Clinicopathological and prognostic significance of circulating tumor cells in patients with lung cancer: a meta-analysis

Background The prognostic significance of circulating tumor cells in patients with lung cancer is controversial. Therefore, we aimed to comprehensively and quantitatively assess the prognostic role of CTCs in patients with lung cancer. Methods The relevant literature was searched using PubMed, the Cochrane database and the China National Knowledge Internet database (up to June 2016). Using Review Manager 5.1.2, a meta-analysis was performed using hazard ratio (HR), odds ratio (OR) and 95% confidence interval (CI) as effect values. Results Thirty studies comprising 2,060 patients with lung cancer were analyzed. The pooled HR values showed that circulating tumor cells were significantly correlated with overall survival (HR =2.63, 95% CI [2.04, 3.39]) and progression-free survival (HR =3.74, 95% CI [2.49, 5.61]) in these patients. Further subgroup analyses were conducted and categorized by sampling time, detection method, and histological type; these analyses showed the same trend. The pooled OR values showed that circulating tumor cells were associated with non small cell lung cancer stage(OR = 2.11, 95% CI [1.42, 3.14]), small cell lung cancer stage (OR = 10.91, 95% CI [4.10, 29.06]), distant metastasis (OR =7.06, 95%CI [2.82, 17.66]), lymph node metastasis (OR =2.31, 95% CI [1.19,4.46]), and performance status(OR =0.42, 95%CI [0.22, 0.78]). Conclusion The detection of circulating tumor cells in the peripheral blood of patients with lung cancer can be indicative of a poor prognosis.


INTRODUCTION
Lung cancer is one of the deadliest diseases in the world. Less than 15% of lung cancer patients survive for more than 5 years after being diagnosed [1]. Due to its aggressive behavior and greater invasive ability than other types of cancer, the predominant cause of treatment failure in patients with lung cancer is believed to be distant metastases, even during early-stage disease. Approximately 25% to 50% of patients with earlystage non small cell lung cancer (NSCLC) show tumor recurrence, even after tumor resection [2,3]. However, current staging methods are unable to detect such occult metastases prior to the emergence of clinical manifestations [4]. Thus, there is an urgent need for moresensitive prognostic and predictive markers.
Circulating tumor cells (CTCs) can be found in the peripheral blood of patients with cancer. Many studies have demonstrated the potential usefulness of CTCs in predicting patient prognosis for several cancer types [5][6][7]. Many studies have also shown associations between CTCs and poor survival in lung cancer [8][9][10][11]. However, the prognostic significance of CTCs in lung cancer remains controversial, as other studies have failed to show an association between CTCs and poor prognosis [12]. In addition, assessing the potential of using CTCs as a prognostic marker has been complicated by interstudy differences in aspects such as study population, methodology and sampling time. Thus, our meta-analysis aimed to examine the association of CTCs with survival and clinicopathological parameters, and to evaluate the prognostic role of CTCs in patients with lung cancer.

Characteristics of the included studies
After initial literature searches, 153 articles were retrieved, and 4 duplicate articles were excluded. After screening the titles and abstracts, 67 studies remained, and their full texts were assessed for eligibility. Of the eligible studies, 37 studies were excluded because they lacked an outcome of interest. Ultimately, 30 studies were selected for analysis; these comprised 24 studies published in English and 6 studies published in Chinese ( Figure 1).

The prognostic effect (OS and PFS) of CTC detection
The pooled HR values showed a significant correlation between CTCs and OS in patients with lung cancer (HR =2.63, 95% confidence interval (CI) [2.04, 3.39], P<0.00001, I 2 =19%) ( Figure 2). Subsequently, subgroup analyses were conducted after categorization by sampling time, detection method, and histological type to further investigate the prognostic role of CTCs. We found a significant correlation between CTCs and OS in the NSCLC (HR =2.55, 95% CI [1.65, 3.93], Figure 1: A flow chart of the study design. A systematic literature search yielded a total of 153 articles related to the relationship between CTCs and lung cancer. After the screening of titles, abstracts and full texts, 123 articles were excluded for reasons detailed in the main text. A meta-analysis was then performed on 30 studies to assess the clinicopathological and prognostic significance of CTCs in patients with lung cancer.

