A nomogram based on glycomic biomarkers in serum and clinicopathological characteristics for evaluating the risk of peritoneal metastasis in gastric cancer

Abstract

Background: Peritoneal metastasis (PM) in gastric cancer (GC) remains an untreatable disease, and is difficult to diagnose preoperatively. Here, we aim to establish a novel prediction model.

Methods: The clinicopathologic characteristics of a cohort that included 86 non-metastatic GC patients and 43 PMGC patients from Zhongshan Hospital were retrospectively analysed to identify PM associated variables. Additionally, mass spectrometry and glycomic analysis were applied in the same cohort to find glycomic biomarkers in serum for the diagnosis of PM. A nomogram was established based on the associations between potential risk variables and PM.

Results: Overexpression of 4 N-glycans (H6N5L1E1: m/z 2620.93; H5N5F1E2: m/z 2650.98; H6N5E2, m/z 2666.96; H6N5L1E2, m/z 2940.08); weight loss ≥ 5 kg; tumour size ≥ 3 cm; signet ring cell or mucinous adenocarcinoma histology type; poor differentiation; diffuse or mixed Lauren classification; increased CA19-9, CA125, and CA724 levels; decreased lymphocyte count, haemoglobin, albumin, and pre-albumin levels were identified to be associated with PM. A nomogram that integrated with five independent risk factors (weight loss ≥ 5 kg, CA19-9 ≥ 37 U/mL, CA125 ≥ 35 U/mL, lymphocyte count < 2.0 * 10 ~ 9/L, and H5N5F1E2 expression ≥ 0.0017) achieved a good performance for diagnosis (AUC: 0.892, 95% CI 0.829-0.954). When 160 was set as the cut-off threshold value, the proposed nomogram represented a perfectly discriminating power for both sensitivity (0.97) and specificity (0.88).

Conclusions: The nomogram achieved an individualized assessment of the risk of PM in GC patients; thus, the nomogram could be used to assist clinical decision-making before surgery.

Keywords: Gastric cancer; Glycomic analysis; MALDI-TOF–MS; Nomogram; Peritoneal metastasis.

Fig. 1

Diagram of study population and cohort selection, with exclusion criteria described on the right-hand side of the diagram. AFP alpha-fetoprotein, CEA carcinoembryonic antigen, CA19-9, CA125, CA724 carbohydrate antigen 19-9, 125, and 724

Fig. 2

Representative mass spectra, scatter plot and ROC (Receiver Operating Characteristic) curve analysis of top four most prominently different expression of glycoforms between non-metastatic GC patients and GC patients with peritoneal metastasis. a 2620.93 m/z (H6N5L1E1). b 2650.98 m/z (H5N5F1E2). c 2666.96 m/z (H6N5E2). d 2940.08 m/z (H6N5L1E2) were identified. GC gastric cancer, PM peritoneal metastasis, AUC area under ROC curve, H hexose, N N-acetylhexosamine, F fucose, L α 2,3-sialic acid, E α 2,6-sialic acid. ***P < 0.001

Fig. 3

Forest Plot of multivariable logistic analysis identified independent risk factors for peritoneal metastasis in gastric cancer patients. CA19-9, CA125, CA724: carbohydrate antigen 19-9, 125, and 724; H hexose, N N-acetylhexosamine, F fucose, L α 2,3-sialic acid, E α 2,6-sialic acid

Fig. 4

Establishment and validation of nomogram to predict the risk of peritoneal metastasis in gastric cancer patients. a A nomogram integrating with weight loss, CA19-9, CA-125, lymphocytes count number, and H5N5F1E2 level was created to predict risk of peritoneal metastasis. b Calibration curve for nomogram-predicted and actual probability of having peritoneal metastasis. c Sensitivity and specificity for predicting peritoneal metastasis of nomogram model and other single variables was compared by the area under the Receiver operating characteristic (ROC) curve (AUC) and tested by Delong. Delong. Clarke-Pearson test. CA19-9, CA125, CA724: carbohydrate antigen 19-9, 125, and 724; H hexose, N N-acetylhexosamine, F fucose, L α 2,3-sialic acid, E α 2,6-sialic acid. ***P < 0.001

Fig. 5

Clinical significance of nomogram to predict peritoneal metastasis. a Total number of nomogram points in non-peritoneal metastasis (Mean ± SEM: 118.6 ± 7.93) and peritoneal metastasis (Mean ± SEM: 216.4 ± 8.77) patients were demonstrated by box plot. The box plot shows the full range of variation (error bars: min and max) with the line representing median. b When 160 was set as the cut-off value determined by ROC analysis and Youden index, nomogram had the best sensitivity (0.97) and specificity (0.88). Youden index = Sensitivity + Specificity − 1. c Positive rate (97%), negative rate (88%), false positive rate (12%), and false negative rate (3%) of the nomogram stratified into low-risk group (total number of nomogram points ≤ 160) and high-risk group (total number of nomogram points > 160). d Proportion of patients with or without peritoneal metastasis was demonstrated in low-risk group and high-risk group. PM peritoneal metastasis; ***P < 0.001