Combined Use of Tumor Markers in Gastric Cancer: A Novel Method with Promising Prognostic Accuracy and Practicality

The effect of a single tumor marker on the prognosis of gastric cancer patients is not ideal. This study explored a novel prognostic assessment method for gastric cancer (GC) patients using a combination of three important tumor markers (CEA, CA72-4, and CA19-9). Data from 1966 GC patients who underwent curative gastrectomy at Sun Yat-Sen University Cancer Center (Guangzhou, China) were included. Hazard ratios (HR) for all factors for overall survival (OS) were analyzed by Cox regression. A nomogram and calibration curve were used to establish the survival prediction model. The prediction accuracy was evaluated with the concordance index (C-index). All patients were divided into four groups (C0–C3) according to the number of elevated tumor markers. The 5-year OS rates of the patients in preoperative groups C0–C3 were 83.8% (81.3–86.4%), 72.8% (68.5–77.4%), 58.9% (50.4–68.9%), and 18.5% (4.0–33.0%), respectively, and those in postoperative groups C0–C3 were 82.1% (79.4–84.8%), 76.1% (72.2–80.3%), 57.6% (48.4–68.5%), and 16.8% (5.1–28.5%), respectively, with significant differences between each C0–C3 subgroup in both preoperative and postoperative cohorts. Multivariate analysis showed that preoperative (HR: 6.001, 95% CI: 3.523–10.221) and postoperative (HR: 8.149, 95% CI: 4.962–13.528) elevated tumor markers were independent risk factors for GC patients. The C-index for the combined use of tumor markers was 0.65–0.66, which was higher than that for using a single tumor marker (0.53–0.56). The combined use of tumor markers significantly improved the prognostic value compared with using a single tumor marker. The survival prediction model including the combined tumor markers was accurate and effective.

Gastric cancer (GC) is the second most common tumor of the gastrointestinal tract. 1 It is also the fifth most common cancer and the third leading cause of cancer death worldwide. 2If GC is diagnosed at an early stage, surgery alone may be curative.However, for patients with locally advanced GC, despite receiving radical surgery and perioperative chemotherapy, their prognosis remains poor. 3,4The tumor-node-metastasis (TNM) staging system of the American Joint Committee on Cancer (AJCC) is recognized as the most reliable standard prognosticator for GC patients. 5owever, the substantial intergroup heterogeneity among patients using the AJCC TNM staging system (i.e., survival of patients in the same TNM stage might differ substantially) shows the need for more accurate stage grouping criteria to improve the prognosis estimation of GC patients, which is needed for surgical evaluation, postoperative treatment, and follow-up planning. 6umerous studies have demonstrated the clinical significance of serum tumor markers in the diagnosis, prognosis, and recurrence monitoring of gastrointestinal cancers. 7ifferent tumor markers have different values in different tumors.According to previous studies, the three most important tumor markers in GC are carcinoembryonic antigen (CEA), carbohydrate antigen 72-4 (CA72-4) and carbohydrate antigen 19-9 (CA19-9). 8These three indicators have a high positive expression rate and good prognostic value in GC and are widely used in monitoring GC patients. 9However, most of the previous studies focused on the clinical significance of preoperative serum tumor marker levels to predict the prognosis of GC patients and focused on single or dual tumor markers, which have shown limited prediction efficiency. 10,11Moreover, the current literature lacks a comparison between preoperative and postoperative levels, as well as the combined application of multiple tumor markers for the prognosis evaluation of GC patients.
Therefore, this study used a large sample of data to systematically evaluate the prognostic value of three tumor markers (CEA, CA72-4, and CA19-9) in pre-and postoperative settings to explore a more accurate and practical prognostic assessment method for GC patients.

