Prognostic value of CD44v9 expression in human cancers: A systematic review and meta-analysis.

BACKGROUND: The prognostic significance of CD44 variant-9 (CD44v9) expression in human cancers has been investigated in several studies, however, definite conclusion has not be reached. The aim of this systematic review and meta-analysis was to evaluate the prognostic significance of CD44v9 expression in various cancers.
METHODS: Three common databases were searched and retrieved studies were assessed using the inclusion and exclusion criteria. The further analyses for overall survival (OS), recurrence-free survival (RFS), and clinicopathological parameters were performed.
RESULTS: Fifteen studies containing 1633 cancer patients were included into this research. Patients with positive CD44v9 expression tended to have shorter OS (hazard ratio [HR] = 1.93, 95% confidence interval [CI] = 1.48-2.52, P < .01) and RFS (HR = 3.60, 95% CI = 1.52-8.53, P < .01) when compared with patients with negative CD44v9 expression. Positive CD44v9 expression was associated with larger tumor size (P = .04), deeper tumor invasion (P < .01), earlier lymph node metastasis (P < .01), and more advanced clinical stage (P < .01) when compared with negative CD44v9 expression.
CONCLUSION: Positive CD44v9 expression predicted worse prognosis in human cancers compared with negative CD44v9 expression. CD44v9 expression could serve as a prognostic factor of human cancers.

Introduction

Despite great improvement of diagnosis and therapies in recent years, cancer remains a major public health problem worldwide.[ A great number of new cancer cases and cancer-related deaths occur annually.[ In view of the poor prognosis of cancer cases, especially cases at advanced clinical stage, researchers begin to seek novel cancer biomarkers to predict the prognosis and improve the decision-making on therapies.[ However, most cancer biomarkers are not satisfactory up to now.[ Therefore, it is urgently needed to seek new biomarkers for human cancers.

CD44 is a family of transmembrane glycoprotein receptors, which bind to hyaluronic acid and link with multiple cellular functions, such as cell adhesion, migration, and invasion.[ CD44 has several isoforms generated through the alternative splicing of 10 variant exons.[ Among variant isoforms, CD44 standard and CD44 variant-6 expression have been proved to be associated with the prognosis of several cancers.[ Recently, increasing evidence showed CD44 variant-9 (CD44v9) might play an important role in the cancer progression, however, definite conclusion has not been reached on account of the contradictory results.[ Therefore, this systematic review and meta-analysis aimed to summarize the current evidence to evaluate the prognostic value of CD44v9 expression in various cancers.

Materials and methods

Literature search and selection

This study has been approved by the institutional review board of our hospital. PubMed, Embase, and Web of Science were comprehensively searched up to July 13, 2018. The subject terms and literature strategy were as follows: (“prognosis” OR “prognostic” OR “progression” OR “survival”) AND (“cancer” OR “tumor” OR “neoplasm”) AND (“CD44 variant 9” OR “CD44v9”). There was no restriction on the language. All relevant studies were further evaluated using inclusion and exclusion criteria by 2 investigators independently, and any dispute was solved by discussing with the third investigator.

Inclusion and exclusion criteria

The study would be included into this systematic review and meta-analysis if it met the following inclusion criteria: patients were diagnosed as cancers; patients were divided into positive or high CD44v9 expression group and negative or low CD44v9 expression group; study design was retrospective study or randomized controlled trial; study provided overall survival (OS), disease-specific survival (DSS), cancer-specific survival (CSS), recurrence-free survival (RFS), distant-metastasis-free survival (DMFS), progression-free survival (PFS), clinicopathological parameters or other prognostic variables; study had sufficient data to be extracted. Following studies would be excluded: duplicated patients, reviews, comments, letters, case reports, conference abstracts, unpublished articles, animal or cell experiments.

