Supplementary Appendix This appendix has been provided by the authors to give readers additional information about their work. Supplement to: Zauber AG, Winawer SJ, O Brien MJ, et al. Colonoscopic polypectomy and long-term prevention of colorectal-cancer deaths. N Engl J Med 2012;366:687-96.
SUPPLEMENT Microsimulation Modeling Estimate of the Impact of Colonoscopic Polypectomy on Colorectal Cancer Mortality in the National Polyp Study Online Appendix: Table of Contents Summary Page 1 Introduction Page 1 Methods Page 2 Supplement Figure 1. Graphical representation of natural history of disease as modeled by MISCAN-Colon model. Page 3 Results Page 5 Discussion Page 5 References Page 8 Supplement Table 1. Characteristics of all adenoma patients at initial polypectomy as observed in the National Polyp Study and as simulated in the MISCAN-Colon model (n=2602). Supplement Table 2. Deaths from Colorectal Cancer in the Observed Adenoma Cohort with Polypectomy Compared to Simulated Adenoma Cohort without Polypectomy. Supplement Appendix Table 1. Main assumptions in the MISCAN-Colon model in simulating the National Polyp Study adenoma cohort. Page 12 Page 13 Page 14 Participating Centers in the National Polyp Study. Page 16
Colonoscopic Polypectomy and Long-Term Prevention of Colorectal-Cancer Deaths (Zauber et al, N Engl J Med 2012) SUPPLEMENT: Microsimulation Modeling Estimate of the Impact of Colonoscopic Polypectomy on Mortality from Colorectal Cancer in the National Polyp Study SUMMARY In the National Polyp Study we compared the observed deaths from colorectal cancer in the adenoma cohort in which all detectable adenomas had been removed with that of a comparable simulated adenoma cohort but without polypectomy. There were 12 observed deaths from colorectal cancer in the adenoma cohort compared to 145 expected deaths from the disease based on the simulated rates for the adenoma cohort without polypectomy (MISCAN-Colon); the standardized mortality ratio was 0.08 (95% confidence interval [CI], 0.04 to 0.14). This represents a 92% reduction in mortality from colorectal cancer in the adenoma cohort with polypectomy compared with a hypothetical cohort of patients with adenomas detected but never removed. INTRODUCTION In the paper 1 we assessed the long-term impact of colonoscopic polypectomy and subsequent surveillance on mortality from colorectal cancer in newly diagnosed adenoma patients in the National Polyp Study (NPS) and demonstrated a large impact of colonoscopic polypectomy. We did so by comparing the mortality from colorectal cancer in the NPS adenoma cohort with an age, sex, race, and calendar year adjusted estimate from the general population 1
SEER9 rates and with a concurrent control group of patients with non-adenomatous polyps. 1 However, as discussed in the paper, a more meaningful comparison group for the NPS adenoma cohort with polypectomy would be a clinical comparison of an adenoma cohort without polypectomy. This is not an ethical or clinical option because of the known potential for adenoma progression to carcinoma. Instead, in this supplement, we addressed this comparison using a microsimulation model of mortality from colorectal cancer for a cohort of patients with adenomas in which the adenomas were ascertained but not removed and the natural history of the adenoma-carcinoma sequence proceeded without intervention. METHODS The deaths from colorectal cancer of the NPS patients with adenomas were compared to the expected deaths from colorectal cancer from a simulated adenoma cohort without polypectomy. We determined the expected number of deaths from colorectal cancer for a cohort of patients with adenomas by using a validated microsimulation model (MISCAN-Colon) to predict the natural history of individuals with adenomas with the same age and baseline disease characteristics as the NPS adenoma cohort, but without any polypectomy. 2 REFERENCE GROUP OF MICROSIMULATION MODEL OF ADENOMA COHORT WITHOUT POLYPECTOMY MISCAN-Colon Microsimulation Model The Department of Public Health at Erasmus MC, the Netherlands, developed the MISCAN-Colon microsimulation model in collaboration with the National Cancer Institute to assess the effect of different interventions on colorectal cancer. The model structure and inputs and the data sources that informed the development of the model can be found in Supplement 2
