Perioperative serum carcinoembryonic antigen: Powerful marker for prognostic prediction and adjuvant chemotherapy decision-making in patients with stage II and III colorectal cancer

Abstract

BACKGROUND

Assessment of the prognosis, follow-up monitoring, and adjuvant treatment decision-making for patients with stage II and III colorectal cancer (CRC) are controversial, as CRC harbors tremendous heterogeneity. Carcinoembryonic antigen (CEA) is an important tumor marker; however, the use of this marker in the management of CRC has not garnered adequate attention.

AIM

To determine the significance of perioperative CEA levels in prognostic stratification and treatment decision making to provide personalized diagnosis and treatment for patients with stage II and III CRC.

METHODS

Patients in the training and validation cohorts were diagnosed with primary stage II or III CRC. Preoperative CEA (pre-CEA) and postoperative CEA (post-CEA) were collectively defined as perioperative CEA. Kaplan-Meier (K-M) survival analyses were used to describe patient survival. Cox stepwise regression analysis based on Akaike information criterion was used to determine the prognostic value of clinicopathological characteristics. Nomograms were developed to predict the probability of overall survival (OS) and disease-free survival (DFS). Annual hazard curves and pie charts were used to demonstrate the features of recurrence or metastasis. Differences were considered statistically significant at P < 0.05.

RESULTS

A total of 2496 and 1293 patients were included in the training and validation cohorts, respectively. K-M analysis indicated that patients with elevated perioperative CEA had poorer OS and DFS, with post-CEA being an independent prognostic factor for OS and DFS. Nomograms based on factors associated with prognosis were constructed, which showed good predictive ability for 3-, 5-, and 7-year OS and DFS. Patients with elevated perioperative CEA were more likely to have recurrence or metastasis, and the period of the second year after surgery was the peak time of recurrence or metastasis. OS and DFS were significantly worse in patients without adjuvant chemotherapy when they had elevated perioperative CEA. Adjuvant chemotherapy could significantly improve the OS of patients with elevated perioperative CEA. Patients with elevated post-CEA who received XELOX could achieve better OS and DFS.

CONCLUSION

Perioperative CEA demonstrate sufficient sensitivity in the prognosis prediction and follow-up of patients with stage II and III CRC. Furthermore, perioperative CEA, especially post-CEA, show promise in guiding adjuvant chemotherapy, suggesting potential for further study.

Key Words: Colorectal cancer; Perioperative carcinoembryonic antigen; Prognostic marker; Recurrence; Metastasis; Adjuvant chemotherapy

Core Tip: Perioperative carcinoembryonic antigen (CEA) is an important prognostic predictor and stratification marker, and it is reasonable to apply perioperative CEA to guide postoperative follow-up of stage II and III colorectal cancer (CRC) patients. The nomogram model based on perioperative CEA can well predict the 3-year, 5-year, and 7-year overall survival and disease-free survival of stage II and III CRC patients. Perioperative CEA can be used to assess recurrence or metastasis features of stage II and III CRC patients. Perioperative CEA, especially postoperative CEA, has the potential to guide adjuvant chemotherapy in stage II and III CRC patients, which is worthy of further study.

INTRODUCTION

Colorectal cancer (CRC) poses a major burden on public health and economic development globally due to its high morbidity and mortality[1]. The tumor-node-metastasis (TNM) staging system is an important parameter guiding the prognosis and treatment of CRC, with a higher TNM stage indicating a worse prognosis for patients. Based on this system, patients categorized as stage III and those categorized as stage II with high risk factors (HRFs) are recommended to receive fluorouracil-based adjuvant chemotherapy to improve prognosis[2].

However, with the standardization of follow-up work, researchers have found significant heterogeneity among these patients. Studies have highlighted considerable differences in the prognosis of patients at the same stage[3-7], and even the 5-year overall survival (OS) and disease-free survival (DFS) of patients with stage IIB/IIC cancer were lower than those of patients with stage IIIA cancer[8,9]. Furthermore, not all patients received adjuvant therapy benefit, and some patients can be cured with surgery alone[10-12]. When patients received standardized treatment, 15%-30% of them are unexpected to relapse[13-16]. However, some researchers have reported that some patients receiving adjuvant chemotherapy exhibited a lower peak of recurrence rates and a delayed peak month of recurrence rates vs those not receiving this treatment[17]. This finding reflects a shortcoming of the TNM system, emphasizing the need for more accurate and precise biomarkers to overcome the limitations of TNM in prognostic assessment and treatment decision-making.

To address these drawbacks, researchers have devised novel strategies and models to facilitate personalized diagnosis and treatment of patients with stage II and III CRC. For example, circulating tumor DNA (ctDNA) can be used as both a prognostic biomarker and a potential tool to guide adjuvant therapy[13,18-20]. Regrettably, this biomarker has not been widely disseminated owing to limitations in testing platforms and the associated costs[13]. Compared with the complexity and high cost of ctDNA, biomarkers that are inexpensive, easily detectable, and broadly accessible may offer greater advantages for precise stratification and personalized treatments.

Serum carcinoembryonic antigen (CEA) is a classical tumor marker that meets the above requirements[21]. It is an ideal tool to guide the diagnosis, treatment, and prognosis of cancer. Several studies have confirmed its suitability for accurate diagnosis and precise prognostic evaluation[22-24]. However, CEA has not received enough attention for use as personalized treatment. This multicenter retrospective study was conducted to determine the significance of perioperative CEA in prognostic stratification and treatment decision making to provide personalized diagnosis and treatment for patients with stage II and III CRC.

MATERIALS AND METHODS

This study was approved by the ethics committee of Harbin Medical University (No. HMUIRB2024014) and the Liaoning Cancer Hospital and Institute (No. 2019G0209), and was conducted following the Reporting Recommendations for Tumor Marker Prognostic Studies guidelines[25].

Patients

Patients with primary stage II and III CRC confirmed based on pathological diagnosis were enrolled in this study. Patients in the training cohort were selected from the Harbin Medical University Cancer Hospital from January 2014 and December 2017, whereas those in the external validation cohort were selected from the Liaoning Cancer Hospital and Institute from January 2010 and December 2019. All patients met the following inclusion criteria: (1) Underwent radical resection surgery without any neoadjuvant therapy; (2) Had complete clinicopathological and follow-up information available; and (3) Had serum CEA values available. Patients were excluded if anyone accompanied other malignancy at initial diagnosis (Figure 1).

Figure 1 Flowchart of patient selection. CRC: Colorectal cancer; CEA: Carcinoembryonic antigen.

