Incidence of Recurrence and Time to Recurrence in Stage I to III Colorectal Cancer

This cohort study investigates stage-specific rates and timing of colorectal cancer recurrence using data from Danish national health registries from 2004 to 2019.


C
olorectal cancer (CRC) represents a considerable health burden worldwide. 1The standard of care for patients with nonmetastatic CRC (Union for International Cancer Control [UICC] TNM stages I-III) is surgery, with clinical staging informing the use of neoadjuvant therapy and histopathological staging informing the use of adjuvant chemotherapy. 2,3Although surgery is intended to be curative treatment, approximately 20% of patients experience recurrence of disease.Therefore, all patients with stages I to III CRC are offered postoperative surveillance.
Over the past 2 decades, several initiatives aimed at reducing the risk of recurrence and improving survival of patients with CRC have been implemented in most countries, including Denmark 4 (eFigure 1 in Supplement 1).Initiatives include increased focus on improving primary treatment with regularly updated national recommendations, 5 a multidisciplinary team approach, 6,7 enhanced recovery after surgery protocols, 8 and implementation of new surgical procedures; the latter include central vascular ligation and dissection along the embryological fascia as the underlying principle of complete mesocolic excision 9 and total mesorectal excision, 10 with increased lymph node yield.5][16][17] Lastly, population-based screening programs have been implemented in several countries, 18 identifying patients with asymptomatic CRC. 19While most of the initiatives have been associated with better outcomes individually, nationwide population-based studies are needed to investigate the implications of these initiatives for the risk of recurrence.
The aim of this study was to determine colon and rectal cancer stage-specific rates of recurrence and, for those with recurrence, the time from surgery to recurrence (TSTR) in a nationwide cohort of patients who underwent surgery with curative intent for nonmetastatic CRC from 2004 to 2019 by comparing 3 calendar periods of surgery with curative intent: 2004 to 2008, 2009 to 2013, and 2014 to 2019.A secondary aim was to describe the risk of recurrence in colon and rectal cancers detected through screening.

Design and Study Population
This nationwide registry-based cohort study included all patients who underwent surgery with curative intent for UICC TNM stages I to III CRC between January 1, 2004, and December 31, 2019.The study was approved by the Danish Colorectal Cancer Group (DCCG) and the Danish Data Protection Agency (Central Denmark Region) and adhered to the European Union General Data Protection Regulations. 20The study was based on anonymized registry data and therefore was exempt from review and informed consent in accordance with the General Data Protection Regulations.The study was reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. 21tients were identified through the DCCG Database, which is a prospective national clinical quality database with information on all patients treated for first-time CRC in Denmark. 22 timeline and description of the management of nonmetastatic CRC in Denmark from 2004 to 2019 are provided in eFigure 1 and the eMethods in Supplement 1.
Patients with a diagnosis of cancer other than CRC or nonmelanoma skin cancer (NMSC) prior to curative CRC surgery were excluded.Patients were also excluded if they emigrated, died, or were diagnosed with a new primary cancer other than CRC or NMSC or metastasis of unspecified origin within 180 days of the index CRC surgery.
Patients were followed up from 180 days after surgery until the following event (whichever came first): recurrence (event), second primary cancer (competing event), death (competing event), emigration (censoring event), 5 years postoperatively (censoring event), or January 1, 2023 (censoring event), when prospectively collected data from the national health care registries were retrieved (eTable 1 in Supplement 1).Hence, patients undergoing curative surgery from 2004 to 2017 had 5 years of follow-up, those undergoing curative surgery in 2018 had 4 years of follow-up, and those undergoing curative surgery in 2019 had 3 years of follow-up.

Data Sources and Identification of Recurrence
Individual-level data were obtained from Danish health care registries, 23 including the Danish Cancer Register (DCR), 24 the Danish National Patient Registry (DNPR), 25 and the Danish Pathology Registry (DPR). 26Data records were linked using the unique 10-digit civil registration number issued to each Danish resident by the Danish Civil Registration System. 27he DCCG Database provided information on the date of CRC diagnosis, date of surgery, and patient characteristics, including participation in the National Danish Colorectal Cancer Screening Program, which started in 2014 (linked from the Danish Colorectal Cancer Screening Database 28 ).The DNPR contains administrative and clinical data and provides information on chemotherapeutic treatment and diagnosis of metastases.The DCR contains data on the incidence of malignant neoplasms.The DPR provides electronically recorded data on all biological specimens according to national guidelines for uniform registration using the Danish version of the Systematized Nomenclature of Medicine (SNOMED) codes. 29ecurrence was detected using a previously described algorithm, 30 with a few optimizations 31

Covariates
The dates of curative surgery were grouped into Eligible patients were categorized by sex (male or female), age group at curative surgery (<55, 55-64, 65-74, 75-84, or ≥85 years), pathological UICC stage (I, II, or III), Charlson Comorbidity Index score (0, 1-2, or ≥3; higher scores indicated a higher number of comorbidities), 32 site of CRC tumor (colon or rectum), and surgical priority (elective or emergency).Disease stages were categorized according to the fifth edition 33 of the UICC TNM classification from 2004 to 2016 and the eighth edition 34 from 2017 to 2019.Patients in the 2014 to 2019 period were categorized according to whether the cancer was detected by screening (screening detected) or not (nonscreening detected).

