Study Characteristics
The original search yielded 2566 records in PubMed, Medline and Web of Science. Of these, 1087 duplicate studies were excluded. We excluded 1395 records after reading the titles and abstracts. After reviewing the full texts, 52 articles were finally included in this study [12-63]. The selection flowchart of this study is shown in Figure 1.
In total, 52 articles (ranging from 1997 to 2019) from Asia, Europe and South America (China, Japan, Korea, Taiwan, Iran, Brazil, Korea, Germany, Italy, Turkey, Poland and France) that reported 68 arms comprising 4158 GC patients were included (sample size median: 59(22–810), mean: 85) (Table 1). Of these 68 arms, 40 arms used the RT-PCR method and 8 arms used the CellSearch System. The sampling time was classified as the preoperative in 45 arms, the intraoperative in 10 arms as well as the postoperative in 9 arms. Meanwhile, four studies did not report the sampling time. Twenty-one articles[12, 16, 20, 21, 23, 24, 26, 29, 30, 32-34, 37-40, 42, 48, 50-52] with 24 arms provided HRs or HRs could be obtained by calculation for OS, PFS or DFS to perform the meta-analysis. Eighteen of the remaining 21 articles [16, 20, 21, 23, 24, 26, 29, 30, 32-34, 38, 40, 42, 48, 50-52] with 21 arms provided HRs for OS. Nine articles [12, 16, 20, 21, 23, 24, 29, 30, 40] with 11 arms provided HRs for PFS, and 3 articles [37, 39, 52] provided HRs for DFS.
Quality Assessment
The quality of the 52 included studies (13 studies in the low-quality categories and the other 39 studies in the high-quality categories) was evaluated on the basis of the NOS (Table 2).
Diagnosis
Diagnosis of CTCs with the parameters of patients
We analyzed the basic parameters of patients, TNM staging, histologic type and invasion in our meta-analysis. The results of the pooled analysis on the parameters of GC patients are summarized in Table 3.
The CTCs positivity rate according to the basic parameters of patients. The meta-analysis of all relevant researches on age indicated that the >60 age group had a remarkably higher CTCs positivity rate than the ≤60 age group (OR=1.48, 95% CI (1.13,1.92), I2=0%). Then, we performed subgroup analysis on sample time (preoperative: OR=1.45, 95% CI (1.00,2.11), I2=16%; intra/postoperative: OR=1.45, 95% CI (0.94,2.25), I2=0%). A similar trend was also discovered in the subgroup analysis of sex. The meta-analysis of relevant studies on sex indicated that the male group had a higher CTCs positivity rate than the female group without significant difference (OR=1.07, 95% CI (0.90,1.27), I2=0%). Then, we performed subgroup analysis on sample time (preoperative: OR=1.15, 95% CI (0.94,1.40), I2=0%; intra/postoperative: OR=0.87, 95% CI (0.62,1.22), I2=2%).
The CTCs positivity rate according to the TNM stage. To further analyze the correlation between the CTCs positivity rate and TNM stage of GC patients, we extracted the TNM stage data from the included studies. The pooled results on TNM stage indicated that the III/IV group had a notably higher CTCs positivity rate than the I/II group (OR=2.73, 95% CI (1.95,3.82), I2=65%). Then, subgroup analysis was performed on sample time (preoperative: OR=2.41, 95% CI (1.56,3.73), I2=71%; intra/postoperative: OR=3.36, 95% CI (2.34,4.82), I2=0%). In addition, the pT category, LN3 (lymph node metastasis), and pM (distant metastasis) of all studies were also separately analyzed. The pT3/T4 group had a notably higher CTCs positivity rate than the pT1/T2 group (overall: OR=2.20, 95% CI (1.64,2.95), I2=54%; preoperative: OR=1.39, 95% CI (1.05,1.83), I2=22%; intra/postoperative: OR=3.93, 95% CI (2.69,5.75), I2=17%). Simultaneously, there was a significant difference between the lymph node metastasis positive and negative groups (overall: OR=2.45, 95% CI (1.77,3.40), I2=55%; preoperative: OR=1.96, 95% CI (1.34,2.85), I2=60%; intra/postoperative: OR=3.42, 95%CI (2.08,5.60), I2=53%), as well as the distant metastasis positive and negative groups (overall: OR=1.97, 95% CI (1.43,2.72), I2=2%; preoperative: OR=1.85, 95% CI (1.31,2.61), I2=0%; intra/postoperative: OR=2.77, 95% CI (1.18,6.47), I2=16%).
TheCTCs positivity rate according to the histologic type. The degree of tumor differentiation and Lauren classification in GC can be indicated by the detection of CTCs. The degree of tumor differentiation was divided into three degrees: well, moderately, and poorly differentiated. And, Lauren classification was divided into three types: intestinal, diffuse, and hybrid type. The poorly differentiated group had a higher CTCs positivity rate than the well/moderately differentiated group (overall: OR=1.91, 95% CI (0.77,4.71), I2=85%; preoperative: OR=1.00, 95% CI (0.30,3.30), I2=86%; intra/postoperative: OR=5.88, 95% CI (2.25,15.35), I2=65%), as well as the Lauren classification diffuse/hybrid type group and intestinal group (overall: OR=1.77, 95% CI (0.70,4.44), I2=79%; preoperative: OR=0.97, 95% CI (0.56,1.70), I2=34%; intra/postoperative: OR=14.52, 95% CI (0.65,323.45), I2=87%), suggesting an higher CTCs positivity rate in the poorly differentiated group and Lauren classification diffuse/hybrid type group, but had no statistically significant.
