Clinical Significance of Serum CA125, CA19-9, CA72-4, and Fibrinogen-to-Lymphocyte Ratio in Gastric Cancer With Peritoneal Dissemination

Chao Huang, Zitao Liu, Xiao Li, Yongqiang Xia, Jun Huang, Hongliang Luo, Zhen Zong, Zhengming ZhuFrontiers in Oncology 2019 · DOI

Abstract

Background: Relevant serum tumor markers have been indicated to be associated with peritoneal dissemination (PD) of gastric cancer (GC). Fibrinogen has been shown to play an important role in the systemic inflammatory response (SIR) and in tumor progression. However, the clinical significance of the fibrinogen-to-lymphocyte ratio (FLR) in GC with PD has not been studied. Methods: The clinical data of 391 patients with GC were collected, including 86 cases of PD. Then, 1:3 matching was performed by propensity score matching (PSM), and the clinical data of the matched 344 patients were analyzed by univariate and multivariate conditional logistic regression. Classification tree analysis was used to obtain the decision rules and a random forest algorithm to extract the important risk factors of PD in GC. A nomogram model for risk assessment of PD in GC was established. Results: Univariate analysis showed that the factors related to PD in GC were: CA125, CA19-9, CA72-4, FLR, NLR, ALR, PLR, and CEA. Conditional logistic regression found that CA125, CA19-9, and FLR were independent risk factors. The accuracy, sensitivity, and specificity of the model were 91%, 89.5%, and 79.5%, respectively. Conclusion: CA125, CA19-9, and FLR are significant risk factors for GC with PD. The decision rules and nomogram model constructed by these markers can correctly predict the risk of PD in GC.

Introduction

Gastric cancer (GC) remains a leading cause of cancer-related mortality worldwide. One of its most devastating progression routes is peritoneal dissemination (PD)—where cancer cells spread to the lining of the abdominal cavity. If PD is present, radical surgery is usually ineffective, and the focus shifts to palliative care. However, detecting PD is notoriously difficult. Standard imaging like CT or PET scans often misses small nodules (under 5mm), leading to a high rate of "missed" diagnoses. While diagnostic laparoscopy is accurate, it is an invasive surgical procedure. There is a critical need for a reliable, non-invasive blood test to identify which patients are likely to have PD before they reach the operating table.

Methodology

The study analyzed clinical data from 391 patients diagnosed with gastric cancer at The Second Affiliated Hospital of Nanchang University, China. To ensure a fair comparison between patients with and without PD, the researchers used Propensity Score Matching (PSM) at a 1:3 ratio, resulting in a matched cohort of 344 patients (86 PD cases and 258 non-PD cases).

The team employed a multi-staged statistical approach:

  1. Univariate and Multivariate Logistic Regression: To identify independent risk factors.
  2. Classification Tree Analysis & Random Forest Algorithms: To rank the importance of different variables and establish clear decision rules.
  3. Nomogram Construction: A visual tool was created to allow clinicians to calculate a patient's risk score based on specific blood markers.

The standout feature of this study is the inclusion of the Fibrinogen-to-Lymphocyte Ratio (FLR), a marker of the body's systemic inflammatory response, alongside traditional carbohydrate antigens (CA125, CA19-9, and CA72-4).

Key Insights

While tumor markers like CA125 are well-known, this paper highlights that the FLR is a potent independent predictor. Fibrinogen promotes tumor growth and migration, while lymphocytes are key to anti-tumor immunity. A high FLR suggests a pro-tumor inflammatory environment.

Variable importance ranking using the Random Forest algorithm highlights CA125 and CA72-4 as top predictors.

The study identified specific "cut-off" values that are more sensitive for PD than standard clinical "normal" ranges:

  • CA125 > 17.3 U/ml
  • CA19-9 > 27.315 U/ml
  • FLR > 2.555

Results

The combined predictive model (Nomogram) outperformed individual markers significantly. The synergy of combining these markers into a single model significantly boosts predictive power.

