- Email: [email protected]
Colorectal Cancer Incidence and Mortality After Removal of Adenomas During Screening Colonoscopies
Journal Pre-proof Colorectal Cancer Incidence and Mortality After Removal of Adenomas During Screening Colonoscopies Paulina Wieszczy, Michal F. Kaminski, Robert Franczyk, Magnus Loberg, Jarek Kobiela, Maria Rupinska, Bartlomiej Kocot, Maciej Rupinski, Oyvind Holme, Urszula Wojciechowska, Joanna Didkowska, David Ransohoff, Michael Bretthauer, Mette Kalager, Jaroslaw Regula PII: DOI: Reference:
S0016-5085(19)41344-9 https://doi.org/10.1053/j.gastro.2019.09.011 YGAST 62893
To appear in: Gastroenterology Accepted Date: 19 September 2019 Please cite this article as: Wieszczy P, Kaminski MF, Franczyk R, Loberg M, Kobiela J, Rupinska M, Kocot B, Rupinski M, Holme O, Wojciechowska U, Didkowska J, Ransohoff D, Bretthauer M, Kalager M, Regula J, Colorectal Cancer Incidence and Mortality After Removal of Adenomas During Screening Colonoscopies, Gastroenterology (2019), doi: https://doi.org/10.1053/j.gastro.2019.09.011. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 by the AGA Institute
Current high-risk group: ≥10 mm or HGD* or ≥3 or villous
About 40% of adenoma patients are high-risk
Novel high-risk group: ≥20 mm or HGD*
About 10% of adenoma patients are high-risk
Amount of surveillance may be reduced without increasing the risk of colorectal cancer and death
*high-grade dysplasia
Manuscript Number: GASTRO-D-19-00169
Title: Colorectal Cancer Incidence and Mortality After Removal of Adenomas During Screening Colonoscopies Short title: Colorectal cancer risk after adenoma removal
Paulina Wieszczy1,2,10, Michal F Kaminski1,2,3, Robert Franczyk2,4, Magnus Loberg3,5, Jarek Kobiela6, Maria Rupinska1,2, Bartlomiej Kocot1, Maciej Rupinski1,2, Oyvind Holme3,9, Urszula Wojciechowska7, Joanna Didkowska7, David Ransohoff8, Michael Bretthauer3,5, Mette Kalager*3,5, Jaroslaw Regula*1,2
1
Department of Gastroenterology, Hepatology and Clinical Oncology, Centre of Postgraduate Medical
Education, Warsaw, Poland 2
Department of Oncological Gastroenterology and Department of Cancer Prevention, the Maria
Sklodowska-Curie Memorial Cancer Center and Institute of Oncology, Warsaw, Poland 3
Clinical Effectiveness Research Group, Institute of Health and Society, University of Oslo and Department
of Transplantation Medicine Oslo University Hospital, Oslo, Norway Oslo, Norway 4
Department of Descriptive and Clinical Anatomy, Center of Biostructure Research, Medical University of
Warsaw, Poland 5
Department of Transplantation Medicine and K. G. Jebsen Center for Colorectal Cancer Research, Oslo
University Hospital, Oslo, Norway 6
Department of General, Endocrine and Transplant Surgery, Medical University of Gdansk, Gdansk, Poland
7
National Cancer Registry of Poland, the Maria Sklodowska-Curie Memorial Cancer Center and Institute of
Oncology, Warsaw, Poland 8
Departments of Medicine and Epidemiology, University of North Carolina at Chapel Hill, Chapel Hill NC,
USA 9
Department of Medicine, Sorlandet Hospital Kristiansand, Kristiansand, Norway
1
10
Frontier Science Foundation, Boston, MA, USA
* MK and JR contributed equally.
Grant Support The study was supported by research grants from the Centre of Postgraduate Medical Education, the Foundation for Polish Science, the Innovative Economy Operational Programme, the Polish Foundation of Gastroenterology, the Polish Ministry of Health, the KG Jebsen Foundation, and the Norwegian Research Council. PW received stipends from the Centre of Postgraduate Medical Education, Polpharma Scientific Foundation and Fulbright Commission. The funders had no role in the design, analyses or interpretation of the study. Correspondence Paulina Wieszczy, MSc Department of Gastroenterology, Hepatology and Clinical Oncology Centre of Postgraduate Medical Education Roentgen Street 5, Warsaw, Poland Phone: +48 22 546 30 58; Fax: +48 22 546 30 57 Email: [email protected] Disclosures We report no potential conflicts of interest relevant to this article. Author Contributions The study was conceived by PW, MFK, RF, MK and JR. All authors contributed to the development of the protocol and data collection. PW, MFK, MK and JR analyzed the data. PW wrote the first draft of the manuscript. All authors contributed to the interpretation of the data and the drafting of the paper. PW, MFK and JR had full access to all study data, and take responsibility for its integrity, and the accuracy of data analyses. Word count: 412 (abstract); 3,002 (running text) Number of tables: 3 tables + 2 figures + appendix 2
Abstract: Background & Aims: Recommendation of surveillance colonoscopy should be based on risk of colorectal cancer and death after adenoma removal. We aimed to develop risk classification system based on colorectal cancer incidence and mortality following adenoma removal.
Methods: We performed a multicenter population-based cohort study of 236,089 individuals (median patient age, 56 years; 37.8% male) undergoing screening colonoscopies with adequate bowel cleansing and cecum intubation at 132 centers in the Polish National Colorectal Cancer Screening Program, from 2000 through 2011. Subjects were followed for a median 7.1 years and information was collected on colorectal cancer development and death. We used recursive partitioning and multivariable Cox models to identify associations between colorectal cancer risk and patient and adenoma characteristics (diameter, growth pattern, grade of dysplasia and number of adenomas). We developed a risk classification system based on standardized incidence ratios, using data from the Polish population for comparison. The primary endpoints were colorectal cancer incidence and colorectal cancer death.
Results: We identified 130 colorectal cancers in individuals who had adenomas removed at screening (46.5 per 100,000 person-years) vs 309 in individuals without adenomas (22.2 per 100,000 person-years). Compared with individuals without adenomas, adenomas ≥20 mm in diameter and high-grade dysplasia were associated with increased risk of colorectal cancer (adjusted hazard ratios, 9.25; 95% CI, 6.39–13.39 and 3.58; 95% CI, 1.96–6.54, respectively). Compared with the general population, colorectal cancer risk was higher or comparable only for individuals with adenomas ≥20 mm in diameter (standardized incidence ratio [SIR], 2.07; 95% CI, 1.40–2.93) or with high-grade dysplasia (SIR, 0.79; 95% CI, 0.39–1.41), whereas for individuals with other adenoma characteristics the risk was lower (SIR, 0.35; 95% CI, 0.28–0.44). We developed a high-risk classification based on adenoma size ≥20 mm or high-grade dysplasia (instead of the current high-risk classification cutoff of ≥3 adenomas or any adenoma with villous growth pattern, highgrade dysplasia, or ≥10 mm in diameter). Our classification system would reduce the number of individuals classified as high-risk and requiring intensive surveillance from 15,242 (36.5%) to 3980 (9.5%), without 3
increasing risk of colorectal cancer in patients with adenomas (risk difference per 100,000 person-years, 5.6; 95% CI, –10.7 to 22.0).
Conclusions: Using data from the Polish National Colorectal Cancer Screening Program, we developed a risk classification system that would reduce the number of individuals classified as high risk and require intensive surveillance more than 3-fold, without increasing risk of colorectal cancer in patients with adenomas. This system could optimize the use of surveillance colonoscopy.
KEY WORDS: colon cancer, tumor, neoplasm, early detection
4
What you need to know: BACKGROUND AND CONTEXT: Recommendations for surveillance colonoscopy should be based on risk of colorectal cancer and death after adenoma removal. We developed a risk classification system based on colorectal cancer incidence and mortality following adenoma removal. NEW FINDINGS: Using data from the Polish National Colorectal Cancer Screening Program, we developed a risk classification system that would reduce the number of individuals classified as high risk (who require intensive surveillance) more than 3-fold, without increasing risk of colorectal cancer in patients with adenomas. LIMITATIONS: Further prospective studies are needed to validate our findings. IMPACT: This system could optimize the use of surveillance colonoscopy.
Lay Summary: Risk of colorectal cancer is increased in patients found to have large adenomas (more than 20 mm in diameter) or high-grade dysplasia during screening colonoscopies. These individuals should receive intensive surveillance for colorectal cancer.
5
INTRODUCTION Colonoscopy screening aims to reduce colorectal cancer incidence and mortality by removal of adenomas and early detection of cancer (1, 2). Because individuals who have had adenomas removed are believed to be at increased risk of developing new adenomas or cancer in the future (3-6), current guidelines recommend frequent colonoscopy surveillance for these patients (7-11). As adenomas have a high prevalence in average-risk populations (>20%) (12, 13), with increased screening activity around the world, the number of individuals referred for surveillance is rapidly increasing. Increasing demand for surveillance limits the availability of colonoscopy resources for diagnostic and screening purposes. There is a need to optimize the use of surveillance colonoscopy based on well quantified risk of colorectal cancer and death after adenoma removal.
Only two randomized trials comparing different surveillance intervals after adenoma removal are available (14, 15). One of these trials used surrogate endpoints (adenomas, not cancer incidence or mortality) (14) and the other was small in size (15). Third large randomized trial comparing different surveillance intervals has been launched, but results will not be available before 2029 (16). Most observational studies investigating surveillance after adenoma removal are hampered by small sample size, use of surrogate endpoints, lack of data to verify quality of colonoscopy and adenoma characteristics or immortal time bias (3, 6, 17-22).
Current surveillance recommendations categorize adenomas into high-risk or low-risk groups according to size, number, growth pattern and grade of dysplasia (7-11). The evidence for this classification system is weak and may not optimally discriminate high-risk requiring intensive surveillance from low-risk patients requiring colonoscopy after 5-10 years, leading to wasted resources and poor cost-effectiveness of current surveillance recommendations.
6
We took advantage of a nationwide high-quality screening colonoscopy program (23) to examine risk factors for colorectal cancer after adenoma removal, and to compare the novel risk model with the current risk model.
METHODS Study design Between 2000 and 2011, the Polish National Colorectal Cancer Screening Program offered colonoscopy as a primary screening tool (23). A total of 132 screening centers were participating in the program. All inhabitants aged 50-66 years with no symptoms of colorectal cancer were advised to undergo colonoscopy screening every 10 years. Individuals with family history of colorectal cancer were eligible starting at age 40 years. Data from all screening colonoscopies are stored in a dedicated database, and all individuals are identified by a national identifier. The following information is registered in the screening database: sex, age, family history of colorectal cancer, intubation depth at screening colonoscopy, quality of bowel preparation, number, location and characteristics of each adenoma (diameter, growth pattern, grade of dysplasia) detected at screening, presence of sessile serrated polyps/traditional serrated adenomas, and completeness of removal as assessed by endoscopist and/or histopathologist. Most lesions > 1 cm in size are removed in referral centers during a dedicated procedure, and the data are registered in the screening database (23). Every year, quality checks by external auditors are performed in the screening database and missing data are completed.
Each individual who had an adenoma removed at screening received a written surveillance recommendation, based on US guidelines (24, 25). Individuals with 3 or more adenomas, any adenoma with villous growth pattern, high-grade dysplasia, or size ≥10 mm in diameter were defined as high-risk group and individuals with 1 or 2 tubular adenomas <10 mm in diameter without high-grade dysplasia were defined as low-risk group. In the high-risk group surveillance was recommended 3 years after first colonoscopy (intensive surveillance) and in the low-risk group, 5-10 years (non-intensive surveillance). For patients who underwent piecemeal resection of an adenoma, the first recommended surveillance was 2-3 months after the resection. 7
We retrieved data on all individuals who underwent screening colonoscopy between January 1st, 2000 and December 31st, 2011 from the screening database. We excluded individuals with colorectal cancer detected at screening colonoscopy, a diagnosis of familial adenomatous polyposis or Lynch syndrome, and those with suboptimal quality of colonoscopy (cecum intubation failure; inadequate bowel cleansing, defined as “poor”, or “very poor” on the Aronchick scale (26); incompletely removed adenomas; or lack of histological polyp assessment) (Appendix Figure 1). We considered all polyps removed within a year after screening as polyps removed at screening.
By linkage, we retrieved date of colorectal cancer diagnosis (histologically verified colorectal adenocarcinoma; ICD-10 codes C18-C20), and date and cause of death from the Polish Cancer Registry and the Population Registry, using the national identifier. Population level data on colorectal cancer incidence and mortality were obtained from the National Cancer Registry in Poland (27).
Colorectal cancer incidence and colorectal cancer death were the primary endpoints of the study. Observation was initiated at the time of screening colonoscopy and censored at the time of death or end of follow-up (December 31st, 2014), whichever occurred first. The study was approved by the institutional ethics committee (53/PB/2017).
Statistical analysis Risk In estimates of risk, the comparator is not obvious, as a consequence we compared the risk in the different risk groups to both individuals without adenomas (Cox proportional hazard ratios) and the general population (standardized incidence ratios). We calculated standardized colorectal cancer incidence and incidence-based mortality ratios as the observed number of colorectal cancer or deaths divided by the expected number (details can be found in the Appendix). The absolute risk was calculated as a number of events per 100,000 person-years of follow-up. 8
Building a novel classification risk system We used the data for individuals with any adenoma removed at screening to identify the novel risk classification system and used multivariable Cox models to identify associations between colorectal cancer risk and patient and adenoma characteristics (7-11), i.e. age (<60, ≥60 years old), sex, family history of colorectal cancer (I⁰ relatives, none), adenoma diameter (<10 mm, 10-19 mm, ≥20 mm), growth pattern (tubular vs. villous or tubulo-villous), grade of dysplasia (high vs. low-grade), number of adenomas (1-2, ≥3). Adenoma diameter was categorized into 3 groups as it was shown that there is a terminal digit preference bias in polyp size measurement in endoscopy (28). Also number of adenomas was categorized into 2 groups because of low number of cancers among patients with ≥5 adenomas. We used the recursive partitioning method to develop the novel classification system (29). The algorithm was as follows: in a multivariable model, the risk factor with the highest hazard ratio (HR) for colorectal cancer were identified. We then divided the dataset into two parts; all individuals with that risk factor were defined as one group, and all other individuals were kept for further analyzes and used to build a second multivariable model (stepwise process, step 1-2, see Appendix). This process was iterated until no more risk groups could be separated out. We used forward stepwise regression and likelihood-ratio test, with a P-value less than 0.1 for inclusion into the multivariable models and reported adjusted HR and 95% confidence intervals (CI) for all variables included in the models.
