Pre-Chemotherapy D-Dimer Levels as Predictors of Survival Outcomes in Advanced Gastric Cancer

Introduction

Gastric cancer causes around 770 000 deaths worldwide each year, and is the fourth-leading cause of cancer-associated mortality [1]. The 5-year overall survival (OS) rate of patients with advanced gastric cancer (AGC) has improved from 23% to 45%, but the prognosis remains unsatisfactory [2–4].

For AGC, systemic chemotherapy is a standard therapy, but biomarkers for predicting gastric cancer outcome are scarce. The main treatment option for AGC that cannot be resected is systemic chemotherapy. During the last decade, use of antibodies against epidermal growth factor receptors and antibodies against vascular endothelial growth factor have improved OS rates. It is significantly simpler and less costly to measure prognostic serum biomarkers for AGC than to use tissue-based biomarkers. In the past few decades, biomarkers have been explored to predict the occurrence and OS of AGC.

Patients with malignant tumors have hypercoagulability as a physiological characteristic. It is essential for tumor angiogenesis that fibrin undergoes extra-cellular remodeling. D-dimer is the result of tissue plasminogen activator degrading cross-linked fibrin factor XIIIa through generating plasmin from plasminogen. By producing a monoclonal antibody that recognizes neither fibrinogen degradation nor non-cross-linked fibrin, it has been easier to investigate human D-dimer levels. Despite the absence of thrombosis, patients with advanced-stage tumors often have a systemic hypercoagulable state [5,6]. Patients (above 90%) with metastatic lesions had abnormal clotting or fibrinolysis, containing antithrombin-III (AT-III) complexes fibrinopeptides A (FPA) and D-dimer [7]. Venous thromboembolism (VTE) is an often-overlooked cause of mortality and morbidity in cancer patients can it is easily prevented and treated [8]. Specifically, a high risk of VTE is strongly associated with gastric cancer [9]. Blood flow stasis, endothelial damage, and hypercoagulability are all associated with VTE in cancer patients [10].

In cancer patients, D-dimer is a biomarker that can be used to diagnose and treat thrombosis, but it is rarely used to identify tumors. The predictive and prognostic value of D-dimer in AGC needs to be validated. Research shows that thrombin activatable fibrinolysis inhibitor (TAFI) and thrombin-antithrombin (TAT) complex levels [11], along with the stage IV patients’ D-dimer levels, were elevated in 52 gastric cancer patients. One study involving 1178 patients of lifetime beyond 2 years discovered that a subgroup of 50 gastric cancer patients had higher plasma D-dimers, which was linked to poorer OS and a notable risk factor for death [12]. Fibrinolytic activity induced by plasmin produced D-dimer as a cross-linked fibrin degradation product. Researchers have found that D-dimers advance cell proliferation and provoke angiogenesis in addition to affecting cellular signaling systems [13], as well as induce tumor growth and spread by motivating cancer cells to adhere to cells of endothelium, affecting platelets and the extra-cellular matrix (ECM) [14].

No long-term research has examined the linkage exists in plasma D-dimer levels and OS in AGC patients. This is the first study to analyze the effect of D-dimer on prognosis of patients with advanced gastric cancer by using propensity matching. We corrected the error due to different distributions of covariates by using Cox proportional hazard regression and propensity score matching. The goal was to estimate how PDL affects the prognosis of patients with AGC.

Material and Methods

PATIENT SELECTION:

Patients newly diagnosed with histologically substantiated advanced gastric malignancy receiving chemotherapy at the Sixth Medical Center, Chinese PLA General Hospital between January 2019 and December 2023 were identified from a retrospective archival database of electronic records. They all had metastatic gastric cancer and had IV stage tumors. The inclusion criteria were: 1) gastric cancer patients aged ≥18 years and with pathologically and/or computed tomography (CT) proven AGC; 2) no prior palliative therapies (containing chemotherapy and radiotherapy); and 3) followed up at least once. The exclusion criteria were: 1) breastfeeding or pregnant women; 2) previous malignancy diagnosed, a concurrent malignancy was present, or secondary tumors; 3) medical histories associated with thromboembolism, familial coagulopathy, active infections, or disseminated intravascular coagulation; 4) anti-coagulant and anti-aggregate therapies; 5) missing data on PDL; 6) underwent adjuvant chemotherapy after surgical resection.

