Downloaded from http://gut.bmj.com/ on November 14, 2015 - Published by group.bmj.com
Gut Online First, published on December 15, 2014 as 10.1136/gutjnl-2014-307729 GI cancer
ORIGINAL ARTICLE
Ten-year incidence of colorectal cancer following a negative screening sigmoidoscopy: an update from the Colorectal Cancer Prevention (CoCaP) programme V Paul Doria-Rose,1 Theodore R Levin,2,3 Albert Palitz,3 Carol Conell,2 Noel S Weiss4,5 ▸ Additional material is published online only. To view please visit the journal online (http://dx.doi.org/10.1136/ gutjnl-2014-307729). For numbered affiliations see end of article. Correspondence to Dr V Paul Doria-Rose, Health Services and Economics Branch, Applied Research Program, Division of Cancer Control and Population Sciences, National Cancer Institute, 9609 Medical Center Drive, Room 3E438, Bethesda, MD 20892-9762, USA; [email protected] Received 24 May 2014 Revised 15 November 2014 Accepted 20 November 2014
ABSTRACT Objective To examine the rates of colorectal cancer (CRC) following a negative screening sigmoidoscopy. Design Cohort study. Setting An integrated healthcare delivery organisation in California, USA. Participants 72 483 men and women aged 50 years and above who had a negative screening sigmoidoscopy between 1994 and 1996. Those at elevated risk of CRC due to inflammatory bowel disease, prior polyps or CRC, or a strong family history of CRC were excluded. Main outcome measures Incidence rates of distal and proximal CRC. Standardised Incidence Ratios were used to compare annual incidence rates of distal and proximal CRC in the cohort to expected rates based on Surveillance, Epidemiology, and End Results data. Additionally, rate ratios (RR) and rate differences (RD) comparing the incidence rate of distal CRC in years 6+ postscreening with that in years 1–5 were calculated. Results Incidence rates of distal CRC were lower than those in the San Francisco Bay area population at large during each of the first 10 years postsigmoidoscopy screening. However, the incidence of distal CRC rose steadily, from 3 per 100 000 in the first year of follow-up to 40 per 100 000 in the 10th year. During the second half of follow-up, the rate of distal CRC was twice as high as in the first half (RR 2 .08, 95% CI 1.38 to 3.16; RD 14 per 100 000 person-years, 95% CI 6 to 22). Conclusions Though still below population levels, the incidence of CRC during years 6–10 following a negative sigmoiodoscopy is appreciably higher than during the first 5 years.
INTRODUCTION
To cite: Doria-Rose VP, Levin TR, Palitz A, et al. Gut Published Online First: [please include Day Month Year] doi:10.1136/gutjnl2014-307729
Randomised controlled trials have demonstrated the ability of flexible sigmoidoscopy screening to prevent colorectal cancer (CRC) incidence and mortality.1–5 While these trials provide strong evidence of screening effectiveness, the most appropriate interval for screening is less clear-cut. Current US screening guidelines for average-risk individuals recommend either screening colonoscopy every 10 years or screening sigmoidoscopy every 5 years.6 7 Several studies have addressed the question of the appropriate interval for endoscopic screening by examining patterns of CRC incidence or mortality following a baseline sigmoidoscopy,8–11 colonoscopy12–16 or unspecified lower endoscopy.17–19
Significance of this study What is already known on this subject?
▸ Several prior studies have examined patterns in the incidence of colorectal cancer (CRC) following sigmoidoscopy or colonoscopy. ▸ However, the absolute risk of cancer following a negative examination has been estimated in few large cohorts of patients receiving screening endoscopies.
What are the new findings?
▸ The incidence of distal CRC remains low for at least a decade following a negative screening sigmoidoscopy. ▸ However, the relative risk of distal cancer doubles in the second 5-year period following screening as compared with the first 5-year period.
How might it impact on clinical practice in the foreseeable future?
▸ These results will allow clinicians to provide estimates of the absolute risk of CRC following a negative screening sigmoidoscopy, and will inform the choice of appropriate screening interval.
While the low incidence of CRC seems to be longlasting (ie, 10 or more years) following endoscopy, there are several limitations of existing studies, such as the inclusion of non-screening endoscopies,11–16 18 19 which may define a group at a greater underlying risk of CRC, the use of patient self-report to determine the occurrence and timing of endoscopy,10 17 and small sample size.9 In previous work, we used data from Kaiser Permanente Northern California (KPNC) to define a cohort of over 70 000 individuals with a negative screening sigmoidoscopy, and observed a low incidence of CRC during the first 5 years postscreening.20 To address the potential impact of a 10-year versus a 5-year screening interval, we now update these findings to include additional data over a total of 10.5 years of follow-up. We examined whether a large increase in the risk of CRC occurred during the second 5-year period postsigmoidoscopy.
