Ki-67 labeling index can be used as a prognostic marker in gastrointestinal stromal tumor: a systematic review and meta-analysis
Int J Biol Markers 2016; 31(2): e204 - e210
Article Type: ORIGINAL RESEARCH ARTICLE
AuthorsJung-Soo Pyo, Guhyun Kang, Jin Hee Sohn
The aim of this study was to investigate the prognostic relevance of the Ki-67 labeling index (LI) in gastrointestinal stromal tumor (GIST) through a systematic review, meta-analysis and diagnostic test accuracy review.
The study included 1,967 GIST cases from 24 eligible studies. We investigated the correlation between high Ki-67 LI and survival and the proper criteria for high Ki-67 LI. In addition, a diagnostic test accuracy review was conducted to evaluate the predictive role of high Ki-67 LI for higher risk of tumor recurrence.
A high Ki-67 LI was significantly correlated with worse disease-free survival (DFS) (hazard ratio [HR] 3.658, 95% confidence interval [CI] 2.687-4.979, p<0.001) and overall survival (OS) (HR 3.730, 95% CI 2.819-4.936, p<0.001). With regard to DFS and OS, the subgroup with a cutoff value of >4% for high Ki-67 LI had a higher HR than the subgroup with a ≤4% cutoff. In the diagnostic test accuracy review, a high Ki-67 LI was significantly correlated with higher risk of tumor recurrence (pooled sensitivity = 0.44, pooled specificity = 0.87, area under the curve on the summary receiver operating characteristic curve = 0.656).
Our results showed that a high Ki-67 LI was significantly correlated with worse prognosis and higher risk of tumor recurrence in GIST. Further prospective studies of the prognostic role of Ki-67 LI are necessary prior to application in clinical practice.
- • Received on 07/09/2015
- • Accepted on 26/10/2015
- • Available online on 27/11/2015
- • Published online on 28/05/2016
This article is available as full text PDF.
Most gastrointestinal stromal tumors (GISTs) have at least some potential for malignancy. The AJCC staging system uses tumor size and mitotic rate as the staging parameters for these tumors (1). While the maximum tumor diameter is easy to determine, there is a possible bias in evaluating the number of mitoses (2). Uncontrolled proliferation is one of the chief characteristics of tumor cells and can be assessed using various methods, including counting mitoses under a microscope. However, stringent criteria are needed for defining mitosis in order to avoid counting pyknotic or dyskaryotic nuclei as mitoses (1).
One of the most widely practiced measurements is the immunohistochemical assessment of Ki-67 antigen, a nuclear marker expressed in all phases of the cell cycle other than G0 (3, 4). Previous investigations have demonstrated that the percentage of Ki-67-positive tumor cells, also known as the Ki-67 labeling index (LI), can be used to risk stratify patients with GIST (5-6-7). However, Ki-67 LI is not an established prognostic factor in clinical practice or in the management of patients with GIST. In addition, Ki-67 LI can be affected by various factors, such as differences in immunohistochemical methods, evaluation methods and observers.
In the present study, we investigated the suitable optimal criteria of high Ki-67 LI and its correlation with survival in 1,967 GISTs from 24 eligible studies. Moreover, a diagnostic test accuracy review was conducted to confirm the predictive role of high Ki-67 LI for high risk of tumor recurrence.
Materials and methods
Published studies search and selection criteria
Relevant articles were obtained by searching the PubMed and MEDLINE databases through April 30, 2015. Searches were performed using the following keywords: “gastrointestinal stromal tumor(s),” “Ki-67” (also “Ki 67” and “Ki67”) and “MIB-1” (also “MIB 1” and “MIB1”). The title and abstract of all searched articles were screened for exclusion. Two independent investigators first screened titles, abstracts and keywords to determine if the given information of each study met the criteria described below, and then they retrieved full-text publications of all potentially eligible articles. Review articles were also screened to identify additional eligible studies. The search results were then scanned according to the following inclusion and exclusion criteria: (1) Ki-67 LI investigated in human GIST tissue; (2) the correlation between Ki-67 LI and survival rate was included; (3) case reports were excluded; and (4) English or Korean language publication.
