Ki-67/MKI67 as a Predictive Biomarker for Clinical Outcome in Gastric Cancer Patients: an Updated Meta-analysis and Systematic Review involving 53 Studies and 7078 Patients

Gastric cancer (GC) threatens human health worldwide and we performed this meta-analysis to evaluate the clinical value of Ki-67/MKI67 in patients with GC. The combined hazard ratio (HR), odds ratio (OR) and 95% confidence interval (95% CI) were calculated to assess the relationships of Ki-67/MKI67 expression with prognoses and clinicopathological characteristics. Genes co-expressed with MKI67 were collected for Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and protein-protein interaction (PPI) network analyses. In total, 53 studies with 7078 patients were included in this study. The pooled HRs indicated that an elevated expression of Ki-67/MKI67 predicted an unfavorable overall survival (HR: 1.54, 95% CI: 1.33-1.78, P<0.0001) and disease-free survival (HR: 2.28, 95% CI: 1.43-3.64, P<0.0001) in GC patients. Additionally, in patients with advanced GC, a high Ki-67/MKI67 expression was also significantly connected with OS (HR: 1.37, 95% CI: 1.18-1.60, P<0.0001). The combined ORs showed that Ki-67/MKI67 expression was related to TNM stage (stage III/IV versus stage I/II: OR=1.93, 95% CI=1.34-2.78, P<0.0001), tumor differentiation (poor versus well/moderate: OR=1.94, 95% CI=1.32-2.85, P=0.001), lymph node metastasis (yes versus no: OR=1.67, 95% CI=1.23-2.25, P=0.001), distant metastasis (yes versus no: OR=1.67, 95% CI=1.24-2.26, P=0.001) and tumor invasion depth (T3/T4 versus Tis/T1/T2: OR=1.98, 95% CI=1.60-2.44, P<0.0001). The results of GO, KEGG pathway and PPI network analyses indicated that Ki-67/MKI67 may be involved in the development of GC via influencing P53 signaling pathway. Ki-67/MKI67 could be a potential indicator to predict the prognosis of patients with GC and identify high-risk cases. Detecting Ki-67/MKI67 expression in clinic may be helpful in optimizing individual treatment and further improving the survival expectancy of patients with GC.


Background
Gastric cancer (GC) is one of the most common and aggressive malignancies worldwide according to the World Health Organization [1]. In the US, about 28,000 new cases and 10,960 deaths occurred in 2017 [2]. While in China, GC is the second main cause of cancer-related death with an estimated 498,000 deaths occur annually [3]. In spite of the improvements in diagnosis and treatment [4], currently, the long-term survival rate for a large number of GC patients is still dismal, with a 5-year survival rate of less than 20% [5]. Most patients have developed tumor metastasis when they are diagnosed with GC, and the high tumor Ivyspring International Publisher metastasis rate leads to unfavorable survival outcome in patients with GC [6]. Identifying a biomarker for early diagnosis and clinical outcome prediction is important to understand the development of GC and further improve the prognosis in patients with GC.
Antigen Ki-67, also known as MKI67 (antigen identified by monoclonal antibody Ki-67), is a cell proliferation-related protein that encoded by gene MKI67. Ki-67/MKI67 exists in G1, S, G2 and mitosis phases of cell cycle [7] and can be put into clinical practice by acting as an index to evaluate cell proliferation by immunohistochemical staining [8]. A large number of studies have reported the clinical and prognostic value of Ki-67/MKI67 in various malignancies, such as breast cancer [9], cervical cancer [10], glioma [11], colorectal cancer [12] and hepatocellular carcinoma [13]. Our research group previously conducted systematic meta-analyses to explore the prognostic role of Ki-67/MKI67 in lung cancer [14], hepatocellular carcinoma [13], cervical cancer [15], glioma [16] and colorectal cancer [17] and found that Ki-67/MIB-1 could be an independent prognostic biomarker in these types of carcinoma.
Although two previous meta-analyses have reported the potential of Ki-67/MKI67 as a predictive biomarker in patients with GC [18,19], studies based on more cases and stronger evidence are still needed to corroborate the prognostic and clinicopathological role of Ki-67/MKI67 for GC patients. Thus, we conducted a comprehensive literature search and performed a systematic analysis on the prognostic value and clinical significance of Ki-67/MKI67 in GC. Simultaneously, we investigated the underling action mechanism by which Ki-67/MKI67 affects cell proliferation and promotes tumor development by performing bioinformatics analysis on genes co-expressed with MKI67.

