This retrospective study was approved by the Institutional Review Board of our hospital (No.2020KT121). The requirement of informed consent was waived off.

The present study enrolled 111 consecutive gastric cancer patients with synchronous PM, who were admitted to our institution and treated with systemic chemotherapy between September 2009 and December 2018. The inclusion criteria were as follows: (i) histopathologically confirmed gastric adenocarcinoma; (ii) availability of complete medical records; (iii) synchronous PM diagnosed by baseline CT or diagnostic laparoscopy and cytology (21); and (iv) patients treated with systemic chemotherapy. The exclusion criteria were as follows: (i) unavailable baseline abdominal and pelvic CT (n = 19); (ii) baseline CT performed >30 days before chemotherapy (n = 9); (iii) any abdominal invasive procedures performed before baseline CT (the peritoneal scar caused by invasive examinations may mimic the PM, n = 11); (iv) for patients diagnosed by diagnostic laparoscopy, CT was performed beyond 30 days before the laparoscopy (n = 2); (v) multiple primary cancers (n = 2); (vi) liver cirrhosis (n = 1); and (vii) previous abdominal inflammatory diseases (n = 1). Finally, 66 patients were selected for further analyses. Among them, 57 patients were diagnosed by baseline CT, while the remaining nine were diagnosed by diagnostic laparoscopy. The primary clinical and pathological features of these patients, which included their age, gender, ECOG PS, histology type, degree of differentiation, chemotherapy regimens, the location of the primary tumor, and the Borrmann type, were collected from their medical records ( Table 1 ).

The posttreatment follow-up of the patients was performed every three months in the hospital, outpatient clinic, or via telephone. These follow-ups lasted until the death of the patients or the cut-off date (March 22, 2021), whichever was earlier. The outcome of the present study was the OS, which was defined as the time from the diagnosis of PM to the death of the patients or their last follow-up.

All patients underwent abdominal and pelvic contrast-enhanced CT examinations, using either the LightSpeed 64 VCT (GE Medical Systems, Milwaukee, WI, USA) or Discovery CT750 HD scanner (GE Medical Systems), after fasting for more than 6 h. To reduce gastrointestinal motility, 10 mg of anisodamine (654-2; Hangzhou Minsheng Pharma, China) was intramuscularly administered 15–20 min prior to the CT examination. Next, 6 g of gas-producing crystals were orally administered with 10 mL of warm water shortly before the CT examination to distend the stomach. The patients were scanned in the supine position. The scan range was from the diaphragmatic dome to the lower margin of the pubis. The following imaging parameters were used: peak tube voltage, 120 kVp; tube current, automatic; collimation thickness, 64.000×0.625 mm; helical pitch, 0.984:1.000, and reconstructed thickness, 5 mm. A nonionic contrast material was injected through the antecubital vein at a rate of 3.5 mL/s (1.5 mL/kg of body weight; iohexol: 300 mg I/mL; Omnipaque, GE Healthcare). The arterial and venous phase scans were performed at 40 s and 70 s after the contrast media injection. The multiplanar reconstruction (MPR) images were obtained with a slice thickness of 5 mm.

Two radiologists (doctor A and doctor B, with 15 and 4 years of experience in gastric cancer imaging, respectively), who were blinded to the prognosis data of patients, independently reassessed the baseline multiplanar CT images of all patients to evaluate the following CT characteristics and assess interobserver reproducibility. Finally, Any discrepancy was resolved through a third radiologist (doctor C, with 20 years of experience in gastric cancer imaging).

