling index in breast cancer: Japan Breast Cancer Research Group Ki-67 Ring Study. Cancer Sci 2013;104:1539-43.
[16] Barnett CC, Varma DG, El-Naggar AK, et al. Limitations of size as a criterion in the evaluation of adrenal tumors. Surgery 2000;128:973-83.
[17] Fassnacht M, Johanssen S, Quinkler M, et al, for the German Adrenocortical Carcinoma Registry Group and the European Network for the Study of Adrenal Tumors. Limited prognostic value of the 2004 International Union against Cancer Staging Classification for adrenocortical carcinoma: proposal for a revised TNM classification. Cancer 2009;115:243-50.
ACCEPTED MANUSCRIPT
Yamazaki et al
[18] Stojadinovic A, Ghossein RA, Hoos A, et al. Adrenocortical carcinoma: clinical, morphologic, and molecular characterization. J Clin Oncol 2002;20:941-50.
[19] Gudlaugsson E, Skaland I, Janssen EAM, et al. Comparison of the effect of different techniques for measurement of Ki-67 proliferation on reproducibility and prognosis prediction accuracy in breast cancer. Histopathology 2012;61:1134-44.
[20] Tang LH, Gonen M, Hedvat C, Modlin IM, Klimstra DS. Objective quantification of the Ki-67 proliferative index in neuroendocrine tumors of the gastroenteropancreatic system: a comparison of digital image analysis with manual methods. Am J Surg Pathol 2012;36:1761-70.
[21] Jones RL, Salter J, A’hern R, et al. Relationship between oestrogen receptor status and proliferation in predicting response and long-term outcome to neoadjuvant chemotherapy for breast cancer. Breast Cancer Res Treat 2010;119:315-23.
[22] Reid MD, Bagci P, Ohike N, et al. Calculation of the Ki-67 index in pancreatic neuroendocrine tumors: a comparative analysis of four counting methodologies. Mod Pathol 2015;2:686-94.
LANUEVA EINER
ACCEPTED MANUSCRIPT
Fig. 1 Image examples of before (A) and after (B) DIA. Cells with blue nuclei are negative, whereas cells with red nuclei are positive for Ki-67 immunoreactivity. Arrows mark endothelial cells of intratumoral vessels.
Fig. 2 A, Correlation of Ki-67 LI in hot spot performed by seven pathologists (01-07). “Hot spot” was selected by individual evaluation. There was a significant correlation among the values of Ki-67 LI counted by the seven pathologists. All pairs ** P ≤ .0001. B, Correlation of Ki-67 LI in hot spots between MA (performed by seven pathologists) and DIA. There was a significant correlation of Ki-67 LI by DIA and by MA (all of the seven pathologists). All pairs ** P ≤ .0004. C, Correlation of Ki-67 LI in average fields performed by seven pathologists (01-07). There was significant correlation among the values. All pairs ** P ≤ .0005. D, Correlation of Ki-67 LI in average fields by MA and DIA. There was a significant correlation of Ki-67 LI between DIA and MA performed by 01, 06, and 07 (01 P = . 0294; 06 P = . 0151; 07 P = . 0151). However, MA performed by 02-05 showed no significant correlation with DIA (O2 P = . 0746; O3 P = .1736; O4 P = .0794, O5 P = .0746). *P < .05. O: Observer.
Fig. 3 A, Comparison of Ki-67 LI in hot spot and average fields by MA. The value in hot spots was significantly higher than that in average fields by MA (23.3% vs. 10.1%; P = . 003). Ki-67 LI
ACCEPTED MANUSCRIPT
Yamazaki et al
value was characterized as the average value among 01-07. B, Comparison of Ki-67 LI in hot spots and average fields by DIA. The values in hot spots were significantly higher by DIA than in average fields (18.2% vs. 6.0%; P = . 003).
Fig. 4 The Ki-67 LI was not significantly correlated with tumor size by any measurement performed. A, MA in hot spots. B, DIA in hot spots. C, MA in average fields. D, DIA in average fields. The Ki-67 LI value was characterized as the average value assigned by 01-07.
