Exploration of between-hospital variation and the basis for decision making in management of malignant colorectal polyps- a Danish population-based study

Abstract

Background

Finding an unexpected carcinoma in an endoscopically resected colorectal polyp poses a dilemma regarding the subsequent management strategy. Proceeding to surgery with formal segmental bowel resection is associated with a low recurrence risk but substantial morbidity and mortality, whereas surveillance without surgery entails low morbidity but a higher risk of recurrence. Clinical guidelines are based on histopathological risk factors (HRF), but pathology data are often incomplete, and national databases have revealed marked practice variation between hospitals.

We aimed to explore between-hospital variation and the basis for treatment decisions after endoscopic resection of malignant colorectal polyps in Denmark.

Methods

A national cohort of colorectal cancer patients from 2016-2020 was extracted from national clinical and pathology registers. Patients undergoing local resection only (surveillance group) and local resection followed by subsequent bowel resection (surgery group) were compared in uni- and multivariable analyses stratified by reported HRF. Patient- and hospital-related factors were included as covariates with particular focus on between-hospital variation.

Results

Overall, 2,188 patients were analyzed, 1,277 in the surveillance group and 911 in the surgery group. Multivariable analyses showed that male sex, older age, comorbidity, lower performance status and left colon or rectum tumor location were significantly associated with surveillance, most even in the presence of HRF. Long higher education and certain hospitals were significantly associated with bowel resection. Predictors of bowel resection despite absence of HRF were certain hospitals and active smoking. In the surgery group without HRF conclusive information was missing in up to 70% of the pathology reports regarding certain HRF. Preoperative image-based overstaging may have resulted in a higher rate of cancer-free specimens. Overall, 63% of bowel resection specimens were cancer-free.

Conclusion

We found marked between-hospital practice variation in management strategy for malignant colorectal polyps, even in adjusted analyses. Among the probable explanations were missing or incomplete pathology data and suspicion of more advanced disease based on clinical staging. Consistency in hospital practice, completeness of the pathology reports and overall better collaboration of the multidisciplinary team are needed to improve the decision-making process in patients with endoscopically removed malignant colorectal polyps.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12957-026-04368-6.

Introduction

Implementation of bowel cancer screening has resulted in the detection of more colorectal carcinomas at an early stage. Often, the carcinoma is merely a small focus detected by the pathologist in a polyp initially judged as benign by the endoscopist and removed endoscopically. In many cases, the patient will be cured by the endoscopic treatment alone. However, some patients are at increased risk of recurrence owing to residual tumor either in situ or in the regional lymph nodes. A decision must thus be made between either close endoscopic and radiologic surveillance (SV) or proceeding with a subsequent segmental bowel resection (SBR), a so-called ’completion’ resection. Several national and international guidelines are available to guide these decisions, based mainly on the presence or absence in the local resection specimen of certain histopathological risk factors (HRF) for lymph node involvement and recurrence, but applied, of course, in combination with an overall assessment of the patient’s general condition and surgical risk [1–4].

The treatment decision constitutes a dilemma, however. The SV strategy relies on local resection alone with close follow-up. It is therefore associated with a low procedure-related risk, but at the same time entails frequent hospital visits and a risk of locoregional recurrence that may worry both patients and health care providers. In the largest meta-analysis, with at least 12 months follow-up, recurrence after endoscopic resection alone was 7–9,5% in malignant colorectal polyps with presence of HRF and 0,7 − 1,2% in malignant colorectal polyps without reported HRF [5].

The SBR strategy, on the other hand, minimizes the risk of locoregional recurrence but entails a formal bowel resection with substantial perioperative morbidity and mortality as well as long-term sequelae [6]. However, the rate of residual cancer in bowel specimens from SBR varies between 17.5 and 43% [7, 8], meaning that the majority of SBR operations are unnecessary from an oncological point of view.

The situation is further complicated by the fact that large cohort studies have found that histopathological information is incomplete in many patients [9–11].

Despite universal free healthcare and national guidelines in place, recent Danish national quality reports have shown marked between-hospital practice variation regarding the management of malignant colorectal polyps [12]. The proportion of SBR for malignant colorectal polyps varied from 14% to 100% between hospitals in 2018 [12]. It is not known whether this practice variation can be explained by variations in case mix [13] or represents a more systematic (hospital- or clinician-level) variation in treatment practices or attitudes. Furthermore, the relationship between treatment choices and the availability of relevant histopathological information has not previously been elucidated.

This study aimed to explore the between-hospital practice variation described above by including a larger cohort of patients from 2016 to 2020 and furthermore, by merging these with data from other national databases to explore in more depth the variation and the basis for decision-making in the management of malignant colorectal polyps in Denmark.

Methods

Data sources and linkage

This study was based on data from the Danish Colorectal Cancer Group Database, which is the national bowel cancer database in Denmark, founded and maintained by the Danish Colorectal Cancer Group (DCCG) [14]. It holds perioperative and clinical data on almost all (> 98%) patients with a first-time diagnosis of colorectal cancer (CRC) [15]. Registration of patients in the database is mandatory for all surgical departments. The Danish quality databases are anchored in the Healthcare Quality Institute (SundK). Additional data, including all histopathology, were provided by the National Pathology Database (NPD) [16] and Statistics Denmark (DST). The data were merged and hosted by OPEN Registry through OPEN (Open Patient data Explorative Network, Odense University Hospital, Region of Southern Denmark). The link between all datasets was the unique personal ID number (CPR) assigned to all citizens with permanent residence in Denmark.

Population

Data processing

Data on 22,864 patients with a first-time CRC from the calendar years 2016–2020 (5 years) were extracted from the NBCD. Based on the reporting surgeon’s classification, 30% were rectal cancers and 70% colon cancers.

We excluded patients with no surgery, emergent surgery, synchronous colorectal cancer, only palliative procedures (stent, stoma, bypass, etc.), or combined endoscopic-laparoscopic surgery (CELS). The remaining patients were classified into those having a formal bowel resection (n = 14,896) and those having a local resection (polypectomy, endoscopic resection, etc.) as their definitive surgical treatment (n = 1,429). In the latter group, 380 were transanal procedures (transanal endoscopic microsurgery (TEM), transanal endoscopic operation (TEO), etc.). From 2016 onwards, surgeons had to register in the NBCD whether a bowel resection was a SBR following a previous endoscopic local resection of an adenocarcinoma. Based on these registrations, we could exclude patients undergoing a formal bowel resection as the primary surgical procedure and define two treatment strategies: (1) Formal bowel resection as ‘completion’ procedure (n = 1,033); and (2) Local resection only (n = 1,429). The patients in these two groups constitute our study population and were denoted “subsequent bowel resection” (SBR) group and “surveillance” (SV) group, respectively. All histopathological information from the years 2016–2020 and from a location in the colon or rectum of these 2,462 patients was requested from the NPD. Details are illustrated in the STROBE diagram in Fig. 1.