Study Population and Ethics
The data from GC patients who underwent curative gastrectomy at the Department of Gastric Surgery at the Sun Yat-Sen University Cancer Center (SYSUCC, Guangzhou, China) between January 2007 and December 2017 were assessed.Patients were included if they had histologically confirmed gastric adenocarcinoma, had no evidence of distant metastasis, underwent curative gastrectomy with D1/D2 lymph node dissection, and had complete clinical information and follow-up records.Patients were excluded if they had other malignant tumors, received neoadjuvant chemotherapy or radiotherapy, or lacked the pre-or postoperative tumor marker data.In total, 1966 GC patients were eligible for this study.Tumor staging was performed according to the eighth edition of the AJCC Staging Manual. 5he demographic and clinicopathologic characteristics of the patients, including sex, age at diagnosis, body mass index, Borrmann type, pathological type, tumor size, TNM stage, adjuvant chemotherapy, and tumor marker value, were recorded for analysis.This study was performed in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board at SYSUCC.All patients provided written informed consent before enrollment.

Patient Follow-Up
All enrolled patients underwent routine outpatient followups after surgery.They were followed up every 3-6 months within 2 years after surgery, every 6-12 months from 3 to 5 years after surgery, and every year after 5 years.Our median follow-up time was 45.9 (range: 1-180) months.The main follow-up examinations included tumor marker determination, computed tomography (CT) examination, ultrasound examination, and endoscopy, among others, to determine the patient's survival and recurrence status.

Statistical Analysis
Statistical Product Service Solutions (SPSS) software (version 26.0, IBM, USA) was used to analyze the data, and R (version 4.0.4,Revolution Analytics, USA) was used to generate the graphs.Categorical variables are shown as numbers and percentages.Continuous variables are presented as means ± standard deviations, or median with range.Survival curves for overall survival (OS) and recurrence-free survival (RFS) were generated using the Kaplan-Meier method, and the log-rank test was used to compare survival times.OS was calculated as the time from the date of surgery to death from any cause or the last follow-up.Cox regression analysis was used to assess the hazard ratios (HRs) of all factors for OS, and factors with a p value < 0.05 or with a high level of importance in the clinical diagnosis were included Combined Use of Tumor Markers … in the multivariate analysis.Nomogram and calibration curve were used to establish the survival prediction model of GC patients.The concordance index (C-index) was used to evaluate the prediction accuracy of the regression analysis model.

Clinical Features
Based on the above inclusion and exclusion criteria, a total of 1966 patients, including 1300 males (66.1%) and 666 females (33.9%), were enrolled in this study.The clinical characteristics of all patients and each cohort are summarized in Table 1.Patients older than 65 years accounted for 22.2% of all patients.The differentiation degree of tumors was mainly low (66.5%), and highly differentiated tumors appeared only in a few patients (0.9%).More than 70% of patients received adjuvant chemotherapy after surgery, while the remaining did not undergo any postoperative treatment due to the early staging (22%) or personal requirements.The pre-and postoperative cohorts had similar population distributions in age, gender, BMI, etc.There were differences in the clinicopathological characteristics and TNM stage between each cohort.Patients with more elevated pre-or postoperative tumor markers had relatively more advanced TNM stages and larger tumor size.
By integrating several important pieces of clinical information and the number of elevated tumor markers in patients with gastric cancer, we established a nomogram prognostic model (Fig. 3a).All patients were divided into a points-low group and a points-high group by nomogram total points.The overall survival time of the points-low group was significantly longer than that of the points-high group (HR: 6.321, 95% CI: 4.823-8.282,P < 0.001; Fig. 3b).Using the calibration curve, it was found that the survival prediction model with the elevated number of tumor markers had a good fit.The C-index was used to evaluate the prediction accuracy of the prognostic model of GC patients by calculating the discrimination between the predicted value and the actual accuracy of the Cox regression model in the survival analysis.The C-indexes of different predictive indicators are shown in Fig. 4. The C-index for a single tumor marker ranged from 0.53 to 0.56.Among the three tumor markers, the C-index of preoperative and postoperative CA72-4 (both 0.56) was the highest.The combined use of tumor markers had a higher C-index (0.65-0.66) for prognostic analysis than a single tumor marker.Moreover, the combination of TNM staging and preoperative (0.76) or postoperative (0.74) cohorts could further improve the C-index for the survival analysis.There was no significant difference between preoperative and postoperative tumor markers in predicting patient survival (P = 0.12).

