Non-Curative Treatment Choices in Colorectal Cancer: Predictors and Between-Hospital Variations in Denmark: A Population-Based Register Study

Simple Summary Despite universal free healthcare and national treatment guidelines, between-hospital variation in treatment choices for colorectal cancer has been reported in several countries. It is unknown whether this variation may be ascribed to simple variations in clinical case mix or to differences in socioeconomic status or even attitudes and traditions among patients and healthcare professionals. This study examined factors associated with non-curative treatment choices and with refraining from recommended chemotherapy in colorectal cancer in Denmark in 2009–2018. We found that non-curative surgical treatment was associated with being old and frail and having widespread cancer and low weight. Not having chemotherapy was also associated with previous treatment complications and living alone. Marked variations in non-curative treatment between hospitals were found, even after taking a wide range of plausible explanations into account. The reason for these variations is unknown and requires further examination. Abstract Background: Variations in treatment choices have been reported in colorectal cancer (CRC). In the context of national recommendations, we aimed to elucidate predictors and between-hospital variations in refraining from curatively intended surgery and adjuvant chemotherapy in potentially curable colorectal cancer. Methods: A total of 34,116 patients diagnosed with CRC from 2009 to 2018 were included for analyses on non-curative treatment in this register-based study. Subsequently 8006 patients were included in analyses on adjuvant treatment. Possible predictors included patient-, disease-, socioeconomic- and perioperative-related factors. Logistic regressions were utilized to examine the predictors of a non-curative aim of treatment and no adjuvant chemotherapy. Results: The predictors of non-curative treatment were high age, poor performance, distant metastases and being underweight. Predictors for no adjuvant treatment were high age, poor performance, kidney disease, postoperative complications and living alone. For both outcomes we found between-hospital variations to be present. Conclusions: Non-curative overall treatment and refraining from adjuvant chemotherapy were associated with well-known risk factors, but the former was also associated with being underweight and the latter was also associated with living alone. Marked between-hospital variations were found and should be examined further.


Introduction
The cornerstone in the curative treatment of colorectal cancer (CRC) is surgery.Of the almost 4000 patients diagnosed with CRC in Denmark in 2022, 68% were treated surgically with curative intent [1].Treatment choices rest on national and European guidelines which have been in place for at least two decades [2][3][4][5], but the final decisions are made together with the individual patient, taking a multitude of patient-and disease-related factors, as well as patient preferences, into account.
The most common reason for refraining from curatively intended surgery is disseminated disease, although palliative surgery may be needed to manage acute problems such as bowel obstruction.Sometimes patients themselves decline curative treatment, and in other cases the surgeon will recommend against major surgery based on an assessment of performance status, age and comorbidity.In these cases, the choice may be compromised oncologic resection or no resection at all [6].
Although the aforementioned characteristics are the major predictors for the noncurative treatment of CRC, factors such as socioeconomic status and psychiatric disease could be important as well, as they are related to adverse outcomes [7][8][9].Finally, clinical experience, traditions and attitudes may vary between individual surgeons, hospitals, and even countries.Majano et al. examined differences between four Northern European countries in CRC survival and were able to relate them to marked differences in resection rates [10].
The use of adjuvant therapy may also vary.Commonly accepted guidelines recommend adjuvant chemotherapy after curatively intended surgery for Union for International Cancer Control (UICC) stage III cancer [2][3][4].The evidence of benefit from adjuvant therapy in UICC stage II CRC is more conflicting, but the European Society for Medical Oncology (ESMO) together with the Danish guidelines suggest adjuvant therapy if specific high risk features (e.g., pT4 tumor, fewer than 12 lymph nodes examined in the specimen) are present together with proficient mismatch repair (pMMR) status and favorable performance status [2][3][4].Present guidelines suggest adjuvant treatment, after relevant risk assessment [2][3][4], consisting of fluoropyrimidines alone or in combination with oxaliplatin.In stage II colon cancer with high risk features, the Danish guidelines suggest fluoropyrimidine monotherapy, while the addition of oxaliplatin is recommended in stage III in patients below the age of 70 [2,3].For rectal cancer, adjuvant chemotherapy is only recommended in those without neoadjuvant radiochemotherapy [2,4].Those with upper rectal cancer have the same indications as colon cancer, while for mid and lower rectal cancer fluoropyrimidines as a monotherapy are recommended [2,4].In certain cases, local radiotherapy can be applied [4].
However, in spite of national guidelines for both the surgical and oncological treatment of CRC, recent annual reports by the Danish Colorectal Cancer Group have found marked between-hospital variations in, e.g., the proportion of patients having surgery for rectal cancer (45-88%) [1] and adjuvant chemotherapy for stage III colon cancer (67-92% and 55-93%) [1,11].The reason for these variations is unknown, but it may be speculated to reflect the differences in case mix between hospitals that we have shown previously [9].
The aim of this study was to elucidate predictors and between-hospital variations in refraining from curatively intended surgery and adjuvant chemotherapy in potentially curable colorectal cancer.

