IJC International Journal of Cancer
Clinicopathological and prognostic significance of circulating tumor cells in patients with gastric cancer: A meta-analysis Xuanzhang Huang*, Peng Gao*, Jingxu Sun, Xiaowan Chen, Yongxi Song, Junhua Zhao, Huimian Xu and Zhenning Wang
The prognostic significance of circulating tumor cells (CTCs) and disseminated tumor cells (DTCs) in patients with gastric cancer (GC) is controversial. The aims of our meta-analysis are to assess its correlation with clinicopathological characteristics and prognostic significance in GC. PubMed, Embase, the Cochrane database, the Science citation index, the CNKI database and the references of relevant studies were systematically searched (up to November, 2013). Using the random-effect model, the meta-analysis was completed with odds ratio (OR), risk ratio, hazard ratio (HR) and 95% confidence intervals (CI) as effect values. Twenty-six studies containing 2,566 patients with GC were analyzed. The overall analysis showed that the incidence difference of tumor cells (CTCs/DTCs) was significant when comparing the stage I/II group to the stage III/IV group (OR 5 0.36, CI [0.23, 0.56]), the Lauren diffuse group to the intestinal group (OR 5 2.06, CI [1.06, 4.00]), the poorly differentiated group to the well/moderate group (OR 5 1.65, CI [1.10, 2.50]), the lymphatic involvement positive group to the positive group (OR 5 2.92, CI [1.00, 8.55]). The detection of CTCs/DTCs was significantly related with the disease-free survival of patients (HR 5 3.42, CI [2.39, 4.91]) and the detection of CTCs in peripheral blood was significantly related with the overall survival of patients (HR 5 2.13, CI [1.13, 4.03]). Our meta-analysis indicates that detection of CTCs/DTCs is associated with prognosis for patients with GC and thus could act as a basis for GC staging.
Gastric cancer (GC) is the fourth most frequently diagnosed cancer and second leading cause of cancer death worldwide.1 Curative resection remains common and the main treatment for GC. Even though the clinical prognosis for GC has been improved by the development of surgery and adjuvant chemoradiotherapy,2 the long-term survival of advanced GC patients is highly unsatisfactory and hindered by recurrence and distant metastasis. In recent years, extensive studies have demonstrated that tumor cells in bone marrow (BM, i.e., disseminated tumor
cells [DTCs]) and peripheral blood (PB, i.e., circulating tumor cells [CTCs]),3 which were first reported by Ashworth in 1869,4 are related to tumor metastasis and relapse. Currently, highly sensitive and specific diagnostic methods have been developed and used for tumor cell detection, including reverse-transcriptase polymerase chain reaction (RT-PCR), immunocytochemistry (ICC), flow cytometry (FCM), and the CellSearch system that targets either tumor-associated mRNA or cytokeratin (CK) isozymes.5 Many researchers using these
Key words: circulating tumor cells, gastric cancer, prognosis, clinicopathological characteristics, meta-analysis Abbreviations: AJCC: American Joint Committee on Cancer; BM: bone marrow; CTCs: circulating tumor cells; DFS: disease-free survival; DTCs: disseminated tumor cells; GC: gastric cancer; HR: hazard ratio; ICC: immunocytochemistry; intra: intraoperative; JCGC: Japanese Classification of Gastric Carcinoma; NR: not reported; OR: odds ratio; OS: overall survival; PB: peripheral blood; pre: preoperative; post: postoperative; RR: relative risk; RT-PCR: reverse-transcriptase polymerase chain reaction; SEN: sensitivity analysis; UICC: Union for International Cancer Control Additional Supporting Information may be found in the online version of this article. Conflicts of interest: The sponsors had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data and analyses. Xuanzhang Huang, Peng Gao, Jingxu Sun, Xiaowan Chen, Yongxi Song, Junhua Zhao, Huimian Xu and Zhenning Wang declare that they have no conflicts of interest and all authors approve the final decision to submit for publication. *X.H. and P.G. contributed equally to this work Grant sponsor: National Science Foundation of China; Grant numbers: 81201888, 81372549 and 81172370; Grant sponsor: Specialized Research Fund for the Doctoral Program of Higher Education; Grant number: 20122104110009; Grant sponsor: Project of Science and Technology of Shenyang; Grant number: F12-193-9-08 DOI: 10.1002/ijc.28954 History: Received 29 Jan 2014; Accepted 2 May 2014; Online 6 May 2014 Correspondence to: Dr. Zhenning Wang, Department of Surgical Oncology and General Surgery, The First Hospital of China Medical University, Shenyang 110001, People’s Republic of China, Tel.: [186-24-83283556], Fax: 1[86-24-22703578], E-mail:
[email protected] C 2014 UICC Int. J. Cancer: 136, 21–33 (2015) V
Cancer Cell Biology
Department of Surgical Oncology and General Surgery, First Hospital of China Medical University, Heping District, Shenyang City, People’s Republic of China
22
Clinicopathological and prognostic significance
Cancer Cell Biology
What’s new? Tumor cells that make their way into the bloodstream or bone marrow are known, respectively as circulating tumor cells (CTCs) and disseminated tumor cells (DTCs). In gastric cancer, both CTCs and DTCs are possible indicators of poor outcome, though their prognostic significance is controversial. Following assessment of correlations between CTCs, DTCs, and gastric cancer prognosis, the authors of the present meta-analysis conclude that the detection of the wayward cells could assist with gastric cancer staging. Correlations were significant specifically between overall survival and CTC detection in peripheral blood and between disease-free survival and both CTC and DTC detection.
