Colorectal cancer (CRC) is the third most common

Transcription

1 GASTROENTEROLOGY 2010;139: CLINICAL The Reduction in Colorectal Cancer Mortality After Colonoscopy Varies by Site of the Cancer HARMINDER SINGH,*, ZOANN NUGENT,, ALAIN A. DEMERS,, ERICH V. KLIEWER,,, SALAHEDDIN M. MAHMUD,, and CHARLES N. BERNSTEIN* *Internal Medicine and University of Manitoba IBD Clinical and Research Centre, Winnipeg, Manitoba; Community Health Sciences, University of Manitoba, Winnipeg, Manitoba; Epidemiology and Cancer Registry, CancerCare Manitoba, Winnipeg, Manitoba; and Cancer Control Research, British Columbia Cancer Agency, Vancouver, British Columbia, Canada This article has an accompanying continuing medical education activity on page e12. Learning Objective: Upon completion of reading this article, successful learners will be able to appreciate the effect of colonoscopy on deaths from colon cancers occurring in different parts of the colon, and will consider potential causes for the differences. See related articles, PohlHetal, on page 858 and Hassan C et al, on page 865 in CGH. BACKGROUND & AIMS: There are limited data on mortality from colorectal cancer (CRC) among patients who have received colonoscopy examinations. We sought to determine CRC mortality among persons undergoing colonoscopies compared with the general population. METHODS: We identified all individuals who had a colonoscopy as their first lower gastrointestinal endoscopy from April 1, 1987, to September 30, 2007 (24,342 men and 30,461 women), based on information from Manitoba s provincial physicians billing claims database. Patients were followed until March 31, 2008, death, or migration out of the province (310,718 person-years). Individuals less than 50 years of age or more than 80 years of age at the index colonoscopy or those with prior sigmoidoscopy, inflammatory bowel disease, resective colorectal surgery, or CRC were excluded. CRC mortality after the index colonoscopy was compared with that of the general population by standardized mortality ratios (SMRs). Stratified analyses were performed to determine CRC mortality for different age groups at index colonoscopy and sex, duration of follow-up, medical specialty of the endoscopist, and site of CRC. RESULTS: There was a 29% reduction in overall CRC mortality (SMR, 0.71; 95% confidence interval [CI], ], a 47% reduction in mortality from distal CRC (SMR, 0.53; 95% CI, ), and no reduction in mortality from proximal CRC (SMR, 0.94; 95% CI, ). The reduction in mortality from distal CRC remained significant for follow-up beyond 10 years (SMR, 0.53; 95% CI, ). CON- CLUSIONS: In Manitoba, colonoscopies significantly reduce mortality from CRC, but the benefit is not uniform for colorectal tumors that arise in different areas of the colon. Keywords: Screening; Prevention; Right Colon; Rectosigmoid. Colorectal cancer (CRC) is the third most common cancer among both men and women and continues to be the second most common cause of cancer-related deaths and premature mortality in North America. 1 The slow development of most CRCs provides a long window of opportunity to detect the precursor lesions and early stage CRCs, thereby preventing development of CRC and reducing the death rate because of CRC. Even though there has been a recent focus on CRC incidence, diseasespecific mortality reduction is the most widely accepted end point for cancer screening. 2 Colonoscopy has become the preferred modality for CRC screening in the United States 3 and is considered an acceptable strategy in many other countries. 4 However, there are limited direct data on CRC mortality after colonoscopy. A recent case-control study from Ontario reported that, after a colonoscopy, there was no reduction in the mortality from CRCs occurring in the proximal colon. 5 Approximately 16% 28% of CRCs in Ontario are classified as site unspecified/overlapping, 6 in contrast to Abbreviations used in this paper: CI, confidence interval; CRC, colorectal cancer; GI, gastrointestinal; IBD, inflammatory bowel disease; ICD-9-CM, International Classification of Diseases, 9th revision, Clinical Modification; MCR, Manitoba Cancer Registry; MH, Manitoba Health; NPS, National Polyp Study; PHIN, personal health identification number; RCT, randomized control trials; SMR, standardized mortality ratios by the AGA Institute /$36.00 doi: /j.gastro

2 October 2010 COLONOSCOPY AND COLORECTAL CANCER MORTALITY % 6% in other jurisdictions, 7,8 and thus there could have been some CRC site misclassification in the Ontario study. It is also unclear whether the Ontario study design could have influenced the findings; patients in the study could have had other lower gastrointestinal (GI) endoscopies prior to their index colonoscopy, which could have potentially biased the results in either direction. Patients having repeated lower GI endoscopies might be higher risk individuals, such as those with a family history of CRC or a history of documented colonic polyps. On the other hand, those with multiple normal lower GI endoscopies will be at lower risk of being diagnosed with CRC. No other study has evaluated the CRC mortality after colonoscopy from proximal and distal CRCs separately. 9 Indeed, there is only one other large study evaluating the effect of colonoscopy on CRC mortality. 10 This previous case-control study, performed on male veterans in the United States (a higher risk group), had reported an approximate 55% reduction in mortality from CRC. To verify the validity and generalizability of their findings, these results need to be replicated in other populations and with alternate (eg, cohort) designs. 