ROC curve for the combined model (CA125+CA72-4+CA19-9+FLR) showing high sensitivity and specificity.

The key performance metrics for the model were:

  • Accuracy: 91%
  • Sensitivity: 89.5%
  • Specificity: 79.5%
  • Area Under the Curve (AUC): 0.910

For comparison, the decision rules alone (based on the classification tree) yielded a specificity of 94.0%, making them highly reliable for ruling out PD when scores are low.

Conclusion and Future Work

The decision rules and nomogram model constructed by CA125, CA19-9, CA72-4, and FLR can correctly predict the risk of PD in GC. However, as a single-center retrospective study, there is an inherent risk of selection bias. While 391 patients is a solid start, 86 PD cases is a relatively small sample for complex algorithmic modeling. Other inflammatory markers like Interleukin-6 (IL-6) or C-reactive protein (CRP) were not included. Future large-scale, multicenter prospective randomized controlled trials are needed to validate these findings across diverse populations.

Summary updated

Abstract

<b>Background:</b> Relevant serum tumor markers have been indicated to be associated with peritoneal dissemination (PD) of gastric cancer (GC). Fibrinogen has been shown to play an important role in the systemic inflammatory response (SIR) and in tumor progression. However, the clinical significance of the fibrinogen-to-lymphocyte ratio (FLR) in GC with PD has not been studied. <b>Methods:</b> The clinical data of 391 patients with GC were collected, including 86 cases of PD. Then, 1:3 matching was performed by propensity score matching (PSM), and the clinical data of the matched 344 patients were analyzed by univariate and multivariate conditional logistic regression. Classification tree analysis was used to obtain the decision rules and a random forest algorithm to extract the important risk factors of PD in GC. A nomogram model for risk assessment of PD in GC was established by using the rms package of R software. <b>Results:</b> Univariate analysis showed that the factors related to PD in GC were: carbohydrate antigen (CA) 125 (<i>P</i> < 0.0001), CA19-9 (<i>P</i> < 0.0001), CA72-4 (<i>P</i> < 0.0001), FLR (<i>P</i> < 0.0001), neutrophil-to-lymphocyte ratio (NLR) (<i>P</i> < 0.0001), albumin-to- lymphocyte ratio (ALR) (<i>P</i> < 0.0001), platelet-to-lymphocyte ratio (PLR) (<i>P</i> = 0.013), and carcinoembryonic antigen (CEA) (<i>P</i> = 0.031). Conditional logistic regression found that CA125 (OR: 1.046; <i>P</i> < 0.0001), CA19-9 (OR: 1.002; <i>P</i> < 0.0001), and FLR (OR: 1.266; <i>P</i> = 0.024) were independent risk factors for GC with PD. The accuracy, sensitivity, specificity, positive predictive value and negative predictive value of the decision rules for detecting PD of GC were 89.5, 77.4, 94.0, 82.8, and 91.8%, respectively. According to the important variables identified by the classification tree and random forest algorithm, the risk assessment model of PD in GC was established. The accuracy, sensitivity, and specificity of the model were 91, 89.5, and 79.5%, respectively. <b>Conclusion:</b> CA125 > 17.3 U/ml, CA19-9 > 27.315 U/ml, and FLR > 2.555 were the risk factors for GC with PD. The decision rules and nomogram model constructed by CA125, CA19-9, CA72-4, and FLR can correctly predict the risk of PD in GC.
BibTeX
@inproceedings{huang2019clinical,
  author = {Chao Huang and Zitao Liu and Xiao Li and Yongqiang Xia and Jun Huang and Hongliang Luo and Zhen Zong and Zhengming Zhu},
  title = {{Clinical Significance of Serum CA125, CA19-9, CA72-4, and Fibrinogen-to-Lymphocyte Ratio in Gastric Cancer With Peritoneal Dissemination}},
  booktitle = {Frontiers in Oncology},
  year = {2019},
  doi = {10.3389/fonc.2019.01159},
}
Figure