methods have reached the conclusion that the presence of CTCs/DTCs is a poor prognostic indicator for breast, colorectal cancer and melanoma but not for GC.6,7 This raises the question of whether the detection of CTCs/DTCs in a patient with GC indicates poor prognosis.8,9 However, a pooled analysis demonstrating an association between CTCs/ DTCs detection and prognosis has not yet been performed. The aims of our study were to use meta-analysis to indicate the correlation between CTCs/DTCs and the clinicopathological characteristics of gastric tumors and estimate whether detection of CTCs/DTCs can act as a clinical predictor for patients with GC.
Material and Methods Literature search
PubMed, Embase, the Cochrane database, the Science citation index and the CNKI database were systematically searched for studies addressing the clinicopathological and prognostic relationship between CTCs/DTCs and GC without language, publication or time restrictions (up to November, 2013). Moreover, we screened the references of the relevant studies (reviews and included studies) to check for potentially relevant articles. The main search terms were “CTCs,” “micrometastasis,” “DTCs,” “BM,” “PB,” “portal blood,” “mesenteric blood” and “gastric cancer/stomach cancer/gastrointestinal cancer.”
fewer than 20, (ii) the samples came from lymph nodes or the peritoneal cavity, (iii) the outcomes of interest were not reported and it was impossible to calculate outcomes from the originally published data or (iv) repeated studies were based on the same database or patients. To avoid the influence of redundant studies, we checked all of the authors and organizations, and evaluated the accrual period and community of patients enrolled for each study. Data extraction and assessment of study quality
Two reviewers (X.Z. Huang and P. Gao) evaluated each study and extracted data independently, and any disagreements were resolved via discussion. Output for this meta-analysis included the first author, year of publication, characteristics of the study population (i.e., number, sex, age, accrual period and community), tumor clinicopathological characteristics, neoadjuvant or postoperative chemoradiotherapy, curative surgery, sample site (BM, PB or MPB), sample time (preoperative, intraoperative or postoperative), detection method, target genes/antigens, rate of CTCs/DTCs positivity, follow-up period, cutoff point, prognostic value (overall survival [OS] and disease-free survival [DFS]) and hazard ratio (HR). The quality of studies was evaluated according to the Newcastle– Ottawa scale (NOS).11 Statistical methods and subgroup/sensitivity analysis
Inclusion criteria
To keep our analysis accurate and reliable, eligible studies were selected based on the after criteria. (i) The clinicopathological or prognostic significance of CTCs/DTCs detection in GC patients with at least one of the outcome measures of interest was reported in the study or calculated from published data. (ii) RT-PCR, FCM, ICC or CellSearch detection methods were used to detect tumorspecific genes/antigens (i.e., CEA mRNA, CK mRNA and surviving mRNA) that had a high sensitivity and specificity for CTCs/ DTCs.10 (iii) Samples were collected from the BM, PB or mesenteric/portal blood (MPB). When several studies were reported from the same authors or organizations, the meta-analysis enrolled the most recent or highest quality study only if the most recent one did not fit the inclusion criteria. Exclusion criteria
Studies were excluded if (i) the number of patients with GC in all clinical parameters of the comparative analysis was
Statistical analysis was performed using Review Manager 5.2 software (Copenhagen: The Nordic Cochrane Centre; The Cochrane Collaboration, 2012), and quality of reporting of meta-analyses was utilized as a guideline for this metaanalysis.12 The estimated risk ratio (RR) and HR were used to evaluate the prognostic effect (relapse/metastasis events, DFS and OS), and the estimated odds ratio (OR) was used to summarize the association between CTCs/DTCs detection and gastric tumor clinicopathological characteristics. If the HR and its variance were not reported directly in the original study, these values were calculated from available reported data using software designed by Tierney et al.13 All statistical values were combined with a 95% confidence interval (95% CI) and the p-value threshold was set at 0.05. The random-effects mode was used to perform the analysis, because this model obtained more conservative results than the fixed-effect model and better fit the multicenter clinical studies due to C 2014 UICC Int. J. Cancer: 136, 21–33 (2015) V
Huang et al.
23
Correlation of tumor cells with clinicopathological parameters The incidence of tumor cells in the TNM stage. The meta-
Figure 1. Selection of studies. Flow diagram showing the selection process for the including studies.
the existence of heterogeneity.14 Heterogeneity was calculated using a Q test, and the I2 value shows the degree of heterogeneity. Publication bias was tested using a funnel plot. The overall analysis was completed by evaluating all the relevant studies according to different clinicopathological parameters and prognostic outcomes. Simultaneously, subgroup analysis was performed in each clinicopathological parameter on the basis of the detection method (PCR, no-PCR), sample time (preoperative, intra/postoperative) and sample site (BM, PB). To avoid the same patients being analyzed repetitively and to maintain good accuracy and precision for the pooled analysis, sensitivity analyses were compiled under criteria that chose one study arm from the BM/PB and one study arms from preoperative/(intra/postoperative) in redundant studies to obtain the minimum or maximum 95% CI.