11 We conducted a population-based cohort study in Manitoba, Canada, to determine CRC mortality after colonoscopy. The primary objective of our study was to determine the difference in CRC mortality reduction from proximal and distal CRCs after colonoscopy. Our secondary objectives were to determine (1) the time duration of reduction in CRC mortality after colonoscopy and (2) the reduction in CRC mortality in the different age groups undergoing colonoscopy. Patients and Methods Description of Data Sources This study was performed using several of Manitoba Health s (MH) administrative databases and the Manitoba Cancer Registry. MH is the publicly funded health insurance agency providing comprehensive universal health insurance to all residents of Manitoba. There is minimal nonparticipation because the residents do not pay premiums to register for insured benefits; there are no copayments, deductibles, or income eligibility criteria. Only inmates of federal facilities, members of the armed forces, and members of the Royal Canadian Mounted Police are not covered by the provincial plan. MH maintains several electronic administrative databases to monitor the services provided and to provide reimbursement to the health care providers for the services rendered. These databases include the Medical Claims database, which includes all physician billings, Hospital Discharge Abstract database of all provincial hospital admissions; and MH Population Registry, a demographic, vital status, and migration status database of all permanent residents in the province. The accuracy and comprehensiveness of these administrative databases have been previously established in multiple studies. 12,13 The Medical Claims database contains information for each physician service, including the patients identification, date of service, single diagnosis (3 digit, International Classification of Diseases, 9th revision, Clinical Modification [ICD-9-CM] code) and service tariff code. The Hospital Discharge database includes for each hospitalization, the patient s identification, dates of admission and discharge, details of attending physicians, up to 16 ICD-9-CM diagnoses, and 12 procedures (Manitoba hospitals switched to International Classification of Diseases, 10th Revision, with Canadian Enhancements [ICD- 10-CA] and Canadian Classification of Interventions on April 1, Since then, the Hospital Discharge database records up to 25 diagnosis codes and 20 procedure codes performed during the hospitalization). Occurrence of CRC and death from CRC was determined from the Manitoba Cancer Registry (MCR), a population-based database recording all cancers diagnosed in residents of the province since Reporting to the MCR is mandated by law. The coding and capture of cancer data are audited regularly by the North American Association of Central Cancer Registries and the Canadian Cancer Registry. The quality of the MCR cancer registrations is high with consistently high levels of reporting completeness and histologic verification. 14,15 The MCR receives reports from Vital Statistics on cause of death and investigates all cases where the reported cause of death is cancer. Since 1984, all permanent residents of Manitoba have been assigned a unique personal health identification number (PHIN) by MH. With the PHIN as a key identifier, longitudinal health service use and outcomes in the province can be determined by deterministic linkage of health utilization files and other databases such as the MCR. To protect patient anonymity, all linkages in this study were performed using scrambled PHINs. Study Cohort Individuals undergoing colonoscopy were identified from the MH s physicians claims database (Figure 1). Because all endoscopists in the province have always been reimbursed on a fee-for-service basis, there is a financial impetus to report the procedures when performed. MH intermittently audits the clinical records for the services provided to ensure there is no overbilling. Over the years, there have been no major changes in the tariff codes for lower gastrointestinal (GI) endoscopy, other than that a separate billing code for colonoscopy with polypectomy using hot biopsy forceps was established in The physician claims database was also used to identify lower GI endoscopy, prior resective colorectal surgery and inflammatory bowel disease (IBD), and the medical speciality of the endoscopists. IBD was identified using a previously validated administrative data

3 1130 SINGH ET AL GASTROENTEROLOGY Vol. 139, No. 4 index colonoscopies diagnostic for CRC; this approach is similar to that adopted in previous studies. 5,9,10 Using sensitivity analysis, others have indicated the validity of this approach. 