Clinicopathologic and Prognostic Value of Serum Carbohydrate Antigen 19-9 in Gastric Cancer: A Meta-Analysis

Background. The clinical value of carbohydrate antigen (CA) 19-9 in gastric cancer is controversial. We evaluated the clinicopathologic and prognostic value of CA 19-9 in gastric cancer. Methods. A literature search was conducted in PubMed and Embase databases. Odds ratios (ORs), risk ratios (RR), hazard ratios (HRs), and 95% confidence intervals (CIs) were used as effect measures. Results. Thirty-eight studies were included. Results showed that there were significant differences in the incidence of high CA 19-9 levels between stages III/IV and I/II groups (OR = 3.36; 95% CI = 2.34–4.84), the pT3/T4 and pT1/T2 groups (OR = 2.40; 95% CI = 1.60–3.59), the lymph node-positive and node-negative groups (OR = 2.91; 95% CI = 2.21–3.84), the metastasis-positive and metastasis-negative groups (OR = 2.76; 95% CI = 1.12–6.82), and vessel invasion-positive and invasion-negative groups (OR = 1.66; 95% CI = 1.11–2.48). Moreover, CA 19-9 was significantly associated with poor overall survival (HR = 1.83; 95% CI = 1.56–2.15), disease-free survival (HR = 1.85; 95% CI = 1.16–2.95), and disease-specific survival (HR = 1.33; 95% CI = 1.10–1.60) in gastric cancer. Conclusions. Our meta-analysis showed that CA 19-9 indicates clinicopathologic characteristics of gastric cancer and is associated with a poor prognosis.


Introduction
Gastric cancer remains the fourth most commonly diagnosed cancer and third leading cause of cancer deaths worldwide [1]. Gastric cancer is often diagnosed at an advanced stage and the survival rate is low [2]. The tumor-node-metastasis (TNM) classification is the most important prognostic factor in gastric cancer, but it is still difficult to obtain complete prognostic information [3]. Therefore, it is important to identify other markers, which should be simple, feasible, and less costly, for the assessment of clinicopathologic characteristics and prediction of prognosis.
As a type of tumor-associated antigen for gastrointestinal cancer, carbohydrate antigen (CA) 19-9 is a sialylated derivative of Lewis a blood group antigen [4] and is not applied to TNM staging of gastric cancer according to the American Joint Committee on Cancer (7th edition) [5]. Most research has focused on pancreatic cancer [6][7][8]. Therefore, the clinical value of CA 19-9 in gastric cancer remains controversial and has not been confirmed [9][10][11].
The aim of this meta-analysis was to evaluate the relationship between CA 19-9 and clinicopathologic characteristics and the prognostic value of CA 19-9 in gastric cancer.

Literature Search.
A systematic literature search was conducted for relevant studies using PubMed and Embase databases. The search strategy included the following terms: "carbohydrate antigen 19-9," "carbohydrate antigen 199," "CA 19-9," "CA 199," "gastric cancer," and "stomach cancer." A manual search of the reference lists of the retrieved studies and reviews was also performed to identify potential studies.

Eligibility Criteria.
Studies were included if the following inclusion criteria were met: (1) sample for measuring CA 2 Disease Markers 19-9 was obtained from serum of gastric cancer; (2) studies reported the clinicopathologic or/and prognostic values of CA 19-9; and (3) outcome measures could be extracted directly or calculated from published data indirectly. Only the most informative study was included if there were several duplicated studies based on the same patient population.

Data Extraction and Quality Assessment.
Studies were reviewed and data were independently extracted by two reviewers (Yong-xi Song and Xuan-zhang Huang). The following data were extracted: first author; country and year of publication; sample size; patients characteristics; followup period; sampling time; cut-off value; and tumor clinicopathologic characteristics and prognostic value (overall survival (OS), disease-free survival (DFS), and disease-specific survival (DSS)). The quality of the included studies was assessed by Newcastle-Ottawa Scale (NOS) criteria [12]. Any disagreements on the data extraction and quality assessment were resolved by comprehensive discussion.

