Preoperative tumor marker burden score predicts survival in colorectal cancer according to tumor size

Abstract

Background

Reliable preoperative biomarkers are needed to improve risk stratification in colorectal cancer (CRC). The prognostic value of a combined score based on multiple serum tumor markers remains underexplored.

Methods

This retrospective study included 307 CRC patients undergoing curative surgery. A preoperative risk score was defined by the number of elevated markers (AFP, CA19-9, CA125, CA724, CA153) above institutional cut-offs. Patients were stratified into low-risk (≤ 1 elevated marker) and high-risk (> 1 elevated marker) groups. Associations with 5-year disease-free survival (DFS) and overall survival (OS) were analyzed using Kaplan-Meier and Cox regression. Subgroup analysis by tumor size (≤ 4 cm vs. >4 cm) was performed.

Results

A low-risk score (≤ 1 elevated marker) was independently associated with improved 5-year OS (adjusted HR = 0.60, 95% CI: 0.37–0.97, P = 0.038) and showed a strong trend for better DFS (adjusted HR = 0.65, 95% CI: 0.42-1.00, P = 0.051) after adjusting for clinicopathological factors. Strikingly, this prognostic association was significant only in patients with tumors ≤ 4 cm (for OS: HR = 0.19, P = 0.003; for DFS: HR = 0.34, P = 0.010) but not in those with larger tumors (P > 0.05). The interaction between tumor size and the marker score on OS approached statistical significance (P for interaction = 0.051).

Conclusion

A simple preoperative multi-marker score is an independent prognostic factor for survival in CRC. Its predictive power is particularly strong in patients with smaller tumors (≤ 4 cm), suggesting a valuable role in refining preoperative risk assessment for this subgroup.

Introduction

Colorectal cancer (CRC) remains a majorglobal health burden, representing one of the leading causes of cancer-related morbidity and mortality worldwide [1, 2]. Despite advancements in screening and treatment modalities, prognostic heterogeneity among patients persists, particularly in locally advanced stages, underscoring the urgent need for reliable biomarkers to guide risk stratification and therapeutic decision-making [3, 4]. Conventional clinicopathological parameters such as tumor stage, grade, and lymph node status have long served as cornerstone prognostic indicators; however, their predictive accuracy is often insufficient to capture the full spectrum of disease behavior, prompting the exploration of novel molecular and serum-based markers [5, 6]. In this context, tumor markers have emerged as valuable tools for reflecting tumor biological activity, with certain markers demonstrating significant associations with disease progression and patient outcomes [1, 7].

Among the plethora of tumor markers, carbohydrate antigens such as CA125, CA199, CA724, and CA153, along with alpha-fetoprotein (AFP), have been extensively investigated in various gastrointestinal malignancies [7, 8]. These markers are often elevated in CRC and have been implicated in processes such as cell adhesion, metastasis, and immune modulation [1, 9, 10]. Importantly, the clinical utility of these markers may extend beyond diagnosis to prognostication, as aberrant expression patterns have been correlated with advanced disease stages and poorer survival in several studies [4, 5]. For instance, elevated levels of CA199 have been consistently associated with liver metastasis and reduced survival in CRC, while CA125 has been linked to peritoneal dissemination and aggressive tumor biology [11]. Nevertheless, the prognostic significance of these markers when evaluated in combination remains underexplored, particularly in the specific context of neoadjuvant therapy settings.

Chronic inflammation is a well-established enabler of carcinogenesis and progression in CRC. Systemic alterations in innate immune cells, particularly monocytes and macrophages, are increasingly documented. Macrophage secretion of the pro-inflammatory cytokine IL-1β was significantly higher in patients with tumors extending beyond the muscularis propria and in those with stage IV disease, suggesting a tumor-driven reprogramming of innate immunity associated with advanced stages [12]. This systemic inflammatory state, potentially measurable through a panel of inflammatory markers like CA125 and CA153, could contribute to the definition of a high-risk phenotype. Moreover, the efficacy of the anti-tumor immune response is crucial.