Study Design and Population
This study is a register-based Danish national cohort study.The data were extracted from the Danish Colorectal Cancer Group (DCCG) database [12], the National Patient Registry (NPR) under the Danish National Health Authority [13] and Statistics Denmark (SD) [13].
The cohort consists of 44,471 adults with a first-time diagnosis of CRC in the years 2009-2018.The cohort and covariates are thoroughly described in Rattenborg et al. [9].

Specific Primary Exclusion Criteria for This Study
In this study we focused on patients with potentially curable disease.Therefore, we added the following exclusion criteria (Figure 1):

•
Incompatibility between the tumor location and resection procedure registered in the DCCG (e.g., primary tumor in the right colon and rectal resection), or synchronous cancer;

•
Patients in whom curative surgery was not an option (e.g., patients who died before surgery or patients not offered surgery because of disseminated disease); • Non-elective surgery;

•
Rare and poorly characterized aims of treatment or an unspecified aim of treatment; • The presence of distant metastasis (UICC stage IV) was not an exclusion criterion, provided that the surgeon had not registered disseminated disease as the reason for no surgery.

The Stratification of the Cohort by Overall Treatment Goal
Patients were stratified into four different groups based on their intended treatment registered in the DCCG database: operative treatment with curative intent (OT-CUR); operative treatment with compromised or palliative intent (OT-NCUR); non-operative treatment because the patient declined (NOT-NO); and non-operative treatment due to comorbidity (NOT-CO).'Compromised surgery' was defined as a suboptimal oncological resection chosen to minimize the surgical trauma in a frail patient.The 'palliative surgery' category had rather few elective patients registered and therefore it was fused with the 'compromised surgery' category for our purpose.Data were collected from the DCCG.

Specific Secondary Exclusion Criteria for Adjuvant Oncological Treatment
Patients in the OT-CUR group and with UICC stage II or III disease and with at least one high-risk feature for benefit (inclusion criteria) of adjuvant therapy, according to national recommendations [2,5], were eligible for analysis, if no exclusion criteria were present.Or, in other words, patients in whom adjuvant chemotherapy was indicated according to national guidelines were eligible for analysis.These changed over the decade studied, and for each patient, we applied the criteria in force at the time of that patient's operation (Supplementary Table S1).Patients > 80 years old were excluded, since it was (by the authors) considered a general rule of thumb for no benefit of adjuvant therapy during the whole period.Patients only having local or unspecified resections were excluded from this analysis, as well as (in some years) patients with deficient mismatch repair (dMMR) status, age > 75 years and World Health Organization (WHO) performance status (PS) > 2 (Figure 1 and Supplementary Table S1).Recommended adjuvant chemotherapy regimes were, through the whole period, fluoropyrimidine, and possibly leucovorin, and oxaliplatin, depending on the aforementioned risk assessment [2,5].We grouped included patients as having any or no adjuvant chemotherapy treatment.Data on adjuvant chemotherapy were collected via the DCCG from the NPR.