Patient Survival and Adjuvant Chemotherapy in Stage II/ III Patients
In this study, we also analyzed the relationship between receiving adjuvant chemotherapy and patient survival.The 5-year OS rate for patients receiving adjuvant chemotherapy in stage II/III patients was 71.9% (95% CI: 68.9-75.0%),while the 5-year OS rate for patients without adjuvant chemotherapy in stage II/III patients was 66.8% (95% CI: 58.9-75.8%).We further analyzed the survival difference between C0-C3 groups with or without adjuvant chemotherapy (Figure S4).In patients without adjuvant chemotherapy, there was no significant difference in survival between the cohort subgroups (P > 0.05 for the log-rank test); and in patients with adjuvant chemotherapy,  the difference was significant (P < 0.05 for the log-rank test).This was consistent in the preoperative and postoperative cohorts.

DISCUSSION
Tumor markers play an important role in the diagnosis and prognosis of various cancers. 13Serum CEA, CA72-4, and CA19-9 are the most clinically used tumor markers in GC. 14 At the moment, relevant studies often explore the prognostic value of only a single tumor marker, and only a few studies have explored the prognostic effect of two tumor markers at the same time. 11,15,16Several studies suggested that the combined use of tumor markers can increase prognostic efficacy. 17,18However, no studies have been conducted to integrate the three tumor markers mentioned above.Through systematic analysis of a large sample, this study conducted an integrated analysis of these three tumor markers in GC patients and found that the method for calculating an elevated number of tumor markers could easily and effectively distinguish the survival and prognosis of each group of patients.
The expression of these markers is usually elevated after GC diagnosis and affects GC prognosis. 15The positive rates of CEA, CA72-4, and CA19-9 have been variably reported in previous studies, ranging from 15% to 30%.In this present study, the positive rates of preoperative CEA, CA72-4, and CA19-9 were also consistent with previous studies. 8,19,20We found that patients with increased tumor markers had shorter OS and RFS rates than those with normal tumor markers, consistent with previous studies. 21,22Further, this study stratified the patients according to the number of elevated tumor markers.It was noted that the OS and RFS decreased successively as the number of tumor markers increased.Therefore, the patients could be divided into 4 groups with different prognoses.Previous studies have reported that the proportion of elevated tumor markers in early GC is relatively low and that patients with early GC have relatively promising prognoses.Thus, the prognostic significance of tumor markers in early GC is relatively limited. 21In our study, the distinction is particularly pronounced in stage III GC rather than stage I or II GC.In GC, the prognostic value of preoperative and postoperative tumor markers is controversial.4][25] Our study found that the prognostic value between preoperative and postoperative tumor markers was not different, whether used alone or in combination.
In prognostic studies, tumor markers are often used in entire populations or advanced GC. 26,27 Lin et al. 23 proposed that the combination of tumor markers and TNM staging could improve prediction accuracy.In our study, the C-index of each single tumor marker ranged from 0.53 to 0.56, while the combined use of tumor markers can improve the C-index to 0.66.9][30] Due to the lack of a particularly effective single tumor marker in GC, the application of tumor markers is limited.However, the combined multi-tumor marker method can significantly improve prediction accuracy and has more potential for further development.In addition, a more effective and simple prognostic analysis system can be established by combining this method with traditional TNM staging.
Postoperative adjuvant chemotherapy is an important factor affecting the prognosis of patients, and previous studies have shown that patients receiving adjuvant chemotherapy tend to have a longer survival time. 31We also explored the predictive value of tumor markers for adjuvant chemotherapy in patients with stage III GC, and found that in patients receiving adjuvant chemotherapy, the difference in the number of elevated tumor markers was significant, but in patients without chemotherapy, there was no significant difference.This suggests that the number of elevated tumor markers is related to the patient's sensitivity to chemotherapy, and patients with a low number of tumor markers have higher sensitivity and a better therapeutic response to chemotherapy.Some studies have found that in addition to traditional tumor markers, such as the fibrinogen-to-lymphocyte ratio (FLR) and circulating DNA (ctDNA), the comprehensive application of multiple indicators is a potential direction for further research. 32,33he tumor marker prediction model established in this study provides indications not only for prognosis determination, but also for patient treatment plan selection and improvement of follow-up frequency.Neoadjuvant chemotherapy has become a hot topic in gastric cancer treatment in recent years, as it can reduce surgical risks and improve overall patient survival. 335][36] By scoring preoperative tumor markers, patients with higher scores (C2/C3) who have worse prognoses can be used as a reference for deciding on neoadjuvant treatment.According to the NCCN Clinical Practice Guidelines in Oncology for Gastric Cancer (Version 2.2022), follow-up examinations for gastric cancer patients include medical history, physical examination, blood tests, enhanced CT scans, endoscopy, etc. 37 Based on the postoperative tumor marker scores, the follow-up frequency and invasive examination requirements can be adjusted within a certain range for high-scoring (C2-C3) and lowscoring (C0-C1) patients.This allows for a more personalized approach to patient follow-up, improving recurrence detection rates while reducing the frequency of invasive examinations.
We acknowledge that our study has some limitations.First, our study was retrospective.Although we tried to reduce the error by increasing the sample size, selection bias still existed.Second, patients were included over a large span of years.Updates to guidelines and technological progress have resulted in certain differences in surgical methods, chemotherapy methods, and detection methods among patients.These differences are difficult to eliminate and may have affected the results of our analysis.Third, many factors might influence tumor marker levels: the influencing factors of different tumor markers are numerous and unclear, and some were not included in our data analysis.Additionally, the collection period for each patient was not exactly the same, so there may be some interference.We are also planning to carry out prospective, multicenter studies to provide more clinically valuable results, hoping to better use tumor markers for prognostic analysis of patients in the future.

