The Prognostic Roles and Clinical Features of CD44v9 in Human Solid Cancers—A Meta-Analysis

Background: CD44 is the primary receptor for hyaluronic acid and serves as a marker for cancer stem cells. CD44v9 is one of CD44’s variants and takes part in cancer’s growth and metastasis. However, the prognostic roles and clinical features of CD44v9 in cancers remain unclear. Therefore, we conducted this meta-analysis to summarize the prognostic signicance and clinical features of CD44v9 in human solid cancers. Methods: we systematically searched all of related studies in PubMed, the Web of Science, Embase and Cochrane library up to June 2020. We analyzed the pooled hazard ratios (HRs) and odds ratios (ORs) with corresponding 95% condence intervals (CIs) to assess the prognostic functions and clinical features of CD44v9 in various human solid cancers. Results: In this meta-analysis, we included 1705 cancer patients among 12 studies. Results indicated that high expression of CD44v9 was signicantly related to poorer overall survival (OS) (HR=1.60, 95%CI 1.28-1.99, P<0.0001), recurrence-free survival/progression-free survival/disease-free survival (RFS/PFS/DFS).( HR=1.81, 95%CI 1.16-2.84, P=0.009) and disease-specic survival/cancer-specic survival (DSS/CSS) (HR=2.93, 95%CI 1.69-5.10, P<0.001). At the same time, we also found that high expression of CD44v9 increased the possibility of lymphoid inltrates (OR=1.59, 95%CI 1.16-2.20, P=0.005), vascular invasion (OR=1.57, 95%CI 1.11-2.22, P=0.010) and higher TNM stage (OR=1.63, 95%CI 1.19-2.23, P=0.002). Conclusion: Our results demonstrate that CD44v9 overexpression is associated with worse OS, RFS/PFS/CFS and DSS/CSS in patients


Background
Cancer has long been one of the most di cult problems in global health. Statistically, there were 18.1 million people diagnosed cancer and caused 9·6 million deaths worldwide in 2018 [1]. The median cancer-related cost of state-level is estimated at $3.7 billion in America [2]. So, it is urgent to strengthen the early diagnosis and treatment of cancer.
Recently, more and more evidences indicate that there is heterogeneity in cancer cells [3,4]. That is, the cells in the same cancer tissue are not completely the same, most of the cells do not have the ability to differentiate, and only a small number of cells have the self-renewal ability similar to the stem cells in normal tissues, such as embryos and bone marrow, and can differentiate into mature cancer cells. This group of cells is regarded as cancer stem cells (CSCs) [5]. CSCs can produce speci c markers in different cancers, such as the CD34 + HLA-DR-surface marker phenotype in leukemic stem cells, CD44 + ESA + surface marker phenotype in breast stem cells and so on [6]. Therefore, these surface markers may provide better evidences for the diagnosis and treatment of cancers.
CD44 is a member of the single-stranded transmembrane glycoprotein family and is overexpressed in some CSCs [7,8]. As an adhesion molecule, CD44 binding HA is involved in tumor proliferation, adhesion, migration, and invasion [9]. CD44 contains both standard (CD44s) and variant (CD44v) types. The CD44 standard form is produced by exons 1-5 and 16-20, and the CD44v1-10 by exons 6-15 [10]. Many studies have shown that CD44 is closely related to the occurrence and development of cancers [11][12][13][14]. CD44v9 is one of CD44 variants. Although it has been reported that overexpression of CD44v9 is associated with poor cancer survival and may be a therapeutic target for cancers [15][16][17][18], it's still controversial [19]. Therefore, we conducted this systemic meta-analysis in order to reveal the relationship between CD44v9 and human solid cancers.

Search strategy:
We searched primary literatures by keyword plus free words from electronic databases such as PubMed, Web of science, Embase and Cochrane library until June 30, 2020. The speci c PubMed search strategy as following: ("Neoplasms" OR "Neoplasia" OR "Neoplasias" OR "Neoplasm" OR "Tumors" OR "Tumor" OR "Cancer" OR "Cancers" OR "Malignancy" OR "Malignancies" OR "malignant neoplasms" OR "malignant neoplasm" OR "neoplasm malignant" OR "neoplasms malignant" OR "benign neoplasms" OR "neoplasms benign" OR "benign neoplasm" OR "neoplasm benign" OR) and ("CD44v9 antigen" OR "CD44(v9)" OR hyaluronan receptors" OR "CD44v9" OR CD44 variant 9). All the searched literatures and related references were evaluated for inclusion in our meta-analysis. If studies continue on the same batch samples, we included the largest one.