Fig. 5 The Ki-67 LI correlated significantly with the score of the Weiss criteria by all measurements. A, MA in hot spots, P = . 0033. B, DIA in hot spots, P = . 0009. C, MA in average fields, P = . 0036. D, DIA in average fields, P = . 0283.
Fig. 6 On the basis of the ENSAT criteria, patients were divided into two groups: Ki-67 LI <10% and >10% by all measurements performed. A, MA in hot spots, P = . 02. B, DIA in hot spots, P = . 07. C, MA in average fields, P = . 05. D, DIA in average fields, P = 96.
ACCEPTED MANUSCRIPT
Yamazaki et al
A
100um
ACCEPTEL
B
100um
: Negative nuclei
: Positive nuclei
ACCEPTED MANUSCRIPT
Yamazaki et al
Fig. 2
A
All pairs; P<0.01 **
C
All pairs; P<0.01 **
01
50
01
01
30
20
01
20
10
0
O2 50
02
02
30
20
02
20
10
·
0
O3 50
03
03
30
03
20
Ki67 LI (%)
Ki67 LI (%)
0
G
O4 50
04
04
30
04
20
0
O5 50
05
05
30
05
20
0
06 50
05
06
30
06
20
10
0
O7 50
07
07
30 888 O
07
20
:
20 50
20 50
20
50
20
50
20 50
20 50
20 50
0
01 02 03 04
05
06
07
D
20 40 0
20
40
0
20
40
0
20
40
0
20
40
D
20 40
0
20
40
01
02
03
04
05
06
07
Ki67 LI (%)
B
D
Ki67 LI (%)
*
DIA
Ki67 LI (%)
50
Ki67 LI
DIA (%)
30
2
0
0
20
40
0
20
40
0
20
40
0
20
40
0
20
40
0
20
0 40
0
20
40
20
50
20 50
20
50
20 50
20
50
20 50
20 50
01
02
03
04
05
06
07
01
02
03
04
05
06
07
MA
MA
Ki67 LI
ACC
Ki67 LI (%)
ACCEPTED MANUSCRIPT
Yamazaki et al
A
80
70
*
60
23.3%
50
Ki67 LI (%)
40
·
30
10.1%
20
10
0
Average
Hot spots
ACCEPT
B
80
60
Ki67 LI (%)
40
18.2%
6.0%
20
0
Average
Hot spots
ACCEPTED MANUSCRIPT
Yamazaki et al
A
B
17.5
17.5
15
15
Tumor size (cm)
12.5
10
Tumor size (cm)
12.5
10
7.5
7.5
5
5
0
0
0
20
40
60
0
20 40 60
Ki67 LI (%)
Ki67LI (%)
C
D
17.5
17.5
15
15
Tumor size (cm)
12.5
10
Tumor size (cm)
12.5
10
7.5
7.5
5
5
0
0
0
10 20 30 40 Ki67 LI (%)
ACCE
0
10
3.0
50
Ki67 LT(%)
Figure 4
ACCEPTED MANUSCRIPT
Yamazaki et al
Fig 5.
A
B
8
… .......
8
7
7
6
6
Weiss criteria
5
Weiss criteria
5
4
4
3
3
2
2
0
0
0
20 40 60 Ki67 LI (%)
0
20
40
60
Ki67 LI (%)
C
D
8
8
7
7
Weiss criteria
6
Weiss criteria
6
5
5
4
4
3
3
2
2
0
0
0 10
30
ACCE
50
0 10
30
50
Ki67 LI (%)
Ki67 LI (%)
ACCEPTED MANUSCRIPT
Yamazaki et al
Fig 6.
