Article Text

Original research
Interval cancer after colonoscopy in the Austrian National Screening Programme: influence of physician and patient factors
  1. Elisabeth Waldmann1,2,
  2. Daniela Penz1,2,
  3. Hana Šinkovec3,
  4. Georg Heinze3,
  5. Christoph Rinner4,
  6. Lena Jiricka1,3,
  7. Barbara Majcher1,2,
  8. Anna Hinterberger1,2,
  9. Michael Trauner1,
  10. Monika Ferlitsch1,2
  1. 1 Department of Internal Medicine III, Division of Gastroenterology and Hepatology, Medical University of Vienna, Vienna, Austria
  2. 2 Quality Assurance Working Group, Austrian Society of Gastroenterology and Hepatology, Vienna, Austria
  3. 3 Center for Medical Statistics, Informatics and Intelligent Systems, Institute of Clinical Biometrics, Medical University of Vienna, Vienna, Austria
  4. 4 Center for Medical Statistics, Informatics and Intelligent Systems, Institute of Medical Information Management, Medical University of Vienna, Vienna, Austria
  1. Correspondence to Professor Monika Ferlitsch, Dept. of Internal Medicine III, Div. of Gastroenterology and Hepatology, Medical University of Vienna, Vienna, Vienna, Austria; monika.ferlitsch{at}meduniwien.ac.at

Abstract

Objective Postscreening colorectal cancer (PSCRC) after screening colonoscopy is associated with endoscopists’ performance and characteristics of resected lesions. Prior studies have shown that adenoma detection rate (ADR) is a decisive factor for PSCRC, but correlations with other parameters need further analysis and ADR may change over time.

Design Cohort study including individuals undergoing screening colonoscopy between 1/2008 and 12/2019 performed by physicians participating in a quality assurance programme in Austria. Data were linked with hospitalisation data for the diagnosis of PSCRC (defined as CRC diagnosis >6 months after colonoscopy). ADR was defined dynamically in relation to the time point of subsequent colonoscopies; high-risk groups of patients were those with an adenoma ≥10 mm, or with high-grade dysplasia, or villous or tubulovillous histology, or a serrated lesion ≥10 mm or with dysplasia, or colonoscopies with ≥3 lesions. Main outcome was PSCRC for each risk group (negative colonoscopy, hyperplastic polyps, low-risk and high-risk group of patients) after colonoscopy by endoscopists with an ADR <20% compared with endoscopists with an ADR ≥20%.

Results 352 685 individuals were included in the study (51.0% women, median age 60 years) of which 10.5% were classified as high-risk group. During a median follow-up of 55.4 months, 241 (0.06%) PSCRC were identified; of 387 participating physicians, 19.6% had at least one PSCRC (8.4% two or more). While higher endoscopist ADR decreased PSCRC incidence (HR per 1% increase 0.97, 95% CI 0.95 to 0.98), affiliation to the high-risk group of patients was also associated with higher PSCRC incidence (HR 3.27, 95% CI 2.36 to 4.00). Similar correlations were seen with regards to high-risk, and advanced adenomas. The risk for PSCRC was significantly higher after colonoscopy by an endoscopist with an ADR <20% as compared with an endoscopist with an ADR ≥20% in patients after negative colonoscopy (HR 2.01, 95% CI 1.35 to 3.0, p<0.001) and for the high-risk group of patients (HR 2.51, 95% CI 1.49 to 4.22, p<0.001).

Conclusion A dynamic calculation of the ADR takes into account changes over time but confirms the correlation of ADR and interval cancer. Both lesion characteristics and endoscopists ADR may play a similar role for the risk of PSCRC. This should be considered in deciding about appropriate surveillance intervals in the future.

  • colonic adenomas
  • colorectal cancer screening

Data availability statement

All data relevant to the study are included in the article or uploaded as online supplemental information.

Statistics from Altmetric.com

Significance of this study

What is already known on this subject?

  • Postscreening colorectal cancer is associated with endoscopists’ performance and characteristics of resected lesions. However, the relative impact of those two features needs further study.

What are the new findings?

  • To characterise performance quality of an endoscopist, we introduced a novel, dynamic calculation of the adenoma detection rate (ADR), which accounts for the possible changes over time.

  • Using this dynamic ADR and a more liberal time definition of postscreening colorectal cancer, we confirmed that both were correlated. HR per 10% increase in ADR 0.7 (95% CI 0.60 to 0.82).

  • The impact of both adenoma parameters and endoscopist quality (ADR) appeared to be of similar importance.

  • The rate of advanced adenomas showed a similar correlation and does not appear to be a superior quality parameter compared with overall ADR.