Inclusion and exclusion criteria.
The included studies need to conform to the following criteria: 1) Study needs to assess prognosis about the expression of CD44v9 in cancers. So, overall survival (OS), progression-free survival (PFS), recurrencefree survival (RFS), disease-free survival (DFS), cancer speci c survival (CSS) or disease speci c survival (DSS) should be reported. 2) Related hazard ratios (HRs) with 95% con dence intervals (CIs) or Kaplan-Meier survival curves should be included.
3) The included studies analyze the relationship between the positive and negative expression of CD44v9 and disease characteristics.
The following studies should be excluded: 1) Sample size less than 50. 2) Reviews, case reports, letters, editorials, meeting abstracts and animal trial.

Date extraction and quality assessment:
Literature screening, data extraction and analysis were performed by two researchers (Yuanxiu Deng and Jie Wang) independently. Any disagreement would be solved by other two researchers (Shenhui Ji and Lu Huang). The following information was extracted from each study: rst author, published year, original country, included period, cancer type, average age, sample size, follow-up time, assay method, cut-off value, clinical features, NOS scale, survival outcome and HR with corresponding 95% CI. The software Getdate Graph Diditizer 2.26 was used to extract relevant data to calculate HR with corresponding 95% CI for survival analysis presented by Kaplan-Meier curve. If the article used univariate and multivariate analyses to report the HR value, we choose the univariate analysis. We used STATA15.1 software to calculate ORs with corresponding 95% CI from the extracted date of positive and negative expression number of CD44v9 and disease characteristics. We used the Newcastle-Ottawa scale (NOS) to evaluate the quality of eligible studies. The NOS score ranges from 0 to 9, and any score no less than 6 can be regarded as high quality.

Statistical Analysis Methods
We pooled the HRs with 95% CI to assess the prognostic role of overexpressed CD44v9 in human solid cancers. The ORs with 95% CI was combined to evaluate clinicopathological characteristics. We used the Higgins I² statistics and Chi-squared test to evaluate heterogeneity among the included literatures. When the P value is more than 0.10 and the I² value is less than 50%, it is thought to be no inter-study heterogeneity in the eligible articles. Then a xed-effects model was applied for pooling the data. Otherwise, the random-effects model would be used. If there were heterogeneity, subgroup analysis was used to explore the sources of heterogeneity. We used Begg funnel plot test to quantify the existence of publication bias. Obvious publication bias should be considered when P value is less than 0.1. Sensitivity analysis was used to evaluate the stability of the results from the included articles. All data analysis was analyzed using STATA15.1 software. When P-value is less than 0.05, the results are regarded as statistically signi cant.

Characteristics of the included literatures
In the 12 articles, there were 1705 subjects with eight kinds of cancers included. Eight articles were published in Japan, the remaining published in Korea, China, Netherland and Germany, respectively. Among these articles, three are for gastric cancer [20][21][22], two for breast cancers [19,23], one for esophageal squamous cell carcinoma [24], one for upper tract urothelial carcinoma [25], one for Hepatocellular carcinoma [26], one for pancreatic carcinoma [27], two for bladder cancer [28,29], one for Colorectal Cancer [30]. The sample sizes range from 62 to 333. All the included literatures performed survival analysis among positive and negative expression of CD44v9 in patients with cancers, using OS, PFS, RFS, DFS DSS, CSS. Since there were few references that contain RFS, PFS, DFS, DSS and CSS, we combined RFS, PFS and DFS together, DSS and CSS together for analysis. All the included literatures showed high quality because NOS score was no less than 6. The main information of the twelve articles is presented in Table 1. We used subgroup analysis to explore the source of heterogeneity for OS, RFS/PFS/DFS and DSS/CSS respectively. We performed this analysis from NOS score, sample size and HR value acquisition method. NOS scores no less than 8 were assigned to the high group, while the rest were assigned to the low group.
We de ned small sample events below 110, large sample events no less than 110. As shown in Fig3 Table 2 Meta-analysis of CD44v9 and clinicopathological features in cancer patients