Clinical information and demographic characteristics

Clinical information included: (1) Sex, age, height, weight, smoking history, drinking history, and other demographic characteristics; (2) Disease information, such as concomitant diseases, tumor sites, surgery time and approaches, postoperative pathological reports, and available CEA values; and (3) Adjuvant therapy and follow-up information. The pathological stage was defined according to the criteria of the 8^th American Joint Committee on Cancer manual for CRC.

Perioperative CEA was defined from 2 dimensions: Preoperative CEA (pre-CEA) and postoperative CEA (post-CEA). Pre-CEA data were extracted from the tumor-marker test report when they were initially diagnosed as CRC, and post-CEA data were extracted from the tumor-marker test report within 1 month after surgery and before receiving any adjuvant therapy. To facilitate clinical applications of the results and conclusions in this study, the cutoff values for pre-CEA and post-CEA were defined as 5 ng/mL, which are consistent clinical values[26,27]. To determine the significance of the dynamic changes in CEA, the CEA difference (CEA-diff) was defined to represent the dynamic change in CEA, which was defined as post-CEA minus pre-CEA, and the cutoff value was 0 ng/mL.

Statistical analysis

Data were analyzed using R software (version 4.0.3). The study endpoints were OS and DFS. OS was defined as the period from surgery to death from any cause, or the last contact, whereas DFS was defined as the period from surgery to local recurrence, distant metastasis, or death, whichever occurred first. Kaplan-Meier (K-M) survival analysis was used to describe patient survival. Cox stepwise regression analysis based on Akaike information criterion (AIC) was used to determine the prognostic value of the clinicopathological characteristics. Results from Cox analyses are presented as hazard ratios (HRs), 95% confidence intervals (CIs), and P values. Differences were considered significant at P < 0.05. Nomograms including possible prognostic factors were developed to predict the probability of the 3-, 5-, and 7-year OS and DFS.

Decision curve analyses (DCAs) of nomograms were also performed to compare the net benefits of the nomograms and evaluate the clinical applicability of the models. Receiver operating characteristic (ROC) curves and area under the curve (AUC) were used to test the predictive performance of the model. AUC > 0.9 indicated a strong predictive ability; AUC between 0.7 and 0.9 indicated good predictive ability; and AUC between 0.5 and 0.7 suggested moderate predictive ability[28]. Confirmatory analysis is used to determine the robustness of the results. The annual hazard rates of recurrence or metastasis were estimated using the kernel smoothing method. Annual hazard curves and pie charts were used to visually demonstrate the temporal and spatial features of recurrence and metastasis, respectively[29,30].

RESULTS

Patient characteristics

A total of 3789 patients with primary stage II and III CRC were enrolled in this study, which included 2496 patients in the training cohort and 1293 patients in the external validation cohort. The clinicopathological characteristics are presented in Table 1. The mean age was 66 years (range 27-91 years) in the training cohort and 61 years (range 21-87 years) in the validation cohort, with 1513 (60.6%) and 769 (59.5%) patients, respectively, in these cohorts being male. In the training cohort, the 3-, 5-, and 7-year OS rates were 89.4% (88.2%-90.7%), 79.5% (77.9%-81.2%), and 70.5% (68.4%-72.8%). The 3-, 5-, and 7-year DFS rates were 81.0% (79.4%-82.6%), 73.4% (71.6%-75.2%), and 66.9% (64.7%-69.1%). In the validation cohort, the 3-, 5-, and 7-year OS rates were 91.5% (89.9%-93.1%), 80.9% (78.2%-83.6%), and 77.6% (74.5%-80.9%). The 3-, 5-, and 7-year DFS rates were 85.8% (83.9%-87.9%), 78.1% (75.5%-80.9%), and 73.9% (70.6%-77.3%).

Table 1 Clinicopathological characteristics of patients, n (%).

Characteristics	Training cohort (n = 2496)	Validation cohort (n = 1293)	P value
Sex			0.518
Male	1513 (60.6)	769 (59.5)
Female	983 (39.4)	524 (40.5)
Age, mean (SD)	65.64 (10.14)	60.73 (10.33)	< 0.001
Height, mean (SD)	166.37 (9.16)	167.52 (28.99)	0.069
Weight, mean (SD)	66.58 (12.43)	66.25 (11.90)	0.432
Smoking			< 0.001
Negative	1370 (54.9)	889 (68.8)
Positive	1126 (45.1)	404 (31.2)
Drinking			0.853
Negative	1963 (78.6)	1021 (79.0)
Positive	533 (21.4)	272 (21.0)
Ileus			< 0.001
Negative	2139 (85.7)	1181 (91.3)
Positive	357 (14.3)	112 (8.7)
Diabetes			< 0.001
Negative	2338 (93.7)	1159 (89.6)
Positive	158 (6.3)	134 (10.4)
Hypertension			< 0.001
Negative	2304 (92.3)	1013 (78.3)
Positive	192 (7.7)	280 (21.7)
Differentiation			< 0.001
Well	7 (0.3)	173 (13.4)
Moderate	2060 (82.5)	1023 (79.1)
Poor	429 (17.2)	97 (7.5)
Signet-ring cell carcinoma			0.022
Negative	2444 (97.9)	1280 (99.0)
Positive	52 (2.1)	13 (1.0)
Mucinous carcinoma			< 0.001
Negative	2060 (82.5)	1132 (87.5)
Positive	436 (17.5)	161 (12.5)
Location			< 0.001
Rectum	1232 (49.4)	797 (61.6)
Left colon	684 (27.4)	267 (20.6)
Right colon	580 (23.2)	229 (17.7)
Vascular cancer embolus			0.051
Negative	2260 (90.5)	1196 (92.5)
Positive	236 (9.5)	97 (7.5)
Perineural invasion			0.397
Negative	2056 (82.4)	1080 (83.5)
Positive	440 (17.6)	213 (16.5)
Stage			< 0.001
II	1285 (51.5)	780 (60.3)
III	1211 (48.5)	513 (39.7)
Microsatellite stability			< 0.001
MSI-H	178 (7.1)	30 (2.3)
MSS	804 (32.2)	504 (39.0)
No detection	1514 (60.7)	759 (58.7)
Recurrence or metastasis site			0.002
No recurrence or metastasis	2158 (86.5)	1171 (90.6)
Liver	121 (4.8)	52 (4.0)
Lung	101 (4.0)	32 (2.5)
Recurrence	25 (1.0)	14 (1.1)
Abdominal pelvic and viscera in abdominal pelvic	46 (1.8)	16 (1.2)
Others (brain, bone, superficial lymph nodes, etc.)	45 (1.8)	8 (0.6)
Adjuvant chemotherapy type			< 0.001
Folfox	134 (5.4)	254 (19.6)
Xelox	981 (39.3)	391 (30.2)
Monotherapy	105 (4.2)	191 (14.8)
No adjuvant chemotherapy	1276 (51.1)	457 (35.3)

MSI-H: Microsatellite instability-high; MSS: Microsatellite stable.

