Significance of CT attenuation and F-18 fluorodeoxyglucose uptake of visceral adipose tissue for predicting survival in gastric cancer patients after curative surgical resection

This study was approved by the Institutional Review Board of our university. Because of the retrospective nature of the study, the requirement to obtain informed consent was waived. We retrospectively reviewed the medical records of patients who were diagnosed with gastric cancer and underwent surgical resection of the gastric cancer at our medical center between March 2012 and December 2016. Of these patients, 117 patients were finally enrolled in the study according to the following inclusion criteria: patients who (1) were histopathologically diagnosed with AGC (pT2-T4), (2) underwent pre-operative FDG PET/CT, and (3) subsequently underwent curative surgical resection of gastric cancer. We excluded patients (1) who were diagnosed with early gastric cancer (pT1) (n = 195), (2) who had distant metastatic lesions on staging work-up (n = 37), (3) in whom microscopic or macroscopic peritoneal seeding metastases were found during surgical exploration or malignant cells were found on peritoneal washing cytology (n = 9), (4) who had a previous history of malignancy or major abdominal surgery (n = 5), (5) who had CT images that were in appropriate for analyzing VAT (n = 3), and (6) who were not followed up for 2 years after surgery without event (n = 8). We also excluded two patients who had a rare histopathological type of gastric cancer (adenosquamous cell carcinoma).

All patients in the study underwent pre-operative examinations including physical examination, gastroduodenoscopy, contrast-enhanced abdominopelvic CT scan, FDG PET/CT scan, and blood test. Afterwards, curative surgical resection, subtotal or total gastrectomy with at least D2 lymph node dissection, was performed according to the treatment guidelines of the Japanese Gastric Cancer Association [1]. The median interval between FDG PET/CT and operation was 2.0 days (range 1.0–21.0 days). With surgical specimens, histopathological stages of all patients were assessed in accordance with the seventh edition of the American Joint Committee on Cancer staging guidelines. Gastric cancers were categorized into five histopathological subtypes: papillary adenocarcinoma, well-differentiated and moderately differentiated tubular adenocarcinoma, poorly differentiated adenocarcinoma, signet-ring cell carcinoma, and mucinous carcinoma [17]. Furthermore, all tumors were classified into two microscopic growth types according to the Lauren classification: intestinal and non-intestinal types. Diffuse, mixed, and non-classifiable types were included in the non-intestinal types [18]. According to the results of histopathologic assessment, adjuvant chemotherapy after surgery was recommended for patients who were diagnosed with TNM stage II and III. For adjuvant chemotherapy, capecitabine plus oxaliplatin (XELOX), TS-1 plus cisplatin, 5-fluorouracil plus cisplatin, or 5-fluorouracil plus oxaliplatin (FOLFOX) were performed. After curative surgery, follow-up examinations including gastroduodenoscopy, contrast-enhanced abdominopelvic CT scan, and blood tests (including serum carcinoembryonic antigen and cancer antigen 19–9) were performed every 6–8 months in the first 3 years and, subsequently, every 10–12 months. For patients who received adjuvant treatment, follow-up examinations were performed mid-chemotherapy and after completion of the adjuvant treatment. Afterwards, those patients with adjuvant treatment underwent routine follow-up programs. Cases with cancer recurrence during follow-up were categorized into three groups according to the site of cancer recurrence: locoregional recurrence (remnant stomach, site of anastomosis, gastric bed, regional lymph node, or peritoneum around the surgical site), peritoneal recurrence (peritoneal seeding nodule, increased peritoneal density, massive ascites, abnormally increased intestinal wall thickness, peribiliary infiltration, or Krukenberg tumor) and distant recurrence (distant lymph node and organ metastasis) [2].

All FDG PET/CT scans were performed using a Biograph mCT 128 scanner (Siemens Healthcare, Knoxville, TN, United States). Before FDG PET/CT scan, all patients were instructed to fast for at least 6 h and only patients who showed blood glucose level of less than 200 mg/dL before FDG administration underwent PET/CT scanning. Sixty minutes after the intravenous injection of FDG at a dose of approximately 4.07 MBq/kg, FDG PET/CT scanning was performed from the skull base to the proximal thigh. Non-contrast-enhanced CT scan was initially performed for attenuation correction at 100 mA and 120 kV_p. Subsequently, PET scanning was performed at 1.5 min per bed position in the three-dimensional acquisition mode. PET images were reconstructed with a point-spread-function based Gauss and Allpass filter algorithm and time-of-flight reconstruction with attenuation correction.