Data extraction and quality assessment

For each eligible study, following information was extracted: first author, publication year, country, sample size, sex, type of cancer, clinical stage, CD44v9 expression of patients, detection methods of CD44v9 expression, clinical outcomes and analysis model of OS. As for prognostic variables, such as OS, RFS, PFS and DMFS, hazard ratio (HR), and corresponding 95% confidence interval (CI) were directly obtained from original studies. While, these data could also be obtained from survival curves in original studies if these data were not directly provided.[ Newcastle–Ottawa scale (NOS) was utilized to evaluate the quality of included studies.[ Studies were considered to be with high quality when NOS score was ≥6. Data extraction and quality assessment were completed by 2 investigators independently, and any dispute was solved by discussing with the third investigator.

Statistical analysis

Statistical analysis was performed by Review Manager 5.3 (The Cochrane Collaboration, Copenhagen, Denmark) and Stata 12.0 software (Stata, College Station, TX). HR and corresponding 95% CI were used to detect the relationship between CD44v9 expression and prognostic variables. Odds ratio (OR) and 95% CI were pooled to determine the association between CD44v9 expression and clinicopathological parameters. Chi-square test and I2 statistic were utilized to assess the inter-study heterogeneity. A fixed-effect model would be used when heterogeneity was not significant (I2 ≤ 50% or Pheterogeneity > .10). If not, a random-effect model would be applied (I2 > 50% or Pheterogeneity > .10). Forest plot was generated to show the overall effects. Funnel plot and Begg test were conducted to evaluate the publication bias. Sensitivity analysis was performed to check the robustness of pooled results.

Results

The follow chart of literature search and selection was shown in Fig. 1. A total of 641 articles were initially retrieved from 3 common databases. After removal of duplicates, 415 articles remained for further evaluation. Then, 388 articles were directly excluded by scanning titles or abstracts. After this, 27 articles were assessed by evaluating full-texts and 12 articles were excluded for the following reasons: 6 papers for irrelevant to this topic, 3 papers for reviews, and 3 papers for without sufficient data. At last, 15 studies were finally included into this systematic review and meta-analysis.[

Figure 1

Follow chart of literature search and selection.

Characteristics of included studies

Fifteen studies containing 1633 cancer patients were included into this meta-analysis[ (Table 1). There were 755 patients in positive CD44v9 expression group and 848 patients in negative CD44v9 expression group. Regarding to countries, 10 studies were conducted in Japan,[ 2 studies in China,[ 1 study in Austria,[ 1 study in Korea,[ and 1 study in India.[ With respect to types of cancer, 11 types of cancer were investigated, including head and neck cancer,[ multiple myeloma,[ gastric cancer,[ upper tract urothelial cancer,[ hepatocellular carcinoma,[ colorectal cancer,[ bladder cancer,[ pancreatic cancer,[ gallbladder cancer,[ breast cancer,[ and laryngeal cancer.[ Besides, 11 studies reported the clinical stage of patients,[ while 4 studies did not report the information about the clinical stage.[ As for detection methods, the expression level of CD44v9 was evaluated using immunohistochemistry in 11 studies[ and using other methods in 4 studies.[ As for clinical outcomes, 12 studies reported clinicopathological parameters,[ 6 studies reported OS,[ 2 studies reported DSS,[ 2 studies reported CSS,[ 4 studies reported RFS,[ 1 study reported PFS[ and 1 study reported DMFS.[ Additionally, regarding to the analysis model of survival, 7 studies used the multivariate analysis model[ and 3 studies used the univariate analysis model.[ NOS score was ≥6 in all studies, which suggested that all studies were with relatively high quality.[

Table 1

Characteristics of studies included in this meta-analysis.

Prevalence of positive CD44v9 expression

As shown in Fig. 2, all studies were included into the analysis for the prevalence of positive CD44v9 expression.[ A random-effect model was utilized on account of the obvious heterogeneity among included studies (I2 = 91%, Pheterogeneity < 0.01). The prevalence of positive CD44v9 expression in human cancers was 0.45 (95% CI, 0.38–0.53).