Correlation between CTCs and clinicopathological parameters
The pooled odds ratio (OR) values showed that there was a significant correlation between CTCs and tumor stage in patients with lung cancer. As shown in Table 3, the incidence of CTC detection in patients with stage III/IV was higher than that in patients with stage I/ II NSCLC (OR = 2.11, 95% CI [1.42,3.14], P=0.0002, I 2 = 20%). Similarly, the incidence of CTC detection in extensive SCLC was higher than that in limited SCLC (OR = 10.91, 95% CI [4.10, 29.06], P<0.00001, I 2 = 4%). However, the subgroup analyses of studies using the RT-PCR method showed no significant correlation between CTCs and tumor stage in either NSCLC patients or SCLC patients.
We found that the presence of CTCs was significantly increased in lung cancer patients with distant metastasis (OR =7.06, 95%CI [2.82, 17.66], P<0.0001, I 2 = 46%). Further subgroup analyses conducted and categorized by sampling time, detection method, and histological type showed the same trend. The presence of CTCs was also significantly increased in lung cancer patients with lymph node metastasis (OR = 2.31, 95%CI [1.19, 4.46], P=0.01, I 2 = 19%), but the subgroup analyses showed a significant correlation between CTCs and lymph node metastasis only in the pretreatment subgroup. Moreover, all the studies included in this analysis pertained to NSCLC. We also found that CTCs were associated with performance status (OR = 0.42, 95%CI [0.22, 0.78], P=0.006, I 2 = 0%). Lower performance scores corresponded to lower CTC incidence. However, the subgroup analyses showed no significant correlation in the post-treatment, PCR or NSCLC subgroups.
Furthermore, pooled analyses of tumor size, performance status, smoking status, and patient age revealed no significant correlation between these clinicopathological parameters and CTCs.

Test of heterogeneity
Except for the 'non-PCR on distant metastasis' subgroup (I 2 = 70%) and the 'PCR on lymph node metastasis' subgroup (I 2 = 58%), the heterogeneity among all the included studies was not significant. However, when

Sensitivity analyses
We performed sensitivity analyses to test the robustness of the pooled results. The pooled HR was not significantly altered when any individual study was removed. Moreover, the pooled OR was not significantly influenced when any individual study was removed, with the exception of lymph node metastasis. The pooled OR of lymph node metastasis  Chen TF  1  1  1  1  0  0  1  1  1  7   Hiltermann TJ  1  1  1  1  0  0  1  1  1  7 Sher YP was significantly altered by removal of the study [17] that was the source of heterogeneity.

Publication bias
As shown in Figure 4, funnel plots showed no evidence of publication bias. In addition, Egger's and Begg's tests were examined to detect publication bias in our article. The results of both Egger's and Begg's tests showed no evidence of publication bias (P>0.05).

DISCUSSION
Although chemoradiotherapy and surgery have been widely used, lung cancer metastasis and recurrence frequently occur. The poor overall survival of patients and the complex heterogeneity of the disease are significant challenges for therapeutic intervention. Therefore, biomarkers that can be used to identify lung cancer recurrence or metastasis are needed to facilitate timely diagnosis and effective treatment strategies for lung cancer patients. CTCs, which are released by primary tumors or metastatic tumors, have been recognized as the cause of tumor metastasis or recurrence [38,39]. However, the clinicopathological and prognostic significance of CTC detection in patients with lung cancer is not clear. In this meta-analysis, we provide strong evidence that CTCs are significantly associated with poor OS and PFS in lung cancer patients, irrespective of sampling time, detection method, and histological type. All the pooled HRs were above 2.0 in our study. These results demonstrate that a CTC-high status indicates poor prognosis in lung cancer patients; these patients may need more-aggressive treatment that is assessed frequently and closely monitored.
According to the pooled ORs in our meta-analysis, CTCs were associated with tumor stage, lymph node metastasis, distant metastasis, and performance status in patients with lung cancer. The results indicated that CTCs can be predictors of disease progression, and may be used to estimate the degree of malignancy and metastatic ability in lung cancer. However, the analysis of lymph node metastasis showed that the correlation occurred only in the pretreatment subgroup. It is generally believed that lymph node metastasis occurs prior to blood-borne metastasis, but the detection of CTCs in patients with early tumors indicates that blood-borne metastasis can occur before lymph node metastasis. In one study [24], the incidence of CTC detection was higher in patients with lymph node metastasis than in those free of lymph node metastasis. However, in other studies [17,19,26,29], the incidence of CTC detection was not correlated with lymph node metastasis. Thus, the correlation between CTCs and lymph node metastasis may require further investigation.

(Continued)
We analyzed studies reporting the detection of CTCs in peripheral blood before and after treatment. The results from these two sampling time were consistent, except for the correlation of CTCs with lymph node metastasis and performance status. Therefore, CTC detection may offer doctors a simpler, less-invasive method that can be used at an earlier stage of disease (relative to other methods) to estimate disease progression and predict the prognosis of patients before treatment. In recent years, various new CTC assay methodologies have been developed, including RT-PCR, immunocytochemistry, and the CellSearch System, for example. Each method has its advantages and disadvantages. We obtained different results for the PCR and non-PCR subgroups in analyses of tumor stage, lymph node metastasis, and performance status. CTCs were not associated with these clinicopathological parameters in the PCR subgroup. It thus seems that non-PCR-based methods are best for CTC detection in this context. Several studies have been performed to compare CTC detection methods, but no conclusive results have been obtained as of yet [40]. Therefore, further studies within the same lung cancer patient populations are needed to provide comparative data on the clinical significance of CTCs detected by different methods.