Results Baseline characteristics of included studies
We identified 1,381 studies in this systematic literature search. Sixty-five potential studies were retrieved via the titles and abstracts. Then 39 studies were excluded because of redundancy or the lack of an outcome of interest. The remaining 26 studies, comprised of seven studies for BM,9,15–20 18 studies for PB,8,21–37 one study for BM/PB38 and no studies for MPB, met the selection criteria for the final analysis (Fig. 1). The studies were from Asia, Europe and Africa (China, Germany, Italy, Japan, Korea, Taiwan, Turkey, Poland and Egypt) and containing 2,566 patients with GC (sample size C 2014 UICC Int. J. Cancer: 136, 21–33 (2015) V
analysis of all relevant studies on TNM stage indicated a significantly lower incidence of tumor cells in the stage I/II group relative to the stage III/IV group (OR 5 0.36, 95% CI [0.23,0.56], I2 5 70%). Sensitivity analysis was performed by altering the study arms (the lowest effect, OR 5 0.41, 95% CI [0.27, 0.62], I2 5 62%). Then subgroup analysis were performed on sample site (BM: OR 5 0.27, 95% CI [0.09, 0.77], I2 5 81%; PB: OR 5 0.39, 95% CI [0.23, 0.67], I2 5 68%), detection method (PCR: OR 5 0.33, 95% CI [0.16, 0.66], I2 5 76%) and sample time (preoperative: OR 5 0.36, 95% CI [0.20–0.65], I2 5 73%; intra/postoperative: OR 5 0.40, 95% CI [0.13, 1.29], I2 5 69%). We also observed a similar trend in the subgroup analysis of the BM and PM compartments separately on the basis of sample time and detection method. The results of pooled analysis on GC TNM stage are summarized in Table 3 or Supporting Information Figure 1. Furthermore, pooled analyses of all studies on the pT category, lymph node metastasis and distant metastasis were performed separately. The incidence of CTCs/DTCs was significantly different between the pT1/T2 and pT3/T4 groups (overall: OR 5 0.38, 95% CI [0.24, 0.63], I2 5 55%; BM: OR 5 0.53 95% CI [0.34, 0.84], I2 5 19%; PB: OR 5 0.33, 95% CI [0.15, 0.71], I2 5 61%), as well as lymph node metastasis positive and negative groups (overall: OR 5 2.89, 95% CI [1.92, 4.34], I2 5 60%, sensitivity analysis [SEN]: OR 5 2.60, 95% CI [1.79, 3.79], I2 5 53%; BM: OR 5 3.22, 95% CI [1.34, 7.74], I2 5 71%; PB: OR 5 2.88, 95% CI [1.79, 4.62], I2 5 54%). Also, there was a significant difference in the incidence of CTCs/ DTCs detection between the distant metastasis positive group and distant metastasis negative group (overall: OR 5 1.96, 95% CI [1.28, 3.01], I2 5 16%; BM: OR 5 1.85, 95% CI [1.15, 2.97], I2 5 0%; BP: OR 5 2.31, 95% CI [1.01, 5.30], I2 5 42%). Histologic type. The detection of tumor cells in BM/PB may indicate the degree of tumor differentiation (Table 3) and Lauren type (poorly differentiated vs. well and moderately
Cancer Cell Biology
median: 55 [29–810], mean: 99) and were published between 1991 and 2013. Except for one study on BM that applied RTPCR, the remaining seven studies used ICC. Of these 19 studies on PB, 15 studies used the RT-PCR method. The sampling time was divided into three time points: preoperative in 19 studies, intraoperative in three studies and postoperative in three studies. The sampling time was not reported in two studies. Fourteen studies provided HRs on DFS or OS to perform the metaanalysis. Excluding the OS study by Hiraiwa et al.30 which provided an inadequate follow-up time of only 12 months, nine of the remaining 13 studies8,16–18,26–28,33,37 were available for HRs on OS, and eight studies8,9,15,16,31,33,35,37 were available for HRs on DFS. Meanwhile, the baseline and clinicopathological characteristics and the outcome data of interest are summarized in Table 1. The quality of the 24 included cohort studies was evaluated according to the NOS (Table 2).
15
17
NR/NR
46 (30/16)
NR/NR
NR/NR
810 (546/264)
810 (546/264)
NO/YES
follow-up:26
Mimori et al.38
NO/YES
55 (38/17)
Yie et al.31
NR/NR
M1:YES/YES; M0:NR/NR
41
12 (8/4)
Hiraiwa et al.30
NR/NR
40 (26/14)
Tani et al.29
NR/YES
NR/NR
57 (44/13)
Total: 1652
NR/NR
NO/NO
109 (70/39)
52 (31/21)
NR/NR
NR/NR
R0:41
59 (38/21)
NR/NR
70 (48/22)
Pituch-Noworolska et al.28
Gretschel et al.9
Uen et al.
27
Ikeguchi et al.8
Illert et al.26
46 (30/16)
Seo et al.25
NR/YES
NR/NR
53 (38/15)
Matsunami et al.19
NR/NR
NR/NO
R0:88
651 (35/30)
NR/NR
114 (73/41)
NR/NR
NR/NR
NR/NR
Shin et al.24
de Manzoni et al.