The images were reviewed with the dynamic adjustment of the window width and level on Picture Archiving and Communication Systems (PACS) workstations. The wide window width was adjusted to display the tiny grainy background noise of fat tissues to highlight the PM signs (22) ( Figures 1A, B ). A combination of MPR images (axial, coronal, and sagittal planes) was employed to precisely locate the peritoneal regions ( Figures 1C, D ). The following peritoneal regions were evaluated ( Figure 2 ) (11, 23, 24): (1) gastrosplenic ligament (GSL); (2) gastrohepatic and hepatoduodenal ligaments (GHL and HDL); (3) gastrocolic ligament (GCL); (4) perihepatic visceral peritoneum; (5) mesentery; (6) greater omentum; (7) superior parietal peritoneum; (8) posterior parietal peritoneum; (9) lateral parietal peritoneum; and (10) pelvic parietal peritoneum. For the score, 1 point was given for the presence of typical PM signs on one peritoneal region, and 0 points were given for its absence. The points obtained from all 10 regions were accumulated to determine the total point for one patient. Thus, the CT-PMS was quantified within the range of 0–10. A CT-PMS of 10 indicated that all 10 peritoneal regions had a typical metastasis sign, while that of 0 indicated the absence of metastasis signs in all peritoneal regions. The typical PM signs on CT included diffuse fibrosis strands, peritoneal nodules, omental cake on the visceral peritoneum, and focal or diffuse thickening with enhancement along the parietal peritoneum ( Table 2 ; Figure 2 ) (13–17).

A number of studies have reported that tumor implants to specific peritoneal sites may narrow or even occlude the nearby cavity organs and cause obstruction of the proximal canal, including biliary obstruction (through the PM of the GHL and HDL), and urothelial obstruction (through the PM of posterior parietal peritoneum) (17, 25, 26). Bowel obstruction included either intestine infiltrated by mesentery PM, or colon infiltrated by GCL and greater omentum PM (26, 27). The presence of obstruction of the bile duct, bilateral urothelial tract, and bowel was evaluated ( Figure 3 ). For the bile duct and bilateral urothelial tract, obstruction caused by tumor infiltration may manifest as a focal or diffuse thickening and hyperenhancement of the tract wall, along with the dilatation of the proximal canal. Bowel obstruction included mechanical obstruction, which presents as a focal or diffuse thickening and hyperenhancement of the bowel wall, along with the dilatation of the proximal bowel, and functional obstruction due to tumor infiltration of nerves in the peritoneum, which presents as extensive dilatation without focal irregular thickening and enhancement of the wall (25, 27). If one or more of these tracts presented with the abovementioned manifestations, a triple tract dilatation (TTD) sign was considered to be present. Notably, patients with other causes of dilatation (such as inflammatory disease and cholelithiasis) were considered to be absent from TTD sign.

The presence of extensive lymph node metastasis (ELM) included the enlargement of the lymph node around para-aortic (PAN) and around the celiac artery and its branches (bulky N) (28). PAN was determined when No. 16 lymph node had a short diameter larger than 10 mm. Bulky N was determined when the NO.8, 9, and 11 lymph nodes had a short diameter larger than 30 mm, or at least two adjacent lymph nodes with a short diameter of each larger than 15 mm (28). We recorded the presence of ELM that met the above criteria.

The grade of ascites was scaled as follows (29): grade 0 = ascites not detected by CT scan; grade 1 = ascites located only in the upper or lower abdominal cavity; grade 2 = neither grade 1 nor grade 3; grade 3 = ascites extending throughout the abdominal cavity.

The statistical analysis was conducted using RStudio 3.5.0 (R Foundation for Statistical Computing, Vienna, Austria) and SPSS 20.0 (IBM, Armonk, USA). Continuous variables were presented as the median with interquartile ranges (IQR). Categorical variables were shown as numbers with percentages. The optimal cut-off value for the CT-PMS was determined using the X-tile 3.6.1 (Yale University School of Medicine, New Haven, CT, USA) to build the classified CT-PMS. Survival curves were calculated using Kaplan–Meier analysis and compared using log-rank test. Bonferroni correction was used to adjust P value for multiple comparisons. Univariable and multivariable Cox proportional hazards models were used to identify the prognostic factors for OS. Values with P <.10 on the univariate analysis were included in the multivariate analyses. The agreement between two radiologists was assessed using the kappa statistic for categorical variables and the weighted kappa statistic for ordinal categorical variables. The values of 0.81–1.00, 0.61–0.80, 0.41–0.60, 0.21–0.40, and <0.20 were indicated to be in almost perfect agreement, substantial agreement, moderate agreement, fair agreement, and slight agreement, respectively (30). A statistically significant difference was reported when P <.05 (two-sided).