Publication identification and searching strategy
A comprehensive literature search via PubMed, PMC, Science Direct, Web of Science, Wiley Online, Chinese National Knowledge Infrastructure (CNKI), Chinese WanFang database, Chinese Chongqing VIP and Embase was carried out up to March 5, 2019 based on the following searching terms: (Ki-67 OR Ki67 OR MKI67 OR MIB-1 OR proliferative index OR proliferative activity OR mitotic index OR mitotic activity OR mitotic figure OR labeling index OR mitotic count OR proliferative marker) AND (prognos* OR surviv* OR predict* OR follow-up OR followed-up OR mortality OR outcome OR diagnosis OR diagnostic OR detect) AND (gastric OR stomach OR tummy) AND (cancer OR carcinoma OR tumo* OR neoplas* OR malignan*).

Eligibility criteria for inclusion and exclusion
Two investigators evaluated all of the acquired records independently using the criteria listed below: (1) All patients were distinctly diagnosed with gastric cancer by pathology; (2) The studies were written in English or Chinese; (3) The studies investigated the expression level of Ki-67/MKI67 in gastric cancer and adjacent non-cancerous tissues; (3) The studies reported the relationships of Ki-67/MKI67 expression with prognosis or clinicopathological parameters in gastric cancer patients; (4) The studies provided enough information to estimate the hazard ratios (HRs), odds ratios (ORs) and their 95% confidence intervals (CIs); (5) The latest and most complete study was included when the same patients were reported in different publications; (6) The studies must be performed in humans.
The following publications were excluded: conference abstracts, reviews, meta-analyses, case reports, expert opinions, studies not written in English or Chinese, and studies without sufficient data to calculate HRs, ORs and their 95% CIs. Because of the heterogeneous expression of Ki-67 in GC, tissue microarrays (TMAs) carry the risk of sampling error due to an underestimation and non-representative evaluation of Ki-67 in GC [20]. Therefore, we excluded studies using TMAs.

Data extraction
Two researchers independently performed the data collection from all of the enrolled studies, with any discrepancies settled by discussion with a third reviewer. The relevant information covered the following details: first author and publication year, region, cut-off value, antibody, quality score, statistical method, and HR with corresponding 95% CI.

Quality assessment
Two investigators independently read each enrolled prognosis-related record and scored it using the Newcastle-Ottawa scale (NOS) [21]. Each study was assessed based on the following three perspectives: selection, comparability and outcome. The score ranged from 0 to 9, and high scores mean high quality of the included records. Additionally, the Quality Assessment of Diagnostic Accuracy Studies-2 (QADAS-2) tool was used to assess the quality of diagnostic accuracy studies [22]. The risk of bias and applicability of individual studies were judged as "low", "high" or "unclear" based on the following four critical domains: patient selection, index test, reference standard, and flow and timing. Both reviewers compared their judgments and achieved a consensus by discussing in conference if the results were controversial.