A
B
1.0
1.0
0.8
*
0.8
0.6
-: Ki67<10%(n=5)
0.6
Surviva
-: Ki67>10% (n=13)
Surviva
:Ki67<10%(n=7)
I rate
₹ 0.4
I rate
0.4
-: Ki67>10% (n=11)
0.2
0.2-
0.0
0.0
0
1000
2000
3000
4000
5000
0
1000
2000
3000
4000
5000
Survival periods (days)
Survival periods (days)
C
D
1.0
1.0
0.8
0.8
0.6
0.6
Ki67<10%(n=14)
Surviva
I rate
Ki67<10%(n=9)
Ki67>10%(n=4)
0.4
Ki67>10% (n=9)
Surviva I rate
20.4
0.2
0.2
0.0
0.0
0
1000
2000
3000
4000
5000
0
1000
2000
3000
4000
5000
Survival periods (days)
ACC
Survival periods (days)
| C as e | A g e | S e | Late ralit | Tu mo | TNM , | Hormo ne | F | A CT | U F C | DH EA- | Adjuva nt chemot herapy | W eis | Sur viv | Out com | Ce ns or co de |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| x | y | r siz e | stage | functio | H | S | s | al | e | ||||||
| ning | SC or | (d) | |||||||||||||
| (c m) | e | ||||||||||||||
| 1 | 1 | F | R | 10. | TXN | F/A | 21 | <4. | Not | 8 | 520 6 | Aliv e | 1 | ||
| 1 | 5 | 0M0 | .7 2 | 0 | availab le | ||||||||||
| 2 | 6 | F | L | 11 | T2N | 30 | 17 | Not | 7 | 709 | Aliv e | 1 | |||
| 2 | 1M0, stage III | .2 | 8 | availab le | |||||||||||
| 3 | 7 | M | L | 11 | T2N | 15 | <2. | Mitota ne | 7 | 161 5 | Aliv e | 1 | |||
| 2 | OMO, stage II | .2 | 0 | ||||||||||||
| 4 | 5 | M | L | 4.3 | TXN OMO | Non-fu | 19 | 35. | Mitota ne | 6 | 142 0 | Aliv e | 1 | ||
| 6 | nctioni ng | .8 | 7 | ||||||||||||
| 5 | 0 | F | R | 4.6 | TXN | 3. | Null | 8 | 498 6 | Aliv e | 1 | ||||
| 1M0 | 86 | ||||||||||||||
| 6 | 5 4 | F | R | 9.5 | TXN 1M1 | 25 | <5. | EDP + weekly TXL | 8 | 165 | Dea d | 0 | |||
| .9 8 | 0 | ||||||||||||||
| 7 | 7 | M | R | 5.4 | T2N | F/E | 21 | <5. | Mitota | 7 | 882 | Dea d | 0 | ||
| 7 | 0M0, stage II | .6 | 0 | ne | |||||||||||
| 8 | 3 7 | F | L | 9 | T3/4 N0M 0, | F/A | 29 | <2. | 400 0 | Not availab le | 8 | 140 | Aliv e | 1 | |
| .2 | 0 |
ACCEPTED MANUSCRIPT
Yamazaki et al
| stage III | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 9 | 7 | F | R | 8.5 | T3N 0M0, stage III | Non-fu nctioni ng | 18 .9 | 41. 3 | 1 2 1 | 52 | Mitota PT ne | 8 | 156 4 | Aliv e | 1 |
| 5 | |||||||||||||||
| 1 0 | 2 | M | L | 4.5 | T1N 0M0, stage I | Non-fu | 6 | 35 | 1 0. 4 | 71 | Mitota ne | 5 | 139 0 | Aliv e | 1 |