Statistical Analysis
Patient demographics and treatment characteristics are presented with continuous data as medians (IQRs) and with categorical variables as the total number and percentage.Multivariable analysis was performed to investigate the association between calendar period of curative surgery and the risk of CRC recurrence, adjusted for patient and treatment characteristics.Risk of recurrence was calculated as 1-, 3-, and 5-year cumulative incidence function (CIF) estimates, with second primary cancer and death treated as competing events.Cumulative incidence function curves were constructed using the Aalen-Johansen estimator for visualization, grouped by the calendar period of curative surgery, and stratified by tumor site and UICC TNM stage.The association between calendar period of curative surgery and the risk of recurrence was estimated using the Fine-Gray competing risks method, with second primary cancer and death treated as competing events, and was reported as subdistribution hazard ratios (sHRs) with 95% CIs. 35Tumor site-and stage-specific sHRs were adjusted for age, sex, and Charlson Comorbidity Index score.
For patients with CRC recurrence, the TSTR was calculated as the time between the date of curative CRC surgery and the date on which at least 1 of the 4 indicators in the algorithm defined recurrence.The difference in TSTR between UICC stages was estimated as a time ratio (with 95% CIs) using an accelerated failure time model including tumor site, age, sex, and Charlson Comorbidity Index score and rightcensoring data for patients at the time of a second primary cancer, death, emigration, 5 years postoperatively, or January 1, 2023, whichever came first.
A subgroup analysis was performed to assess the association between patients with screening-detected CRC and those with nonscreening-detected CRC.The risk of recurrence was calculated and 1-, 3-, and 5-year CIF estimates of recurrence and sHRs adjusted for age, sex, and Charlson Comorbidity Index score were reported.

Results
We identified 45 274 patients who underwent curative surgery for pathological UICC stages I to III CRC between 2004 and 2019 (eFigure 3 in Supplement 1).After applying exclusion criteria (approximately 15% of patients excluded; eResults in Supplement 1), the final cohort consisted of 34 166 patients (median Demographic and clinical characteristics of patients with colon or rectal cancer stratified by calendar period of surgery are summarized in Table 1.From 2004 to 2019, the distribution of UICC stages shifted toward a higher proportion of stage I CRC and a lower proportion of stage II CRC (eFigure 4 in Supplement 1).The shift was most prominent from 2014 to 2019, consistent with implementation of a population-wide CRC screening program in 2014.

Time From Surgery to Recurrence
The proportion of diagnosed recurrences was highest in the first 3 years after surgery in all 3 calendar periods (eFigure 6

Recurrences Associated With Screening Status
In the 2014 to 2019 period, 43.6% of patients with stage I CRC, 20.5% of those with stage II CRC, and 22.5% of those with stage III CRC were diagnosed through screening (eTable 4 in Supplement 1).Patients with screening-detected CRC more often had lower-stage disease and were more often male, were younger, and had fewer comorbidities at curative surgery compared with those with nonscreening-detected CRC.

Discussion
From 2004 to 2019, many initiatives were implemented at a national level in Denmark to improve the treatment of patients with CRC (eFigure 1 in Supplement 1).7][38][39] While the combined implications of these initiatives have been associated with improved survival, 4,40 their implications for risk of recurrence remain unexplored.The results of the present study show that the CRC recurrence rate has decreased substantially and continuously from 2004 to 2019, even after stratifying results by tumor site and UICC stage.
In 2014, population-based CRC screening via fecal immunochemical testing was implemented in Denmark and identified a group of younger patients with asymptomatic CRC.Consistent with reports from other countries, 19 we saw that the screening prevalence round from 2014 to 2017 was associated with an increase in the number of patients with diagnosed CRC, with a shift toward lower UICC stages at diagnosis. 41Our comparison of screening-detected and nonscreening-detected CRC revealed lower rates of not only overall recurrence but also stagespecific recurrence in cancers detected by screening.Speculating whether screening was the main factor in the low stagespecific recurrence rates observed from 2014 to 2019, we explored the recurrence rate in patients whose cancer was not detected through screening.Their recurrence rate in the 2014 to 2019 period was still lower than recurrence rates observed in the 2 earlier periods, suggesting that other initiatives implemented in this period also had implications for the risk of recurrence (eFigure 1 in Supplement 1).
Additionally, detection through screening was associated with lower T category within each UICC stage.This change toward lower T category contributed to the lower stagespecific recurrence rates observed for the patients with screening-detected CRC, as advanced T category is a known factor associated with CRC recurrence. 42The T category-specific recurrence rates were also lower in patients with screeningdetected vs nonscreening detected CRC.The persistent lower recurrence rate in those with screening-detected CRC could be due to a further downstaging of tumor burden beyond T category, such as tumor size, or to distinct differences in the biological characteristics, such as growth rate.Further studies are needed to address this finding.
The main goal of follow-up protocols is to detect recurrence early to maximize the patient's survival.The European Society for Medical Oncology recommends a combination of clinical assessments, regular measurements of carcinoembryonic antigen levels, computed tomography (CT), and a colonoscopy with higher frequencies for the first 3 years after curative surgery and for a total duration of 5 years in patients with nonmetastatic CRC. 14,43However, substantial variation has been found between national surveillance guidelines, 44 and a Cochrane Review 45 found no benefit of intensified surveillance on overall survival.Denmark participated in the international multicenter randomized clinical trial COLOFOL, 46 which did not show improved 5-year overall survival or CRC-specific survival associated with more frequent CT scans and carcinoembryonic antigen measurements (5 times vs 2 times), which is why Denmark follows a surveillance program with 2 CT scans.We found that the 5-year CIF of recurrence is now below 7% and 10% for TNM stage I colon and rectal cancers, respectively (Table 2), and is even lower for cancers detected by screening.Furthermore, TSTR was longer for patients with stage I CRC than for patients with stage II and particularly stage III CRC.Due to this difference in the pattern of recurrence for patients with stage I disease, less intensive surveillance, or even no surveillance, may be noninferior to current guidelinesespecially when also considering health-related quality of life, late sequelae, and cost-effectiveness. 47Interventional, preferably randomized studies are needed to explore shifting from one-size-fits-all surveillance toward a personalized surveillance strategy. 48

