Differences in survival and recurrence of colorectal cancer by stage across population‐based European registries

Recurrence after colorectal cancer resection is rarely documented in the general population while a key clinical determinant for patient survival. We identified 8785 patients with colorectal cancer diagnosed between 2010 and 2013 and clinically followed up to 2020 in 15 cancer registries from seven European countries (Bulgaria, Switzerland, Germany, Estonia, France, Italy, and Spain). We estimated world age‐standardized net survival using a flexible cumulative excess hazard model. Recurrence rates were calculated for patients with initially resected stage I, II, or III cancer in six countries, using the actuarial survival method. The proportion of nonmetastatic resected colorectal cancers varied from 58.6% to 78.5% according to countries. The overall 5‐year net survival by country ranged between 60.8% and 74.5%. The absolute difference between the 5‐year survival extremes was 12.8 points for stage II (Bulgaria vs Switzerland), 19.7 points for stage III (Bulgaria vs. Switzerland) and 14.8 points for Stage IV and unresected cases (Bulgaria vs. Switzerland or France). Five‐year cumulative rate of recurrence among resected patients with stage I–III was 17.7%. As compared to the mean of the whole cohort, the risk of developing a recurrence did not differ between countries except a lower risk in Italy for both stage I/II and stage III cancers and a higher risk in Spain for stage III. Survival after colorectal cancer differed across the concerned European countries while there were slight differences in recurrence rates. Population‐based collection of cancer recurrence information is crucial to enhance efforts for evidence‐based management of colorectal cancer follow up.


| INTRODUCTION
Colorectal cancer is a major public health problem in developed countries because of its incidence and severity. 1It is the most frequent cancer after breast cancer, being responsible for approximately 520,000 new cases diagnosed in 2020 in Europe, accounting for 12.9% of all cancer diagnoses, and 250,000 deaths (12.6% of all cancer deaths).European average 5-year survival for patients diagnosed with colon and rectum cancer in 1999-2007 was 57% and 56%, respectively. 24][5] The geographic survival differences observed in Europe by population based studies can be partly associated to differences in diagnosis and treatment. 2To our knowledge, although patients may suffer after colorectal resection from recurrence, either local or distant visceral metastases, no population-based study has yet compared occurrence of recurrence between European countries.These data are key clinical determinants and provide healthcare planners with relevant information to improve cancer control.For this purpose, monitoring outcomes by stage at diagnosis is required to better understand differences across countries in the effectiveness of treatment, cancer care and survival.
Within the EUROCARE project, we carried out a "high-resolution" study (HIGHCARE project, [http://www.hrstudies.it/])using data collected by European cancer registries (CRs) to measure international differences in colorectal cancer management and outcomes.The objective of this study was to compare in a European populationbased cohort the 5-year risk of recurrence and the 5-year net survival according to stage at diagnosis of colorectal cancers.

| Population
Eighteen CRs from nine countries with national (Belgium, Bulgaria and Estonia) or regional (France, Germany, Italy, Poland, Spain and Switzerland) cancer registration, supplied data for this High-Resolution study.Data from regional CR were pooled by country.In order to draw unbiased samples with respect to the whole incidence series, the High-Resolution study protocol required CR to draw at least 500 invasive colorectal cancers adult (15 years or more) cases diagnosed between 2010 and 2013, with topography code C18-C20 classified according to the International Classification of Diseases for Oncology (ICD-O-3). 6One of the following sampling procedures was used by CR: (i) one (or more) year of complete incidence, for CR covering a small-medium area; (ii) an administratively defined sub-area (e.g. one region in a country, one health district in a region), for CR covering a large area (e.g.national CR); (iii) a defined time period of incidence for CR covering a large area (e.g.national CR).The Figure S1 shows the process of identifying eligible cases.Most CRs obtained the required (or higher) number of cases, by collecting all incident cases in one or more years of the study period.The process resulted in slightly less than 500 cases in the Belgium (492 incident cases) and Greater Poland CR (493 incident cases), and in some French and Spanish CR covering smaller areas or at lower incidence (Figure S1).These CR were included in the study in consideration of the small deviation from the requested sample size was small and in the case of regional CR, these were pooled by country.Trained CRs personnel accessed the clinical records and collected the clinical information required by the study protocol.Pseudonymized data were transmitted directly by each registry through a dedicated web-portal and were stored in a secure dedicated server.
The database of the Highcare project included 10,458 colorectal adenocarcinomas from 18 CR (Figure S1).For the present study, cases from CR with more than 25% missing data on stage at diagnosis (Belgium, Kielce and Greater Poland, N = 1554) were excluded.Additionally, colorectal cancers recorded by death certificate only (107 cases) or within autopsy (12 cases) were excluded.In the end, 8785 patients provided by 15 CR from 7 countries were included for descriptive and net survival analyses.Only nonmetastatic patients initially surgically resected for cure were considered at risk of local or distal recurrence, resulting to 4795 patients included in the recurrence analysis.Registries with 5-year clinical follow-up missing in more than 25% of cases (Geneva and Calvados) were excluded for the recurrence analyses.