Validation of the model was performed using bootstrap re-sampling method with 1,000 bootstrap samples. Current classification model and novel model were compared using Harrell's Concordance Statistic.
Sensitivity analyses Since removal of adenomas detected at surveillance colonoscopy may alter the risk of cancer, we obtained dates of colonoscopies and polypectomies reimbursed by the National Health System for all individuals with a screening colonoscopy in 2007-2008 (these data were unavailable for the rest of the cohort). Data on surveillance included procedure code (colonoscopy/colonoscopy with polypectomy), but not histology 9
reports. To account for the change in risk of adenoma removal, we did a sensitivity analysis including the cohort of individuals screened in 2007-2008 and censored follow-up for all patients with adenoma removal at the date of the polypectomy. Since the cancer risk can be altered only by adenoma removal, censoring on first polypectomy enabled to remove the effect of adenoma removal on cancer risk (see details in the Appendix). Finally, to examine the impact of sessile serrated adenomas/polyps on our results, we performed a sensitivity analysis where all individuals diagnosed with such lesions were excluded.
Overall issues For all analyses p-values less than 0.05 were considered statistically significant. All tests were two-sided. No sample size calculations were performed before the study. All analyses were performed with Stata software, version 13.1 (Stata Corporation, College Station, Texas, USA). Figures were prepared with R statistical software, version 3.0.1 (R Development Core Team, Vienna, Austria).
RESULTS Screening colonoscopy population characteristics A total of 265,688 individuals underwent screening colonoscopy in the program between 2000 and 2011. Altogether 29,599 (11.1%) individuals were excluded, leaving 236,089 individuals for analysis (Appendix Figure 1). Median age was 56 years (interquartile range 52-60 years) and 89,358 (37.8%) were males (Table 1).
The median follow-up was 7.1 years, with a maximum of 14.4 years. We identified 439 colorectal cancers during 1,669,684 person-years of observation and 132 colorectal cancer deaths during 1,671,266 personyears of observation (Table 1). In the current risk stratification system 15,242 (36.5%) individuals were classified as high-risk and 26,536 (63.5%) as low-risk.
Current risk classification system 10
Comparison with individuals without adenomas Individuals with low-risk adenomas had a 49% higher risk of colorectal cancer compared to individuals without adenomas (age adjusted HR 1.49; 95%CI 1.13-1.98, P=0.005). The ratio for individuals with highrisk adenomas was almost 3 times higher compared to individuals without adenomas (age adjusted HR 2.94; 95%CI 2.28-3.81; P<0.001) (Figure 1A, Appendix Table 2).
Comparison with general population In the current risk classification system, individuals with adenomas had significantly lower risk than general population. Individuals with low-risk adenomas had standardized incidence ratio of 0.35 (95%CI 0.26-0.45) and individuals with high risk adenomas had a standardized incidence ratio of 0.65 (95%CI 0.51-0.82) (Table 2).
Building a novel risk classification system Only individuals with adenomas In the multivariable model including all individuals with adenomas, adenoma size ≥20mm in diameter was associated with the highest risk of colorectal cancer (HR 4.83; 95%CI 3.05-7.63, P<0.001) (step 1 in Appendix Table 3 and Appendix Figure 2). After removing these individuals with adenomas ≥20mm, highgrade dysplasia had the highest risk (HR 2.06; 95%CI 1.08-3.92, P=0.028) (step 2). Neither number of adenomas (≥3 versus 1-2) nor growth pattern (tubulo-villous or villous versus tubular) were independent risk factors.
Novel risk classification system Comparison with individuals without adenomas Compared to individuals without adenomas, individuals with adenomas <20mm in diameter without highgrade dysplasia had 1.5 times higher risk of colorectal cancer (age adjusted HR 1.55; 95%CI 1.22-1.96; P<0.001), individuals with adenomas <20mm in diameter with high-grade dysplasia had 3 times higher risk of colorectal cancer (age adjusted HR 3.58; 95%CI 1.96-6.54; P<0.001) and individuals with adenomas 11
≥20mm in diameter had 9 times higher risk (age adjusted HR 9.25; 95%CI 6.39-13.39; P<0.001). Similar patterns were found for colorectal cancer death (Table 3). Bootstrap method confirmed the significance of novel risk groups (Appendix Table 5).
In the sensitivity analysis including individuals screened in 2007-2008 where individuals were censored at the time of polypectomy, we confirmed the findings and individuals with adenomas ≥20mm in diameter had 12 times higher risk of colorectal cancer than individuals with no adenomas (age adjusted HR of 12.25; 95%CI 6.32-23.76; P<0.001). Similar patterns were found for colorectal cancer death (Table 3). Exclusion of patients with sessile serrated adenomas/polyps did not affect the results (Appendix Table 6).
Comparison with general population In the novel risk classification system, individuals with adenomas <20mm in diameter and with no highgrade dysplasia had a significantly lower risk of colorectal cancer (standardized incidence ratio 0.35; 95%CI 0.28-0.44). Individuals with adenomas <20mm in diameter with high-grade dysplasia had colorectal cancer risk comparable to the general population (standardized incidence ratio 0.79; 95%CI 0.39-1.41), whereas individuals with adenomas ≥20mm in diameter had higher risk (2.07; 95%CI 1.40-2.93) (Table 2).
Novel versus current risk classification Harrell's Concordance Statistic showed fair prediction ability with comparable values for novel and current risk classification model (62.6% vs. 62.7%). We classified individuals with adenomas ≥20mm in diameter or adenomas with high-grade dysplasia as high-risk in the novel risk classification system since these individuals had higher colorectal cancer risk comparing with individuals with no adenomas and at the same time their colorectal cancer risk was not lower than the general population (Tables 2 and 3). A total of 3,980 (9.5%) individuals were classified as high-risk compared to 15,242 (36.5%) individuals in the current risk classification system, which is 74% less. The risk of colorectal cancer per 100,000 person-years among individuals who were classified as high-risk in the current risk classification system (≥3 adenomas, any adenoma with villous growth pattern, high-grade dysplasia, or size ≥10 mm in diameter) and as low-risk 12
group in the novel risk classification system (<20mm in diameter and no high-grade dysplasia) was 38.8 (30 cancers in 11,262 individuals), whereas absolute risk among individuals classified as low-risk in the current system (1-2 tubular adenomas <10 mm in diameter without high-grade dysplasia) was 33.2 (58 cancers in 26,478 individuals). This gives the risk difference between these groups of 5.6 per 100,000 person-years (95%CI -10.7-22.0) (Figure 2, Appendix Tables 7 and 8).
DISCUSSION We have developed and internally validated a novel risk-stratification system for the prediction of colorectal cancer incidence and death in individuals after adenoma removal. Our novel risk stratification system had similar predictive abilities as current model, but was more efficient in discriminating risk. The novel risk stratification system may help to optimize the use of surveillance colonoscopy resources without increasing the risk of cancer.
The novel risk system classifies individuals into three distinct categories based on adenoma size and grade of dysplasia: 1) adenomas ≥20mm in diameter, 2) adenomas <20 mm with high-grade dysplasia and 3) adenomas <20 mm without high-grade dysplasia. Compared to the general population, individuals with adenomas ≥20mm in diameter have higher risk, individuals with adenomas <20 mm with high-grade dysplasia slightly lower, and individuals with <20 mm without high-grade dysplasia have half the risk. Compared to individuals without adenomas, all categories have higher risk. This is in line with a recent observational study where size ≥20mm and high-grade dysplasia were found to be independent risk factors regardless of surveillance (22).
There is a striking lack of studies with colorectal cancer incidence and mortality as endpoints to support current adenoma surveillance strategies (7-9, 16). Most studies that inform today’s guidelines use surrogate endpoints combining advanced adenomas and cancer. These endpoints are difficult to interpret due to overdiagnosis and lead-time bias (21). The strengths of our study include nationwide design, large size,
13
high-quality colonoscopies, complete database of screening colonoscopies, and use of definitive endpoints (colorectal cancer incidence and mortality).
The Polish Cancer Registry is estimated to be 94% complete (30). The actual number of colorectal cancers could therefore be slightly different from reported. Poland has a public health care system and very little private health care. Thus, although we cannot exclude that some individuals might have had surveillance colonoscopy/polypectomy outside the public health system (not traceable in the National Health Found database), this number is likely not exceeding 5%. However, there is no reason to believe that this should introduce bias, since the Cancer Registry completeness and access to private health services should not vary among the analyzed groups. Exclusion of patients with inadequate bowel preparation and incomplete colonoscopy reduced the risk of missing adenomas, and likely better reflects the risk of cancer in patients after adenoma removal (4, 31).
Limitations of our study include the lack of an external validation dataset. However, we are not aware of any large enough validation cohort we could use. Although colorectal cancer incidence is regarded as the clinically most compelling endpoint in colorectal cancer screening and surveillance studies, it is potentially biased by lead time and overdiagnosis (32). Therefore, we also analyzed colorectal cancer mortality which is not affected by these biases, and the results were comparable. As in most studies, we did not perform a formal validation of polyp size assessment. However, polyp size assessment is part of our endoscopy training curriculum and was consistent over time. Thus, major misclassification seems unlikely.
One may argue that our results can be explained by recommended surveillance or lack thereof (24). However, the results from our sensitivity analysis confirmed our findings. Indeed, our results indicate that the novel risk classification system may not be altered by surveillance. We excluded individuals with suboptimal colonoscopy, and the Polish program has shown good quality (33). Thus, poor bowel preparation or high rates of incomplete removal of adenomas are not likely to bias our results. Given that our results
14
regarding current risk classification are in line with other studies, inaccuracy of our findings regarding novel risk classification seems highly improbable (21, 34).
Since our cohort consists of individuals that volunteered to undergo screening colonoscopy, the underlying colorectal cancer risk may differ from that in the general population. Baseline risk of colorectal cancer during the follow-up in our cohort may seem to be low (26.3 per 100,000 person-years), but we included only high-quality exams of relatively young (median age 56 years) and predominantly female individuals. However, standardized incidence rates are comparable to those reported in recent studies (22). Further, the observed adenoma detection rate of 17.7% and serrated adenoma detection rate of 1.0% is low compared to today’s standard. Endoscopists with low adenoma detection rates may remove large polyps less effectively, which can confound the current result. However, the risk of such confounding of our results is low, because in the Polish screening program, most large polyps are removed in a referral center.
By applying the novel risk classification in surveillance guidelines, the burden of colonoscopies could be reduced considerably. To allow policy makers, physicians and individual patients to make an informed choice about surveillance strategies according to resources, values and preferences, we have estimated the absolute risk of colorectal cancer for the different categories of high-risk, and shown how the novel definition of high-risk alter the number of individuals in need of intensive surveillance (Figure 2). Optimized use of surveillance may help to shift limited colonoscopy resources to screening and diagnostics.
In summary, we showed that by applying our new high-risk definition, only 10% of individuals with adenomas may need intensive surveillance without compromising the risk for the rest of the adenoma patients.
15
ACKNOWLEDGMENTS The authors thank Hans‑Olov Adami, M.D., Ph.D., from the Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm and Marcin Polkowski from Department of Gastroenterology, Hepatology and Clinical Oncology, the Maria Sklodowska-Curie Memorial Cancer Center and Institute of Oncology, Warsaw, Poland for their editorial advice.
16
REFERENCES 1.
Winawer SJ, Zauber AG, Ho MN, O'Brien MJ, Gottlieb LS, Sternberg SS, et al. Prevention of
Colorectal Cancer by Colonoscopic Polypectomy. N Engl J Med. 1993;329(27):1977-81. 2.
Zauber AG, Winawer SJ, O'Brien MJ, Lansdorp-Vogelaar I, van Ballegooijen M, Hankey BF, et al.
Colonoscopic polypectomy and long-term prevention of colorectal-cancer deaths. N Engl J Med. 2012;366(8):687-96. 3.
Yamaji Y, Mitsushima T, Ikuma H, Watabe H, Okamoto M, Kawabe T, et al. Incidence and
recurrence rates of colorectal adenomas estimated by annually repeated colonoscopies on asymptomatic Japanese. Gut. 2004;53(4):568-72. 4.
Loeve F, van Ballegooijen M, Boer R, Kuipers EJ, Habbema JD. Colorectal cancer risk in adenoma
patients: a nation-wide study. Int J Cancer. 2004;111(1):147-51. 5.
Cottet V, Jooste V, Fournel I, Bouvier AM, Faivre J, Bonithon-Kopp C. Long-term risk of colorectal
cancer after adenoma removal: a population-based cohort study. Gut. 2012;61(8):1180-6. 6.
Martinez ME, Baron JA, Lieberman DA, Schatzkin A, Lanza E, Winawer SJ, et al. A pooled analysis
of advanced colorectal neoplasia diagnoses after colonoscopic polypectomy. Gastroenterology. 2009;136(3):832-41. 7.
Atkin WS, Valori R, Kuipers EJ, Hoff G, Senore C, Segnan N, et al. European guidelines for quality
assurance in colorectal cancer screening and diagnosis. First Edition--Colonoscopic surveillance following adenoma removal. Endoscopy. 2012;44 Suppl 3:SE151-63. 8.
