Gastric cancer is one of the most common types of cancer worldwide and is the second leading cause of cancer-related death, with approximately 700,000 deaths annually [
]. Although major improvements have been achieved in the early detection and screening of gastric cancer, many individuals are still diagnosed with advanced-stage gastric cancer every year, which underscores the poor prognosis of the disease [
]. Therefore, a practical method that can precisely predict the survival outcome of patients with gastric cancer is essential, because stratification of patients with potential survival outcomes could influence the treatment decision.
During the 1980s, positron emission tomography (PET) was incorporated into the clinical practice [
]. FDG-PET uses
F–FDG), a glucose analog, as tracer to evaluate the metabolic status of the morphological lesions. In order to quantify a lesion’s metabolic activity, standardized uptake value (SUV) is introduced to clinical practice. The SUV value provides a semi-quantitative analysis and description of the radioactivity in a lesion [
]. In practical work, a circular region of interest placed in the FDG-accumulating area was selected to obtain the SUV value. Because of the increased glycolytic activity of cancer cells, this imaging technique has been recently used for the detection of primary and metastatic lesions in the field of oncology, particularly in gastric cancer [
Furthermore, recent studies [
] have shown a significant relationship between prognosis and pretreatment PET imaging. This finding revealed that patients with a high standardized uptake value (SUV) had a worse prognosis than individuals with low SUV. This was confirmed in several types of cancer, including esophageal cancer and non-small cell lung cancer [
]. However, some studies [
] presented controversial conclusions for gastric cancer and a comprehensive analysis of the association between SUV and prognosis of gastric cancer have not yet been conducted. Therefore, this meta-analysis aimed to assess whether high SUV can be used as a prognosis predictor in patients with gastric cancer.
We systematically searched the databases PubMed, EMBASE, the Cochrane library, and Web of Science for relevant articles from January 1975 to February 2016. We used the keywords “gastric cancer”,” stomach neoplasm”,” gastric carcinoma”,” stomach cancer”,” PET”,” positron emission tomography”,” 18F- FDG”,” 18-Fluoro-deoxy-glucose”, “F-18-fluorodeoxyglucose” and “2-Fluoro −2-deoxy-D-glucose” to summarize our search strategy. Moreover, we expanded our search by screening the references of relevant studies for additional studies that might be useful in our meta-analysis.
Inclusion criteria and exclusion criteria
To keep our analysis accurate and reliable, we used the following inclusion criteria: (i) The studies reported at least one of the following outcome measures of interests: overall survival, recurrence-free survival and progression-free survival; (ii) a PET scan was performed prior to treatments, including chemotherapy, surgical therapy, and radiotherapy; (iii) studies only published in English with full-texts available were included. (iv) Studies contained a clear description of the PET protocol and reported the SUVmax or SUV mean of 18F–FDG. When several studies from the same authors or institutions were available, the meta-analysis included the most recent or highest-quality study.
Studies were excluded for the following reasons: (i) the prognostic information of patients was not reported in the studies; (ii) the hazard ratio (HR) could not be calculated considering the originally published data; (iii) the studies included patients diagnosed with gastro-esophageal junction carcinoma or gastrointestinal cancer; (iv) Studies were excluded if they only focused on the SUV of metastatic lymph nodes, surgical anastomoses or distant metastatic sites rather than primary tumor lesions.
Data extraction and assessment of the study quality
Two investigators (Z.H. Wu and J.H. Zhao) independently reviewed the enrolled studies. Any discrepancies were presented to a third author and resolved through discussions among these investigators. The primary elements extracted consisted of the following: (1) the FDG avidity, which was defined as the focally increased 18F–FDG uptake exceeding the surrounding normal tissue [
]; (2) types of SUV, correction of SUVs, definition of threshold SUVs; (3) HR associated with the FDG uptake value for overall survival (OS), recurrence-free survival (RFS), progression-free survival (PFS), and their respective 95% confidence interval (CI). The main outcomes of our analysis were the pooled HRs for OS, RFS, and PFS.