Doria-Rose VP, et al. Gut 2014;0:1–7. doi:10.1136/gutjnl-2014-307729
1
Copyright Article author (or their employer) 2014. Produced by BMJ Publishing Group Ltd (& BSG) under licence.
Downloaded from http://gut.bmj.com/ on November 14, 2015 - Published by group.bmj.com
GI cancer METHODS Study population, data sources and follow-up KPNC is a large, integrated healthcare delivery system in California, serving more than 3.3 million members from Santa Rosa to the north, Sacramento to the east and Fresno to the south. In 1994, KPNC initiated the Colorectal Cancer Prevention (CoCaP) programme, which offered a screening flexible sigmoidoscopy at least once every 10 years to all plan members aged 50 years or older. During this time, sigmoidoscopy screening was vigorously promoted within the KPNC health system through active inreach to patients at the time of an office visit for other reasons. Sigmoidoscopy exams were performed almost exclusively without sedation; bowel preparation was generally two enemas performed prior to the procedure. For the first 3 years of the CoCaP programme, sigmoidoscopy report data from all KPNC facilities were entered into a computerised database. These data included several items based on patient self-report: family history of CRC, history of prior sigmoidoscopy and colonoscopy, and personal history of CRC and colorectal polyps. Additionally, endoscopists reported the depth of insertion of the sigmoidoscope, whether the exam was limited by the presence of stool, and the number and size of any polyps identified. Both the patient and the endoscopist reported whether the sigmoidoscopy was performed due to symptoms or for screening. Other sources of data included KPNC membership databases, which provided information on patient demographics (age, sex and race/ethnicity) and enrolment dates, and various inpatient and outpatient databases which documented dates and types of medical procedures and diagnoses occurring during the follow-up period. Finally, the KPNC Tumour Registry identified all cases of CRC, collecting information on date of diagnosis, location within the colorectum, and stage. Using the CoCaP database, we previously defined a cohort of 72 483 KPNC members who had a screening sigmoidoscopy negative for both polyps and cancer between 1994 and 1996; approximately 80% of the total number of those screened were included based on their negative exam.21 Cohort inclusion and exclusion criteria, and cancer incidence during the first 5 years after negative sigmoidoscopy have previously been described.20 Briefly, cohort members were men and women aged 50 years and older, who were considered to be at average risk of developing CRC. Patients were excluded from the cohort if either the patient or the endoscopist indicated that the sigmoidoscopy was not for screening, if the endoscopist noted that the examination was being performed due to inflammatory bowel disease or if the patient reported a personal history of colorectal polyps/ cancer and/or a high-risk family history of CRC (defined as one first-degree relative diagnosed with CRC before age 55 years or CRC in multiple first-degree relatives). Patient follow-up concluded at the time of diagnosis of distal (rectum and sigmoid colon) or proximal (other colon sites excluding the appendix) CRC, death, health plan membership termination or subsequent colorectal procedures including sigmoidoscopy, colonoscopy or barium enema. Sigmoidoscopy resulted in censoring only in the analysis of distal CRC incidence, since the occurrence of sigmoidoscopy would have negligible impact on the subsequent incidence of proximal cancer. Colorectal examinations resulted in censoring because the goal of this type of analysis is to determine the amount of time required after a negative screen for CRC to be diagnosed, and subsequent exams could alter this time course due to additional interventions (ie, polypectomy). However, procedures occurring within 6 months prior to CRC diagnosis were considered to be part of the diagnosis process 2
and did not result in censoring. This updated analysis includes all CRC diagnoses and censoring events through 10.5 years of follow-up.