29 (19/10)
18
52
Noh et al.23
62
NR/YES
NR/NR
total: 180
34 (18/16)
NR/NR
NR/NR
CRT
109(63/46)
97(NR)
Number (M/F)
Majima et al.22
Schott et al.
Yeh et al.21
Jauch et al.16
Schlimok et al.
Reference
NR
NR
NR
58 (26–77)
NR
Mean:57
Mean:65(1); 64(2)
NR
NR
NR
60 (34–84)
66.3 (26–86)
NR
69 (41–87)
58 (31–78)
58 (31–78)
64 (35–78)
NR
NR
67 (27–87)
58 (33–74)
NR
NR
57 (31–81)
NR
64 6 11
NR
Age mean 6 SD/ Median (range)
PB
BM
PB
PB
PB
PB
PB
PB
BM
BM
PB
PB
PB
PB
PB
PB
BM
PB
BM
BM
PB
PB
BM
PB
BM
BM
BM
SS
Table 1. Baseline characteristics and design variables of the including studies
pre
pre
NR
NR
pre
post
pre
pre
pre
pre
intra
post
pre
pre
post
pre
pre
pre
pre
pre
post
pre
pre
pre
intra
intra
intra
ST
RT-PCR
RT-PCR
RT-PCR ELISA
RT-PCR ELISA
CellSearch
RT-PCR
RT-PCR
FACS,ICC
ICC
ICC
RT-PCR
RT-PCR
RT-PCR
RT-PCR
RT-PCR
RT-PCR
APAAP-ICC
RT-PCR
ICC
ICC
RT-PCR
RT-PCR
ICC
RT-PCR
APAAP-ICC
APAAP-ICC
APAAP-ICC
Method
MT1–MMP
MT1–MMP
survivin
survivin
EpCAM,CK8,CK19
hTERT,MUC1
hTERT,MUC1
CD45,CK8, CK18,CK19
CK18(CK2)
CK18(CK2)
MUC1,c-Met
CEA
CK20
CK20
CEA
CEA
pan-CK
CEA,MUC-1, c-MET hTERT
CK8,CK19 (CAM5.2)
CK8,CK19 (CAM5.2)
CEA
CK19,CK20
CEA (C1P83), CAl9-9,Kl-1, Ra96,17-1A
CK19
CK18(CK3)
CK18(CK2)
CK18(CK2)
Target antigen/ gene
185/810
34/810
12/26
25/55
17/41
2/12
6/40
31/57
27/165
15/109
43/52
27/59
15/41
28/70
9/46
24/46
16/53
30/65
24/88
33/114
16/29
5/52
30/62
7/34
95/180
55/109
34/97
Rate(1)
Cancer Cell Biology
NR(Max>24)
NR(Max>24)
Max:36 or until relapse
NR
NR
NR
NR
NR
Max:56
Max:56
26–39
20.1(2–31)
20(1–57)
20(1–57)
Mean:12
Mean:12
17.5 (3–29.7)
NR
NR
30(9–49)
Max:48
NR
Max:48
NR
NR
30.6 6 15.1
Mean:25
Follow up mean 6 SD/ median (range)
NR
NR
2000 WHO
2000 WHO
AJCC
Sixth UICC
Sixth UICC
Fifth UICC
NR
NR
Sixth AJCC
Second JCGC
Fourth UICC
Fourth UICC
Fifth AJCC
Fifth AJCC
Fifth UICC
NR
Fifth UICC
Fifth UICC
NR
Fifth UICC
Fourth UICC
NR
Fourth UICC
Fourth UICC
NR
TNM edition
NR
NR
DFS(9)
NR
NR3
NR
NR
OS(10)
NR
DFS(10)
OS(10)
OS;DFS (10)
OS(9)
NR
NR
NR
NR
NR
OS(10)
NR
NR
NR
OS(10)
NR
NR
OS;DFS (9)
DFS (9)
HR (method NO.)
NR
NR
DFS
NR
OS
NR
NR
OS
NR
DFS
OS
OS DFS
OS
NR
NR
NR
NR
NR
OS
NR
NR
OS
OS
NR
NR
DFS OS
DFS
OM
YES
YES
NR
NR
NR
YES
NR
NR
109R0
YES
Part YES
55R0
YES
41R0
YES
YES
NR
NR
YES
88R0
YES
NR
41R0
NO
Part YES
YES
YES
CR
284
98 (63/35)
Kutun et al.34
Cao et al.35
NO/NR
148 (99/49)
40 (28/12)
Uenosono et al.37
Wang et al.20
NR/NR
NO/28 YES
NO/74 YES
NR/NR
NO/YES
NR/NR
CRT
65 (46/19)
Ito et al.
36
30 (16/14)
Saad et al.33
30 (15/15)
Number (M/F) (31–86)
57 (40–73)
NR
58.8 (33–76)
NR
NR
55.0 6 10; (31–72)
BM
PB
PB
PB
PB
PB
PB
SS
pre
pre
pre
NR
pre
pre
pre
ST
ICC MACS
CellSearch
ICC
RT-PCR ELISA
RT-PCR
RT-PCR
RT-PCR
Method
CK,HEA(CK-FITC, HEA125)
EpCAM,CK8 CK19
GFP,EpCAM, CD45
survivin
CEA,CK19
CK18
CEA
Target antigen/ gene
18/40
16/148
24/65
45/98
15/28
15/30
11/30
Rate(1)
NR(Max>24)
31.6(4–72)
NR
47.5(36.5–56.0)
NR
NR
NR
Follow up mean 6 SD/ median (range)
NR
Seventh UICC
Seventh UICC
Fifth UICC
NR
Seventh UICC
NR
TNM edition
NR
OS(in text); DFS(10)
NR
DFS(11)
NR
OS(9); DFS(10)
NR
HR (method NO.)