  • How might these results change the focus of research or clinical practice?

  • Decision about surveillance intervals after screening colonoscopy may include endoscopists’ ADR, dynamically calculated, in addition to lesion characteristics.

Introduction

Screening colonoscopy aims at detecting and removing premalignant colonic lesions, and therefore, reduce incidence and mortality of colorectal cancer.1–3 Surveillance colonoscopy intervals after screening colonoscopy are scheduled according to number, size and histopathological characteristics of lesions detected.4 Although endoscopic screening is usually performed with high-quality standards, postscreening colorectal cancer (PSCRC) still occurs in a significant number of patients.5 The underlying risk factors are poorly understood. PSCRCs are cancers diagnosed between screening colonoscopy and scheduled time of surveillance colonoscopy. They develop as consequences of overlooked lesions, distinct tumour features associated with rapid tumour progression and/or inadequate polyp resection.6–8 Data on associated technical issues (eg, lesion detection rates, bowel preparation quality), patients’ (eg, female sex, higher age), respectively, polyp/tumour characteristics (eg, localisation, morphology, molecular pathology) are discussed controversially.5 6 9–11 The main surrogate parameter used to determiner colonoscopy outcome quality has been adenoma detection rate (ADR).7 8 12 Most studies investigated technical issues rather than characteristics of premalignant lesions prior to occurrence of PSCRC. However, characteristics of preexisting lesions are associated with future tumour growth.4

Therefore, the aim of the present study was to evaluate the impact of endoscopic performance and lesion characteristics on the incidence of PSCRC. We decided to use a dynamic model of ADR calculation to accommodate for changes in quality parameters over time.

Methods

Study population

Austria has an opportunistic colorectal cancer screening programme offering faecal occult blood testing starting by the age of 40 years or screening colonoscopy to asymptomatic individuals at average risk for colorectal cancer starting by the age of 50 years. The database also comprises colonoscopies of individuals 30–49 years with a positive family history of CRC or if they expressed fear of cancer, but individuals with hereditary cancer syndromes (eg, FAP or Lynch syndrome) or chronic inflammatory bowel disease are by definition not average-risk screening and were therefore not included in the study cohort. All screening colonoscopies included in the present study were performed within a nationwide quality assurance programme. Participation in this programme is on voluntary basis and approximately half of all endoscopic centres participate. Details on this project were published previously.13–15 Physicians are contractually committed to provide all screening colonoscopies performed. Since data are collected prospectively informed consent by the patients for data transmission is required. Data are published only in an anonymised form; hence it is unlikely that informed consent is refused. However, due to a general data protection regulation, the possibility of data monitoring is limited to those individuals who provided informed consent. Screening colonoscopy report forms include patients’ sex and age as well as lesions count, size, localisation, macroscopic appearance and histopathological workup as well as scheduled time of surveillance colonoscopy.

Study design and outcomes

The present study considered individuals who underwent screening colonoscopy within The Austrian National colorectal cancer screening programme and included colonoscopy data from physicians voluntarily participating in the quality assurance programme between January 2008 and December 2019. Patients were followed from the date of screening colonoscopy (baseline) to the first of the following events: (1) the subsequent surveillance colonoscopy, (2) a diagnosis of colorectal cancer or (3) the end of the follow-up period. Main outcome was PSCRC for each risk group (negative colonoscopy, hyperplastic polyps, low-risk and high-risk group of patients) after colonoscopy by endoscopists with an ADR <20% compared with endoscopists with an ADR ≥20%.

Definition and data assessment of PSCRC

A PSCRC was defined as colorectal cancer diagnosed at least 6 months after screening colonoscopy and before the date of an actual surveillance colonoscopy. The definition of PSCRC or interval cancer is not consistent in the literature.7 8 16 We chose this definition because it reflects prevalence of PSCRC in clinical routine, which is not only influenced by physician but also by patient related factors. Studies with an upper limit of 36 months16 would miss, that is, PSCRC 50 months after resection of low-risk polyps or after negative colonoscopy.

To obtain data on the diagnosis of PSCRC, we linked the screening colonoscopy report forms with hospitalisation data provided by The Main Association of the Austrian Social Security Institutions that comprises date of admission and discharge as well as diagnosis codes, coded by the International Classification of Diseases (ICD-10 and ICD-9, respectively) on an individual basis. For linkage of both registries the colonoscopy database and the hospital registry, individuals were identified by a unique pseudonym consisting of one to four numbers by The Main Association of the Austrian Social Security Institutions, the umbrella organisation of all insurance institutions in Austria. We merged the datasets by a numerical patient identifier provided by the The Main Association of the Austrian Social Security Institutions in both datasets. Since insurance coverage is close to 100% in Austria, the merged dataset is highly unlikely to miss data because the respective insurance company registers every hospital admission through the bill of expenses. We considered the ICD codes associated with malignant neoplasm of the colon and rectum (figure 1). The outcomes (PSCRC detected at hospital admissions) were queried in December 2019.