5 Outcomes The study cohort was followed up to death, migration from the province, or end of the study period (March 31, 2008). CRCs (and death from CRC) were identified using ICD-9-CM codes (used from 1956 to 2001 in the MCR) , , and and ICD-10-CA codes (used from 2002 onward) C18.0, C18.2 C18.9, C19, C20, and C26.0. We included cases with CRC who were listed to have died from another cancer (lung, primary unknown, liver) in records from Vital Statistics but had no record of diagnosis of another cancer in MCR. Proximal CRCs included cancers occurring in and proximal to the splenic flexure (ICD-9 CM codes 153.0, 153.1, 153.4, 153.6, and 153.7; ICD-10-CA codes C18.0, C18.2 C18.5). Distal CRCs included cancers distal to the splenic flexure (ICD-9 CM codes 153.2, 153.3, 154.0, and 154.1; ICD-10-CA codes C18.6, C18.7, C19, and C20). We assigned the subsite of CRC in the colon for CRC mortality as that recorded at the time of cancer diagnosis. Figure 1. Identification of the study cohort. algorithm, 16 which also relied on data from MH s hospital discharges database. We identified all individuals having a colonoscopy as their first lower GI endoscopy between April 1, 1984, and September 30, Individuals having their first colonoscopy between April 1, 1987, and September 30, 2007, were included in the study. Manitoba s databases can be reliably linked from April 1, 1984, onward, and, therefore, a minimum 3-year lead-in time period was used to identify and exclude individuals with history of prior lower GI endoscopy, IBD, or resective colorectal surgery. They were excluded because their risk for CRC is likely different than that for the average risk individuals. All included individuals were between the age of 50 and 80 years on the date of the index colonoscopy. They also had to be registered with MH for at least 3 years prior to the index colonoscopy. In addition, individuals with a history of CRC prior to the index colonoscopy, as identified from the MCR, were also excluded because the study evaluated the benefit of colonoscopy prior to CRC diagnosis. We excluded index colonoscopies associated with a CRC diagnosis within 6 months so as to exclude Statistical Analysis The CRC mortality in the colonoscopy cohort was compared with the age-, sex-, and calendar year-adjusted CRC mortality rates in the rest of the Manitoba population; individuals with CRC diagnosed prior to the age of the individuals under consideration in the study cohort were excluded from the calculation of the CRC mortality rate in the Manitoba population. Results were expressed as standardized mortality ratios (SMRs), calculated by indirect standardization. 17 In stratified analyses, the SMRs were calculated for men and women separately. We also stratified by age at colonoscopy (50 64, 65 74, and years). The SMRs were calculated for mortality from all CRCs, proximal CRCs, and distal CRCs. Individuals with synchronous or metachronous CRCs were included in the analysis of mortality from proximal and distal CRC, only if all CRCs in an individual occurred in proximal or distal colon, respectively. Confidence intervals (CIs) were calculated assuming a Poisson distribution. 18 We also determined the SMR for mortality from CRC after colonoscopy with no features of colorectal neoplasia (CRC and/or colorectal polyps), ie, negative colonoscopy. Similar to prior studies (including ours), we defined negative colonoscopy as a colonoscopy without any additional billable procedures (MH tariff code 3185) and without a diagnosis of CRC within 6 months of the index procedure. 6,19 22

4 October 2010 COLONOSCOPY AND COLORECTAL CANCER MORTALITY 1131 Sensitivity Analysis Cause of death listed on death certificates can be inaccurate. 23 We have an advantage in Manitoba because all suspected cancer-related deaths are investigated by cancer registrars to determine whether such individuals did have the respective cancer diagnosis. Moreover, in previous studies, verification of the cause of cancer death by independent experts, instead of those obtained from death certificates, did not markedly change the estimate of mortality reduction. 24 However, we conducted a sensitivity analysis to assess the potential impact of misclassification of cause of death. To achieve that, all the analyses were repeated after classifying all deaths within 5 years of the CRC diagnosis as deaths related to CRC. Most individuals with CRC who survive 5 years beyond the date of the cancer diagnosis are believed to have been cured of their cancer, 25 so deaths occurring after 5 years of diagnosis are less likely to be related to CRC. In addition, for all CRCs diagnosed in Manitoba during the study years, we calculated the conditional relative survival ratio by the life table method, using the SEER Survival System. 26 Conditional survival indicates the future survival when the cohort has already survived a given amount of time. Relative survival adjusts for the expected survival, calculated from the experience in the general population, with adjustment for age, gender, and calendar year of follow-up. The calculation of the conditional relative survival ratios over the follow-up time allowed us to determine the point at which mortality for individuals with CRC returned to the same level as that of the general population, which is also sometimes referred to as time to statistical cure. 27 We then repeated our SMR analyses, classifying all deaths before the statistical cure as deaths related to CRC. Data management and analyses were performed using SAS version 9.1 (SAS Institute, Cary, NC). This study was approved by the University of Manitoba s Health Research Ethics Board and Manitoba Health s Health Information and Privacy Committee. Results Between April 1, 1987, and September 30, 2007, there were 115,212 individuals in Manitoba who had a colonoscopy as their first lower GI endoscopy. Of these 54,803 individuals (24,342 men; 30,461 women) were included in the study cohort (Figure 1) and contributed a total of 310,718 person-years of follow-up time in the study. The median and