NR
OS RFS
OS
DFS
NR
DFS OS
NR
OM
NR
NR
Part YES
YES
YES
YES
NR
CR
3
2
Cancer Cell Biology
The number of peripheral blood that used for analyzing hTERT expression was largest, and the sensitivity and specificity of hTERT was relatively better. Only provides for the 165 patients who underwent resection of the tumor from primary article. The HR for OS was unavailable because the inadequate follow-up time was only approximate 12 months. 4 Only group II could provide several data of interest. Abbreviations: CRT: chemoradiotherapy(pre/postoperation); Rate(1): Rate of CTC/DTC positive patients, n/N(%); SS: sample site; ST: sample time; OM: outcome measured; CR: curative resection; NR: not reported; CTCs: circulating tumor cells; DTCs: disseminated tumor cells; pre: preoperative; intra: intraoperative; post: postoperative; BM: bone marrow; PB: peripheral blood; OS: overall survival; DFS: disease-free survival; RR: relative risk; OR: odds ratio; HR: hazard ratio; UICC: Union for International Cancer Control; AJCC: American Joint Committee on Cancer; JCGC: Japanese Classification of Gastric Carcinoma; WHO: World Health Organization; ICC: Immunocytochemistry; RT-PCR: reverse-transcriptase polymerase chain reaction; Max: maximum; APAAP: alkaline phosphatase antialkaline phosphatase; ELISA: enzyme-linked immunosorbent assay; MACS: magnetic activated cell sorting; FACS: fluorescence-activated cell sorter; CK: cytokeratin; CAl9-9: carbohydrate antigen 19-9; CEA: carcinoembryonic antigen; MUC-1: mucin-1; C-Met: c-met protooncogene product; hTERT: human telomerase reverse transcriptase; pan-CK: pancytokeratin; EpCAM: epithelial cell adhesion molecule; MT1-MMP: membrane type 1 matrix metalloproteinase; GFP: green fluorescent protein; FITC: fluorescein isothiocyanate; HEA: human epithelial antigen.
1
32
Piva et al.
Reference
Age mean 6 SD/ Median (range)
Table 1. Baseline characteristics and design variables of the including studies (Continued)
26
Clinicopathological and prognostic significance
Table 2. The assessment of the risk of bias in each Cohort study using the Newcastle–Ottawa scale Comparability (0–2)
Selection (0–4) REC
SNEC
AE
DO
SC
AF
AO
FU
AFU
Total
Schlimok et al.15
0
1
1
1
0
0
0
1
0
4
Jauch et al.
1
1
1
1
0
0
1
1
1
7
Yeh et al.21
0
1
1
1
0
0
0
0
0
3
Schott et al.
1
1
1
1
0
0
1
1
0
6
Majima et al.22
0
1
1
1
0
0
0
0
0
3
Noh et al.
1
1
1
1
0
0
1
1
0
6
de Manzoni et al.18
1
1
1
1
0
0
1
1
1
7
Matsunami et al.
1
1
1
1
0
0
1
0
1
6
Seo et al.25
1
1
1
1
0
0
0
0
1
5
Illert et al.
1
1
1
1
0
0
1
1
1
7
Ikeguchi et al.8
0
1
1
1
0
0
1
0
1
5
Uen et al.
1
1
1
1
0
0
1
0
1
6
Gretschel et al.9
1
1
1
1
0
0
1
1
1
7
Pituch-Noworolska et al.
1
1
1
1
0
0
1
1
1
7
Tani et al.29
0
1
1
1
0
0
1
0
1
5
Hiraiwa et al.
0
1
1
1
0
0
1
0
0
4
Yie et al.31
1
1
1
1
0
0
0
1
0
5
Mimori et al.
1
1
1
1
0
0
0
0
0
4
Saad et al.33
1
1
1
1
0
0
1
0
1
6
1
1
1
1
0
0
0
0
0
3
1
1
1
1
0
0
0
1
0
5
Ito et al.
1
1
1
1
0
0
1
0
0
5
Uenosono et al.37
1
1
1
1
0
0
1
1
0
6
1
1
1
1
0
0
0
0
1
5
16
17
23
19
26
27
28
Cancer Cell Biology
Outcome (0–3)
Study
30
38
Kutun et al.
34
Cao et al.35 36
20
Wang et al.
