Endoscopy 2019; 51(08): 733-741
DOI: 10.1055/a-0919-4803
Original article
© Georg Thieme Verlag KG Stuttgart · New York

Risk of post-colonoscopy colorectal cancer in Denmark: time trends and comparison with Sweden and the English National Health Service

Lasse Pedersen
1  Department of Surgical Gastroenterology, Aalborg University Hospital, Aalborg, Denmark
,
Roland Valori
2  Department of Gastroenterology, Gloucestershire Hospitals, Gloucester, United Kingdom
,
Inge Bernstein
1  Department of Surgical Gastroenterology, Aalborg University Hospital, Aalborg, Denmark
,
Karen Lindorff-Larsen
3  Nordsim: Center for Skills Training and Simulation, Aalborg University Hospital, Aalborg, Denmark
,
Charlotte Green
4  Department of Emergency, Aarhus University Hospital, Aarhus, Denmark
,
Christian Torp-Pedersen
5  Department of Health Science and Technology, Aalborg University, Aalborg, Denmark
› Author Affiliations
Further Information

Corresponding author

Lasse Pedersen, MD
Department of Surgical Gastroenterology
Aalborg Universitetshospital
Hobrovej 18-22
Aalborg 9100
Denmark   

Publication History

submitted: 16 November 2018

accepted after revision: 10 April 2019

Publication Date:
07 June 2019 (eFirst)

 

Abstract

Background The post-colonoscopy colorectal cancer (PCCRC) rate is a key quality indicator for colonoscopy. Previously published PCCRC rates have been difficult to compare owing to differences in methodology. The primary aim of this study was to compare Danish PCCRC rates internationally and to calculate Danish PCCRC rates using the World Endoscopy Organization (WEO) consensus method for future comparison. The secondary aim was to identify factors associated with PCCRC.

Methods National registries were used to examine the risk of PCCRC. The Danish 3-year rate of PCCRC (PCCRC-3yr) was calculated using previously published methods from England, Sweden, and the WEO. Poisson regression analysis was performed to identify factors associated with PCCRC.

Results The Danish PCCRC-3yr was significantly higher than the rate in the English NHS (relative risk [RR] 1.12, 95 % confidence interval [CI] 1.05 – 1.19) and Sweden (RR 1.15, 95 %CI 1.06 – 1.24). The Danish PCCRC-3yr based on the WEO consensus method fell from 22.5 % in 2001 to 7.9 % in 2012. The multivariable Poisson regression model found PCCRC to be significantly associated with diverticulitis (RR 3.25, 95 %CI 2.88 – 3.66), ulcerative colitis (RR 3.44, 95 %CI 2.79 – 4.23), hereditary cancer (age < 60 years: RR 7.39, 95 %CI 5.77 – 9.47; age ≥ 60 years: RR 3.81, 95 %CI 2.74 – 5.31), and location in the transverse (RR 1.57, 95 %CI 1.28 – 1.94) and ascending colon (RR 1.85, 95 %CI 1.64 – 2.08).

Conclusions The PCCRC-3yr was higher in Denmark than in comparable countries. Differences in colonoscopist training, background, and certification are possible contributing factors. A review of colonoscopist training and certification in Denmark, and continuous audit and feedback of colonoscopist performance may reduce PCCRC-3yr.


#

Introduction

The incidence of colorectal cancer (CRC) in Denmark is among the highest in the world [1], and with almost 5000 cases annually, it has become the third most common cancer in Denmark [2]. The preferred diagnostic procedure is colonoscopy, either for symptomatic patients or as a follow-up screening test because it can both diagnose cancer and remove pre-cancerous lesions. Unfortunately, CRC is known to occur after a negative colonoscopy (a colonoscopy that does not diagnose cancer). CRC occurring after a negative colonoscopy is known as interval or post-colonoscopy colorectal cancer (PCCRC) [3]. For benchmarking purposes, the World Endoscopy Organization (WEO) recommends a time frame of PCCRC occurring within 3 years of a colonoscopy (PCCRC-3yr) [4] [5].