Given the limitations of relying on single markers, there is growing interest in developing integrated scoring systems that combine multiple biomarkers to enhance prognostic precision [4]. The concept of assigning scores based on the number of elevated tumor markers—termed tumor marker burden—has gained traction in several cancer types. This approach operates on the premise that simultaneous elevation of multiple markers may indicate more extensive disease or greater biological aggressiveness, potentially corresponding to worse clinical outcomes [13]. In CRC, however, the prognostic value of a multi-marker scoring system has not been thoroughly investigated. Moreover, whether such a scoring system retains its predictive power across different tumor stages or sizes remains an open question, necessitating further stratified analyses.

Methods

Study population and setting

This retrospective cohort study initially reviewed the medical records of all patients diagnosed with primary CRC who underwent curative-intent surgery at the First Affiliated Hospital of Soochow University between September 2013 and December 2020. The inclusion criteria were: (1) histopathologically confirmed adenocarcinoma of the colon or rectum; (2) availability of complete preoperative serum tumor marker measurements within 2 weeks prior to surgery; (3) underwent standard radical resection (R0). The exclusion criteria were: (1) patients with a history of other malignancies; (2) patients with incomplete baseline data; (3) patients lost to follow-up.

After applying these criteria, a total of 307 patients were included in this study following the exclusion of 10 cases with a history of other malignancies, 5 cases due to lack of baseline data and 7 cases lost to follow-up. (Fig. 1).

Fig. 1.

Flow diagram showing the patient selection process

Baseline demographic and clinical information

Preoperative peripheral blood samples were collected from all patients after an overnight fast. The panel of AFP, CA19-9, CA125, CA724, and CA153 was selected because these markers are routinely measured in clinical practice for gastrointestinal malignancies and have been validated to correlate with CRC progression and prognosis [9, 14–16]. Serum levels were measured using standardized chemiluminescent immunoassays. The institutional upper limits of normal were: AFP (25.0 ug/L), CA19-9 (37.0 U/mL), CA125 (35.0 U/mL), CA724 (8 U/mL), CA153 (25 U/mL). Carcinoembryonic antigen (CEA) was excluded from the panel due to incomplete preoperative data availability in a substantial proportion of patients, to avoid selection bias and ensure data integrity. Patients were stratified into low-risk (≤ 1 elevated marker) and high-risk (> 1 elevated marker) groups. Baseline variables included age, gender, tumor location, tumor size, depth of invasion (T stage), lymph node metastasis (N stage), distant metastasis (M stage), postoperative chemotherapy, differentiation grade, tumor deposits, vascular invasion and nerve invasion.

Subgroup stratification

Subgroup analysis by tumor size (≤ 4 cm vs. >4 cm) was performed. The 4 cm cutoff was predefined based on published prognostic studies in colorectal cancer, not derived from median tumor size or post-hoc optimal-cutoff analysis in the present cohort [17, 18].

Follow-up and definition of prognostic endpoints

Postoperative follow-up was conducted according to the National Comprehensive Cancer Network (NCCN) guidelines. Patients were typically evaluated every 3–6 months for the first 3 years and annually thereafter. The primary prognostic endpoint was Overall Survival (OS), defined as the time from the date of surgery to the date of death from any cause or the last confirmed date of follow-up. The secondary endpoint was Disease-Free Survival (DFS), defined as the time from surgery to the first event of disease recurrence (local or distant) or death from any cause.

Statistical analysis

Categorical variables were presented as frequencies and percentages and compared using the Chi-square test or Fisher’s exact test. Continuous variables were expressed as medians with interquartile ranges and compared using the Mann-Whitney U test. Patients who were lost to follow-up or alive at the last follow-up were defined as censored observations; non-informative censoring was assumed in the Cox proportional hazards model. Survival curves for OS and DFS were generated using the Kaplan-Meier method and compared by the log-rank test. Univariate and multivariate Cox proportional hazards regression models were employed to identify independent prognostic factors. Hazard ratios (HR) and their 95% confidence intervals (CI) were calculated. After adopting stepwise regression to select the model combination with the smallest AIC, age, sex, TNM, differentiation, tumor deposits were included in the multivariate regression mode2.

At the same time, we included interaction terms in the cox regression model to study the association between the number of elevated tumor indicators and the 5-year DFS and OS of patients at different tumor size.