| 9 | nctioni ng | ||||||||||||||
| 1 1 | 4 | F | L | 3.1 | T1N 1M0, stage I | F | 18 .6 | 2 | 2 3 8 | 14 | Mitota ne | 5 | 114 0 | Dea d | 0 |
| 5 | |||||||||||||||
| 1 | 4 | F | L | 5.5 | T3N 1M0, stage III | F | 23 .5 | 2 | 6 1 0 | 16 | Mitot ane, EDP | 8 | 891 | Dea d | 0 |
| 2 | 9 | ||||||||||||||
| 1 3 | 5 | F | L | 4 | TIN | F | 22 | <5 | 1 | Mitota | 6 | 957 | Dea d | 0 | |
| 0 | OMO, stage I | .2 | 6 9 0 | ne, E+CB DCA | |||||||||||
| 1 4 | 2 | F | L | 15 | T4N | F/A/E | 40 | <5 | 1 | 192 0 | EP/Mit | 8 | 199 | Dea d | 0 |
| 6 | 1M1, stage IV | .3 | 2 1 0 | otane, RFA | |||||||||||
| 1 5 | 4 | M | R | 16 | T4N | A/E | 14 .6 | 10. 6 | 1 1. 1 | 878 | Mitota ne, (EDP/ M) | 8 | 140 | Dea d | 0 |
| 8 | 1M1, stage IV | ||||||||||||||
| 1 | 5 | M | R | 13 | T4N | F/A/E | 33 | <5 | 8 | 527 | EDP/ | 8 | 119 | Dea d | 0 |
| 6 | 4 | 1M1, stage | .9 | 3 1 | Mitota ne, | 7 |
ACCEPTED MANUSCRIPT
Yamazaki et al
| IV | RFA, TAE | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 7 | 6 1 | FR | 8 8.5 | T4N 1M0, stage III | F/A | 24 .6 | <2 | 4 7 9 | 291 | Mitota ne, radiati on | 8 | 486 | Aliv e | 1 | |
| 1 8 | 5 | FL | 3.8 | T1N 0M0, stage I | A | 1. 6 | <2 | 131 | Mitot | 8 | 179 | Aliv e | 1 | ||
| 9 | |||||||||||||||
| ane |
Abbreviations: F, cortisol; A, androgen; E, estrogen; EDP; TXL; CBDCA; EP; RFA; M; TAE.
ACCEPTED MA
ACCEPTED MANUSCRIPT
| Case | O1 | O2 | O3 | O4 | O5 | O6 | O7 | MA | DIA | |
|---|---|---|---|---|---|---|---|---|---|---|
| (01-07) | ||||||||||
| 1 | Hot spot | 4.5 | 3.5 | 1.0 | 4.2 | 4.6 | 5.1 | 4.3 | 3.9 | 8.7 |
| Average | 1.0 | 1.3 | 0.9 | 0.9 | 1.5 | 0.9 | 1.8 | 1.2 | 5.3 | |
| 2 | Hot spot | 7.4 | 16.7 | 5.9 | 7.8 | 7.5 | 6.1 | 7.9 | 8.5 | 7.7 |
| Average | 1.6 | 4.0 | 5.7 | 3.9 | 1.1 | 5.0 | 4.1 | 3.6 | 2.8 | |
| 3 | Hot spot | 5.8 | 14.0 | 6.0 | 3.2 | 5.9 | 5.2 | 6.2 | 6.6 | 6.8 |
| Average | 1.6 | 2.8 | 2.4 | 1.4 | 3.3 | 2.4 | 2.8 | 2.4 | 4.8 | |
| 4 | Hot spot | 2.3 | 2.0 | 1.1 | 1.4 | 3.9 | 1.7 | 1.8 | 2.0 | 1.6 |
| Average | 0.9 | 1.8 | 1.3 | 0.7 | 0.9 | 0.8 | 1.2 | 1.1 | 0.6 | |
| 5 | Hot spot | 41.7 | 42.7 | 60.4 | 51.3 | 58.4 | 26.8 | 53.2 | 47.8 | 45.1 |
| Average | 22.2 | 33.3 | 33.2 | 23.7 | 18.3 | 21.8 | 20.1 | 24.6 | 20.6 | |