Statistical analysis
To determine the ability of Ki-67/MKI67 in discriminating tumor from non-tumorous gastric tissues, we performed summary receiver operating characteristic curve (SROC) analysis and calculated the area under the curve (AUC) value, the diagnostic OR and the corresponding sensitivity and specificity by using Meta-DISc [23].
The HRs and 95% CIs were combined to measure the effects of Ki-67/MKI67 expression on overall survival (OS) and disease-free survival (DFS). We directly extracted the HRs and 95% CIs from the original studies if the survival data were provided; otherwise, we calculated them by using Engauge Digitizer Version 4.1 on the basis of the Kaplan-Meier survival curves. Three investigators read the survival curves independently to improve the precision in the extraction of HRs and 95% CIs. Then we combined the individual HRs into a pooled HR by using Stata 12.0 (StataCorp, College Station, TX, USA) to investigate the correlation between Ki-67/MKI67 expression and prognosis in patients with GC. Subgroup analyses were conducted to determine whether different regions, antibodies, cut-off values and statistical methods resulted in the differences of the overall results. An observed HR>1 and it's 95% CI not crossing 1 indicated that a high expression of Ki-67/MKI67 predicted a poor outcome in gastric cancer patients.
The impacts of Ki-67/MKI67 on clinicopathological parameters were considered statistically significant if the 95% CI did not cross 1. The data calculations were conducted by using Stata 12.0.
To evaluate the stability of the pooled results, we conducted sensitivity analysis by excluding individual studies successively. The χ 2 test (Chi squared test; Chi 2 ) and the I 2 test (Higgins I-squared test; I 2 ) were used to analyze the heterogeneity among studies. If an inter-study heterogeneity existed (P<0.05 and/or I 2 ≥50%), a random-effects model was applied; otherwise, a fixed-effects model was chosen. Begg's funnel plot and Egger's test were used to assess the publication bias of the eligible studies. All P<0.05 (tested by two-sided) were considered statistically significant.

Expression level of Ki-67/MKI67 in GC
Three studies [35,40,44] with 384 patients provided sufficient data to compare the expression of Ki-67/MKI67 in gastric cancerous and adjacent non-cancerous tissues. The potential risk of bias and applicability of the three studies are shown in  expectancy (data not shown). We found that the group with a cut-off value less than 0.25 showed the highest HR, and the group with a cut-off value more than 0.25 showed the lowest HR. For OS, the HRs in high and low cut-off value groups were 1.38 (95% CI: 1.17-1.63, P<0.0001) and 1.77 (95% CI: 1.37-2.28, P<0.0001), respectively. Two studies did not provide the cut-off value of Ki-67/MKI67 expression, and the pooled HR of the two articles was 2.00 (95% CI: 1.36-2.93, P<0.0001). For DFS, the overall results suggested that highly expressed Ki-67/MKI67 was related to worse survival outcome both in high cut-off value group (HR: 1.51, 95% CI: 1.08-2.11, P=0.017) and low cut-off value group (HR: 3.76, 95% CI: 2.14-6.60, P<0.0001).

Clinicopathological significance of Ki-67/MKI67 for GC patients
The clinicopathological value of Ki-67/MKI67 expression for GC patients was assessed (

Sensitivity analysis
Results of sensitivity analysis showed that the combined HRs and ORs corresponding to the successive exclusion of individual reports were not significantly altered, indicating that the pooled results of the current meta-analysis were stable (Figs. 6-8).

Publication bias
The Begg's funnel plots are presented in Fig. 9. No publication bias existed in all situations (Table 3  and Table 4).

Potential action mechanism of MKI67 in GC
In total, 440 genes co-expressed with MKI67 in GC were collected for GO, KEGG pathway and PPI network analyses. The top five GO functional annotations and top ten KEGG pathways are presented in Fig. 10. The results of GO enrichment analyses indicated that the main biological functions of MKI67 were related to cell division, nucleoplasm and protein binding. The results of KEGG pathway analysis revealed that these co-expressed genes were involved in P53 signaling pathway.