Until recently, there has been no universally accepted method of calculating the PCCRC-3yr, and rates vary considerably depending on the method used to calculate them [4] [6] [7] [8] [9]. Danish registries enable calculation of the PCCRC-3yr using the different methodologies used previously to calculate the PCCRC-3yr, including those of the English National Health Service (NHS) and Sweden. This allows a direct comparison between colonoscopy services at a national level.

The primary aim of this study was to evaluate the Danish PCCRC-3yr and compare it with the English NHS and the Swedish health system. The Danish PCCRC-3yr based on the WEO method is presented for future reference [5]. The secondary aim was to investigate factors associated with PCCRC.


#

Methods

Data sources

Data for this population-based cohort study were obtained from two primary sources: the Danish Cancer Registry and the Danish National Patient Registry [10] [11]. The Danish Cancer Registry contains information about every Danish cancer event with mandatory reporting since 1987. The data include date of diagnosis, histology, stage (Dukes classification until 2003, TNM classification from 2004), and International Classification of Disease 10 (ICD-10). Data dating from 1978 – 2003 were originally coded according to ICD-7, but have subsequently been recoded to ICD-10 by the Danish Cancer Registry. The Danish National Patient Registry contains information about every contact with the public Danish health service regarding administrative data, diagnosis, treatments, procedures, and examinations. Data from the Danish National Prescription Registry were further used to calculate the Charlson Comorbidity Index [12] [13]. It is possible to link individuals between the Danish Cancer Registry, the Danish National Patient Registry, and the Danish National Prescription Registry using a unique encrypted number based on the social security number. As Danish health care is based on a universal, free-for-all, public health care system there is little room for any private practice, creating optimal conditions for registries to be as complete as possible.


#

Identifying cancers, colonoscopies, and comorbidities

A master dataset containing CRC and colonoscopies was obtained by searching the Danish Cancer Registry and the Danish National Patient Registry from 1 January 1998 to 31 December 2015. CRC was identified using Danish ICD-10 codes DC18, DC19, and DC20. Colonoscopies were identified using Nordic Medico-Statistical Committee Classification of Surgical Procedure Codes UJF32 and UJF35 Colonoscopy without/with biopsy. A comorbidity database was constructed containing information about every case of Crohn’s disease, ulcerative colitis, and diverticulitis (using Danish ICD-10 DK50, DK51, and DK57, respectively) diagnosed to 31 December 2015. Hereditary CRC disease was defined as one of the following: DZ848A1 – familial history of hereditary nonpolyposis CRC; DZ800 – familial history with cancer in the gastrointestinal tract; DD126B – hereditary polyposis coli; or DD126F – familial adenomatous polyposis. Comorbidities present at the time of the colonoscopy were used for further analysis. The Charlson Comorbidity Index was calculated based on ICD-10 codes according to the method of Quan et al. 1 year prior to the date of colonoscopy [12] [14].


#

Denmark compared with England

The comparison between the Danish and English NHS PCCRC-3yr was based on a previously published paper by Morris et al. [4]. In the original paper, English NHS PCCRC-3yr rates were calculated using four methods previously described by Bressler, Cooper, le Clercq, and Singh, respectively [6] [7] [8] [9]. The Cooper method was used for Danish comparison as it best resembles the method used by Forsberg, which is the method used in Sweden [15]. The English NHS PCCRC-3yr was based on the year of cancer diagnosis between 2001 and 2010, and limited to the first primary CRC. A PCCRC was defined as a CRC diagnosed 6 – 36 months after a negative colonoscopy. The nominator was PCCRC and the denominator was all CRCs within 36 months after colonoscopy (i. e. CRCs detected 6 – 36 months after a negative colonoscopy and CRCs detected at colonoscopy or within 6 months following colonoscopy).


