Correlation between baseline magnetic resonance imaging features and serum carcinoembryonic antigen levels in patients with primary rectal cancer

Abstract

BACKGROUND

Serum carcinoembryonic antigen (CEA) levels and magnetic resonance imaging (MRI) findings are widely used for the diagnosis and treatment of rectal cancer; however, research investigating their correlation remains limited.

AIM

To investigate the correlation between baseline MRI features and serum CEA levels in patients diagnosed with primary rectal cancer.

METHODS

Eighty patients (age: 42-78 years) diagnosed with primary rectal cancer were enrolled. Baseline MRI examinations were performed to evaluate tumor size, T stage, circumferential resection margin status, extramural vascular invasion (EMVI), and lymph node metastasis. Serum CEA levels were concurrently measured. Statistical methods were used to analyze correlations.

RESULTS

Tumor size, T stage, EMVI, and lymph node metastasis were significantly correlated with serum CEA levels (P < 0.05). Multivariate analysis identified T stage and lymph node metastasis as independent factors influencing serum CEA levels.

CONCLUSION

This study confirmed the correlation between baseline MRI features and serum CEA levels in patients with primary rectal cancer, highlighting their potential utility for precise diagnosis, staging, and prognostic evaluation.

Key Words: Rectal cancer; Serum carcinoembryonic antigen; Magnetic resonance imaging; Tumor; Diagnosis

Core Tip: The correlation between magnetic resonance imaging (MRI) features and serum carcinoembryonic antigen (CEA) levels for diagnosing and treating rectal cancer currently remains under investigation. Accordingly, the present study explored this association in 80 patients diagnosed with primary rectal cancer at the authors’ hospital between July 2022 and August 2024. Patients were grouped according to clinical characteristics and treatment plans, and analysis confirmed a strong correlation between baseline MRI features and serum CEA levels.

INTRODUCTION

Colorectal cancer, particularly rectal cancer, is among the most common malignancies of the digestive tract and poses a significant threat to human health[1]. In recent years, the incidence of rectal cancer has been increasing annually due to lifestyle changes and aging populations[2,3]. According to global cancer statistics, rectal cancer ranks among the leading malignancies in terms of both incidence and mortality worldwide, with a substantial number of new cases and deaths reported each year[4,5]. In China, rectal cancer remains a critical public health issue that imposes a heavy burden on patients, families, and society.

There are multiple challenges to the diagnosis and treatment of rectal cancer. Early stage rectal cancer often presents with nonspecific symptoms, leading to delayed diagnosis and a high proportion of patients diagnosed at advanced stages of disease, thereby missing optimal treatment opportunities[6,7]. Surgery remains the primary treatment modality; however, for cases involving locally advanced or metastatic disease, monotherapy using surgery yields suboptimal outcomes, necessitating multimodal approaches, such as chemoradiotherapy[8]. Although essential, these treatments are associated with physical and psychological distress, variable efficacies, and high recurrence rates. Consequently, identifying effective diagnostic and prognostic biomarkers to improve early detection and therapeutic outcomes has become a focal point in oncology research[9,10].

Serum carcinoembryonic antigen (CEA), a widely used tumor biomarker, plays a pivotal role in colorectal cancer management. CEA is an acidic glycoprotein with embryonic antigenic properties and is minimally expressed in healthy adults, but is markedly elevated in individuals diagnosed with colorectal cancer due to tumor cell proliferation and secretion. Studies suggest that CEA not only aids in auxiliary diagnosis but is also correlated with disease staging, therapeutic response evaluation, prognosis prediction, and recurrence monitoring[11-13]. For example, serum CEA levels are closely associated with tumor size, depth of invasion, and lymph node metastasis, with higher levels typically observed in the advanced stages of disease. Postoperative decline(s) in CEA levels may indicate therapeutic efficacy, whereas persistent or recurrent elevation often signals tumor relapse or metastasis. Nevertheless, CEA lacks specificity for rectal cancer because its levels are influenced by confounding factors such as smoking, alcohol consumption, and inflammation, thus limiting its clinical utility[14-16].

