Is Colonoscopy Alone Adequate for Surveillance in Stage I Colorectal Cancer?
Article information
Abstract
Purpose
While colonoscopy is the standard surveillance tool for stage I colorectal cancer according to National Comprehensive Cancer Network guidelines, its effectiveness in detecting recurrence is debated. This study evaluates recurrence risk factors and patterns in stage I colorectal cancer to inform comprehensive surveillance strategies.
Materials and Methods
A retrospective analysis of 2,248 stage I colorectal cancer patients who underwent radical surgery at Samsung Medical Center (2007-2018) was conducted. Exclusions were based on familial history, prior recurrences, preoperative treatments, and inadequate data. Surveillance included colonoscopy, laboratory tests, and computed tomography (CT) scans.
Results
Stage I colorectal cancer patients showed favorable 5-year disease-free survival (98.3% colon, 94.6% rectum). Among a total of 1,467 colon cancer patients, 26 (1.76%) experienced recurrence. Of the 781 rectal cancer patients, 47 (6.02%) experienced recurrence. Elevated preoperative carcinoembryonic antigen levels and perineural invasion were significant recurrence risk factors in colon cancer, while tumor budding was significant in rectal cancer. Distant metastasis was the main recurrence pattern in colon cancer (92.3%), while rectal cancer showed predominantly local recurrence (50%). Colonoscopy alone detected recurrences in a small fraction of cases (3.7% in colon, 14.9% in rectum).
Conclusion
Although recurrence in stage I colorectal cancer is rare, relying solely on colonoscopy for surveillance may miss distant metastases or locoregional recurrence outside the colorectum. For high-risk patients, we recommend considering regular CT scans alongside colonoscopy. This targeted approach may enable earlier recurrence detection and improve outcomes in this subset while avoiding unnecessary scans for the low-risk majority.
Introduction
Colorectal cancer is one of the most common cancers worldwide. The introduction of the screening system for colorectal cancer has downstaged pathologic staging [1,2]. Stage I colorectal cancer shows favorable oncologic outcome after radical surgery, with disease-free survivals of 92%-100% in colon cancer and 89%-96% in rectal cancer [3-7]. Due to this favorable oncologic outcome, surveillance after radical surgery of stage I colorectal cancer is simpler than that for stage II-IV colorectal cancer. The National Comprehensive Cancer Network guideline and American Society of Colon and Rectal Surgeons guideline recommended only colonoscopy for surveillance after radical surgery of stage I colon cancer [8]. According to the National Comprehensive Cancer Network guidelines, it is recommended to undergo a colonoscopy one year following surgery. Should an advanced adenoma be identified during this examination, another colonoscopy is advised 1 year later. Conversely, if no advanced adenoma is found, subsequent colonoscopies are recommended 3 and 5 years post-surgery for stage I colorectal cancer [9]. In the European Society of Medical Oncology clinical practice guideline, surveillance by TNM stage is not distinct, and the routine surveillance and follow-up for colorectal cancer comprise colonoscopy within the first year and regular serum carcinoembryonic antigen (CEA) tests every 6 months in the first 3 years, and a minimum of two computed tomography (CT) scans of the chest, abdomen, and pelvis in the first 3 years for distant metastases [10]. With this background, the current study was performed to evaluate the recurrence rate and recurrence patterns of stage I colorectal cancer to help determine whether colonoscopy alone is adequate for surveillance in stage I colorectal cancer.
Materials and Methods
1. Study design and patients
This retrospective study included all patients with colorectal cancer who underwent radical surgery and were diagnosed with stage I at Samsung Medical Center between January 2007 and December 2018. The exclusion criteria were as follows: patients with familial cancer; patients being treated for recurrent disease; patients who underwent preoperative treatment; patients who underwent local excision; patients lacking pathologic data; and patients without sufficient follow-up. The preoperative diagnosis was performed via colonoscopy as well as chest and abdomen-pelvic CT.