Statistical analysis Descriptive statistics were examined for demographic and sigmoidoscopy exam characteristics of cohort members. χ2 Tests were used to determine whether the stage distribution of colorectal adenocarcinomas differed according to cancer site (distal or proximal) or time since the negative screen (years 1–5 vs 6+). The overall proportion of subjects diagnosed with CRC during follow-up was estimated using the Kaplan–Meier product limit estimator. To examine the temporal pattern of CRC incidence following a negative screening sigmoidoscopy, annual rates of distal and proximal colorectal adenocarcinomas were calculated by dividing the number of colorectal adenocarcinoma cases by the total person-years at risk for each year postscreening. This analysis uses a broader definition of adenocarcinoma than was used in the 5-year follow-up of this cohort (see online supplementary appendix 1).20 Further, updates to the KPNC Tumour Registry after the data were extracted for the 5-year follow-up also contributed to the net addition of more cases. As a result, the incidence rates we report here over the first 5 years of follow-up are higher than those previously reported.20 However, the general pattern of incidence (very low initially, increasing gradually over 5 years) was consistent regardless of adenocarcinoma definition. The duration of low CRC risk following a negative sigmoidoscopy was estimated in two separate ways using both external and internal comparisons. First, standardised incidence ratios (SIR) and 95% CIs22 were used to compare the observed annual incidence rates of distal and proximal colorectal adenocarcinomas with expected rates based on incidence in the San Francisco Bay area as a whole, as reported to the Surveillance, Epidemiology, and End Results (SEER) Tumour Registry for the years 1994–2006.23 Standardisation was performed on the basis of sex, age (in 5-year categories through age 84 years and 85 years and over), and calendar year. Second, we compared incidence rates of distal colorectal adenocarcinomas over all person-time accruing during the second half of follow-up (years 6 plus) to those in the first 5 years of follow-up; rate ratios (RR), rate differences (RD) and their associated 95% CI were calculated,22 providing an estimate of the amount of disease that could be prevented with a 5-year screening interval as compared with a 10-year screening interval. These RRs and RDs were calculated without adjusting for age, since examining patterns of cancer incidence to help inform the appropriate timing for rescreening should be based on the unadjusted, absolute risk of disease; for this purpose, it is not relevant whether any increase in risk over time is due to ageing of the cohort, increased time elapsed since the negative screening exam or a combination of the two. RRs and RDs were calculated both overall and for subgroups based on demographic and sigmoidoscopy exam characteristics. No subjects had missing data for any of the key variables used for the calculation of crude and adjusted CRC incidence rates; for subgroup analyses, when a substantial proportion of the data for any variable were missing (race/ethnicity, endoscopist specialty), results are reported separately for the missing/unknown category. Patterns of colorectal procedures were also examined. One concern for this analysis was that, as cohort members returned for rescreening later in follow-up, there would be an artefactual increase in cancer incidence reflecting prevalent cases diagnosed (earlier than they otherwise would have been) at rescreening. Doria-Rose VP, et al. Gut 2014;0:1–7. doi:10.1136/gutjnl-2014-307729
Downloaded from http://gut.bmj.com/ on November 14, 2015 - Published by group.bmj.com
GI cancer This was a particular concern at approximately 10 years postscreening, since KPNC recommends a 10-year screening interval for sigmoidoscopy. Though the indication for the initial negative sigmoidoscopy was available in the CoCaP database, indication for any subsequent censoring procedures was not. However, KPNC maintains a clinical database of all International Classification of Diseases (ICD)-9 diagnoses/V codes for all inpatient and outpatient visits that cover the entire study period. This enabled us to classify censoring procedures as ‘diagnostic’ or ‘not diagnostic’, based on whether there were symptoms that could lead to endoscopy/barium enema and V codes indicating screening (see online supplementary appendix 2). Any relevant symptom codes in the 6 months prior to the procedure (including day of exam) resulted in the classification as diagnostic, unless a V code for screening was also present during that time. We considered any procedures within the 6 months before or after the anniversary of the initial screening sigmoidoscopy to be conducted at approximately that anniversary date; for example, all exams occurring from 9.5 through 10.5 years of follow-up were classified as 10-year exams.