Hassan C, Quintero E, Dumonceau JM, Regula J, Brandao C, Chaussade S, et al. Post-polypectomy
colonoscopy surveillance: European Society of Gastrointestinal Endoscopy (ESGE) Guideline. Endoscopy. 2013;45(10):842-64. 9.
Levin B, Lieberman DA, McFarland B, Andrews KS, Brooks D, Bond J, et al. Screening and
surveillance for the early detection of colorectal cancer and adenomatous polyps, 2008: a joint guideline from the American Cancer Society, the US Multi-Society Task Force on Colorectal Cancer, and the American College of Radiology. Gastroenterology. 2008;134(5):1570-95.
17
10.
Cairns SR, Scholefield JH, Steele RJ, Dunlop MG, Thomas HJ, Evans GD, et al. Guidelines for
colorectal cancer screening and surveillance in moderate and high risk groups (update from 2002). Gut. 2010;59(5):666-89. 11.
van Heijningen EM, Lansdorp-Vogelaar I, Steyerberg EW, Goede SL, Dekker E, Lesterhuis W, et al.
Adherence to surveillance guidelines after removal of colorectal adenomas: a large, community-based study. Gut. 2015;64(10):1584-92. 12.
Ferlitsch M, Reinhart K, Pramhas S, Wiener C, Gal O, Bannert C, et al. Sex-specific prevalence of
adenomas, advanced adenomas, and colorectal cancer in individuals undergoing screening colonoscopy. JAMA. 2011;306(12):1352-8. 13.
Brenner H, Altenhofen L, Stock C, Hoffmeister M. Expected long-term impact of the German
screening colonoscopy programme on colorectal cancer prevention: Analyses based on 4,407,971 screening colonoscopies. European Journal of Cancer. 2015;51(10):1346-53. 14.
Winawer SJ, Zauber AG, O'Brien MJ, Ho MN, Gottlieb L, Sternberg SS, et al. Randomized
Comparison of Surveillance Intervals after Colonoscopic Removal of Newly Diagnosed Adenomatous Polyps. N Engl J Med. 1993;328(13):901-6. 15.
Kronborg O, Jørgensen OD, Fenger C, Rasmussen M. Three randomized long-term surveillance
trials in patients with sporadic colorectal adenomas. Scand J Gastroenterol. 2006;41(6):737-43. 16.
Jover R, Bretthauer M, Dekker E, Holme O, Kaminski MF, Loberg M, et al. Rationale and design of
the European Polyp Surveillance (EPoS) trials. Endoscopy. 2016;48(6):571-8. 17.
Chung SJ, Kim YS, Yang SY, Song JH, Kim D, Park MJ, et al. Five-year risk for advanced
colorectal neoplasia after initial colonoscopy according to the baseline risk stratification: a prospective study in 2452 asymptomatic Koreans. Gut. 2011;60(11):1537-43. 18.
Lieberman DA, Weiss DG, Harford WV, Ahnen DJ, Provenzale D, Sontag SJ, et al. Five-year colon
surveillance after screening colonoscopy. Gastroenterology. 2007;133(4):1077-85. 19.
Cafferty FH, Wong JM, Yen AM, Duffy SW, Atkin WS, Chen TH. Findings at follow-up
endoscopies in subjects with suspected colorectal abnormalities: effects of baseline findings and time to follow-up. Cancer J. 2007;13(4):263-70. 18
20.
Saini SD, Kim HM, Schoenfeld P. Incidence of advanced adenomas at surveillance colonoscopy in
patients with a personal history of colon adenomas: a meta-analysis and systematic review. Gastrointestinal endoscopy. 2006;64(4):614-26. 21.
Loberg M, Kalager M, Holme O, Hoff G, Adami HO, Bretthauer M. Long-term colorectal-cancer
mortality after adenoma removal. N Engl J Med. 2014;371(9):799-807. 22.
Atkin W, Wooldrage K, Brenner A, Martin J, Shah U, Perera S, et al. Adenoma surveillance and
colorectal cancer incidence: a retrospective, multicentre, cohort study. The Lancet Oncology. 2017;18(6):823-34. 23.
Kaminski MF, Kraszewska E, Rupinski M, Laskowska M, Wieszczy P, Regula J. Design of the
Polish Colonoscopy Screening Program: a randomized health services study. Endoscopy. 2015. 24.
Winawer SJ, Fletcher RH, Miller L, Godlee F, Stolar MH, Mulrow CD, et al. Colorectal cancer
screening: clinical guidelines and rationale. Gastroenterology. 1997;112(2):594-642. 25.
Winawer S, Fletcher R, Rex D, Bond J, Burt R, Ferrucci J, et al. Colorectal cancer screening and
surveillance: clinical guidelines and rationale-Update based on new evidence. Gastroenterology. 2003;124(2):544-60. 26.
Aronchick C, Lipshutz W, Wright S, Dufrayne F, Bergman G. Validation of an instrument to assess
colon cleansing. The American journal of gastroenterology. 1999;94:2667. 27.
National Cancer Registry. Available from: http://onkologia.org.pl/raporty/.
28.
Plumb AA, Nickerson C, Wooldrage K, Bassett P, Taylor SA, Altman D, et al. Terminal digit
preference biases polyp size measurements at endoscopy, computed tomographic colonography, and histopathology. Endoscopy. 2016;48(10):899-908. 29.
Zhang H. Recursive Partitioning and Applications. 2nd ed. ed. Singer BH, editor. New York, NY:
Springer New York; 2010. 30.
Didkowska J, Wojciechowska U. Cancer in Poland in 2013. Warsaw: National Cancer Registry of
Poland, the Maria Sklodowska-Curie Memorial Cancer Center and Institute of Oncology; 2015.
19
31.
Froehlich F, Wietlisbach V, Gonvers JJ, Burnand B, Vader JP. Impact of colonic cleansing on
quality and diagnostic yield of colonoscopy: the European Panel of Appropriateness of Gastrointestinal Endoscopy European multicenter study. Gastrointestinal endoscopy. 2005;61(3):378-84. 32.
Bretthauer M, Kalager M. Principles, effectiveness and caveats in screening for cancer. The British
journal of surgery. 2013;100(1):55-65. 33.
Kaminski MF, Wieszczy P, Rupinski M, Wojciechowska U, Didkowska J, Kraszewska E, et al.
Increased Rate of Adenoma Detection Associates With Reduced Risk of Colorectal Cancer and Death. Gastroenterology. 2017;153(1):98-105. 34.
Click B, Pinsky PF, Hickey T, Doroudi M, Schoen RE. Association of Colonoscopy Adenoma
Findings With Long-term Colorectal Cancer Incidence. Jama. 2018;319(19):2021-31.
20
Figure legends Figure 1. Cumulative hazard rates for colorectal cancer in current and novel risk groups and no adenoma group (panel A – current risk groups; panel B – novel risk groups) The graph shows cumulative hazard rates for colorectal cancer in the current and novel risk groups and in the no adenoma group. Current risk groups are defined as: low-risk (1-2 tubular adenomas <10 mm in diameter without high-grade dysplasia); high-risk (3 or more adenomas, any adenoma with villous growth pattern, high-grade dysplasia, or size ≥10 mm in diameter). Novel risk groups are defined as: adenomas ≥20 mm in diameter; adenomas <20 mm in diameter with high-grade dysplasia; adenomas <20 mm in diameter without high-grade dysplasia. Abbreviations: HGD – high-grade dysplasia.
Figure 2. Risk of colorectal cancer in different risk group definitions One circle represents approximately 418 individuals. Red circles represent proportion of individuals in the novel high-risk group (≥20 mm in diameter or high-grade dysplasia), light blue circles represent proportion of individuals who are moved from the current high-risk group (3 or more adenomas, any adenoma with villous growth pattern, high-grade dysplasia, or size ≥10 mm in diameter) to the novel low-risk group (<20 mm in diameter without high-grade dysplasia), dark blue circles represent proportion of individuals in current low-risk group (1-2 tubular adenomas <10 mm in diameter without high-grade dysplasia). Corresponding bars show colorectal cancer risk per 100,000 person-years. Risk difference between current low-risk and those moved to the novel low-risk group was 5.6 (95% CI -10.7-22.0). Abbreviations: p-yrs – person-years.
21
Table 1. Baseline characteristic of individuals included in the study Individuals, no. (%)
CRC, no.
CRC risk, %
P-years
236,089
439
0.19
1,669,684
CRC risk per 100,000 p-yrs 26.3
40-59
171,081 (72.5)
261
0.15
1,226,992
60-66
65,008 (27.5)
178
0.27
Female
146,731 (62.2)
260
Male
89,358 (37.8)
None
CRC death, no.
CRC death risk, %
P-years
132
0.06
1,671,226
CRC death risk per 100,000 p-yrs 7.9
21.3
66
0.04
1,227,891
5.4
442,692
40.2
66
0.10
443,375
14.9
0.18
1,046,372
24.8
73
0.05
1,047,259
7.0
179
0.20
623,312
28.7
59
0.07
624,007
9.5
189,714 (80.4)
358
0.19
1,333,197
26.9
112
0.06
1,334,516
8.4
I⁰ relatives
46,375 (19.6)
81
0.17
336,487
24.1
59
0.13
336,750
17.5
Any adenoma
41,778 (17.7)
130
0.31
279,585
46.5
34
0.08
280,059
12.1
< 10 mm
33,927 (81.2)
76
0.22
225,548
33.7
20
0.06
225,804
8.9
10-19 mm
5,753 (13.8)
23
0.40
39,578
58.1
6
0.10
39,660
15.1
≥ 20 mm
2,098 (5.0)
31
1.48
14,459
214.4
8
0.38
14,594
54.8
Tubular
33,803 (80.9)
96
0.28
224,166
42.8
27
0.08
224,506
12.0
Villous/tubulo-villous
7,975 (19.1)
34
0.43
55,419
61.4
7
0.09
55,553
12.6
Variable Total Age group
Sex
Family history of CRC
Size of adenoma
Growth pattern
Grade of dysplasia
Low
39,290 (94.0)
104
0.26
262,293
39.7
27
0.07
262,648
10.3
High
2,488 (6.0)
26
1.05
17,293
150.3
7
0.28
17,410
40.2
1-2
38,693 (92.6)
120
0.31
258,002
46.5
33
0.09
258,429
12.8
3-4
2,439 (5.8)
9
0.37
16,920
53.2
1
0.04
16,960
5.9
≥5
646 (1.5)
1
0.15
4,664
21.4
0
0
4,669
0
Low-risk
26,536 (63.5)
58
0.22
174,743
33.2
17
0.06
174,929
9.7
High-risk
15,242 (36.5)
72
0.47
104,842
68.7
17
0.11
105,129
16.2
Number of adenomas
Current risk groups
Abbreviations: CRC – colorectal cancer, p-yrs – person-years of observation. Low-risk refers to individuals with 1-2 tubular adenomas <10 mm in diameter without highgrade dysplasia. High-risk refers to individuals with 3 or more adenomas, or any adenoma with villous growth pattern, high-grade dysplasia, or size ≥10 mm in diameter.
Table 2. Comparison with the general population: Standardized incidence ratio for colorectal cancer and standardized incidence based mortality ratio for colorectal cancer death No. of No. of No. of SIR No. of expected SMR Variable observed expected observed CRC [95% CI] CRC deaths [95% CI] CRCs CRCs deaths All
439
1,441
0.30 [0.28-0.33]
132
653
0.20 [0.17-0.24]
No adenoma
309
1163
0.27 [0.24-0.30]
98
525
0.19 [0.15-0.23]
Low-risk group
58
168
0.35 [0.26-0.45]
17
76
0.22 [0.13-0.36]
High-risk group
72
110
0.65 [0.51-0.82]
17
51
0.33 [0.19-0.53]
<20 mm, no HGD
88
249
0.35 [0.28-0.44]
23
114
0.20 [0.13-0.30]
<20 mm, HGD
11
14
0.79 [0.39-1.41]
3
6
0.50 [0.10-1.46]
≥20 mm
31
15
2.07 [1.40-2.93]
8
7
1.14 [0.49-2.25]
Current risk stratification system
Novel risk groups definition
Abbreviations: CRC – colorectal cancer, SIR – standardized incidence ratio, SMR – standardized incidence based mortality ratio. Low-risk refers to individuals with 1-2 tubular adenomas <10 mm in diameter without high-grade dysplasia. High-risk refers to individuals with 3 or more adenomas, or any adenoma with villous growth pattern, high-grade dysplasia, or size ≥10 mm in diameter.
Table 3. Comparison with individuals without adenomas: Multivariable hazard ratios for risk of colorectal cancer (CRC) and death (CRC death) for the novel risk group definition Number of HR† HR† CRC P-value CRC death P-value individuals [95% CI] [95% CI] Primary analyses Novel risk groups definition No adenoma
194,311
309
1.00
98
1.00
<20 mm, no HGD
37,798
88
1.55 [1.22-1.96]
<0.001
23
1.32 [0.84-2.09]
0.227
<20 mm, HGD
1,881
11
3.58 [1.96-6.54]
<0.001
3
2.97 [0.94-9.38]
0.064
≥20 mm
2,099
31
9.25 [6.39-13.39]
<0.001
8
7.45 [3.62-15.33]
<0.001
No adenoma
58,362
73
1.00
30
1.00
<20 mm, no HGD
12,335
23
1.52 [0.95-2.42]
0.083
9
1.85 [1.23-2.79]
0.003
<20 mm, HGD
493
2
3.47 [0.85-14.15]
0.083
0
<0.001
≥20 mm
745
10
12.25 [6.32-23.76]
<0.001
2
8.53** [4.91-14.84]
Sensitivity analyses* Novel risk groups definition
†
Multivariable HR adjusted for age group. Sex and family history of colorectal cancer were tested for inclusion, but were not found significant. *Only individuals screened in 2007-2008 with censoring on polypectomy. **Categories ≥20 mm and <20 mm, HGD are collapsed as one category due to estimability. Abbreviations: HGD=highgrade dysplasia, CRC=colorectal cancer.