All patients underwent radical resection of R0 resection for primary tumor with regional lymph node (LN) dissection. The TNM stage was determined by American Joint Committee on Cancer 8th edition [11]. Histopathologic features were examined by two gastrointestinal pathologists. Positive tumor budding was defined as the presence of single tumor cells or small clusters of up to five cells in the tumor stroma [12].
Patients underwent chest and abdomen-pelvis CT every 6 months for 5 years after the operation. Colonoscopy was done one year after surgery and then 3 and 5 years after. Routine lab analyses including CEA and carbohydrate antigen 19-9 were performed every 3 months for the first year after surgery and then were subsequently performed every 6 months for 5 years after surgery.
2. Statistical analysis
Statistical analyses were performed using R Statistical Software ver. 3.6.3 (Foundation for Statistical). Chi-square test or Fisher’s exact test were used to analyze categorical variables whereas two-sample t-test was used to analyze continuous variables that did not violate the normality assumption and Wilcoxon rank sum test was used to analyze continuous variables that did violate the normality assumption. The Kaplan-Meier method was used to analyze survival curves, and differences between the curves were evaluated using the long-rank test. Cox proportional hazard models were used to identify factors associated with an increased risk of recurrence and death. Firth’s penalized Cox regression models were used for some variables due to the sparsity of some levels. Variables with a p-value < 0.1 in the univariable model were included in multivariable analysis. A p-value ≤ 0.05 in multivariable analysis was considered to be statistically significant.
3. Nomogram model building and validation
Multivariable logistic regression analysis was conducted to examine the relationships between various factors and recurrence of stage I colon and rectal cancer. The final model was chosen based on variables that showed p < 0.10 on a backward step-down selection process.
Nomogram validation consisted of discrimination and calibration. Discrimination refers to a nomogram model’s ability to correctly distinguish between two classes of outcomes. First, for internal validation, we used a resampling method with 100 bootstrap samples of the same size as the original data to estimate the bias-corrected or over-fitting corrected predictive accuracy of the model, which is expressed as the concordance index (c-index). Second, we assessed calibration, which compares the predicted probability with the observed outcome in 10 groups partitioned according to the deciles of the predicted probabilities.
Results
1. Clinicopathologic characteristics and survival outcomes of stage I colorectal cancer patients
In total, this study analyzed data from 2,248 patients with stage I colorectal cancer who were treated with radical surgery, and they had a median follow-up of 66 months (range, 6 to 193 months). This analysis included 1,467 cases of colon cancer and 781 cases of rectal cancer. Table 1 describes the patients’ characteristics. The median age was 59 years old (range, 27 to 90 years) and 58.7% were male. In terms of pathologic T category, T1 comprised 53.4% of cases, T2 comprised 46.6%, and 12.1% had harvested LN less than 12. Five-year disease-free survival (DFS) was 98.3% in colon cancer patients and 94.6% in rectal cancer patients (Fig. 1). Five-year overall survival (OS) was 99.7% in colon cancer patients and 99.2% in rectal cancer patients (Fig. 1).
Baseline clinicopathologic features of stage I colorectal cancer patients
Disease-free survival (A) and overall survival (B) of stage I colon and rectal cancer.
In colon cancer patients with recurrence, preoperative CEA level was significantly higher, and there was also higher positive perineural invasion and tumor budding (p=0.002, p=0.025, and p=0.039) (Table 2). In rectal cancer, patients with recurrence had higher pathologic T category, lower numbers of harvested LN, higher positive lymphovascular invasion, and tumor budding (p=0.040, p=0.030, p < 0.001, and p < 0.001) (Table 3).