Prognostic value of perioperative CEA

K-M survival analysis was performed to evaluate the association of perioperative CEA and prognosis. K-M survival analysis of the training cohort indicated that patients with elevated pre-CEA (OS: P < 0.001, DFS: P < 0.001) and elevated post-CEA (OS: P < 0.001, DFS: P < 0.001) demonstrated significantly poorer OS and DFS (Figure 2A-D). Cox regression analysis suggested perioperative CEA as a prognostic factor. However, when CEA was defined as a categorical variable, only pos-CEA demonstrated a higher association with OS and DFS (OS: HR = 2.338, 95%CI: 1.949-2.805, P < 0.001; DFS: HR = 2.406, 95%CI: 2.031-2.850, P < 0.001) (Supplementary Table 1). Cox stepwise regression analysis based on AIC revealed post-CEA as an independent prognostic factor for OS (HR = 1.0008, 95%CI: 1.0004-1.0012, P < 0.001) and DFS (HR = 1.0007, 95%CI: 1.0002-1.0011, P = 0.0022) (Tables 2 and 3). Similar statistically significant results were noted for the validation cohort (Figure 2E-H, Tables 4 and 5, Supplementary Table 2). These findings highlight the significant and stable prognostic values of pre- and post-CEA. To verify the robustness of the results, we included the cohorts as a variable in the model and found that the cohort source itself has no independent predictive value for the outcome, and its inclusion has minimal impact on the other variables in the model (Supplementary Table 3).

Figure 2 Kaplan-Meier survival analysis according to perioperative carcinoembryonic antigen. A-D: Kaplan-Meier curves for overall survival (OS) and disease-free survival (DFS) for patients with different pre-carcinoembryonic antigen (CEA) (A and B) and post-CEA (C and D) in the training cohort; E-H: Kaplan-Meier curves of OS and DFS for patients with different pre-CEA (E and F) and post-CEA (G and H) in the validation cohort. pre-CEA: Preoperative carcinoembryonic antigen; post-CEA: Postoperative carcinoembryonic antigen.

Table 2 Univariate and multivariate stepwise Cox regression analysis (stepped by Akaike information criterion) of overall survival in training cohort.

Variables	Level	Univariate analysis			Multivariate analysis
		HR	95%CI	P value	HR	95%CI	P value
Sex	Female (reference = male)	0.9630	0.8180-1.1300	0.6550
Age	Age (continuous)	1.0300	1.0200-1.0400	< 0.001	1.0330	1.0234-1.0417	< 0.001
Weight	Weight (continuous)	0.9980	0.9920-1.0000	0.5340
Height	Height (continuous)	1.0000	0.9950-1.0100	0.4010	1.0072	0.9987-1.0158	0.0950
Smoking	Positive (reference = negative)	1.0100	0.8620-1.1900	0.8890
Drinking	Positive (reference = negative)	1.1400	0.9460-1.3700	0.1690
Ileus	Positive (reference = negative)	1.3600	1.1100-1.6700	0.0032	1.3589	1.1019-1.6758	0.0041
Diabetes	Positive (reference = negative)	1.6000	1.2200-2.0900	0.0006
Hypertension	Positive (reference = negative)	1.2100	0.9210-1.5800	0.1730
Differentiation	Moderate (reference = well)	0.6260	0.2010-1.9500	0.4190	0.6163	0.1971-1.9266	0.4052
	Poor (reference = well)	0.9190	0.2920-2.8900	0.8850	0.8002	0.2535-2.5260	0.7039
Signet-ring cell carcinoma	Positive (reference = negative)	1.8500	1.1700-2.9100	0.0086
Mucinous carcinoma	Positive (reference = negative)	1.2300	1.0100-1.5000	0.0417	1.2369	1.0115-1.5125	0.0383
Location	Left colon (reference = rectum)	0.7820	0.6470-0.9450	0.0110	0.7762	0.6404-0.9407	0.0098
	Right colon (reference = rectum)	0.7540	0.6120-0.9290	0.0081	0.7744	0.6256-0.9586	0.0189
Perineural invasion	Positive (reference = negative)	1.8700	1.560-2.2400	< 0.001	1.6107	1.3309-1.9492	< 0.001
Vascular cancer embolus	Positive (reference = negative)	1.8700	1.4900-2.3400	< 0.001	1.3868	1.0981-1.7514	0.0060
Microsatellite stability	MSS (reference = MSI-H)	1.1100	0.7610-1.6100	0.5940
	No detection (reference = MSI-H)	1.6500	1.1600-2.3600	0.0056
AJCC stage	Stage III (reference = stage II)	1.9300	1.6400-2.2700	< 0.001	1.6919	1.4248-2.0090	< 0.001
Pre-CEA	Pre-CEA (continuous)	1.0030	1.0020-1.0040	< 0.001
Post-CEA	Post-CEA (continuous)	1.0010	1.0010-1.0010	< 0.001	1.0008	1.0004-1.0012	< 0.001
Pre-CA 19-9	Pre-CA 19-9 (continuous)	1.0000	1.000-1.000	0.0001
Post-CA19-9	Post-CA 19-9 (continuous)	1.0000	1.000-1.000	< 0.001

pre-CEA: Preoperative carcinoembryonic antigen; post-CEA: Postoperative carcinoembryonic antigen; AJCC: American Joint Committee on Cancer; CI: Confidence interval; HR: Hazard ratio; MSI-H: Microsatellite instability-high; MSS: Microsatellite stable.

Table 3 Univariate and multivariate stepwise Cox regression analysis (stepped by Akaike information criterion) of disease-free survival in training cohort.