TheCTCs positivity rate according to the invasion. The lymphatic invasion-positive group had a higher CTCs positivity rate than the lymphatic invasion-negative group (overall: OR=1.80, 95% CI (1.26,2.57), I2=38%; preoperative: OR=1.27, 95% CI (0.99,1.63), I2=0%; intra/postoperative: OR=3.29, 95% CI (1.97,5.47), I2=0%), vessel invasion-positive group (overall: OR=2.29, 95% CI (1.54,3.40), I2=61%; preoperative: OR=1.35, 95% CI (0.88,2.07), I2=37%; intra/postoperative: OR=3.41, 95% CI (2.20,5.29), I2=30%) and perineural invasion-positive group (overall: OR=3.32, 95%CI (1.86,5.93), I2=0%; preoperative: OR=2.45, 95%CI (1.16,5.19), I2=0%; intra/postoperative: OR=5.21, 95%CI (2.09,12.98), I2=30%) than in the negative group, indicating that tumors in GC patients who are CTC positive may be more likely to spread.
The survival of GC patients
To analyze the survival of GC patients, we extracted 21 studies that provided data for OS with 1,673 GC patients, 11 studies for PFS with 811 GC patients, and 3 studies for DFS with 280 GC patients. The pooled results on prognostic effects of CTCs detected in GC patients indicated that CTC-positive status detected in GC patients was associated with poor OS (HR =1.94, 95% CI (1.64,2.30), P≤0.001; I2=33%, Figure 2a), PFS (HR =2.45, 95% CI (1.65,3.64), P≤0.001; I2=71%, Figure 2b), and DFS (HR =2.78, 95% CI (1.89,4.10), P≤0.001; I2=0%, Figure 2c).
According to the difference in the variables (publication year, country, patients’ number, sampling time, detection methods, CTC positive rate, quality of the article), we performed subgroup analyses and the results is demonstrated in Table 4. In the included studies for OS, the median publication year was 2013, the median number of patients was 65 and the median positive rate of the patients was 46.2%, respectively. In the included studies for PFS, the median publication year was 2016, the median number of patients was 62 and the median positive rate of the patients was 42.9%, respectively. The summary analysis results demonstrated that CTC positivity had a significant impact on the prognosis of OS and PFS in most subgroups.
CTC detection and DCR
6 studies assessed the association between CTCs positivity and DCR in patients receiving chemotherapy/chemoradiotherapy. When pooling the estimated RR, CTC-positive patients had a poor response to chemoradiotherapy compared with CTC-negative patients (RR =0.63, 95% CI (0.44,0.91)), as shown in Figure 3.
Subgroup analysis and publication bias
Heterogeneity analysis showed that PFS in the pooled survival effects has significant heterogeneity (I2= 71%, P≤0.001). To explore potential sources of heterogeneity, we used subgroup analysis according to the difference in the variables including the publication year, country, patients’ number, sampling time, detection methods, positive rate, and quality of the article. Eventually, we concluded that the main source of the heterogeneity in PFS came from the intra/postoperative set of Matsusaka et al [40] and Pernot et al [20]. When we divided detection methods into PCR, cytological methods, and detected methods (eg. in situ hybridization assay, morphology-based enrichment coupled with RT-PCR, colorimetric membrane-array) subgroups, we found the interesting results. In our Meta-analysis, Compared with the PCR method (OR=1.51, 95% CI (1.37, 1.67)) (Figure 4), the CellSearchTM System (OR=1.38, 95% CI (1.29, 1.48)) had a lower CTCs positivity rate, but it’s ability to count CTCs number and correlated with the increased risk of death(OS: (RT-PCR: HR=1.57, 95%CI (1.28,1.92)), (CellSearchTM System: HR=2.02, 95%CI (1.63, 2.50))). It confirms that the CellSearchTM System is a detected method of high reproducibility, accuracy, linearity, and reliability. Non-CellSearchTM System cytological methods (OR=1.72, 95% CI (1.52, 1.96)) had a higher CTCs positivity rate than PR-PCR and CellSearchTM System. The other methods (OR=1.74, 95% CI (1.42, 2.14)) had a higher CTCs positivity rate than RT-PCR(OR=1.51, 95% CI (1.37, 1.67)), as well as its ability to count CTCs number and their correlation with the increased risk of death (OS: (RT-PCR: HR=1.57, 95%CI (1.28,1.92)), (Others: HR= 2.22, 95%CI (1.44, 3.41))). We found that the increased risk of death (OS) for CTC-positive patients in cytological methods and other methods subgroup was more conspicuous compared with PCR subgroup.
We used funnel plots to detect publication bias, as shown in Figure 5. All of the funnel plots of included studies showed the symmetrical distribution. Thus, no significant publication bias was found in the meta-analyses of OS, PFS, and DFS.