#

Denmark compared with Sweden

The comparison between the Danish and Swedish PCCRC-3yr was based on a previously published paper by Forsberg et al. [15]. The data used had a different structure, as the denominator was changed from cancer to colonoscopies. Our master dataset was recoded to meet identical specifications and time frame (year of colonoscopy 2001 – 2010 with 3-year CRC follow-up). The time frames for both detected CRC and PCCRC were similar in the study by Forsberg to those described above: 0 – 6 months and 6 – 36 months, respectively. In cases of multiple colonoscopies, only the first colonoscopy performed in the interval (0 – 36 months) was considered in the analysis. Individuals with CRC, Crohn’s disease, or ulcerative colitis diagnosed before the first colonoscopy were excluded. In cases of multiple CRCs, only the first CRC diagnosis was used. Using the Forsberg method, the PCCRC-3yr is calculated by dividing the number of individuals in the PCCRC group (6 – 36 months) by the number of individuals with a colonoscopy followed by a CRC (0 – 36 months).


#

Factors associated with PCCRC

For the current study, the dataset was limited to include colonoscopies from 2001 to 2012 with a CRC diagnosed within 3 years from colonoscopy (i. e. CRC from 2001 to 2015). In cases of multiple colonoscopies, only the first colonoscopy within 3 years of a CRC diagnosis was used for further analysis. The outcome variable was the occurrence of a colonoscopy resulting in a PCCRC (CRC 6 – 36 months after a negative colonoscopy) or a detected CRC (0 – 6 months after colonoscopy). The predictor variables were: year of colonoscopy, age (at colonoscopy), sex, site of tumor, hereditary CRC, diverticular disease, Crohn’s disease, ulcerative colitis, tumor stage, and Charlson Comorbidity Index. Tumor site was based on Danish ICD-10 and grouped into: rectum and sigmoid (DC187, DC19 + DC20), left flexure and descending colon (DC185 + DC186), transverse colon (DC184), right colon (DC180 – DC183), and not otherwise specified (DC188 + DC189). Age was grouped into five categories: 18 – 49, 50 – 59, 60 – 69, 70 – 79, and ≥ 80 years at time of colonoscopy. Year of colonoscopy was grouped into two groups: 2001 – 2006 and 2007 – 2012. Tumor stage was defined as tumor size (T1/T2 /Dukes A vs. T3/T4/Dukes B), metastasis (yes/no/unknown), or regional lymph nodes (yes/no/unknown).

The Poisson regression model was used to test for an interaction between age and hereditary CRC, as the article by Forsberg suspected that an increased PCCRC risk among young individuals was caused by hereditary CRC [15].


#

Risk of PCCRC over time: WEO consensus method

The PCCRC-3yr was calculated according to the WEO consensus method recently presented by Rutter et al. in the WEO consensus statements on post-colonoscopy and post-imaging CRC [5]. CRC and procedure codes were identified as previously described, but ICD-10 C181 (malignant neoplasms of the appendix) was excluded. With this method, each individual is allowed multiple CRCs and multiple colonoscopies; however, for each CRC, only one colonoscopy in the detected CRC group and one colonoscopy in the PCCRC group are allowed. The PCCRC-3yr using the WEO consensus method is calculated by dividing the number of PCCRCs (6 – 36 months) by the total number of detected CRCs (0 – 6 months) and PCCRCs.


#

Statistical methods

Confidence intervals (CIs) for rates were calculated assuming a Poisson distribution. Relative risks (RRs) when comparing countries were calculated according to Altman [16]. A multivariable Poisson regression model was constructed to identify factors associated with PCCRC. Poisson regression was performed using SAS 9.4 (SAS Institute, Cary, North Carolina, USA) with the PROC GENMOD procedure. In the Poisson regression model there was an interaction between hereditary CRC and age, and for this reason the effect of hereditary CRC was stratified by age above and below 60 years.