| 6 | Hot spot | 51.1 | 37.6 | 28.1 | 51.7 | 76.1 | 55.1 | 47.7 | 49.6 | 78.9 |
| Average | 10.6 | 17.3 | 15.4 | 14.2 | 32.3 | 21.5 | 9.5 | 17.3 | 19.1 | |
| 7 | Hot spot | 38.6 | 40.9 | 37.7 | 16.1 | 26.1 | 25.8 | 21.8 | 29.6 | 28.1 |
| Average | 11.8 | 9.8 | 8.1 | 2.6 | 5.5 | 11.3 | 12.5 | 8.8 | 8.2 | |
| 8 | Hot spot | 36.1 | 41.0 | 25.8 | 11.8 | 14.9 | 31.3 | 12.5 | 24.8 | 18.1 |
| Average | 17.1 | 16.5 | 15.0 | 3.3 | 2.1 | 7.5 | 8.2 | 10.0 | 10.7 | |
| 9 | Hot spot | 8.5 | 27.5 | 21.8 | 11.8 | 11.2 | 10.6 | 9.4 | 14.4 | 16.8 |
| Average | 1.1 | 5.1 | 4.7 | 2.7 | 3.4 | 4.9 | 0.1 | 3.1 | 5.4 | |
| 10 | Hot spot | 7.8 | 12.8 | 6.5 | 2.1 | 9.5 | 3.5 | 4.3 | 6.6 | 6.2 |
| Average | 1.8 | 2.9 | 1.8 | 1.3 | 1.8 | 1.6 | 2.4 | 1.9 | 0.9 | |
| 11 | Hot spot | 11.6 | 26.8 | 7.9 | 7.8 | 15.6 | 8.8 | 8.9 | 12.5 | 7.6 |
| Average | 2.5 | 4.6 | 6.2 | 3.1 | 4.1 | 4.1 | 1.3 | 3.7 | 1.4 | |
| 12 | Hot spot | 24.1 | 31.4 | 23.3 | 11.4 | 14.3 | 11.7 | 14.4 | 18.6 | 10.4 |
| Average | 11.7 | 12.7 | 9.3 | 9.5 | 10.3 | 7.9 | 8.6 | 10.0 | 2.5 | |
| 13 | Hot spot | 9.5 | 28.4 | 20.8 | 5.9 | 8.9 | 10.6 | 8.8 | 13.3 | 4.3 |
| Average | 1.1 | 9.2 | 2.5 | 1.5 | 1.9 | 2.0 | 1.7 | 2.8 | 0.2 | |
| 14 | Hot spot | 22.6 | 40.2 | 30.8 | 26.0 | 32.1 | 29.9 | 27.7 | 29.9 | 10.9 |
| Average | 17.0 | 21.8 | 18.0 | 10.5 | 13.3 | 18.0 | 4.1 | 14.7 | 2.3 | |
| 15 | Hot spot | 25.8 | 53.2 | 42.8 | 29.0 | 49.5 | 32.8 | 33.4 | 38.1 | 13.3 |
| Average | 10.2 | 24.4 | 34.3 | 19.3 | 22.1 | 24.0 | 12.9 | 21.0 | 3.4 |
ACCEPTED MANUSCRIPT
Yamazaki et al
| 16 | Hot spot | 26.3 | 38.7 | 39.9 | 34.2 | 32.1 | 40.3 | 34.5 | 35.1 | 13.3 |
| Average | 3.2 | 15.0 | 19.2 | 7.6 | 9.8 | 8.9 | 6.5 | 10.0 | 0.2 | |
| 17 | Hot spot | 25.8 | 27.1 | 37.9 | 29.2 | 29.9 | 39.5 | 46.1 | 33.6 | 16.4 |
| Average | 11.7 | 15.6 | 16.4 | 12.9 | 8.4 | 3.7 | 8.9 | 11.1 | 1.9 | |
| 18 | Hot spot | 35.1 | 64.1 | 40.0 | 39.8 | 50.2 | 45.6 | 39.4 | 44.9 | 33.4 |
| Average | 19.0 | 34.5 | 38.8 | 38.9 | 43.0 | 30.7 | 31.7 | 33.8 | 18.0 |
NOTE. MA (01-07) is the average value of Ki-67 labeling index by manual analysis by seven pathologists.
Abbreviations: DIA, digital image analysis; MA, manual analysis; O, observer.