Comparison of baseline clinicopathologic features of stage I colon cancer patients between recurrent patients and non-recurrent patients
Comparison of baseline clinicopathologic features of stage I rectal cancer patients between recurrent patients and non-recurrent patients
2. Recurrence patterns and treatment for recurrence in stage I colorectal cancer patients
The recurrence patterns and treatment modalities are outlined in Table 4. The median times to recurrence were 25 months (range, 5 to 72 months) for colon cancer and 22 months (range, 6 to 50 months) for rectal cancer. Among the 26 patients diagnosed with recurrence in colon cancer, distant metastasis (92.3%) was the predominant recurrence type. By contrast, local recurrence (50%) was the most common type in rectal cancer.
Recurrence pattern and treatment type of stage I colorectal cancer
Among cases of colon cancer, one patient experienced recurrence at the anastomosis site, while another patient had recurrence at regional LN metastasis. Distant metastasis was diagnosed in 24 patients, with the liver being the most prevalent site, followed by the lung. Among the 47 patients with recurrence in rectal cancer, 24 exhibited local recurrence, 18 had distant metastasis, and five presented with both local recurrence and distant metastasis. Pelvic wall recurrence was the most frequent, followed by recurrence at the anastomosis site. In cases of distant metastasis, the lung was the primary site, followed by the liver.
Among colon cancer patients with recurrence, 12 (46.2%) underwent surgery and 11 received perioperative chemotherapy. Ten patients (38.5%) underwent chemotherapy or radiotherapy for recurrent cases. Local treatments were administered to two patients (7.7%): one underwent endoscopic mucosal resection for anastomosis site recurrence, while the other underwent radiofrequency ablation. For rectal cancer patients with recurrence, 30 (63.8%) underwent surgery and 26 received perioperative chemotherapy or radiotherapy. Fourteen patients (29.8%) underwent chemotherapy and/or radiotherapy for recurrence.
3. Risk factors for recurrence in stage I colorectal cancer patients
In multivariable analysis for colon cancer, preoperative CEA and perineural invasion were found to be risk factors for recurrence (p=0.049 and p=0.031) (Table 5). Moreover, age and tumor budding were prognostic factors for OS (p=0.003 and p=0.036). In rectal cancer, only tumor budding was found to be a risk factor for recurrence in multivariable analysis (p=0.011) (Table 6). Age, body mass index, and tumor budding were also prognostic factors for OS in multivariable analysis (p < 0.001, p=0.002, and p=0.040).
Univariable and multivariable analyses of prognostic factors for recurrence in stage I colon cancer
Univariable and multivariable analyses of prognostic factors for recurrence in stage I rectal cancer
4. Nomogram for recurrence risk for stage I colon cancer
The results of multivariable logistic regression analysis of risk factors potentially related to recurrence were included in the nomogram (Fig. 2A). The nomogram underwent internal validation, and the estimated 5-year survival probabilities in internal validation were found to be 98.7% in the low-risk group, 98.3% in the intermediate-risk group, and 94.6% in the high-risk group; these values were similar to the 5-year survival probabilities of the cohort (99.7% in the low-risk group, 98.0% in the intermediate-risk group, and 93.6% in the high-risk group) (Fig. 2B and C). C-index of 0.684, an Integrated Calibration Index (ICI) of 0.049, an E50 of 0.033, and an E90 of 0.125, which indicated good agreement between the predicted and observed probabilities of recurrence.
(A) Nomogram for recurrence risk in stage I colon cancer. Recurrence-free survival curve according to risk group (B) and 5-year recurrence-free survival probability (C) according to risk group in observed group and estimated group. CEA, carcinoembryonic antigen; PNI, perineural invasion.
5. Nomogram for recurrence risk for stage I rectal cancer
The results of multivariable logistic regression analysis of the risk factors potentially related to recurrence were included in the nomogram (Fig. 3A). The nomogram underwent internal validation, and the estimated 5-year survival probabilities in internal validation were found to be 96.7% in the low-risk group, 91.4% in the intermediate-risk group, and 82.6% in the high-risk group, which were similar to the 5-year survival probabilities of the cohort (97.3% in the low-risk group, 90.2% in the intermediate-risk group, and 77.5% in the high-risk group) (Fig. 3B and C). The C-index of 0.684, an ICI of 0.050, an E50 of 0.036, and an E90 of 0.100, all indicated good agreement between the predicted and observed probabilities of recurrence.