Comorbidity
Comorbidity was reported on an overall level as an aggregated Charlson comorbidity index [15] score (0, 1, 2, 3+) in descriptive tables with updated weights [16].International Classification of Disease 10th edition (ICD-10) codes for CRC were excluded.Only dichotomous comorbidity variables were included in the regression analyses, based on ICD-10 and Anatomical Therapeutic Chemical Classification System (ATC) codes, as reported recently [9].In brief, the included somatic domains were cardiovascular disease, chronic pulmonary disease, diabetes, dementia, liver disease, kidney disease, chronic nerve disease, other cancer or tumors and connective tissue disease, and the included psychiatric domains were affective disorders, schizophrenia spectrum disorders, disorder of adult personality and behavior and disorders due to psychoactive substance abuse.ICD-10 codes were collected from the NPR and ATC codes from the prescription database at SD.

Disease-Related Factors
The primary tumor was defined as located in either the colon or the rectum.Clinical presentation with distant metastases at the time of diagnosis was included (yes, no, unknown), as well as information on pretreatment discussion at a multidisciplinary team (MDT) conference (yes, no, unknown).The hospital responsible for the definitive treatment was included in the analyses.These hospitals were identified by letters A to Q, ordered by total patient volume (low to high).Data were collected from the DCCG.

Covariates Included Only in Regression Models for Adjuvant Chemotherapy
In addition to the aforementioned variables, the following variables were of interest in the adjuvant chemotherapy analysis: mismatch repair (MMR) status was categorized as proficient MMR (pMMR), deficient MMR (dMMR) or unknown and collected from the DCCG.In the study period, the most common neoadjuvant radio-chemotherapy regime for rectal cancer was 50.4 Gray in 28 fractions and fluoropyrimidine, applied mainly for T4 and T3 (depending on suspected involvement of margins) [2,5].Neoadjuvant therapy was categorized as no, yes or unknown for those with rectal cancer only and was collected from the DCCG and via the DCCG from the NPR.Data on postoperative medical and surgical complications within 30 days after surgery were included (no, yes or unknown) and collected from the DCCG.

Statistical Methods
Descriptive statistics were applied to examine between-hospital variations in outcome variables.For group comparisons (e.g., between two different proportions) we estimated 95% confidence intervals (CI).
In order to examine predictors for non-curative treatment aims, a multivariable multinomial logistic regression model was utilized.The outcome variable was the aim of treatment with the OT-CUR group as the comparison level.The relevant covariates were included.We applied the Hosmer-Lemeshow test with twenty groups to investigate the goodness of model fit.Results are presented as relative risk ratios (RR) with corresponding 95% CI.
In order to examine predictors for refraining from adjuvant chemotherapy, a multivariable logistic regression model was built for colon and rectal cancer, respectively.The possible outcomes were any kind of adjuvant chemotherapy, with no adjuvant chemotherapy as the comparison level.As mentioned briefly above, exclusion criteria in force at the time of the operation of each particular patient were used to define the population with the indication for adjuvant chemotherapy.Since some of these exclusion criteria may have predicted non-treatment with adjuvant chemotherapy before they were included in national recommendations, we included these criteria in our analysis.To balance the analysis we recoded age, MMR and neoadjuvant therapy to non-applicable (n-a) in the periods when they were exclusion criteria.The exact distribution (without n-a) is reported.Results from this model are presented as odds ratios (OR) with corresponding 95% CI.Sensitivity analyses with a reduction of hospitals to tertiles (low, medium, high) of the total volume of patients were done in order to examine if hospital volume was a predictor.
Missing data were included for all analyses and treated as an unknown level for each variable.We did not perform imputation.Data were stored and managed on the secure servers of Statistics Denmark, using Stata IC/17 (StataCorp LCC, 4905 Lakeway Drive, College Station, TX, USA) for analysis.Data were only extracted after anonymization.