A total of 66 gastric cancer patients with synchronous PM were included in the present study. The median follow-up period was 319 days (IQR: 171–457 days). All patients had died by the time of study closure. The median age of these patients was 57 years (IQR: 48–64 years). Most of these patients had diffuse-type (n = 39, 59.10%) and poor differentiation (n = 48, 72.73%) on histopathology. Furthermore, all patients received palliative systemic chemotherapy, alone or in combination with supportive care. The clinicopathological characteristics of these patients are summarized in Table 1 . The most commonly involved region was the greater omentum (n = 35, 53.00%), followed by the GCL (n = 31, 47.00%), lateral parietal peritoneum (n = 22, 33.30%), posterior parietal peritoneum (n = 19, 28.80%), mesentery (n = 17, 25.80%), GHL and HDL (n = 17, 25.80%), GSL (n = 16, 24.20%), pelvic parietal peritoneum (n = 12, 18.20%), superior parietal peritoneum (n = 6, 9.10%), and perihepatic visceral peritoneum (n = 2, 3.00%).

The optimal cut-off value for the CT-PMS was 2 by X-tile. There were 39 patients with a CT-PMS of 0–2 and 27 patients with a CT-PMS of 3–10. The median OS of patients with a CT-PMS of 0–2 was 338 days (IQR: 255–546 days), while that of the remaining patients was 258 days (IQR: 100–390 days). A significant difference in the OS was perceived between patients with CT-PMS of 0–2 and 3–10 (P = .02, log-rank test; Figure 4A ).

Eight patients presented with TTD sign. Among these patients, three had biliary dilatation, three had bilateral urothelial dilatation, and the other two had bowel dilatation. The median OS of patients with the presence of TTD sign was 97 days (IQR: 73–292 days), while that of the remaining patients was 335 days (IQR: 186–460 days). A borderline significant difference in the OS was perceived between patients with and without TTD sign (P = .07, log-rank test; Figure 4B ). Furthermore, we divided patients into three groups including five patients with both CT-PMS of 3–10 and the presence of TTD sign, 25 patients with either CT-PMS of 3–10 (n=22), or the presence of TTD sign (n=3) and 36 patients with only CT-PMS of 0–2. A significant difference in the OS was perceived between patients with CT-PMS of 0–2 and patients with both CT-PMS of 3–10 and the presence of TTD sign (P = .00, log-rank test; P value has been adjusted by Bonferroni correction and P <.02 [0.05/3] was considered statistically significant difference; Figure 4C ).

Eighteen patients presented with ELM. Among these patients, ten patients had both PAN and Bulky N, five had Bulky N, and the other three had PAN. The median OS of patients with ELM was 252 days (IQR: 104–390 days), while that of the remaining patients was 340 days (IQR: 186–476 days). A borderline significant difference in the OS was perceived between patients with and without ELM (P = .07, log-rank test; Figure 4D ). Furthermore, we divided patients into three groups including 11 patients with both CT-PMS of 3–10 and ELM, 23 patients with either CT-PMS of 3–10 (n=16), or ELM (n=7), and 32 patients with CT-PMS 0–2. A significant difference in the OS was perceived between patients with CTPMS of 0–2 and patients with both CT-PMS of 3–10 and ELM (P = .01, log-rank test; P value has been adjusted by Bonferroni correction and P <.02 [0.05/3] was considered statistically significant difference; Figure 4E ).