#
#

Results

Denmark compared with England

From 2001 to 2010, 39 100 Danish individuals were diagnosed with first primary CRC, of whom 11 483 individuals had undergone colonoscopy within 3 years of the diagnosis. A total of 992 PCCRCs were identified. The Danish PCCRC-3yr using the Cooper method was 8.6 % (95 %CI 8.1 % – 9.2 %). The Danish PCCRC-3yr was significantly higher than the reported PCCRC-3yr of 7.7 % (95 %CI 7.6 % – 7.9 %) from the English NHS (RR 1.12, 95 %CI 1.05 – 1.19) [4].


#

Denmark compared with Sweden

From 2001 to 2010, 415 991 colonoscopies were performed in Denmark on 295 952 individuals. When excluding individuals with Crohn’s disease, ulcerative colitis, and age < 18 years, a total of 11 369 individuals were diagnosed with CRC within 3 years of a colonoscopy, of whom 1027 individuals were diagnosed with CRC within 6 – 36 months of a colonoscopy. The Danish PCCRC-3yr using the Forsberg method was 9.0 % (95 %CI 8.5 % – 9.5 %). The corresponding Swedish numbers were 16 319 individuals diagnosed with CRC within 3 years of a colonoscopy and 1286 diagnosed with CRC within 6 – 36 months of a colonoscopy [15]. The Swedish rate was 7.9 % (95 %CI 7.5 % – 8.3 %). The Danish PCCRC-3yr was significantly higher than the Swedish rate (RR 1.15, 95 %CI 1.06 – 1.24).


#

Factors associated with PCCRC

Baseline characteristics are summarized in [Table 1] and results from the multivariable Poisson regression model are shown in [Table 2]. The risk of PCCRC in hereditary CRC differed between patients above and below 60 years of age (P for interaction 0.02). Hereditary CRC was highly associated with PCCRC both in patients aged < 60 years (RR 7.39, 95 %CI 5.77 – 9.47) and in those aged ≥ 60 years (RR 3.81, 95 %CI 2.74 – 5.31). No significant association between PCCRC and age groups in general was found when using the age group > 80 years as a reference.

Table 1

Characteristics of Danish individuals and colorectal cancer occurring < 36 months after colonoscopy: 2001 – 2012.

PCCRC[1]

Detected CRC[2]

All cancers

n

%

n

%

n

Mean age at diagnosis, years

72.4

70.5

70.6

Year of colonoscopy

  • 2001 – 2006

615

11.7

4620

88.3

5235

  • 2007 – 2012

766

7.5

9416

92.5

10182

Age, years

  • 18 – 49

80

12.0

588

88.0

668

  • 50 – 59

133

7.0

1759

93.0

1892

  • 60 – 69

363

8.2

4045

91.8

4408

  • 70 – 79

516

9.7

4802

90.3

5318

  • ≥ 80

289

9.2

2842

90.8

3131

Sex

  • Male

680

8.5

7296

91.5

7976

  • Female

701

9.4

6740

90.6

7441

Charlson Comorbidity Index

  • 0

999

8.3

11036

91.7

12035

  • 1

172

10.8

1424

89.2

1596

  • 2

135

11.7

1014

88.3

1149

  • ≥ 3

75

11.8

562

88.2

637

Colon tumor site

  • Rectum/sigmoid

475

6.2

7210

93.8

7685

  • Splenic flexure/descending

57

7.6

692

92.4

749

  • Transverse

97

10.3

846

89.7

943

  • Cecum/ascending/hepatic flexure

610

11.8

4574

88.2

5184

  • Colon – not otherwise specified

142

16.6

714

83.4

856

Tumor size

  • T1/T2/Dukes A

264

10.3

2303

89.7

2567

  • T3/T4/Dukes B

716

7.4

8969

92.6

9685

  • Unknown

401

12.7

2764

87.3

3165

Lymph nodes metastasis

  • Yes

396

7.3

5018

92.7

5414

  • No

528

8.4

5724

91.6

6252

  • Unknown

457

12.2

3294

87.8

3751

Solid metastasis

  • Yes

288

9.6

2708

90.4

2996

  • No

795

7.8

9396

92.2

10191

  • Unknown

298

13.4

1932

86.6

2230

Crohn’s disease

  • Yes

26

21.0

98

79.0

124

  • No

1355

8.9

13938

91.1

15293

Ulcerative colitis

  • Yes

82

32.5

170

67.5

252

  • No

1299

8.6

13866

91.4

15165

Hereditary cancer (age < 60 years)