The following items were collected from the full texts of selected articles by J. Pyo and were verified by G. Kang: publication date, first author’s name, clone name/source and dilution, cutoff for assessing high Ki-67 LI, and data allowing estimation of the impact of Ki-67 expression on disease-free survival (DFS) and/or overall survival (OS). Whenever possible, data were extracted according to the NIH risk-group stratification system (based on tumor size and mitotic index). We did not define a minimum number of patients needed to include a study in our analysis. Data associated with survival were extracted after a 60-month follow-up period. When the authors reported the same patient population in several publications, only the most recent or complete study was included in the analysis. Any disagreements were resolved by consensus.
Ki-67 LI was considered high or low according to the cutoff values provided by the authors. For quantitative aggregation of survival results, the impact of Ki-67 on survival was analyzed according to hazard ratios (HRs) using 1 of 3 methods. In studies not quoting the HR or its confidence interval (CI), these variables were calculated from the presented data using the following parameters: the HR point estimate, the log-rank statistic or its p value, and the
The HRs were then combined into an overall HR using Peto’s method (9). In the current meta-analysis, the random-effect model was more suitable than the fixed-effect model, because all eligible studies used various methods and cutoff values for evaluation of Ki-67 LI. In addition, the random-effect model is suitable to estimate the mean of an effect distribution. Therefore, the pooled estimates in the current meta-analysis were analyzed using the random-effect model. Heterogeneity between studies was assessed using the Q and I2 statistics. For assessment of publication bias, Begg’s funnel plot and Egger’s test were performed. If significant publication bias was found, the fail-safe N and trim-and-fill tests were additionally conducted to confirm the degree of publication bias. The results were 2-sided and considered statistically significant at p<0.05. These analyses were performed using Comprehensive Meta-Analysis (CMA) version 2.0 (Biostat).
To confirm the correlation between high Ki-67 LI and risk stratification, we performed a diagnostic test accuracy review using the Meta-Disc program (version 1.4) (10). The forest plots for sensitivity and specificity and the summary receiver operating characteristic (SROC) curve were obtained through diagnostic test accuracy review. The value of the area under the curve (AUC) on SROC and the diagnostic odds ratio (OR) were calculated using the Meta-Disc program.
Characteristics of the studies
Using the search keywords we identified 211 potentially relevant publications at the level of the abstract and title. Studies were excluded if any of the following criteria applied: 1) incomplete or insufficient information; 2) non-English; 3) duplicate publication; 4) performed using animal models or cell lines; 5) review or case report; and 6) study of another disease. Applying the inclusion and exclusion criteria, we eventually selected 24 reports for the meta-analysis (
Main characteristics of eligible studies
|Study||Antibody Corp.||Dilution||Cutoff (%)||N||Ki-67 LI|
|N = number of patients; LI = labeling index; ND = no description.|
|Artigiani Neto 2012 (11)||Dako||300||0||55||28||27|
|Calabuig-Farinas 2011 (12)||Dako||50||5||108||58||50|
|Carrillo 1997 (13)||Immunotech||ND||22||59||12||47|
|Cerski 2011 (14)||Dako||50||3||84||28||57|
|Di Vizio 2008 (15)||Dako||16000||10||94||21||73|
|Fujita 2012 (16)||Dako||100||4.9||47||15||32|
|Gelen 2003 (17)||Immunotech||80||10||50||22||28|
|Hu 2008 (18)||Dako||ND||4||49||32||17|
|Huang 2009 (19)||Dako||100||5||285||80||205|
|Imamura 2007 (20)||Dako||100||4.6||95||ND||ND|
|Jeong 2014 (21)||Novocastra||250||5||32||19||13|
|Liang 2012 (22) (Small bowel)||Invitrogen||50||5||42||13||29|
|Liu 2013 (23)||Abcam||100||10||113||39||74|
|Lu 2013 (24)||Abcam||1000||5||111||53||58|
|Nakamura 2005 (25)||Dako||100||4.92||80||21||37|
|Ogino 2013 (7)||Dako||ND||10||22||9||13|
|Rudolph 1998 (26)||ND||ND||10||52||22||30|
|Sanchez Hidalgo 2007 (27)||ND||25||10||29||8||21|
|Song 2009 (28)||Dako||80||1||135||93||42|
|Tsai 2008 (29)||Dako||ND||5||49||15||34|
|Wang 2014 (30)||Cell Signaling Tech.||200||1||84||35||49|
|Wang 2001 (31)||Immunotech||50||5||44||23||21|
|Wong 2003 (32)||Dako||20||5||108||22||86|
|Wu 2006 (33)||Dako||ND||10||85||5||80|
Flow chart of study search and selection methods.