| Variables
The High-Resolution study protocol with the list of required variables is available in the project website (http://www.hrstudies.it/protocols.html).Collected data included sex, age at diagnosis, primary colon or rectal location, tumor resection, chemotherapy administration and TNM staging.Colon cancers arised from the caecum to the sigmoid (C18), and rectal cancer from the rectosigmoid junction to the rectal ampulla (C19 and C20) according to the International Classification of Diseases for Oncology (ICD-O-3). 6Cancer extent at diagnosis was coded according to the pathological TNM classification of malignant tumors 7 in four stages: stage I limited to the colorectal wall (T1/T2, N0, M0), stage II extension beyond the colorectal wall (T3/T4, N0, M0), stage III lymph node involvement (any T, N1/N2, M0), and stage IV spread to distant organs.Unresected nonmetastatic patients could not be staged according to pathological staging.Due to small numbers and similar prognosis, unresected patients (N = 331) were pooled with metastatic stage IV patients.Stage was missing in 8.6% of the 8785 cases.Vital status and follow-up date were completed through an active follow-up and resulted in a proportion of patients lost to follow-up below 8% for all countries.
The probability of recurrence (locoregional or distant) was computed for resected colorectal cancer patients with primary stage I, II, or III.Locoregional recurrences were defined as recurrences in the site of the original primary tumor, at the bowel anastomosis or in lymph nodes that would classify as regional lymph nodes according to TNM classification.Distant recurrences were defined by any histological, morphological and clinical evidence of distant metastases according to TNM classification.

| Statistical analysis
Association between categorical data was analyzed using a two-sided chi-square test for heterogeneity.Due to limited power regarding the frequency of recurrence, stage I and stage II were pooled in relapse analyses.Excess Mortality Rate methods (Pohar-Perme and modelbased methods) take competing risk of death from other causes into account in order to provide an estimator that can be interpreted as survival in a hypothetical world without competitive risks. 8Net survival is recommended for geographical comparisons because competing causes of death may vary widely between countries or regions.It was estimated using the Excess Mortality hazard, which is the mortality due directly or indirectly to cancer and is the difference between the mortality hazard observed in the studied cohort and the expected mortality hazard.Expected mortality was that of the corresponding general population and was retrieved from lifetables of all-cause mortality stratified on year of age, sex, calendar year and geographic area covered by the cancer registry.Patients followed-up less than 5 years for vital status contributed up to the follow-up date.World agestandardized net survival was estimated firstly by country using the Pohar-Perme nonparametric estimator 9 and secondly modelled for each stage by country, adjusted on age and sex, using the flexible cumulative excess hazard model proposed by Nelson et al. 10 The inter-country difference in model-based net survival was tested by likelihood ratio test.The association between the probability of death and patients and tumor characteristics was estimated using adjusted Excess Mortality Rate Ratio derived from multivariable model with likelihood ratio tests.For each country, the Excess Mortality Rate Ratio was given as marginal effect (reference = mean of the study cohort).In stage II, European guidelines recommend chemotherapy only in intermediate-and high-risk patients. 11An additional multivariate Excess Mortality Rate model was fitted for stage II adjusting for individual chemotherapy administration in order to account for geographical heterogeneity of chemotherapy administration.
Recurrence rates were calculated using the actuarial survival method.The curves were compared using the log-rank test.Patients who died from intercurrent disease were censored at time of death.
Patients clinically followed-up less than 5 years contributed up to the follow-up date.A multivariate analysis was performed using a flexible cumulative hazard model.Hazard ratio of relapse was given as marginal effect (reference = mean).STATA©, release 17 (STATA corp., College Station, Texas) was used for all analyses.

