Cumulative Incidence, Risk Factors, and Overall Survival of Disease Recurrence after Curative Resection of Stage II–III Colorectal Cancer: A Population-based Study

Abstract Real-world data are necessitated to counsel patients about the risk for recurrent disease after curative treatment of colorectal cancer. This study provided a population-based overview of the epidemiology of recurrent disease in patients with surgically resected stage II/III colorectal cancer. Patients diagnosed with stage II/III primary colorectal cancer between July and December 2015 were selected from the Netherlands Cancer Registry (N = 3,762). Cumulative incidence of recurrent disease was estimated, and multivariable competing risk regression was used to identify risk factors for recurrent disease in patients with primary colon and rectal cancer. Moreover, overall survival (OS) after diagnosis of recurrent colorectal cancer was estimated. Median clinical follow-up was 58 months (Q1–Q3: 22–62). Five-year cumulative incidence of recurrent disease was 21.6% [95% confidence interval (CI): 20.0–23.2] and 30.0% (95% CI: 28.3–33.5) for patients with primary colon and rectal cancer, respectively. Stage III disease and incomplete resection margin in patients with primary colon cancer and extramural vascular invasion in patients with primary rectal cancer were strongly (HR ≥ 2) associated with recurrent disease. Median OS of patients with distant, locoregional, or the synchronous combination of distant and locoregional recurrent disease was 29, 27, and 13 months, respectively (P < 0.001). Patients with distant recurrences limited to liver or lung showed a median OS of 46 and 48 months, respectively. The incidence of recurrent disease was higher in patients with rectal cancer than in patients with colon cancer, predominantly due to higher rates of distant recurrences. OS after recurrent disease was impaired, but subgroups of patients diagnosed with recurrent disease limited to one site showed statistically significantly longer OS. Significance: Population-based data on recurrent colorectal cancer are rare, but crucial for counseling patients and their physicians. This large nationwide, population-based study provides an up-to-date overview of the epidemiology of recurrent disease in patients with stage II and III primary colon and rectal cancer treated with surgical resection.


Introduction
Colorectal cancer is a disease with a major global burden (1).Almost 4 out of 5 patients with colorectal cancer are diagnosed when the disease has not yet influencing OS (14)(15)(16).Besides, disease stage of the primary tumor and molecular factors are also strongly associated with OS after diagnosis of recurrent disease (8).
Accurate and up-to-date epidemiologic information about disease recurrence is of importance to inform patients on their recurrence risk during colorectal cancer follow-up.Population-based, nationwide studies are best suited to provide such estimates, because they are less prone to selection bias and thus allow for comprehensive interpretation.However, population-based studies on recurrent colorectal cancer are rare, as registration of diagnoses of recurrent colorectal cancer requires extensive effort.
The Netherlands Cancer Registry has recently collected recurrence data of patients with primary stage II and III colorectal cancer with 5 years follow-up after resection.Using these data, the aim of this study was to assess the cumulative incidence of recurrent disease and factors associated with diagnosis of recurrent disease in patients with stage II and III primary colon or rectal cancer who underwent surgical resection, as well as to assess OS after diagnosis of recurrent colorectal cancer.

Data Collection
Patient and tumor data were obtained from the Netherlands Cancer Registry (NCR), managed by the Netherlands Comprehensive Cancer Organisation (IKNL).Trained data managers extracted data from electronic patient files in all Dutch hospitals, after notification of newly diagnosed patients through the automated national pathological archive (PALGA).All patients diagnosed with stage II and III colorectal cancer between July 1 and December 31 of 2015 were included.The additional data collection about recurrent disease was performed in 2021.No data were available about the treatment of recurrent colorectal cancer.Annual linkage of the NCR with the Dutch personal records database on February 1, 2022 provided the vital status.The Privacy Review Board of the NCR approved this study.

Study Design and Population
In this retrospective, population-based cohort study, the following patients were excluded: patients in whom the primary tumor was not treated with a surgical resection, patients with an appendiceal, overlapping or unspecified localization of the primary tumor, patients with neuroendocrine tumor morphology and patients without follow-up data (Supplementary Fig. S1).In patients diagnosed with more than one synchronous primary colorectal tumor, the characteristics of the tumor with the highest disease stage were used.