Discussion
In the present study, we firstly investigated the ability of Ki-67/MKI67 in discriminating tumor from normal controls. We found that an over-expressed Ki-67/MKI67 could differentiate gastric cancer patients from normal subjects. However, the result should be interpreted with caution because only three relevant studies were included and an inter-study heterogeneity among the three studies was discovered. Further researches with larger sample scales are necessary to validate our result.
A number of studies have indicated that Ki-67/MKI67 is a predictive marker for tumor deterioration and its expression shows an association with unfavorable prognosis and malignant clinical phenotypes in various cancers, such as bladder cancer [77], non-small cell lung cancer [78], hepatocellular carcinoma [13], cervical cancer [10] and glioma [16]. The clinical and prognostic value of Ki-67/MKI67 in GC has also been investigated previously. Luo et al. [19] has conducted a meta-analysis based on 29 studies and demonstrated that a high Ki-67/MKI67 expression predicted poor OS and DFS in patients with GC. The authors also discovered that an increased Ki-67/MKI67 expression was related to Lauren's classification and tumor size. Liu et al. [18] also performed a meta-analysis and found that a highly expressed Ki-67/MKI67 predicted an inferior OS in 3825 patients with GC. Compare with Luo's and Liu's studies, our study included much more samples, resulting in more persuasive and stronger conclusions. A total of 53 studies with 7078 patients were enrolled in our meta-analysis. We demonstrated the prognostic and clinicopathological value of Ki-67/MKI67 for GC patients. Especially, we revealed the potential of Ki-67/MKI67 as a prognostic biomarker in advanced gastric cancer patients.
Metastasis is the most important cause of cancer-related death in patients with GC. Identifying a marker to predict the presence of metastasis and stratify patients into different risk groups is helpful in improving the prognosis of GC patients. Here, an up-regulation of Ki-67/MKI67 was discovered in patients with advanced TNM stage, poor tumor differentiation, serosa and neighboring organs invasion, lymph node metastasis and distant metastasis, suggesting that Ki-67/MKI67 could function as an indicator to predict gastric cancer progression and identify high-risk patients, thereby optimizing individual treatment management and improving the prognosis of patients with GC.
Ki-67/MKI67 is a well-known cell cycle-related protein. A recent study has corroborated that specific Ki-67/MKI67 splice variants promote cancer progression by influencing cell cycle [79]. It is reported that P53 exert inhibitory effect on Ki-67 promoter by regulating P53-and SP1-dependent pathways [80]. Interesting, we also demonstrated the close link between Ki-67/MKI67 and P53 signaling pathway. We found that MKI67 interacted with genes RRM2, CCNB2, GTSE1, CDK1, CCNB1 and CHEK1, which are critical molecules those are involved in P53 signaling network and related to cell cycle progression. Based on previous studies and our evidence, we hypothesize that MKI67 may promote the initiation and progression of GC via interacting with P53 signaling pathway. Further in vitro and in vivo experiments are necessary to confirm our conclusions. Fig. 6. Result of sensitivity analysis from a random-effects model for overall survival (a) and disease-free survival (b). Three vertical lines indicated the pooled hazard ratio (HR) with 95% confidence interval (CI) calculated from all included studies. Each dotted horizontal line belongs to an independent meta-analysis. And the middle circle and two sides' short vertical lines represents the pooled HR and its 95% CI corresponding to the successive exclusion of each study.   There were still several deficiencies and limitations in our study. First, an obvious inter-study heterogeneity among the included studies was discovered in the present meta-analysis. There were no unified standards for the antibody concentration, IHC staining degree and cut-off value in different studies, which may be part of the reasons for the heterogeneity. Additionally, the differences in clinicopathological characteristics such as age, gender, tumor stage may also lead to the heterogeneity. A random-effects model was applied to eliminate the effect of the heterogeneity on our results. Remarkably, although the heterogeneity existed in our study, the results of sensitivity analyses indicated that no individual researches influenced the pooled HRs and ORs. Second, in 24 studies, HRs with 95% CIs had to be calculated from the Kaplan-Meier survival curves, which may decrease the reliability and accuracy of our meta-analysis. To reduce the deviation, three researchers independently extracted HRs from the Kaplan-Meier curves; the similarity between the copy curves drawn by the extracted data and original curves was the standard to identify which HRs could be used. Third, our meta-analysis depended on fully published studies written in English or Chinese, which may contribute to potential language bias. In addition, studies with positive results are more likely to be published in magazines. Thus, we should not ignore the potential bias even though the results from Begg's and Egger's tests showed no publication bias in the present study.

Conclusions
Our meta-analysis provides evidence that Ki-67/MKI67 not only could be a potential prognostic biomarker in clinic for GC patients but also could be an indicator to predict GC progression and to identify high-risk cases, thereby optimizing individual treatment management and improving the prognosis of GC patients. More interesting, we also found that gene MKI67 probably promoted the occurrence and development of GC by influencing P53 signaling pathway. Further well-designed studies are required to validate our conclusions.