Two-sided P values of 0.05 were considered statistically significant. All statistics were conducted with SAS 9.4 software (SAS Institute Inc., Cary, NC). Figures were drawn by R software (R Development Core Team 2014, www.r-project.org).

Results

A total of 307 CRC patients were finally included with the mean age was 59.0 ± 11.2 years old. 223 patients had a 5-year disease-free survival, and 265 patients had a 5-year overall survival. Patients of tumor size ≥ 4 cm was 175 (57.0%). Patients with high risk level were more likely to be higher T stage, N stage, TNM stage, high-level differentiation, higher rate of tumor deposits. Comparison of data of all patients is shown in Table 1.

Table 1.

Baseline characteristics of patients undergoing colorectal cancer surgery

Characteristic	Risk level			p-value
	Overall	Low N = 248	High N = 59
	N = 307
Age, y, mean ± SD	59.0 ± 11.2	59.4 ± 10.7	57.4 ± 12.9	0.273
Male, n (%)	169 (55.0)	136 (54.8)	33 (55.9)	0.879
Postoperative chemotherapy, n (%)	254 (83.0)	204 (82.6)	50 (84.7)	0.692
Rectal tumor, n (%)	143 (46.7)	117 (47.4)	26 (44.1)	0.648
Tumor size ≥ 4 cm, n (%)	175 (57.0)	131 (52.8)	44 (74.6)	0.002
T, n (%)				< 0.001
T1/T2	48 (15.6)	46 (18.5)	2 (3.4)
T3	182 (59.3)	151 (60.9)	31 (52.5)
T4	77 (25.1)	51 (20.6)	26 (44.1)
N, n (%)				< 0.001
N0	144 (46.9)	126 (50.8)	18 (30.5)
N1	108 (35.2)	87 (35.1)	21 (35.6)
N2	55 (17.9)	35 (14.1)	20 (33.9)
TNM, n (%)				0.006
I	43 (14.0)	41 (16.5)	2 (3.4)
II	101 (32.9)	85 (34.3)	16 (27.1)
III	163 (53.1)	122 (49.2)	41 (69.5)
Low-level differentiation, n (%)	36 (11.7)	22 (8.9)	14 (23.7)	0.001
Tumor deposits, n (%)	44 (14.3)	31 (12.5)	13 (22.0)	0.060
Vascular tumor thrombus, n (%)	22 (7.2)	15 (6.0)	7 (12.1)	0.152
Nerve invasion, n (%)	17 (5.5)	12 (4.8)	5 (8.6)	0.334
Disease-free survival, n (%)	223 (72.6)	189 (76.2)	34 (57.6)	0.004
Overall survival, n (%)	265 (86.3)	221 (89.1)	44 (73.6)	0.003

Table 2 illustrates the association between high-risk patients and 5-year prognosis in colorectal cancer patients, presenting hazard ratios (HRs) with 95% confidence intervals (CIs) and P values across unadjusted and adjusted models. For DFS: In the unadjusted model, compared with high-risk patients, low-risk patients had a positive association with better DFS (P = 0.002). After adjusting for age and sex (Model 1), the association remained significant (P = 0.001). With further adjustment for TNM stage, differentiation, and tumor deposits (Model 2), the association marginally existed (P = 0.051). For OS, in the unadjusted model, low-risk patients showed a markedly higher possibility of OS (P = 0.003). This association persisted in Model 1 (P = 0.002) and Model 2 (P = 0.038). Kaplan-Meier curves of overall survival stratified by number of elevated tumor indicators (≥ 1 vs. 1) over 5 years could be showed in Fig. 2.

Table 2.

The association between Low-risk patients and 5-year prognosis in patients with colorectal cancer

Group	Unadjusted		Model 1		Model 2
	HR (95% Cl)	P value	HR (95% Cl)	P value	HR (95% Cl)	P value
DFS
Low-risk patients	2.06 (1.29–3.30)	0.002	2.13 (1.33–3.41)	0.001	1.63 (0.99–2.66)	0.051
OS
Low-risk patients	2.57 (1.36–4.84)	0.003	2.62 (1.39–4.95)	0.002	1.98 (1.03–3.78)	0.038

Model 1: Adjusted for age, sex

Model 2: Adjusted age, sex, TNM, differentiation, tumor deposits

Fig. 2.