Figure 1

Patients flow. PSCRC, postscreening colorectal cancer.

Dynamic calculation of ADR, AADR and hrADR

In order to disentangle the ADR of an endoscopist from the result of a colonoscopy for a patient treated by this endoscopist, ADR of an endoscopist was dynamically defined. In particular, ADR was recomputed each time an endoscopist performed a colonoscopy as the proportion of colonoscopies with at least one adenoma detected out of all previous screening colonoscopies performed by this physician (online supplemental figure S1). The first 29 examinations of each endoscopist, or examinations from endoscopists performing less than 30 examinations were excluded in any further analysis. Similarly, advanced ADR (AADR) and high-risk ADR (hrADR) were dynamically calculated. An advanced adenoma was defined as adenoma ≥10 mm, or with high-grade dysplasia, or villous or tubulovillous histology, or a serrated lesion ≥10 mm or with dysplasia. hrADR was defined as the proportion of screening colonoscopies with adenomas harbouring characteristics of an advanced adenoma or colonoscopies with ≥3 lesions, corresponding to the definition of the high-risk group in current surveillance guidelines.4 Details on histological features were assessed for the most advanced lesion detected.

Supplemental material

Risk groups

Based on findings from screening colonoscopy, we stratified patients into risk groups following the postpolypectomy colonoscopy surveillance guidelines of the European Society of Gastrointestinal Endoscopy. The high-risk group of patients are recommended to undergo surveillance colonoscopy after 3 years, all other individuals after 10 years.4 In the present study, the high-risk group of patients comprised adenomas ≥10 mm, or with high-grade dysplasia, or villous or tubulovillous histology, or a serrated lesion ≥10 mm or with dysplasia, or colonoscopies with ≥3 lesions. The low-risk group of patients included individuals with one or two tubular adenomas <10 mm, or with low-grade dysplasia, or serrated lesions<10 mm, or without dysplasia. At the beginning of the study period, lesions’ count was assessed as 1, 2–4 and >4. Screenees with 2–4 lesions detected were considered low-risk group of patients. The other two groups included patients where hyperplastic polyps only were detected at screening colonoscopy and patients with negative screening colonoscopy, respectively.

Exclusion criteria

Patients with missing hospital admission date or hospital discharge date (which was used to define PSCRC), with follow-up shorter than 6 months, or who had cancer detected at screening colonoscopy, and patients with a history of colorectal cancer were excluded from the study (figure 1).

For regression analysis, we further excluded endoscopists with <30 colonoscopies and the first 29 examinations of each endoscopist.

Statistical analysis

Continuous variables are described by median and IQR and categorical variables by absolute frequencies and percentages. All variables were completely recorded except for quality of bowel preparation, which was not assessed in the database until 2012 and not used other than for descriptive purposes. We evaluated the PSCRC incidence by the method of Kaplan-Meier estimation. Similarly, we estimated the incidence of PSCRC stratified by the risk group and ADR quartiles, respectively. We updated each endoscopist’s ADR, AADR and hrADR dynamically at each examination. Such we accounted for the possible learning effect of endoscopists over time and did not include information from the future to determine these detection rates. Patients were censored at the end of the follow-up period or at the actual date of a surveillance colonoscopy.

We used multivariable Cox proportional hazards model to associate PSCRC incidence with the patient-specific risk factors sex, age, cecal intubation status and risk group and the endoscopist specific quality indicator ADR. We tested by using likelihood ratio test whether the presence of interactions between ADR and risk groups could improve the model. Similarly, we investigated whether ADR modelled nonlinearly, as a restricted cubic spline function with three knots, could better describe its association with the outcome (online supplemental figure S2). Moreover, we described the association of ADR with the relative hazard similarly to Corley et al,8 where ADR was grouped according to quintiles. For comparability, we used the same category definition.8 The robust sandwich variance estimate was used to account for within-physician clustering. The proportional hazards assumption was tested by evaluating the correlation of scaled Schoenfeld residuals with time and time-covariate interactions were introduced into the model in case of non-proportionality. To assess ranking of importance of prognostic factors in the final model, in particular of ADR and risk groups, we used standardised regression coefficients, which were defined as the log HR of a variable multiplied with the empirical SD of that variable. Additionally, we refitted the model by splitting the group of patients with hyperplastic polyps by size of the polyps (<5 mm and ≥5 mm, respectively) to detect whether there were non-proportional influences of hyperplastic polyp size on the risk of PSCRC. Similar analysis was repeated by substituting ADR in the model by AADR and hrADR, respectively. Two-sided p values less than 5% were considered statistically significant. Statistical analysis was performed using R 3.6.3 with packages survival17 and rms.18