Appendix Table 1, in previous publications, 3-5 and also in a standardized model profile. 6 MISCAN-Colon is one of the models in the Cancer Incidence and Surveillance Modeling Network (CISNET) of the United States National Cancer Institute (http://cisnet.cancer.gov/). A graphic representation of the natural history in the model is given in Supplement Figure 1. Screening Adenoma States Preclinical Cancer States Clinical Cancer States no lesion adenoma 5 mm adenoma 6-9 mm adenoma 10 mm preclinical stage I preclinical stage II preclinical stage III preclinical stage IV clinical stage I clinical stage II clinical stage III clinical stage IV death colorectal cancer Supplement Figure 1. Graphical representation of natural history of disease as modeled by MISCAN-Colon model. The opportunity to intervene in the natural history through screening is noted. In brief, the MISCAN-Colon model simulates the relevant lifetimes of a large population of individuals from birth to death, first without screening and subsequently with the changes that would occur under the implementation of a screening and surveillance program. Colorectal cancer arises in this population according to the adenoma-carcinoma sequence. 7,8 More than one adenoma can occur in an individual and each can independently develop into colorectal cancer. Adenomas progress in size from small (1-5 mm) to medium (6-9 mm) to large (10+ mm). Most adenomas will never grow into cancer (non-progressive adenomas). These adenomas either stay 3
6-9 mm in size or continue to grow to 10 mm or larger. Some adenomas (progressive adenomas) may eventually become malignant, transforming to a stage I cancer. The cancer may then progress from stage I to stage IV. In every stage there is a probability of the cancer being diagnosed because of symptoms versus alternatively progressing without symptoms into the next stage. Meanwhile, a person may die of other causes at any time, precluding developing or dying from colorectal cancer. The survival after clinical diagnosis depends on the stage in which the cancer was detected. MISCAN-Colon simulation of the National Polyp Study cohort We used the MISCAN-Colon microsimulation model to simulate a population of adenoma patients of the same age, sex and baseline adenoma characteristics as the NPS 2 (Supplement Table 1). The participants in the NPS were subjects who had been referred for initial colonoscopy and had adenomatous polyps diagnosed and removed. The MISCAN-Colon model was adapted to this situation by applying a fictitious screening test to the general population to select individuals with positive test results who were referred to colonoscopy evaluation and had adenomas detected at diagnostic colonoscopy but the adenomas were not removed. The fictitious screening test can be regarded as a mixture of fecal occult blood test (FOBT) or sigmoidoscopy. 2 These individuals with positive screening tests constituted the simulated trial population referred for colonoscopy (Figure 1 1 ). As in the NPS, simulated individuals with a history of colorectal cancer or colorectal cancer detected at the baseline diagnostic colonoscopy were excluded from the trial population. The sensitivity and age distribution of the fictitious screening test was adjusted to reproduce the age distribution, the distribution of adenomas over the distal and proximal colon, and the size and multiplicitydistribution of adenomas at initial polypectomy in the NPS. 2 4