FOLLOW-UP:

The study enrolled 134 patients after excluding 16 patients, as shown in Figure 1. The control group consisted of 89 patients with advanced gastric malignancy and low plasma D-dimer level. The data from both groups are summarized in Table 1. The study was reviewed and approved by the Ethics Committee of the Sixth Medical Center, Chinese PLA General Hospital. There were 134 AGC patients at the Sixth Medical Center, Chinese PLA General Hospital between January 2019 and December 2023 who were eligible for this study. Follow-up data were obtained by review of hospital records, communication with patients’ families, and reviews of the Sixth Medical Center, Chinese PLA General Hospital Registry. Patients were followed up until June 30, 2024. OS time is the interval between when gastric cancer was diagnosed and last follow-up or death.

ENZYME-LINKED FLUORESCENT IMMUNOASSAYS FOR D-DIMER LEVELS:

Among advanced gastric malignancy patients undergoing chemotherapy or radiotherapy, venous blood samples were collected and enzyme-linked fluorescent immunoassays and the miniVidas device (BioMerieux SA) were used to measure D-dimer levels. D-dimer levels over 1.5 ug/ml were considered HPD.

STATISTICAL ANALYSES:

Comparing categorical data was done applying chi-square tests or Fisher’s exact tests. Contrasting continuous data were assesses using the t test or Mann-Whitney U test. The survival benefits of these 2 treatments were compared using Kaplan-Meier survival curves and log-rank test. Multivariate and univariate analyses were performed using Cox proportional hazard regression models. The Wilcoxon signed-rank test was used to compare the samples of the 2 pairs. Analysis of multivariate survival data was conducted after univariate survival variables with P values >0.05 were incorporated. Hazard ratios (HR) and corresponding 95% confidence intervals (CI) were calculated for each predictor of survival. Since patients were not assigned to LPD or HPD at random, we reduced selection bias and balanced potential confounders. Applying propensity score matching (PSM), the following variables were included in the propensity model: sex, age, tumor location, tumor size, histology, pathological diagnosis, chemotherapy cycles, CEA, CA199, and CA724. P<0.05 (2-sided) was regarded as indicating a statistically significant difference. The propensity score indicates the conditional probability of a subject receiving a treatment given a vector of covariates, and is often used in non-randomized studies to adjust selection bias. A caliper of 0.02 was applied to the logit of the propensity score’s standard deviation. Each of the statistical analyses above was carried using SPSS software, version 25.0 (IBM, Chicago, IL, USA).

Results

THE CLINICAL AND PATHOLOGICAL FEATURES OF THE ENTIRE STUDY SERIES PRIOR TO MATCHING:

The vast majority (145 [96.7%]) of the 150 patients were followed up at least once. There were 134 patients included in the analysis after 16 patients were excluded. The follow-up time was 12.0 months on average (range: 3–50) (Figure 1). The 134 included patients consisted of 99 males (73.9%) and 35 females (26.1%). The median age was 63 years, with a range of 29–79 years. There were 26 of the 150 recruited patients who were not tested for HER-2 status, and 15 (12.1%) of these 124 patients were HER-2 positive, and these 15 patients received trastuzumab for chemotherapy. A total of 134 patients received chemotherapy using 5-fluorouracil, leucovorin, and oxaliplatin (FOLFOX6) or capecitabine and oxaliplatin (XELOX) or paclitaxel and cisplatin (TP) or docetaxel, oxaliplatin and S-1(DOS). Prior to the first treatment evaluation, 2 cycles of chemotherapy had been finished. In conformity with the Response Evaluation Criteria in Solid Tumors, version 1.1, computed tomography (CT) or magnetic resonance imaging (MRI) was used to determine treatment response. The objective responses were classified as partial responses (PRs), stable diseases (SDs), and progressive diseases (PDs).

Two groups of patients were categorized according to their pre-chemotherapy plasma D-dimer level (PDL): the low pretreatment D-dimer (LPD) group consisted of 89 patients with <1.5 ug/ml, and the high pretreatment D-dimer (HPD) group was composed of 45patients with PDL ≥1.5 ug/ml. A comparison of clinicopathologic variables was conducted between the 2 groups, as demonstrated in Table 1. Neither group had any statistically significant differences in sex, age, pathological diagnosis, tumor size, histology, or chemotherapy cycles. Patients in the LPD group were more likely to have malignancies situated in the upper third of the body (P=0.025) than in the HPD group. Among all patients, there was a median PFS of 7.5 months (with a 95% CI of 5.422–9.578) and a median OS of 13.8 months (with a 95% CI of 12.251–15.349). Kaplan-Meier curves for PFS and OS are shown in Figure 2A, 2B. None of the 134 patients achieved CR, 44 patients achieved PR, and 64 patients were SD. An objective response rate (ORR) of 32.84% and disease control rate (DCR) of 80.60% were achieved.