Ethics and Permisson
This study was approved by the DCCG and Danish Clinical Quality Program (DCCG-2018-03-08a) and the Danish Data Protection Agency (jr.no 18/15252).No other approvals were required under Danish law [18].

Results
For the overall analyses, 34,116 patients were included in this study (Figure 1).

The Characteristics of Overall Treatment Aims
Table 1 shows an overview of the included cohort, stratified by the aim of treatment.The majority (90%) had OT-CUR, with five percent having OT-NCUR.Three percent had NOT-NO and three percent had NOT-CO.During the study period, the proportion of patients receiving curative treatment was consistent.The proportion of patients who had OT-NCUR decreased from eight to three percent over the study period, while proportions of NOT-NO and NOT-CO were relatively constant during the study period.

The Characteristics of Patients with a Curative Treatment Aim
The patients in the OT-CUR group were generally younger, had a lower ASA score and PS, lower Charlson scores, higher educational level, higher annual household income and were more often cohabiting (Table 1).The majority of OT-CUR patients had a segmental resection, while 6% had a local resection (e.g., endoscopic mucosal resection).Also, a few patients (<1%) in this group did not actually receive any curative operative treatment although the intention of the treatment before surgery was curative.

The Characteristics of Patients with a Non-Curative Treatment Aim
The non-curatively treated patients were characterized by higher age, ASA, PS and Charlson score, short educational level and lower annual household income, and were more frequently living alone (Table 1).
OT-CUR and OT-NCUR were more common among patients with colon cancer rather than rectum cancer, while NOT-NO and NOT-CO were less often found in colon cancer patients (Table 1).The most predominant types of surgery in OT-NCUR were resection (47%), relief of obstruction (42%), local resections (7%) or only exploration (4%) (not shown in tables).
Looking at the data descriptively, non-curative treatment varied from 8 to 13% between hospitals as seen in Figure 2. OT-NCUR treatment varied from 2 to 7%.NOT-NO and NOT-CO varied only a few percentages between hospitals, 2 to 4%, and 1 to 4%, respectively.

The Predictors of Non-Curative Treatment in a Multinomial Model
The highest RR values among the predictors of OT-NCUR were distant metastases and advanced age, ASA or PS (Table 2).To a lesser degree BMI, dementia, other cancer disease at diagnosis and rectal location were predictors of OT-NCUR.Between-hospital variations in RR varied significantly, with RR as low as 0.3.The highest RR among the predictors of both NOT-NO and NOT-CO were high age, ASA and PS (Table 2).Also, rectal cancer, year of diagnosis, distant metastases, BMI and living alone were predictors of both non-operative aims.A range of comorbidities were, not surprisingly, predictors of NOT-CO, while low income was a predictor of NOT-NO.Between-hospital variations in RR shared patterns for NOT-NO and NOT-CO, with RR up to 2.8 for NOT-NO and between 0.3 and 2.1 for NOT-CO.In a sensitivity analysis, where hospitals were contracted to tertiles (1st, 2nd and 3rd) of the total volume of patients, the only outcome which was markedly different with volume was NOT-CO, with a RR for the low volume tertile of 0.61 (95% CI 0.49-0.77)using the high volume tertile as the comparison level (not shown in table).

Adjuvant Chemotherapy
For analyses on adjuvant chemotherapy, 8006 patients with an indication for adjuvant chemotherapy were included (Figure 1).An overview is seen in Table 3.The majority of patients received adjuvant therapy (72%), with the largest proportion in patients with colon cancers.The majority (81%) of UICC stage III received adjuvant treatment, while only 42% of stage II had treatment.Those who did not receive adjuvant chemotherapy were generally of higher age, had a higher ASA and/or Charlson score, had a shorter education, had a lower annual household income, were living alone and had more often had complications within 30 days after surgery.The unadjusted between-hospital variations are seen from Figure 3.