Secular trends in treatment. The NPS adenoma cohort was followed over calendar years 1980-2003. We adjusted the natural history model without polypectomy intervention for the changes in relative survival due to improvements in treatment over this time period. The MISCAN-Colon model had simulated the United States population from 1975 to 2000 with respect to mortality from colorectal cancer as a function of changes in risk factor behavior, screening adherence, and treatment. 9 We estimated that 3% of the change in mortality from colorectal cancer in that time period was due to changes in treatment; 9 therefore we reduced the MISCAN-Colon natural history model estimates of deaths from colorectal cancer by 3% to adjust for these secular trends over time. RESULTS There were 12 observed deaths from colorectal cancer in the NPS adenoma cohort compared to 145 expected deaths from the disease based on the simulated adenoma cohort rates without polypectomy (MISCAN-Colon); the standardized mortality ratio was 0.08 (95% confidence interval [CI], 0.04 to 0.14). (Supplement Table 2). This represents a 92% reduction in CRC mortality in the adenoma cohort compared with a hypothetical cohort of patients with adenomas detected but never removed, so that the progression of adenomas to colorectal cancer and to death from colorectal cancer was not impeded by intervention. There was a comparable size reduction for the first 10 years following initial polypectomy as well as for the 10 or more years after polypectomy (Supplement Table 2). DISCUSSION Comparison of the observed deaths from colorectal cancer in the NPS adenoma cohort with colonoscopic polypectomy to that expected in a simulated hypothetical cohort without 5
removal of their adenomas suggests that a very large reduction in mortality from colorectal cancer (92%) could be achieved by intervening in the adenoma-carcinoma sequence with removal of adenomas. These results supplement the results presented with comparisons to a non-adenoma internal control group and the general population SEER rates to suggest that colonoscopic polypectomy in conjunction with surveillance provides a marked decrease in mortality from colorectal cancer. 1 The MISCAN-Colon model has been validated and shown high concordance with outcomes of observed clinical trials. The validity of the model is based on observational data before the introduction of screening, such as clinical incidence and mortality from colorectal cancer 10 and the size and multiplicity distribution of adenomas in autopsy and colonoscopy studies. The external validity has further been tested on the results of the National Polyp Study, 2 and other large (randomized) screening and surveillance studies, such as the CoCap sigmoidoscopy study, 3 and the Minnesota Colon Cancer Control Study. 3 In addition, the model was able to explain observed incidence and mortality trends in the United States when accounting for risk factor behaviors, screening practice and chemotherapy treatment. 9 One of the most important assumptions in the modeling is the distribution of the dwell time of the adenomas and that of preclinical cancers. This is generally an unobservable parameter but one which affects the projections of the model, especially with respect to appropriate surveillance times following initial adenoma detection. In general the MISCAN-Colon model has a shorter dwell time than other microsimulation models for colorectal cancer. 11 6
In conclusion, the microsimulation modeling approach suggests that colonoscopic polypectomy in an adenoma population should provide a significant reduction (92%) in mortality from colorectal cancer compared to no polypectomy. 7