Figure 2

Meta-analysis for the prevalence of positive CD44v9 expression. CD44v9 = CD44 variant-9.

Association between CD44v9 expression and OS

Six studies reported OS,[ 2 studies reported DSS,[ and 2 studies reported CSS,[ and all of them were pooled to assess the association between CD44v9 expression and OS (Fig. 3). There was no heterogeneity among included studies (I2 = 0%, Pheterogeneity = .44), as a result, a fixed-effect model was used. The results showed patients with positive CD44v9 expression tended to have significantly shorter OS compared with patients with negative CD44v9 expression (HR = 1.93, 95% CI, 1.48–2.52, P < .01).

Figure 3

Meta-analysis for the association between CD44v9 expression and OS. CD44v9 = CD44 variant-9, OS = overall survival.

To comprehensively assess the relationship between CD44v9 expression and OS in human cancers, we conducted the subgroup analyses (Table 2). The significant association between positive CD44v9 expression and shorter OS remained in the majority of subgroup analyses (P < .05) except for the subgroup analysis for univariate analysis model (P = .06).

Table 2

Subgroup analyses for the association between CD44v9 expression and OS.

Association between CD44v9 expression and RFS, PFS, or DMFS

As shown in Fig. 4, 4 studies reported RFS,[ 1 study reported PFS,[ and 1 study reported DMFS.[ Compared with negative CD44v9 expression, positive CD44v9 expression was obviously associated with shorter RFS (HR = 3.60, 95% CI, 1.52–8.53, P < .01; I2 = 82%, Pheterogeneity < .01) and PFS (HR = 3.35, 95% CI, 1.08–10.40, P = .04; I2 = 76%, Pheterogeneity = .04). However, no distinct relationship between CD44v9 expression and DMFS was observed (HR = 2.23, 95% CI, 0.22–22.60, P = .50).

Figure 4

Meta-analysis for the association between CD44v9 expression and RFS, PFS, or DMFS. CD44v9 = CD44 variant-9, DMFS = distant-metastasis-free survival, PFS = progression-free survival, RFS = recurrence-free survival.

Association between CD44v9 expression and clinicopathological parameters

As listed in Table 3, there was no significant relationship between CD44v9 expression and age (P = .40), sex (P = .46), or tumor differentiation (P = .09). However, positive CD44v9 expression was obviously associated with larger tumor size (P = .04), deeper tumor invasion (P < .01), earlier lymph node metastasis (P < .01), and more advanced clinical stage (P < .01) compared with negative CD44v9 expression in human cancers.

Table 3

Meta-analysis for the association between CD44v9 expression and clinicopathological parameters.

Publication bias and sensitivity analysis

Begg test showed significant publication bias regarding to the prevalence of positive CD44v9 expression (P = .03) (Fig. 5). No publication bias was observed in the association between CD44v9 expression and OS (P = .53) (Fig. 6). Funnel plots indicated there was no distinct publication bias in terms of RFS (Fig. 7A), age (Fig. 7B), sex (Fig. 7C), tumor differentiation (Fig. 7D), tumor size (Fig. 7E), depth of invasion (Fig. 7F), lymph node metastasis (Fig. 7G), and clinical stage (Fig. 7H). Sensitivity analysis for the association between CD44v9 expression and OS confirmed the robustness of results (Fig. 8).

Figure 5

Begg test for the prevalence of positive CD44v9 expression. CD44v9 = CD44 variant-9.

Figure 6

Begg test for the prevalence of association between CD44v9 expression and OS. CD44v9 = CD44 variant-9, OS = overall survival.

Figure 7

Funnel plots for the RFS and clinicopathological parameters (A, RFS; B, age; C, sex; D, tumor differentiation; E, tumor size; F, depth of invasion; G, lymph node metastasis; H, clinical stage). RFS = recurrence-free survival.