Fig. 1.

STROBE diagram

Histopathology – specimen types

The NPD dataset contained 7,082 observations on 2,461 patients, while one patient from the SV group was missing. Each observation represented a pathology report requested on a single form, but often describing several subspecimens (e.g., polyps, biopsies) collected from the same patient on the same occasion. Available information included the date of request, date of receipt of the specimen at the Pathology department, and date of reporting, as well as all formal pathology encoding according to the Danish Pathology Coding System (www.patobank.dk/snomed). We based all statistical analyses on the pathology codes, but ad hoc access to the free text of the reports enabled us to deal with doubtful or inconsistent procedure dates, specimen types and specimen locations. In particular, analysis of histopathological findings was based exclusively on the codes as reported by the pathologist, with the very few exceptions explained below. All extraction of pathology data was performed by two senior authors (HBR and JL).

We excluded reports without anatomical codes for colon or rectum, as well as those without codes for either a relevant carcinoma or an explicit code for “no residual tumor”. We split the remaining 3,859 observations into individual sets of pathology codes for each of the up to 21 subspecimens described in each report and discarded subspecimens without codes for either a relevant carcinoma or “no residual tumor”. This resulted in 3,938 observations for further analysis, of which 79 observations were for two or more sub-specimens from the same pathology report. These observations will hereafter be referred to as specimens, and were classified into local specimens (i.e., biopsies and specimens obtained by any local procedure, including TEM, etc.) and bowel specimens (i.e., specimens obtained by formal bowel resection).

Tumor locations were encoded as Caecum, Ascending, Right Flexure or Transverse Colon; or Left Flexure, Descending or Sigmoid Colon; or Unspecified Colon, Unspecified Right Colon or Unspecified Left Colon; or, finally, Rectum. The locations were further classified into Right-sided Colon, Left-sided Colon, and Rectum. We excluded 11 specimens with uncertain tumor locations.

By definition, a local resection of a colorectal carcinoma, which is not followed up by a ‘completion’ bowel resection within 120 days, is regarded as the definitive treatment of the carcinoma. Whereas, if a formal resection of the bowel segment is performed within 120 days, this ‘completion’ resection is regarded as the definitive treatment.

Tumor sidedness and identification of the definitive specimen

For each patient, we defined the specimen produced by the definitive surgical treatment of the case as the definitive specimen.

In the SBR group, we identified the definitive specimen (i.e., bowel specimen) by matching the type and date of the definitive procedure, as recorded by the surgeon, with the tumor location and date of receipt at the Pathology department, as recorded by the pathologist. In case of inconsistencies, we accessed the free text of the pathology reports and tried to resolve conflicts and correct obvious errors. If doubt or conflict remained, we excluded the specimen and/or patient. Left-sided specimens removed by a resection involving the rectum, and rectum specimens removed by a sigmoid resection, remained, however, since the surgical distinction between a low sigmoid resection and a high anterior resection of the rectum is notoriously difficult. SBR patients without bowel specimens were excluded, leaving all SBR patients with one definitive specimen. Subsequently, we matched the sidedness and location of each patient’s other specimens with the location of the definitive specimen. Again, any conflicts were either resolved or led to exclusion of the specimens and/or patients (e.g., patients with two carcinomas).

In the SV group, we matched the sidedness of the local specimens with the tumor location registered by the surgeon at the definitive procedure, since local resection procedures are generally not specific for a particular bowel segment, except for transanal procedures. Conflicts, uncertainties or errors were either resolved or led to exclusion of specimens or patients.

Relationship between specimens and treatment decisions

In the SBR group, pathology reports concerning specimens received at the Pathology department < 120 days earlier, and sent at least two days earlier than the definitive procedure, were regarded as the basis for the decision to proceed with ‘completion’ surgery and were coded accordingly. Patients without any local specimens before the definitive operation were excluded.

In the SV group, all available specimens received in the Pathology department up to 120 days after the definitive procedure were regarded as the definitive histopathology of the case and thus part of the basis for the decision not to proceed with SBR. Eighteen patients from the SV group died within 120 days of the definitive procedure and could therefore not be assigned to the SV group with certainty and were excluded. We also excluded 86 patients whose definitive procedure date was less than 120 days before 31 December 2020, since pathology data from 2021 were incomplete.

The abovementioned data processing left us with the following population for extraction of detailed histopathological features:

• ◦ Subsequent Bowel Resection group

  • ◦ 911 patients with 1,884 specimens.
  • ◦ No. of specimens prior to definitive (i.e., basis for decision), 973; median no. per patient, 1 (range 1–5; 1 in 855 patients; 2 + in 56 patients).

• ◦ Surveillance group

  • ◦ 1,277 patients with 1,544 specimens.
  • ◦ Median no. of specimens/patient, 1 (range, 1–3).

Extraction of data on histopathological features and risk factors

Specimens of both local resections and subsequent bowel resections were classified into histopathological type (adenocarcinoma, poorly differentiated carcinoma, mucinous carcinoma, signet cell carcinoma, medullary carcinoma) or “no residual tumor” based on the pathology codes. One specimen had ambiguous encoding and was recoded as “no residual tumor” (no remaining primary, but pN+).

Likewise, information on pT and pN categories, involvement of resection margin, tumor depth, tumor grade, venous and lymphatic invasion, tumor budding, nervous invasion, and mismatch repair (MMR) deficiency was extracted and coded. Coding inconsistencies in five patients were resolved by manual review of the reports and recoded in three patients.

Tumor grade was only considered for glandular adenocarcinomas, whereas the other tumor types were regarded as high grade and thus independent histopathological risk factors in themselves. Mismatch repair deficiency (dMMR) was recorded when expression of one or more of the MMR genes (MSH2, MLH1, MSH6, PMS2) was completely absent, i.e. patients with partial or reduced expression were regarded as MMR proficient (pMMR). Tumor budding (Bd) was only considered for glandular adenocarcinomas.

Based on the Danish national guidelines [1], the histopathological characteristics shown in Table 1 were defined as histopathological risk factors (HRF) for residual cancer in situ and/or in lymph nodes, and thus an indication for SBR when present in a malignant colorectal polyp specimen. The guidelines also stated that missing or non-conclusive parameters could pose an increased risk of residual cancer, in which case the indication for SBR should be discussed at a multidisciplinary team (MDT) meeting. These guidelines became available in 2018 as the first comprehensive Danish recommendations on management of malignant polyps. Although our study population spanned the years 2016–2020, we chose to define the HRF in accordance with these guidelines in order to apply consistent criteria for the presence or absence of HRF on the entire cohort.

Table 1.