Magnetic resonance imaging (MRI), a non-invasive imaging modality, has emerged as a cornerstone in the diagnosis and staging of rectal cancer. Due to its superior soft-tissue resolution, MRI enables precise visualization of the rectal wall layers, tumor morphology, size, depth of invasion, and involvement of adjacent structures, thereby guiding accurate staging. For example, MRI distinguishes early stage (T1 or T2) from locally advanced (T3 or T4) tumors, identifies lymph nodes or distant metastases, and provides a basis for tailored treatment strategies. Patients with T1/T2 tumors, extramural invasion < 5 mm, and no nodal metastasis may undergo total mesorectal excision (TME) directly[17,18]. In contrast, patients with extramural invasion > 5 mm and 1-3 metastatic lymph nodes typically undergo short-course radiotherapy followed by TME, whereas patients with extramural invasion > 5 mm, positive circumferential resection margin (CRM), or ≥ 4 metastatic lymph nodes require long-course chemoradiotherapy before surgery. Additionally, MRI is indispensable for postoperative surveillance, enabling early detection of recurrence.

Despite the individual clinical utility of serum CEA levels and MRI in rectal cancer management, existing studies investigating their correlation are sparse and predominantly limited by small cohorts or inconsistent imaging protocols. Elucidating this relationship could refine the clinical interpretation of CEA, enhance diagnostic accuracy, and optimize therapeutic decision-making, ultimately improving patient outcomes and quality of life[19,20]. Using a larger sample size (n = 80) and standardized MRI acquisition techniques, the present study aimed to strengthen the supportive evidence in this under-researched area.

MATERIALS AND METHODS

Study population

Data from 80 patients, who were diagnosed with primary rectal cancer at the authors’ hospital between July 2022 and August 2024, were included. Detailed demographic and clinical data (age, sex, tumor location, etc.) were recorded. Patients were stratified into subgroups according to their clinical characteristics and treatment protocols; for example: (1) T stage: T1-T2 vs T3-T4; (2) N stage: N0 vs N1-N2; and (3) CRM status: CRM-negative vs CRM-positive. This stratification facilitated an in-depth exploration of the correlations between the MRI features and serum CEA levels, thereby ensuring the accuracy and reliability of the findings.

Inclusion/exclusion criteria

Inclusion criteria: (1) Pathologically confirmed primary rectal cancer; (2) Availability of baseline MRI examinations performed before treatment with images meeting the quality standards for analysis; and (3) No history of antitumor therapy (surgery, chemotherapy, or radiotherapy) before MRI.

Exclusion criteria: (1) History of another malignancy; (2) Severe dysfunction of vital organs (heart, liver, and kidneys) contraindicating MRI; (3) MRI images with significant artifacts or poor quality that could impair analysis; and (4) Psychiatric disorders or cognitive impairment.

Baseline MRI examination

All patients underwent MRI using a Discovery MR750 scanner (GE Healthcare, Madison, WI, United States), which provides a high-field strength and resolution for detailed visualization of the rectal anatomy and surrounding tissues. Before imaging, patients fasted for 4-6 hours to minimize intestinal content, which could interfere with image quality.

Imaging protocols

T1-weighted imaging: Spin-echo (SE) sequence with the following parameters: (1) Repetition time (TR): 400-700 ms; (2) Echo time (TE): 8-15 ms; (3) Slice thickness: 3-5 mm; (4) Slice gap: 0.5-1 mm; (5) Field of view: 300-380 mm × 300-380 mm; and (6) Matrix: 256 × 256-320 × 320.

T2-weighted imaging: (1) Fast SE sequence, with TR 3000-5000 ms; (2) TE: 80-120 ms; and (3) Other parameters identical to those of T1-weighted imaging (T1WI).

Diffusion-weighted imaging: A single-shot echo-planar imaging sequence, with b-values of 0 and 1000 s/mm², TR of 3000-6000 ms, TE of 60-100 ms, and other parameters matching T1WI.

Contrast-enhanced MRI was performed when clinically indicated. A gadolinium-based contrast agent (e.g., gadopentetate dimeglumine) was administered intravenously at 0.1 mmol/kg, at an injection rate of 1.5-2.0 mL/second. Postcontrast scans were acquired during the arterial, venous, and delayed phases. Image analysis was performed independently by 2 experienced radiologists blinded to clinical data and serum CEA levels.

Key evaluations

Tumor size was measured on T2-weighted imaging (T2WI) images at the largest tumor cross section, with the average of the longest and shortest diameters recorded.

T stage: Determined as per the Union for International Cancer Control, 8^th Edition TNM staging criteria: (1) T1: Tumor invades submucosa; (2) T2: Tumor invades muscularis propria; (3) T3: Tumor extends beyond the bowel wall; (4) T4: Tumor invades adjacent organs/structures; and (5) N stage: Lymph nodes were classified as metastatic if short-axis diameter ≥ 5 mm, irregular in shape, or heterogeneous in signal intensity (Figure 1).