Publication bias and sensitivity analysis among eligible studies
We used Begg's test for possible publication bias in quantitative analysis for survival analysis. As shown in Fig 5A-C, no publication bias exists (P 0.1). Furthermore, sensitivity analysis was used to verify the stability of the meta-analysis for CD44v9 in the OS, RFS/PFS/DFS and DSS/CSS (Fig 6A-C). The results showed that each of the included articles had no effect on the overall stability, that is, the meta-analysis results were stable.

Discussion
Within cancer cells, CD44v9 interacts with the subunit cystine/glutamate antiporter (xCT) to promote intracellular synthesis of the antioxidant glutathione (GSH) [31]. This response can enhance the defense against reactive oxygen species (ROS) and promote tumor growth [32,33]. Some anti-tumor drugs targeting this mechanism may be more effective in controlling tumor progression [29,31,34].
Many scholars have studied the relationship between CD44v9 and human cancer prognosis. For example, Suwannakul's study reveals that CD44v9 is a novel marker of human cholangiocarcinoma stem cells and is closely associated with the development of in ammation-related cancers [15]. Their subsequent researches con rm that CD444v9 overexpression induces a stem-like phenotype that promotes proliferation, migration, invasion, and redox balance of cancer cells in cholangiocarcinoma [35]. However, the effect of CD44v9 in human solid cancers remains unclear. Takaki's study shows that CD44v9 is highly expressed in tumors with EGFR mutations, especially in early-stage lung adenocarcinoma, but not signi cantly associated with advanced lung adenocarcinoma [36]. Hirata reports that among patients with early gastric cancer, the recurrence rate in those with CD44v9 overexpression is signi cantly higher than those with low expression [37]. But a meta-analysis suggests that CD44v9 overexpression is closely related to 5-year CSS and might be a therapeutic target for advanced tumors [10]. Sato observes that patients with low expression of CD44v9 have a poorer ve-year survival rate than those with high expression and have a higher risk of lymphatic metastasis among primary squamous cell carcinoma of the tongue. This may be related to the adhesion of CD44v9. When the expression of CD44v9 decreased, the adhesion weakened, which promoted the metastasis of tumor cells through lymphatics [38].
In this meta-analysis, our results reveal that overexpression of CD44v9 is closely related to the poor OS, RFS/PFS/DFS and DSS/CSS. However, when NOS score, HR extraction method and sample size were used for subgroup analysis, there was no statistical signi cance between CD44v9 and RFS/PFS/DFS. The main reason is that in these studies, HR values were analyzed by single factor analysis or extracted from the survival curve, ignoring the in uence of confounding factors. And the fact that only a few of the literature contain this information is another major factor. Furthermore, according to statistical data, high expression of CD44v9 was positively correlated with lymphatic in ltration, vascular in ltration and high TNM stage, with no or little heterogeneity. There was no publication bias and the results were stable.
There still exist several limitations in this meta-analysis. Firstly, when OS, RFS/PFS/DFS, and DSS/CSS were analyzed respectively, only a few articles provided survival analysis. More studies are needed to provide relevant data to make the results more convincing in the future. Secondly, the CD44v9 immunohistochemistry cut-off value was de ned differently. Thus, more speci c studies are needed to de ne CD44v9 positive expression criteria. Thirdly, most of the studies were from Japan. And extensive research is needed to see whether the results are applicable to other regions. Finally, some HR values were obtained indirectly from Kaplan-Meier survival curves, with unavoidable statistical bias.

Conclusion
To our knowledge, this is the rst meta-analysis to systematically analyze the clinical characteristics and prognostic roles of CD44v9 in human solid cancers. Our results suggest that elevated expression of CD44v9 may be a signi cant predictor of poor prognosis in these patients. CD44v9 might be a targeted molecular marker for the diagnosis and treatment of cancers in the future.