Figure 8

Sensitivity analysis for the association between CD44v9 expression and OS. CD44v9 = CD44 variant-9, OS = overall survival.

Discussion

Although plenty of publications demonstrated the potential prognostic significance of CD44v9 expression in human cancers, definite conclusion has not been obtained because of controversial results.[ In our study, we discovered the prevalence of positive CD44v9 expression was 0.45 in human cancers. More importantly, we found positive CD44v9 expression is an unfavorable factor for OS and RFS in various cancers. Positive CD44v9 expression was also related to larger tumor size, deeper tumor invasion, earlier lymph node metastasis, and more advanced clinical stage compared with negative CD44v9 expression. Therefore, positive CD44v9 expression might predict worse prognosis compared with negative CD44v9 expression in human cancers. Many publications have explored the underlying mechanisms of prognostic significance of CD44v9 expression in cancers.[ Yamakawa et al[ found CD44v9 might enhance pentose phosphate pathway flux and maintain glutathione levels in gastric cancer cell. Kobayashi et al[ discovered CD44v9 promoted the tumor invasion and induced the worse prognosis via epithelial-mesenchymal transition in bladder cancer. Kakehashi et al[ study indicated CD44v9 was a potential biomarker of tumor-initiating stem-like cells in liver cancer patients related to Nrf2-mediated resistance to oxidative stress. Maruyama et al[ observed positive CD44v9 expression was associated with lower Ki-67 positivity and cleaved caspase 3 positivity in duodenal cancer, and they also detected reduced mitotic activity in CD44v9 positive cells.

There were several highlights in our study. First, to our knowledge, this study was the first meta-analysis to determine the prognostic value of CD44v9 expression in cancers. Second, the subgroup analysis for the association between CD44v9 expression and OS was performed, which provided comprehensive evidence on this topic. Third, heterogeneity was small in the majority of analyses in our study, which guaranteed the accuracy of results. Fourth, a total of 1633 patients were included into this study, which could provide a relatively convincing conclusion. Nonetheless, our study was not without limitations. First, similar to other meta-analyses,[ the detection methods of CD44v9 expression level varied a lot among different studies, which might add the bias to the results. Second, although we did not set up the restriction on countries when conducting the literature search, most of included studies were performed in Asia, as a result, the conclusion of our study was hard to be generalized into other continents. Third, several factors might affect the prognosis of cancers, such as treatments, combined diseases, and response to chemotherapy, however, individual's information was unavailable for us, which was an inherent shortcoming for all meta-analyses. Fourth, the heterogeneity was large (e.g., meta-analysis of prevalence of positive CD44v9 expression) and sample size was small (e.g., meta-analysis of DMFS) in some analyses, which might reduce the accuracy of results. Fifth, previous researchers and editors tended to publish positive results of CD44v9 expression in human cancers, as a result, this selection bias would undoubtedly affect the results of our meta-analysis based on the previous studies.

Conclusion

Positive CD44v9 expression was associated with shorter OS, shorter RFS, larger tumor size, deeper tumor invasion, earlier lymph node metastasis, and more advanced clinical stage compared with negative CD44v9 expression in cancers. Therefore, CD44v9 expression could serve as a novel prognostic factor of human cancers.

Author contributions

Data curation: Li Zeng, Yitian Chen, Chengwei Tang.

Formal analysis: Li Zeng.

Investigation: Li Zeng, Chengwei Tang.

Methodology: Li Zeng, Yitian Chen, Chengwei Tang.

Resources: Li Zeng, Chengwei Tang.

Software: Li Zeng, Chengwei Tang.

Supervision: Chengwei Tang.

Validation: Yitian Chen, Chengwei Tang.

Writing – original draft: Li Zeng, Yitian Chen, Chengwei Tang.

Writing – review & editing: Li Zeng, Yitian Chen, Chengwei Tang.