Kaplan-Meier Curves of overall survival stratified by number of elevated tumor indicators (≥ 1 vs. < 1) over 5 years

Subgroup analysis for the primary outcome

This subgroup analysis evaluates the association between low-risk patients and outcomes (DFS: disease-free survival; OS: overall survival) according to tumor size, with results as Fig. 3: For tumor size > 4 cm: The hazard ratio (HR) was 1.36 (95%CI: 0.76–2.45), with a P-value of 0.296 (not statistically significant). For tumor size ≤ 4 cm: The HR was 2.98 (95% CI: 1.28–6.91), with a P-value of 0.010 (statistically significant). The P for the interaction between tumor size subgroups in DFS was 0.238 (no significant interaction). For tumor size > 4 cm: The HR was 1.32 (95% CI: 0.59–2.95), with a P-value of 0.499 (not statistically significant). For tumors ≤ 4 cm: The HR was 5.31 (95% CI: 2.67–16.2), with a P-value of 0.003 (statistically significant). The P for the interaction between tumor size subgroups in OS was 0.051 (approaching statistical significance).

Fig. 3.

The impact of elevated tumor indicators < 1 on endpoint event according to tumor size

In summary, among low-risk patients, tumor size ≤ 4 cm is significantly associated with worse DFS and OS, while no significant association is observed for tumors > 4 cm (P > 0.05). The interaction P-value for OS was 0.051 (marginally significant), and the subgroup findings should be interpreted with caution.

Discussion

Our study demonstrates that a preoperative tumor marker burden score (≤ 1 elevated markers) is an independent prognostic factor for OS in CRC patients, with a strong trend for DFS. The predictive value is most prominent in tumors ≤ 4 cm.

This composite score, derived from AFP, CA19-9, CA125, CA724, and CA153, provides a pragmatic and readily accessible tool for preoperative risk stratification. The biological plausibility of this finding is rooted in the concept that a higher burden of elevated circulating tumor markers may reflect a greater degree of tumor biological aggressiveness, increased metastatic potential, or a more pronounced host-tumor interaction favoring progression [19]. Aberrant glycosylation is a hallmark of cancer and is intimately associated with malignant phenotypes such as invasion and migration; the serum levels of glycoproteins like CA19-9 and CA125 are direct reflections of this pathological reprogramming, which also contributes to shaping an immunosuppressive tumor microenvironment [20–22]. Consequently, a lower score likely signifies a less aggressive tumor biology, translating into more favorable long-term outcomes.

The stratification of patients based on tumor size revealed a significant interaction, wherein the prognostic utility of the tumor marker score was particularly pronounced in patients with tumors ≤ 4 cm. In this subgroup, a score of ≤ 1 was robustly associated with superior DFS and OS, whereas this association was absent in patients with larger tumors. This differential effect suggests that the biological information captured by serum markers may be most clinically relevant in earlier-stage or locally confined disease, where the balance between tumor virulence and host response is more finely poised. In larger tumors, the overwhelming tumor burden and associated complex molecular landscape, potentially characterized by higher genomic instability [23] or distinct metabolic reprogramming [24], might eclipse the prognostic signal provided by serum markers alone. The absence of a significant interaction effect for DFS (P = 0.051 for OS interaction) further hints that the influence of tumor size and marker count on survival mechanisms may be more complex for overall survival, possibly involving factors beyond mere disease recurrence, such as treatment tolerance or non-cancer-related mortality.

The near-significant interaction for OS (P= 0.051) warrants careful consideration. It suggests that the combined effect of primary tumor size and systemic tumor marker burden may differentially impact the ultimate survival endpoint through mechanisms not fully captured by DFS. Larger tumors often present with more advanced T-stage and may be associated with a higher likelihood of occult metastatic disease or a systemic milieu that promotes progression. This is supported by research indicating that factors like lymph node count and the metastatic lymph node ratio provide superior prognostic information compared to traditional staging, reflecting the intensity of the host immune response and the extent of metastatic spread [25]. Furthermore, the tumor microenvironment plays a critical role; for instance, the density of stromal mast cells has been identified as an independent prognostic factor for overall survival, suggesting that local immune contexture significantly influences patient outcomes [26]. Therefore, in patients with large tumors, even a low serum marker score might be insufficient to counterbalance the adverse prognostic implications of a bulky primary lesion and its associated microenvironment.

