Prognostic Value of Circulating Tumour Cells detected with the CellSearch System in Esophageal Cancer Patients: A Systematic Review and Meta-Analysis

Background: Esophageal carcinoma (EC) is the seventh-most prevalent tumor in the world, which is still one of the primary causes of tumor-related death. Identifying noteworthy biomarkers for EC is particularly signicant in guiding effective treatment. Recently, circulating tumor cells (CTCs) in peripheral blood (PB) were intensively discussed as prognostic markers in patients with EC. However, an ongoing controversy still exists regarding the prognostic signicance of CTCs determined by the CellSearch system in EC sufferers. This meta-analysis was designed to approach this topic. Methods: We systematically conducted searches using PubMed, Medline, Web of Science and the Cochrane Library for relevant studies, which were published through February 20, 2020. Using the random-effects model, our study was performed in Review Manager software, with odds ratios (ORs), risk ratios (RRs), hazard ratios (HRs) and 95% condence intervals (CIs) as the effect values. Results: Totally 7 articles were nally included in this study. For clinicopathological characteristics, the pooled results on TNM stage indicated that the III/IV group had higher rate of CTCs compared with the I/II group (OR=1.36, 95% CI: 0.68-2.71, I 2 =0%). Incidence of CTCs was higher in patients with T3/T4 stage (OR=2.92, 95% CI: 1.31-6.51, I 2 =0%) and distant metastasis group (OR=5.18, 95% CI: 2.38-11.25, I 2 =0%) compared to patients with T1/T2 stage or non-metastatic group. The pooled analysis revealed that CTC positivity detected in EC patients was correlated with poor overall survival (OS) (HR=2.83, 95% CI:1.99-4.03, I 2 =0%) and relapse-free survival (RFS) (HR=4.71, 95% CI:2.73-8.13, I 2 =0%). When pooling the estimated RR, a poor therapeutic response to chemoradiotherapy was discovered in patients with CTC positivity (RR=1.99, transition EpCAM CTCs using immunomagnetic ways deplete EpCAM high cells, favorable for the correlation between EpCAM low CTCs and clinical outcomes of patients (36-38). Results from a pilot study in patients with metastatic lung cancer did not indicate any signicant association between the incidence of EpCAM low CTCs and overall survival (OS). Similar results were found in a research including 97 metastatic non-small-cell lung cancer patients. In other types of cancers such as prostate cancer and breast cancer, the incidence of ≥ 5 EpCAM low CTCs was not signicantly associated with prognosis of patients, contrary to the presence of ≥ 5 EpCAM high . Additionally, previous study also indicated that EpCAM high CTCs from colorectal cancer approximately account for 89% (39), which indicates the numbers of CTCs detected with CellSearch system were more than the missed EpCAM low CTCs, and correspondingly the signicance of EpCAM high CTCs were higher. Collectively, these studies all suggested that although CTCs with a mesenchymal phenotype may not be detected by the CellSearch system, obvious signicance of EpCAM low CTCs in predicating prognosis in cancer patients was not indicated. and RFS in most subgroups. When pooling the HR for OS, CTC-positive status detected in EC patients was correlated with poor OS (HR=2.83, 95% CI: 1.99-4.03, I 2 =0%) and DFS (HR=4.71, 95% CI: 2.73-8.13, I 2 =0%). Patients with CTC positivity have a worse prognosis than those with CTC negativity. Moon DH suggested that although CTCs detected with the CellSearch system are an independent prognostic marker, it remains to be elucidated whether they can be considered a predictive marker for therapy (40). However, Riethdorf S indicated that the dynamic monitoring of CTCs with the CellSearch system might help to predict therapeutic ecacy in cancer (41). Then, we extracted data and analysed the DCR of chemotherapy in patients with EC, and the pooled analysis demonstrated that the DCR of the CTC positivity was lower than that of the CTC negativity (RR=1.99, 95% CI:1.73-2.29, I 2 =60%). Because CTCs could more likely escape from the primary tumour and enter into peripheral blood when the biological control by the primary tumour was not functioning and internal milieu altered, tumour recurrence after surgical treatment was more likely to appear in CTC-positive patients (42). Therefore, CTC detection can be regarded as an effective evaluation tool for assessing chemoradiotherapy ecacy and monitoring tumor recurrence in many solid tumors (43-45), including EC. Moreover, our meta-analysis demonstrated that CTC positivity remarkably with TNM staging, pT category, and distant metastasis. EC patients with stage III-IV have higher CTCs-incidence I-II the correlation between the incidence of CTCs and clinical stage only in three 221 40 CTCs-positive.