A Cumulative Colorectal Cancer Hazard [%]
5.0
High−risk Low−risk No adenoma
4.0
3.0
2.0
1.0
0 0 No. at Risk High−risk
2
4
6
8
10
5,028 7,217 72,409
2,144 2,862 32,344
6
8
10
1,400 1,224 22,900
680 742 10,823
282 252 4,472
Years 15,242 26,536 194,311
Low−risk No adenoma B
12,723 21,686 166,638
9,843 15,681 103,221
≥ 20 mm < 20 mm, HGD < 20 mm, noHGD No adenoma
5.0 Cumulative Colorectal Cancer Hazard [%]
15,049 26,323 192,993
4.0
3.0
2.0
1.0
0 0 No. at Risk ≥ 20 mm < 20 mm, HGD < 20 mm, noHGD
2
4 Years
2,009 1,881 37,798
2,059 1,853 37,460
1,766 1,579 31,064
High−risk defined as ≥ 20 mm or HGD ● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
Risk/100,000 p−yrs
150 100
50 0
New Current Risk among low−risk those moved high−risk
No. of subjects in risk groups
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
Risk/100,000 p−yrs
Current high−risk defined as HGD or ≥ 3 or villous or ≥ 10 mm
No. of subjects in risk groups
150 100 50 0
Current low−risk
Current high−risk
Supplemental material Contents Appendix 1. Standardized incidence and mortality rates Appendix 2. Sensitivity analysis Appendix Table 1. Baseline characteristic of individuals Appendix Table 2. Risk of colorectal cancer and colorectal cancer death for the current risk group definition Appendix Table 3. Risk factors of colorectal cancer in individuals with adenomas – steps of building new risk classification system Appendix Table 4. Multivariable hazard ratios for risk of colorectal cancer and colorectal cancer death for the novel risk group definition Appendix Table 5. Bootstrap model estimates for novel risk group definition Appendix Table 6. Comparison with individuals without adenomas: Multivariable hazard ratios for risk of colorectal cancer (CRC) and death (CRC death) for the novel risk group definition. Sensitivity analysis for individuals without serrated adenomas/polyps. Appendix Table 7. Colorectal cancer risk according to current and novel risk classification system Appendix Table 8. Comparison with individuals without adenomas: Multivariable hazard ratios for risk of colorectal cancer (CRC) for the risk groups based on the current and novel definitions. Appendix Figure 1. Follow-up flow-chart Appendix Figure 2. Model building based on characteristics of adenomas
1
Appendix 1. Standardized incidence and mortality rates We calculated standardized colorectal cancer incidence ratios as the observed number of colorectal cancers divided by the expected number. The expected number of colorectal cancers was defined as the product of the colorectal cancer rate in the general population and the number of person-years at risk in the study population. Incidence ratios were standardized according to 5-years age groups and sex. We computed 95% CI for standardized colorectal cancer incidence ratios under the assumption that the observed number of colorectal cancers followed a Poisson distribution. Using a similar analytical approach, we obtained values of standardized colorectal cancer mortality ratio. The expected number of deaths was calculated as number of deaths that would be observed among individuals who had cancer diagnosis after the date of screening colonoscopy (incidence based mortality). Results of this analysis can be found in Table 2.
Appendix 2. Sensitivity analysis Of the individuals screened in 2007-2008 who had adenomas ≥20mm (745 individuals), 414 (55.6%) had colonoscopy and 211 (28.3%) had polypectomy with a median time to polypectomy of 3.3 years. Among individuals with adenomas <20mm with high-grade dysplasia (493 individuals), 227 (46.0%) had colonoscopy and 109 (22.7%) of had polypectomy with a median time to polypectomy of 3.3 years. Among individuals with adenomas <20mm without high-grade dysplasia (12,335 individuals), 3,861 (31.3%) had colonoscopy and 1,557 (12.6%) had polypectomy with a median time to polypectomy of 4.4 years. The results from the analysis of colorectal cancer risk and death according to novel risk classification system, where individuals were censored at the time of polypectomy, were in line with the results from the primary analysis. Individuals with adenomas ≥20mm in diameter had 12 times higher risk of colorectal cancer (age adjusted HR of 12.25; 95%CI 6.32-23.76; P<0.001) than individuals with no adenomas.
2
Appendix Table 1. Baseline characteristic of individuals
40-59
60-66
89,358
171,081
65,008
Family history of CRC, N (%) 46,375
41,778 (17.7)
21,015 (23.5)
27,517 (16.1)
14,261 (21.9)
8,412 (18.1)
<10 mm
33,927 (81.2)
16,889 (80.4)
22,548 (81.9)
11,379 (79.8)
6,763 (80.4)
10-19 mm
5,753 (13.8)
3,073 (14.6)
3,648 (13.3)
2,105 (14.8)
1,188 (14.1)
≥20 mm
2,098 (5.0)
1,071 (5.1)
1,321 (4.8)
777 (5.4)
461 (5.5)
Tubular Villous/tubulovillous Grade of dysplasia
33,803 (80.9)
16,941 (80.6)
22,450 (81.6)
11,353 (79.6)
6,767 (80.4)
7,975 (19.1)
4,092 (19.5)
5,067 (18.4)
2,908 (20.4)
1,645 (19.6)
Low
39,290 (94.0)
19,693 (93.7)
25,960 (94.3)
13,330 (93.5)
7,877 (93.6)
High
2,488 (6.0)
1,340 (6.4)
1,557 (5.7)
931 (6.5)
535 (6.4)
1-2
38,693 (92.6)
19,101 (90.9)
25,733 (93.5)
12,960 (90.9)
7,761 (92.3)
3-4
2,439 (5.8)
1,507 (7.2)
1,408 (5.1)
1,031 (7.2)
497 (5.9)
376 (1.4)
270 (1.9)
154 (1.8)
Variable All Any adenoma
All, N (%)
Male sex, N (%)
236,089
Age group, N (%)
Size of adenoma
Growth pattern
Number of adenomas
≥5 646 (1.5) 425 (2.0) Abbreviation: CRC – colorectal cancer, N - number
3
Appendix Table 2. Multivariable hazard ratios for risk of colorectal cancer and colorectal cancer death for the current risk group definition Number of HR CRC HR CRC P-value P-value individuals [95% CI] death [95% CI] Current risk groups definition† No adenoma
194,311
309
1.00
Low-risk group
26,536
58
1.49 [1.13-1.98]
High-risk group
15,242
72
2.94 [2.28-3.81]
98
1.00
0.005
17
1.48 [0.88-2.46]
0.145
<0.001
17
2.16 [1.29-3.62]
0.004
†
HR adjusted for age group. Sex and family history of colorectal cancer were tested for inclusion, but were not found significant. Abbreviations: CRC=colorectal cancer.
4
Appendix Table 3. Risk factors of colorectal cancer in individuals with adenomas – steps of building new risk classification system HR* N CRC P-value [95% CI] Step 1† Diameter < 10 mm
33,927
76
1.00
10-19 mm
5,753
23
1.51 [0.94-2.42]
0.089
≥ 20 mm
2,098
31
4.83 [3.05-7.63]
<0.001
Low
39,290
104
1.00
High
2,488
26
2.16 [1.35-3.46]
40-59
27,517
72
1.00
60-66
14,261
58
1.60 [1.13-2.26]
Low
37,799
88
1.00
High
1,881
11
2.06 [1.08-3.92]
40-59
26,196
56
1.00
60-66
13,484
43
1.57 [1.05-2.33]
< 10 mm
33,927
76
1.00
10-19 mm
5,753
23
1.52 [0.94-2.45]
Grade of dysplasia
0.001
Age
0.008
Step 2‡ Grade of dysplasia
0.028
Age
0.027
Diameter
0.088
Step 3$ Age 40-59
1.00
60-66
1.07 [1.02-1.12]
0.002
Family history of CRC Other st
1 degree
1.00 1.75 [1.08-2.84]
0.002
*HR reported for variables found significant in stepwise selection. Age group, sex, family history of colorectal cancer, adenoma diameter, grade of dysplasia, growth pattern and no. of adenomas were tested for inclusion into model. †Models built using observations for all subject with adenomas. ‡Models built using observations for subject with adenomas <20 mm in diameter. $ Models build using observations for subjects with adenomas <20 mm in diameter and no high-grade dysplasia. Abbreviations: CRC=colorectal cancer
5
Appendix Table 4. Multivariable hazard ratios for risk of colorectal cancer and colorectal cancer death for the novel risk group definition Colorectal cancer Model coefficient
HR* [95% CI]
Colorectal cancer death P-value
Model coefficient
HR* [95% CI]
P-value
Age 40-59 60-66
1.00 0.60
1.82 [1.50-2.20]
1.00 <0.001
1.03
2.79 [1.98-3.93]
<0.001
Novel risk group No adenoma
1.00
1.00
<20 mm, no HGD
0.44
1.55 [1.22-1.96]
<0.001
0.28
1.32 [0.84-2.09]
0.227
<20 mm, HGD
1.28
3.58 [1.96-6.54]
<0.001
1.09
2.97 [0.94-9.38]
0.064
≥20 mm
2.22
9.25 [6.39-13.29]
<0.001
2.01
7.45 [3.62-15.33]
<0.001
*HR reported for variables found significant in stepwise selection. Age group, sex and family history of colorectal cancer were tested for inclusion into the models. Abbreviations: HGD=high-grade dysplasia.
6
Appendix Table 5. Bootstrap model estimates for novel risk group definition Colorectal cancer HR [95% CI]
Colorectal cancer death P-value
HR [95% CI]
P-value
Age 40-59
1.00
60-66
1.82 [1.49-2.20]
1.00 <0.001
2.80 [1.99-3.90]
<0.001
Novel risk group No adenoma
1.00
1.00
<20 mm, no HGD
1.55 [1.22-1.95]
<0.001
1.32 [0.84-2.10]
0.226
<20 mm, HGD
3.60 [1.88-6.28]
<0.001
2.97 [0-Inf]
0.908
≥20 mm
9.21 [6.36-13.46]
<0.001
7.46 [3.42-16.28]
<0.001
7
Appendix Table 6. Comparison with individuals without adenomas: Multivariable hazard ratios for risk of colorectal cancer (CRC) and death (CRC death) for the novel risk group definition. Sensitivity analysis for individuals without serrated adenomas/polyps. Number of HR† HR† CRC P-value CRC death P-value individuals [95% CI] [95% CI] Novel risk groups definition No adenoma
194,311
309
1.00
98
1.00
<20 mm, no HGD
35,621
87
1.62 [1.27-2.05]
<0.001
23
1.40 [0.89-2.20]
0.151
<20 mm, HGD
1,836
11
3.66 [2.01-6.69]
<0.001
3
3.05 [0.96-9.62]
0.058
≥20 mm
2,027
31
9.53 [6.58-13.79]
<0.001
7
6.68 [3.10-14.40]
<0.001
†
Multivariable HR adjusted for age group. Sex and family history of colorectal cancer were tested for inclusion, but were not found significant. Abbreviations: HGD=high-grade dysplasia,
CRC=colorectal cancer.
8
Appendix Table 7. Colorectal cancer risk according to current and novel risk classification system. Novel risk classification Current risk classification
<20 mm and no HGD
≥20 mm or HGD
Total
CRC=58 CRC=0 CRC=58 no-CRC=26,478 no-CRC=0 no-CRC=26,478 Low-risk P-yrs=174,743 P-yrs=0 P-yrs=174,743 CRC risk per 100,000 p-yrs=33.2 CRC risk per 100,000 p-yrs =NA CRC risk per 100,000 p-yrs=33.2 CRC=30 CRC=42 CRC=72 no-CRC=11,232 no-CRC=3,938 no-CRC=15,170 High-risk P-yrs=77,250 P-yrs=27,592 P-yrs=104,842 CRC risk per 100,000 p-yrs=38.8 CRC risk per 100,000 p-yrs=152.2 CRC risk per 100,000 p-yrs=68.7 CRC=88 CRC=42 CRC=130 no-CRC=37,710 no-CRC=3,938 no-CRC=41,648 Total P-yrs=251,993 P-yrs=27,592 P-yrs=279,585 CRC risk per 100,000 p-yrs=34.9 CRC risk per 100,000 p-yrs=152.2 CRC risk per 100,000 p-yrs=46.5 CRC=no. of individuals with colorectal cancer diagnosis during the follow-up time, no-CRC=no. of individuals without colorectal cancer diagnosis during the follow-up time, CRC=colorectal cancer, p-yrs – person-years of observation
9
Appendix Table 8. Comparison with individuals without adenomas: Multivariable hazard ratios for risk of colorectal cancer (CRC) for the risk groups based on the current and novel definitions. Number of HR† CRC P-value individuals [95% CI] No adenoma
194,311
309
1.00
Current low-risk*
26,536
58
1.49 [1.13-1.98]
0.005
Novel low-risk, current high-risk**
11,262
30
1.66 [1.14-2.42]
0.008
Novel high-risk***
3,980
42
6.54 [4.74-9.04]
<0.001
†
Age adjusted. *1-2 tubular adenomas <10 mm in diameter without high-grade dysplasia. **3 or more adenomas, any adenoma with villous growth pattern, or size 10-19 mm in diameter. ***≥20 mm in diameter or high-grade dysplasia.
10
Appendix Figure 1. Follow-up flow-chart Abbreviations: CRC=colorectal cancer, f-up=follow-up.