Our results align with the growing emphasis on integrating multiple biomarkers for refined prognostication. The development of combined prognostic models that incorporate molecular markers, such as SLC44A4 expression, with clinical parameters has been shown to improve predictive ability beyond standard clinicopathological features alone [27]. Similarly, machine learning approaches are being leveraged to identify powerful discriminative biomarkers for early detection and prognosis, as demonstrated with genes like SULF1 and CXCL8 [28]. Our simple, cost-effective serum marker score operates on a similar integrative principle, translating multiple biological signals into a single prognostic index. This approach is complementary to more complex molecular techniques, such as ctDNA methylation assays for preoperative identification of lymph node metastasis, and may be particularly valuable in resource-limited settings [29].

The potential clinical applications of our findings are multifold. Preoperatively, this score could aid in patient counseling and inform the intensity of preoperative staging investigations. For patients with a low score and a small tumor, a less aggressive surgical strategy might be considered, whereas a high score could prompt a more extensive search for metastatic disease or influence decisions regarding the extent of lymphadenectomy, especially given that lymph node metastasis status is crucial for surgical management [30]. Furthermore, it could help stratify patients for adjuvant therapy trials, potentially identifying a subgroup with small tumors but a high marker score who might benefit from more intensive treatment. The score might also serve as a dynamic marker for monitoring treatment response and detecting recurrence, though this requires longitudinal validation.

Several limitations of our study must be acknowledged. Firstly, the retrospective and single-center nature of the analysis introduces the potential for selection bias. The relatively favorable 5-year survival rates in this cohort are attributed to all patients undergoing curative R0 resection, high rate of postoperative chemotherapy, and standardized perioperative management. Single-center retrospective design may introduce selection bias, which should be considered when comparing with other populations. Secondly, while we included five common markers, other promising serum biomarkers or novel molecular signatures, such as those derived from dietary restriction-responsive genes [31] or identified through Mendelian randomization studies of metabolites [32], were not evaluated and might provide additional prognostic information. The inclusion of CEA, a well-established marker in CRC, could have strengthened the panel, and its absence is a notable limitation. Also, We did not consider inflammatory indices, which might provide complementary prognostic information and improve the comprehensiveness of the marker panel. Thirdly, the precise biological mechanisms linking the elevation of each specific marker to tumor progression are not fully elucidated and likely vary, encompassing roles in adhesion, invasion, and immune modulation. Finally, our unweighted score based on the count of elevated markers prioritizes clinical simplicity and practicability; however, we acknowledge that ignoring the magnitude of marker elevation may lead to loss of prognostic information. Future studies may develop weighted scoring systems to better reflect biological gradients.

Conclusion

In conclusion, our study establishes that a preoperative tumor serum marker score of ≤ 1 is an independent prognostic factor for improved overall survival in colorectal cancer patients and shows a strong trend for disease-free survival. The prognostic strength of this score is particularly evident in patients with smaller tumors (≤ 4 cm), highlighting the interplay between local tumor burden and systemic biological aggression. This easily obtainable score provides a practical tool for enhancing preoperative risk stratification and could serve as a valuable adjunct to existing staging systems, ultimately contributing to more personalized treatment strategies for patients with colorectal cancer.

Authors’ contributions

DT and HH wrote the main manuscript text. JC prepared Figs. 1, 2 and 3. XL prepared Table 1, and 2. All authors reviewed the manuscript.

Funding

Suzhou Strengthening Health through Science and Education Project (grant: QNXM2024012).

Data availability

The datasets generated and/or analysed during the current study are not publicly available due to the datasets are owned by the institution only but are available from Jie Chen on reasonable request.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki. The study protocols were approved by the Ethics Committee of the First Affiliated Hospital of Soochow University (approval No. 2023 − 559). Due to the retrospective design of the study, the requirement for written informed consent was waived by the ethics committee. All patient data were de-identified and anonymized to protect patient confidentiality prior to data analysis.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.