Supplemental material

Results

Baseline characteristics

A total of 385 708 screening colonoscopies were performed between January 2008 and December 2019 of which 352 685 could be included in the study (see figure 1 for further details); 51.0% were women, median age was 60 years (IQR 54–68). Median ADR was 21.8% (IQR 16.1–28.7) and the coecum was intubated in 97.0%. 10.5% of patients were classified as high-risk group of patients, 12.7% as low-risk group of patients, in 14.4% hyperplastic polyps only were detected and 62.5% had a negative colonoscopy (no lesions detected). Median follow-up time obtained by a reverse Kaplan-Meier method was 55.4 months (IQR 33.9–85.7). Patient’s and physician’s characteristics are presented in table 1.

Table 1

Baseline characteristics of patients and physicians included in the study

Characteristics of PSCRC

During the follow-up, 241 PSCRCs were identified. The cumulative PSCRC incidence at 5 years was 0.08% (95% CI 0.07% to 0.09%); 1 year after negative colonoscopy 0.015% developed PSCRC (3 years: 0.045%, 5 years: 0.068%). 1 year after negative colonoscopy by an endoscopist with an ADR≥20%–0.007% (3 years: 0.027%, 5 years: 0.045%) developed PSCRC whereas colonoscopy performed by an endoscopist with an ADR <20%–0.023% (3 years: 0.063%, 5 years: 0.089%) developed PSCRC.

Incidence of PSCRC 1 year after colonoscopy where hyperplastic polyps were detected is 0%, however, 3 years after colonoscopy where hyperplastic polyps were detected incidence of PSCRC is 0.041% (0.026% after colonoscopy by an endoscopist with an ADR≥20% and 0.067% by an endoscopist with an ADR <20%). Incidence of PSCRC 5 years after colonoscopy where hyperplastic polyps were detected is 0.067% (0.065% after colonoscopy by an endoscopist with an ADR≥20% and 0.074% by an endoscopist with an ADR <20%).

One year after colonoscopy assigned to low-risk group of patients 0.007% developed PSCRC (3 years: 0.040%, 5 years: 0.057%). One year after colonoscopy in this risk group by an endoscopist with an ADR≥20%–0.007% (3 years: 0.040%, 5 years: 0.051%) developed PSCRC. Incidence of PSCRC 1 year after colonoscopy by an endoscopist with an ADR <20% was 0.008% (3 years: 0.042%, 5 years: 0.071%). In high-risk group of patients 1 year after colonoscopy 0.042% developed PSCRC (3 years: 0.135%, 5 years: 0.184%). 1 year after colonoscopy and assigned to the high-risk group of patients by an endoscopist with an ADR≥20% incidence of PSCRC is 0.029% (3 years: 0.105%, 5 years: 0.140%), 1 year after colonoscopy by an endoscopist with an ADR <20%–0.080% developed PSCRC (3 years: 0.223%, 5 years: 0.307%).

Figures 2 and 3 show the cumulative PSCRC incidence according to ADR and risk groups of patients, respectively. Figure 4 shows the cumulative PSCRC incidence considering both the ADR (<20% or≥20%) and risk groups of patients. 49.4% of patients with PSCRC were female; the median age at screening colonoscopy was 67 years (IQR 59–74). The coecum was reached in 93.0%. The median time between screening colonoscopy and first hospitalisation due to PSCRC was 24.5 months (IQR 13.0–39.9). 47% of PSCRCs were located proximal to the sigmoid colon, 37% in the sigmoid colon or rectum; in 16% the localisation was not specified.

Figure 2

Cumulative postscreening colorectal cancer (PSCRC) incidence stratified by adenoma detection rate (ADR) quartiles. ADR was defined dynamically and was recomputed each time an endoscopist performed a colonoscopy. ADR <16.2%: red line ADR 16.2%–21.8%: orange line ADR 21.9%–28.7%: green line ADR >28.7%: blue line.