interquartile range age at the index colonoscopy was 62 (56 70) years for men and 62 (55 70) years for women. There were 32,306 individuals between the ages of 50 and 80 years with a negative index colonoscopy, who were followed up for a total of 182,249 person-years. Overall, there was a 29% reduction in CRC mortality, with the largest reduction (39%) occurring between the 5 Table 1. Standardized Mortality Ratios for Overall CRC Mortality After Colonoscopy in Manitoba, Number of CRC deaths Observed Expected SMR 95% CI Overall Time since index colonoscopy 6 36 mo y y y Sex Male Female Age at index colonoscopy (y) Specialty of the endoscopist Gastroenterology General surgery Internal medicine General practice to 10 years of follow-up (Table 1). Individuals between the age of 75 and 80 years at the index colonoscopy had no reduction in deaths from all CRC, as compared with the rest of the individuals in the same age group in the general population; however, when stratified for site of CRC, all age groups had similar (42% 53%) reduction in deaths from distal CRC (Table 2). There was no reduction in deaths from proximal CRC (compared with the general population), even when stratified by sex or time since index colonoscopy (Table 2). There was a 47% reduction in deaths from distal CRC, which remained statistically significant for follow-up beyond 10 years. We repeated the analysis, with a modified definition of proximal CRCs to include CRCs in and proximal to the descending colon and restricting the distal CRCs to those occurring in the rectum, rectosigmoid, and sigmoid colon (parts of colon more likely to be examined during unsedated flexible sigmoidoscopy). This variation in definition of proximal and distal CRCs did not appreciably change the reduction seen in SMRs (Table 3). Following a negative colonoscopy, there was approximately a 10% greater reduction in overall CRC mortality than after all colonoscopies (Table 4). Analysis of the conditional relative annual survival rates suggested that individuals diagnosed with CRC in the province between 1987 and 1997 had no significant excess mortality beyond 6 years following CRC diagnosis (Figure 2). Classifying all deaths within 5 or 6 years of the CRC diagnosis as deaths related to CRC had no appre-

5 1132 SINGH ET AL GASTROENTEROLOGY Vol. 139, No. 4 Table 2. Standardized Mortality Ratios for Mortality From Proximal and Distal CRC, After Colonoscopy in Manitoba, Proximal CRC Distal CRC Observed Expected SMR 95% CI Observed Expected Overall Time since index colonoscopy 6 36 mo y y y Sex Male Female Age at index colonoscopy (y) Specialty of the endoscopist Gastroenterology General surgery Internal medicine General practice SMR 95% CI ciable effect on the SMRs (other than for the colonoscopies performed by gastroenterologists) (Table 5 for 6 years). We also analyzed the data based on the specialty of the endoscopist. When the index colonoscopy was performed by gastroenterologists, there was a trend toward a larger reduction in SMRs and a nonsignificant reduction in mortality from proximal CRCs, which became significant when the definition of proximal and distal CRCs was modified (as described above) (Tables 1, 2, and 3). There was a significant reduction in deaths from proximal CRCs after performance of negative colonoscopies by Table 3. Standardized Mortality Ratios for Mortality From Proximal and Distal CRC, After Colonoscopy in Manitoba, Proximal CRC Distal CRC Observed Expected SMR 95% CI Observed Expected Overall Time since index colonoscopy 6 36 mo y y y Sex Male Female Age at index colonoscopy (y) Specialty of the endoscopist Gastroenterology General surgery Internal medicine General practice NOTE. Analysis with modified definition of distal CRC to include only cancers of rectum, rectosigmoid, and sigmoid colon. SMR 95% CI

6 October 2010 COLONOSCOPY AND COLORECTAL CANCER MORTALITY 1133 Table 4. Standardized Mortality Ratios for CRC Mortality After Negative Colonoscopy in Manitoba, , Stratified for Site of CRC Proximal CRC Distal CRC Observed Expected SMR 95% CI Observed Expected Overall Time since index colonoscopy 6 36 mo y y y Sex Male Female Age at index colonoscopy (y) Specialty of the endoscopist Gastroenterology General surgery Internal medicine General practice SMR 95% CI gastroenterologists (Table 4). There was no reduction in deaths from CRC at either proximal or distal site when the index colonoscopy was performed by a general practice physician. Discussion This population-based cohort study suggests that reduction in CRC mortality after a colonoscopy in usual clinical practice is mostly limited to deaths from CRCs in the distal colon. This is a consistent finding irrespective of the time duration after the index colonoscopy, sex of the individuals undergoing colonoscopy, and whether or not the index colonoscopy is a negative colonoscopy. Reduction in mortality from distal CRC is sustained for Figure 2. Conditional relative median annual survival (and 95% CIs) for all individuals diagnosed with colorectal cancer in Manitoba between 1987 and follow-up beyond 10 years from the index colonoscopy. Individuals between the age of 75 and 80 years at time of the index colonoscopy have a similar reduction in deaths from distal CRC as those between 50 and 75 years of age. Sensitivity analysis suggests that these findings are likely not affected by potential misclassification of the cause of death on death certificates or variation in definition of proximal and distal CRCs. There may be greater reduction in CRC mortality when the index colonoscopy is performed by gastroenterologists. The variability in reduction by specialty of the endoscopist warrants further exploration and consideration in regard to training requirements for physicians who perform colonoscopy. Despite differences in study design (cohort vs case control) and study population (exclusion of individuals with a lower GI endoscopy prior to the index colonoscopy and restriction to individuals between the ages of 50 and 80 years at the index colonoscopy in our study), our finding of lower reduction of deaths from proximal CRC is remarkably consistent with the findings from the Ontario study, which had also found a 31% reduction in overall CRC mortality after colonoscopy. 