The Predictors of Adjuvant Chemotherapy in a Logistic Regression Model
The multivariable logistic regression tables for having any adjuvant treatment for colon and rectal cancer respectively are shown in Table 4.The most significant predictors associated with not receiving adjuvant therapy (and thereby having a low OR for receiving adjuvant therapy) for both colon and rectum were as follows: high age, ASA III and PS 1 or 2-4, respectively, together with kidney disease, postoperative complications, living alone and earlier years of surgery.For colon specifically, dementia, liver disease, other cancer and dMMR were also predictors, while obesity was found to be a predictor of receiving treatment.For rectum, nerve disease, neoadjuvant treatment and being underweight were predictors of no treatment.The adjusted OR for hospitals varies from 0.6 to 4.5 and 1.8 to 5.4 for colon and rectum, respectively.In sensitivity analyses, a low volume tertile was a predictor in colon (OR 0.77, 95% CI 0.65-0.92),but not in rectum.It should be noted that for rectum, the two hospitals with the lowest numbers had only 11 and 44 cases in the whole period, but even without these, the resulting OR of a low volume tertile was still insignificant.Some other variables were significantly associated with the outcome, but had a 95% CI very close to 1.

A Review of Aim and Results
We aimed to elucidate predictors and between-hospital variations in refraining from curatively intended surgery in potentially curable colorectal cancer.We analysed this in a cohort study of patients diagnosed with CRC during the years 2009-2018, using the Danish national registers.Similarly, we aimed to examine predictors and between-hospital variations in refraining from adjuvant chemotherapy recommended in national CRC treatment guidelines.We wish to emphasize that we are confident that the actual treatment choices in each case were made for good reasons, and we did not intend to judge hospitals whose treatment patterns seemed to deviate from those of others.We only wanted to identify overall predictors for the treatment choices and to investigate whether these predictors could explain between-hospital variation in adjusted analyses.
Generally speaking, the most clinically significant results were those whose coefficients (RR or OR) deviated markedly from 1 (e.g., 0.75 or 1.25), but also had 95% confidence intervals which deviated markedly from 1 (e.g., not including 0.90 or 1.10).In light of this, we found, for the multivariable analysis of overall treatment aim, that advanced age, ASA and PS, as well as distant metastases at diagnosis and being underweight, were predictors of all non-curative treatment aims.Apart from BMI, all of these are wellestablished predictors for adverse outcomes.For OT-NCUR, dementia and other cancer were also predictors, compared with an aim of curatively intended surgery.Specifically for non-operative treatment, having a rectal tumor, being in the lower two quartiles of household income and living alone were predictors compared with OT-CUR.For NOT-CO, dementia, liver disease and other cancer were found to be predictors, which is not surprising.The between-hospital variations in the overall aim of treatment were found to vary significantly even after multivariable adjustments for all three non-curative aims of treatment.It may be noted that the treatment choices did not seem complementary, i.e., hospitals with a significantly positive coefficient for NOT-NO generally also had a positive coefficient for NOT-CO.
We found that refraining from adjuvant treatment was predicted by high age and ASA, kidney disease, postoperative complications and living alone, both for colon and rectum cancer.Age and ASA are markers for patient frailty, and kidney disease is important for tolerance of adjuvant chemotherapy [3].Postoperative complications can lead to a prolonged hospital stay and deranged performance and thereby affect the presumed effect and tolerability of chemotherapy [19].For colon cancer specifically, PS, dementia and dMMR were also predictors.For rectum cancer specifically, nerve disease, neoadjuvant treatment and being underweight were also predictors.The between-hospital variations in use of adjuvant chemotherapy were significant even after multivariable adjustments in both colon and rectum cancer.