REFERENCES 1. Zauber A, Winawer S, O'Brien M, et al. Colonoscopic polypectomy and long-term prevention of colorectal cancer deaths. N Engl J Med 2012; 366, February 23, 2012. 2. Loeve F, Boer R, Zauber AG, et al. National Polyp Study data: evidence for regression of adenomas. International journal of cancer 2004;111:633-9. 3. Loeve F, Boer R, Ballegooijen Mv, Oortmarssen GJv, Habbema JDF. Final Report MISCAN-COLON Microsimulation Model for Colorectal Cancer. Rotterdam: Department of Public Health, Erasmus University; 1998. 4. Loeve F, Boer R, van Oortmarssen GJ, van Ballegooijen M, Habbema JDF. The MISCAN-COLON simulation model for the evaluation of colorectal cancer screening. Comput Biomed Res 1999;32:13-33. 5. Loeve F, Brown ML, Boer R, van Ballegooijen M, van Oortmarssen GJ, Habbema JD. Endoscopic colorectal cancer screening: a cost-saving analysis. Journal of the National Cancer Institute 2000;92:557-63. 6. Vogelaar I, van Ballegooijen M, Zauber AG, Boer R, Oortmarssen GJv, Loeve F. Department of Public Health. Erasmus MC. Model Profiler of the MISCAN-Colon microsimulation model for colorectal cancer. In; 2004. https://cisnet.flexkb.net/mp/pub/cisnet_colorectal_sloankettering_profile.pdf (accessed October 12, 2011). 7. Morson B. The polyp-cancer sequence in the large bowel. Proc R Soc Med 1974;67:451-7. 8. Muto T, Bussey HJ, Morson BC. The evolution of cancer of the colon and rectum. Cancer 1975;36:2251-70. 8
9. Edwards BK, Ward E, Kohler BA, et al. Annual report to the nation on the status of cancer, 1975-2006, featuring colorectal cancer trends and impact of interventions (risk factors, screening, and treatment) to reduce future rates. Cancer 2010;116:544-73. 10. Surveillance Epidemiology, and End Results (SEER) Program (http://www.seer.cancer.gov) SEER* Stat Database: Incidence - SEER 9 Regs Public-Use, Nov2010 Sub (1973-2008) <Katrina/Rita Population Adjustment> - Linked To County Attributes - Total U.S., 1969-2009 Counties, National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch, released April 2011, based on the November 2010 submission. In; 2011. 11. Kuntz K, Lansdorp-Vogelaar I, Rutter C, et al. A systematic comparison of microsimulation models of colorectal cancer: the role of assumptions about adenoma progression. Journal of Medical Decision Making 2011;31:530-9. 12. Arminski TC, McLean DW. Incidence and distribution of adenomatous polyps of the colon and rectum based on 1.000 autopsy examinations.. Dis Colon Rectum 1964;7:249-61. 13. Blatt LJ. Polyps of the colon and rectum: incidence and distribution. Dis Colon Rectum 1961;4:277-82. 14. Bombi JA. Polyps of the colon in Barcelona, Spain. Cancer 1988;61:1472-6. 15. Chapman I. Adenomatous polypi of large intestine: incidence and distribution. Ann Surg 1963;157:223-6. 16. Clark JC, Collan Y, Eide TJ, et al. Prevalence of polyps in an autopsy series from areas with varying incidence of large-bowel cancer. International journal of cancer 1985;36:179-86. 17. Jass JR, Young PJ, Robinson EM. Predictors of presence, multiplicity, size and dysplasia of colorectal adenomas. A necropsy study in New Zealand. Gut 1992;33:1508-14. 9
18. Johannsen LG, Momsen O, Jacobsen NO. Polyps of the large intestine in Aarhus, Denmark. An autopsy study. Scand J Gastroenterol 1989;24:799-806. 19. Rickert RR, Auerbach O, Garfinkel L, Hammond EC, Frasca JM. Adenomatous lesions of the large bowel. Cancer 1979;43:1847-57. 20. Vatn MH, Stalsberg H. The prevalence of polyps of the large intestine in Oslo: an autopsy study. Cancer 1982;49:819-25. 21. Williams AR, Balasooriya BA, Day DW. Polyps and cancer of the large bowel: a necropsy study in Liverpool. Gut 1982;23:835-42. 22. Walter SD, Day NE. Estimation of the duration of a pre-clinical disease state using screening data. American journal of epidemiology 1983;118:865-86. 23. Gyrd-Hansen D, Sogaard J, Kronborg O. Analysis of screening data: colorectal cancer. Int J Epidemiol 1997;26:1172-81. 24. Launoy G, Smith TC, Duffy SW, Bouvier V. Colorectal cancer mass-screening: estimation of faecal occult blood test sensitivity, taking into account cancer mean sojourn time. International journal of cancer 1997;73:220-4. 25. Hardcastle JD, Thomas WM, Chamberlain J, et al. Randomised, controlled trial of faecal occult blood screening for colorectal cancer. Results for first 107,349 subjects. Lancet 1989;1:1160-4. 26. Kronborg O, Fenger C, Olsen J, Bech K, Sondergaard O. Repeated screening for colorectal cancer with fecal occult blood test. A prospective randomized study at Funen, Denmark. Scand J Gastroenterol 1989;24:599-606. 27. Hixson, L. J., M. B. Fennerty, et al. (1990). Prospective study of the frequency and size distribution of polyps missed by colonoscopy. J Natl Cancer Inst 1990;82: 1769-1772. 10