Statistical Analysis.
Odds ratios (ORs) and 95% confidence intervals (CIs) were used as a measure to evaluate the relationship between CA 19-9 and tumor clinicopathologic characteristics. For the relationship between CA 19-9 and prognosis, we used hazard ratios (HRs) and 95% CIs as effect measures. HRs and 95% CIs were extracted directly or calculated from available data using the methods designed by Tierney if the values were not reported directly [13]. For the studies that reported several results separately based on different subgroups, we pooled multiple effect values into an estimated value for further meta-analysis.
The Cochran test and 2 statistics were used to evaluate heterogeneity, and statistically significant heterogeneity was defined as < 0.10 and/or 2 > 50% [14]. We used a random effects model if significant heterogeneity existed; otherwise, a fixed effects model was used. The sources of heterogeneity were explored by metaregression and subgroup analyses. Metaregression was only performed when the number of studies was greater than ten, considering the accuracy of the results [15,16]. Publication bias was assessed using Begg's and Egger's tests, and trim-and-fill analysis was performed to assess the effect of publication bias [17][18][19].
All statistical analyses were conducted using STATA software (version 12.0; Stata Corporation, College Station, TX, USA), and a two-sided value < 0.05 was considered statistically significant.

Baseline Characteristics of Eligible Studies.
A total of 1244 studies were initially identified, of which 943 were excluded after reviewing titles and abstracts. Then, 263 studies were excluded after full-text review. Finally, 38 studies were included (Figure 1) [9][10][11].

CA 19-9 and DFS.
Seven studies evaluated the prognostic effect of CA 19-9 on DFS [9,11,29,34,36,43,48]. Our results indicated that a significantly poor DFS existed in the high CA 19-9 group (HR = 1.85; 95% CI = 1.16-2.95; 2 = 60.2%; Figure 3). The results for preoperative CA 19-9 levels were marginally significant (HR = 1.96; 95% CI = 1.00-3.85; 2 = 72.9%). Similarly, subgroup analyses for cut-off value    Overall (I 2 = 75.8%), p = 0.000) Figure 2: The estimated hazard ratio (HR) was summarized for the association between carbohydrate antigen 19-9 and overall survival. The left-hand column lists the first author of each study; the middle column graphically displays the effect measure for each study incorporating confidence intervals (the black solid point represents effect measure and horizontal line represents 95% confidence intervals), the area of gray square is proportional to the study's weight in the meta-analysis, and the pooled measure of effect is plotted as a diamond; the right-hand column numerically displays the effect measure, 95% confidence intervals, and weight of each study. and NOS score also showed that CA 19-9 was a significant poor prognostic factor for DFS (Table 2).

CA 19-9 and DSS.
Six studies evaluated the prognostic effect of CA 19-9 on DSS [25,37,39,41,42,51]. Our results indicated that the high CA 19-9 group was associated with a significantly poor DSS (HR = 1.33; 95% CI = 1.10-1.60; 2 = 17.5%; Figure 4). Including the studies using a cut-off value of 37 U/mL, we observed a similar result ( Table 2). Overall (I 2 = 60.2%), p = 0.020) Figure 3: The estimated hazard ratio (HR) was summarized for the association between carbohydrate antigen 19-9 and disease-free survival. The left-hand column lists the first author of each study; the middle column graphically displays the effect measure for each study incorporating confidence intervals (the black solid point represents effect measure and horizontal line represents 95% confidence intervals), the area of gray square is proportional to the study's weight in the meta-analysis, and the pooled measure of effect is plotted as a diamond; the right-hand column numerically displays the effect measure, 95% confidence intervals, and weight of each study.

CA 19-9 and Vessel and Lymphatic Invasion.
Four [21,35,48,52] and three studies [48,50,52] assessed the relationship between CA 19-9 and vessel invasion and between CA 19-9 and lymphatic invasion, respectively. Our results showed a significantly higher incidence of high CA 19-9 levels in the vessel invasion-positive group compared with the vessel invasion-negative group, and there was no significant relationship between the CA 19-9 level and lymphatic invasion ( Table 2).