Histopathological risk factors for increased risk of residual cancer and indication for subsequent surgery when present in a malignant colorectal polyp specimen, according to the Danish national guidelines [1]

Characteristic/Factor	High risk	Low risk
Tumor grade/type	• Medullary carcinoma • Undifferentiated carcinoma • Signet ring cell carcinoma • Mucinous carcinoma • High-grade glandular adenocarcinoma	• Low- or medium-grade glandular adenocarcinoma
pT category	pT > 1	pT = 1
Resection margin	R1 (≤ 1 mm)	R0 (> 1 mm)
Tumor budding	Bd2-3 (glandular adenocarcinoma)	Bd0-1 (glandular adenocarcinoma)
Lymphatic invasion	+	−
Venous invasion	+	−
Tumor depth	Kikuchi sm2-3, Haggitt 3–4	Kikuchi sm1, Haggitt 1–2

In some patients in the SBR and SV groups, the decision to proceed to, or refrain from, ‘completion’ resection was based on more than one local specimen. We considered a patient positive for a histopathological risk factor if this was reported in at least one of the patient’s local specimens. The resection margin, however, was an exception from this rule, as a safe margin may sometimes be obtained by a ‘redo’ local resection, e.g. an incomplete polypectomy followed by a full-thickness TEM. After manual review, we recoded 22 patients (all in SV group) with involved or uncertain margins and more than one specimen in SV to “no histopathological risk factor for margin overall” when the final local specimen had a safe margin. As expected, 16 of these patients had a TEM or other transanal resection, and one had an endoscopic full-thickness resection.

Research questions and statistical analysis plan

Our aim was to answer the following research questions:

1. Do management strategies for malignant colorectal polyps vary between Danish hospitals, and if so:

   • 1.1.
     Which patient- and hospital-related factors contribute to this variation?
   • 1.2.
     Do hospitals differ after adjustment for relevant covariates?

2. Can the between-hospital differences be ascribed to differences in the hospitals’ propensity for bowel resection in low-risk cases, or for surveillance in high-risk cases?

3. What information on histopathological risk factors was available in the local specimens to serve as basis for the decisions made?

4. What other information available at the multidisciplinary team meeting may potentially have influenced the treatment decision?

5. To what extent did histopathological risk factors reported in the local specimens predict metastatic lymph nodes or residual tumor in situ, and was there any concordance between the histopathological risk factors reported in the local specimens and the bowel specimens?

Treatment decisions and their associations with patient- and hospital-related factors

Baseline and summary data on management strategies across hospitals are presented as counts and percentages in tables and figures.

The relationship between management strategy and available patient- and hospital-related factors were analyzed by univariable and multivariable logistic regressions, including the following preselected covariates:

• Demographics, comorbidity: Age group in years (< 60, 60–74, ≥ 75), sex (female, male), Body Mass Index (BMI) category (< 30, ≥ 30), smoking status (never, former, active), alcohol consumption (≤ 14 units/week,> 14 units/week), heredity (known/none known), American Society of Anesthesiologists classification (ASA 1,2,3,4), Eastern Cooperative Oncology Group (ECOG) Performance Status Scale (PS 0, 1, 2, 3 or 4), Charlson Comorbidity Index (CCI 0,1,2,≥3).

• Socioeconomics: Household income group (highest, middle or lowest tertile stratified by 10-year age group), highest acquired education level grouped into 3 levels (primary school, short higher education (e.g. secondary school or vocational education), long higher education (e.g. Bachelor’s degree or higher) [17], cohabitation (living alone, cohabiting), occupation (employed, unemployed, retired), immigration status (Danish, Immigrant or descendant).

• Tumor location: (right colon, left colon, sigmoid/high rectum, low rectum, or rectum not otherwise specified (NOS))

• Structural factors: Treating hospital (anonymized Danish hospitals A-Q), discussion at a multidisciplinary team (MDT) meeting prior to choice of management strategy (yes, no).

The selected covariates are either well-known predictors of surgical risk and outcomes and thus relevant for treatment choices and characterization of case mix [13], or variables particularly relevant for the present purpose (e.g., discussion at an MDT meeting). The continuous variables age and BMI were categorized according to clinical practice: < 60 years = “young”, 60–74 = typical age at diagnosis, ≥ 75 = older patient; BMI ≥ 30 = obese.

Groups and subgroups for analysis

Groups (SBR and SV)

The two main study groups (SBR and SV) were included in an overall analysis of between-hospital variation. Following initial univariable analyses, any apparent between-hospital variation in management strategy was analyzed in an adjusted analysis including the above-mentioned covariates to elucidate whether the variation could be explained by differences in case mix.

Subgroups

Subsequently, we explored these differences further by separately analyzing the management choices in patients with and without known HRF in their local specimens, respectively. For that purpose, we defined the following four subgroups:

• SBR-HRF+: Patients in SBR group with HRF present (617 patients).

• SBR-HRF-: Patients in SBR group without documented HRF present (i.e. unclear indication for surgery) (294 patients).

• SV-HRF-: Patients in SV group without documented HRF present (563 patients).

• SV-HRF+: Patients in SV groups with HRF present (i.e. apparent deviation from the national guidelines) (714 patients).

We thus explored the decisions to proceed with SBR in patients without known histopathological risk factors (HRF), and the decisions to abstain from SBR despite known presence of HRF (i.e. SV).

Patients without known HRF (n = 857), and patients with known HRF (n = 1331) were thus analyzed separately by cross-tabulation and by regressing unclear indication or apparent deviation from national guidelines (i.e., SBR in HRF- and SV in HRF+, respectively) on the same covariates as above.

To evaluate the robustness of our findings we performed a post-hoc sensitivity analysis for all three multivariable analyses described above. The same models were used, but covariates with more than 5% missing data and/or p-values > 0.1 were excluded from analysis. All levels in the covariate “Hospital” were preserved.

The basis for treatment decisions: histopathology and other factors

To explore the basis for the treatment decisions, we first analyzed the distribution of the standard histopathological risk factors (present or absent, as well as missing/not reported) in each of the four subgroups.

Information on histopathological risk factors extracted from the pathology report could be either conclusive (i.e. explicitly stating whether the HRF was present or not), inconclusive (explicitly stating that presence of HRF could not be assessed by the pathologist), or missing (none of the former). The HRF- status of a patient (i.e., no known presence of an HRF) could therefore be based on conclusive (negative) information, inconclusive information or missing information, whereas HRF+ status could only be based on positive detection of a histopathological risk factor. This means that the HRF- status is only an indication as to whether a HRF was known to be present at the time of decision-making and not evidence of its absence.

Subsequently, we explored whether additional clinical or histopathological information available at the MDT meeting might potentially have influenced the treatment decisions, by comparing the frequencies in the four subgroups of the following: clinical tumor category > cT1, suspected lymph node (cN1, cN2) or distant (cM1) metastasis, MMR deficiency (dMMR), and nervous invasion.