We evaluated the quality of the enrolled studies according to the Newcastle–Ottawa scale (NOS) [
]. Studies that scored ≥7 of a maximum possible score of nine were regarded as high-quality trials whereas those scored ≥5 were recognized as moderate-quality trials. Furthermore, to systematically assess the methodological quality and ensure that the enrolled studies were accurate and reliable, we further evaluated them using a quality scale that was applied in a previous study [
]. This scale was composed of four categories: scientific design, generalizability, analysis of results and analysis of PET reports [
]. Each category contained several items, and each item was assigned values zero, one or two. And each category had a maximum score of 10 points.
Review manager software version 5.2 (Cochrane Collaboration) was used to analyze the data collected from each study. To evaluate the prognostic effect representatively, we used the HR or estimated relative risk (RR) and their corresponding 95% CI as the effect variable. In cases in which we could not acquire the HR and its 95% CI explicitly, several relatively accurate methods reported by Tierney et al. [
] were used to calculate these values using data available in the literature.
The heterogeneity among the studies was calculated using the Cochrane Q-test and a value of I
indicated the degree of heterogeneity [
]. In cases of lack of significant heterogeneity (I
< 50%) among the studies, a fixed effect model was chosen for the meta-analysis [
]. Otherwise, a random effects model was used [
]. Publication bias was examined via the analysis of funnel plots [
]. In our meta-analysis, we calculated the pooled HR for OS, RFS, and PFS.
In recent decades, FDG-PET has been widely used in clinical practice for staging patients with cancer and for detecting local and distant metastasis [
]. In recent years, several studies [
] in the field of gastric cancer have focused on the prognostic value of metabolic activity detected by pretreatment FDG-PET. However, whether the SUV of primary tumors is a prognostic factor in patients with gastric cancer remains unclear. Some studies [
] that investigated the prognostic value of SUV in gastric cancer found a significant prognostic value of high SUV whereas other studies [
] did not find any evident relationship between SUV and prognosis.
A meta-analysis was the statistical pooling of the outcomes identified in individual studies. Therefore, it can increase the precision of the estimated effect of the individual studies and consequently elucidate the relationship between the observed variables and the outcomes and can eventually be applied in clinical practice [
]. In the present meta-analysis, the analysis of the pooled HR for OS indicated that patients with a high SUV had higher risk of death than those with low SUV on the basis of the threshold values. Moreover, we found that high SUV was an important factor for predicting RFS. These findings are important because this is the first meta-analysis of studies with controversial opinions on the prognostic value of SUV in primary lesions of gastric cancer. To date, patient characteristics like tumor size, cancer staging, and the status of local or distant metastasis have been widely acknowledged as significant prognostic factors for gastric cancer [
]. As SUV values held an advantage that they can reflect the metabolic status of lesion compared with other diagnostic methods, SUV and patient characteristics can be synergistically used to predict prognosis. Meaningfully, a previous study [
] showed combining SUV value and pT stage could increase the value of SUV for predicting prognosis. Therefore, our study provided a direction towards studying on the prognostic role of combining SUV and patients characteristics like pT stage or others.
The standardized uptake value (SUV) was introduced for quantitative analysis. For calculating the SUV value, regions of interest were selected from primary tumor lesion in the trans-axial PET image where the lesion seemed to have the most intense FDG uptake. In that way, SUVmean represents the mean value of the SUVs within the selected regions of interest and SUVmax is the largest value among the SUVs of the selected regions of interest. And both SUVmean and SUVmax can reflect the SUV values of the tumor lesions. Our subgroup analysis based on SUV types indicated that there is a significant relationship between high SUV and poor prognosis in the subgroup applying the SUVmax as the SUV value of tumor lesion. However, in the subgroup using the SUVmean, the relationship is not significant. This can be explained by that when using the SUVmean, it is more likely for us to neglect the larger SUVs among the regions of interest. This outcome also reminded us that we should give priority to SUVmax when designing a study focusing on the SUV value and prognosis.
The enrolled studies used several methods to determine the threshold SUV. Some studies [
] used the median or mean value (other methods) as the cutoff value because they argued that the ROC method tended to generate many false-positive results. Other studies [
] used ROC curves. In addition, the results of the subgroup analysis indicated that, in the studies that applied the ROC method, the patients with high SUV had a pooled 1.77-fold higher risk of death (Fig.