5 We have extended the findings of the Ontario study and addressed some of the critiques of that study by other authors 11 by examining risks among individuals of different age groups at colonoscopy, at different durations of followup, and specifically assessing the effect of negative colonoscopy. Our initial hypothesis that there would be a larger and significant reduction in mortality from proximal CRC with long-term follow-up is not supported by our results.

7 1134 SINGH ET AL GASTROENTEROLOGY Vol. 139, No. 4 Table 5. Standardized Mortality Ratios for Mortality From Proximal and Distal CRC, After Colonoscopy in Manitoba, Proximal CRC Distal CRC Observed Expected SMR 95% CI Observed Expected Overall Time since index colonoscopy 6 36 mo y y y Sex Male Female Age at index colonoscopy (y) Specialty of the endoscopist Gastroenterology General surgery Internal medicine General practice SMR 95% CI NOTE. Analysis with classification of all deaths within 6 years of the CRC diagnosis as deaths related to CRC. These study results and the results from Ontario should not be interpreted as there being unequivocally no reduction of mortality because of proximal CRC after colonoscopies performed in usual clinical practice. Both of these studies may be biased by the inclusion of some high-risk individuals because these administrative data sets, for instance, could not assess for family history of CRC. Furthermore, it is possible the colonoscopy reduced the risk of proximal CRC mortality in these individuals to that in the general population hence actually achieving a benefit. Therefore, while the colonoscopy cohort may contain a greater percentage of at-high-risk persons for CRC than the general population, the finding of comparable mortality rates in the 2 cohorts suggests there may possibly be some beneficial impact on proximal CRC mortality. With the same caveats in mind, the marked reduction in distal CRC mortality is all the more impressive. Hence, the importance of these studies is the demonstration of markedly lower magnitude of reduction in deaths because of proximal CRC than that because of distal CRCs after colonoscopies performed in usual clinical practice. Lower reduction of mortality from proximal CRC could be due to a higher miss rate of proximal CRC on colonoscopy. 28 A higher miss rate of proximal CRC may mean that, when ultimately diagnosed, the cancer is more advanced and less curable. Proximal neoplasia may be more likely to be missed on colonoscopy because of the differences in the characteristics of the lesions in the proximal and distal colon. Flat appearing lesions, which are more likely to be missed, are more common in the proximal colon. 29,30 Other potential reasons for a higher miss rate of proximal CRC include worse bowel preparation in the proximal colon and incomplete colonoscopies. 31 Experienced endoscopists do have a high detection rate of flat lesions, even with regular endoscopes, 32 and, therefore, the recent emphasis on quality of colonoscopy examinations could lead to a higher detection rate of colonic lesions and perhaps higher CRC mortality reduction than seen after colonoscopy performed many years ago. Split dosing of the bowel preparation, where 33% 50% of the cleansing agent is given on the day of the procedure, can also result in higher detection of colonic neoplasia. 33 However, there are also differences in tumor biology between proximal and distal CRC, with proximal CRC carrying a worse prognosis. 34,35 In this context, it is important to recall that other modalities for CRC screening, including the newer immunochemical fecal tests, do have a lower effectiveness for proximal CRC outcomes. 36,37 No other study has evaluated the differences in CRC mortality by colon subsite after all colonoscopy. A recent cohort study from a single center in Indiana reported a 65% statistically nonsignificant reduction in CRC mortality after colonoscopy was performed on healthy professionals and their spouses when compared with the expected calculated rates from the US general population. 38 This study is from a tertiary referral center and hence may not be widely applicable. 39 It is interesting to note that, in this Indiana study, 6 of the 7 cases with CRC detected after the baseline colonoscopy were in the proximal colon. Our results are consistent with a recent study