Histopathological findings in the resection specimens

Finally, histopathological findings in local specimens and bowel specimens from the SBR group were compared. First, to assess to what extent presence of HRF in the local specimens predicted metastatic lymph nodes or residual tumor in situ, and second, whether there was concordance between the presence of HRF in local and bowel specimens from each patient.

Results

Baseline characteristics

The study cohort consisted of 2188 patients who had a malignant colorectal polyp removed endoscopically in Denmark during the years 2016–2020. The definitive management strategy was subsequent bowel resection (SBR) in 911 patients (42%) and surveillance (SV) in 1277 (58%).

The commonest characteristics were being male, in the age group 60–74 years, with a BMI under 30, with low comorbidity or performance status, co-living, retired from work and with short higher education (see Table 2).

Table 2.

Patient characteristics of patients with malignant colorectal polyps by management strategy

Variable / Management strategy	Surveillance (% of non-missing)	Subsequent bowel resection (% of non-missing)	Total (% of non-missing)	Missing (%)
NUMBER (%)	1277 (58.4)	911 (41.6)	2188 (100.0)	0 / 2188 (0.00)
GENDER, n (%)				0 / 2188 (0.00)
Female	491 (38.4)	426 (46.8)	917 (41.9)
Male	786 (61.6)	485 (53.2)	1271 (58.1)
AGE MEDIAN, years	71	67	69	0 / 2188 (0.00)
AGE GROUP, n (%)				0 / 2188 (0.00)
< 60 years	185 (14.5)	190 (20.9)	375 (17.1)
60–74 years	705 (55.2)	589 (64.7)	1294 (59.1)
≥75 years	387 (30.3)	132 (14.5)	519 (23.7)
BMI CATEGORY, n (%)				156 / 2188 (7.13)
< 30	887 (78.1)	683 (76.2)	1570 (77.3)
≥30	249 (21.9)	213 (23.8)	462 (22.7)
SMOKING STATUS, n (%)				387 / 2188 (17.69)
Never	376 (39.1)	377 (44.9)	753 (41.8)
Active	144 (15.0)	132 (15.7)	276 (15.3)
Former	441 (45.9)	331 (39.4)	772 (42.9)
WEEKLY ALCOHOL USE, n (%)				368 / 2188 (16.82)
0–14 units	807 (83.5)	726 (85.1)	1533 (84.2)
>14 units	160 (16.5)	127 (14.9)	287 (15.8)
CO-HABITATION STATUS, n (%)				2 / 2188 (0.09)
Co-living	863 (67.6)	650 (71.5)	1513 (69.2)
Living alone	414 (32.4)	259 (28.5)	673 (30.8)
INCOME TERTILE BY AGE, n (%)				87 / 2188 (3.98)
Lowest tertile	417 (34.2)	306 (34.7)	723 (34.4)
Middle Tertile	414 (33.9)	284 (32.2)	698 (33.2)
Highest Tertile	389 (31.9)	291 (33.0)	680 (32.4)
ETHNICITY, n (%)				2 / 2188 (0.09)
Danish	1219 (95.5)	871 (95.8)	2090 (95.6)
Immigrant OR descendant	58 (4.5)	38 (4.2)	96 (4.4)
HIGHEST ACQUIRED EDUCATION, n (%)				34 / 2188 (1.55)
Primary school	383 (30.6)	233 (25.8)	616 (28.6)
Short higher education following primary	644 (51.4)	450 (49.9)	1094 (50.8)
Long higher education following primary	225 (18.0)	219 (24.3)	444 (20.6)
OCCUPATION, n (%)				1 / 2188 (0.05)
Affiliated to labour market	306 (24.0)	312 (34.3)	618 (28.3)
Not affiliated to labour market *	33 (2.6)	26 (2.9)	59 (2.7)
Retirement or early retirement	938 (73.5)	572 (62.9)	1510 (69.0)
HEREDITY, n (%)				715 / 2188 (32.68)
None known	678 (89.4)	590 (82.5)	1268 (86.1)
Known	80 (10.6)	125 (17.5)	205 (13.9)
CCI** CATEGORY, n (%)				0 / 2188 (0.00)
0	707 (55.4)	600 (65.9)	1307 (59.7)
1	219 (17.1)	126 (13.8)	345 (15.8)
2	177 (13.9)	110 (12.1)	287 (13.1)
≥3	174 (13.6)	75 (8.2)	249 (11.4)
ASA***, n (%)				140 / 2188 (6.40)
1	274 (24.0)	252 (27.8)	526 (25.7)
2	591 (51.7)	512 (56.6)	1103 (53.9)
3 and 4 merged****	278 (24.3)	141 (15.5)	419 (20.5)
PS*****, n (%)				156 / 2188 (7.13)
PS 0	785 (68.3)	711 (80.5)	1496 (73.6)
PS 1	237 (20.6)	139 (15.7)	376 (18.5)
PS 2	95 (8.3)	26 (2.9)	121 (6.0)
PS 3 or 4	32 (2.8)	7 (0.8)	39 (1.9)
TUMOR LOCATION, n (%)				0 / 2188 (0.00)
Right Colon	89 (7.0)	149 (16.4)	238 (10.9)
Left Colon	40 (3.1)	36 (4.0)	76 (3.5)
Low Rectum	427 (33.4)	140 (15.4)	567 (25.9)
Sigmoid or High Rectum	639 (50.0)	561 (61.6)	1200 (54.8)
Rectum Not Otherwise Specified	82 (6.4)	25 (2.7)	107 (4.9)
HOSPITAL, n (%)				0 / 2188 (0.00)
A	12 (0.9)	9 (1.0)	21 (1.0)
B	22 (1.7)	39 (4.3)	61 (2.8)
C	18 (1.4)	44 (4.8)	62 (2.8)
D	9 (0.7)	60 (6.6)	69 (3.2)
E	27 (2.1)	65 (7.1)	92 (4.2)
F	56 (4.4)	38 (4.2)	94 (4.3)
G	64 (5.0)	53 (5.8)	117 (5.3)
H	86 (6.7)	35 (3.8)	121 (5.5)
I	93 (7.3)	42 (4.6)	135 (6.2)
J	83 (6.5)	55 (6.0)	138 (6.3)
K	89 (7.0)	53 (5.8)	142 (6.5)
L	97 (7.6)	50 (5.5)	147 (6.7)
M	137 (10.7)	48 (5.3)	185 (8.5)
N	87 (6.8)	102 (11.2)	189 (8.6)
O	135 (10.6)	54 (5.9)	189 (8.6)
P	108 (8.5)	88 (9.7)	196 (9.0)
Q	154 (12.1)	76 (8.3)	230 (10.5)
DISCUSSED AT A MULTIDISCIPLINARY MEETING, n (%)				7 / 2188 (0.32)
No	158 (12.4)	103 (11.3)	261 (12.0)
Yes	1113 (87.6)	807 (88.7)	1920 (88.0)

*unemployed, welfare payment, education, other**CCI Charlson Comorbidity Index, ***ASA American Society of Anesthesiologists classification, **** Cells with fewer than 5 patients are not shown on their own. ASA 3 and 4 are therefore merged in one cell, *****PS Eastern Cooperative Oncology Group Performance Status Scale *

Data completeness was high, except for information on heredity, smoking and alcohol consumption where up to 33% were missing.