  • Yes

36

64.3

20

35.7

56

  • No

1345

8.8

14016

91.2

15361

Hereditary cancer (age ≥ 60 years)

  • Yes

23

43.4

30

56.6

53

  • No

1358

8.8

14006

91.2

15364

Diverticulitis

  • Yes

270

27.4

715

72.6

985

  • No

1111

7.7

13321

92.3

14432

Surgical specialty

  • Yes

1113

8.8

11505

91.2

12618

  • No

268

9.6

2531

90.4

2799

Hospital region

  • West

704

8.7

7429

91.3

8133

  • East

677

9.3

6607

90.7

7284

University hospital

  • Yes

215

12.0

1573

88.0

1788

  • No

1166

8.6

12463

91.4

13629

PCCRC, post-colonoscopy colorectal cancer.

1 CRC diagnosed 6 – 36 months after a negative colonoscopy.


2 CRC diagnosed < 6 months after colonoscopy.


Table 2

Multivariable Poisson regression analysis for post-colonoscopy colorectal cancer in Denmark.

RR

95 %CI

P value

Year of colonoscopy

  • 2007 – 2012

1.00

  • 2001 – 2006

1.55

1.40 – 1.71

< 0.001

Age, years

  • 18 – 49

1.24

0.98 – 1.58

0.07

  • 50 – 59

0.88

0.72 – 1.08

0.23

  • 60 – 69

1.09

0.94 – 1.26

0.27

  • 70 – 79

1.14

1.00 – 1.31

0.05

  • ≥ 80

1.00

Sex

  • Male

1.00

  • Female

1.00

0.91 – 1.10

0.99

Charlson Comorbidity Index

  • 0

1.00

  • 1

1.20

1.03 – 1.40

< 0.05

  • 2

1.25

1.06 – 1.48

< 0.01

  • ≥ 3

1.24

0.98 – 1.56

0.07

Colon tumor site

  • Rectum/sigmoid

1.00

  • Splenic flexure/descending

1.23

0.93 – 1.62

0.14

  • Transverse

1.57

1.28 – 1.94

< 0.001

  • Cecum/ascending/hepatic flexure

1.85

1.64 – 2.08

< 0.001

  • Colon – not otherwise specified

2.08

1.74 – 2.49

< 0.001

Tumor size

  • T1 /T2 /Dukes A

1.00

  • T3 /T4 /Dukes B

0.70

0.61 – 0.81

< 0.001

  • Unknown

0.86

0.71 – 1.04

0.12

Lymph nodes metastasis

  • No

1.00

  • Yes

0.94

0.83 – 1.07

0.33

  • Unknown

1.20

0.99 – 1.47

0.07

Solid metastasis

  • No

1.00

  • Yes

1.13

0.98 – 1.31

0.09

  • Unknown

1.30

1.11 – 1.52

< 0.01

Crohn’s disease

  • No

1.00

  • Yes

1.44

0.97 – 2.16

0.07

Ulcerative colitis

  • No

1.00

  • Yes

3.44

2.79 – 4.23

< 0.001

Hereditary cancer (age < 60 years)

  • No

1.00

  • Yes

7.39

5.77 – 9.47

< 0.001

Hereditary cancer (age ≥ 60 years)

  • No

1.00

  • Yes

3.81

2.74 – 5.31

< 0.001

Diverticulitis

  • No

1.00

  • Yes

3.25

2.88 – 3.66

< 0.001

Surgical specialty

  • No

1.00

  • Yes

1.12

0.98 – 1.28

0.09

Hospital region

  • West: North, Central and Southern region.

1.00

  • East: Zealand and the Capital region.