| RESULTS
The mean age at diagnosis of colorectal cancer was 71.1 years (standard deviation, SD = 12.2) varying slightly according to country, from 69.1 years (SD = 10.6) in Bulgaria to 72.2 years (SD = 10.5) in Germany.Table 1

| Survival
The overall 5-year standardized net survival was 60.8% [95% CI: 55. 3  1A and Table S1 depict the comparison of modelled 5-year net survival of colorectal cancer between countries according to stage at diagnosis.There was no difference in net survival rates for cancers diagnosed at stage I while the magnitude of the differences between countries increased with stage.
The absolute difference between the 1-year survival extremes, was 6.4 points for stage II (Bulgaria vs. Switzerland), 6.9 points for stage III (Bulgaria vs. Switzerland), and 16.9 points for Stage IV and unresected cases cancers (Bulgaria vs. Switzerland or France).
Corresponding figures at 5 years were 12.8 points for stage II (Bulgaria vs. Switzerland), 19.7 points for stage III (Bulgaria vs. Switzerland), and 14.8 points for Stage IV and unresected cases (Bulgaria vs. Switzerland or France).Figure 1B graphically underlines that the geographical differences in stage-specific net survival were reduced after adjustment for sex, age, and cancer location in stages III and IV/unresected.As compared to the mean of all included countries, Bulgaria, Germany and Spain presented higher adjusted excess mortality rates in stage II (Figure 2).There T A B L E 1 Description of the studied population according to country in patients with colorectal cancer diagnosed between 2011 and 2013.Adjusting the model for individual chemotherapy administration suppressed the excess risk of death for Germany whereas it remained significant for Bulgaria and Spain (Figure S3).The risk of recurrence increased slightly with age at diagnosis of primary cancer.Women were less prone to develop a recurrence than men in both stage I/II and stage III colorectal cancers.In patients with stage I/II, recurrence was lower in colon than in rectal cancer.As compared to the mean of the six countries, the risk of developing a recurrence within the 5 years following the resection of the primary colorectal cancer did not differ between countries except a lower risk in Italy for both stage I/II and stage III cancers and a higher risk in Spain for stage III (Table S2).

| DISCUSSION
To the best of our knowledge, this is the first population-based study comparing recurrences rates after colorectal resection across T A B L E 2 Five-year net survival and 1-, 3-, and 5-year cumulative recurrence rates in nonmetastatic resected colorectal cancer by country during the 2011-2013 period.European countries, using areas covered by CRs.Our study highlights that net survival rates of colorectal cancer differed across the concerned European countries while there were slight differences in recurrence rates.
The distribution of the stages showed some differences between the countries included in the study, that may be partly explained by country-specific diagnostic procedures, screening implementation and management.3][14] Organized mass screening programme was implemen- were analyzed for patients diagnosed during 2000-2007. 16The proportion of distant disease ranged between 19% in Australia and UK and 31% in Denmark for colon and 17% in Australia and 27% in Denmark for rectal cancer.In our study, the proportions were higher, ranging from 22% to 41%.Five-year relative survival of patients diagnosed with colon cancer between 2010 and 2014 in Estonia was 90% for stage I, 86% for stage II, 71% for stage III and 15% for stage IV.The corresponding survival figures for rectal cancer were 92%, 75%, 70%, and 12%. 17In our study, the figures were similar for all stages except stage II presenting higher 5-year survival, without clear explanation.In a selected screened Italian population, the 5-year relative survival was similar to our study for stages I (97%) and IV (16%) but lower for stage II (83%) and III (71%). 18Social environment and inequalities in access to optimal prevention and care could partly explain some of the observed differences in survival. 19ile population-based studies comparing survival of colorectal cancer between countries are conducted regularly in Europe, this is, to our knowledge, the first comparing European risks of recurrence within 5 years of initial management using common standardized collection, coding, and analysis procedures.After adjustment for sex, age and cancer location, there were slight differences in recurrence rates among countries.In rectal cancer, the state of the circumferential resection margin after neoadjuvant treatment 20 and the total mesorectum excision are associated to the recurrence rate, 21 but it is possible that the proportion of patients who benefited from the technique varied by country.It is sometimes difficult for clinicians and pathologists to differentiate a cancer recurrence from a second primary colorectal cancer, so that misclassifications may exist and differ between countries, even with common coding rules.Data from the Swedish Colorectal Cancer Registry showed that 15% of patients with a radically operated colon cancer diagnosed between 2010 and 2018 have had a recurrence, a proportion similar to that of our study, and a 3-and 5-year recurrence risk of respectively 14% and 18%. 22In a population-based Dutch study, 16%, 13%, and 20% of patients with resected right-sided, left sided and rectal stage I-III cancer diagnosed in 2015 presented with disease recurrence 3 years after resection which is similar to the figures observed in our study. 23Although cancer recurrence is a crucial outcome in colorectal cancer care, it is not systematically and actively recorded in most population-based CRs. 24 colon and rectal cancers could be due to the different staging procedures adopted in stage III cancers, with more intensive imaging used in rectum, which makes the prevalence of undetected metastases lower than in colon cancers and reverse the intrinsic higher probability of recurrence of rectal cancer at least for stage III.In the next future, easier access to administrative databases could help in the registration of recurrences.Yet, events are often identified using the treatments administered, which concern only a part of the individuals. 25e geographical comparisons of our study were limited by its relatively small sample size which did not allow for separate analysis of colon and rectal cancer, and by the lack of data regarding major prognostic factors acting at initial diagnosis or after recurrence, such as surgery modalities, post relapse treatments, performance status or comorbidities. 26Furthermore, in countries with regional registry coverage, the registry areas included in our study might differ from nationwide data in terms of health resources and organization of oncological care.The overall net survival estimations may differ from that of low resolution (i.e.not including stage) previous studies 2,3 based on wider population.However, our objective was to allow comparisons of stage specific survival between countries, applying a uniform statistical methodology to the whole dataset with available data on stage.The relatively high proportion of missing information regarding clinical outcomes 5 years after diagnosis is unavoidable in large long-term population-based studies.