Current management strategies in Danish hospitals

Between-hospital variation in overall management strategy

The national crude percentage of Danish patients managed by SV was 58.4%, while 41.6% of patients had a SBR, as indicated with a blue and red dashed line in Fig. 2. These proportions varied markedly between hospitals, however, as shown in Fig. 2, where hospitals are ranked by total number of malignant colorectal polyps from A (lowest) to Q (highest). The ratio between SBR and SV varied from 9:60 (0.15) to 137:48 (2.85).

Fig. 2.

Management strategies for malignant colorectal polyps in anonymized Danish hospitals ranked by total number of malignant colorectal polyps in ascending order (A-Q)

Factors associated with overall management strategy

Univariable regression analysis (Table S1 in supplementary material) showed that male sex, age ≥ 75 years, Charlson Comorbidity Index (CCI) > 0, ASA > 2, PS > 0, former smoker status, being retired from work, and tumor location in the rectum or left colon were significantly associated with choosing SV as management strategy (i.e. OR < 1). On the other hand, hereditary predisposition to colorectal cancer, certain hospital departments (C, D and E) and long higher education were statistically significantly associated with SBR management (i.e. OR > 1).

In the multivariable analyses (Table S2 in supplementary material) male sex, age > 75 years, CCI 1 and ≥ 3, PS > 0, and tumor location in the rectum or left colon were still significantly associated with SV (i.e. OR < 1). Long higher education and hospital department (C, D and E) were still significantly associated with SBR (i.e. OR > 1). The model included all the covariates listed in the Methods section. Table 3 shows only the statistically significant covariates.

Table 3.

Multivariable logistic regression analysis of patient characteristics and hospital variables significantly associated with choice of surgery as management strategy*

Management strategy group reference: subsequent bowel resection
Variables	OR	95% CI	P-value
GENDER, ref.** female
Male	0.698	0.558–0.875	0.002
AGE GROUP, ref.** age < 60
Age ≥ 75	0.423	0.269–0.665	0.0001
CCI***, ref.** CCI 0
CCI 1	0.716	0.527–0.974	0.033
CCI ≥ 3	0.655	0.447–0.96	0.03
PS****, ref.** PS0
PS 1	0.697	0.507–0.959	0.027
PS 2	0.362	0.201–0.651	0.001
PS 3/4	0.305	0.11–0.847	0.023
TUMOR LOCATION, ref** Right Colon
Left Colon	0.388	0.204–0.743	0.004
Low Rectum	0.127	0.085–0.19	0.0001
Sigmoideum/High Rectum	0.431	0.301–0.617	0.0001
Rectum Not Otherwise specified	0.183	0.099–0.338	0.0001
HOSPITAL, ref.** A*****
C	11.848	3.19-44.008	0.0001
D	8.482	2.411–29.837	0.001
E	4.178	1.326–13.166	0.015
HIGHEST ACQUIRED EDUCATION, Ref.** Primary school
Long higher education	1.59	1.143–2.21	0.006

* Based on the full model but only the statistically significant variables are illustrated in the table, **Reference, ***CCI Charlson Comorbidity Index, ****PS Eastern Cooperative Oncology Group Performance Status Scale, *****Hospital A had proportions of SBR and SV closest to the Danish mean and was selected as reference

Between-hospital variation in management strategy by known presence of HRF

In principle, any deviation of a particular hospital’s relative proportions of SBR and SV from the national average may be the result of either a propensity for SBR in HRF- patients (i.e. potential “overtreatment”) or a propensity for SV in HRF+ patients (potential “undertreatment”), or a combination. We therefore analyzed these proportions separately for the HRF- and HRF+ groups.

Figure 3 shows the percentage choosing surgery with unclear indications (SBR in HRF- patients), or apparent deviation from national guidelines (SV in HRF+ patients), by HRF group in Danish hospitals. Low hospital volume seems associated with a tendency to operate on a larger proportion of patients, irrespective of HRF group, whereas a larger volume seems particularly associated with choosing SV in many HRF+ patients.

Fig. 3.

Percentage of patients undergoing bowel resection with unclear indication or not undergoing surgery despite presence of histopathological risk factors, in Danish hospitals. Red bars: percentage of patients receiving subsequent bowel resection in the group without known histopathological risk factors. Blue bars: percentage of patients receiving surveillance in the group with known histopathological risk factors

Factors associated with unclear indication or apparent deviation from national guidelines by risk group

The associations of patient- and hospital-related covariates with deviation from the recommended treatment strategy or unclear indication for surgery were analyzed separately for the HRF- and HRF+ groups as sub-groups, to elucidate between-hospital practice variation.

Patients with known presence of histopathological risk factors (HRF+)

In the HRF+ group the univariable analysis (Table S3 in the supplementary material) showed that male sex, age ≥ 75 years, CCI > 0, ASA > 1, PS > 0, active or former smoker status and tumor location in the rectum were significantly associated with SV as definitive treatment despite presence of histopathological risk factors in the local resection specimens. Living alone and retirement were also significantly associated with SV. Hospital departments I and M were strongly associated with SV and therefore deviating from the national guidelines. Hereditary predisposition to colorectal cancer, short and long higher education, and hospital department D were, on the other hand, strongly associated with adhering to the guidelines, i.e. with an OR significantly < 1.

Age ≥ 75 years, CCI 1 and ≥ 3, PS > 0, and tumor location in the rectum remained as statistically significant predictors of SV in the multivariable analysis. On the other hand, long higher education was significantly associated with SBR i.e. adhering to the guidelines (Table 4, which shows only significant associations, or Table S4 in Supplementary Material). No significant between-hospital variation was found in the multivariable analysis.

Table 4.