A Discussion of Related Studies
As shown in Table 1, we found a distribution between treatment strategies of 90% (OT-CUR), 5% (OT-NCUR), 3% (NOT-NO) and 3% (NOT-CO).We have no knowledge of other studies that are directly comparable.Giesen et al. had a mean of 95% resected patients for non-metastatic CRC in the Netherlands [20].We found being underweight to be a predictor of non-curative treatment.In an earlier study, we found obesity to be protective of 90-day mortality [9], which seems in line with our findings in the present study.Axt et al. found that weight loss was associated with postoperative complications [21], unfortunately we have no information on pretreatment weight loss.We also found that living alone was associated with non-operative treatment aims.It is well established that lower socioeconomic status is associated with an adverse outcome of CRC in populationbased studies [7,22,23] and this correlates with our findings of low income being associated with NOT-NO.In our study, not all socioeconomic factors were predictors.As shown by others, the effect of each socioeconomic factor does not necessarily point in the same direction due to different causal mechanisms and associations with specific covariates and outcomes [24,25].We found variations between hospitals in all three non-curative treatment aims, however with rather low numbers, especially for OT-NCUR.Giesen et al. found that resection rates vary between hospitals in the Netherlands, but that they did not translate into differences in mortality [20].In a Swedish study, Ljunggren et al. found variations in metastatic surgery for CRC between university hospitals and non-university hospitals, but no difference in the hospital volume of patients [26], while an English study by Downing et al. found that hospitals participating in interventional studies had better survival than those who did not [27].We did not include data on university hospitals nor participation in interventional studies.Regarding adjuvant chemotherapy treatment, Babaei et al. found between-hospital variations in stage II with risk factors (17-38%) and stage III (55-68%) in a register-based study conducted in the Netherlands, Sweden and Belgium [28].If you ask the oncologists about adherence to national guidelines on adjuvant therapy in the Netherlands, 66% and 84% agreement was found for stage II and III, respectively [29].Shared decision-making is on the rise, also in Denmark, but is not implemented in guidelines yet [30].Probably this had little effect on between-hospital variations in our study of patients from the years 2009 to 2018.The goodness of fit (GOF) test of the overall analysis returned a very low p-value (Table 2) and the interpretation of this is important.A significant GOF test is not surprising in a large cohort and just implies to the reader that on a population level (or hospital-specific level), these relative risk ratios are plausible, but should not be used on an individual level.
We also found that 28% of the patients with an indication for adjuvant chemotherapy did not have any adjuvant treatment, with differences between stage II (58%) and III (19%).DCCG has set a desired target level for commencing adjuvant therapy of 80-90% for stage III CRC [1], which is in line with our findings, as seen from Table 3.In the latest annual report from the United Kingdom, 61% of stage III colon cancer received a major resection followed by adjuvant chemotherapy in 2019 (latest pre-COVID-19 pandemic numbers) [31].We found that refraining from adjuvant treatment was predicted by living alone, which is in line with the socioeconomic indicators found in a recent systematic review and metaanalysis [32].Our findings of variations between hospitals are in line with annual reports from the United Kingdom and Denmark, although the latter is with rather low numbers per hospital [1,31].

Strength and Limitations
This study has some limitations.First, the period these patients are collected from is rather old (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018) and this can affect the generalizability of our results.Secondly, we have missing data on some variables such as reason for no referral for adjuvant therapy evaluation, WHO performance status (which is missing in the majority of patients as seen in Table 1) and clinical TN category at diagnosis (although metastases at diagnosis is included).Third, it is worth mentioning that we pooled ypT and pT categories in the analyses on adjuvant therapy.Those with a good response on neoadjuvant therapy probably have less of an effect from adjuvant therapy and vice versa [2], and this could affect the adjuvant treatment decision.We have insufficient data on this and we suspect this to have little impact on our overall conclusions.Fourth, we did not include genetic markers, such as KRAS.ESMO guidelines do not recommend the inclusion of these, due to a lack of utility in the adjuvant treatment decision-making process [3].The strengths of this study are mainly its size, including a large cohort with almost 100% coverage of the national population [33], but also the prospective collection of data (although retrospectively extracted) from the national registries.