11
Appendix Figure 2. Model building based on characteristics of adenomas The graph shows the steps of obtaining new risk classification system. First, basing on the results of multivariable model the set was divided according to adenoma diameter (≥20 mm or <20 mm). Then, individuals with adenomas ≥20 mm in diameter were excluded and analysis were redone. Remaining individuals were divided according to grade of dysplasia. In the last step, only individuals with adenomas <20 mm in diameter with low-grade dysplasia were analyzed.
12
S0016-5085(19)41344-9 https://doi.org/10.1053/j.gastro.2019.09.011 YGAST 62893
To appear in: Gastroenterology Accepted Date: 19 September 2019 Please cite this article as: Wieszczy P, Kaminski MF, Franczyk R, Loberg M, Kobiela J, Rupinska M, Kocot B, Rupinski M, Holme O, Wojciechowska U, Didkowska J, Ransohoff D, Bretthauer M, Kalager M, Regula J, Colorectal Cancer Incidence and Mortality After Removal of Adenomas During Screening Colonoscopies, Gastroenterology (2019), doi: https://doi.org/10.1053/j.gastro.2019.09.011. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 by the AGA Institute
Current high-risk group: ≥10 mm or HGD* or ≥3 or villous
About 40% of adenoma patients are high-risk
Novel high-risk group: ≥20 mm or HGD*
About 10% of adenoma patients are high-risk
Amount of surveillance may be reduced without increasing the risk of colorectal cancer and death
*high-grade dysplasia
Manuscript Number: GASTRO-D-19-00169
Title: Colorectal Cancer Incidence and Mortality After Removal of Adenomas During Screening Colonoscopies Short title: Colorectal cancer risk after adenoma removal
Paulina Wieszczy1,2,10, Michal F Kaminski1,2,3, Robert Franczyk2,4, Magnus Loberg3,5, Jarek Kobiela6, Maria Rupinska1,2, Bartlomiej Kocot1, Maciej Rupinski1,2, Oyvind Holme3,9, Urszula Wojciechowska7, Joanna Didkowska7, David Ransohoff8, Michael Bretthauer3,5, Mette Kalager*3,5, Jaroslaw Regula*1,2
1
Department of Gastroenterology, Hepatology and Clinical Oncology, Centre of Postgraduate Medical
Education, Warsaw, Poland 2
Department of Oncological Gastroenterology and Department of Cancer Prevention, the Maria
Sklodowska-Curie Memorial Cancer Center and Institute of Oncology, Warsaw, Poland 3
Clinical Effectiveness Research Group, Institute of Health and Society, University of Oslo and Department
of Transplantation Medicine Oslo University Hospital, Oslo, Norway Oslo, Norway 4
Department of Descriptive and Clinical Anatomy, Center of Biostructure Research, Medical University of
Warsaw, Poland 5
Department of Transplantation Medicine and K. G. Jebsen Center for Colorectal Cancer Research, Oslo
University Hospital, Oslo, Norway 6
Department of General, Endocrine and Transplant Surgery, Medical University of Gdansk, Gdansk, Poland
7
National Cancer Registry of Poland, the Maria Sklodowska-Curie Memorial Cancer Center and Institute of
Oncology, Warsaw, Poland 8
Departments of Medicine and Epidemiology, University of North Carolina at Chapel Hill, Chapel Hill NC,
USA 9
Department of Medicine, Sorlandet Hospital Kristiansand, Kristiansand, Norway
1
10
Frontier Science Foundation, Boston, MA, USA
* MK and JR contributed equally.
Grant Support The study was supported by research grants from the Centre of Postgraduate Medical Education, the Foundation for Polish Science, the Innovative Economy Operational Programme, the Polish Foundation of Gastroenterology, the Polish Ministry of Health, the KG Jebsen Foundation, and the Norwegian Research Council. PW received stipends from the Centre of Postgraduate Medical Education, Polpharma Scientific Foundation and Fulbright Commission. The funders had no role in the design, analyses or interpretation of the study. Correspondence Paulina Wieszczy, MSc Department of Gastroenterology, Hepatology and Clinical Oncology Centre of Postgraduate Medical Education Roentgen Street 5, Warsaw, Poland Phone: +48 22 546 30 58; Fax: +48 22 546 30 57 Email: [email protected] Disclosures We report no potential conflicts of interest relevant to this article. Author Contributions The study was conceived by PW, MFK, RF, MK and JR. All authors contributed to the development of the protocol and data collection. PW, MFK, MK and JR analyzed the data. PW wrote the first draft of the manuscript. All authors contributed to the interpretation of the data and the drafting of the paper. PW, MFK and JR had full access to all study data, and take responsibility for its integrity, and the accuracy of data analyses. Word count: 412 (abstract); 3,002 (running text) Number of tables: 3 tables + 2 figures + appendix 2
Abstract: Background & Aims: Recommendation of surveillance colonoscopy should be based on risk of colorectal cancer and death after adenoma removal. We aimed to develop risk classification system based on colorectal cancer incidence and mortality following adenoma removal.
Methods: We performed a multicenter population-based cohort study of 236,089 individuals (median patient age, 56 years; 37.8% male) undergoing screening colonoscopies with adequate bowel cleansing and cecum intubation at 132 centers in the Polish National Colorectal Cancer Screening Program, from 2000 through 2011. Subjects were followed for a median 7.1 years and information was collected on colorectal cancer development and death. We used recursive partitioning and multivariable Cox models to identify associations between colorectal cancer risk and patient and adenoma characteristics (diameter, growth pattern, grade of dysplasia and number of adenomas). We developed a risk classification system based on standardized incidence ratios, using data from the Polish population for comparison. The primary endpoints were colorectal cancer incidence and colorectal cancer death.
Results: We identified 130 colorectal cancers in individuals who had adenomas removed at screening (46.5 per 100,000 person-years) vs 309 in individuals without adenomas (22.2 per 100,000 person-years). Compared with individuals without adenomas, adenomas ≥20 mm in diameter and high-grade dysplasia were associated with increased risk of colorectal cancer (adjusted hazard ratios, 9.25; 95% CI, 6.39–13.39 and 3.58; 95% CI, 1.96–6.54, respectively). Compared with the general population, colorectal cancer risk was higher or comparable only for individuals with adenomas ≥20 mm in diameter (standardized incidence ratio [SIR], 2.07; 95% CI, 1.40–2.93) or with high-grade dysplasia (SIR, 0.79; 95% CI, 0.39–1.41), whereas for individuals with other adenoma characteristics the risk was lower (SIR, 0.35; 95% CI, 0.28–0.44). We developed a high-risk classification based on adenoma size ≥20 mm or high-grade dysplasia (instead of the current high-risk classification cutoff of ≥3 adenomas or any adenoma with villous growth pattern, highgrade dysplasia, or ≥10 mm in diameter). Our classification system would reduce the number of individuals classified as high-risk and requiring intensive surveillance from 15,242 (36.5%) to 3980 (9.5%), without 3
increasing risk of colorectal cancer in patients with adenomas (risk difference per 100,000 person-years, 5.6; 95% CI, –10.7 to 22.0).
Conclusions: Using data from the Polish National Colorectal Cancer Screening Program, we developed a risk classification system that would reduce the number of individuals classified as high risk and require intensive surveillance more than 3-fold, without increasing risk of colorectal cancer in patients with adenomas. This system could optimize the use of surveillance colonoscopy.
KEY WORDS: colon cancer, tumor, neoplasm, early detection
4
What you need to know: BACKGROUND AND CONTEXT: Recommendations for surveillance colonoscopy should be based on risk of colorectal cancer and death after adenoma removal. We developed a risk classification system based on colorectal cancer incidence and mortality following adenoma removal. NEW FINDINGS: Using data from the Polish National Colorectal Cancer Screening Program, we developed a risk classification system that would reduce the number of individuals classified as high risk (who require intensive surveillance) more than 3-fold, without increasing risk of colorectal cancer in patients with adenomas. LIMITATIONS: Further prospective studies are needed to validate our findings. IMPACT: This system could optimize the use of surveillance colonoscopy.
Lay Summary: Risk of colorectal cancer is increased in patients found to have large adenomas (more than 20 mm in diameter) or high-grade dysplasia during screening colonoscopies. These individuals should receive intensive surveillance for colorectal cancer.
5
INTRODUCTION Colonoscopy screening aims to reduce colorectal cancer incidence and mortality by removal of adenomas and early detection of cancer (1, 2). Because individuals who have had adenomas removed are believed to be at increased risk of developing new adenomas or cancer in the future (3-6), current guidelines recommend frequent colonoscopy surveillance for these patients (7-11). As adenomas have a high prevalence in average-risk populations (>20%) (12, 13), with increased screening activity around the world, the number of individuals referred for surveillance is rapidly increasing. Increasing demand for surveillance limits the availability of colonoscopy resources for diagnostic and screening purposes. There is a need to optimize the use of surveillance colonoscopy based on well quantified risk of colorectal cancer and death after adenoma removal.
Only two randomized trials comparing different surveillance intervals after adenoma removal are available (14, 15). One of these trials used surrogate endpoints (adenomas, not cancer incidence or mortality) (14) and the other was small in size (15). Third large randomized trial comparing different surveillance intervals has been launched, but results will not be available before 2029 (16). Most observational studies investigating surveillance after adenoma removal are hampered by small sample size, use of surrogate endpoints, lack of data to verify quality of colonoscopy and adenoma characteristics or immortal time bias (3, 6, 17-22).
Current surveillance recommendations categorize adenomas into high-risk or low-risk groups according to size, number, growth pattern and grade of dysplasia (7-11). The evidence for this classification system is weak and may not optimally discriminate high-risk requiring intensive surveillance from low-risk patients requiring colonoscopy after 5-10 years, leading to wasted resources and poor cost-effectiveness of current surveillance recommendations.
6
We took advantage of a nationwide high-quality screening colonoscopy program (23) to examine risk factors for colorectal cancer after adenoma removal, and to compare the novel risk model with the current risk model.
METHODS Study design Between 2000 and 2011, the Polish National Colorectal Cancer Screening Program offered colonoscopy as a primary screening tool (23). A total of 132 screening centers were participating in the program. All inhabitants aged 50-66 years with no symptoms of colorectal cancer were advised to undergo colonoscopy screening every 10 years. Individuals with family history of colorectal cancer were eligible starting at age 40 years. Data from all screening colonoscopies are stored in a dedicated database, and all individuals are identified by a national identifier. The following information is registered in the screening database: sex, age, family history of colorectal cancer, intubation depth at screening colonoscopy, quality of bowel preparation, number, location and characteristics of each adenoma (diameter, growth pattern, grade of dysplasia) detected at screening, presence of sessile serrated polyps/traditional serrated adenomas, and completeness of removal as assessed by endoscopist and/or histopathologist. Most lesions > 1 cm in size are removed in referral centers during a dedicated procedure, and the data are registered in the screening database (23). Every year, quality checks by external auditors are performed in the screening database and missing data are completed.
Each individual who had an adenoma removed at screening received a written surveillance recommendation, based on US guidelines (24, 25). Individuals with 3 or more adenomas, any adenoma with villous growth pattern, high-grade dysplasia, or size ≥10 mm in diameter were defined as high-risk group and individuals with 1 or 2 tubular adenomas <10 mm in diameter without high-grade dysplasia were defined as low-risk group. In the high-risk group surveillance was recommended 3 years after first colonoscopy (intensive surveillance) and in the low-risk group, 5-10 years (non-intensive surveillance). For patients who underwent piecemeal resection of an adenoma, the first recommended surveillance was 2-3 months after the resection. 7
We retrieved data on all individuals who underwent screening colonoscopy between January 1st, 2000 and December 31st, 2011 from the screening database. We excluded individuals with colorectal cancer detected at screening colonoscopy, a diagnosis of familial adenomatous polyposis or Lynch syndrome, and those with suboptimal quality of colonoscopy (cecum intubation failure; inadequate bowel cleansing, defined as “poor”, or “very poor” on the Aronchick scale (26); incompletely removed adenomas; or lack of histological polyp assessment) (Appendix Figure 1). We considered all polyps removed within a year after screening as polyps removed at screening.
By linkage, we retrieved date of colorectal cancer diagnosis (histologically verified colorectal adenocarcinoma; ICD-10 codes C18-C20), and date and cause of death from the Polish Cancer Registry and the Population Registry, using the national identifier. Population level data on colorectal cancer incidence and mortality were obtained from the National Cancer Registry in Poland (27).
Colorectal cancer incidence and colorectal cancer death were the primary endpoints of the study. Observation was initiated at the time of screening colonoscopy and censored at the time of death or end of follow-up (December 31st, 2014), whichever occurred first. The study was approved by the institutional ethics committee (53/PB/2017).
Statistical analysis Risk In estimates of risk, the comparator is not obvious, as a consequence we compared the risk in the different risk groups to both individuals without adenomas (Cox proportional hazard ratios) and the general population (standardized incidence ratios). We calculated standardized colorectal cancer incidence and incidence-based mortality ratios as the observed number of colorectal cancer or deaths divided by the expected number (details can be found in the Appendix). The absolute risk was calculated as a number of events per 100,000 person-years of follow-up. 8
Building a novel classification risk system We used the data for individuals with any adenoma removed at screening to identify the novel risk classification system and used multivariable Cox models to identify associations between colorectal cancer risk and patient and adenoma characteristics (7-11), i.e. age (<60, ≥60 years old), sex, family history of colorectal cancer (I⁰ relatives, none), adenoma diameter (<10 mm, 10-19 mm, ≥20 mm), growth pattern (tubular vs. villous or tubulo-villous), grade of dysplasia (high vs. low-grade), number of adenomas (1-2, ≥3). Adenoma diameter was categorized into 3 groups as it was shown that there is a terminal digit preference bias in polyp size measurement in endoscopy (28). Also number of adenomas was categorized into 2 groups because of low number of cancers among patients with ≥5 adenomas. We used the recursive partitioning method to develop the novel classification system (29). The algorithm was as follows: in a multivariable model, the risk factor with the highest hazard ratio (HR) for colorectal cancer were identified. We then divided the dataset into two parts; all individuals with that risk factor were defined as one group, and all other individuals were kept for further analyzes and used to build a second multivariable model (stepwise process, step 1-2, see Appendix). This process was iterated until no more risk groups could be separated out. We used forward stepwise regression and likelihood-ratio test, with a P-value less than 0.1 for inclusion into the multivariable models and reported adjusted HR and 95% confidence intervals (CI) for all variables included in the models.