Variables	Level	Univariate analysis			Multivariate analysis
		HR	95%CI	P value	HR	95%CI	P value
Sex	Female (reference = male)	0.9560	0.8220-1.1100	0.5560
Age	Age (continuous)	1.0200	1.0100-1.0300	< 0.001	1.0219	1.0141-1.0298	< 0.001
Weight	Weight (continuous)	1.0000	0.9950-1.0100	0.9010
Height	Height (continuous)	1.0100	0.9970-1.0100	0.1810	1.0087	1.0004-1.0171	0.0411
Smoking	Positive (reference = negative)	0.9490	0.8190-1.1000	0.4880
Drinking	Positive (reference = negative)	1.1800	0.9950-1.4000	0.0564
Ileus	Positive (reference = negative)	1.4800	1.2300-1.7900	< 0.001	1.4622	1.2071-1.7712	0.0001
Diabetes	Positive (reference = negative)	1.9400	1.5300-2.4600	< 0.001	1.5164	1.1925-1.9285	0.0007
Hypertension	Positive (reference = negative)	1.4300	1.1300-1.8200	0.0034
Differentiation	Moderate (reference = well)	0.7580	0.2430-2.3600	0.6320	0.7395	0.2368-2.3092	0.6035
	Poor (reference = well)	1.1000	0.3500-3.4400	0.8740	0.9363	0.2974-2.9481	0.9105
Signet-ring cell carcinoma	Positive (reference = negative)	1.8000	1.1700-2.7800	0.0078
Mucinous carcinoma	Positive (reference = negative)	1.2400	1.0300-1.4900	0.0206	1.2125	1.0070-1.4599	0.0420
Location	Left colon (reference = rectum)	0.8000	0.6720-0.9520	0.0122	0.7983	0.6689-0.9528	0.0126
	Right colon (reference = rectum)	0.7800	0.6440-0.9440	0.0107	0.7979	0.6553-0.9715	0.0246
Perineural invasion	Positive (reference = negative)	1.7900	1.5100-2.1300	< 0.001	1.5455	1.2952-1.8441	< 0.001
Vascular cancer embolus	Positive (reference = negative)	1.6700	1.3500-2.0700	< 0.001	1.2415	0.9940-1.5506	0.0565
Microsatellite stability	MSS (reference = MSI-H)	1.1000	0.7800-1.5600	0.5810
	No detection (reference = MSI-H)	1.6700	1.2100-2.3200	0.0020
AJCC stage	Stage III (reference = stage II)	1.9800	1.7000-2.3000	< 0.001	1.7244	1.4716-2.0205	< 0.001
Pre-CEA	Pre-CEA (continuous)	1.0030	1.0020-1.0040	< 0.001
Post-CEA	Post-CEA (continuous)	1.0010	1.0010-1.0020	< 0.001	1.0007	1.0002-1.0011	0.0022
Pre-CA 19-9	Pre-CA 19-9 (continuous)	1.0000	1.0000-1.0000	< 0.001	1.0006	1.0000-1.0013	0.0578
Post-CA 19-9	Post-CA 19-9 (continuous)	1.0000	1.0000-1.0000	< 0.001

pre-CEA: Preoperative carcinoembryonic antigen; post-CEA: Postoperative carcinoembryonic antigen; AJCC: American Joint Committee on Cancer; CI: Confidence interval; HR: Hazard ratio; MSI-H: Microsatellite instability-high; MSS: Microsatellite stable.

Table 4 Univariate and multivariate stepwise Cox regression analysis (stepped by Akaike information criterion) analyses of overall survival in validation cohort.

Variables	Level	Univariate analysis			Multivariate analysis
		HR	95%CI	P value	HR	95%CI	P value
Sex	Female (reference = male)	0.9700	0.7330-1.2900	0.8340
Age	Age (continuous)	1.0400	1.0200-1.0500	< 0.001	1.0339	1.0192-1.0489	< 0.001
Weight	Weight (continuous)	0.9930	0.9800-1.0000	0.2410
Height	Height (continuous)	0.9780	0.9610-0.9960	0.0191	0.9828	0.9649-1.0011	0.0650
Smoking	Positive (reference = negative)	0.9580	0.7070-1.3000	0.7840
Drinking	Positive (reference = negative)	0.7210	0.4920-1.0600	0.0946
Ileus	Positive (reference = negative)	1.2500	0.8420-1.8500	0.2680	1.5257	0.9942-2.3413	0.0532
Diabetes	Positive (reference = negative)	1.5100	1.0200-2.2600	0.0420
Hypertension	Positive (reference = negative)	1.4300	1.0500-1.9700	0.0256
Differentiation	Moderate (reference = well)	0.4860	0.3470-0.6820	< 0.001	0.5418	0.3803-0.7720	0.0007
	Poor (reference = well)	0.6200	0.3450-1.1100	0.1100	0.8004	0.4386-1.4608	0.4683
Signet-ring cell carcinoma	Positive (reference = negative)	5.2500	2.4600-11.200	< 0.001
Mucinous carcinoma	Positive (reference = negative)	1.1200	0.7730-1.6100	0.5590	1.0618	0.7103-1.5873	0.7699
Location	Left colon (reference = rectum)	0.8850	0.6230-1.2600	0.4950	0.8688	0.5923-1.2744	0.4718
	Right colon (reference = rectum)	1.0900	0.7730-1.5500	0.6130	0.9103	0.6058-1.3680	0.6512
Perineural invasion	Positive (reference = negative)	1.4300	1.0200-2.0200	0.0403	1.1350	0.7867-1.6376	0.4982
Vascular cancer embolus	Positive (reference = negative)	2.5600	1.7400-3.7500	< 0.001	1.9304	1.2780-2.9159	0.0018
Microsatellite stability	MSS (reference = MSI-H)	0.5320	0.2400-1.1800	0.1200
	No detection (reference = MSI-H)	0.7360	0.3440-1.5700	0.4270
AJCC stage	Stage III (reference = stage II)	2.0500	1.5500-2.7100	< 0.001	1.7645	1.3127-2.3717	0.0002
Pre-CEA	Pre-CEA (continuous)	1.0030	1.0020-1.0040	< 0.001
Post-CEA	Post-CEA (continuous)	1.0020	1.001-1.0030	0.0005	1.0015	1.0004-1.0025	0.0058
Pre-CA 19-9	Pre-CA 19-9 (continuous)	1.0000	1.0000-1.0000	< 0.001
Post-CA 19-9	Post-CA 19-9 (continuous)	1.0000	1.0000-1.0000	< 0.001

pre-CEA: Preoperative carcinoembryonic antigen; post-CEA: Postoperative carcinoembryonic antigen; AJCC: American Joint Committee on Cancer; CI: Confidence interval; HR: Hazard ratio; MSI-H: Microsatellite instability-high; MSS: Microsatellite stable.

Table 5 Univariate and multivariate stepwise Cox regression analysis (stepped by Akaike information criterion) of disease-free survival in validation cohort.