1.02

0.92 – 1.13

0.69

University Hospital

  • No

1.00

  • Yes

1.14

0.99 – 1.30

0.07

RR, relative risk; CI, confidence interval.

The risk of a PCCRC increased when moving through each colon segment using the rectum/sigmoid colon as a reference. Higher risk was found in the transverse colon (RR 1.57, 95 %CI 1.28 – 1.94) and in the ascending colon and cecum (RR 1.85, 95 %CI 1.64 – 2.08). PCCRC was less likely when the tumor was large (T3/T4/Dukes B vs. T1/T2/Dukes A: RR 0.70, 95 %CI 0.61 – 0.81). PCCRC was not associated with an increased risk of lymph node or solid metastasis. Diseases such as diverticulitis (RR 3.25, 95 %CI 2.88 – 3.66), ulcerative colitis (RR 3.44, 95 %CI 2.79 – 4.23), and a Charlson Comorbidity Index of 1 or 2 (RR 1.20, 95 %CI 1.03 – 1.40 and RR 1.25, 95 %CI 1.06 – 1.48, respectively) were also significantly associated with PCCRC.


#

Risk of PCCRC over time: WEO consensus method

Biannual results are shown in [Fig. 1], with falling PCCRC-3yr over time, from 22.5 % in 2001 to 7.9 % in 2012. Annual numbers of detected CRC and PCCRC colonoscopies are available in [Table 3] for future comparison with Danish rates.

Zoom Image
Fig. 1 Danish post-colonoscopy colorectal cancer (PCCRC) occurring within 3 years of a negative colonoscopy, based on the consensus method of the World Endoscopy Organization [5].
Table 3

Numbers of Danish post-colonoscopy colorectal cancer and diagnosed cancer colonoscopies based on the consensus method of the World Endoscopy Organization [5].

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

PCCRC colonoscopies

84

112

103

125

149

160

152

190

170

170

165

166

Diagnosed cancer colonoscopies[1]

289

314

378

1196

1311

1444

1481

1461

1598

1771

1817

1947

PCCRC-3yr[2]

22.5

26.3

21.4

9.5

10.2

10.0

9.3

11.5

9.6

8.8

8.3

7.9

PCCRC, post-colonoscopy colorectal cancer.

1 CRC detected at colonoscopy or within 6 months of colonoscopy.


2 PCCRC-3yr calculated as PCCRC/(diagnosed cancer + PCCRC) × 100 %.



#
#

Discussion

This nationwide population-based cohort study is the first to directly compare PCCRC-3yr rates between countries using equivalent methodologies. Furthermore, the analysis confirms previous associations and identifies new factors associated with PCCRC. Our finding of over-representation of PCCRC in both the transverse and ascending colon/cecum is in line with previous studies that found increased risk of PCCRC in the proximal colon [4] [6] [7] [8] [9] [15]. Procedure-related factors are likely to play a part, as the right colon is more difficult to cleanse with oral agents, landmarks are less clear, and it is more technically challenging to reach the proximal colon and maintain proper scope position. Additionally, there is increasing evidence that polyps and cancers in the right colon might also represent a different entity. PCCRC is associated with microsatellite instability and CpG island methylator phenotype (CIMP) [17]. The precursor lesions for CIMP-positive tumors are suspected to be sessile serrated adenomas (SSAs) [17]. SSAs are known to occur more frequently in the right colon and their flat appearance makes them more difficult to detect during colonoscopy [18].

We also found an association between PCCRC and a previous diagnosis of diverticulitis, which is in line with previous studies [4] [6] [15] [19]. In general, diverticular disease is known to make a colonoscopy more difficult, with increased patient pain, impaired views in the affected areas, and the risk of mistaking neoplastic tissue for an area of severe inflammation [19]. Long-standing chronic inflammation in diverticulitis has been suggested as a risk for CRC; however, the finding has been disputed by others, and might be a result of misclassification and intensive surveillance [20] [21] [22].