Perspectives
Refraining from curative treatment is more widespread in some hospitals than in others, even when adjusting for various predictors.No desirable proportions per hospital are applicable, since these discussions are complex with patients at a very vulnerable point in their lives and the decisions are a subject of personal preferences.Further elaboration into this research area requires further data on the impact of treatment choices on survival, morbidity and quality of life.Also, research into tools aiding shared decision-making could probably promote further equity for these patients in making well-informed decisions.

Conclusions
An intended non-curative aim of treatment in colorectal cancer was associated with age, performance status and distant metastases.Refraining from indicated adjuvant chemotherapy was also associated with age and performance status, but also with living alone, kidney disease and postoperative complications.Between-hospital variations in treatment choices were found, even in adjusted analyses, and should be examined further.

Supplementary Materials:
The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/cancers16020366/s1,Table S1: Overview of inclusion and exclusion criteria and changes over the study period regarding national recommendation of treatment with adjuvant oncological treatment after curative re-section of stage II and III colorectal cancer.Informed Consent Statement: Since this is a register-based study, no consent from patients was required under Danish law [17].

Figure 2 .
Figure 2. Proportions of non-curative aim of treatment by hospital (A-Q).Abbreviations: OT, operative treatment; NOT, non-operative treatment.

Figure 3 .
Figure 3. Proportion of patients receiving adjuvant chemotherapy by hospital (A-Q).Hospital A and B only treat colon cancer.

Table 1 .
Characteristics of 34,116 patients with colorectal cancer diagnosed in the years 2009-2018 in Denmark by overall treatment aim.
In order to avoid showing identifiable data some numbers have been changed to <N. 1 One unit = 12 g of pure ethanol.2InternationalStandard Classification of Education (ISCED) 2011.Abbreviations: ASA, American Society of Anesthesiologists; WHO, World Health Organization; MDT, treatment decisive multidisciplinary team conference; OT-CUR, operative treatment with curative intent; OT-NCUR, operative treatment with compromised/palliative intent; NOT-NO, non-operative treatment due to patient decline; NOT-CO, non-operative treatment due to patient comorbidity.

Table 2 .
Multivariable multinomial logistic regression of a non-curative aim of treatment with curative surgical treatment as the comparison level for 34,116 colorectal cancer patients.
1 One unit = 12 g of pure ethanol.2InternationalStandard Classification of Education (ISCED) 2011.p-values < 0.05 in bold.Abbreviations: RR, relative risk ratio; CI, confidence interval; ASA, American Society of Anesthesiologists; WHO, World Health Organization; MDT, treatment decisive multidisciplinary team; OT-CUR, operative treatment with curative intent; OT-NCUR, operative treatment with compromised/palliative intent; NOT-NO, non-operative treatment due to patient decline; NOT-CO, non-operative treatment due to patient comorbidity.Goodness of fit test: p-value= 0.000.

Table 3 .
Overview of 8006 colorectal cancer patients with an indication for adjuvant chemotherapy in the years 2009-2018.

Table 3 .
Cont.In order to avoid showing identifiable data some numbers are changed to <n. 1 One unit = 12 g of pure ethanol.2InternationalStandard Classification of Education (ISCED) 2011.Abbreviations: ASA, American Society of Anesthesiologists; WHO, World Health Organization; MDT, treatment decisive multidisciplinary team; pMMR, proficient mismatch repair; dMMR, deficient mismatch repair.

Table 4 .
Multivariable logistic regression of receiving adjuvant chemotherapy for 8006 stage II or III colorectal cancer patients with an indication for adjuvant chemotherapy treatment.