Assessment of Publication Bias and Heterogeneity.
The results of Begg's and Egger's tests showed no significant publication bias, except in the HRs for OS. The trim-andfill analyses indicated that publication bias could not impact on the results for OS. Metaregression analyses showed that publication year may contribute to heterogeneity in the analyses of lymph node and hepatic metastases (lymph node: coefficient = −0.041, standard error = 0.019, and = 0.046; hepatic metastasis: coefficient = −0.113, standard error = 0.021, and = 0.001).

Discussion
Gastric cancer is a global health problem with a low survival rate. Tumor-associated markers are urgently needed for the Overall (I 2 = 17.5%), (p = 0.301) Figure 4: The estimated hazard ratio (HR) was summarized for the association between carbohydrate antigen 19-9 and disease-specific survival. The left-hand column lists the first author of each study; the middle column graphically displays the effect measure for each study incorporating confidence intervals (the black solid point represents effect measure and horizontal line represents 95% confidence intervals), the area of gray square is proportional to the study's weight in the meta-analysis, and the pooled measure of effect is plotted as a diamond; the right-hand column numerically displays the effect measure, 95% confidence intervals, and weight of each study.
diagnosis of cancer and assessment of prognosis. Nevertheless, gastric cancer-specific markers have not yet been established, and the unsatisfactory specificity and sensitivity limit clinical utility. Although many studies have been conducted to evaluate the clinical value of CA 19-9 in gastric cancer, there is still no general agreement. Our meta-analysis provides important and valuable evidence for the individualized treatment for gastric cancer.
Our results indicate that CA 19-9 is associated with clinicopathologic characteristics, including tumor stage, pT category, lymph node metastasis, distant metastasis, and vessel invasion, and is feasible for gastric tumor staging. Moreover, our results provide evidence that CA 19-9 can be used to predict the prognosis of gastric cancer. Similar results were obtained in the subgroup analyses stratified by sample time, cut-off value, and study quality.
Recently, several studies have reported that CA 19-9 is associated with gastric cancer metastasis and prognosis [24,49]. Our results also obtained similar evidence. Metastatic patients had a 2.76-fold elevated level of CA 19-9 compared to patients without metastasis. Moreover, the CA 19-9 level was more frequently elevated in patients with peritoneal and hepatic metastases (2.20-and 3.13-fold, resp.). Our results also indicated that an elevated level of CA 19-9 is associated with poor prognosis in gastric cancer, including poor OS, DFS, and DSS. Marrelli et al. demonstrated that stage I patients with elevated CA 19-9 levels had a similar prognosis to stage II with normal levels, and the prognosis of stage II with elevated levels was similar to stage III with normal levels [49]. Thus, the CA 19-9 level had a very important and valuable prognostic value in patients with gastric cancer. Nevertheless, the intrinsic mechanisms by which the elevated level of CA 19-9 can result in poor prognosis are unclear. The most plausible explanation may be that CA 19-9, as a ligand of E-selectin and an intercellular adhesion molecule, plays a crucial role in intercellular adhesion of tumor cells to vascular endothelial cells and then contributes to tumor invasion and metastasis [55,56]. Indeed, experimental studies have demonstrated that cells expressing CA 19-9 may have a greater capacity of invasion and metastasis [57][58][59]. In addition, Tabuchi et al. suggested that CA 19-9 may be drained by the thoracic duct of the lymphatic system via node metastasis or invasive lymphatics [60]; thus the elevated level of CA 19-9 may prompt the presence of micrometastases, which can lead to subsequent relapse/metastasis. Further studies are needed to investigate the potential mechanisms underlying the association between CA 19-9 and tumor metastasis and prognosis.