Previous studies have shown conflicting results regarding the association between age and PCCRC. Bressler et al. found that older age was associated with an increased risk of PCCRC, whereas other studies have found no association with age [6] [15] [23]. Forsberg et al. found an increased risk of PCCRC in the age group of 18 – 30 years and suspected that hereditary cancers might be part of the explanation [15]. Our analysis supports the Swedish hypothesis. When adjusting for interaction between hereditary cancer and age there was no significant association between any age group and the risk of PCCRC, but borderline associations were found in both the 18 – 50 years and 70 – 79 years age groups. Hereditary cancer disease was associated with a higher risk of PCCRC both among younger (< 60 years) and older (≥ 60 years) age groups. The exact reasons for this association cannot be established from this study, but hereditary cancers are known to have a different biological pathway, occur earlier in life, and be fast growing [24].

Ulcerative colitis was associated with an increased risk of PCCRC but Crohn’s disease was not. In previous studies, inflammatory bowel disease (IBD) has often been excluded before the initial analysis, mainly because IBD-associated cancers are considered a different entity, with different genomic alterations compared with CRC that is not related to IBD [6] [7] [15] [25]. It is unknown whether IBD affects the adenoma – carcinoma sequence directly or whether the higher risk of PCCRC is procedure related. The mucosa is often abnormal in patients with long-standing IBD and subtle lesions might be missed, and the short surveillance intervals recommended for high risk patients will inevitably raise the chance of finding a cancer within 3 years. Short surveillance intervals are unlikely to be a reason for higher rates of PCCRC in IBD in the Danish cohort because, until 2014, the Danish guidelines did not require such intensive surveillance. Of course, it is possible that the patients most at risk were having frequent colonoscopies at shorter intervals to assess disease and guide treatment, and this will accentuate PCCRC-3yr rates.

From a Danish perspective, it is concerning that the PCCRC-3yr was significantly higher than in both Sweden and the English NHS. Since a nationwide colonoscopy quality survey (performed in 1999 and reported in 2004) demonstrated poor results in the English NHS, there have been many quality improvement initiatives in the English NHS that might explain some of the differences [26]. A certification process for both newly independent screening colonoscopists and those screening fecal occult blood tests, a national training infrastructure, and participation of endoscopy services in a unit accreditation scheme that requires services to monitor colonoscopist key performance indicators and act on poor performance, are just some of the initiatives that have not occurred in Denmark [27] [28]. A closer look into the colonoscopist background and training reveals differences that may also be relevant. In England and the USA, gastroenterologists perform most of the colonoscopies, whereas in Canada, Sweden, and Denmark surgeons perform a higher proportion of colonoscopies. Our study did not find any difference in the risk of PCCRC between colonoscopies performed in a surgical unit and those performed in a nonsurgical unit; however, previous studies have found increased risk of PCCRC with nongastroenterologists, surgeons, and family physicians [7] [29] [30]. For years, at least in Denmark, colonoscopy might have been regarded as a minor procedure among surgeons, just one among many procedures that needed to be mastered during surgical training. As described by Rabeneck et al., the time spent on endoscopy training by gastroenterologists is almost 16 months vs. 2 months for surgeons during their residency/training program in Canada [30]. Such differences between surgeons and gastroenterologists do not exist in Denmark: there are no official endoscopy training programs during residency. Current training to become a surgeon in Denmark requires 200 colonoscopies during the 5-year training program, whereas gastroenterologists have no minimum requirements regarding colonoscopies. Gastroenterology trainees are required to pass a 2-day colonoscopy course and perform to acceptable standards; however, few gastroenterologists have performed more than 100 colonoscopies at the end of their 5-year specialist training program (personal correspondence with chief education gastroenterologists, North Region, Denmark).