Definitions of Diagnoses and Outcomes
The 7th edition of the Unio Internationale Contra Cancrum TNM (tumornode-metastasis) classification was used as reference for the clinical and pathologic tumor stage (17).Pathologic stage was used, but in patients treated with neoadjuvant treatment and in patients in whom pathologic staging was missing, clinical stage was used.
The standard of the International Classification of Disease-Oncology (ICD-O-3) was used as reference for the tumor localization and morphology (18).
Recurrent disease was defined as any recurrence of colorectal cancer which presented after resection of the primary tumor.A locoregional recurrence (LRR) was defined as recurrent colorectal cancer near the site of the resection, or in lymph nodes defined as regional lymph nodes by the TNM classification.A distant recurrence (DR) was defined as recurrent colorectal cancer not meeting the definition of LRR, and localization was registered according to the third edition of the ICD-O-3.
Clinical follow-up time was measured from the date of resection of the primary tumor until detection of the first recurrence of colorectal cancer, death or last clinical contact with the hospital.Follow-up until vital status was measured from date of resection of the primary tumor until date of vital status (i.e., deceased or alive) registered in the Dutch personal records database.

Statistical Analyses
The primary outcomes of this study were cumulative incidence of recurrent colorectal cancer, factors associated with diagnosis of recurrent colorectal cancer and OS after recurrent colorectal cancer.First, an overview of the population characteristics was provided using descriptive statistics.Second, the cumulative incidence of recurrent disease at 1, 3, and 5 years after primary tumor resection was estimated for patients with primary colon and rectal cancer, stratified for primary disease stage.The cumulative incidence was also estimated separately for LRR only, DR only or both LRR and DR.Cumulative incidence was adjusted to account for the competing risk of death, and each function was compared using Gray test (21).
Third, univariable and multivariable associations between patient-and tumorrelated factors and recurrent disease were estimated using competing risk regression according to the subdistribution hazards approach (22).Again, death was considered a competing risk and the outcomes were presented for primary colon and rectal cancer.Age, ASA physical status, morphology, differentiation grade, the number of assessed lymph nodes, and resection margin were categorized to enhance clinical interpretation or to improve statistical power due to an otherwise small group size.All factors included in the univariable model were included in the multivariable model, except for emergency resection because of possible collinearity with bowel obstruction at presentation.Fourth, OS after diagnosis of recurrent colorectal cancer was estimated using the Kaplan-Meier method, and differences between survival curves in localization and pattern were tested using the log-rank test.

Missing Data and Software
To reduce bias due to missing data, multiple imputation with chained equations was used (23).Additional information on the handling of missing data by multiple imputation is presented in Supplementary Data S1.Sensitivity analyses without multiple imputation were performed for the risk factors and were reported in Supplementary Table S5.
All statistical analyses were performed using R version 4.1.3with the "cmprsk" and "survival" package.The "mice" package was used for the multiple imputation.P < 0.05 was regarded as statistically significant.

FIGURE 1
Cumulative incidence function for diagnosis of recurrent disease in patients with primary colon (A) and rectal cancer (B), stratified for primary disease stage and presented with the corresponding 95% confidence bands.

Data Availability
The data that support the findings of this study are available from the IKNL.Restrictions apply to the availability of these data, which were used under license for this study.Data are available from the corresponding author, T.C. Boute, with the permission of IKNL.

Patient and Tumor Characteristics
In total, 3,762 patients were included in this study (Supplementary Fig. S1).Median age of all included patients was 69 years (Q 1 -Q 3 : 63-77) and 54.8% was male.The primary tumor was located in the colon in 70.9% of patients, and in the rectum in 29.1% of patients.Patient and tumor characteristics, stratified between patients with primary colon and rectal cancer, are presented in Table 1.The majority of patients with primary colon cancer were diagnosed with stage II disease (52.5%), while stage III disease was more common in patients with primary rectal cancer (70.6%).Patients with primary colon cancer with stage II and III disease were treated with adjuvant chemotherapy in 8.1% and 63.6% of cases, respectively (Supplementary Table S1).Patients with primary rectal cancer with stage II and III disease were treated with neoadjuvant (chemo)radiation in 52.5% and 79.0% of cases, respectively.

Patient and Tumor Factors Associated with Diagnosis of Recurrent Disease in Patients with Primary Colon and Rectal Cancer
The results of the univariable and multivariable analyses for the risk of diagnosis with recurrent disease in patients with primary colon and rectal cancer are included in Supplementary Tables S3 and S4, respectively, along with 5-year cumulative incidence rates stratified for the included factors.