Variables	Level	Univariate analysis			Multivariate analysis
		HR	95%CI	P value	HR	95%CI	P value
Sex	Female (reference = male)	1.1300	0.8830-1.4500	0.3300
Age	Age (continuous)	1.0300	1.0200-1.0400	< 0.001	1.0268	1.0140-1.0398	< 0.001
Weight	Weight (continuous)	0.9910	0.9800-1.0000	0.0908
Height	Height (continuous)	0.9770	0.9610-0.9920	0.0034	0.9783	0.9624-0.9944	0.0084
Smoking	Positive (reference = negative)	0.9360	0.7140-1.2300	0.6330
Drinking	Positive (reference = negative)	0.7840	0.5630-1.0900	0.1500
Ileus	Positive (reference = negative)	1.3000	0.9090-1.8600	0.1500	1.5541	1.0495-2.3014	0.0277
Diabetes	Positive (reference = negative)	1.5400	1.0800-2.2000	0.0163	1.3728	0.9581-1.9672	0.0842
Hypertension	Positive (reference = negative)	1.4600	1.1100-1.9400	0.0076
Differentiation	Moderate (reference = well)	0.5430	0.3990-0.7390	0.0001	0.6312	0.4566-0.8726	0.0054
	Poor (reference = well)	0.7020	0.4190-1.1800	0.1790	0.9253	0.5435-1.5752	0.7748
Signet-ring cell carcinoma	Positive (reference = negative)	3.3600	1.5800-7.1200	0.0016
Mucinous carcinoma	Positive (reference = negative)	0.9560	0.6750-1.3500	0.8010
Location	Left colon (reference = rectum)	0.9310	0.6820-1.2700	0.6550	0.9343	0.6638-1.3148	0.6965
	Right colon (reference = rectum)	1.0100	0.7370-1.3900	0.9350	0.9149	0.6299-1.3287	0.6403
Perineural invasion	Positive (reference = negative)	1.3500	0.9910-1.8400	0.0573	1.1080	0.7997-1.5350	0.5377
Vascular cancer embolus	Positive (reference = negative)	2.2500	1.5700-3.2300	< 0.001	1.8155	1.2354-2.6679	0.0024
Microsatellite stability	MSS (reference = MSI-H)	0.6150	0.2960-1.2800	0.1930
	No detection (reference = MSI-H)	0.7550	0.3710-1.5300	0.4370
AJCC stage	Stage III (reference = stage II)	1.8800	1.4600-2.4100	< 0.001	1.6515	1.2721-2.1439	0.0002
Pre-CEA	Pre-CEA (continuous)	1.0030	1.0010-1.0040	< 0.001
Post-CEA	Post-CEA (continuous)	1.0010	1.0010-1.0020	0.0007	1.0012	1.0003-1.0022	0.0100
Pre-CA 19-9	Pre-CA 19-9 (continuous)	1.0000	1.0000-1.0000	< 0.001	1.0014	1.0005-1.0024	0.0041
Post-CA 19-9	Post-CA 19-9 (continuous)	1.0000	1.0000-1.0000	< 0.001

pre-CEA: Preoperative carcinoembryonic antigen; post-CEA: Postoperative carcinoembryonic antigen; AJCC: American Joint Committee on Cancer; CI: Confidence interval; HR: Hazard ratio; MSI-H: Microsatellite instability-high; MSS: Microsatellite stable.

T stage accounts for a high percentage of the prognosis of patients with stage II and III CRC and is an important factor in treatment decision-making[31,32]. Accordingly, the prognostic predictive ability of perioperative CEA and T stage by ROC curves was compared to further elucidate the prognostic value of perioperative CEA. AUCs in both the training and validation cohorts showed that post-CEA had a stronger ability to predict prognosis than T stage or pre-CEA (Supplementary Figure 1), indicating its importance in providing additional prognostic information for CRC.

Nomograms for prognosis

Considering the significant association between perioperative CEA and prognosis, nomograms based on factors selected using Cox stepwise regression analysis were constructed to predict the prognosis more accurately (Figures 3A and B, 4A and B). In both the cohorts, the calibration curves exhibited a high agreement for the 3-, 5-, and 7-year OS (C-index: 0.658, 95%CI: 0.646-0.670) and DFS (C-index: 0.652, 95%CI: 0.642-0.663) between the prediction by nomograms and the actual observations (Figures 3C and D, 4C and D). The ROC curves and AUCs indicated the good predictive ability of the models for the 3-, 5-, and 7-year OS (in the training cohort, the AUC for OS for 3, 5, and 7 years was 0.657, 0.679, and 0.683, respectively; in the validation cohort, the AUC for OS for 3, 5, and 7 years was 0.700, 0.671, and 0.680, respectively), and DFS (in the training cohort, the AUC of DFS for 3, 5, and 7 years was 0.666, 0.680, and 0.681, respectively; in the validation cohort, the AUC for DFS for 3, 5, and 7 years were 0.684, 0.655, and 0.703, respectively) (Figures 3E and F, 4E and F). The DCA suggested that the models showed positive net benefits in predicting the 3-, 5-, and 7-year OS and DFS (Figures 3G and H, 4G and H). These results indicated the feasibility of the model.

Figure 3 Nomograms for prognosis in the training cohort. A and B: Nomograms of overall survival (OS) and disease-free survival (DFS) based on the training cohort; C: Calibration curves for the nomogram of OS at 3, 5, 7 years in the training cohort; D: Calibration curves for the nomogram of DFS at 3, 5, 7 years in the training cohort; E: Receiver operating characteristic (ROC) curves for the nomogram of OS at 3, 5, 7 years in the training cohort; F: ROC curves for the nomogram of DFS at 3, 5, 7 years in the training cohort; G and H: Decision curve analyses for the nomograms of OS (G) and DFS (H) at 3, 5, 7 years in the training cohort. OS: Overall survival; DFS: Disease-free survival; AUC: Area under the curve; post-CEA: Postoperative carcinoembryonic antigen; AJCC: American Joint Committee on Cancer; CI: Confidence interval.

Figure 4 The performance of the nomograms in the validation cohort. A and B: Nomograms of overall survival (OS) and disease-free survival (DFS) based on the training cohort; C: Calibration curves for the nomogram of OS at 3, 5, 7 years in the validation cohort; D: Calibration curves for the nomogram of DFS at 3, 5, 7 years in the validation cohort; E: Receiver operating characteristic (ROC) curves for the nomogram of OS at 3, 5, 7 years in the validation cohort; F: ROC curves for the nomogram of DFS at 3, 5, 7 years in the validation cohort; G and H: Decision curve analyses (DCA) for the nomograms of OS (G) and DFS (H) at 3, 5, 7 years in the validation cohort. OS: Overall survival; DFS: Disease-free survival; AUC: Area under the curve; post-CEA: Postoperative carcinoembryonic antigen; AJCC: American Joint Committee on Cancer; CI: Confidence interval.

Recurrence and metastasis patterns based on perioperative CEA

After ascertaining the significant and stable prognostic value of perioperative CEA, the recurrence or metastasis pattern was then explored based on perioperative CEA to guide the follow-up in a more reasonable and personalized manner.

