Colorectal cancer (CRC) ranks as the world’s second most deadly malignancy (1). Despite advancements in surgical techniques and integrated therapies, the clinical outcomes for CRC patients remain unsatisfactory. Approximately fifteen percent of stage II patients and thirty percent of stage III patients experience recurrence even after radical resection (2, 3). The high recurrence and mortality rates have increasingly drawn attention to the need for individualized treatment and prognosis of this disease. Clinicians currently rely on the TNM staging system to predict and assess the prognosis of patients with colorectal cancer (4). While the current staging system provided essential insights into tumour behavioral characteristics, it doesn’t fully encompass vital determinants of patient prognosis, such as age, serum tumour markers(STMs) and so on. Consequently, there’s an imperative demand to unearth novel markers for individualized prognostic assessment, empowering clinicians to offer more precise counsel on survival forecasts and therapeutic approaches for CRC patients.

Owning to its simplicity and cost-effectiveness, tumour marker detection is extensively performed in medical institutions. Carcinoembryonic antigen (CEA) and carbohydrate antigen 19-9 (CA19-9) represent the primary STMs for preoperative evaluation and postoperative follow-up examination of CRC patients. CEA is an acidic glycoprotein associated with oncogenic advancement (5). Some clinical guidelines recommend CEA as a prognostic biomarker for CRC and endorse its routine measurement after radical resection in CRC patients (6, 7). CA19-9 is closely linked to recurrence and survival in colorectal cancer (8). Significantly, combined tumour marker testing has significantly improved predictive accuracy compared with single marker testing (9). Concurrently, the number of positive tumour markers(NPTMs) is gaining attention (10, 11). Previous research has demonstrated its feasibility as a prognostic factor for stage II-III CRC (12). However, while this study has underscored the impact of NPTMs before surgery on prognosis, the significance of postoperative STMs remains underexplored. Recently, some researches have paid attention to the role of postoperative STMs and found that they are also promising indicators (9, 13, 14). It has also been shown that the number of positive tumour markers before and after treatment is important for the prognosis of rectal cancer (15, 16). Therefore, we believe that combining both preoperative and postoperative CEA and CA19-9 measures might enhance predictive accuracy.

While clinical predictive models are endorsed for estimating the recurrence and survival of diverse malignancies due to their utility and comprehensiveness (17, 18), no research has incorporated NPTMs before and after surgery into these models for stage I-III CRC. Recognizing the vital prognostic implications of NPTMs, we evaluated the association of preoperative and postoperative CEA and CA199 with OS and DFS in patients with stage I-III CRC who underwent radical resection. Age, sex, pN stage, NPTMs before and after surgery were chosen to construct the clinical prediction models of overall survival(OS) and disease-free survival(DFS). Additionally, we further compared the clinical value of these models with that of the AJCC 8th TNM stage.

Researches on biomarkers of gastrointestinal cancer have been widely concerned. A study on dogs has explored the important role of lipopolysaccharide in intestinal carcinogenesis (22), and Li et al.’ s study on mice suggested that the secretory protein cathepsin K can be used as a new predictive biomarker for CRC (23). In addition, previous studies have confirmed the potential prognostic value of absolute quantification of free circulating DNA (24)and long non-coding RNA plasmacytoma variant translocation 1 (25) in CRC patients as biomarkers. Serum CEA and CA19-9 are common and cost-effective biomarkers in clinical practice and they are instrumental in predicting the prognosis of CRC, holding significant value in both pre- and post-operation (26, 27). However, previous researches have primarily paid attention to the prognostic significance of preoperative STMs (28–30), with little attention given to postoperative CEA and CA19-9. Study has demonstrated that in terms of predicting survival duration, combined tumour markers assessments hold an advantage over single marker tests (31). In recent years, NPTMs, introduced as a novel reference index, has exhibited profound prognostic potential (11, 15). Given their significant clinical value, this study categorized patients according to NPTMs, assessed the prognostic significance of combined STMs detection before and after surgery, and subsequently developed clinical prediction models.

In this study, we found that the increased NPTMs before and after surgery were associated with a poor prognosis of CRC. Furthermore, NPTMs before surgery was closely related to the TNM stage and tumour location, consistent with prior findings (11). Studies reported that patients with normal STMs after surgery possessed a notably better prognosis compared with patients with abnormal STMs (13, 15). This research also supported this result. Compared with patients with normal postoperative CEA and CA19-9, patients with both tumour markers positive postoperatively had approximately a 4.2-fold increased risk of death and a 3.0-fold increased risk of recurrence. We also discovered that both preoperative and postoperative positive CEA and CA19-9 were more likely to occur in population with higher pTNM stage, higher pN stage, and those with neural/vascular invasion. For these patients, a more intensive follow-up strategy should be implemented.