There was a significantly lower PFS and OS among HPD patients than among LPD patients (mPFS: 6.0 vs 8.7 months, P=0.015; mOS: 12.2 vs 15.1 months, P=0.037) (Figures 3A, 4A, Table 2). A survival analysis with univariate and multivariate variables is demonstrated in Table 3. The univariate analysis discovered a significant impact on OS for chemotherapy cycle, CA199, CA724, and D-dimer levels. Chemotherapy cycle and D-dimer levels independently predicted PFS through multivariate analysis. The chemotherapy cycle and CA724 levels were independently associated with survival. However, PDL was not significantly associated with OS (with a hazard ratio (HR) of 1.362, 95% CI of 0.851–2.181, P=0.198). The correlation between D-dimer levels and chemotherapy response before PSM is shown in Table 4. In accordance with the first response evaluation, patients with PD had an increased mean D-dimer by 1.72 ug/mL compared with patients with PR and SD (P=0.006). By contrast, the mean D-dimer increased by 1.21 mg/mL in 26 PD patients during the first response evaluation, although the difference was not statistically significant (P=0.113).

PATIENT CHARACTERISTICS AFTER PROPENSITY SCORE MATCHING:

A propensity score-based one-to-one matching method was used to select 43 patients for each group. As a result of the propensity score analysis, the characteristics are indicated in the right columns of Table 5. A total of 43 patients in the LPD were matched with 43 patients in the HPD as a result of covariate adjustment. In the matched study series, there was a median PFS of 6.3 months (with a 95% CI of 5.002–7.598) and a median OS of 13.6 months for all 89 patients (95% CI: 12.209–14.991). The Kaplan-Meier curve showing PFS and OS is shown in Figure 2C, 2D. However, the OS time for LPD and HPD differed significantly. The OS of the HPD group was remarkably lower than in the LPD group (mOS: 12.2 vs 15.1 months, P=0.032) (Table 6, Figure 4B), but the PFS did not significantly vary between the 2 groups (mPFS: 6.0 vs 7.3 months, P=0.182) (Figure 3B). After PSM, only D-dimer levels (HR 1.746, 95% CI: 1.040–2.932; P=0.035) and chemotherapy cycle (HR 0.277, 95% CI: 0.160–0.478; P=0.000) showed significant associations with OS in univariate analysis. After the multivariate adjustment, the predictive value still existed (Table 3). As determined by multivariable survival analysis, D-dimer levels were independently associated with OS (HR 1.711, 95% CI 1.109–2.875; P=0.042) and chemotherapy cycle (HR 0.280, 95% CI 0.163–0.483; P=0.000). Other variables, including age, sex, pathological diagnosis, tumor location, tumor size, CEA, CA199, and CA724, showed no significant associations with PFS or OS after PSM (Table 3). The P values are calculated by Wilcoxon signed-rank test on the difference in D-dimer levels between at the pretreatment and the first response evaluation in Table 4. When PD patients were compared with PRs or SDs, their mean D-dimer increased by 1.91ug/mL (P=0.039). Conversely, 26 patients with PD had an increase in mean D-dimer of 2.21 mg/mL during the first response evaluation. However, statistical significance was not achieved by this difference (P=0.387). And AGC patients may benefit from the application of D-dimer levels as a predictor of chemotherapy response.

Discussion

LIMITATIONS:

The small sample size is a primary limitation, and the statistical power may therefore be affected. Affecting the conclusion of the present study of D-dimer patients before chemotherapy were not usually measured after admission to hospital. Owing to its retrospective nature, selection bias existed in this research inevitably. Few patients with metastatic gastric cancer did not have their D-dimer measured as part of their baseline assessments.

Conclusions

Plasma D-dimer analysis is an inexpensive, noninvasive, and simple method that can be a useful guide in predicting the dissemination and prognosis of advanced gastric cancer.