(A) Nomogram for recurrence risk in stage I rectal cancer. Recurrence-free survival curve according to risk group (B) and 5-year recurrence-free survival probability (C) according to risk group in observed group and estimated group. LN, lymph node; LVI, lymphovascular invasion.
Discussion
This study aimed to assess the risk factors for recurrence in stage I colon and rectal cancers and develop a nomogram for predicting recurrence risk in these patients. Stage I colon cancer exhibited a recurrence rate of 1.8%, while rectal cancer exhibited a recurrence rate of 6%. Distant metastasis was the predominant type of recurrence in stage I colon cancer, while local recurrence was more common in stage I rectal cancer.
Elevated preoperative CEA levels have traditionally been considered to be a risk factor for colorectal cancer recurrence, particularly when CEA exceeds the normal range (≥ 5 ng/mL) [13,14]. CEA, which is a complex glycoprotein associated with cell adhesion, is secreted by various solid tumors, including 90% of colon tumors [15,16]. The results of prior studies have suggested that preoperative CEA levels are not predictive in stage I compared to stages II or III, which is possibly due to CEA concentration correlating with tumor size and advanced stage [17-19]. This might be due to the fact that the concentration of CEA is correlated with the tumor size and advanced stage [20,21]. Therefore, various efforts have been made to determine the optimal cut-point values for CEA as a prognostic factor for recurrence. Baqar et al. [13] demonstrated the role of preoperative CEA in colorectal cancer, showing that high-normal CEA (≥ 2.5 ng/mL) was associated with poorer survival than low-normal CEA (< 2.5 ng/mL). Margalit et al. [22] identified an optimal preoperative CEA cut-point of 2.35 ng/mL for predicting survival in stage I and II colon cancer. In our study, while the results of continuous analysis revealed that CEA levels were prognostic factors for recurrence in stage I colon cancer, the results of categorical analysis (divided into < 5 ng/mL and ≥ 5 ng/mL) did not demonstrate significance. On the other hand, a CEA level higher than 1.3 ng/mL was identified as a negative prognostic factor for recurrence in the multivariable analysis. Moreover, the risk of recurrence increased by a factor of 1.091 with each unit increase in the logarithm of CEA levels. Consequently, using a threshold of 5 ng/mL to categorize patients in stage I colorectal cancer may overlook individuals at risk of relapse. This suggests the need for a lower baseline CEA level of 1.3 ng/mL for more accurate risk stratification in stage I colorectal cancer. Perineural invasion emerged as a significant prognostic factor for recurrence in stage I colon cancer, aligning with previous findings [3,23,24]. Similarly, tumor budding proved to be a prognostic factor for survival in stage I rectal cancer, which was also consistent with earlier studies [3,25].
Recurrences of colorectal cancer are most commonly observed within the first 3 years following radical resection. As a result, an intensive follow-up protocol is typically implemented for stage II and III patients during this period [26,27]. However, in contrast to the corresponding rates of stage II and III, the recurrence rate in stage I colorectal cancer gradually and steadily increases over 5 years. Cumulative recurrence rates were recorded at 48.6%, 66.2%, and 91.9% at 2, 3, and 5 years post-surgery, thus aligning with the findings of previous studies [7,26]. Accordingly, it is advisable to conduct surveillance for at least 5 years post-surgery to facilitate early detection and treatment of recurrence in stage I colorectal cancer.