Systematic review and meta-analysis
Correlation between high Ki-67 LI and survival
Among the eligible studies, 15 and 11 analyzed subsets provided data for the correlation between Ki-67 LI and DFS or OS, respectively. A high Ki-67 LI was significantly associated with worse DFS (HR 3.658, 95% CI 2.687-4.979, p<0.001,
Forest plot diagrams of hazard ratio for correlations between high Ki-67 labeling index and disease-free survival (
Next, we conducted a subgroup analysis based on Ki-67 LI to evaluate the optimal criteria of high Ki-67 LI. Accordingly, eligible studies were divided into categories based on the cutoff value. The results of the subgroup analysis based on the criteria for high Ki-67 LI are shown in
Forest plot diagrams of hazard ratios according to criteria of high Ki-67 labeling index. (
Diagnostic test accuracy review
To confirm the diagnostic accuracy of the predictive role of high Ki-67 LI for higher (high and intermediate) risk, we performed a diagnostic test accuracy review. Among the eligible studies, 10 studies showed a correlation between Ki-67 LI and risk stratification of tumor recurrence (15, 18, 21, 23-24-25, 29, 30, 34, 35). The pooled sensitivity and specificity of these 10 studies were 0.44 (95% CI 0.41-0.48) and 0.87 (95% CI 0.84-0.91), respectively (
Pooled sensitivity (
Summary receiver operating characteristic (SROC) curve of high Ki-67 labeling index for predicting high and intermediate risks of tumor recurrence.
The recurrence and prognosis of GISTs can be clinically predicted using tumor size and mitotic rate. Ki-67 LI has been widely used as a prognostic marker in various tumors such as neuroendocrine tumors, soft tissue tumors and breast tumors. However, the prognostic relevance and diagnostic standardization of Ki-67 LI in GISTs have not yet been fully elucidated. The present study is the first systematic review and meta-analysis of published studies for evaluation of the role of Ki-67 LI as a prognostic marker in GIST.
In clinical practice, evaluation for tumor recurrence distinguishes high, intermediate, low and very low risks through the combination of tumor size and mitotic rate. In addition, there are different staging criteria between primary tumor locations, regardless of similar parameters. The mitotic rate is expressed as the number of mitotic figures per 50 high power fields. However, unlike tumor size, counting mitoses under a microscope can vary due to numerous causes including inter- and intra-observer differences, differentiation with pyknosis, and observed tumor area. Because the standardization for evaluation of mitosis has not been fully elucidated in GIST, risk evaluation among studies may show discrepancies. Prognostic parameters based on IHC, such as Ki-67, can be useful in decreasing such discrepancies. Previous studies have reported a correlation between Ki-67 LI and mitotic rate in GIST (16, 18, 23, 36, 37); however, this conclusion is controversial. The current meta-analysis showed that high Ki-67 Li was significantly concordant with higher mitotic count (pooled sensitivity = 0.60, 95% CI 0.51-0.69; pooled specificity = 0.91, 95% CI 0.88-0.94; AUC in SROC = 0.977; data not shown). In addition, previous studies reported that Ki-67 LI was an independent parameter for prediction of prognosis by multivariate analysis (7, 11-12-13-14-15-16-17-18-19-20-21-22-23-24-25-26-27-28-29-30-31-32-33). Taken together, these results suggest that Ki-67 LI can be used as an alternative marker of the mitotic rate.