| CONCLUSION
This cooperative study provides unprecedented comparisons in recurrence rates of colorectal cancer by stage in Europe.Population-based collection of detailed data on cancer diagnosis, treatment, and recurrences is crucial to enhance efforts for evidence-based management of colorectal cancer follow up.
before the study period in Italy and France (respectively in 2003 and 2006), in 2014 in Spain, 2016 in Estonia and 2019 in Germany.In Switzerland, mandatory health insurance reimburse colorectal cancer screening by either colonoscopy or fecal occult blood test since 2013.No screening programme was set up in Bulgaria.A recent populationbased Danish study showed reduction in recurrence risk in patients with screening-detected colorectal cancer compared to nonscreening detected cancer (5-year cumulative rate for stage III 21.7% vs. 27.1%). 15This study included patients diagnosed between 2004 and 2019, which is comparable to our study period.The figure of 27% for nonscreening detected cancer is similar to the figures we observed in countries without organized mass screening.This may also partly explain the low recurrence risk observed in Italy.In the future, it will be interesting to study the impact of screening implementation on colorectal outcomes.The geographical variations in the distribution of stage at diagnosis may only partly explain the observed differences in 5-year survival which persisted within each stage.The low proportion of stage IV/unresected and the high proportion of stage III patients observed in Bulgaria might reflect differences in diagnostic examinations, which could partly explain through a Will Rogers phenomenon differences in survival, together with lower chemotherapy administration.Net survival was not associated with the frequency of chemotherapy administration by stage in each country.Stage II colorectal cancers represents a heterogeneous group regarding chemotherapy administration guidelines. 11While follow-up is an option for low-risk stage II patients, chemotherapy is recommended for intermediateand high-risk patients by European guidelines.Adjusting for individual chemotherapy administration reduced the geographical disparities in excess mortality rate.Data from population-based CRs in six high-income countries Obtaining such complete information requires time consuming methods, such as consulting medical files or directly contacting physicians.The detection of metastases at diagnosis or during follow up depends on the sensitivity of the diagnostic procedures used, therefore impacting both stage at diagnosis and cumulative incidence of recurrence.Differences in care practice can explain the situation observed in Bulgaria, where the lowest proportion of stage IV was associated with the lowest survival for both stage III and IV, suggesting that part of the metastases was undetected and thus stage III included some of the less lethal metastatic cases.This shift could have increased the average lethality of both stage III and stage IV.Similarly, the difference in recurrence for shows the distribution of the patients and tumor characteristics according to country.Colorectal cancer was more frequent in men than in women except in Switzerland (48.7% vs. 51.3%)and Estonia (48.2% vs. 51.8%).The colon/rectum location ratio varied from 1.4 in Estonia to 2.3 in Italy ( p < .001).The repartition of stages among colorectal cancers was overall 19.2%, 27.1%, 22.6%, and 31.0%for stages I, II, III, and IV/unresected, respectively.It was heterogeneous across countries ( p < .001),especially for Estonia which exhibited the lowest proportions of stage I (12.5%) and the highest proportion of stage IV/unresected (41.4%).Bulgaria presented the highest proportions of stages II (34.1%) and III (26.9%) and the lowest proportion of stage IV/unresected (21.5%).In Germany, the proportion of stage II was high (32.8%)and that of stage IV/unresected was low (24.2%).