Clinicopathological parameters
We found significant heterogeneity in the non-PCR subgroup on distant metastasis and in the PCR subgroup on lymph node metastasis. In these two subgroups, CTCs were detected before treatment using the same detection methods. However, the optimal cut-off values for CTC detection were obviously different for the two subgroups. In addition, the markers of CTC detection were not uniform in the PCR-subgroup studies that investigated the association between CTCs and lymph node metastasis. We propose that these two factors might be the principal causes of heterogeneity.
This study has some notable limitations. First, our meta-analysis was limited to the published scientific literature, and univariate data were also included in the present meta-analysis because multivariate survival analysis data were not available. Second, the CTC detection assays varied in our study, and included different endpoints, cut-off values, and experimental designs. Moreover, we excluded some papers that did not calculate OS and PFS, which may have influenced the results to some degree [18,25,41].
In conclusion, this meta-analysis indicates that the detection of CTCs in peripheral blood may be an indicator of patient prognosis, and provides evidence that CTC detection can be used to estimate the degree of malignancy and metastatic ability in patients with lung cancer. In the future, high-quality, well-designed and large-scale multicenter studies are needed to further substantiate these findings.

Literature search
PubMed, the Cochrane database and the China National Knowledge Internet database were searched for studies pertaining to the clinicopathological and prognostic relationship between CTCs and lung cancer without language, publication or time restrictions (up to June 2016). The main search terms were "lung or pulmonary or pulmonic or pneumonic or pneumal" and "cancer or tumor or tumour or carcinoma or neoplasm(s)" and "CTC(s) or circulating tumor cell(s) or circulating cancer cell(s) or circulating epithelial cell(s) or micrometastasis". Furthermore, relevant articles were identified from references cited in the retrieved articles and in review articles by manual searching.

Selection criteria
Eligible studies were included if they met the following criteria: (i) CTCs were detected in lung cancer patients; (ii) samples were collected from peripheral blood; and (iii) at least one of the outcome measures of interest was reported in the study or calculated from published data. When several studies were reported from the same authors or organizations, the meta-analysis included the most recent study (or the highest-quality study if the most recent study did not fit the inclusion criteria).
Studies were excluded if they met any of the following criteria: (i) the number of patients with lung cancer was fewer than 20; (ii) repeated studies were based on the same database or patients; or (iii) they provided insufficient data.

Data extraction and assessment of study quality
Two independent reviewers evaluated each study and extracted data independently, and any disagreements were resolved via discussion. We performed two types of analysis. The first type of analysis determined whether CTC status was associated with OS or PFS. The second type of analysis determined whether CTC status was correlated with clinicopathological parameters, which included tumor size, lymph node metastasis, distant metastasis, NSCLC stage(III/IV vs. I/II), SCLC stage (extensive disease vs. limited disease), gender, age, smoking and performance status. Data for multivariate survival analyses reported in the included articles were included in this meta-analysis. If these data were not available, then univariate analytical data were included. The quality of studies was evaluated according to the NOS [42], and studies with an NOS score≥ 5 were considered to be of high quality.

Statistical analysis
Statistical analysis was performed using Review Manager 5.1.2software. The estimated HR was used to evaluate the prognostic effect (OS and PFS), and the estimated OR was used to summarize the association between CTC detection and the clinicopathological characteristics of lung cancer. If the HR and its variance were not reported directly in the original study, then these values were calculated from the available reported data using software designed by Tierney et al. [43]. All statistical values were combined with a 95% CI, and the P-value threshold was set at 0.05. The random-effects mode was used to perform the analysis, as this model produced more conservative results than did the fixedeffects model, and it was a better fit for the multicenter clinical studies owing to the existence of heterogeneity [44]. Heterogeneity was calculated using a Q test, and the I 2 value represented the degree of heterogeneity. Publication bias was tested using a funnel plot, and by Egger's and Begg's tests, in Stata 12.0 software. The overall analysis was completed by evaluating all the relevant studies according to different clinicopathological parameters and prognostic outcomes. Further subgroup analyses were conducted and categorized by sampling time (pretreatment and post-treatment), detection method (PCR and non-PCR), and histological type (NSCLC and SCLC). Sensitivity analyses were performed by excluding one study at a time to evaluate the influence of single studies on summary effect values.

Abbreviations
CI, confidence intervals CTCs, circulating tumor cells HR, hazard ratio NOS, Newcastle-Ottawa scale NSCLC, non small cell lung cancer OR, odds ratio OS, overall survival PFS, progression-free survival RT-PCR, reverse-transcriptase polymerase chain reaction SCLC, small cell lung cancer