Biopsy is needed if preoperative therapy is being considered for unresectable or metastatic GIST. In addition, [for high diagnostic accuracy of] fine-needle aspiration (FNA) biopsy for GISTs, Ki-67 LI has been shown to predict the malignant potential in (37, 38). On the other hand, a recent study showed that the mean mitotic count of tissue obtained from endoscopic ultrasound-guided core biopsies was significantly lower than that in resected specimens (39). The main reason for this discrepancy is uneven distribution of mitotic figures within a tumor; in addition, mitotically active areas can be missed in a small biopsy sample (40). Despite these limitations, the National Comprehensive Cancer Network (NCCN) recommends Ki-67 IHC analysis when the mitotic index is based on fewer than 50 HPF (i.e., in small biopsy samples), which could further support the proliferation rate as determined by the mitotic rate (41). In our diagnostic test accuracy review, the specificity was as high as 0.87. If low Ki-67 LI is identified, the probability of low or very low risk of tumor recurrence might be very high, regardless of tumor size. Therefore, Ki-67 LI might be useful for risk stratification in small biopsy and FNA specimens.
Our meta-analysis showed that high Ki-67 LI was associated with a higher risk of recurrence and a worse survival at variable cutoff values, which raises the question of an optimal Ki-67 cutoff point. Until now, there has been a lack of consensus on scoring for Ki-67 IHC and an appropriate cutoff point for positivity or high expression in GISTs. Many cutoff values have been used in previous studies, varying from a few percent to more than 20%. These cutoff values have limited value outside the studies from which they were derived and the laboratories that performed them (4). In the absence of standardized IHC methodology, we were unable to achieve an ideal cutoff point that could be used in routine pathology practice. However, with regard to which part of the tumor should be assessed (e.g., hot spots, overall average), it seems reasonable that Ki-67 LI be obtained from the tumor area with the highest nuclear labeling on screening, much like the grading of neuroendocrine tumors (1). The criteria used for Ki-67 LI are 2% and 20% in neuroendocrine tumors. However, more detailed criteria would be needed for evaluation of the usefulness of Ki-67 LI in GIST. Subgroup analysis could be helpful for confirmation of the proper cutoff value in determining high Ki-67 LI. Interestingly, the difference in HRs for DFS was largest between subgroups using a 4% cutoff (
There are a number of limitations to the present study. First, with the current meta-analysis we could not evaluate the differences in prognostic roles between surgical specimens and other types of specimens, such as small biopsy and FNA specimens, due to insufficient information in the eligible studies. Further prospective studies are needed to investigate the prognostic role of Ki-67 LI in small biopsy and FNA specimens of GIST. Second, the follow-up periods differed between eligible studies. To avoid bias from follow-up periods, survival data were extracted after a 60-month follow-up period. Although the follow-up period had no effect on the correlation between Ki-67 LI and survival, the correlation between Ki-67 LI and survival could differ from those in previous reports. Third, the present study included only 1 study using image analysis (13). To assess the effect on HR, sensitivity analysis was conducted, which showed that Carrillo’s study (13) had no effect on the pooled HR (HR 3.517, 95% CI 2.679-4.618). Although the concordance of Ki-67 LI was good between microscopic observation and image analysis in a study using a cutoff of 10% in GISTs (7), additional analysis of the correlation between microscopic observation and image analysis could not be performed due to the small number of studies using image analysis. Lastly, analysis related to specific organs of the gastrointestinal tract, such as stomach and small bowel, could not be performed due to insufficient information from eligible studies.
In conclusion, the current meta-analysis showed that high Ki-67 LI was significantly correlated with worse survival and high risk of tumor recurrence. Further prospective studies for diagnostic standardization and correlation between Ki-67 LI and mitotic rate are required prior to application in various tumor locations and various types of specimens.
Edge SB; American Joint Committee on Cancer. 7th ed. New York: Springer 2010; 648.