Figure 3

Cumulative postscreening colorectal cancer (PSCRC) incidence stratified by risk group high-risk group*: red line hyperplastic polyps: orange line low-risk group**: green line negative colonoscopy: blue line * adenomas ≥10 mm or with high-grade dysplasia or villous or tubulovillous histology, or a serrated lesion ≥10 mm or with dysplasia, or colonoscopies with ≥3 lesions. ** 1–2 tubular adenomas <10 mm, or with low-grade dysplasia, or serrated lesions<10 mm, or without dysplasia.

Figure 4

Cumulative postscreening colorectal cancer (PSCRC) incidence stratified by adenoma detection rate (ADR) and risk group. ADR was defined dynamically and was recomputed each time an endoscopist performed a colonoscopy. High-risk group: adenomas≥10 mm, or with high-grade dysplasia, or villous or tubulovillous histology, or a serrated lesion ≥10 mm or with dysplasia, or colonoscopies with ≥3 lesions. Low-risk group: 1–2 tubular adenomas<10 mm, or with low-grade dysplasia, or serrated lesions<10 mm, or without dysplasia.

Risk factors for PSCRC

After exclusion of endoscopists with <30 colonoscopies and the first 29 examinations of each endoscopist, 336 419 screening colonoscopies and 234 PSCRC were included in the regression analysis (figure 1). A multivariable Cox model including age, sex, cecal intubation status, risk group and ADR (table 2) revealed a significant association between PSCRC and ADR (HR per 1% increase 0.97, 95% CI 0.95 to 0.98, p<0.001) as well as PSCRC and high-risk group of patients (HR compared with negative colonoscopy 3.27, 95% CI 2.36 to 4.53, p<0.001). By taking patients’ age, sex, cecal intubation status and risk group into account, a patient examined by an endoscopist with 1% higher ADR was expected to have an approximately 3% lower risk of PSCRC (95% CI 2 to 5). However, modelling ADR as a restricted cubic spline with three knots revealed there might be a non-linear association between ADR and incident PSCRC (p=0.045) (online supplemental figure 2). Consequently, for better comparability of results, we grouped ADR into categories as defined in Corley et al,8 where ADR quintiles were used (figure 5). For patients treated by endoscopists with ADRs between 23.9% and 28.4%, 28.4% and 33.5% and ≥33.5%, respectively, as compared with endoscopists with ADRs below 19.1% the risk of PSCRC was significantly reduced (HR 0.57, 95% CI 0.38 to 0.87, p=0.010, HR 0.3, 95% CI 0.16 to 0.54, p<0.001 and HR 0.36, 95% CI 0.20 to 0.63, p<0.001), after adjustment for covariates. There was, however, no evidence of a clear threshold where the risk of PSCRC decreased.

Figure 5

Association of adenoma detection rate (ADR) and adjusted HRs for postscreening colorectal cancer (PSCRC).

Table 2

Multivariable Cox model analysing the impact of patients’ age and sex, assessment to hyperplastic polyps, low or high-risk group, cecal intubation and adenoma detection rate (ADR) of the endoscopists on the occurrence of postcolonoscopy colorectal cancer (PSCRC)

The association between PSCRC and increased patient age was also highly significant (HR per 10 years increase 1.79, 95% CI 1.54 to 2.08, p<0.001). The influence of negative cecal intubation status appeared non-proportional in time (p=0.013) and was strongly predictive for PSCRCs occurring in the first 21 months after colonoscopy (HR 3.93, 95% CI 2.15 to 7.19, p<0.001) but not for those occurring later on (HR 0.84, 95% CI 0.31 to 2.32, p=0.741).

Notably, no significant association could be detected between PSCRC and female sex (HR compared with male sex 1.02, 95% CI 0.76 to 1.39, p=0.878), or for the low-risk group of patients compared with negative colonoscopy (HR 0.92, 95% CI 0.57 to 1.49; 0.732), or when hyperplastic polyps were detected at screening colonoscopy as compared with negative colonoscopy (HR 1.16, 95% CI 0.78 to 1.71, p=0.470). The impact of sex, age, risk groups, ADR modelled linearly and cecal intubation rate on the occurrence of PSCRC is summarised in table 2. Splitting the group of patients with hyperplastic polyps according to the size of polyps (<5 mm and ≥5 mm, respectively) did not show any non-proportional influences of hyperplastic polyp size on the risk of PSCRC.

An analyses substituting ADR with AADR and hrADR showed a significant association with PSCRC for both parameters (AADR per 1% increase HR 0.96, 95% CI 0.92 to 1.00, p=0.034, hrADR per 1% increase HR 0.94, 95% CI 0.91 to 0.97, p<0.001).