Figure 1 Representative magnetic resonance imaging images of rectal cancer T staging. A: T1; B: T2; C: T3; D: T4. Panels A-D correspond to T1-T4 stages, respectively, showing tumor invasion depth and adjacent tissue involvement on T2-weighted imaging and contrast-enhanced T1-weighted imaging sequences. CE-T1WI: Contrast-enhanced T1-weighted imaging; T2WI: T2-weighted imaging.

CRM status: CRM-positive was defined as a tumor-to-margin distance < 1 mm according to the European Society for Medical Oncology guidelines. Extramural vascular invasion (EMVI) was defined as the presence of tumor signals within the perirectal vessels on T2WI, according to the National Comprehensive Cancer Network criteria. Tumor location was classified as the upper, middle, or lower rectum based on the anatomical segments. Discrepancies between radiologists were resolved through consensus discussion.

Serum CEA measurement

Fasting venous blood samples (5 mL) were collected from patients on the morning of their baseline MRI examination using vacuum tubes containing a separation gel. The samples were centrifuged at 3000 ×g for 10 minutes to isolate serum for CEA analysis. CEA levels were measured via electrochemiluminescence immunoassay (ECLIA) using a laboratory analyzer Cobas e601 (Roche Diagnostics, Rotkreuz, Switzerland) and Roche Diagnostics CEA assay kits. ECLIA yields high sensitivity, specificity, and rapid turnaround. All procedures were performed in strict accordance with manufacturer’s instructions. Quality control was performed every 24 hours using Roche PreciControl Tumor Markers 1 and 2 to ensure accuracy and reproducibility.

RESULTS

Baseline characteristics of patients

Data from 80 patients [44 male (55.0%), 36 female (45.0%)] diagnosed with primary rectal cancer were included in this study, with baseline characteristics summarized in Table 1. The age of the patients ranged from 42 years to 78 years, with a mean ± SD age of 62.3 ± 8.7 years. The body mass index (BMI) ranged from 18.5 kg/m² to 29.6 kg/m², with a mean BMI of 24.1 ± 3.4 kg/m². Regarding tumor location, 12 (15.0%) cases were classified as upper rectal cancer, 34 (42.5%) as middle rectal cancer, and 34 (42.5%) as lower rectal cancer. Statistical analysis revealed no significant differences in age, sex, BMI, or other baseline characteristics between the groups (all P > 0.05), indicating balanced intergroup comparability. These findings support the validity of the subsequent correlation analyses (Table 1).

Table 1 Baseline characteristics of patients.

Characteristic	n	Percentage
Age (years)	42-78
Average value (mean ± SD)	62.3 ± 8.7
Gender
Male	44	55.0
Female	36	45.0
Body mass index (kg/m²)	18.5-29.6
Average value (mean ± SD)	24.1 ± 3.4
Tumor location
Upper rectum	12	15.0
Middle rectum	34	42.5
Lower rectum	34	42.5

Univariate analysis of baseline MRI features and serum CEA levels

Results of the univariate analysis of MRI features and serum CEA levels are summarized in Table 2. Tumor size exhibited a significant positive correlation with serum CEA levels (Spearman’s correlation coefficient r = 0.456, P < 0.01), indicating that larger tumors were associated with higher serum CEA levels. Similarly, T stage was significantly positively correlated with serum CEA levels (r = 0.387, P < 0.01), suggesting a gradual increase in CEA levels with advancing T stage. A positive CRM was positively correlated with serum CEA level (r = 0.325, P < 0.05), implying that patients with CRM involvement had relatively elevated CEA levels. EMVI also exhibited a positive correlation with serum CEA levels (r = 0.298,P < 0.05), with higher CEA levels in patients with EMVI. Lymph node metastasis was significantly associated with increased serum CEA levels (r = 0.356, P < 0.01), with CEA levels being markedly higher in patients with lymph node metastasis than in those without lymph node metastasis. These preliminary findings demonstrate a close correlation between baseline MRI features and serum CEA levels in patients with primary rectal cancer, providing a foundation for further investigation (Table 2).

Table 2 Univariate analysis of magnetic resonance imaging features and serum carcinoembryonic antigen levels.

Magnetic resonance imaging feature	n	Serum carcinoembryonic antigen level (ng/mL) (mean ± SD)	r value	P value
Tumor size (cm)			0.456	< 0.01
≤ 4	40	5.2 ± 1.8
> 4	40	8.3 ± 2.5
T-stage			0.387	< 0.01
T1-T2	30	4.8 ± 1.6
T3-T4	50	8.1 ± 2.9
Circumferential resection margin			0.325	< 0.05
Positive	28	8.9 ± 3.3
Negative	52	5.2 ± 2.1
Extramural vascular invasion			0.298	< 0.05
Yes	35	7.8 ± 2.7
No	45	5.3 ± 2.3
Lymph node metastasis			0.356	< 0.01
Yes	40	8.5 ± 2.8
No	40	4.9 ± 1.9

Tumor size: Patients were stratified according to median tumor diameter (4 cm). The < 4 cm group exhibited significantly lower CEA levels than the > 4 cm group (P < 0.01).