8 October 2010 COLONOSCOPY AND COLORECTAL CANCER MORTALITY 1135 from Germany, which had found that the reduction in prevalence of advanced colorectal neoplasia after colonoscopy was limited to the left side of the colon. 40 We found that, although there was no reduction in overall CRC mortality among individuals between the ages of 75 and 80 years at the index colonoscopy (as compared with the general population of the same age), the reduction in mortality from distal CRC was similar in the different age groups. This discrepancy is likely due to a higher proportion of CRC occurring in the proximal colon in older individuals, 41 and, as this study shows, colonoscopy in usual practice having lower effectiveness in reducing mortality from proximal CRC. Although most studies on CRC screening modalities have excluded individuals older than 75 years of age, our results suggest that, if more effective strategies could be developed for detecting/preventing proximal CRC, individuals between the age of 75 and 80 years should also be considered for CRC screening. Our results suggest that endoscopic screening could prevent deaths from distal CRC in this age group. Ideally, benefits and disadvantages of colonoscopy should be established in randomized control trials (RCT). A large RCT on colonoscopy has just started in Europe but has not accrued all the subjects yet, so the results will not be available for many years. 42 In the interim, observational studies will continue to provide information on benefits of colonoscopy. Moreover, even after RCTs are performed, the benefit in usual clinical practice can only be determined by population-based cohort studies, similar to ours. Our study also suggests that colonoscopy could be associated with a large reduction in deaths from proximal CRC if the index colonoscopy was performed by gastroenterologists in routine clinical practice. Whereas this finding needs further evaluation, this is consistent with one of our recent studies in which we found the risk of diagnosis of CRC in the first 3 years (ie, early/missed CRC) following a negative colonoscopy was higher if the index colonoscopy was performed by nongastroenterologists. 21 The results of our study should be interpreted in the context of the study strengths and limitations. This is a population-based study, and, hence, the results are more likely to be generalizable than those from centers with specialist expertise in endoscopy. Manitoba s administrative databases have been validated in previous studies. The information in the MCR has been consistently found to be of high quality, including high levels of ascertainment, completeness, and histologic verification. The proportion of cancer cases diagnosed in Manitoba based on death certificates only or with subsite of colon unspecified (6% vs 16% 28% in Ontario) is lower than elsewhere in Canada. 5,6,8 Although our data are more direct than many of the previous studies because we studied CRC mortality after colonoscopy, our comparisons are using SMRs to the risk in the general population and are therefore indirect for the different categories of variables used for stratification. We did not have clinical details, such as indications, extent and quality of the colonoscopy examination, or information on several risk factors for CRC. Because we did not have information on indications, we used a 6-month window to exclude index examinations diagnostic for CRC, a widely used and accepted approach. 5,9 Our results are, therefore, applicable to those without CRC diagnosed within 6 months of the index colonoscopy and for follow-up beyond 6 months of the initial procedure. With regard to indications, it is important to recall that, even though the seminal National Polyp Study (NPS), which established the benefits of colonoscopy, recruited patients referred for symptoms, abnormal test results, and/or positive family history, NPS had demonstrated a marked reduction in CRC incidence as compared with the general population in the United States. 43 The predictive value for most symptoms for CRC is low, and, therefore, even though a colonoscopy may be performed for symptoms, it will also serve as a screening colonoscopy. 44 The NPS had also excluded individuals with CRC diagnosed at the index colonoscopy examinations in their analysis. We did not have information on lower GI endoscopies performed prior to 1984 because the Manitoba Health electronic databases started in 1984, and, therefore, individuals in our colonoscopy cohort could have had lower GI endoscopies prior to However, lower GI endoscopies were performed much less frequently in that time period, including the early 1980s in Manitoba. 8 Even though many experts consider disease-specific mortality as the outcome of interest for cancer screening, there are inaccuracies inherent in determining cause of death. Therefore, we conducted a sensitivity analysis including for all deaths within 5 years of CRC diagnosis as deaths related to CRC. The results were consistent with our primary analysis. There is no reason to expect that the limitations of our study should have a differential effect on the risk of mortality from proximal and distal CRCs. In conclusion, this study suggests that, after a colonoscopy in usual clinical practice, there is significant reduction in CRC mortality; however, the benefit is much larger for reduction in mortality from distal CRC and less evident for proximal CRC. Further research is required to determine the factors contributing to the disparity in CRC reduction rates between proximal and distal CRC mortality. Finally, the reduction in CRC mortality was not uniform across endoscopist specialty of practice. This too requires further exploration. It may reflect different approaches to training or different approaches to practice.