Two board-certified nuclear medicine physicians retrospectively evaluated FDG PET/CT images using a United States Food and Drug Administration-approved medical image viewer, OsiriX MD 10.0 software (Pixmeo, Geneva, Switzerland) without knowing clinical outcomes of the patients. First, FDG uptake of primary gastric cancer lesions was measured. A volume of interest (VOI) was manually drawn over the primary cancer lesion on transaxial PET/CT images and the maximum standardized uptake value (SUV) of primary cancer lesion (SUVmaxT) was measured. SUV was calculated as [decay corrected activity (kBq) per tissue volume (mL)]/[injected FDG activity (kBq) per body mass (g)]. In patients who had gastric cancer lesions with no discernibly increased FDG uptake, VOI was drawn in accordance with the tumor location seen on contrast-enhanced abdominopelvic CT images and gastroduodenoscopy. Afterwards, CT attenuation and FDG uptake of VAT on PET/CT images were measured (Fig. 1). To avoid the influence of the primary cancer lesion, three consecutive slices of the CT images at the level of the L4–L5 vertebrae, which is remote from the gastric cancer lesion, were selected for image analysis. VAT area, which is defined as an area with a CT attenuation range between − 200 and − 50 Hounsfield units (HU) in the intra-abdominal fat tissue, on those three slices of CT images was automatically identified (Fig. 1b) [12, 19]. Mean value of CT attenuation of the VAT area was measured (Fig. 1c). The VAT area on CT images was then exported to the corresponding PET images to measure FDG uptake of VAT. Physiologic FDG activities in the urine, bowels, and vessels that could affect the measurement of FDG uptake of VAT due to bowel movement during PET/CT scanning as well as the spillover FDG activity were manually removed. The mean SUV of the VAT area was then measured (Fig. 1d).

To assess the reproducibility of measurement of VAT attenuation and SUV between the two readers, concordance correlation coefficients for VAT parameters were calculated. From the blood tests results obtained at staging work-up, the neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR) were calculated for each patient. The Komogorov–Smirnov test was performed to evaluate the normality of distribution for continuous variables. Pearson’s correlation coefficients were calculated to evaluate the relationship between VAT parameters and other continuous variables. The Kruskal–Wallis test and Mann–Whitney U test were performed to evaluate differences of VAT parameters between patient groups. For survival analysis, the significance of the predictive value of each variable for RFS, peritoneal RFS, and OS was investigated using Cox proportional hazards regression test for univariate and multivariate analyses. Bonferroni adjustment was applied to correct for multiple testing. Survival time was defined as time from the day of curative surgical resection to the day of recurrence, peritoneal recurrence, or death. Patients who had no recurrence or survived were censored at the day of the last follow-up visit at our medical center. For continuous variables, optimal cut-off values were selected from the receiver-operating-characteristic (ROC) curve analysis, and all continuous variables were categorized into two groups according to those cut-off values. Survival curves for VAT parameters were estimated using the Kaplan–Meier method to calculate cumulative RFS, peritoneal RFS, and OS. All statistical analyses were performed using MedCalc Statistical Software version 18.10.2 (MedCalc software bvba, Ostend, Belgium). A p value of less than 0.05 was considered statistically significant.

Clinical characteristics of the enrolled 117 patients are summarized in Table 1. Among these 117 patients, 90 patients (76.9%) showed increased FDG uptake in the gastric cancer lesion, whereas the remaining 27 patients (23.1%) showed no discernible increase in FDG uptake of the primary tumor lesion. Further, the SUVmaxT of those 17 patients was measured based on contrast-enhanced CT and gastroduodenoscopy findings. At the time of analysis, 42 patients (35.9%) experienced cancer recurrence after curative surgical resection and 23 patients (19.7%) died. The median duration of follow-up was 37.0 months (range 2.6–77.2 months). Of the 42 patients with cancer recurrence, both peritoneal recurrence and distant recurrence were found in 20 patients (47.6%) and locoregional recurrence was found in 10 patients (23.8%). Eight patients (19.0%) had two sites of recurrence at the time of cancer recurrence diagnosis (peritoneal and distant recurrence, 4 patients; locoregional and peritoneal recurrence, 2 patients; locoregional and distant recurrence, 2 patients). Of the 42 patients showing recurrence, 14 patients (33.3% of the patients with recurrence) experienced peritoneal recurrence without locoregional or distant recurrence.

On the assessment of reproducibility of VAT parameters, there was substantial agreement between the two readers for the measurement of VAT attenuation (concordance correlation coefficient: 0.994, 95% confidence interval: 0.990–0.997) and VAT SUV (concordance correlation coefficient: 0.988, 95% confidence interval: 0.979–0.995).