Considering the current controversies regarding the signi cance of the CellSearch system-detected CTCs in prognosis of EC patients, in our study, we systematically analysed data obtained in published literatures and summed up the potential clinicopathological and prognostic signi cance of the CellSearch system-detected CTCs in EC patients.

Search Strategy
We systematically searched PubMed, Medline, Web of Science and the Cochrane Library for relevant studies, which published through February 20, 2020. The following key words were used: "Circulating tumor cells", "CTCs", "CellSearch system" and "esophageal cancer". We

Eligibility Criteria and Quality Assessment
To be included in the meta-analysis, articles were selected based on the following criteria: (i) the articles only using the CellSearch system to detect CTCs and investigate the prognostic signi cance of CTC in EC patients; (ii) the article reported at least one noteworthy outcome indicator of CTCs, or the outcome could be calculated, based on data extracted from the published data; and (iii) the samples were collected from peripheral blood. Articles were excluded based on the following criteria: (i) the article was published in languages other than English; (ii) the number of EC patients and samples was less than ten; (iii) samples were collected from lymph nodes, bone marrow, or the abdominal cavity; (iv) non-human experiments; (v) reviews, case reports, comments, letters, and meeting records; (vi) EC and CTCs were not studied; and (vii) unable to obtain enough data through article reports and data calculations.
We evaluated the quality of the included literature with the Newcastle-Ottawa Scale(NOS), recommended by the Cochrane Library (19), according to three categories: (i) study group selection; (ii) comparability of groups; and (iii) outcome of interest. The full score was 9, and 1-4 points indicated low-quality, while 5-9 points were considered high-quality.

Data Extraction
Two reviewers independently used a standardized form to extract the data from the included studies: rst author's name, publication year, country of patients, characteristics of patients (number, sex, age), sampling time, detection markers, detection rate, histology, prognostic value, hazard ratio (HR) and disease control rate (DCR) of chemotherapy, and any disputes or differences were settled by a third independent investigator. For studies with multiple arms, each arm was considered an independent data set. The tumors DCR were evaluated in accordance with the Response Evaluation Criteria in Solid Tumors (RECIST) guideline (20

Statistical Analysis
We used Review Manager software (RevMan, version 5.3, The Nordic Cochrane Centre, The Cochrane Collaboration, London, UK) to analyze the data in our meta-analysis. The estimated odds ratios (ORs) from the included studies were used to assess the association between CTC detection and different clinicopathologic features of EC. To statistically assess the prognostic effects of CTCs, we extracted the HR and 95% con dence interval (CI) of overall survival (OS) and relapse-free survival (RFS) from the included studies. If HRs, 95% CIs, or P-values were not directly provided in the original literature, the estimated HR was used to assess prognostic effects based on the method described by Tierney et al (21), and HR>1 re ects further disease progression or more deaths in the patients with CTC positivity. Furthermore, the estimated risk ratio (RR) was calculated to assess the DCR. We pooled the extracted HRs together in Review Manager. All statistical values were combined with 95% CIs, and all P values were two sided whose threshold was considered statistically signi cant when it was less than 0.05. Heterogeneity among the studies was tested using Cochran Q test and I 2 statistic. Signi cant heterogeneity was considered when P ≤0.1 or I 2 ≥50% (22), and in these cases, a random-effects model was used. Simultaneously, according to the differences in the data retrieved, subgroup analyses were performed, such as for the age of patients, sex of patients, histology, and clinicopathological signi cance.
Publication bias was evaluated using a funnel plot.

Study Characteristics
The initial search yielded 32 records in PubMed, Medline, Web of Science and the Cochrane Library. Of these, 16 duplicate studies were excluded. We excluded 6 records after reading the titles and abstracts. After reviewing the full texts, 7 articles were nally included in this study (11,12,(23)(24)(25)(26)(27). The selection owchart of this study is shown in Figure 1.

Quality Assessment
Among the 7 studies included: 2 studies (26,27) were of low quality and the other 5 studies (11,12,(23)(24)(25) were of high quality, evaluated based on the NOS (Table 2). Table 2 The assessment of the risk of bias in included studies using the Newcastle-Ottawa scale. adequacy of follow-up of cohorts (≥90%). "1" means that the study is satisfied the item and "0" means the opposite situation.

Diagnosis CTC detection and clinicopathological features
We extracted and analyzed clinicopathological variables from the included articles in our meta-analysis when they were mentioned in at least 3 studies. The results of the pooled ORs of the parameters of EC patients, which were used to evaluated the potential correlations between the detection of CTCs and clinicopathological parameters, are summarized in Table 3. We extracted and analyzed eight clinicopathological features according to the criteria mentioned above.