Validation of the model was performed using bootstrap re-sampling method with 1,000 bootstrap samples. Current classification model and novel model were compared using Harrell's Concordance Statistic.
Sensitivity analyses Since removal of adenomas detected at surveillance colonoscopy may alter the risk of cancer, we obtained dates of colonoscopies and polypectomies reimbursed by the National Health System for all individuals with a screening colonoscopy in 2007-2008 (these data were unavailable for the rest of the cohort). Data on surveillance included procedure code (colonoscopy/colonoscopy with polypectomy), but not histology 9
reports. To account for the change in risk of adenoma removal, we did a sensitivity analysis including the cohort of individuals screened in 2007-2008 and censored follow-up for all patients with adenoma removal at the date of the polypectomy. Since the cancer risk can be altered only by adenoma removal, censoring on first polypectomy enabled to remove the effect of adenoma removal on cancer risk (see details in the Appendix). Finally, to examine the impact of sessile serrated adenomas/polyps on our results, we performed a sensitivity analysis where all individuals diagnosed with such lesions were excluded.
Overall issues For all analyses p-values less than 0.05 were considered statistically significant. All tests were two-sided. No sample size calculations were performed before the study. All analyses were performed with Stata software, version 13.1 (Stata Corporation, College Station, Texas, USA). Figures were prepared with R statistical software, version 3.0.1 (R Development Core Team, Vienna, Austria).
RESULTS Screening colonoscopy population characteristics A total of 265,688 individuals underwent screening colonoscopy in the program between 2000 and 2011. Altogether 29,599 (11.1%) individuals were excluded, leaving 236,089 individuals for analysis (Appendix Figure 1). Median age was 56 years (interquartile range 52-60 years) and 89,358 (37.8%) were males (Table 1).
The median follow-up was 7.1 years, with a maximum of 14.4 years. We identified 439 colorectal cancers during 1,669,684 person-years of observation and 132 colorectal cancer deaths during 1,671,266 personyears of observation (Table 1). In the current risk stratification system 15,242 (36.5%) individuals were classified as high-risk and 26,536 (63.5%) as low-risk.
Current risk classification system 10
Comparison with individuals without adenomas Individuals with low-risk adenomas had a 49% higher risk of colorectal cancer compared to individuals without adenomas (age adjusted HR 1.49; 95%CI 1.13-1.98, P=0.005). The ratio for individuals with highrisk adenomas was almost 3 times higher compared to individuals without adenomas (age adjusted HR 2.94; 95%CI 2.28-3.81; P<0.001) (Figure 1A, Appendix Table 2).
Comparison with general population In the current risk classification system, individuals with adenomas had significantly lower risk than general population. Individuals with low-risk adenomas had standardized incidence ratio of 0.35 (95%CI 0.26-0.45) and individuals with high risk adenomas had a standardized incidence ratio of 0.65 (95%CI 0.51-0.82) (Table 2).
Building a novel risk classification system Only individuals with adenomas In the multivariable model including all individuals with adenomas, adenoma size ≥20mm in diameter was associated with the highest risk of colorectal cancer (HR 4.83; 95%CI 3.05-7.63, P<0.001) (step 1 in Appendix Table 3 and Appendix Figure 2). After removing these individuals with adenomas ≥20mm, highgrade dysplasia had the highest risk (HR 2.06; 95%CI 1.08-3.92, P=0.028) (step 2). Neither number of adenomas (≥3 versus 1-2) nor growth pattern (tubulo-villous or villous versus tubular) were independent risk factors.
Novel risk classification system Comparison with individuals without adenomas Compared to individuals without adenomas, individuals with adenomas <20mm in diameter without highgrade dysplasia had 1.5 times higher risk of colorectal cancer (age adjusted HR 1.55; 95%CI 1.22-1.96; P<0.001), individuals with adenomas <20mm in diameter with high-grade dysplasia had 3 times higher risk of colorectal cancer (age adjusted HR 3.58; 95%CI 1.96-6.54; P<0.001) and individuals with adenomas 11
≥20mm in diameter had 9 times higher risk (age adjusted HR 9.25; 95%CI 6.39-13.39; P<0.001). Similar patterns were found for colorectal cancer death (Table 3). Bootstrap method confirmed the significance of novel risk groups (Appendix Table 5).
In the sensitivity analysis including individuals screened in 2007-2008 where individuals were censored at the time of polypectomy, we confirmed the findings and individuals with adenomas ≥20mm in diameter had 12 times higher risk of colorectal cancer than individuals with no adenomas (age adjusted HR of 12.25; 95%CI 6.32-23.76; P<0.001). Similar patterns were found for colorectal cancer death (Table 3). Exclusion of patients with sessile serrated adenomas/polyps did not affect the results (Appendix Table 6).
Comparison with general population In the novel risk classification system, individuals with adenomas <20mm in diameter and with no highgrade dysplasia had a significantly lower risk of colorectal cancer (standardized incidence ratio 0.35; 95%CI 0.28-0.44). Individuals with adenomas <20mm in diameter with high-grade dysplasia had colorectal cancer risk comparable to the general population (standardized incidence ratio 0.79; 95%CI 0.39-1.41), whereas individuals with adenomas ≥20mm in diameter had higher risk (2.07; 95%CI 1.40-2.93) (Table 2).
Novel versus current risk classification Harrell's Concordance Statistic showed fair prediction ability with comparable values for novel and current risk classification model (62.6% vs. 62.7%). We classified individuals with adenomas ≥20mm in diameter or adenomas with high-grade dysplasia as high-risk in the novel risk classification system since these individuals had higher colorectal cancer risk comparing with individuals with no adenomas and at the same time their colorectal cancer risk was not lower than the general population (Tables 2 and 3). A total of 3,980 (9.5%) individuals were classified as high-risk compared to 15,242 (36.5%) individuals in the current risk classification system, which is 74% less. The risk of colorectal cancer per 100,000 person-years among individuals who were classified as high-risk in the current risk classification system (≥3 adenomas, any adenoma with villous growth pattern, high-grade dysplasia, or size ≥10 mm in diameter) and as low-risk 12
group in the novel risk classification system (<20mm in diameter and no high-grade dysplasia) was 38.8 (30 cancers in 11,262 individuals), whereas absolute risk among individuals classified as low-risk in the current system (1-2 tubular adenomas <10 mm in diameter without high-grade dysplasia) was 33.2 (58 cancers in 26,478 individuals). This gives the risk difference between these groups of 5.6 per 100,000 person-years (95%CI -10.7-22.0) (Figure 2, Appendix Tables 7 and 8).
DISCUSSION We have developed and internally validated a novel risk-stratification system for the prediction of colorectal cancer incidence and death in individuals after adenoma removal. Our novel risk stratification system had similar predictive abilities as current model, but was more efficient in discriminating risk. The novel risk stratification system may help to optimize the use of surveillance colonoscopy resources without increasing the risk of cancer.
The novel risk system classifies individuals into three distinct categories based on adenoma size and grade of dysplasia: 1) adenomas ≥20mm in diameter, 2) adenomas <20 mm with high-grade dysplasia and 3) adenomas <20 mm without high-grade dysplasia. Compared to the general population, individuals with adenomas ≥20mm in diameter have higher risk, individuals with adenomas <20 mm with high-grade dysplasia slightly lower, and individuals with <20 mm without high-grade dysplasia have half the risk. Compared to individuals without adenomas, all categories have higher risk. This is in line with a recent observational study where size ≥20mm and high-grade dysplasia were found to be independent risk factors regardless of surveillance (22).
There is a striking lack of studies with colorectal cancer incidence and mortality as endpoints to support current adenoma surveillance strategies (7-9, 16). Most studies that inform today’s guidelines use surrogate endpoints combining advanced adenomas and cancer. These endpoints are difficult to interpret due to overdiagnosis and lead-time bias (21). The strengths of our study include nationwide design, large size,
13
high-quality colonoscopies, complete database of screening colonoscopies, and use of definitive endpoints (colorectal cancer incidence and mortality).
The Polish Cancer Registry is estimated to be 94% complete (30). The actual number of colorectal cancers could therefore be slightly different from reported. Poland has a public health care system and very little private health care. Thus, although we cannot exclude that some individuals might have had surveillance colonoscopy/polypectomy outside the public health system (not traceable in the National Health Found database), this number is likely not exceeding 5%. However, there is no reason to believe that this should introduce bias, since the Cancer Registry completeness and access to private health services should not vary among the analyzed groups. Exclusion of patients with inadequate bowel preparation and incomplete colonoscopy reduced the risk of missing adenomas, and likely better reflects the risk of cancer in patients after adenoma removal (4, 31).
Limitations of our study include the lack of an external validation dataset. However, we are not aware of any large enough validation cohort we could use. Although colorectal cancer incidence is regarded as the clinically most compelling endpoint in colorectal cancer screening and surveillance studies, it is potentially biased by lead time and overdiagnosis (32). Therefore, we also analyzed colorectal cancer mortality which is not affected by these biases, and the results were comparable. As in most studies, we did not perform a formal validation of polyp size assessment. However, polyp size assessment is part of our endoscopy training curriculum and was consistent over time. Thus, major misclassification seems unlikely.
One may argue that our results can be explained by recommended surveillance or lack thereof (24). However, the results from our sensitivity analysis confirmed our findings. Indeed, our results indicate that the novel risk classification system may not be altered by surveillance. We excluded individuals with suboptimal colonoscopy, and the Polish program has shown good quality (33). Thus, poor bowel preparation or high rates of incomplete removal of adenomas are not likely to bias our results. Given that our results
14
regarding current risk classification are in line with other studies, inaccuracy of our findings regarding novel risk classification seems highly improbable (21, 34).
Since our cohort consists of individuals that volunteered to undergo screening colonoscopy, the underlying colorectal cancer risk may differ from that in the general population. Baseline risk of colorectal cancer during the follow-up in our cohort may seem to be low (26.3 per 100,000 person-years), but we included only high-quality exams of relatively young (median age 56 years) and predominantly female individuals. However, standardized incidence rates are comparable to those reported in recent studies (22). Further, the observed adenoma detection rate of 17.7% and serrated adenoma detection rate of 1.0% is low compared to today’s standard. Endoscopists with low adenoma detection rates may remove large polyps less effectively, which can confound the current result. However, the risk of such confounding of our results is low, because in the Polish screening program, most large polyps are removed in a referral center.
By applying the novel risk classification in surveillance guidelines, the burden of colonoscopies could be reduced considerably. To allow policy makers, physicians and individual patients to make an informed choice about surveillance strategies according to resources, values and preferences, we have estimated the absolute risk of colorectal cancer for the different categories of high-risk, and shown how the novel definition of high-risk alter the number of individuals in need of intensive surveillance (Figure 2). Optimized use of surveillance may help to shift limited colonoscopy resources to screening and diagnostics.
In summary, we showed that by applying our new high-risk definition, only 10% of individuals with adenomas may need intensive surveillance without compromising the risk for the rest of the adenoma patients.
15
ACKNOWLEDGMENTS The authors thank Hans‑Olov Adami, M.D., Ph.D., from the Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm and Marcin Polkowski from Department of Gastroenterology, Hepatology and Clinical Oncology, the Maria Sklodowska-Curie Memorial Cancer Center and Institute of Oncology, Warsaw, Poland for their editorial advice.
16
REFERENCES 1.
Winawer SJ, Zauber AG, Ho MN, O'Brien MJ, Gottlieb LS, Sternberg SS, et al. Prevention of
Colorectal Cancer by Colonoscopic Polypectomy. N Engl J Med. 1993;329(27):1977-81. 2.
Zauber AG, Winawer SJ, O'Brien MJ, Lansdorp-Vogelaar I, van Ballegooijen M, Hankey BF, et al.
Colonoscopic polypectomy and long-term prevention of colorectal-cancer deaths. N Engl J Med. 2012;366(8):687-96. 3.
Yamaji Y, Mitsushima T, Ikuma H, Watabe H, Okamoto M, Kawabe T, et al. Incidence and
recurrence rates of colorectal adenomas estimated by annually repeated colonoscopies on asymptomatic Japanese. Gut. 2004;53(4):568-72. 4.
Loeve F, van Ballegooijen M, Boer R, Kuipers EJ, Habbema JD. Colorectal cancer risk in adenoma
patients: a nation-wide study. Int J Cancer. 2004;111(1):147-51. 5.
Cottet V, Jooste V, Fournel I, Bouvier AM, Faivre J, Bonithon-Kopp C. Long-term risk of colorectal
cancer after adenoma removal: a population-based cohort study. Gut. 2012;61(8):1180-6. 6.
Martinez ME, Baron JA, Lieberman DA, Schatzkin A, Lanza E, Winawer SJ, et al. A pooled analysis
of advanced colorectal neoplasia diagnoses after colonoscopic polypectomy. Gastroenterology. 2009;136(3):832-41. 7.
Atkin WS, Valori R, Kuipers EJ, Hoff G, Senore C, Segnan N, et al. European guidelines for quality
assurance in colorectal cancer screening and diagnosis. First Edition--Colonoscopic surveillance following adenoma removal. Endoscopy. 2012;44 Suppl 3:SE151-63. 8.
Hassan C, Quintero E, Dumonceau JM, Regula J, Brandao C, Chaussade S, et al. Post-polypectomy
colonoscopy surveillance: European Society of Gastrointestinal Endoscopy (ESGE) Guideline. Endoscopy. 2013;45(10):842-64. 9.