The role of circulating tumour DNA (ct-DNA) in predicting the prognosis of colorectal cancer has garnered widespread attention (32, 33). Study has reported a correlation between CT-DNA in tumour cells and residual microcancer cells, but its clinical application remains limited due to its high costs (34). Conversely, tumour marker detection is affordable and easy to operate. Konishi et al. reported that patients with elevated postoperative CEA faced a higher hazard of early recurrence, especially within the first year after radical surgery (13). Sonoda et al. found that elevation of CEA post-surgery is independently correlated with an unfavorable prognosis in stage II-III CRC (14). In this study, we found that NPTMs before and after surgery were independent prognostic factors for OS and DFS in patients with stage I-III colorectal cancer. The elevation of tumour marker levels postoperatively may suggest the presence of unrecognized residual minute cancer cells at the time of surgery or in postoperative radiological examinations, which raises the possibility of relapse (35, 36). Therefore, in clinical practice, it is essential not only to perform combined tumour markers testing before surgery for colorectal cancer patients but also to pay attention to postoperative combined tumour markers testing. Patients with positive tumour markers might benefit from comprehensive treatment and require followed-up regularly.

TNM stage is commonly used for prognosis prediction and assessment, but its ability to predict patient outcomes may be limited (37). Nomogram is a powerful graphical prediction tool that illustrates the likelihood of a specific event occurring based on multiple variables (17). Compared with the TNM stage, the nomogram is more intuitive and easier to understand. Moreover, it can incorporate more risk factors, significantly enhancing the accuracy of prediction. Previous studies have developed CRC-related survival prediction models based on STMs (38, 39), but they were limited in sample size and did not focus on the prognostic significance of post-surgical STMs. This might have restricted their predictive accuracy to some extent. Therefore, this study constructed two more comprehensive clinical prediction models based on NPTMs to help clinicians predict and evaluate the prognosis. Both nomograms included five variables: sex, age at surgery, pN stage, NPTMs before and after surgery. The pN stage and NPTMs after surgery had significant effects on total scores of two models.

The C-index results indicated that the discriminative ability of the two models is significantly superior to the TNM stage. Risk stratification analysis revealed that the predictive models for OS and DFS exhibited commendable discrimination proficiency. Further, the results of DCA underscored that our models exhibited superior performance in clinical decision-making compared with TNM stage. The calibration curves for both groups also confirmed the strong concordance between the predictive model and the actual outcomes. NRI and IDI are two statistical indicators used to assess the enhanced performance of predictive models. Through their comparative analysis, we can discern the differential performance of various models and select the optimal one (40, 41). Within our investigation, both NRI and IDI metrics indicated that the novel models had superior accuracy and discriminatory ability in forecasting 3-year and 5-year OS and DFS for CRC patients. To summarize, both prognostic models exhibited strong predictive efficacy and clinical applicability, and can be utilized in clinical settings to forecast the prognosis of stage I-III CRC patients.

This study had several limitations. First of all, both the discovery cohort and validation cohort were established through random grouping, which could lead to imbalances at baseline. For instance, distinct variations in clinicopathological characteristics were observed in both cohorts (age, P < 0.05). Secondly, for the constructed model, the validation cohort performed better than the discovery cohort in some aspect. Thirdly, since stage IV patients receive different treatment methods from stage I–III patients, we did not include stage IV patients in our study, so the nomograms cannot be applied to stage IV CRC patients. Finally, Some factors that might be associated with prognosis, such as BRAF and KRAS mutation status and nutritional condition, were not considered in our study. Future studies should incorporate more valuable prognostic factors and conduct external validation.

NPTMs, both preoperatively and postoperatively, were closely related to the prognosis of stage I-III colorectal cancer patients. Compared with the AJCC 8th TNM stage, the nomograms based on preoperative and postoperative CEA and CA19-9 demonstrated superior predictive capability and clinical applicability, offering more precise prognosis for colorectal cancer patients. The results of this study suggest that preoperative and postoperative CEA and CA199, are crucial in predicting patients’ prognosis, and both clinicians and patients should be aware of the importance of the preoperative and postoperative testing of these two tumour markers. Therefore, it is not recommended to ignore the testing of these markers for various reasons. Further validation of the nomograms in different cohorts is needed to enhance their generalizability.