Abbreviations: REC: representativeness of the exposed cohort; SNEC: selection of the nonexposed cohort; AE: ascertainment of exposure; DO: demonstration that outcome of interest was not present at start of study; SC: study controls for age, sex; AF: study controls for any additional factors (chemoradiotherapy, curative resection); AO: assessment of outcome; FU: follow-up long enough (36M) for outcomes to occur; AFU: adequacy of follow-up of cohorts (90%). “1” means that the study is satisfied the item and “0” means the opposite situation.
differentiated, overall: OR 5 1.65, 95% CI [1.10, 2.50], I2 5 45%; Lauren diffuse vs. intestinal, overall: OR 5 2.06, 95% CI [1.06, 4.00], I2 5 71%). Then, subgroup analyses on sample site were performed for each histologic classification: Lauren diffuse compared to intestinal (BM: OR 5 1.63, 95% CI [1.05, 2.54], I2 5 0%; PB: OR 5 2.71, 95% CI [0.72, 10.22], I2 5 82%) and poorly differentiated compared to well and moderately differentiated (BM: OR 5 1.28, 95% CI [0.71, 2.33], I2 5 0%; PB: OR 5 1.84, 95% CI [1.06, 3.18], I2 5 56%). A higher incidence of CTCs/DTCs was observed in the vessel involvement positive group and lymphatic involvement positive group compared to the corresponding negative group (overall: ORvessel 5 2.77, 95% CI [1.90, 4.05], I2 5 21%; ORlymphatic 5 2.92, 95% CI [1.00, 8.55], I2 5 89% SEN: ORlymphatic 5 1.78, 95% CI [0.89, 3.55], I2 5 64%). The subgroup analysis on sample site showed similar results (BM: ORvessel 5 2.76, 95% CI [1.66, Vessel and lymphatic invasion.
4.60], I2 5 0%; ORlymphatic 5 1.20, 95% CI [0.49, 2.93], I2 5 70%; PB: ORvessel 5 3.38, 95% CI [1.60, 7.15], I2 5 48%; ORlymphatic 5 4.57, 95% CI [1.07, 19.42], I2 5 85%). In the main, the level of CTCs/DTCs detection was higher in the vessel invasion-positive group, but the level in lymphatic invasion was less clear (Table 3). Relapse/metastasis and death events based on follow-up
The incidence of relapse/metastasis was more likely to increase in CTCs/DTCs positive patients than in CTCs/TCs negative patients (overall: RR 5 2.17, 95% CI [1.50, 3.15], I2 5 73%, SEN: RR 5 2.22, 95% CI [1.44, 3.41], I2 5 76%; BM: RR 5 1.89, 95% CI [1.43, 2.50], I2 5 26%, SEN: RR 5 1.92, 95% CI [1.33, 2.77], I2 5 38%; BP: RR 5 2.38, 95% CI [1.11, 5.12], I2 5 83%, SEN: RR 5 1.98, 95% CI [0.90, 4.34], I2 5 82%). We observed similar results for the overall analysis of all studies on death events in the follow-up period (overall: RR 5 2.17, 95% CI [1.16, 4.07], I2 5 64%), whereas C 2014 UICC Int. J. Cancer: 136, 21–33 (2015) V
Pre
2.90 [1.40, 3.12 [1.82, 2.76 [1.66, 2.76 [1.66, – 6.01] 5.35] 4.60] 4.60] I2 5 55% I2 5 0% I2 5 0% I2 5 0% 4.88 [3.44, 2.44 [1.06, 1.20 [0.49, 1.13 [0.31, – 6.94] 5.62] 2.93] 4.09] I2 5 94% I2 5 60% I2 5 70% I2 5 73%
2.77 [1.90, 2.45 [1.84, – 3.26] 4.05] I2 5 0% I2 5 21%
2.92 [1.00, 3.19 [0.89, – 11.48] 8.55] I2 5 90% I2 5 89%
Vessel invasion (1) vs. (2) (OR)
Lymphatic invasion (1)vs. (2) (OR)
Cancer Cell Biology
–
2.17 [1.16, 2.26 [0.98, – 4.07] 5.25] I2 5 64% I2 5 72%
Death (RR)
4.76 [0.61, 1.67 [0.91, 1.67 [0.91, 1.44 [0.65, – 37.20] 3.08] 3.08] 3.15] I2 5 76% I2 5 53% I2 5 53% I2 5 52%
2.17 [1.50, 2.28 [1.27, 1.87 [1.10, 1.95 [0.97, 2.41 [1.51, 1.89 [1.43, 1.91 [1.20, 2.01 [1.39, – 3.15] 4.08] 3.20] 3.91] 3.84] 2.50] 3.04] I2 5 73% I2 5 77% I2 5 70% I2 5 68% I2 5 77% I2 5 26% I2 5 48% 2.90] I2 5 0%