9 1136 SINGH ET AL GASTROENTEROLOGY Vol. 139, No. 4 References 1. Canadian Cancer Society s Steering Committee. Canadian Cancer Statistics Toronto, Canada: Canadian Cancer Society, Black WC, Haggstrom DA, Welch HG. All-cause mortality in randomized trials of cancer screening. J Natl Cancer Inst 2002;94: Klabunde CN, Lanier D, Nadel MR, et al. Colorectal cancer screening by primary care physicians: recommendations and practices, Am J Prev Med 2009;37: Arditi C, Peytremann-Bridevaux I, Burnand B, et al. Appropriateness of colonoscopy in Europe (EPAGE II). Screening for colorectal cancer. Endoscopy 2009;41: Baxter NN, Goldwasser MA, Paszat LF, et al. Association of colonoscopy and death from colorectal cancer. Ann Intern Med 2009;150: Lakoff J, Paszat LF, Saskin R, et al. Risk of developing proximal versus distal colorectal cancer after a negative colonoscopy: a population-based study. Clin Gastroenterol Hepatol 2008;6: Saltzstein SL, Behling CA. Age and time as factors in the left-toright shift of the subsite of colorectal adenocarcinoma: a study of 213,383 cases from the California Cancer Registry. J Clin Gastroenterol 2007;41: Singh H, Demers AA, Xue L, et al. Time trends in colon cancer incidence and distribution and lower gastrointestinal endoscopy utilization in Manitoba. Am J Gastroenterol 2008;103: Ransohoff DF. How much does colonoscopy reduce colon cancer mortality? Ann Intern Med 2009;150: Muller AD, Sonnenberg A. Protection by endoscopy against death from colorectal cancer. A case-control study among veterans. Arch Intern Med 1995;155: Romagnuolo J, Barkun AN, Manges AR. The effectiveness of colonoscopy in reducing mortality from colorectal cancer. Ann Intern Med 2009;150: Roos LL, Mustard CA, Nicol JP, et al. Registries and administrative data: organization and accuracy. Med Care 1993;31: Robinson JR, Young TK, Roos LL, et al. Estimating the burden of disease. Comparing administrative data and self-reports. Med Care 1997;35: Chen VW, Wu XC. Incidence, cancer in North America, In: Andrews PA, ed. North American Association of Cancer Registries; April Springfield, IL: North American Association of Central Cancer Registries, Inc, Hotes Ellison J, Wu XC, et al, editors. Cancer in North America: Volume 1: incidence. Springfield, IL: North American Association of Central Cancer Registeries Inc, 2006:II Bernstein CN, Blanchard JF, Rawsthorne P, et al. Epidemiology of Crohn s disease and ulcerative colitis in a central Canadian province: a population-based study. Am J Epidemiol 1999;149: Hennekens CH, Buring JE. Epidemiology in Medicine. Boston/ Toronto: Little, Brown and Co, 1997: Esteve J, Benhamou E, Raymond L. Statistical methods in cancer research. Volume IV. Descriptive epidemiology. IARC Sci Publ 1994;128: Singh H, Turner D, Xue L, et al. Risk of developing colorectal cancer following a negative colonoscopy examination: evidence for a 10-year interval between colonoscopies. JAMA 2006;295: Singh G, Gerson LB, Wang H, et al. Screening colonoscopy, colorectal cancer and gender: an unfair deal for the fair sex? Gastrointest Endosc 2007;65:AB Singh H, Nugent Z, Mahmud SM, et al. Predictors of colorectal cancer after negative colonoscopy: a population-based study. Am J Gastroenterol 2010;105: Rabeneck L, Paszat LF, Saskin R. Endoscopist specialty is associated with incident colorectal cancer after a negative colonoscopy. Clin Gastroenterol Hepatol 2010;8: Robinson MH, Rodrigues VC, Hardcastle JD, et al. Faecal occult blood screening for colorectal cancer at Nottingham: details of the verification process. J Med Screen 2000;7: Hardcastle JD, Chamberlain JO, Robinson MH, et al. Randomised controlled trial of faecal-occult-blood screening for colorectal cancer. Lancet 1996;348: Sasieni PD, Adams J, Cuzick J. Avoidance of premature death: a new definition for the proportion cured. J Cancer Epidemiol Prev 2002;7: Merrill RM, Henson DE, Barnes M. Conditional survival among patients with carcinoma of the lung. Chest 1999;116: Yu B, Tiwari RC, Cronin KA, et al. CANSURV: a Windows program for population-based cancer survival analysis. Comput Methods Programs Biomed 2005;80: Bressler B, Paszat LF, Chen Z, et al. Rates of new or missed colorectal cancers after colonoscopy and their risk factors: a population-based analysis. Gastroenterology 2007;132: Soetikno RM, Kaltenbach T, Rouse RV, et al. Prevalence of nonpolypoid (flat and depressed) colorectal neoplasms in asymptomatic and symptomatic adults. JAMA 2008;299: Park DH, Kim HS, Kim WH, et al. Clinicopathologic characteristics and malignant potential of colorectal flat neoplasia compared with that of polypoid neoplasia. Dis Colon Rectum 2008; 51: Rex DK, Eid E. Considerations regarding the present and future roles of colonoscopy in colorectal cancer prevention. Clin Gastroenterol Hepatol 2008;6: Gorgun E, Church J. Flat adenomas of the large bowel: a single endoscopist study. Dis Colon Rectum 2009;52: Unger RZ, Amstutz SP, Seo DH, et al. Willingness to undergo split-dose bowel preparation for colonoscopy and compliance with split-dose instructions. Dig Dis Sci 2010;55: Meguid RA, Slidell MB, Wolfgang CL, et al. Is there a difference in survival between right- versus left-sided colon cancers? Ann Surg Oncol 2008;15: Benedix F, Kube R, Meyer F, et al. Comparison of 17,641 patients with right- and left-sided colon cancer: differences in epidemiology, perioperative course, histology, and survival. Dis Colon Rectum 2010;53: Hundt S, Haug U, Brenner H. Comparative evaluation of immunochemical fecal occult blood tests for colorectal adenoma detection. Ann Intern Med 2009;150: Morikawa T, Kato J, Yamaji Y, et al. A comparison of the immunochemical fecal occult blood test and total colonoscopy in the asymptomatic population. Gastroenterology 2005;129: Kahi CJ, Imperiale TF, Juliar BE, et al. Effect of screening colonoscopy on colorectal cancer incidence and mortality. Clin Gastroenterol Hepatol 2009;7: Robertson DJ. Is screening colonoscopy effective in preventing cancer? Some answers, more questions. Clin Gastroenterol Hepatol 2009;7: Brenner H, Hoffmeister M, Arndt V, et al. Protection from rightand left-sided colorectal neoplasms after colonoscopy: population-based study. J Natl Cancer Inst 2010;102: Rabeneck L, Davila JA, El Serag HB. Is there a true shift to the right colon in the incidence of colorectal cancer? Am J Gastroenterol 2003;98: Nordic Initiative on Colorectal Cancer. Colorectal cancer a pan- Nordic study about the value of colonoscopy screening is required. Lakartidningen 2006;103: Winawer SJ, Zauber AG, O Brien MJ, et al. The National Polyp Study. Design, methods, and characteristics of patients with

10 October 2010 COLONOSCOPY AND COLORECTAL CANCER MORTALITY 1137 newly diagnosed polyps. The National Polyp Study Workgroup. Cancer 1992;70(Suppl 5): Ford AC, Veldhuyzen van Zanten SJ, Rodgers CC, et al. Diagnostic utility of alarm features for colorectal cancer: systematic review and meta-analysis. Gut 2008;57: Received April 12, Accepted June 16, Reprint requests Address requests for reprints to: Harminder Singh, MD, MPH, Section of Gastroenterology, McDermot Avenue, Winnipeg, Manitoba, Canada R3E3P4. fax: (204) Acknowledgments The results and conclusions are those of the authors, and no official endorsement by Manitoba Health is intended or should be inferred. Conflicts of interest The authors disclose no conflicts. Funding Supported by a grant from the Manitoba Health Research Council, and by an ACG Junior Faculty Development Grant (to H.S.). Dr Bernstein is supported in part by a Crohn s and Colitis Foundation of Canada Research Scientist Award and the Bingham Chair in Gastroenterology. Dr Singh is supported in part by an ACG Junior Faculty Development Award.