CTC detection and DCR
Only 2 studies assessed the association between incidence of CTCs and DCR in patients receiving chemotherapy/chemoradiotherapy, and the overall response rate (ORR) was used to assess the response to chemoradiotherapy. When pooling the estimated RR, CTC-positive patients had a poor response to chemoradiotherapy compared with CTC-negative patients (RR=1.99, 95% CI: 1.73-2.29, I 2 =60%), as shown in Figure 3.

Subgroup analysis and publication bias
Regarding the heterogeneity of the pooled survival effects, there was no statistical signi cance in between-study heterogeneity for OS and RFS. We used funnel plots to detect publication bias, as shown in Figure 4. In all comparisons, shape of the funnel plots had a symmetrical distribution. Thus, no signi cant publication bias was found in the meta-analyses of OS and RFS.

Discussion
Although the advanced treatment have been widely adopted in EC patients recently, the presence of spreading and recrudescence of EC are still great challenges for both surgeons and patients (3,28). Due to late diagnosis and limited treatment options, most EC patients have a poor prognosis and high mortality. To prompt timely diagnosis and treatment, biomarkers to determine the recurrent or metastatic status of EC are in great request. Recently, CTCs, detached cells from a primary tumor in PB, have been increasingly investigated for their prognostic value in many tumors. As described in the "seed and soil" theory (29), CTCs are regarded as critical factors for tumors metastasis (30). As demonstrated in many studies, high CTCs was associated with the poor survival in many solid tumours, such as breast cancer (31), bladder cancer(32), ovarian cancer and gastric cancer (33,34). In addition, due to its bene ts of time and cost saving, easy operation and higher speci city and reproducibility, CTC detection from PB can be regarded as an effective evaluation tool for monitoring and assessing treatment effects in EC patients. For EC, several previously published meta-analyses demonstrated the prognostic value of CTCs; however, the assays used to detect CTCs were predominately restricted to polymerase chain reaction (PCR) and immunocytochemistry (ICC) in the included studies (35). Interestingly, the clinical utility of CTC detection with the CellSearch system from the PB of EC patients has been demonstrated in several studies (11,12,(23)(24)(25)(26)(27). Thus, to quantitatively assess the clinical value of CTCs determined using the CellSearch system in EC patients is valuable. It is commonly acknowledged that CTCs detected using CellSearch system are EpCAM high , and EpCAM low CTCs might be missed due to epithelial-mesenchymal transition (EMT). Researchers have described a method to collect EpCAM low CTCs using immunomagnetic ways to deplete EpCAM high cells, which is favorable for investigating the correlation between EpCAM low CTCs and clinical outcomes of patients (36)(37)(38). Results from a pilot study in patients with metastatic lung cancer did not indicate any signi cant association between the incidence of EpCAM low CTCs and overall survival (OS). Similar results were found in a research including 97 metastatic non-small-cell lung cancer patients. In other types of cancers such as prostate cancer and breast cancer, the incidence of ≥5 EpCAM low CTCs was not signi cantly associated with prognosis of patients, contrary to the presence of ≥5 EpCAM high . Additionally, previous study also indicated that EpCAM high CTCs from colorectal cancer approximately account for 89% (39), which indicates the numbers of CTCs detected with CellSearch system were more than the missed EpCAM low CTCs, and correspondingly the signi cance of EpCAM high CTCs were higher. Collectively, these studies all suggested that although CTCs with a mesenchymal phenotype may not be detected by the CellSearch system, obvious signi cance of EpCAM low CTCs in predicating prognosis in cancer patients was not indicated.
As far as we know, this is the rst meta-analysis that focused on the signi cance of CTC determined only using the CellSearch system in EC. ). Patients with CTC positivity have a worse prognosis than those with CTC negativity. Moon DH suggested that although CTCs detected with the CellSearch system are an independent prognostic marker, it remains to be elucidated whether they can be considered a predictive marker for therapy (40). However, Riethdorf S indicated that the dynamic monitoring of CTCs with the CellSearch system might help to predict therapeutic e cacy in cancer (41). Then, we extracted data and analysed the DCR of chemotherapy in patients with EC, and the pooled analysis demonstrated that the DCR of the CTC positivity was lower than that of the CTC negativity (RR=1.99, 95% CI:1.73-2.29, ). Because CTCs could more likely escape from the primary tumour and enter into peripheral blood when the biological control by the primary tumour was not functioning and internal milieu altered, tumour recurrence after surgical treatment was more likely to appear in CTCpositive patients (42). Therefore, CTC detection can be regarded as an effective evaluation tool