Levin B, Lieberman DA, McFarland B, Andrews KS, Brooks D, Bond J, et al. Screening and
surveillance for the early detection of colorectal cancer and adenomatous polyps, 2008: a joint guideline from the American Cancer Society, the US Multi-Society Task Force on Colorectal Cancer, and the American College of Radiology. Gastroenterology. 2008;134(5):1570-95.
17
10.
Cairns SR, Scholefield JH, Steele RJ, Dunlop MG, Thomas HJ, Evans GD, et al. Guidelines for
colorectal cancer screening and surveillance in moderate and high risk groups (update from 2002). Gut. 2010;59(5):666-89. 11.
van Heijningen EM, Lansdorp-Vogelaar I, Steyerberg EW, Goede SL, Dekker E, Lesterhuis W, et al.
Adherence to surveillance guidelines after removal of colorectal adenomas: a large, community-based study. Gut. 2015;64(10):1584-92. 12.
Ferlitsch M, Reinhart K, Pramhas S, Wiener C, Gal O, Bannert C, et al. Sex-specific prevalence of
adenomas, advanced adenomas, and colorectal cancer in individuals undergoing screening colonoscopy. JAMA. 2011;306(12):1352-8. 13.
Brenner H, Altenhofen L, Stock C, Hoffmeister M. Expected long-term impact of the German
screening colonoscopy programme on colorectal cancer prevention: Analyses based on 4,407,971 screening colonoscopies. European Journal of Cancer. 2015;51(10):1346-53. 14.
Winawer SJ, Zauber AG, O'Brien MJ, Ho MN, Gottlieb L, Sternberg SS, et al. Randomized
Comparison of Surveillance Intervals after Colonoscopic Removal of Newly Diagnosed Adenomatous Polyps. N Engl J Med. 1993;328(13):901-6. 15.
Kronborg O, Jørgensen OD, Fenger C, Rasmussen M. Three randomized long-term surveillance
trials in patients with sporadic colorectal adenomas. Scand J Gastroenterol. 2006;41(6):737-43. 16.
Jover R, Bretthauer M, Dekker E, Holme O, Kaminski MF, Loberg M, et al. Rationale and design of
the European Polyp Surveillance (EPoS) trials. Endoscopy. 2016;48(6):571-8. 17.
Chung SJ, Kim YS, Yang SY, Song JH, Kim D, Park MJ, et al. Five-year risk for advanced
colorectal neoplasia after initial colonoscopy according to the baseline risk stratification: a prospective study in 2452 asymptomatic Koreans. Gut. 2011;60(11):1537-43. 18.
Lieberman DA, Weiss DG, Harford WV, Ahnen DJ, Provenzale D, Sontag SJ, et al. Five-year colon
surveillance after screening colonoscopy. Gastroenterology. 2007;133(4):1077-85. 19.
Cafferty FH, Wong JM, Yen AM, Duffy SW, Atkin WS, Chen TH. Findings at follow-up
endoscopies in subjects with suspected colorectal abnormalities: effects of baseline findings and time to follow-up. Cancer J. 2007;13(4):263-70. 18
20.
Saini SD, Kim HM, Schoenfeld P. Incidence of advanced adenomas at surveillance colonoscopy in
patients with a personal history of colon adenomas: a meta-analysis and systematic review. Gastrointestinal endoscopy. 2006;64(4):614-26. 21.
Loberg M, Kalager M, Holme O, Hoff G, Adami HO, Bretthauer M. Long-term colorectal-cancer
mortality after adenoma removal. N Engl J Med. 2014;371(9):799-807. 22.
Atkin W, Wooldrage K, Brenner A, Martin J, Shah U, Perera S, et al. Adenoma surveillance and
colorectal cancer incidence: a retrospective, multicentre, cohort study. The Lancet Oncology. 2017;18(6):823-34. 23.
Kaminski MF, Kraszewska E, Rupinski M, Laskowska M, Wieszczy P, Regula J. Design of the
Polish Colonoscopy Screening Program: a randomized health services study. Endoscopy. 2015. 24.
Winawer SJ, Fletcher RH, Miller L, Godlee F, Stolar MH, Mulrow CD, et al. Colorectal cancer
screening: clinical guidelines and rationale. Gastroenterology. 1997;112(2):594-642. 25.
Winawer S, Fletcher R, Rex D, Bond J, Burt R, Ferrucci J, et al. Colorectal cancer screening and
surveillance: clinical guidelines and rationale-Update based on new evidence. Gastroenterology. 2003;124(2):544-60. 26.
Aronchick C, Lipshutz W, Wright S, Dufrayne F, Bergman G. Validation of an instrument to assess
colon cleansing. The American journal of gastroenterology. 1999;94:2667. 27.
National Cancer Registry. Available from: http://onkologia.org.pl/raporty/.
28.
Plumb AA, Nickerson C, Wooldrage K, Bassett P, Taylor SA, Altman D, et al. Terminal digit
preference biases polyp size measurements at endoscopy, computed tomographic colonography, and histopathology. Endoscopy. 2016;48(10):899-908. 29.
Zhang H. Recursive Partitioning and Applications. 2nd ed. ed. Singer BH, editor. New York, NY:
Springer New York; 2010. 30.
Didkowska J, Wojciechowska U. Cancer in Poland in 2013. Warsaw: National Cancer Registry of
Poland, the Maria Sklodowska-Curie Memorial Cancer Center and Institute of Oncology; 2015.
19
31.
Froehlich F, Wietlisbach V, Gonvers JJ, Burnand B, Vader JP. Impact of colonic cleansing on
quality and diagnostic yield of colonoscopy: the European Panel of Appropriateness of Gastrointestinal Endoscopy European multicenter study. Gastrointestinal endoscopy. 2005;61(3):378-84. 32.
Bretthauer M, Kalager M. Principles, effectiveness and caveats in screening for cancer. The British
journal of surgery. 2013;100(1):55-65. 33.
Kaminski MF, Wieszczy P, Rupinski M, Wojciechowska U, Didkowska J, Kraszewska E, et al.
Increased Rate of Adenoma Detection Associates With Reduced Risk of Colorectal Cancer and Death. Gastroenterology. 2017;153(1):98-105. 34.
Click B, Pinsky PF, Hickey T, Doroudi M, Schoen RE. Association of Colonoscopy Adenoma
Findings With Long-term Colorectal Cancer Incidence. Jama. 2018;319(19):2021-31.
20
Figure legends Figure 1. Cumulative hazard rates for colorectal cancer in current and novel risk groups and no adenoma group (panel A – current risk groups; panel B – novel risk groups) The graph shows cumulative hazard rates for colorectal cancer in the current and novel risk groups and in the no adenoma group. Current risk groups are defined as: low-risk (1-2 tubular adenomas <10 mm in diameter without high-grade dysplasia); high-risk (3 or more adenomas, any adenoma with villous growth pattern, high-grade dysplasia, or size ≥10 mm in diameter). Novel risk groups are defined as: adenomas ≥20 mm in diameter; adenomas <20 mm in diameter with high-grade dysplasia; adenomas <20 mm in diameter without high-grade dysplasia. Abbreviations: HGD – high-grade dysplasia.
Figure 2. Risk of colorectal cancer in different risk group definitions One circle represents approximately 418 individuals. Red circles represent proportion of individuals in the novel high-risk group (≥20 mm in diameter or high-grade dysplasia), light blue circles represent proportion of individuals who are moved from the current high-risk group (3 or more adenomas, any adenoma with villous growth pattern, high-grade dysplasia, or size ≥10 mm in diameter) to the novel low-risk group (<20 mm in diameter without high-grade dysplasia), dark blue circles represent proportion of individuals in current low-risk group (1-2 tubular adenomas <10 mm in diameter without high-grade dysplasia). Corresponding bars show colorectal cancer risk per 100,000 person-years. Risk difference between current low-risk and those moved to the novel low-risk group was 5.6 (95% CI -10.7-22.0). Abbreviations: p-yrs – person-years.
21
Table 1. Baseline characteristic of individuals included in the study Individuals, no. (%)
CRC, no.
CRC risk, %
P-years
236,089
439
0.19
1,669,684
CRC risk per 100,000 p-yrs 26.3
40-59
171,081 (72.5)
261
0.15
1,226,992
60-66
65,008 (27.5)
178
0.27
Female
146,731 (62.2)
260
Male
89,358 (37.8)
None
CRC death, no.
CRC death risk, %
P-years
132
0.06
1,671,226
CRC death risk per 100,000 p-yrs 7.9
21.3
66
0.04
1,227,891
5.4
442,692
40.2
66
0.10
443,375
14.9
0.18
1,046,372
24.8
73
0.05
1,047,259
7.0
179
0.20
623,312
28.7
59
0.07
624,007
9.5
189,714 (80.4)
358
0.19
1,333,197
26.9
112
0.06
1,334,516
8.4
I⁰ relatives
46,375 (19.6)
81
0.17
336,487
24.1
59
0.13
336,750
17.5
Any adenoma
41,778 (17.7)
130
0.31
279,585
46.5
34
0.08
280,059
12.1
< 10 mm
33,927 (81.2)
76
0.22
225,548
33.7
20
0.06
225,804
8.9
10-19 mm
5,753 (13.8)
23
0.40
39,578
58.1
6
0.10
39,660
15.1
≥ 20 mm
2,098 (5.0)
31
1.48
14,459
214.4
8
0.38
14,594
54.8
Tubular
33,803 (80.9)
96
0.28
224,166
42.8
27
0.08
224,506
12.0
Villous/tubulo-villous
7,975 (19.1)
34
0.43
55,419
61.4
7
0.09
55,553
12.6
Variable Total Age group
Sex
Family history of CRC
Size of adenoma
Growth pattern
Grade of dysplasia
Low
39,290 (94.0)
104
0.26
262,293
39.7
27
0.07
262,648
10.3
High
2,488 (6.0)
26
1.05
17,293
150.3
7
0.28
17,410
40.2
1-2
38,693 (92.6)
120
0.31
258,002
46.5
33
0.09
258,429
12.8
3-4
2,439 (5.8)
9
0.37
16,920
53.2
1
0.04
16,960
5.9
≥5
646 (1.5)
1
0.15
4,664
21.4
0
0
4,669
0
Low-risk
26,536 (63.5)
58
0.22
174,743
33.2
17
0.06
174,929
9.7
High-risk
15,242 (36.5)
72
0.47
104,842
68.7
17
0.11
105,129
16.2
Number of adenomas
Current risk groups
Abbreviations: CRC – colorectal cancer, p-yrs – person-years of observation. Low-risk refers to individuals with 1-2 tubular adenomas <10 mm in diameter without highgrade dysplasia. High-risk refers to individuals with 3 or more adenomas, or any adenoma with villous growth pattern, high-grade dysplasia, or size ≥10 mm in diameter.
Table 2. Comparison with the general population: Standardized incidence ratio for colorectal cancer and standardized incidence based mortality ratio for colorectal cancer death No. of No. of No. of SIR No. of expected SMR Variable observed expected observed CRC [95% CI] CRC deaths [95% CI] CRCs CRCs deaths All
439
1,441
0.30 [0.28-0.33]
132
653
0.20 [0.17-0.24]
No adenoma
309
1163
0.27 [0.24-0.30]
98
525
0.19 [0.15-0.23]
Low-risk group
58
168
0.35 [0.26-0.45]
17
76
0.22 [0.13-0.36]
High-risk group
72
110
0.65 [0.51-0.82]
17
51
0.33 [0.19-0.53]
<20 mm, no HGD
88
249
0.35 [0.28-0.44]
23
114
0.20 [0.13-0.30]
<20 mm, HGD
11
14
0.79 [0.39-1.41]
3
6
0.50 [0.10-1.46]
≥20 mm
31
15
2.07 [1.40-2.93]
8
7
1.14 [0.49-2.25]
Current risk stratification system
Novel risk groups definition
Abbreviations: CRC – colorectal cancer, SIR – standardized incidence ratio, SMR – standardized incidence based mortality ratio. Low-risk refers to individuals with 1-2 tubular adenomas <10 mm in diameter without high-grade dysplasia. High-risk refers to individuals with 3 or more adenomas, or any adenoma with villous growth pattern, high-grade dysplasia, or size ≥10 mm in diameter.
Table 3. Comparison with individuals without adenomas: Multivariable hazard ratios for risk of colorectal cancer (CRC) and death (CRC death) for the novel risk group definition Number of HR† HR† CRC P-value CRC death P-value individuals [95% CI] [95% CI] Primary analyses Novel risk groups definition No adenoma
194,311
309
1.00
98
1.00
<20 mm, no HGD
37,798
88
1.55 [1.22-1.96]
<0.001
23
1.32 [0.84-2.09]
0.227
<20 mm, HGD
1,881
11
3.58 [1.96-6.54]
<0.001
3
2.97 [0.94-9.38]
0.064
≥20 mm
2,099
31
9.25 [6.39-13.39]
<0.001
8
7.45 [3.62-15.33]
<0.001
No adenoma
58,362
73
1.00
30
1.00
<20 mm, no HGD
12,335
23
1.52 [0.95-2.42]
0.083
9
1.85 [1.23-2.79]
0.003
<20 mm, HGD
493
2
3.47 [0.85-14.15]
0.083
0
<0.001
≥20 mm
745
10
12.25 [6.32-23.76]
<0.001
2
8.53** [4.91-14.84]
Sensitivity analyses* Novel risk groups definition
†
Multivariable HR adjusted for age group. Sex and family history of colorectal cancer were tested for inclusion, but were not found significant. *Only individuals screened in 2007-2008 with censoring on polypectomy. **Categories ≥20 mm and <20 mm, HGD are collapsed as one category due to estimability. Abbreviations: HGD=highgrade dysplasia, CRC=colorectal cancer.