ADR, risk groups and PSCRC

In order to investigate effect modification, we fitted interactions between ADR and risk group of patients, which were not statistically significant (p=0.686) and were therefore not included in the final model. These results suggest there is a non-negligible effect of the ADR on the outcome regardless of the risk group, and that the patients in the high-risk group have a significantly higher risk of CRC as compared with other groups (see also figure 4).

If screening colonoscopy was performed by an endoscopist with an ADR <20%, the risk for PSCRC was significantly higher as compared with an endoscopist with an ADR ≥20% (HR 2.05, 95% CI 1.49 to 2.80, p<0.001).

Strikingly, for individuals after negative colonoscopy the risk for PSCRC was significantly higher if colonoscopy was performed by an endoscopist with an ADR<20% compared with an endoscopist with an ADR≥20% (HR 2.01, 95% CI 1.35 to 3.0, p<0.001). In the high-risk group of patients risk for PSCRC was also significantly higher after colonoscopy performed by an endoscopist with an ADR<20% as compared with an endoscopist with an ADR≥20% (HR 2.51, 95% CI 1.49 to 4.22, p<0.001). When we stratified the low-risk group of patients and patients with hyperplastic polyps both had an elevated risk when colonoscopy was performed by an endoscopist with an ADR<20% compared with an endoscopist with an ADR≥20%, although neither was statistically significant (HR 1.8, 95% CI 0.80 to 4.02, p=0.155 and HR 1.46, 95% CI 0.73 to 2.92, p=0.290, respectively). The association between ADR and PSCRC for all risk groups is shown in figure 4.

Interestingly, the strength of association with incident PSCRC was similar for ADR and risk group (standardised regression coefficients 0.36 and −0.33, respectively), indicating equal importance for predicting incidence of PSCRC.

Discussion

Currently, surveillance intervals are based on number, size and histology of colorectal lesions resected at the index colonoscopy following the observation that the presence of an adenoma increases the risk for future CRC depending on lesions’ characteristics.4 Two landmark studies revealed another important predictor for PSCRC risk, the endoscopists’ performance quality with ADR as surrogate parameter.7 8 However, the design of previous studies did not allow conclusions with regard to a possible implementation of endoscopists’ ADR into surveillance guidelines since lesions characteristics, the current basis of surveillance intervals, were not included in the model. The present study investigated the impact of lesion characteristics and endoscopy performance on the occurrence of PSCRC simultaneously in a screening population. The results show that PSCRC is associated with both ADR and affiliation to the high-risk group of patients after screening colonoscopy (ie, multiple, or large lesions, advanced dysplasia or villous histology) independently. Interactions between ADR and patient risk group were not statistically significant. Notably, for individuals in the high-risk group and after negative screening colonoscopy performed by an endoscopist with an ADR <20%, the risk for PSCRC was two times increased as compared with an endoscopist with an ADR ≥20%. These findings add to the existing evidence of the importance of endoscopists’ performance quality in screening colonoscopy. Moreover the results provide evidence to consider ADR of the endoscopist for surveillance recommendations in addition to the current practice of taking lesion characteristics into account.

The caveats of uniform surveillance intervals after baseline colonoscopy, independent of endoscopists’ performance, were addressed previously.19 20 Individuals undergoing colonoscopy by a high preforming endoscopist benefit from ‘double protection’ of a true clearing colonoscopy at baseline and shorter surveillance intervals after detection of high-risk adenomas. Those individuals having colonoscopy performed by a low detector are at higher risk for missed adenomas at baseline and return at unjustified long intervals.

In contrast to this previous study,19 our study analysed both the impact of lesions characteristics and endoscopists’ performance on PSCRC rather than advanced neoplasias. Affiliation to the high-risk group of patients at baseline colonoscopy increased the risk of PSCRC independently of ADR of the endoscopist (assessed from his/her previous colonoscopies). The interactions between ADR and patient risk groups were not significant, indicating no effect modifications; hence, low ADR of an endoscopist and high-risk group of patients were independently associated with a higher risk of PSCRC.

In the present study, we introduced a novel, dynamic calculation of the ADR. Former studies defined ADR over all colonoscopies performed annually or within the study period. From the perspective of any given colonoscopy, this indicated that the estimated ADR of the respective endoscopist is based also on his/her future colonoscopies, which have not yet been performed. This concept violates a basic paradigm in predictive statistics that the past should not be explained by the future. Moreover, the dynamically calculated ADR is sensitive to change and may account for a possible learning effect of endoscopists over time.