T-stage: Mean CEA levels were lower in T1-T2 tumors (4.8 ± 1.6 ng/mL) and significantly elevated in T3-T4 tumors (8.1 ± 2.9 ng/mL, P < 0.01), indicating a positive correlation between CEA levels and tumor invasion depth.

CRM: CEA levels were significantly higher in the CRM-positive group (8.9 ± 3.3 ng/mL) compared with the CRM-negative group (5.2 ± 2.1 ng/mL, P < 0.05).

EMVI: Patients with EMVI demonstrated a 47% increase in mean CEA levels compared to those without EMVI (7.8 ± 2.7 ng/mL vs 5.3 ± 2.3 ng/mL).

Lymph node metastasis: The metastasis group exhibited a 73% elevation in CEA levels compared with the non-metastatic group (8.5 ± 2.8 ng/mL vs 4.9 ± 1.9 ng/mL, P < 0.01).

Multivariate analysis

MRI features with statistical significance in the univariate analysis, tumor size, T-stage, CRM status, EMVI, and lymph node metastasis were included as independent variables in a multivariate logistic regression model. The results are summarized in Table 3. Tumor size [odds ratio (OR) = 2.563, 95%CI: 1.562-4.208, P < 0.01], T stage (OR = 1.875, 95%CI: 1.236-2.847, P < 0.01), and lymph node metastasis (OR = 1.684, 95%CI: 1.057-2.685, P < 0.05) were identified as independent risk factors that significantly influenced serum CEA levels. In contrast, CRM status (OR = 1.234, 95%CI: 0.856-1.782, P > 0.05) and EMVI (OR = 1.156, 95%CI: 0.789-1.697, P > 0.05) did not demonstrate independent effects on serum CEA levels in the multivariate analysis. The lack of significance of EMVI may reflect its close association with lymph node metastasis because tumor infiltration into the perirectal vessels often precedes lymphatic dissemination. This collinearity may attenuate the independent predictive value of elevated CEA levels. Nevertheless, EMVI remains a critical marker of advanced disease and systemic spread, consistent with its established role in risk stratification[20]. These results suggest that tumor size, T stage, and lymph node metastasis are critical MRI features that are closely associated with serum CEA levels in patients with primary rectal cancer, exhibiting a significant independent predictive value for evaluating serum CEA levels (Table 3). Tumor size, T-stage, and lymph node metastasis were identified as independent risk factors that were significantly associated with elevated serum CEA levels (all P < 0.05). CRM status (CRM status and EMVI) did not show statistically significant effects on serum CEA levels (P > 0.05).

Table 3 Multivariate logistic regression analysis of factors associated with serum carcinoembryonic antigen levels.

Independent variable	B	SE	Wald	Odds ratio	95%CI	P value
Tumor size	0.942	0.256	13.578	2.563	1.562-4.208	< 0.01
T-stage	0.629	0.201	9.745	1.875	1.236-2.847	< 0.01
Circumferential resection margin status	0.212	0.198	1.146	1.234	0.856-1.782	> 0.05
Extramural vascular invasion	0.145	0.205	0.503	1.156	0.789-1.697	> 0.05
Lymph node metastasis	0.521	0.239	4.736	1.684	1.057-2.685	< 0.05

B: Regression coefficient.

Development and performance evaluation of the predictive model

Model construction: A predictive model was constructed using tumor size, T stage, and lymph node metastasis as independent variables and serum CEA levels as the dependent variable. The model development process and performance metrics are listed in Table 4. The model demonstrated a sensitivity of 76.5% and specificity of 82.3%, indicating its ability to reliably identify patients with elevated serum CEA levels while accurately excluding those with low CEA levels. The area under the curve (AUC) was 0.856 (95%CI: 0.792-0.920), exceeding the threshold of 0.8, confirming the high predictive accuracy of the model for serum CEA levels in patients with primary rectal cancer. The Hosmer-Lemeshow test yielded a P value of 0.765 (P > 0.05), indicating good model calibration and reliability of the predicted outcomes. This predictive model serves as an effective clinical tool to assess serum CEA levels in patients diagnosed with primary rectal cancer, thereby enhancing the precision of treatment planning and prognostic evaluation (Table 4)

Table 4 Construction process of the prediction model and performance analysis indicators.