In our study, the innovative use of both univariable and multivariable analyses to construct a comprehensive nomogram has significantly enhanced our ability to accurately assess recurrence risk in stage I colon and rectal cancer patients. By effectively categorizing patients into low, intermediate, and high-risk groups, the nomogram has proven its efficacy in mirroring actual recurrence rates, thus showcasing its potential as a predictive tool. In recent years, nomograms have emerged as pivotal tools in forecasting colorectal cancer recurrence that can provide valuable guidance in patient management [28-30]. Despite traditional recommendations against routine laboratory and CT surveillance in stage I colorectal cancer due to its historically low recurrence risk, our findings have unveiled a more nuanced reality [8,9]. To elaborate, we discovered that distant metastasis is the most frequent recurrence pattern in stage I colon cancer. In stage I rectal cancer, local recurrences were often situated at the pelvic wall, which is a location that is not adequately monitored by colonoscopy alone. Remarkably, a substantial majority of patients with recurrence, 92.3% in colon and 93.6% in rectal cancer, harbored identifiable risk factors. However, standard colonoscopy detected recurrence in only a small fraction of these cases, with the majority of recurrences being identified through abdomino-pelvic or chest CT scans. The study’s implications are further underscored by the observed 5-year DFS rates of 98.3% for colon cancer and 94.6% for rectal cancer, alongside respective 5-year OS rates of 99.7% and 99.2%. These figures not only align with but also surpass those in previous studies, thus suggesting the benefits of early detection and treatment facilitated by rigorous surveillance protocols. However, it is crucial to balance the benefits of such surveillance with economic considerations, particularly given the low recurrence rate in stage I compared to stages II and III colorectal cancer. Our nomogram serves as a strategic tool in this context, enabling the identification of patients who would most benefit from regular postoperative surveillance. This targeted approach promises to optimize oncological outcomes while preventing unnecessary healthcare expenditures, thus marking a significant step forward in personalized cancer care.
Nevertheless, this study has limitations that must be noted. It was conducted at a single center and relied on retrospective data, which potentially introduced heterogeneity in diagnosis and treatment due to the extended study duration. Moreover, being a high-volume quaternary center, some patients underwent surveillance at peripheral hospitals, potentially leading to loss of patients who developed recurrence after radical surgery. Finally, the lack of access to specific biomarkers, such as microsatellite instability status, Ras mutations, or consensus molecular subtypes, limits the precision of our findings. Despite these limitations, this study is the first to evaluate risk factors for stage I colon and rectal cancer, and it offers a nomogram with which to predict recurrence in these patients.
Stage I colorectal cancer generally has an excellent prognosis, with low recurrence rates. While colonoscopy remains crucial for surveillance, it may miss distant metastases, the predominant recurrence pattern. For high-risk patients—identified by specific factors such as elevated CEA, positive perineural invasion or tumor budding in colon cancer, and T2 category, fewer retrieved LN, positive lymphovascular invasion, or TB in rectal cancer—we recommend considering regular abdomino-pelvic and chest CT scans alongside colonoscopy. This targeted approach may facilitate earlier detection and improved outcomes in this small but significant subset. For stage I patients without these risk factors, standard colonoscopy-based surveillance appears sufficient, given the very low recurrence risk. This stratified approach aims to balance intensive monitoring for high-risk patients with avoiding unnecessary scans for the majority of stage I patients.
Notes
Ethical Statement
This study was approved by the Institutional Review Board of Samsung Medical Center (IRB No. 2023-08-133). The requirement for informed consent was waived by the Institutional Review Board of Samsung Medical Center due to the use of de-identified data from medical records, ensuring that the privacy and confidentiality of all participants were maintained.
Author Contributions
Conceived and designed the analysis: Kim S, Lee WY, Cho YB.
Collected the data: Kim S, Shin JK, Park Y.
Contributed data or analysis tools: Kim S, Huh JW, Yun SH, Cho YB.
Performed the analysis: Kim S, Huh JW, Kim HC, Cho YB.
Wrote the paper: Kim S, Park Y.
Review and edit: Kim S, Shin JK, Park Y, Huh JW, Kim HC, Yun SH, Lee WY, Cho YB.
Conflict of Interest
Conflict of interest relevant to this article was not reported.