Low ADR is significantly associated with PSCRC,7 8 21 irrespective of the risk group. The present study strengthen the ADR as quality measure as it confirmed the influence on PSCRC and additionally showed that incidence rates were increased among all patient groups with the strongest impact for the high-risk group and after negative colonoscopy. The median ADR was 21.8% and therefore adhered to the guidelines at that time.22 Each 1% increase in ADR was associated with a 3% decrease in risk of PSCRC. An explorative analysis similar to one performed in a previous study8 indicated that patients who underwent screening colonoscopy by physicians with an ADR in the highest quartile (>27%) had a significantly lower risk of PSCRC as compared with the lowest quartile. When grouping ADR according to the quintile grouping of Corley,8 our results agreed well with those reported by Corley et al.8 This adds evidence to the recent guidelines, which raised the target ADR to 25%.21 However, we could not identify a clear threshold where the risk of PSCRC decreases.

Other parameters such as AADR were discussed as alternative indicators to ADR.23 The availability of histological adenoma features provided the opportunity to analyse the association with these parameters on the occurrence of PSCRC in a screening colonoscopy population. Both AADR and hrADR were significantly associated with PSCRC. Even though the AADR and the newly introduced hrADR were significantly associated with the risk of PSCRC, we could not identify any advantages of these two measures over the ADR. Therefore, there is no benefit of replacing ADR with new and more complex hrADR in everyday clinical practice.

The definition of PSCRC deviates from the definitions in previous studies as patient observations were censored at the date of the actual surveillance colonoscopy, rather than the scheduled time of surveillance. There were two reasons for this modification. First, it reflects the “real life setting” since the actual surveillance interval often does not correspond to the recommended interval that is, because the patient does not show up, endoscopy capacity is limited etc. Second, censoring at three years for the high-risk group would have led to loss of many events, which are informative for our analysis. The effects of both ADR and risk group on PSCRC were consistent over time and a third of all PSCRC were observed after 3 years. Therefore, data on PSCRCs occurring beyond the 3 years provides valuable information that should not be discarded for estimating these effects.

The percentage of PSCRC (1.36 cases per 10 000 patient-years) was similar to comparable studies with prospective design (2.23–22.38 cases per 10 000 patient-years); these studies analysed patients who underwent screening colonoscopy and developed colorectal cancer at least 6 months after screening but before the scheduled surveillance colonoscopy.7 8 24 25 Studies that identified PSCRC retrospectively, hence analysed patients with colorectal cancer and looked back in time whom of these patients had undergone colonoscopy more than 6 month before diagnosis, showed higher percentage rates of PSCRC.26 27 These studies assessed the prevalence of PSCRC relative to the number of all colorectal cancers. In contrast to other countries with colorectal cancer screening programmes, there is no individual patient identifier in Austria. However, the diagnostic codes assessed at the hospital discharges are a very reliable source since surgery and chemotherapy are solely done in hospitals.

It has been shown that PSCRC is more likely to develop in individuals with an adenoma detected at index colonoscopy.6 Moreover, colorectal cancer mortality was moderately higher among individuals who had had a high-risk adenoma removed in their lives and lower among those who had a low-risk adenoma removed, as compared with the general population.26 Notably, exact size and number, crucial characteristics for the definition of high-risk patients, and screening modality (sigmoidoscopy or colonoscopy) was not documented.26 Importantly, in the present study full information on histology, size and number (exact or grouped) of detected and removed lesions were available and screening colonoscopies only were considered. Additionally, in contrast to the aforementioned study, the present study also considered negative colonoscopies and colonoscopies with hyperplastic lesions detected.

PSCRC risk was not significantly increased in the low-risk group of patients compared with negative colonoscopies, but only in the high-risk group. We found no evidence that low-risk patients in general could benefit from intensified surveillance. This is of clinical importance since the majority of endoscopists recall patients of the low-risk group too early.28 Interestingly, endoscopists with an ADR ≥20% had significantly better adherence to surveillance guidelines as compared with those <20%, who frequently recalled patients too early.28 Strikingly, the present study suggests that shorter surveillance intervals may be considered after a negative colonoscopy whenever screening colonoscopy was done by a low-performing endoscopist.