CT attenuation and SUV of VAT showed a significant positive correlation with each other (p < 0.001, r = 0.799). In correlation analyses of VAT attenuation and SUV with stage, tumor factors, serum inflammatory markers, and clinical outcomes, increased VAT attenuation and SUV were observed in patients with advanced tumors and worse clinical outcomes (Table 2). For T stage and TNM stage, Kruskal–Wallis test showed significant differences of VAT attenuation and SUV between patients with T2, T3, and T4 stages and between patients with stage I, II, and III (p < 0.05). On post hoc analysis, patients with T4 stage and stage III had significantly higher VAT attenuation and SUV than those with T2 and T3 stages and those with stage I and II, respectively (p < 0.05). Patients with lymph node metastasis also showed significantly higher VAT SUV than those with no lymph node metastasis (p = 0.034). For tumor factors, there were significant, but, weak positive correlations between tumor size and both VAT parameters and between SUVmaxT and VAT SUV (p < 0.05). In contrast, no significant differences of VAT attenuation and SUV were shown according to the tumor histopathology and Lauren classification (p > 0.05). For serum inflammatory markers, only PLR showed significant positive correlations with both VAT attenuation and SUV (p < 0.05). The comparisons of VAT attenuation and SUV were also performed between 38 patients who had experienced cancer recurrence at 3-year after surgery and 62 patients who were on regular follow-up at 3-year after surgery with no recurrence, excluding 17 patients who were censored within 3-year after surgery. The results of the comparison analysis revealed that patients with recurrence during the first 3-year after surgery had significantly higher VAT attenuation (p = 0.011, − 89.66 ± 9.23 vs. − 94.11 ± 7.73) and VAT SUV (p = 0.004, 0.54 ± 0.16 vs. 0.45 ± 0.14) than those with no recurrence during the first 3-year.

The prognostic values of VAT attenuation and SUV for predicting RFS, peritoneal RFS, and OS were assessed along with age, sex, tumor factors, serum inflammatory markers, and SUVmaxT. For continuous variables, optimal cut-off values were selected by the ROC curve analysis, which were 60 years for age, 5.0 cm for tumor size, 5.00 mg/dL for serum C-reactive protein (CRP) level, 2.30 for NLR, 173.76 for PLR, 4.65 for SUVmaxT, − 90.60 for VAT attenuation, and 0.46 for VAT SUV. ROC curve analysis revealed that the sensitivity, specificity, positive predictive value, and negative predictive value of VAT attenuation were 54.8%, 70.7%, 51.1%, and 73.6% for recurrence; 60.0%, 67.0%, 27.2%, and 89.0% for peritoneal recurrence; and 78.3%, 73.4%, 41.9%, and 93.2% for death, using a cut-off value of − 90.60. The sensitivity, specificity, positive predictive value, and negative predictive value of VAT SUV were 83.3%, 38.7%, 43.2%, and 80.6% for recurrence; 70.0%, 58.8%, 25.9%, and 90.5% for peritoneal recurrence; and 91.3%, 63.8%, 38.2%, and 96.8% for death, using a cut-off value of 0.46.

On univariate survival analysis, both VAT attenuation and SUV were significantly associated with RFS, peritoneal RFS, and OS (p < 0.05; Table 3). Patients with high VAT attenuation and SUV had significantly worse survival than those with low values (Figs. 2 and 3). In addition to VAT attenuation and SUV, T stage, presence of lymph node metastasis, tumor size, and NLR were significant predictors for RFS, peritoneal RFS, and OS (p < 0.05). SUVmaxT and PLR were associated with only RFS and OS, and tumor histopathology was a significant predictor for only peritoneal RFS (p < 0.05).

The variables that showed statistical significance (p value of < 0.05) for predicting RFS, peritoneal RFS, and OS on univariate analysis were selected for multivariate analysis (Table 4). Bonferroni correction for multiple testing was performed for the multivariate analysis, and p value of less than 0.0083 was considered as statistically significant. VAT attenuation and SUV were evaluated on separate models, because both were significantly associated with each other. For serum inflammatory markers, only NLR was included in the analysis, because NLR and PLR showed significant association with each other. VAT attenuation remained as significant positive predictors for peritoneal RFS (p = 0.008) and OS (p = 0.007) and VAT SUV showed significant association only with OS (p = 0.004). Along with VAT attenuation and SUV, tumor histopathology (p = 0.008) was an independent predictor for peritoneal RFS and T stage (p = 0.006) was an independent predictor for OS. In contrast to peritoneal RFS and OS, both VAT attenuation and SUV failed to show statistical significance for predicting RFS (p > 0.0083) on multivariate analysis, and only T stage showed significant association with RFS (p < 0.0083).