In our meta-analysis, the elevated level rate (sensitivity) of CA 19-9 ranges from 6.8% to 51.7% (median: 30.0%) and the CA 19-9 level was most frequently elevated in advanced stage tumor. Kochi et al. reported that the elevated level of CA 19-9 increased gradually with tumor stages (25.3% in stage I, 10.5% in stage II, 12.6% in stage III, and 51.6% in stage IV) [48]. Similarly, Duraker and Elk reported that the rate of advanced tumors (stages III and IV) was significantly higher  Overall (I 2 = 71.6%), (p = 0.000) Figure 5: The estimated odds ratio (OR) was summarized for the association between carbohydrate antigen 19-9 and tumor stage. The lefthand column lists the first author of each study; the middle column graphically displays the effect measure for each study incorporating confidence intervals (the black solid point represents effect measure and horizontal line represents 95% confidence intervals), the area of gray square is proportional to the study's weight in the meta-analysis, and the pooled measure of effect is plotted as a diamond; the right-hand column numerically displays the effect measure, 95% confidence intervals, and weight of each study.
in patients with elevated CA 19-9 levels than in patients with normal levels [46]. The included patients in the Tachibana [23]. Future studies are needed to explore which kinds of tumor markers can be combined with CA 19-9 to increase diagnostic accuracy obviously.
Compared with the diagnostic value, the sensitivity for predicting recurrence was relatively higher. Marrelli et al. reported that the sensitivity for recurrence of CA 19-9 was 56%, with a specificity of 74% [61]. Kim et al. reported that the sensitivity for recurrence was 68.2% in advanced gastric cancer, with a specificity of 80.0% [34]. Moreover, the combination of CA 19-9 and other tumor markers provided more useful prognostic information. Marrelli et al. reported that the sensitivity increased to 87% when CA 19-9 was combined with CEA and CA 72-4 [61]. Huang et al. obtained similar results on the prediction of recurrence [9]. Thus, CA 19-9, a low-cost and convenient test, may be useful for Disease Markers 9 predicting prognosis and of use as an important indicator for high risk of recurrence.
There is no agreement on the optimal cut-off value for the prognostic value of CA 19-9. Most studies have used a cut-off value of 37 U/mL, but this cut-off value is often used for the diagnosis of gastric cancer and it is unknown whether this cut-off value is appropriate as a prognostic value. Yajima et al. determined the cut-off value of CA 19-9 as a predictor of prognosis via receiver operating characteristic analysis, and the cut-off value was ≥77 U/mL, which was approximately twice the upper limit of the normal range (37 U/mL) [20]. High-quality, well-designed multicenter studies are required to establish an optimal cut-off value for CA 19-9 and determine whether we should use different cut-off values for clinicopathologic and prognostic values.
The strength of the present meta-analysis lies in the inclusion of all relevant studies. Not only did we evaluate the clinicopathologic value of CA 19-9 in patients with gastric cancer, we also assessed the prognostic value of CA 19-9. Moreover, our results and conclusions were confirmed by indepth subgroup analyses. Thus, our results may be valuable for diagnosis, prediction of prognosis, and individual treatment in patient with gastric cancer.
There were several limitations in the present study. First, we could not obtain detailed individual information. Thus, we could not control other potential biases. Second, there was considerable heterogeneity and heterogeneity still could not be eliminated. Third, our study did not obtain a conclusive result regarding the optimal cut-off value and could not evaluate whether chemotherapy would impact the prognostic value of CA 19-9 due to a limited number of included studies. Therefore, future studies are urgently needed to assess the cut-off value of CA 19-9 and effects of chemotherapy on CA 19-9.

Conclusions
Our meta-analysis showed that CA 19-9 may indicate clinicopathologic characteristics of gastric cancer and is associated with poor prognosis. Further high-quality and large-scale studies are required to determine whether CA 19-9 can be used for the individualized treatment for patients with gastric cancer.

Ethical Approval
The study did not violate the local regulations and was approved by the Research Ethics Committee of China Medical University, China.