Coccolini F Catena F Ansaloni L Pinna AD Gastrointestinal stromal tumor and mitosis, pay attention. 2012 18 6 587 588
Gerdes J Lemke H Baisch H Wacker HH Schwab U Stein H Cell cycle analysis of a cell proliferation-associated human nuclear antigen defined by the monoclonal antibody Ki-67. 1984 133 4 1710 1715
Dowsett M Nielsen TO A’Hern R et al; International Ki-67 in Breast Cancer Working Group. 2011 103 22 1656 1664
Hasegawa T Yamamoto S Nojima T et al. Validity and reproducibility of histologic diagnosis and grading for adult soft-tissue sarcomas. 2002 33 1 111 115
Yamaguchi U Hasegawa T Sakurai S et al. Interobserver variability in histologic recognition, interpretation of KIT immunostaining, and determining MIB-1 labeling indices in gastrointestinal stromal tumors and other spindle cell tumors of the gastrointestinal tract. 2006 14 1 46 51
Ogino J Asanuma H Hatanaka Y et al. Validity and reproducibility of Ki-67 assessment in gastrointestinal stromal tumors and leiomyosarcomas. 2013 63 2 102 107
Parmar MK Torri V Stewart L Extracting summary statistics to perform meta-analyses of the published literature for survival endpoints. 1998 17 24 2815 2834
Yusuf S Peto R Lewis J Collins R Sleight P Beta blockade during and after myocardial infarction: an overview of the randomized trials. 1985 27 5 335 371
Zamora J Abraira V Muriel A Khan K Coomarasamy A Meta-DiSc: a software for meta-analysis of test accuracy data. 2006 6 1 31
Artigiani Neto R Logullo AF Stávale JN Lourenço LG Ki-67 expression score correlates to survival rate in gastrointestinal stromal tumors (GIST). 2012 27 5 315 321
Calabuig-Fariñas S López-Guerrero JA Navarro S et al. Evaluation of prognostic factors and their capacity to predict biological behavior in gastrointestinal stromal tumors. 2011 19 4 448 461
Carrillo R Candia A Rodriguez-Peralto JL Caz V Prognostic significance of DNA ploidy and proliferative index (MIB-1 index) in gastrointestinal stromal tumors. 1997 28 2 160 165
Cerski MR Pereira F Matte US et al. Exon 11 mutations, Ki67, and p16(INK4A) as predictors of prognosis in patients with GIST. 2011 207 11 701 706
Di Vizio D Demichelis F Simonetti S et al. Skp2 expression is associated with high risk and elevated Ki67 expression in gastrointestinal stromal tumours. 2008 8 1 134
Fujita A Yamamoto H Imamura M et al. Expression level of the mitotic checkpoint protein and G2-M cell cycle regulators and prognosis in gastrointestinal stromal tumors in the stomach. 2012 460 2 163 169
Gelen T Elpek GO Aksoy NH Ogüş M Keleş N p27 Labeling index and proliferation in gastrointestinal stromal tumors: correlations with clinicopathologic factors and recurrence. 2003 33 7 346 352
Hu TH Tai MH Chuah SK et al. Elevated p21 expression is associated with poor prognosis of rectal stromal tumors after resection. 2008 98 2 117 123
Huang HY Li SH Yu SC et al. Homozygous deletion of MTAP gene as a poor prognosticator in gastrointestinal stromal tumors. 2009 15 22 6963 6972
Imamura M Yamamoto H Nakamura N et al. Prognostic significance of angiogenesis in gastrointestinal stromal tumor. 2007 20 5 529 537
Jeong SY Park WW Kim YS et al. [Prognostic significance of Ki-67 expression in patients undergoing surgical resection for gastrointestinal stromal tumor]. 2014 64 2 87 92 [Article in Korean].