–
–
Relapse/metastasis (RR4)
During follow up
1.87 [0.78, 1.84 [1.06, 1.60 [0.88, – 4.52] 3.18] 2.90] I2 5 0% I2 5 56% I2 5 43%
1.85 [1.03, 1.40 [0.77, 1.28 [0.71, 1.28 [0.71, – 3.33] 2.55] 2.33] 2.33] I2 5 61% I2 5 0% I2 5 0% I2 5 0%
1.65 [1.10, 1.40 [0.96, – 2.04] 2.50] I2 5 21% I2 5 44%
Differentiated: poor vs. well 1 moderate
1.62 [0.79, 4.76 [0.61, 4.76 [0.61, – 3.31] 37.20] 37.20] I2 5 64% I2 5 76% I2 5 76%
3.72 [1.62, 8.53] I2 5 0%
1.10 [0.49, 2.48] I2 5 0%
–
2.79 [0.94, 8.33] I2 5 49%
2.35 [0.63, 8.82] I2 5 73%
0.47 [0.13, 1.73] I2 5 71%
4.76 [0.61, 37.20] I2 5 76%
–
–
13.75 [7.98, 5.15 [0.60, 23.71] 43.92] I2 5 88% I2 5 72%
4.32 [1.00, 18.57] I2 5 61%
2.18 [1.09, 4.35] I2 5 64%
3.93 [0.97, 15.91] I2 5 78%
1.70 [0.42, 6.82] I2 5 33%
1.90 [1.40, 2.38 [1.11, 2.43 [0.61, 2.09 [0.60, 1.96 [0.92, 2.59] 5.12] 9.67] 7.25] 4.17] I2 5 38% I2 5 83% I2 5 84% I2 5 81% I2 5 72%
1.70 [0.69, 4.57 [1.07, 4.57 [1.07, – 4.15] 19.42] 19.42] I2 5 52% I2 5 85% I2 5 85%
2.74 [1.35, 3.38 [1.60, 2.53 [1.54, – 5.59] 7.15] 4.16] I2 5 0% I2 5 48% I2 5 17%
1.63 [1.05, 2.71 [0.72, 1.00 [0.50, – 2.54] 10.22] 1.99] I2 5 0% I2 5 82% I2 5 0%
1.85 [1.15, 2.31 [1.01, 2.14 [0.86, – 2.97] 5.30] 5.33] I2 5 0% I2 5 42% I2 5 48%
2.06 [1.06, 1.29 [0.78, 1.91 [1.13, 3.93 [0.97, 1.35 [0.82, 1.63 [1.05, 1.56 [0.61, 1.60 [0.88, – 4.00] 2.13] 3.24] 15.91] 2.23] 2.54] 4.02] 2 2 2 2 2 2 2 2 I 5 71% I 5 8% I 5 0% I 5 78% I 5 30% I 5 0% I 5 50% 2.93] I 5 0% –
0.26 [0.09, 0.72] I2 5 60%
1.90 [1.19, 2.88 [1.79, 2.85 [1.62, 8.03 [0.30, 3.16 [1.94, 3.05] 4.62] 5.01] 211.81] 5.15] I2 5 14% I2 5 54% I2 5 51% I2 5 86% I2 5 42%
0.53 [0.34, 0.33 [0.15, 0.55 [0.25, – 0.84] 0.71] 1.21] I2 5 19% I2 5 61% I2 5 50%
0.29 [0.07, 1.09] I2 5 72%
No RT-PCR
Detection method
Intra/post RT-PCR
Lauren: diffuse vs. intestinal
Histologic type (OR)
–
1.96 [1.28, 1.86 [1.10, – 3.15] 3.01] I2 5 28% I2 5 16%
Metastasis before surgery (OR)
1.70 [0.42, 2.11 [1.34, 1.85 [1.15, 1.72 [1.00, – 6.82] 3.30] 2.97] 2.96] I2 5 33% I2 5 14% I2 5 0% I2 5 0%
–
Any
PB Sample time
0.50 [0.33, 0.39 [0.23, 0.45 [0.22, 0.21 [0.01, 0.43 [0.23, 0.75] 0.67] 0.90] 4.87] 0.80] I2 5 0% I2 5 69% I2 5 72% I2 5 85% I2 5 69%
No RT-PCR RT-PCR
2.89 [1.92, 3.10 [1.89, 3.28 [0.85, 3.74 [2.12, 2.08 [1.26, 3.22 [1.34, 4.28 [1.32, – 4.34] 5.08] 12.72] 6.62] 3.43] 7.74] 13.92] I2 5 60% I2 5 57% I2 5 77% I2 5 58% I2 5 41% I2 5 71% I2 5 73%
Intra/post
LN3(1) vs. LN(2)(OR)
Pre
–
No RT-PCR Any
Detection method
pT:T1/T2 vs. T3/T4(OR) 0.38 [0.24, 0.49 [0.29, 0.22 [0.02, 0.26 [0.09, 0.49 [0.30, 0.53 [0.34, 0.44 [0.21, – 0.63] 0.84] 2.29] 0.72] 0.80] 0.84] 0.93] I2 5 55% I2 5 40% I2 5 85% I2 5 60% I2 5 44% I2 5 19% I2 5 39%
Intra/post RT-PCR
BM Sample time
–
Pre
Detection method
0.36 [0.23, 0.36 [0.20, 0.40 [0.13, 0.33 [0.17, 0.41 [0.25, 0.27 [0.09, 0.21 [0.05, –2 0.56] 0.65] 1.29] 0.66] 0.68] 0.77] I2 5 70% I2 5 73% I2 5 69% I2 5 76%, I2 5 38% I2 5 81% 0.90] I2 5 83%
StageI/II vs. III/IV(OR1)
Any
Sample time
Any site
Table 3. Detailed results of subgroup analyses for clinicopathological parameters and prognostic significance
3.83 [1.91, – 7.69] I2 5 19%
2
OR: odds ratio. “–” symbol: no results due to insufficient studies. 3 LN: lymph node. 4 RR: relative risk. 5 HR: hazard ratio. 6 OS: overall survival. 7 DFS: disease-free survival. 8 NOS: score for Newcastle–Ottawa scale.