In the training cohort, the recurrence or metastasis rates in patients with elevated and normal pre-CEA were 17.92% and 10.69%, which were 27.58% and 11.04% in patients with high and low post-CEA (Table 6). When stratified based on pre-CEA or post-CEA, in terms of spatial characteristics, the liver and the lungs were the top 2 predominant risk sites of recurrence or metastasis (Figure 5A and B). Moreover, in terms of temporal characteristics, the dynamic hazed curves peaked at the time point of the second year after surgery, suggesting that patients at this time were at a high risk of recurrent or metastasis (Figure 5C and D).

Figure 5 Recurrence and metastasis patterns based on perioperative carcinoembryonic antigen. A: Relative frequencies of recurrence or metastasis sites for patients in the training cohort based on preoperative carcinoembryonic antigen (pre-CEA); B: Relative frequencies of recurrence or metastasis sites for patients in the training cohort based on postoperative carcinoembryonic antigen (post-CEA); C: Annual recurrence or metastasis hazard curves for patients in the training cohort based on pre-CEA; D: Annual recurrence or metastasis hazard curves for patients in the training cohort based on post-CEA; E: Relative frequencies of recurrence or metastasis sites for patients in the validation cohort based on pre-CEA; F: Relative frequencies of recurrence or metastasis sites for patients in the validation cohort based on post-CEA; G: Annual recurrence or metastasis hazard rates for patients in the validation cohort based on pre-CEA; H: Annual recurrence or metastasis hazard rates for patients in the validation cohort based on post-CEA. pre-CEA: Preoperative carcinoembryonic antigen; post-CEA: Postoperative carcinoembryonic antigen.

Table 6 Recurrence or metastasis rates in different carcinoembryonic antigen level.

Level	Recurrence or metastasis rate in training cohort (%)	Recurrence or metastasis rate in validation cohort (%)
Elevated pre-CEA	17.92	10.61
Normal pre-CEA	10.69	8.74
Elevated post-CEA	27.58	19.13
Normal post-CEA	11.04	7.75
Total	13.42	9.36

pre-CEA: Preoperative carcinoembryonic antigen; post-CEA: Postoperative carcinoembryonic antigen.

In the validation cohort, the recurrence or metastasis rates in patients with elevated and normal pre-CEA were 10.61% and 8.74%, respectively, and those of patients with elevated and normal post-CEA were 19.13% and 7.75%, respectively (Table 6). The spatial and temporal characteristics of patients with elevated pre-CEA and post-CEA were similar to those in the training cohort. However, the temporal characteristics of patients with normal pre-CEA and post-CEA demonstrated inconsistencies with those in the training cohort, wherein dynamic hazed curves peaked at the time point of the first year after surgery, and the risk then decreased with each year (Figure 5E-H). In conclusion, within 2 years after surgery, the liver and the lungs were the organs at the highest risk of recurrence or metastasis in patients with elevated CEA, especially those exhibiting elevated post-CEA.

Potential of perioperative CEA in guiding decision-making for adjuvant chemotherapy

As mentioned previously, perioperative CEA is an important value for prognostic guidance, and patients with elevated perioperative CEA have a higher risk of poor prognosis. Thus, the question arises whether it is necessary to provide a more aggressive treatment strategy for such patients to improve their prognosis. Therefore, we proposed a hypothesis of applying perioperative CEA for stratification to guide adjuvant chemotherapy in patients with stage II and III CRC.

To verify our hypothesis, we analyzed whether there was a difference in the prognosis of patients with or without adjuvant chemotherapy when they had different perioperative CEA. When patients received adjuvant chemotherapy, K-M survival analysis in the training cohort revealed that the higher the perioperative CEA, the worse the OS and DFS (all P < 0.001) (Supplementary Figure 2). However, in the validation cohort, only post-CEA showed a significant association with prognosis. In this cohort, the higher the post-CEA, the worse the OS and DFS (OS: Pre-CEA: P = 0.16, post-CEA: P = 0.001; DFS: Pre-CEA: P = 0.17, post-CEA: P < 0.001) (Supplementary Figure 3). When patients did not receive adjuvant chemotherapy, K-M survival analysis in both the cohorts suggested that the higher the perioperative CEA, the worse the OS and DFS (all P < 0.05) (Figure 6). These results suggested that post-CEA was significantly associated with OS and DFS among patients with or without adjuvant chemotherapy.

Figure 6 Potential of perioperative carcinoembryonic antigen in guiding treatment decision making. A-D: Kaplan-Meier curves of overall survival (OS) and disease-free survival (DFS) for patients without adjuvant chemotherapy in the training cohort, grouping based on pre-CEA (A and B), grouping based on post-CEA (C and D); E-H: Kaplan-Meier curves of OS and DFS for patients without adjuvant chemotherapy in the validation cohort, grouping based on pre-CEA (E and F),grouping based on post-CEA (G and H).

Next, the association between perioperative CEA and survival was evaluated between patients with stage II and III CRC who received or did not receive adjuvant chemotherapy. Adjuvant chemotherapy was found to significantly prolong the OS but not the DFS of patients in both cohorts (Supplementary Figure 4).

Whether patients with elevated perioperative CEA could derive survival benefits from adjuvant chemotherapy was clarified next. In both the cohorts, K-M survival analysis (Figure 7) revealed better OS in patients with elevated perioperative CEA after receiving adjuvant chemotherapy (all P < 0.05) but similar results were not obtained with respect to DFS (Figure 7). However, improved prognosis from adjuvant chemotherapy was not observed in patients with normal perioperative CEA (Supplementary Figures 5 and 6).

Figure 7 Impact of adjuvant chemotherapy on survival in patients with elevated perioperative carcinoembryonic antigen. A and B: Kaplan-Meier curves of overall survival (OS) and disease-free survival (DFS) for patients with elevated pre-CEA in the training cohort who did or did not receive adjuvant chemotherapy; C and D: Kaplan-Meier curves of OS and DFS for patients with elevated post-CEA in the training cohort who did or did not receive adjuvant chemotherapy; E and F: Kaplan-Meier curves of OS and DFS for patients with elevated pre-CEA in the validation cohort who did or did not receive adjuvant chemotherapy; G: And H: Kaplan-Meier curves of OS and DFS for patients with elevated post-CEA in the validation cohort who did or did not receive adjuvant chemotherapy.

Based on these results, it could be logically confirmed that adjuvant chemotherapy can significantly improve the OS of patients with elevated perioperative CEA. However, whether adjuvant chemotherapy can improve the DFS of patients with elevated perioperative CEA or whether patients with normal perioperative CEA can derive survival benefits from adjuvant chemotherapy is doubtful.