Multivariable logistic regression analysis of patient characteristics and hospital variables significantly associated with choice of management strategy in surveillance group with histopathological risk factors and subsequent bowel resection group without known histopathological risk factors *

Management strategy group	SV with HRF			SBR without HRF
Variables	P-value	OR	95% CI	P-value	OR	95% CI
GENDER, ref.** female
Male	Non-significant			0.004	0.543	0.358–0.823
AGE GROUP, ref.** age < 60
Age ≥ 75	0.014	2.155	1.172–3.964	0.002	0.279	0.123–0.629
CCI***, ref.** CCI 0
CCI 1	0.047	1.496	1.006–2.226	Non-significant
CCI ≥ 3	0.004	2.059	1.252–3.386	Non-significant
PS****, ref.** PS 0
PS1	0.016	1.670	1.099–2.539	Non-significant
PS2	0.0001	4.062	1.884–8.759	Non-significant
PS3 and 4	0.002	14.039	2.686–73.391	Non-significant
SMOKING STATUS, ref.** Never
Active	Non-significant			0.02	2.143	1.13–4.062
TUMOR LOCATION base, ref** Right Colon
Left Colon	Non-significant			0.015	0.279	0.1-0.781
Low Rectum	0.0001	6.101	3.443–10.81	0.0001	0.051	0.025–0.106
Sigmoideum High Rectum	Non-significant			0.0001	0.276	0.159–0.478
Rectum Not Otherwise Specified	0.015	2.743	1.214–6.196	0.0001	0.05	0.014–0.175
HOSPITAL, ref.** A
C	Non-significant			0.0001	54.218	6.328-464.508
D	Non-significant			0.008	11.124	1.883–65.727
I	Non-significant			0.031	6.578	1.189–36.39
HIGHEST ACQUIRED EDUCATION, ref.** Primary school
Long higher education	0.042	0.636	0.411–0.985	Non-significant

* Based on the full model but only the statistically significant variables are illustrated in the table, **Reference, ***CCI Charlson Comorbidity Index, ****PS Eastern Cooperative Oncology Group Performance Status Scale

Patients without known presence of histopathological risk factors (HRF-)

In the HRF- group the univariable analysis (Table S5 in supplementary material) showed that BMI > 30, hereditary predisposition to colorectal cancer and management at hospital departments C and D were significantly associated with SBR (OR 9.9 and 7.7, respectively). Age ≥ 75 years, male sex and tumor location in the rectum or left colon were significantly associated with SV, i.e. OR significantly < 1.

In the multivariable analysis, active smoking and hospital departments C, D and I (OR 54, 11 and 6, respectively) were significantly associated with SBR. Age ≥ 75, male sex and tumor location in the rectum or left colon remained significantly associated with SV (Table 4, which shows only significant associations, or Table S6 in Supplementary Material).

Results of the post-hoc sensitivity analyses for all three multivariable analyses were consistent with those of the full models, as shown in Table S7, S8 and S9 in the Supplementary Material.

The basis for decisions – histopathological risk factors

Histopathological risk factors reported in the local specimens

The number and distribution of HRF reported in local specimens in each of the four management strategy groups are shown in Tables S10 and S11 in Supplementary Material. In around half the patients (51.4%), 1–2 HRF were reported in their local specimens, whereas 39.2% had no reported HRF. More than 2 HRF were reported in 9.5% overall. Involved resection margin (31%) and tumor depth (26%) were the most commonly reported HRF.

Conclusive information regarding histopathological risk factors in local specimens

We analyzed the actual informational content of the histopathological data for each HRF in each patient and categorized it as either conclusive, inconclusive, or missing, as defined above. Figure 4 shows the number of HRF with conclusive information in each of the management groups. It is evident that in the SBR-HRF- group, two-thirds of the patients had conclusive information on only one of the eight histopathological HRF analyzed here, whereas in the other groups, conclusive information was available on most or all HRF in most patients.

Fig. 4.

Conclusive information on histopathological risk factors in local specimens (0 to 8), by management strategy group (SV-HRF+, SV-HRF-, SBR-HRF-, SBR-HRF+)

Detailed information regarding the distribution of conclusive information regarding HRF in each management group is shown in Supplementary Table S12. For most HRF in the SBR-HRF- group, conclusive information regarding HRF was present in 25–30% and missing in 70–75%. The only HRF with conclusive information in almost 100% of all four groups was tumor type.

Additional histopathological and clinical information present at the MDT meeting

Roughly 90% of cases had been discussed in an MDT meeting, with no significant differences between groups in univariable and multivariable analyses, as shown in tables S4 and S6 in Supplementary Material.

Overall, the proportion of patients with a cT category > cT1 was unexpectedly large, and larger in the SBR group (cT0, 4%; cT1, 60%; cT2-4, 21%; cTx, 15%) than in the SV group (cT0, 6%; cT1, 70%; cT2-4, 10%; cTx, 13%). This was particularly pronounced in the SBR-HRF- group (cT2, 19%; cT3, 12%; cT4, 1.4%). Overall, 32% of patients in the SBR-HRF- group were staged as cT2+, as compared with 5–16% in the other three groups.

Also, a significantly higher frequency of cN + was found in the SBR-RF- group (12.6%) than in the other groups (1–6%), while cM1 was only slightly more common (3.7% vs. 1.8–2.6%). MMR deficiency was also more frequent (11.2%) in the SBR-HRF- group than in the other groups (4–6%). There was no association, however, between cN + and dMMR status. Nervous invasion was rare (< 1%) and not found in the SBR-HRF- group.

The finding of cT2 + in a notable proportion of patients led us to a post-hoc sub-analysis of the uni- and multivariable regression analyses with inclusion of the cT, cN and cM categories as covariates to explore their association with management strategy.

Regarding overall predictors for SBR, both univariable (Table S13 in supplementary material) and multivariable analyses (Table S14 in supplementary material) showed that cT2 + and cN+ status were significantly associated with SBR, whereas cM was not. This extended model analysis showed otherwise only minor deviations from the findings mentioned above.

We also repeated the analyses in the SBR-HRF- and SV-HRF+ subgroups.

In the multivariable analysis of the SBR-HRF- group, cT2, cT3, cT4 and cN1 were significantly associated with SBR. Comparing with the prior model we found that long higher education and Hospital B were now significantly associated with SBR, whereas PS2 was associated with not choosing SBR. (Uni- and multivariable analysis tables S15 and S16 in supplementary material).

In the SV-HRF+ group, cT3 and cN1 were significantly associated with abstaining from SV in the uni- and multivariable analyses. In the multivariable analysis male gender was associated with SV, and Hospital D was associated with not choosing SV. Furthermore, CCI was no longer associated with SV. (Uni- and multivariable analysis tables S17 and S18 in supplementary material)

Overall, the inclusion of the clinical TNM stages in the regression model resulted in male gender, performance status, higher education and certain hospitals remaining as significant predictors in both the SV-HRF + and SBR-HRF- group.

Findings in the bowel resection specimens

In the two SBR groups, the histopathological findings in the local specimens could be compared with those of the subsequent bowel resection specimen.

Lymph node metastases

Overall, lymph node metastases were reported in 19% of the SBR specimens, with similar frequency in the HRF + and HRF- groups (18% versus 21% respectively).