1
HR for DFS (NOS > 5)
8
Pre
Intra/post
–
–
–
4.20 [1.22, – 14.43] I2 5 68%
–
4.98] I2 5 0%
2.94 [1.73,
3.42 [2.39, 6.72 [2.59, 2.85 [1.70, 3.23 [1.87, 3.83 [2.15, 3.77 [1.97, – 4.91] 17.45] 4.79] 5.57] 6.81] 7.21] I2 5 0% I2 5 0% I2 5 0% I2 5 0% I2 5 20% I2 5 38%
No RT-PCR Any
HR for DFS7
Intra/post RT-PCR
BM Sample time
1.58 [0.93, 1.86 [1.01, 0.86 [0.38, 2.98 [1.36, 1.21 [0.62, 0.93 [0.53, 0.98 [0.47, – 3.41] 1.93] 6.54] 2.37] 1.65] 2.03] 2.68] I2 5 66% I2 5 0% I2 5 19% I2 5 69% I2 5 0% I2 5 0% I2 5 58%
Pre
Detection method
HR5 for OS6
Any
Sample time
Any site
Table 3. Detailed results of subgroup analyses for clinicopathological parameters and prognostic significance (Continued)
–
–
–
Any
Pre
–
4.53 [1.04, – 19.76] I2 5 0%
–
3.77 [1.97, 3.37 [1.97, 4.53 [1.04, – 7.21] 5.75] 19.76] I2 5 38% I2 5 0% I2 5 0%
–
3.23 [1.87, 5.57] I2 5 0%
–
–
1.52 [0.41, 5.67] I2 5 84%
No RT-PCR
Detection method Intra/post RT-PCR
Sample time
PB
0.93 [0.53, 2.13 [1.13, 2.41 [1.19, 0.81 [0.14, 2.98 [1.36, 1.65] 4.03] 4.88] 4.61] 6.54] I2 5 0% I2 5 53% I2 5 66% I2 5 0% I2 5 19%
No RT-PCR RT-PCR
Detection method
Cancer Cell Biology
29
Cancer Cell Biology
Huang et al.
Figure 2. Estimated hazard ratios (HR) summary for DFS (a) and OS (b). The estimated hazard ratios (HR) are summarized for DFS (a) and overall survival (b). (a) HR for DFS with CTCs/DTCs detection. (b) HR for overall survival with CTCs/DTCs detection. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
the results were not significant in the BM and PB subgroups (BM: RR 5 1.67, 95% CI [0.91, 3.08], I2 5 53%; PB: RR 5 4.76, 95% CI [0.61, 37.20], I2 5 76%). The prognostic effect (OS and DFS) of CTCs detection
To assess the prognostic effect for detection of tumor cells in GC patients more deeply, a meta-analysis was performed on C 2014 UICC Int. J. Cancer: 136, 21–33 (2015) V
HR for DFS or OS. A poor prognosis for OS in patients with GC was shown by the detection of tumor cells in PB, but not in BM (overall: HR 5 1.58, 95% CI [0.93, 2.68], I2 5 58%; BM: HR 5 0.93, 95% CI [0.53, 1.65], I2 5 0% PB: HR 5 2.13, 95% CI [1.13, 4.03], I2 5 53%). Of the three studies on BM, none showed a significant prognostic effect for OS. Of the six studies on PB, three studies (Illert et al.,26 Saad et al.33
30
Clinicopathological and prognostic significance
Cancer Cell Biology
Discussion
Figure 3. Assessment of publication bias using Funnel plot analysis. (a) Funnel plot analysis of studies on DFS. (b) Funnel plot analysis of studies on OS. Both funnel plots have a symmetrical distribution. Publication bias was not found in the meta-analyses of DFS and OS. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
and Uenosono et al.37 showed that the presence of CTCs was associated with a significant increase in the risk of death. In addition, the detection of tumor cells indicated a poor prognostic effect on DFS (overall: HR 5 3.42, 95% CI [2.39, 4.91], I2 5 0%; BM: HR 5 3.77, 95% CI [1.97, 7.21], I2 5 38%; PB: HR 5 3.37, 95% CI [1.97, 5.75], I2 5 0%). As was shown by subgroup analysis on the sample time and detection method, a significant prognosis of CTCs/DTCs detection was confirmed as follow: between preoperative (overall: HR 5 6.72, 95% CI [2.59, 17.45], I2 5 0%; PB: HR 5 4.53, 95% CI [1.04, 19.76], I2 5 0%) and intra/postoperative (overall: HR 5 2.85, 95% CI [1.70, 4.79], I2 5 0%; BM:HR 5 2.94, 95% CI [1.73, 4.98], I2 5 0%), between RT-PCR (overall: HR 5 3.23, 95% CI [1.87, 5.57], I2 5 0%; PB: HR 5 3.23, 95% CI [1.87, 5.57], I2 5 0%) and no RT-PCR (overall: HR 5 3.83, 95% CI [2.15, 6.81], I2 5 20%; BM: HR 5 3.77, 95% CI [1.97, 7.21], I2 5 38%). Sensitivity analysis was also performed on HR for DFS (Supporting Information Fig. 2) without the low quality studies (NOS score