CONCLUSION
The survival prediction model including the number of elevated tumor markers has good discrimination and calibration, and is significantly better than using a single tumor marker.Combined use of tumor markers has promising prognostic accuracy and practicality and should be used routinely in prognostic analysis in GC patients.The new prognostic model can provide effective guidance for treatment selection and follow-up strategies for patients.

FIG
FIG. 2 Kaplan-Meier analysis for OS and RFS of patients with gastric cancer compared according to the preoperative cohort (C0-C3).a, b Kaplan-Meier analysis for OS and RFS in all patients; c, d Kaplan-Meier analysis for OS and RFS in Stage I patients; e, f Kaplan-Meier analysis for OS and RFS in Stage II patients; g, h Kaplan-Meier analysis for OS and RFS in Stage III patients; * indicates that the p-value of the log-rank test between each pair of groups is less than 0.05.** indicates that the p value of the log-rank test between each pair of groups is less than 0.01 Combined Use of Tumor Markers …

FIG. 3
FIG. 3 (a) The nomogram survival prediction model of gastric cancer patients with important clinical information and elevated number of tumor markers.(b) Kaplan-Meier analysis of OS of the patients

FIG. 4 C
FIG. 4 C-index comparison of survival analysis of patients with different indicators

TABLE 1
General characteristics of 1966 patients with gastric cancer BMI body mass index, TNM tumor, lymph node, metastasis

TABLE 2
Univariate and multivariate analyses of clinicopathological and tumor markers in all 1966 patients with gastric cancer for OS BMI body mass index, OS overall survival * Indicates p value of Cox regression less than 0.05, **Indicates p value of Cox regression less than 0.01

TABLE 3
Survival analysis after stepwise model optimization Model 1 = Adjusted by age + BMI + sex; Model 2 = Model 1 + tumor size + differentiated degree + Borrmann type + adjuvant chemotherapy; Model 3 = Model 2 + TNM stage * Indicates p value of Cox regression less than 0.05, **Indicates p value of Cox regression less than 0.01