Residual tumor in situ

Residual tumor in situ was regarded as present if the pathology report of the bowel specimen contained a code for a carcinoma, as opposed to the code “no residual tumor”. This was found in 24% of the SBR specimens. Notably, this was the case in a much larger proportion of the SBR-HRF- patients (48%) than in the SBR-HRF+ group (13%).

Overall 63% of bowel specimens were cancer free, i.e. contained neither in situ tumor nor lymph node metastases. The proportion of bowel specimens without residual disease was markedly higher in SBR-HRF+ patients (72%) than in the SBR-HRF- group (45%).

Concordance of histopathological risk factors in local specimens and bowel specimens

We compared the reported HRF in the bowel specimens with the local specimens without known HRF (SBR-HRF- group). HRF were present in close to half (45%) of the subsequent bowel resection specimens, despite none being reported in the local specimens. The most commonly identified HRF were: > pT1 (32%), venous invasion (13%) and deep tumor invasion (10%), as illustrated in table S19 in Supplementary Material.

Concordance of preoperative clinical staging and findings in bowel specimens

Overall, the final pT category of the SBR specimens was pT1 in 74%, pT2 in 12%, pT3 in 8%, and pT4 in < 1%. Notably, 32% of SBR-HRF- specimens were pT2-4 as compared with only 15% of SBR-HRF+, furthermore, 37% of SBR patients staged as cT2 + had bowel specimens with pT1 tumors. This was most common in the HRF+ group (SBR-HRF-, 32%; SBR-HRF+, 43%). On the other hand, 87.5% of SBR patients staged as cT1 had pT1 tumors in their bowel specimens. The cT category was also related to residual disease in situ (cT0 or cTx, 24%; cT1, 14%; cT2+, 53%), but to a lesser degree to lymph node involvement (cT0 or cTx, 21%; cT1, 17%; cT2+, 26%). In the cN category 50% of cN + SBR patients had pN0 tumors, while 82% of cN0 SBR patients had pN0 tumors.

Discussion

Choosing the right management strategy once a malignant focus is found in an endoscopically resected polyp continues to be a dilemma. There is a growing acknowledgement worldwide that too many patients undergo ‘completion’ resection producing bowel specimens without residual cancer [7, 8], and that the risk of recurrence after local treatment is relatively small, even when histopathological risk factors are found in the specimen [5]. Better tools for risk stratification are urgently needed.

We analyzed the practice variation between Danish hospitals and sought to elucidate its underlying causes, with the aim of identifying areas for improvement. We confirmed substantial between-hospital practice variation, a proportion of cancer-free bowel resection specimens of almost two thirds, and a large proportion of missing pathology information, but the detailed information available enabled us to contribute additional insight into the decision-making when malignant polyps are encountered.

Between-hospital practice variation

Between-hospital practice variation in management strategy was evident in our study, even after adjustment for plausible patient-related covariates, and also suggested a relationship between hospital volume and treatment choices. The differences may thus be more systematic than random and may reflect variation in clinician attitudes rather than case mix. Similar unexplained variation has been shown for other treatment choices like permanent stomas or refraining from curative treatment [14, 19].

Factors associated with management strategy

Our study showed that overall, male sex, age ≥ 75 years, CCI > 0, PS > 0, and distal tumor location were significantly associated with SV as management strategy. This was in accordance with the largest population based study with multivariable analyses that likewise showed that men were more likely to undergo polypectomy alone compared with women, patients with Charlson-Deyo score of 0 were less likely to undergo polypectomy alone compared with those with a higher score, and that tumors in the left colon were more likely to be managed by polypectomy alone than tumors in the right colon [18]. However, patients above and below 65 years of age were equally likely to undergo polypectomy alone.

Moreover, we found that higher education was significantly associated with SBR as management strategy. Interestingly, a recent multicenter cross-sectional study by Dekkers et al. on patients undergoing local resection as initial treatment for a malignant polyp, found that patients with higher education were more often actively involved in medical decision-making and more often proceeded to SBR [19]. Education level may thus play a role.

The most recent systematic review and meta-analysis on predictors of management strategy for malignant colorectal polyps, however, included 24,609 patients across 15 studies [20] and showed no significant differences in patient characteristics (age and gender) in respect to which management was chosen following polypectomy for malignant colorectal polyps.

Unclear indications for surgery and apparent deviation from the guidelines

The factors associated with SBR without known histopathological risk factors were in line with the results of Zammit et al. [21]. In their multivariable population-based analysis, SBR patients tended to be younger and have right-sided malignant colorectal polyps. Furthermore, we found active smoking to be associated with SBR.

On the other hand, our study showed that male sex, older age, comorbidity, higher performance status and tumor location in the rectum were significantly associated with a more conservative approach, again in agreement with Zammit et al.’s results [21]. This was also in line with a recent population-based Danish study on patients with all tumor stages [22], which showed that inter alia high age, poor performance status, distant metastases and being underweight were predictors for non-curative treatment as the management strategy in patients with potentially curable colorectal cancer. Old age and poor performance status were also predictors for refraining from adjuvant treatment.

The basis for decisions

Seeking explanations for the marked practice variation we tried to elucidate the basis for the treatment decisions in the four groups of patients. We found that while conclusive information on most HRF was available for most patients in three groups – SBR-HRF + and both SV groups – it was missing in the vast majority of SBR-HRF- patients. Furthermore, in most cases information on one or more HRF was simply missing rather than being explicitly encoded as not assessable.

Missing information on histopathological risk factors in malignant colorectal polyps has been reported before [7–9, 18, 23–25]. Zammit et al. found that 71.1% of pathology reports had at least one missing standard parameter [10]. Our most frequently missing histopathological risk factor was tumor budding (missing in 57%) and tumor depth (Haggitt/Kikuchi level) (47%) [20, 21].

Missing crucial information on histopathological risk factors complicates decision-making and patient counseling, and both the patient and the surgeon may perceive surgery as the safest choice if the risks of residual tumor and/or lymph node metastasis are unpredictable. This is not necessarily at variance with national guidelines, as the guidelines in force during the study period recommended considering SBR if one or more histopathological risk factors could not be assessed. Our finding that the frequency of lymph node metastasis in the resection specimens was similar in the SBR-HRF- and SBR-HRF+ groups would support the wisdom of that notion.

However, owing to the availability of the preoperative clinical TNM category (cTNM) in the national database we also explored the relationship between cT, cN and cM categories and the decision to proceed to SBR. Surprisingly, we found a large proportion of patients with clinical T category > cT1 even though all the SBR patients had by definition had a local resection of a malignant polyp first. Less surprisingly, however, we found that cT2 + and cN+ categories were significant predictors of SBR, particularly in the SBR-HRF- group, suggesting that clinical suspicion of more advanced disease may have been a driver in the decision to operate. This might potentially explain at least some of the many SBRs with cancer-free bowel specimens. It is well known from other studies [26] that image-based cT and cN are poor predictors of pT and pN categories with a propensity to overstage, and our comparisons of cTNM with findings in the bowel specimens confirmed that.