There were 10 patients with ascites of grade 0, 26 patients with ascites of grade 1, 23 patients with ascites of grade 2, and seven patients with ascites of grade 3. There was no optimal cut-off value for the grade of ascites by X-tile. The median OS for patients with grade 0 ascites was 315 days (IQR: 157–531 days), for patients with grade 1 ascites was 323 days (IQR: 255–460 days), for patients with grade 2 ascites was 270 days (IQR: 161–390 days), while for the remaining patients with grade 3 ascites was 364 days (IQR: 91–690 days). There was no significant difference in the OS between patients with different grades of ascites ( Figure 4F ).

In the univariate Cox analysis, the age, gender, ECOG PS, histology type, chemotherapy regimen, location of the primary tumor, Bormann type, and the grade of ascites did not have a prognostic impact ( Table 1 ), while the CT-PMS and the classified CT-PMS (dichotomized by the optimal cut-off value of CT-PMS) were associated with the OS (hazard ratio [HR]: 1.10; 95% confidence interval [CI]: 1.00–1.21; P = .049; HR: 1.84; 95% CI: 1.10–3.07; P = .02). Differentiation, the presence of TTD sign, and the presence of ELM showed a predictive trend for OS, but failed to achieve statistical significance (HR: 1.77; 95% CI: 1.00–3.13; P = .05 for differentiation; HR: 1.97; 95% CI: 0.93–4.18; P = .08 for TTD sign; HR: 1.67; 95% CI: 0.96–2.90; P = .07 for ELM). In the multivariate analysis included CT-PMS (without the classified CT-PMS), after controlling for differentiation, TTD sign, and ELM, the CT-PMS was identified to be independently associated with survival (HR: 1.10; 95% CI: 1.00–1.21; P = .049). In the multivariate analysis included classified CT-PMS (without the CT-PMS), after controlling for differentiation, TTD sign, and ELM, the classified CT-PMS was identified to be independently associated with survival (HR: 1.84; 95% CI: 1.10–3.07; P = .02).

The interobserver agreement for assessing PM signs of the individual regions varied from fair to substantial: κ = 0.62 (95% CI: 0.41–0.83) for the GSL, κ = 0.57 (95% CI: 0.34–0.81) for the GHL and HDL, κ = 0.64 (95% CI: 0.45–0.82) for the GCL, κ = 0.21 (95% CI: -0.18 to 0.61) for the perihepatic visceral peritoneum, κ = 0.74 (95% CI: 0.56–0.92) for the mesentery, κ = 0.70 (95% CI: 0.52–0.87) for the greater omentum, κ = 0.42 (95% CI: 0.09–0.75) for the superior parietal peritoneum, κ = 0.64 (95% CI: 0.44–0.84) for the posterior parietal peritoneum, κ = 0.60 (95% CI: 0.39–0.81) for the lateral parietal peritoneum, and κ = 0.58 (95% CI: 0.30–0.85) for the pelvic parietal peritoneum. Although the interobserver agreement was suboptimal in individual peritoneal regions, the summation of the involved regions compensated for the differences. The CT-PMS and the classified CT-PMS evaluated by the two radiologists had a substantial agreement (weighted kappa value [κ _w] = 0.63 [95% CI: 0.54–0.72] for CT-PMS; κ = 0.77 [95% CI: 0.63–0.92] for the classified CT-PMS). The interobserver agreement for assessing tract dilatation varied from substantial to almost perfect: κ = 0.66(95% CI: 0.04–1.28) for bowel dilatation, κ = 0.65 (95% CI: 0.20–1.10) for bile dilatation, κ = 1.00 (95% CI: 1.00–1.00) for urothelial dilatation and κ _w = 0.77 (95% CI: 0.53–1.02) for the TTD sign. The interobserver agreement for assessing ELM was almost perfect: κ = 0.84 (95% CI: 0.68–0.99). The interobserver agreement for assessing ascites grade was substantial: κ _w = 0.79 (95% CI: 0.67–0.91).