Potential of perioperative CEA in guiding the selection of adjuvant chemotherapy strategies

Whether perioperative CEA have any value in guiding the selection of adjuvant chemotherapy strategies was further explored after clarifying that patients with elevated perioperative CEA could derive survival benefits from adjuvant chemotherapy. We explored whether there was a difference in the prognosis of patients with elevated perioperative CEA when they received different kinds of adjuvant chemotherapy was subsequently explored. K-M survival analysis revealed that patients in the training cohort with elevated post-CEA who received XELOX could achieve better OS and DFS (OS: P = 0.00044, DFS: P = 0.07). However, patients with elevated pre-CEA did not have significantly different prognoses (OS: P = 0.28, DFS: P = 0.41) (Figure 8A-D). K-M survival analysis supported that patients in the validation cohort with elevated post-CEA who received XELOX could achieve better OS and DFS (OS: P = 0.00029, DFS: P = 0.0016) and those with elevated pre-CEA had no significantly different prognoses (OS: P = 0.098, DFS: P = 0.15) (Figure 8E-H). These results affirmed the potential of post-CEA in guiding the selection of adjuvant chemotherapy strategies.

Figure 8 Potential of perioperative carcinoembryonic antigen in guiding the selection of adjuvant chemotherapy strategies. A and B: Kaplan-Meier curves of overall survival (OS) and disease-free survival (DFS) for patients with elevated pre-CEA in the training cohort after receiving different adjuvant chemotherapy strategies; C and D: Kaplan-Meier curves of OS and DFS for patients with elevated post-CEA in the training cohort after receiving different adjuvant chemotherapy strategies; E and F: Kaplan-Meier curves of OS and DFS for patients with elevated pre-CEA in the validation cohort after receiving different adjuvant chemotherapy strategies; G and H: Kaplan-Meier curves of OS and DFS for patients with elevated post-CEA in the validation cohort after receiving different adjuvant chemotherapy strategies.

Prognostic value of dynamic changes in perioperative CEA and their role in guiding adjuvant therapy

To evaluate the prognostic significance of the dynamic changes in CEA and their value in guiding adjuvant therapy, CEA-diff was defined to represent the dynamic changes in CEA (defined as post-CEA minus pre-CEA). In the training cohort, findings from K-M survival analysis revealed that patients with CEA-diff < 0 ng/mL could achieve better OS and DFS (OS: P = 0.02, DFS: P = 0.0061) (Supplementary Figure 7A and B). Unfortunately, despite consistent trends observed in the validation cohort, the results were not statistically significant (Supplementary Figure 7C and D). ROC curves and AUC values further indicated that CEA-diff underperformed significantly compared with T staging, pre-CEA, or post-CEA in predicting prognoses (Supplementary Figures 1 and 7E-H). When recurrence and metastasis patterns were predicted based on CEA-diff, the results were similar to those obtained for perioperative CEA (Supplementary Figure 8). Frustratingly, the performance of CEA-diff was utterly dismal when used to guide decision-making for adjuvant chemotherapy (Supplementary Figures 9-11).

DISCUSSION

CEA is a commonly used tumor markers for uninterrupted monitoring during follow-up to predict outcomes and aid in the detection of possible recurrent or metastatic events[33-35]. However, its role in predicting prognoses still lacks consensus, and its role in guiding treatment decision-making is not yet widely recognized. In the prophase studies, our research team confirmed post-CEA as an important prognostic factor that significantly improved the performance of the TNM model for stage III colon cancer. The model combined post-CEA with T and N and was the best prognostic model for stage III colon cancer[23]. For patients with stage II CRC, early post-CEA was a better biomarker for prognosis than T stage and pre-CEA. It demonstrates potential as a biomarker indicating HRF to guide the prognosis and treatment of patients with stage II CRC[22]. This retrospective study was conducted based on a previous study, and our goal of clarifying the value of perioperative CEA in prognosis and adjuvant treatment decision-making was successfully achieved.

One of the main controversies currently surrounding the role of perioperative CEA in prognosis is the detection time of CEA. Initially, researchers focused only on pre-CEA[36-38]. For example, Tarantino et al[36] demonstrated pre-CEA levels as a strong predictor of decreased OS, and they advocated the use of pre-CEA levels as a prognostic factor in the preoperative assessment of patients with rectal cancer. However, in recent years, more researchers have begun to pay attention to the value of post-CEA[23,33,39,40]. Lin et al[33] has confirmed early post-CEA as a better prognostic indicator than pre-CEA in predicting the prognosis of patients with curable CRC. The concept of perioperative CEA was defined in this study for the more objective evaluation of the value of CEA. Here, pre-CEA was applied only to patients who did not receive neoadjuvant therapy. Although neoadjuvant therapy is a standard approach for locally advanced CRC, pre-CEA have also been proven to play an important role in neoadjuvant therapy decision-making[41,42]. However, evidence suggested that serum CEA levels change significantly before and after neoadjuvant therapy[43]. Therefore, we excluded patients who had received neoadjuvant therapy to ensure scientific rigor and rationality of the study. Similarly, based on the half-life of CEA and the characteristics of tumor burden changes after surgery[44,45], post-CEA were defined as CEA levels within 1 month after surgery and before receiving any adjuvant therapy. The cutoff value for the CEA was defined as 5 ng/mL, which is a consistent value routinely used in clinical settings. Although some studies have proposed a new cutoff value for CEA levels[46,47], 5 ng/mL remains a moderate, inclusive value that is recognized by both clinical practitioners and researchers[24,48,49]. Therefore, 5 ng/mL was ultimately selected as the best cutoff value for this study.

In this study, patients with elevated perioperative CEA were found to have a worse prognosis and a higher risk of recurrence and metastasis. The Cox regression analysis, ROC curves and AUC results indicated that compared to pre-CEA or T stage, post-CEA demonstrated more robust performance in prognostic prediction and it was more suitable to be a prognostic marker. These findings highlight the more robust ability of post-CEA for prognostic assessment and prediction. A meta-analysis has emphasized the significance of post-CEA in prognosis[50]. The continuous high level of CEA after surgery indicates the presence of minimal residual disease, an important marker of poor prognosis and adjuvant chemotherapy. It is therefore reasonable to infer that post-CEA performance is better than pre-CEA performance. Additionally, nomograms incorporating post-CEA but not pre-CEA were developed and shown to provide reliable assistance for prognostic assessment. ROC curves and AUCs of nomograms, pre-CEA, post-CEA and T stage also showed the performance of the nomograms in prognostic prediction was the most robust, which emphasized the necessity of constructing nomograms (Figure 3 and Supplementary Figure 1). The occurrence and progression of cancers are influenced by multiple factors. Therefore, nomograms incorporating various statistically significant factors are necessary. Some researchers have developed and validated nomograms to predict survival in CRC. Dai et al[51] have developed nomograms based on pre-CEA/post-CEA levels, whereas the nomograms constructed by Lou et al[52] and Dai et al[53] only include pre-CEA levels. Li et al[54] have emphasized the importance of dynamic changes for post-CEA. In such a situation, identifying the appropriate dimension of CEA that is more valuable for prognosis is an important decision. The nomograms we constructed that includes post-CEA is a complement to this field.