Discussion
This study found a 5-year cumulative incidence of recurrent disease in surgically resected stage II and III primary colon cancer of 13.0% and 30.9%, respectively, in rectal cancer, these numbers were 23.4% and 32.8%, respectively.Patients with rectal cancer were more frequently diagnosed with DR only.Stage III disease and an incomplete resection margin were strongly associated (i.e., HR with ≥2) with diagnosis of recurrent disease in patients with primary colon cancer, while in rectal cancer extramural vascular invasion showed a strong association with recurrence.OS of patients diagnosed with recurrent colorectal cancer in more than one site was poor, while subgroups of patients diagnosed with recurrent colorectal cancer in the liver or lung only showed statistically significantly longer OS than patients with recurrent colorectal cancer in other sites.
Few population-based studies have assessed the cumulative incidence of recurrent disease in primary colon and rectal cancer patients, perhaps due to the additional effort needed to collect the necessary clinical follow-up data for the entire population.Nonetheless, the 5-year cumulative incidence of DR only estimated in the current study (primary colon cancer: 15.5%; primary rectal cancer: 22.3%) was comparable with rates found in other international studies (16%-22%; refs.[24][25][26][27].On the other hand, the 5-year cumulative incidences of LRR and LRR+DR are low in comparison with the (inter)national literature (28)(29)(30)(31).This could be indicative of adequate presurgical assessment of resectability, as locoregional failure is clearly associated with incomplete resection margins and locally advanced tumors (28,32,33).Previous studies have been performed in the Netherlands using data from the first semester of 2015 (11,14,15), but were limited by the availability of 3 years of clinical follow-up data only.While these studies suggest a potential flattening of recurrence curves around the 3-year mark, the current study showed an additional 120 recurrences detected between year 3 and 5.This led to an increase in cumulative incidence from 20.5% in year 3 to 24.1% in year 5. Given that follow-up guidelines recommend a 5-year schedule, the current study presents a more comprehensive report of recurrence incidence.The most straightforward explanation for this difference lies in the classification of disease stage: Qaderi and colleagues used the pathologic TNM classification for all patients (11), regardless of neoadjuvant treatment.The current study reported the clinical instead of the pathologic TNM classification for patients treated with neoadjuvant treatment.Herewith, the results of the current study reflect the risk for recurrent disease in regard to the extent of the disease at presentation, and not after neoadjuvant treatment.
The cumulative incidence of recurrent disease increased rapidly in the first 2 years after resection of the primary colon or rectal cancer, but flattened between the third and fifth year after surgical resection.This finding validates the duration of the advised oncological follow-up for surgically resected colorectal cancer in Western countries (34)(35)(36), which is limited to 5 years after surgical resection.Other studies have demonstrated that the incidence of recurrent disease more than 5 years after surgical resection of primary colon or rectal cancer is exceptionally low (37)(38)(39)(40)(41).
Prognostic factors such as extramural vascular invasion, lymphatic invasion, but also tumor deposits and tumor budding are a focus of ongoing research because of their assumed ability to possibly enhance the TNM classification in regard to risk classification (42)(43)(44).Stage III disease and an incomplete resection margin showed the strongest associations with diagnosis of recurrent disease, but intramural vascular invasion, extramural vascular invasion and lymphatic invasion showed associations of substantial strength as well.Nonrandomized studies have been able to show an OS benefit for adjuvant chemotherapy in patients with stage II primary colon cancer with positive extramural vascular invasion (45) or lymphovascular invasion (46).However, pT4 tumors, <10 investigated lymph nodes, tumor perforation, obstruction, vascular invasion and poor-no tumor differentiation were all classifications for a high-risk stage II tumor, and consequentially and indication for adjuvant chemotherapy in the treatment guideline in the Netherlands in 2014.This guideline was revised in 2019, and based on various population-based studies, the classification of a high-risk stage II tumor was narrowed to pT4 only.Undoubtedly, adjuvant chemotherapy for patients with stage II colon cancer will remain controversial, as is illustrated by the practice variation in the Netherlands and the absent OS benefit on a population base (47).In patients with primary rectal cancer, extramural vascular invasion showed the strongest association with diagnosis of recurrent disease of all included factors, vastly stronger than lymphatic invasion.The presence of MRI-based extramural vascular invasion is currently implemented in the clinical management, and is an absolute indication for neoadjuvant chemoradiation or short-course radiotherapy (35,(48)(49)(50).
Notably, Qaderi and colleagues found a higher recurrent disease risk in patients with stage I-II colon cancer ages between 65 and 74 compared with those ages <65 years old (11).No such association was found in the current study in patients with stage II-III colon cancer.The most likely explanation for the difference in results between the two studies probably lies in the difference in the statistical models.The relationship between age and disease recurrence was adjusted for more confounding factors in the present model [21 vs. 14  This study has several limitations.First, no recurrence data were available for the group of patients with stage I primary colorectal cancer who are generally also treated with surgical resection (i.e., pT2N0M0).Nonetheless, it is known that recurrent colorectal cancer in this group is rare (24).Second, no information was available about the number of recurrences present in each organ/site, about the occurrence of subsequent (re-)recurrences, nor about the treatment of recurrences.However, these treatment data are rarely reported in population-based studies, and generally only collected in clinical trials.These data would have been of value in the stratification of OS after diagnosis of recurrent colorectal cancer, especially in the comparison of different treatment strategies for the management of recurrent disease.Therefore, it would be preferable if future rounds of data collections about recurrent colorectal cancer for cancer registries would include data about treatment of recurrences, while it must be acknowledged that resources for additional data collection are limited.Also, drawing conclusions from such information would be challenging due to underlying mechanisms such as patient preferences and medical history, which could significantly influence treatment outcomes.Third, the nationwide colorectal cancer screening program in the Netherlands was started in 2014, although not yet fully operational in the second semester of 2015.This program consists of biennial faecal immunochemical testing for the population between 55 and 75 years.Only a small cohort of patients (∼10%) received screening in the second semester of 2015 (55), thus the effect is expected to be small.In several years' time, a reassessment of the cumulative incidence, associated factors and OS in the population in which colorectal cancer screening is fully operational is needed.