Liang YM Li XH Li WM Lu YY Prognostic significance of PTEN, Ki-67 and CD44s expression patterns in gastrointestinal stromal tumors. 2012 18 14 1664 1671
Liu LC Xu WT Wu X Zhao P Lv YL Chen L Overexpression of carbonic anhydrase II and Ki-67 proteins in prognosis of gastrointestinal stromal tumors. 2013 19 16 2473 2480
Lu C Liu L Wu X Xu W CD133 and Ki-67 expression is associated with gastrointestinal stromal tumor prognosis. 2013 6 5 1289 1294
Nakamura N Yamamoto H Yao T et al. Prognostic significance of expressions of cell-cycle regulatory proteins in gastrointestinal stromal tumor and the relevance of the risk grade. 2005 36 7 828 837
Rudolph P Gloeckner K Parwaresch R Harms D Schmidt D Immunophenotype, proliferation, DNA ploidy, and biological behavior of gastrointestinal stromal tumors: a multivariate clinicopathologic study. 1998 29 8 791 800
Sánchez Hidalgo JM Muñoz Casares FC Rufian Peña S et al. [Gastrointestinal stromal tumors (GIST): factors predictive of survival after R0-cytoreduction]. 2007 99 12 703 708 [Article in Spanish].
Song Z Wang JL Pan YL Tao DY Gan MF Huang KE Survival and prognostic factors analysis in surgically resected gastrointestinal stromal tumor patients. 2009 56 89 149 153
Tsai MC Lin JW Lin SE Chen HH Lee CM Hu TH Prognostic analysis of rectal stromal tumors by reference of National Institutes of Health risk categories and immunohistochemical studies. 2008 51 10 1535 1543
Wang H Chen P Liu XX et al. Prognostic impact of gastrointestinal bleeding and expression of PTEN and Ki-67 on primary gastrointestinal stromal tumors. 2014 12 1 89
Wang X Mori I Tang W et al. Gastrointestinal stromal tumors: clinicopathological study of Chinese cases. 2001 51 9 701 706
Wong NA Young R Malcomson RD et al. Prognostic indicators for gastrointestinal stromal tumours: a clinicopathological and immunohistochemical study of 108 resected cases of the stomach. 2003 43 2 118 126
Wu TJ Lee LY Yeh CN et al. Surgical treatment and prognostic analysis for gastrointestinal stromal tumors (GISTs) of the small intestine: before the era of imatinib mesylate. 2006 6 1 29
Neves LR Oshima CT Artigiani-Neto R Yanaguibashi G Lourenço LG Forones NM Ki67 and p53 in gastrointestinal stromal tumors-GIST. 2009 46 2 116 120
Zhao WY Xu J Wang M Zhang ZZ Tu L Wang CJ Lin TL Shen YY Liu Q Cao H Prognostic value of Ki67 index in gastrointestinal stromal tumors. 2014 7 5 2298 2304
Yang J Feng F Li M et al. Surgical resection should be taken into consideration for the treatment of small gastric gastrointestinal stromal tumors. 2013 11 1 273
Ando N Goto H Niwa Y et al. The diagnosis of GI stromal tumors with EUS-guided fine needle aspiration with immunohistochemical analysis. 2002 55 1 37 43
Okubo K Yamao K Nakamura T et al. Endoscopic ultrasound-guided fine-needle aspiration biopsy for the diagnosis of gastrointestinal stromal tumors in the stomach. 2004 39 8 747 753
Larghi A Fuccio L Chiarello G et al. Fine-needle tissue acquisition from subepithelial lesions using a forward-viewing linear echoendoscope. 2014 46 1 39 45
Ricci R Chiarello G Vanella G Larghi A On the reliability of mitotic count on biopsy samples of gastrointestinal stromal tumors. 2014 40 4 484 485
Demetri GD von Mehren M Antonescu CR et al. NCCN Task Force report: update on the management of patients with gastrointestinal stromal tumors. 2010 8 Suppl 2 S1 S41, quiz S42-S44.
- Pyo, Jung-Soo [PubMed] [Google Scholar] 1
- Kang, Guhyun [PubMed] [Google Scholar] 2
- Sohn, Jin Hee [PubMed] [Google Scholar] 1, * Corresponding Author (firstname.lastname@example.org)
Department of Pathology, Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine, Seoul - South Korea
Department of Pathology, Inje University Sanggye Paik Hospital, Seoul - South Korea
Jung-Soo Pyo and Guhyun Kang contributed equally to this study