Indicator	Value
Sensitivity (%)	76.5
Specificity (%)	82.3
Area under the curve	0.856
95%CI	0.782-0.930
P value of Hosmer-Lemeshow test	0.765

Stratified analysis of model performance: A further stratified analysis of the predictive model performance is presented in Table 5. Across subgroups stratified according to T stage and lymph node metastasis status, the model yielded an AUC > 0.7, indicating consistent predictive utility in all subgroups. Notably, in the subgroup with T3-T4 stage disease and lymph node metastasis, the AUC reached 0.912 (95%CI: 0.865-0.958), reflecting a high predictive accuracy for serum CEA levels in patients with advanced disease. In contrast, the subgroup with stage T1-T2 disease and no lymph node metastasis exhibited a relatively low but clinically meaningful AUC of 0.735 (95%CI: 0.642-0.828). This stratified analysis comprehensively demonstrates the robust performance of the model across heterogeneous patient populations, providing a reference for clinicians to apply the model based on individual patient profiles (Table 5).

Table 5 Performance analysis of the prediction model in different subgroups.

Subgroup	n	Area under the curve	95%CI
Stage T1-T2 and no lymph node metastasis	18	0.735	0.623-0.847
Stage T1-T2 and with lymph node metastasis	12	0.802	0.698-0.906
Stage T3-T4 and no lymph node metastasis	21	0.834	0.721-0.947
Stage T3-T4 and with lymph node metastasis	19	0.912	0.835-0.989

DISCUSSION

The present study comprehensively investigated the correlation between baseline MRI features and serum CEA levels in patients diagnosed with primary rectal cancer and revealed a significant association between these parameters. In univariate analysis, MRI characteristics, such as tumor size, T-stage, CRM status, EMVI, and lymph node metastasis, were all significantly correlated with serum CEA levels. Larger tumor size was associated with higher serum CEA levels, which is likely attributable to enhanced tumor cell proliferation and metabolic activity in larger lesions, leading to increased CEA secretion into the bloodstream. T stage, which reflects tumor invasion depth, demonstrated a progressive increase in serum CEA levels with advancing stage, suggesting that elevation of CEA levels parallels tumor aggressiveness. A positive CRM, indicative of tumor involvement at the surgical resection margin, was correlated with elevated CEA levels, potentially signaling greater local invasiveness and a higher risk for recurrence or metastasis. EMVI, a pathway for hematogenous tumor spread, has been linked to increased CEA levels, implying systemic dissemination of cancer cells via the circulatory system. Lymph node metastasis, a key prognostic marker, was associated with markedly higher serum CEA levels compared with non-metastatic cases, further validating the role of CEA in assessing metastatic progression[20,21].

Multivariate analysis identified tumor size, T stage, and lymph node metastasis as independent risk factors for elevated serum CEA levels. Tumor size directly reflects tumor burden, with larger tumors likely driving greater CEA production. An advanced T-stage signifies deeper tissue infiltration and higher malignant potential, potentially amplifying CEA expression. Lymph node metastasis, a critical indicator of tumor spread, triggers a significant increase in serum CEA levels, underscoring its utility in monitoring disease progression. Clinically, these findings highlight the importance of integrating tumor size, T-stage, and lymph node status to predict serum CEA levels, thereby informing personalized therapeutic strategies and prognostic evaluations[22,23].

The predictive model, constructed using tumor size, T stage, and lymph node metastasis as independent variables and serum CEA level as the dependent variable, demonstrated high clinical utility. The model achieved a sensitivity of 76.5%, a specificity of 82.3%, and an AUC of 0.856. The Hosmer-Lemeshow test (P = 0.765) confirmed good model calibration. This model provides clinicians with a robust tool for accurately evaluating serum CEA levels in patients diagnosed with primary rectal cancer. By integrating MRI features, physicians can predict serum CEA levels, assess disease severity, and optimize treatment strategies. For example, patients predicted to have elevated CEA levels may benefit from aggressive neoadjuvant chemoradiotherapy for downstaging tumors, thereby improving surgical outcomes and survival rates.

Limitations

Despite its findings, this study had several limitations. The relatively small sample size may have limited the generalizability of the results. Future studies should expand the cohort to include patients from diverse geographical regions and ethnic backgrounds to enhance the reliability and external validity of the results.

CONCLUSION

Results of this study underscore the value of MRI-based risk stratification combined with serum CEA monitoring to guide the individualized management of rectal cancer, ultimately aiming to balance treatment efficacy with quality of life.