) whereas the studies that used other methods found no significant relationship between SUV and prognosis. Moreover, a study suggested that the ROC method could help identify the most appropriate threshold value [
In the current stage, because of inconsistent PET techniques, use of different PET protocols, and differences in patient characteristics depending on the geographical region evaluated, it was extremely difficult for different medical centers to find a consistent threshold value to distinguish patients with high or low SUV. Previous studies pointed out that a value of 2.5 could be used as the cutoff value for tumor delineation [
]. The threshold values in the studies evaluated herein varied between 4.6 and 8.2, indicating that the ROC method was the ideal method to determine the cutoff value; therefore, these studies provided a strategy that allowed the ROC method to be consistently used to calculate the cutoff value. Our study provides a direction and evidence that SUV value is a potential parameter for prognosis predicting. In this respect, the obtaining of a consistent threshold value from all the available studies will better serve the prediction of the prognosis in the future. Therefore, to obtain an accurate standard threshold values in clinical practice, further studies are needed to formulate a systemic PET protocol assessing standard to get a consistent cutoff value and eventually promote the utilization of SUV in predicting prognosis.
During the selection of eligible studies, we chose the studies that investigated disease prognosis and determined the pretreatment SUV. In this manner, the treatment strategies used subsequently in the patients included in these studies were noteworthy because these therapies played a key role in determining prognosis. Our subgroup analysis based on these treatment strategies suggested that high SUV was associated with poor prognosis in both the subgroups and indicated that the prognosis-predicting value of pretreatment SUV was not affected by the subsequent therapies used in patients with gastric cancer.
The analysis of the full texts of the included studies indicated that, in some studies, the authors excluded the patients with a history of diabetes mellitus [
]. Previous studies on bronchial and cervical cancer pointed out that diabetes had no significant influence on the uptake of FDG [
]. However, in other studies [
], a long-term high level of blood glucose had an impact on the uptake of FDG, thereby affecting its detection via PET. Furthermore, some studies [
] suggested that pre-existing diabetes mellitus increased the risk of gastric cancer, which indicates that many patients with gastric cancer may also present with diabetes mellitus. Unfortunately, to date, no published studies have focused on the influence of diabetes on PET performance in patients with gastric cancer. Therefore, further studies are needed to elucidate this influence.
This is the first meta-analysis that evaluated the prognostic value of SUV in patients with gastric cancer. We acknowledge some limitations in our study. First, the number of studies included in our meta-analysis was relatively small and eight papers included this meta-analysis were regarded as moderate quality according to NOS. We expect that more high-quality studies on this issue will be published in the future. Second, because of the low morbidity observed in Western countries, most studies were confined to eastern Asia, which could decrease the representativeness of our results for gastric cancer on a global scale. In addition, we hope that more studies from different centers and different geographical regions will be carried out in the future. Third, as mentioned above, we were unable to find a fixed cutoff value to distinguish the patients with high or low SUV. Therefore, more studies with data from individual patients are needed to obtain standard threshold values for predicting the prognosis of gastric cancer. However, despite these limitations, this is the first meta-analysis that evaluated the prognostic value of SUV in gastric cancer. We found that patients with high pretreatment SUV tended to have poor prognosis. Furthermore, our results suggested that the ROC method could better define a threshold value.
We thank the department of Surgical Oncology of the First Affiliated Hospital of China Medical University for Technical Assistance.
This work was supported by Natural Science Foundation of Liaoning Province (No. 2014029201), Program of Education Department of Liaoning Province (L2014307), the Key Laboratory Programme of Education Department of Liaoning Province (LZ2015076) and Scientific Programme of Science & Technology Department of Liaoning Province (2015225002). The funding bodies have no roles in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.
Availability of data and materials
The data supporting the founding of this paper are presented in this manuscript (i.e. Tables, Figure and Reference).
ZNW, ZHW and JHZ designed the research, analyzed the results and wrote the manuscript. ZHW and JHZ searched the studies and collected the eligible studies. YXS contributed on solving the discrepancies between ZHW and JHZ on selecting studies. PG and XWC extracted and analyzed the data. YXS, JXS and BM wrote parts of the manuscript and revised this manuscript. ZNW contributed to manuscript revision and supervised all the work. All authors read and provided suggestions during manuscript preparation.
The authors declare that they have no competing interests.
Consent for publication
Ethics approval and consent to participate
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.