A Cumulative Colorectal Cancer Hazard [%]
5.0
High−risk Low−risk No adenoma
4.0
3.0
2.0
1.0
0 0 No. at Risk High−risk
2
4
6
8
10
5,028 7,217 72,409
2,144 2,862 32,344
6
8
10
1,400 1,224 22,900
680 742 10,823
282 252 4,472
Years 15,242 26,536 194,311
Low−risk No adenoma B
12,723 21,686 166,638
9,843 15,681 103,221
≥ 20 mm < 20 mm, HGD < 20 mm, noHGD No adenoma
5.0 Cumulative Colorectal Cancer Hazard [%]
15,049 26,323 192,993
4.0
3.0
2.0
1.0
0 0 No. at Risk ≥ 20 mm < 20 mm, HGD < 20 mm, noHGD
2
4 Years
2,009 1,881 37,798
2,059 1,853 37,460
1,766 1,579 31,064
High−risk defined as ≥ 20 mm or HGD ● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
Risk/100,000 p−yrs
150 100
50 0
New Current Risk among low−risk those moved high−risk
No. of subjects in risk groups
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ● ● ●
Risk/100,000 p−yrs
Current high−risk defined as HGD or ≥ 3 or villous or ≥ 10 mm
No. of subjects in risk groups
150 100 50 0
Current low−risk
Current high−risk
Supplemental material Contents Appendix 1. Standardized incidence and mortality rates Appendix 2. Sensitivity analysis Appendix Table 1. Baseline characteristic of individuals Appendix Table 2. Risk of colorectal cancer and colorectal cancer death for the current risk group definition Appendix Table 3. Risk factors of colorectal cancer in individuals with adenomas – steps of building new risk classification system Appendix Table 4. Multivariable hazard ratios for risk of colorectal cancer and colorectal cancer death for the novel risk group definition Appendix Table 5. Bootstrap model estimates for novel risk group definition Appendix Table 6. Comparison with individuals without adenomas: Multivariable hazard ratios for risk of colorectal cancer (CRC) and death (CRC death) for the novel risk group definition. Sensitivity analysis for individuals without serrated adenomas/polyps. Appendix Table 7. Colorectal cancer risk according to current and novel risk classification system Appendix Table 8. Comparison with individuals without adenomas: Multivariable hazard ratios for risk of colorectal cancer (CRC) for the risk groups based on the current and novel definitions. Appendix Figure 1. Follow-up flow-chart Appendix Figure 2. Model building based on characteristics of adenomas
1
Appendix 1. Standardized incidence and mortality rates We calculated standardized colorectal cancer incidence ratios as the observed number of colorectal cancers divided by the expected number. The expected number of colorectal cancers was defined as the product of the colorectal cancer rate in the general population and the number of person-years at risk in the study population. Incidence ratios were standardized according to 5-years age groups and sex. We computed 95% CI for standardized colorectal cancer incidence ratios under the assumption that the observed number of colorectal cancers followed a Poisson distribution. Using a similar analytical approach, we obtained values of standardized colorectal cancer mortality ratio. The expected number of deaths was calculated as number of deaths that would be observed among individuals who had cancer diagnosis after the date of screening colonoscopy (incidence based mortality). Results of this analysis can be found in Table 2.
Appendix 2. Sensitivity analysis Of the individuals screened in 2007-2008 who had adenomas ≥20mm (745 individuals), 414 (55.6%) had colonoscopy and 211 (28.3%) had polypectomy with a median time to polypectomy of 3.3 years. Among individuals with adenomas <20mm with high-grade dysplasia (493 individuals), 227 (46.0%) had colonoscopy and 109 (22.7%) of had polypectomy with a median time to polypectomy of 3.3 years. Among individuals with adenomas <20mm without high-grade dysplasia (12,335 individuals), 3,861 (31.3%) had colonoscopy and 1,557 (12.6%) had polypectomy with a median time to polypectomy of 4.4 years. The results from the analysis of colorectal cancer risk and death according to novel risk classification system, where individuals were censored at the time of polypectomy, were in line with the results from the primary analysis. Individuals with adenomas ≥20mm in diameter had 12 times higher risk of colorectal cancer (age adjusted HR of 12.25; 95%CI 6.32-23.76; P<0.001) than individuals with no adenomas.
2
Appendix Table 1. Baseline characteristic of individuals
40-59
60-66
89,358
171,081
65,008
Family history of CRC, N (%) 46,375
41,778 (17.7)
21,015 (23.5)
27,517 (16.1)
14,261 (21.9)
8,412 (18.1)
<10 mm
33,927 (81.2)
16,889 (80.4)
22,548 (81.9)
11,379 (79.8)
6,763 (80.4)
10-19 mm
5,753 (13.8)
3,073 (14.6)
3,648 (13.3)
2,105 (14.8)
1,188 (14.1)
≥20 mm
2,098 (5.0)
1,071 (5.1)
1,321 (4.8)
777 (5.4)
461 (5.5)
Tubular Villous/tubulovillous Grade of dysplasia
33,803 (80.9)
16,941 (80.6)
22,450 (81.6)
11,353 (79.6)
6,767 (80.4)
7,975 (19.1)
4,092 (19.5)
5,067 (18.4)
2,908 (20.4)
1,645 (19.6)
Low
39,290 (94.0)
19,693 (93.7)
25,960 (94.3)
13,330 (93.5)
7,877 (93.6)
High
2,488 (6.0)
1,340 (6.4)
1,557 (5.7)
931 (6.5)
535 (6.4)
1-2
38,693 (92.6)
19,101 (90.9)
25,733 (93.5)
12,960 (90.9)
7,761 (92.3)
3-4
2,439 (5.8)
1,507 (7.2)
1,408 (5.1)
1,031 (7.2)
497 (5.9)
376 (1.4)
270 (1.9)
154 (1.8)
Variable All Any adenoma
All, N (%)
Male sex, N (%)
236,089
Age group, N (%)
Size of adenoma
Growth pattern
Number of adenomas
≥5 646 (1.5) 425 (2.0) Abbreviation: CRC – colorectal cancer, N - number
3
Appendix Table 2. Multivariable hazard ratios for risk of colorectal cancer and colorectal cancer death for the current risk group definition Number of HR CRC HR CRC P-value P-value individuals [95% CI] death [95% CI] Current risk groups definition† No adenoma
194,311
309
1.00
Low-risk group
26,536
58
1.49 [1.13-1.98]
High-risk group
15,242
72
2.94 [2.28-3.81]
98
1.00
0.005
17
1.48 [0.88-2.46]
0.145
<0.001
17
2.16 [1.29-3.62]
0.004
†
HR adjusted for age group. Sex and family history of colorectal cancer were tested for inclusion, but were not found significant. Abbreviations: CRC=colorectal cancer.
4
Appendix Table 3. Risk factors of colorectal cancer in individuals with adenomas – steps of building new risk classification system HR* N CRC P-value [95% CI] Step 1† Diameter < 10 mm
33,927
76
1.00
10-19 mm
5,753
23
1.51 [0.94-2.42]
0.089
≥ 20 mm
2,098
31
4.83 [3.05-7.63]
<0.001
Low
39,290
104
1.00
High
2,488
26
2.16 [1.35-3.46]
40-59
27,517
72
1.00
60-66
14,261
58
1.60 [1.13-2.26]
Low
37,799
88
1.00
High
1,881
11
2.06 [1.08-3.92]
40-59
26,196
56
1.00
60-66
13,484
43
1.57 [1.05-2.33]
< 10 mm
33,927
76
1.00
10-19 mm
5,753
23
1.52 [0.94-2.45]
Grade of dysplasia
0.001
Age
0.008
Step 2‡ Grade of dysplasia
0.028
Age
0.027
Diameter
0.088
Step 3$ Age 40-59
1.00
60-66
1.07 [1.02-1.12]
0.002
Family history of CRC Other st
1 degree
1.00 1.75 [1.08-2.84]
0.002
*HR reported for variables found significant in stepwise selection. Age group, sex, family history of colorectal cancer, adenoma diameter, grade of dysplasia, growth pattern and no. of adenomas were tested for inclusion into model. †Models built using observations for all subject with adenomas. ‡Models built using observations for subject with adenomas <20 mm in diameter. $ Models build using observations for subjects with adenomas <20 mm in diameter and no high-grade dysplasia. Abbreviations: CRC=colorectal cancer
5
Appendix Table 4. Multivariable hazard ratios for risk of colorectal cancer and colorectal cancer death for the novel risk group definition Colorectal cancer Model coefficient
HR* [95% CI]
Colorectal cancer death P-value
Model coefficient
HR* [95% CI]
P-value
Age 40-59 60-66
1.00 0.60
1.82 [1.50-2.20]
1.00 <0.001
1.03
2.79 [1.98-3.93]
<0.001
Novel risk group No adenoma
1.00
1.00
<20 mm, no HGD
0.44
1.55 [1.22-1.96]
<0.001
0.28
1.32 [0.84-2.09]
0.227
<20 mm, HGD
1.28
3.58 [1.96-6.54]
<0.001
1.09
2.97 [0.94-9.38]
0.064
≥20 mm
2.22
9.25 [6.39-13.29]
<0.001
2.01
7.45 [3.62-15.33]
<0.001
*HR reported for variables found significant in stepwise selection. Age group, sex and family history of colorectal cancer were tested for inclusion into the models. Abbreviations: HGD=high-grade dysplasia.
6
Appendix Table 5. Bootstrap model estimates for novel risk group definition Colorectal cancer HR [95% CI]
Colorectal cancer death P-value
HR [95% CI]
P-value
Age 40-59
1.00
60-66
1.82 [1.49-2.20]
1.00 <0.001
2.80 [1.99-3.90]
<0.001
Novel risk group No adenoma
1.00
1.00
<20 mm, no HGD
1.55 [1.22-1.95]
<0.001
1.32 [0.84-2.10]
0.226
<20 mm, HGD
3.60 [1.88-6.28]
<0.001
2.97 [0-Inf]
0.908
≥20 mm
9.21 [6.36-13.46]
<0.001
7.46 [3.42-16.28]
<0.001
7
Appendix Table 6. Comparison with individuals without adenomas: Multivariable hazard ratios for risk of colorectal cancer (CRC) and death (CRC death) for the novel risk group definition. Sensitivity analysis for individuals without serrated adenomas/polyps. Number of HR† HR† CRC P-value CRC death P-value individuals [95% CI] [95% CI] Novel risk groups definition No adenoma
194,311
309
1.00
98
1.00
<20 mm, no HGD
35,621
87
1.62 [1.27-2.05]
<0.001
23
1.40 [0.89-2.20]
0.151
<20 mm, HGD
1,836
11
3.66 [2.01-6.69]
<0.001
3
3.05 [0.96-9.62]
0.058
≥20 mm
2,027
31
9.53 [6.58-13.79]
<0.001
7
6.68 [3.10-14.40]
<0.001
†
Multivariable HR adjusted for age group. Sex and family history of colorectal cancer were tested for inclusion, but were not found significant. Abbreviations: HGD=high-grade dysplasia,
CRC=colorectal cancer.
8
Appendix Table 7. Colorectal cancer risk according to current and novel risk classification system. Novel risk classification Current risk classification
<20 mm and no HGD
≥20 mm or HGD
Total
CRC=58 CRC=0 CRC=58 no-CRC=26,478 no-CRC=0 no-CRC=26,478 Low-risk P-yrs=174,743 P-yrs=0 P-yrs=174,743 CRC risk per 100,000 p-yrs=33.2 CRC risk per 100,000 p-yrs =NA CRC risk per 100,000 p-yrs=33.2 CRC=30 CRC=42 CRC=72 no-CRC=11,232 no-CRC=3,938 no-CRC=15,170 High-risk P-yrs=77,250 P-yrs=27,592 P-yrs=104,842 CRC risk per 100,000 p-yrs=38.8 CRC risk per 100,000 p-yrs=152.2 CRC risk per 100,000 p-yrs=68.7 CRC=88 CRC=42 CRC=130 no-CRC=37,710 no-CRC=3,938 no-CRC=41,648 Total P-yrs=251,993 P-yrs=27,592 P-yrs=279,585 CRC risk per 100,000 p-yrs=34.9 CRC risk per 100,000 p-yrs=152.2 CRC risk per 100,000 p-yrs=46.5 CRC=no. of individuals with colorectal cancer diagnosis during the follow-up time, no-CRC=no. of individuals without colorectal cancer diagnosis during the follow-up time, CRC=colorectal cancer, p-yrs – person-years of observation
9
Appendix Table 8. Comparison with individuals without adenomas: Multivariable hazard ratios for risk of colorectal cancer (CRC) for the risk groups based on the current and novel definitions. Number of HR† CRC P-value individuals [95% CI] No adenoma
194,311
309
1.00
Current low-risk*
26,536
58
1.49 [1.13-1.98]
0.005
Novel low-risk, current high-risk**
11,262
30
1.66 [1.14-2.42]
0.008
Novel high-risk***
3,980
42
6.54 [4.74-9.04]
<0.001
†
Age adjusted. *1-2 tubular adenomas <10 mm in diameter without high-grade dysplasia. **3 or more adenomas, any adenoma with villous growth pattern, or size 10-19 mm in diameter. ***≥20 mm in diameter or high-grade dysplasia.
10
Appendix Figure 1. Follow-up flow-chart Abbreviations: CRC=colorectal cancer, f-up=follow-up.
11
Appendix Figure 2. Model building based on characteristics of adenomas The graph shows the steps of obtaining new risk classification system. First, basing on the results of multivariable model the set was divided according to adenoma diameter (≥20 mm or <20 mm). Then, individuals with adenomas ≥20 mm in diameter were excluded and analysis were redone. Remaining individuals were divided according to grade of dysplasia. In the last step, only individuals with adenomas <20 mm in diameter with low-grade dysplasia were analyzed.
12