Falling PCCRC-3yr has previously been demonstrated in the English NHS and in Sweden [8] [15]. We also found declining PCCRC-3yr over time in the Danish system. The multivariable analysis showed an RR for PCCRC of 1.55 (95 %CI 1.40 – 1.71) when comparing 2001 – 2006 with 2007 – 2012, indicating better performance over time ([Table 2]). [Fig. 1] shows biannual PCCRC-3yr from 2001 to 2012 using the WEO consensus method. Rates are clearly falling over time. However, it must be remembered that both [Fig. 1] and [Table 2] are based on a cancer-only model. The model compares the number of colonoscopies that diagnosed a CRC (0 – 6 months before CRC) and the number of colonoscopies missing a cancer (6 – 36 months before CRC). This methodology is identical to multiple previous studies but it does come with a risk [4] [6] [7] [8] [9]. If the way colonoscopy is used changes over time, the PCCRC-3yr may decline without any real improvements.

For example, more than a decade ago, it was unusual in Denmark to always require preoperative histological confirmation of CRC. Today, a pre-operative colonoscopy would almost certainly be required to achieve a histological confirmation, resulting in a colonoscopy-detected CRC that would not have registered in previous years. This will have the effect of increasing the denominator and lowering the PCCRC-3yr. Such a phenomenon does seem to have affected the Danish numbers. As seen from [Table 3], the numbers of colonoscopy-detected CRCs tripled between 2003 and 2004, whereas the numbers of PCCRC remained relatively stable. From procedure codes we could identify the main driving force as colonoscopy with biopsies, indicating a pre-operative regime shift (see Appendix A in the online-only Supplementary material). More recent data are less likely to be affected by changes in approach and it is reassuring that they show falling PCCRC-3yr rates over time ([Fig. 1]).

Registries may over-report PCCRC. A study from Pennsylvania showed that 47 % of PCCRCs were due to registry errors [31]. However, the Danish Cancer Registry is known to be very accurate, but there is always a risk of procedures and cancers being miscoded, misclassified, or misdated [10] [11]. A previous Danish study reviewed patient records in 101 interval cancers from one Danish medical center [32]. The time frame for detected CRC was shorter than in our study, with an increased risk of CRC being incorrectly assigned to the PCCRC group. Despite the short time frame for detected CRCs, 89 % of PCCRCs were correctly assigned. Overall, registry errors seem to be less of a concern in the Danish Cancer Registry and Danish National Patient Registry than other registries.

Registries make it possible to explore associations between CRC and colonoscopies, but very little information about the actual procedures is available. There is limited reporting of colonoscopy key performance indicators at local level and no plans for a nationwide colonoscopy database. There are supplemental procedure codes for completeness and bowel preparation but, unfortunately, there is no mandatory reporting [33].

This study has demonstrated that it is possible to compare rates of PCCRC in different jurisdictions. The key finding of higher PCCRC-3yr rates in Denmark compared with England and Sweden prompts the question of whether differences in the provision of services, training of endoscopists, and/or quality improvement practices such as accreditation are influencing rates. Given that 76 % – 86 % of PCCRC cases are considered avoidable [8] [34], there is an urgent need to change practice in order to reduce rates. National improvement programs, important as they might be, will take years to implement and have an effect. However, the data from this and other studies indicate that there are some things that can be done immediately, such as ensuring that patients with the highest risk of PCCRC (such as those with diverticulosis, hereditary cancer, ulcerative colitis, and high comorbidity) have their colonoscopy performed by the most competent colonoscopists. Furthermore, endoscopy services could ensure that all patients with suboptimal visualization of the colon (whether this is due to poor preparation or incomplete procedures) have further colonic imaging, either repeat colonoscopy or computed tomography colonoscopy. Finally, performing root cause analysis (as recommended by WEO) of individual PCCRCs will identify other factors amenable to improvement interventions.


#
#

Competing interests

None


Corresponding author

Lasse Pedersen, MD
Department of Surgical Gastroenterology
Aalborg Universitetshospital
Hobrovej 18-22
Aalborg 9100
Denmark   


Zoom Image
Fig. 1 Danish post-colonoscopy colorectal cancer (PCCRC) occurring within 3 years of a negative colonoscopy, based on the consensus method of the World Endoscopy Organization [5].