for assessing chemoradiotherapy e cacy and monitoring tumor recurrence in many solid tumors (43)(44)(45), including EC. Moreover, our meta-analysis demonstrated that CTC positivity was remarkably correlated with TNM staging, pT category, and distant metastasis. EC patients with stage III-IV have higher CTCs-incidence than patients with stage I-II (OR=1.36). However, the correlation between the incidence of CTCs and clinical stage was only discussed in three included articles with 221 patients and among them 40 are CTCs-positive. Besides, the studies by Woestemeier A (12) provided the limited data of patients with stage I-III. Therefore, although the results indicated there was no signi cance of this difference between patients with stage I-II and those with stage III-IV (P=0.38), with more patients and studies included in the future, the results might suggest signi cant difference between different clinical status and stages. Interestingly, the AC group had a notably higher incidence of CTCs compared with the SCC group (OR=1.86, 95% CI: 0.81-4.26, I 2 =0%) the results is not obviously signi cant (P=0.14), which is consistent with other studies (46). Besides, studies concerning the relative aggressive behavior of AC group and SCC group is rare. And it is relatively di cult to discuss the correlation between the higher incidence of CTCs and aggressive behavior in AC and SCC group, respectively. In summary, the pooled results indicate that the CTCs determined by the CellSearch system have important clinical value in assessing the prognosis of EC patients, guiding treatment decisions, and monitoring treatment effects. For CTC-positive patients, more early aggressive treatment and effective evaluation may be required.
The CellSearch system used for detection of CTCs has more advantages compared to ICC and PCR, including saving time and cost, easy operation and higher speci city and reproducibility for CTC enrichment. Since our meta-analysis of researches utilizing the CellSearch system for detection of CTCs decreased the heterogeneity caused by various detection assays, there was no statistical signi cance in between-study heterogeneity for OS and RFS. Therefore, the detection method is the main source of between-study heterogeneity. In all comparisons, shape of the funnel plots had a symmetrical distribution. Thus, no signi cant publication bias was found in the metaanalyses of OS and RFS.
In addition, clinical consensus still remained equivocal on the optimal cutoff value for predicting the prognosis of EC patients with CTCs. In our meta-analysis, both the cut-off value CTCs ≥1/7.5 ml and CTCs ≥2/7.5 ml seemed to indicate equivalent predicative value, suggesting these two cut-off values are both associated with poor prognosis. However, when we excluded the intra-therapy set of Tanaka et al., a signi cantly higher HR for OS was found with the cut-off value of CTCs ≥2/7.5 ml (HR=3.14, 95% CI:1.82-11.97) than with the cut-off value of CTCs ≥1/7.5 ml (HR =2.89, 95% CI: 1.93,4.31). Therefore, in EC patients, the cut-off value of CTCs ≥2/7.5 ml may be correlated with poorer prognosis than the cut-off value of CTCs ≥1/7.5 ml. Thus, high-quality, well-designed, large-scale multi-centre research is needed to identify the better cut-off value and more appropriate sampling time of CTC detection.
Several limitations remained in our study. First, due to several studies didn't report HRs, the estimated HR was used to assess prognostic effects based on the method described by Tierney et al (21). Second, we used extracted data rather than raw data from individual patients, and we could not correct all clinicopathological parameters according to a consistent standard. Third, we limited our analysis to studies published in English, so the choice of language brings another bias. Fourth, the total amount of patients was relatively small in the metaanalysis. Fourth, there are low patients' number and no multicenter controlled trials in our meta-analysis. Fifth, with the limited data in the included articles, the data considering clinical pathological characteristics and prognosis of AC and SCC group patients were not available separately. Despite these limitations, we still demonstrated that CTC positivity determined using CellSearch system was an indicator of poor prognosis in patients with EC.

Conclusions
Sum up, our meta-analysis indicated that the presence of CTCs determined using the CellSearch system is correlated with the prognosis of EC patients and provided a scienti c foundation for EC staging. Additionally, subgroup analysis indicated that CTC positivity is more associated with a poorer prognosis than CTC negativity. Additionally, the CTCs determined using the CellSearch system can be regarded as an effective evaluation tool for assessing chemoradiotherapy e cacy and monitoring tumour recurrence for EC patients. However, highquality, well-designed, large-scale multi-centre research is needed to verify our results and con rm the clinical value of CTCs determined using the CellSearch system in EC patients.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.   Risk ratio (RR) for DCR.