Residual disease and histopathological risk factors in the bowel specimens

Our study showed that 63% of patients managed with SBR had cancer free bowel resection specimens, meaning that oncologically, their operation was ostensibly unnecessary. Also, overall residual disease in the bowel specimens was much more common in the HRF- group than in the HRF+ group, counterintuitive to what we would expect. This was mainly due to a higher percentage of residual disease in situ in the SBR-HRF- group, despite the absence of reported resection margin involvement. On the other hand, this group had the highest frequency of cT2+.

When we compared local and bowel resection specimens from the same patients, we found histopathological risk factors in 45% of the bowel specimens in the SBR-HRF- group, histopathological risk factors that had gone undetected, or at least unreported, in the corresponding local specimens. This is probably part of the explanation for the high frequency of residual disease in the SBR-HRF- group, but it also emphasizes the risk associated with missing histopathological information in local specimens. Similar lack of information and discordance between local and resection specimens have been reported before by Zammit [11] and Gimon [20].

Strengths and limitations

Among the strengths of this study were the high quality and data completeness of the national Danish databases and the reliable identification of patients across all datasets by the unique personal ID number. Also, as opposed to previous database studies based on limited pathology data, we had access to the complete pathology dataset for each patient, which enabled us to collect all histopathological information that had been available for the MDT decision-making and subsequent discussion with the patient. Furthermore, inconsistencies in the encoding could often be resolved by accessing the full free text of the pathology reports.

However, our analysis of the basis for decisions is less than complete. We did not have access to the patients’ medical records, imaging, or histopathology slides, and we do not know what was discussed at the MDT meetings. We did, however, include all accessible information encompassing patient-related, disease-related, and hospital-related predictors. The cT, cN and cM categories were available, and suggest that preoperative imaging (itself a somewhat poor predictor of risk) may also have influenced the decisions in many patients. Finally, we have no information on how patients were informed, and to what degree shared decision making was practiced [27, 28].

Our definitions of the HRF were based on the first comprehensive Danish guidelines on management of malignant polyps which became available from 2018. They are consistent with most European guidelines and were in force until the end of 2025. Although our cohort spanned 2016–2020, we applied the same HRF criteria to all patients to classify the histopathological basis for decisions consistently. Since we have no information on the timing of knowledge dissemination and local implementation of the new guidelines, we regard this consistent approach as the best possible solution. Finally, it should be emphasized that our goal was to provide a systematic and consistent exploration of the basis for treatment decisions, not to judge whether hospitals complied with national guidelines or not.

Regarding the statistical methods, we cannot rule out that the relatively large number of covariates and significance tests, as well as the potential for collinearity among the covariates, may have influenced the confidence intervals and p-values. Our main findings, however, proved robust in the sensitivity analyses and were consistent with the patterns visible in the figures.

Perspectives and potential for improvement

Decision-making regarding management of malignant colorectal polyps is complex. Guidelines describe that management strategies should be based on the presence of histopathological risk factors, but as shown here, pathology information is very often incomplete. Our study furthermore revealed that other factors, including cTNM categories, also play a part in the decision-making process, but we also confirmed the inaccuracy and overstaging propensity of clinical staging. Moreover, between-hospital practice variation in management strategy was found despite adjustment for plausible patient- and hospital-related covariates. Finally, hospitals with larger patient volumes (and experience) tended to have a more conservative approach. The decision-making process is therefore multifactorial and depends on the entire MDT as well as on the patient.

This suggests that better training and cooperation within the MDT as a whole, and communication skills training in shared decision making itself may potentially improve both the basis for the decisions and the decisions themselves [29]. Recent meta-analyses have provided more realistic and less alarming risk estimates for malignant polyps, and colorectal surgeons need to be aware of this [5, 30]. Increased focus on updated guidelines and quality monitoring from scientific societies and national authorities have improved outcomes in many other areas and would be appropriate measures as well. Finally, we believe that if the real risks of the various treatment options were presented to the patients in a shared decision-making framework, potentially supported by decision aids, more patients would choose wisely for themselves, incorporating their personal preferences [31, 32].

Conclusion

We found marked between-hospital practice variation in management strategies for malignant colorectal polyps, even after adjusting for potential confounders. Overall, 63% of bowel resection specimens were cancer-free. Among the potential explanations were missing or incomplete pathology data and suspicion of more advanced disease based on inaccurate clinical staging. We suggest better training and collaboration of the multidisciplinary team to improve shared decision making in patients with endoscopically removed malignant colorectal polyps.

Abbreviations

SV

Surveillance

SBR

Subsequent segmental bowel resection

NCBD

National Bowel Cancer Database

DCCG

Danish Colorectal Cancer Group

RKKP

the national clinical quality control program

CRC

colorectal cancer

NPD

National Pathology Database

DST

Statistics Denmark

CPR

unique personal ID number

CELS

Combined endoscopic-laparoscopic surgery

TEM

transanal endoscopic microsurgery

TEO

transanal endoscopic operation

MMR

mismatch repair

dMMR

Mismatch repair deficiency

pMMR

MMR proficient

HRF

Histopathological risk factors

Bd

Tumor budding

SBR-HRF+

Patients in SBR with HRF present

SBR-HRF-

Patients in SBR without HRF present

SV-HRF-

Patients in SV without HRF present

SV-HRF+

Patients in SV with HRF present

BMI

Body Mass Index

ASA

American Society of Anesthesiologists classification

ECOG

Eastern Cooperative Oncology Group

PS

Performance Status Scale

CCI

Charlson Comorbidity Index

NOS

Not otherwise specified

MDT

Multidisciplinary team

Authors’ contributions

H.J.W. and H.B.R. conceptualised and designed the work. H.J.W. acquired the data. H.J.W. and H.B.R. performed the analyses. H.J.W., H.B.R., K.D.S, A.E., L.L. interpreted the data, and have drafted or substantively revised the work. HJW and H.B.R. prepared Figs. 1, 2, 3 and 4. All authors reviewed the manuscript. All authors reviewed the change reports before re-submitting the manuscript.

Funding

Open access funding provided by University of Southern Denmark. This study was funded by the University of Southern Denmark; the Surgical Department at Lillebaelt University Hospital of Southern Denmark, Vejle; Danish Cancer Research Foundation, and the Research Fund at Lillebaelt University Hospital.

Data availability

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

Declarations

Ethics approval and consent to participate

Not applicable, as this is a register-based cohort study.

Competing interests

The authors declare no competing interests.