Conclusion
This retrospective, nationwide, population-based study provides an up-to-date overview of the epidemiology of recurrent disease in patients with stage II and III primary colon and rectal cancer treated with surgical resection.The 5-year cumulative incidence of recurrent disease was 13.0% and 30.9% for patients with stage II and III primary colon cancer, respectively, and 23.4% and 32.8% for patients with stage II and III primary rectal cancer, respectively.The majority of recurrent disease was localized distantly, especially in patients with primary rectal cancer.OS differed statistically significantly between the different patterns of recurrence, with patients with isolated metastasis in the liver or lung showing longer OS, while patients with recurrent colorectal cancer synchronously diagnosed in two or more sites had severely impaired OS outcomes.

FIGURE 2 FIGURE 3
FIGURE 2Cumulative incidence functions of distant recurrent disease (i.e., DR only), locoregional recurrent disease (i.e., LRR only) and the synchronous combination of LRR+DR in patients with primary colon (A) and rectal cancer (B), presented with the corresponding 95% confidence bands.

FIGURE 4 FIGURE 5
FIGURE 4 OS plot of patients diagnosed with recurrent colorectal cancer with corresponding 95% confidence bands, stratified for locoregionally recurrent colorectal cancer only (LRR), distantly recurrent colorectal cancer (DR) at one site, DR at two sites, DR at two or more sites, and the synchronous diagnosis of DR+LRR.Vertically dashed lines represent median OS.

The 3 -
year cumulative incidence of recurrent disease after surgical resection of stage II and III colon cancer in the current study (stage II: 10.3%; stage III: 26.7%) was very comparable with that of patients diagnosed in the first semester of 2015 (stage II: 12%-16%; stage III: 24%-31%; ref. 11).However, the incidence rates of recurrence for surgically treated stage II and III rectal cancer seemed lower for patients diagnosed in the second semester of 2015 (stage II: 19.0%; stage III: 29.3%) than in the first semester of 2015 (stage II: 24%; stage III: 38%).

TABLE 1
Patient and tumor characteristics of patients with primary colon and rectal cancer (Continued on the following page )
(53)15) of freedom in current study versus Qaderi and colleagues(11)].The OS of patients diagnosed with recurrent colorectal cancer was clearly related to the pattern of the recurrence: patients diagnosed with LRR only or DR only in one site showed superior OS outcomes in comparison with patients diagnosed with LRR+DR or DR in two or more sites.In clinical practice, patients with synchronous or metachronous metastases limited to one site are treated with curative intent if local treatment of all metastases falls within the clinical possibilities(52).This treatment effect is likely the cause of the relatively long median OS for liver only (48.1 months) and lung only (45.6 months) recurrences estimated in this study.In limited cases, curative-intent treatment is given to patients with synchronously diagnosed LRR and DR(14,15), or distant metastases in two or more sites(53).However, the majority of patients diagnosed with recurrent colorectal cancer in two or more locations are treated with palliative intent or best supportive care, (54) leading to median OS estimates in the current study of less than 18 months.