Another important finding of the present study relates to the impact of hyperplastic polyps on PSCRC. Although hyperplastic lesions are discussed as potential precursor lesions for CRC, the malignant potential is not completely resolved.29 30 The present study cohort showed that patients with hyperplastic lesions detected at screening colonoscopy might have a slightly higher risk for PSCRC as compared with negative colonoscopies. However, the 95% CI failed to exclude parity irrespective of subgrouping by the size of the lesion. Hyperplastic lesions are potential precursor lesions of serrated lesions, which are significantly associated with PSCRC, particularly in the proximal colon.31

The major strengths of the present study are the introduction of a new, dynamic calculation of the ADR, and the design of analysing the impact of both ADR and patient risk groups on PSCRC simultaneously in one large screening colonoscopy cohort. The ADR was dynamically recomputed at each colonoscopy an endoscopist performed, based on all his/her previous colonoscopies. Thus we could adjust for changes in endoscopists’ performance over time (learning curve, age effects) and at the same time avoid confounding the effect of detection of high-risk or low-risk lesions from the baseline quality of the screening colonoscopy. However, a certain minimum number of colonoscopies performed is required for the calculations of ADR and ADR may be imprecise if the number is too small. ADR was initially implemented as quality measure in screening colonoscopy.7 Latest guidelines did not distinguish between indications for colonoscopy,21 which limits the predictive value and comparability of unified thresholds.

Other strengths are the assessment of histological features of resected polyps and the consideration of screening colonoscopies only, rather than diagnostic or therapeutic procedures. The study population consists of individuals at average risk for colorectal cancer 50 years or older, and individuals 30–49 years with family history of colorectal cancer or if they report fear of cancer. However, the group of individuals younger than 50 years was low (2085 individuals (0.59%) 30–39 years, and 8576 individuals (2.43%) 40–49 years). Therefore the results are well generalisable for other screening cohorts.

A limitation is the grouping of lesions count in the colonoscopy report form into the categories 1, 2–4 and >4 at the beginning of the study period. A data analysis of colonoscopies performed within the quality assurance programme after implementation of the exact adenoma count showed that the vast majority of this group comprised colonoscopies with two lesions detected. This analysis can certainly not be fully applied to the cohort in the present study, however, since it includes mostly the same endoscopists and a similar screening population the conclusion seems admissible to the population included in the study. Hence, colonoscopies with 2–4 lesions detected were considered low-risk, corresponding most closely to the guidelines.4 Moreover, bowel preparation quality has not been assessed before 2012. Since its implementation into the colonoscopy report form, preparation quality was adequate in >90% of examinations and therefore according to the guidelines.21 Another limitation is the follow-up of 55 months, which might surrogate an explanation for the higher PSCRC rate in high-risk individuals. However, if screening was performed with high quality, and all existing lesions are detected and completely removed, those individuals should not develop cancer within 3 years.4 Another important limitation is the possibility of missing data. First, because of the lack of an obligatory quality control programme, the study cohort was limited to screening colonoscopies performed within the existing national quality assurance programme in which approximately half of all endoscopic centres participate. Second, screenees must provide written informed consent for data transfer of the colonoscopy report form to the quality assurance programme and usage for scientific purpose in an anonymised form. Since data are processed in an anonymised form, it is unlikely that informed consent is refused, however, there remains a possibility. Third, there is a residual risk for misclassification through wrong ICD code recording. Finally, there is a possibility that individuals developed colorectal cancer and died from it without being hospitalised, in which case the cancer diagnosis world not have been assessed within the study. However, since the entire Austrian population has mandatory insurance coverage, this scenario is highly unlikely.

In summary, patients with high-risk lesions who received screening colonoscopy from an endoscopist with a low ADR had the highest risk for PSCRC. To characterise performance quality of an endoscopist, we introduced a novel, dynamic calculation of the ADR, which accounts for the possible changes over time. Endoscopists who have an ADR <20% should receive interventions to improve performance since the risk for PSCRC was higher in all patient risks groups, even tough the difference was only significant after negative colonoscopy and for high-risk patients. The decision about surveillance intervals after screening colonoscopy may include endoscopists’ ADR in addition to lesion characteristics.

Data availability statement

All data relevant to the study are included in the article or uploaded as online supplemental information.

Ethics statements

Ethics approval

The Ethics Committee of the Medical University of Vienna approved the study (EK 1323/2015).

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Correction notice This article has been corrected since it published Online First. The first sentence of the result section has the study period added.

  • Contributors EW: data management, writing of the manuscript, intellectual input; DP: critical revision of the manuscript, intellectual input; HS: Statistical analysis, critical revision of the manuscript; GH: Statistical analysis, critical revision of the manuscript; CR: Statistical analysis; LJ: Statistical analysis; BM: data management, intellectual input; AH: data management, intellectual input; MT: critical revision of the manuscript, intellectual input; MF: critical revision of the manuscript, intellectual input, data management.

  • Funding The quality certificate for screening colonoscopy (Qualitätszeritfikat Darmkrebsvorsorge) is supported by the Main Association of Statutory Insurance Institutions, The Austrian Society for Gastroenterology and Hepatology and the Austrian Cancer Aid.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research. Refer to the Methods section for further details.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.