28. Rex DK, Cutler CS, Lemmel GT, et al. Colonoscopic miss rates of adenomas determined by back-to-back colonoscopies. Gastroenterology 1997;112:24-8. 29. Winawer SJ, Zauber AG, O'Brien MJ, et al. Randomized comparison of surveillance intervals after colonoscopic removal of newly diagnosed adenomatous polyps. The National Polyp Study Workgroup. N Engl J Med 1993;328:901-6. 11
Supplement Table 1. Characteristics of All Adenoma Patients at Initial Polypectomy as Observed in the National Polyp Study and as Simulated in the MISCAN-Colon Model Characteristic Age NPS Observed Adenoma Cohort (N=2602) MISCAN Model <50 13% 11% 50-59 26% 27% 60-69 36% 40% >70 25% 23% Adenoma Size 10mm 43% 44% 10mm 57% 56% No. of adenomas 1 59% 61% 2 22% 23% 3 19% 16% 12
Supplement Table 2. Deaths from Colorectal Cancer in the Observed Adenoma Cohort with Polypectomy Compared to Simulated Adenoma Cohort without Polypectomy. Adenoma No. Person- NPS Microsimulation Reference Group Cohort Years at Risk MISCAN-Colon Model of Adenoma Subjects with No Polypectomy Observed Expected SMR 95%CI % Reduction P-value CRC Deaths CRC Deaths Without Polypectomy All years 2602 37,073 12 145 0.08 0.04-0.14 92% <0.001 followed < 10 years 2602 22,903 4 63 0.06 0.02-0.15 94% <0.001 10 years 2031 14,170 8 82 0.10 0.05-0.19 90% <0.001 13
SUPPLEMENT APPENDIX Supplement Appendix Table 1. Main Assumptions in the MISCAN-Colon Model in Simulating the National Polyp Study Adenoma Cohort Parameter Value Based on Adenoma incidence in general population Age and sex dependent: Adenoma prevalence in autopsy and colonoscopy studies of 15% in age group 50-59 to 33% in age group 70+ 12-21 and cancer incidence in SEER registry in 1978 10 Distribution of risk for adenomas over the general population Gamma distributed, mean 1, variance 2 Multiplicity distribution of adenomas in autopsy studies 12-21 Duration distributions in preclinical stages Exponential Expert opinion, other cancer models 22-24 Mean duration of non-progressive adenomas Lifelong Expert opinion Mean duration of progressive adenomas 16.4 yrs Expert opinion and references 7-8 Mean duration of preclinical cancer 3.6 yrs Cancer detection rate at first screening and background cancer incidence in FOBT trials 25,26 Probability to develop cancer from 0% Expert opinion removed adenoma 14
Parameter Value Based on Sensitivity of initial and surveillance colonoscopic examination adenoma 5mm: 75% adenoma 6-9mm: 85% Back-to-back colonoscopy studies 27,28 adenoma 10+mm: 95% cancer: 95% Sensitivity of fictitious screening test adenoma 5mm: 15% adenoma 6-9mm: 28% adenoma 10+mm: 99% cancer: 100% Adenoma size at initial polypectomy of all patients included in the National Polyp Study The NPS cohort was constructed to represent patients referred to colonoscopy on the basis of a positive screening test comparable to that of fecal occult blood test or flexible sigmoidoscopy which were the predominant screening reasons for referral to the NPS. Reach of initial and surveillance 100% National Polyp Study design 29 colonoscopic examination 15
PARTICIPATING CENTERS IN THE NATIONAL POLYP STUDY The seven centers participating in the National Polyp study are the Memorial Sloan- Kettering Cancer Center (New York, NY), SJ Winawer (PI); Mt. Sinai Medical Center (New York, NY) JD Waye (PI); Veterans Affairs Medical Center (Minneapolis, MN), JH Bond (PI); Medical College of Wisconsin (Milwaukee, WI), E Stewart and W Hogan (PI s); Massachusetts General Hospital (Boston, MA), S. Hedberg (deceased), F. Ackroyd (PIs) ; Cedars-Sinai Medical Center (Los Angeles, CA), J Panish (PI); and Valley Presbyterian Hospital (Van Nuys, CA), M. Schapiro (PI). The Mallory Institute of Pathology, Boston University School of Medicine (Boston, MA) was the site for pathology review team: MJ O Brien (Boston University), SS Sternberg (Memorial Sloan-Kettering), and LS Gottlieb (Boston University) (deceased). 16