CEA is currently considered an important indicator of postoperative follow-up to promptly determine possible recurrence or metastasis events. However, its accuracy and sensitivity are controversial, and the relationship between CEA and recurrence or metastasis is still uncertain. A systematic review included 42 original studies and assessed the diagnostic accuracy of CEA in determining recurrence after intended curative surgery for primary CRC. Their perspective was that CEA levels did not effectively detect treatable recurrences at an early stage and that a clinically relevant effect on patient mortality is yet to be proven[55]. Another study group has suggested that measuring CEA can, with high accuracy, help identify patients who are likely to demonstrate recurrence only within the first year after surgery[56]. Based on these conflicting reports, we explored recurrence or metastasis patterns based on CEA. We discovered that within 2 years after surgery, the liver and the lungs were the highest risk sites for recurrence or metastasis in patients with elevated CEA, especially those having elevated post-CEA. Similar to our findings, a systematic review reported that in patients with CRC, the liver and the lungs were the most common sites of recurrence or metastasis and that approximately 30%-50% of recurrence or metastasis occurs within the first 2 years[57]. A Japanese study has reported similar findings[58]. Numerous studies have confirmed the temporal characteristics of recurrence and metastasis in patients with CRC[17,59]. Collectively, it can be concluded that frequent surveillance is warranted during the first 2 years after surgery, especially in high-risk areas such as the chest and the abdomen.

In our study, it was also demonstrated that the perioperative CEA, especially post-CEA, had the potential to guide adjuvant chemotherapy in patients with stage II and III CRC. Adjuvant chemotherapy-related decision-making for patients with stage II and III CRC remains controversial[10,60-62]. Therefore, identifying novel markers to guide adjuvant chemotherapy decision-making and making it more personalized is of utmost importance. Some researchers have recently focused on the possibility of applying CEA in guiding adjuvant therapy. However, their study conclusions are not uniform[63-65]. Liu et al[64] found that adjuvant chemotherapy had no substantial survival benefits in stage IIA colon cancer with elevated pretreatment serum CEA levels. The retrospective analysis by Pu et al[63] indicates that elevated post-CEA should be considered a HRF in guiding adjuvant chemotherapy for stage II colon cancer. Inoue et al[66] have reported that patients with T4N0 and T1-3N1 colon cancer and a high pre-CEA can benefit from adjuvant chemotherapy with oxaliplatin for 6 months. Furthermore, Auclin et al[67] used the date of the multicenter international study of oxaliplatin/5-fluorouracil (5-FU)/Leucovorin (LV) in the adjuvant treatment of colon cancer (MOSAIC trial) and reported CEA as a strong prognostic factor for DFS and OS in stage II colon cancer. Only high-risk patients with stage II colon cancer with post-CEA levels > 2.35 ng/mL benefited from the treatment of addition of oxaliplatin to LV + 5-FU (LV5FU2). Our findings are in agree with the conclusions from the above studies, confirming the important role of perioperative CEA, especially post-CEA, in adjuvant treatment decision-making, thereby providing new ideas for this controversial field and indicating that the use of perioperative CEA in guiding adjuvant chemotherapy deserves further exploration.

It is widely recognized that a 3/6 months adjuvant chemotherapy of XELOX or FOLFOX is the standard treatment for CRC patients of stage II with HRFs and CRC patients of stage III. However, considering factors such as toxicity and economic effects, an increasing number of researchers have raised questions regarding the choice of treatment plans and treatment duration. Based on the results of the IDEA and TOSCA studies suggested 3-month adjuvant chemotherapy with XELOX may be considered for stage II CRC patients with HRFs except for T4 and low-risk stage III patients with T1-3N1[5-7]. Encouragingly, in this study, K-M survival analysis revealed that patients with elevated post-CEA who received XELOX could achieve better OS and DFS. This finding may suggest that post-CEA holds a higher priority over pre-CEA in guiding adjuvant therapy decision-making. However, these results still require validation through prospective studies.

Lastly, we determined the monitoring value of dynamic changes in perioperative CEA; however, our attempts did not yield the desired results. Some possible reasons that may have affected our research outcomes and conclusions are documented. First, the timing and frequency of CEA measurement are important yet controversial factors. Studies have conducted long-term and standardized monitoring of CEA. Although reliable results have been obtained for each study, a consistent conclusion regarding the timing and frequency of CEA was not reached[48,49,56,68]. This deficit suggests the requirement of a more scientific design for the timing and frequency of monitoring CEA. Furthermore, a meta-analysis of published articles should be undertaken to determine the differences and common aspects in monitoring CEA across studies. Some researchers have emphasized that, in addition to the dynamic monitoring of CEA, combining the monitoring of CA 19-9 and CA 125 was more helpful in assessing the prognosis of patients with CRC[48]. This finding emphasizes the necessity of the combined monitoring of biomarkers and provides new insights for our subsequent research. Accordingly, the importance of dynamic changes in CEA will be the focus of our subsequent research, wherein the credibility of the study will be enhanced based on more comprehensive and rigorous experimental designs.

To summarize, the value of perioperative CEA in prognostic stratification and follow-up monitoring in patients with stage II and III CRC was explored in this study. More importantly, the potential of perioperative CEA in assisting adjuvant chemotherapy decision-making was clarified. Nevertheless, our study still has some shortcomings. First, we could not verify the value of dynamic changes in CEA. However, it should be noted that this aspect has been verified in some other studies[24,69]. This will therefore be the focus of our future research. Another disadvantage is that our retrospective analysis yields less robust conclusions compared with prospective studies. Accordingly, the results should be verified by conducting prospective studies involving multicenter cohorts. Especially, the proposal of applying perioperative CEA to guide adjuvant therapy requires support from prospective studies, which will be the focus of our team's future research.

CONCLUSION

Perioperative CEA are important prognostic predictor and stratification marker. It is reasonable to apply perioperative CEA in guiding the follow-up of patients with stage II and III CRC. Perioperative CEA, especially post-CEA, are important prognostic marker that has the potential to guide adjuvant chemotherapy in patients with stage II and III CRC, making this topic worthy of further investigation.

ACKNOWLEDGEMENTS

Thanks for all the participants included in this study.