Strengths and Limitations
Strengths of the current study include the nationwide approach, reflecting standardized CRC management rather than the performance of individual centers.Also, the large sample size increased the statistical power and confidence of the reported recurrence rates.
Current knowledge about CRC recurrence is based primarily on intensive follow-up of selected patient cohorts, such as those enrolled in clinical trials or cohorts from individual centers. 49Herein we used registry data to obtain recurrence status, which is an efficient and inexpensive approach to obtain follow-up data compared with medical record review, which is time and resource intensive.Furthermore, we recently validated the algorithm we used and found that it was highly accurate (sensitivity, 94%; specificity, 99%; positive predictive value, 94%; and negative predictive value, 99%) when using contemporary registry data, 31 which is in line with previous validations on historical cohorts. 30Also, the registry data needed for the algorithm is available for the entire cohort, as all Danish citizens have unrestricted access to a public tax-supported health care system. 50t can be considered a limitation that we excluded approximately 15% of patients, particularly patients with previous cancers, as the algorithm cannot discern whether recurrence occurs due to one or the other cancer.Although this exclusion may have introduced selection bias and affected the generalizability of the study, we decided to exclude these patients to improve the internal validity and accuracy of recurrence rates.The algorithm does not diagnose recurrence within 180 days after surgery.This quarantine was used to allow for completion of primary cancer treatment with up to 6 months of adjuvant chemotherapy and to avoid diagnosing synchronous metastases as a recurrence. 51Also, patients who died within 180 days after  surgery were excluded to ensure all patients experienced time at risk of recurrence after 180 days.

Conclusions
This register-based cohort study found substantial reductions in CRC recurrence risk for Danish patients with UICC stages I to III CRC from 2004 to 2019.The risk reductions were seen for all stages and for both colon and rectal cancers; reductions were especially notable in patients with screening-detected CRC but were also seen in patients whose CRC was not diagnosed through screening.Time to recurrence differed according to UICC stage.We believe that the risk of CRC recurrence has become so low in selected patient groups that further research on personalized surveillance protocols is indicated.
3 calendar periods: 2004 to 2008, 2009 to 2013, and 2014 to 2019.The latter period coincides with the introduction of the National Danish Colorectal Cancer Screening Program in 2014.

Figure 2 .
Figure 2. Time From Surgery to Colorectal Cancer Recurrence by Union for International Cancer Control (UICC) Stage 0.06

Figure 3 .
Figure 3. Cumulative Incidence Function Curves for Colorectal Cancer Recurrence According to Cancer Detection Method (Screening or Not Screening) Stratified by Union for International Cancer Control (UICC) Stage

Table 1 .
Demographics and Treatment Characteristics of Patients With Stages I to III Colorectal Cancer Grouped by Tumor Site and Stratified by Calendar Period of Primary Surgery Figure 1.Cumulative Incidence Function Curves for Colorectal Cancer Recurrence Stratified by Tumor Site and Compared by Calendar Period of Curative Surgery b Not reported for patients treated with neoadjuvant oncological therapy.c If the patient underwent resection.Incidence of Recurrence and Time to Recurrence in Stage I to III Colorectal Cancer Original Investigation Research jamaoncology.com(Reprinted) JAMA Oncology January 2024 Volume 10, Number 1 in Supplement 1).Over the 3 calendar periods, the proportion of recurrences diagnosed within the first 3 years after surgery increased (2004-2008, 80.0%; 2009-2013, 82.7%; and 2014-2019, 85.4%).From 2004 to 2008, the distribution of recurrences was dominated by a peak at 1 year postoperatively, while the pattern shifted such that by the 2014 to 2019 period, there were marked peaks at both year 1 and year 3 postoperatively (eFigure 7 in Supplement 1).Stage-specific TSTR revealed a shorter TSTR for patients with stage III CRC than for patients with stage II and particularly for
those with stage I (Figure2).