RESULTS Characteristics of the study cohort are shown in table 1. This analysis is based on 584 106 person-years of follow-up (mean 8.1 per person) that accrued during the 10.5-year follow-up period. During this time, the experience of 45 764 individuals was censored for reasons other than development of CRC or end of the follow-up period; reasons for censoring are shown in table 2. Over the entire follow-up period, 103 cases of distal and 333 cases of proximal colorectal adenocarcinomas were diagnosed. Approximately three-quarters of cancers were proximal in both the first and second halves of follow-up (table 2). For both distal and proximal cancers, stage distribution was also very similar in 1–5 versus 6+ years. However, distal cancers were more likely than proximal cancers to be diagnosed at an earlier stage (42% vs 31% localised, p=0.01) (table 3). During follow-up, 0.9% of subjects were diagnosed with CRC, including 0.6%, 1.5% and 2.7% of those aged
Gut Online First, published on December 15, 2014 as 10.1136/gutjnl-2014-307729 GI cancer
ORIGINAL ARTICLE
Ten-year incidence of colorectal cancer following a negative screening sigmoidoscopy: an update from the Colorectal Cancer Prevention (CoCaP) programme V Paul Doria-Rose,1 Theodore R Levin,2,3 Albert Palitz,3 Carol Conell,2 Noel S Weiss4,5 ▸ Additional material is published online only. To view please visit the journal online (http://dx.doi.org/10.1136/ gutjnl-2014-307729). For numbered affiliations see end of article. Correspondence to Dr V Paul Doria-Rose, Health Services and Economics Branch, Applied Research Program, Division of Cancer Control and Population Sciences, National Cancer Institute, 9609 Medical Center Drive, Room 3E438, Bethesda, MD 20892-9762, USA; [email protected] Received 24 May 2014 Revised 15 November 2014 Accepted 20 November 2014
ABSTRACT Objective To examine the rates of colorectal cancer (CRC) following a negative screening sigmoidoscopy. Design Cohort study. Setting An integrated healthcare delivery organisation in California, USA. Participants 72 483 men and women aged 50 years and above who had a negative screening sigmoidoscopy between 1994 and 1996. Those at elevated risk of CRC due to inflammatory bowel disease, prior polyps or CRC, or a strong family history of CRC were excluded. Main outcome measures Incidence rates of distal and proximal CRC. Standardised Incidence Ratios were used to compare annual incidence rates of distal and proximal CRC in the cohort to expected rates based on Surveillance, Epidemiology, and End Results data. Additionally, rate ratios (RR) and rate differences (RD) comparing the incidence rate of distal CRC in years 6+ postscreening with that in years 1–5 were calculated. Results Incidence rates of distal CRC were lower than those in the San Francisco Bay area population at large during each of the first 10 years postsigmoidoscopy screening. However, the incidence of distal CRC rose steadily, from 3 per 100 000 in the first year of follow-up to 40 per 100 000 in the 10th year. During the second half of follow-up, the rate of distal CRC was twice as high as in the first half (RR 2 .08, 95% CI 1.38 to 3.16; RD 14 per 100 000 person-years, 95% CI 6 to 22). Conclusions Though still below population levels, the incidence of CRC during years 6–10 following a negative sigmoiodoscopy is appreciably higher than during the first 5 years.
INTRODUCTION
To cite: Doria-Rose VP, Levin TR, Palitz A, et al. Gut Published Online First: [please include Day Month Year] doi:10.1136/gutjnl2014-307729
Randomised controlled trials have demonstrated the ability of flexible sigmoidoscopy screening to prevent colorectal cancer (CRC) incidence and mortality.1–5 While these trials provide strong evidence of screening effectiveness, the most appropriate interval for screening is less clear-cut. Current US screening guidelines for average-risk individuals recommend either screening colonoscopy every 10 years or screening sigmoidoscopy every 5 years.6 7 Several studies have addressed the question of the appropriate interval for endoscopic screening by examining patterns of CRC incidence or mortality following a baseline sigmoidoscopy,8–11 colonoscopy12–16 or unspecified lower endoscopy.17–19
Significance of this study What is already known on this subject?
▸ Several prior studies have examined patterns in the incidence of colorectal cancer (CRC) following sigmoidoscopy or colonoscopy. ▸ However, the absolute risk of cancer following a negative examination has been estimated in few large cohorts of patients receiving screening endoscopies.
What are the new findings?
▸ The incidence of distal CRC remains low for at least a decade following a negative screening sigmoidoscopy. ▸ However, the relative risk of distal cancer doubles in the second 5-year period following screening as compared with the first 5-year period.
How might it impact on clinical practice in the foreseeable future?
▸ These results will allow clinicians to provide estimates of the absolute risk of CRC following a negative screening sigmoidoscopy, and will inform the choice of appropriate screening interval.
While the low incidence of CRC seems to be longlasting (ie, 10 or more years) following endoscopy, there are several limitations of existing studies, such as the inclusion of non-screening endoscopies,11–16 18 19 which may define a group at a greater underlying risk of CRC, the use of patient self-report to determine the occurrence and timing of endoscopy,10 17 and small sample size.9 In previous work, we used data from Kaiser Permanente Northern California (KPNC) to define a cohort of over 70 000 individuals with a negative screening sigmoidoscopy, and observed a low incidence of CRC during the first 5 years postscreening.20 To address the potential impact of a 10-year versus a 5-year screening interval, we now update these findings to include additional data over a total of 10.5 years of follow-up. We examined whether a large increase in the risk of CRC occurred during the second 5-year period postsigmoidoscopy.
Doria-Rose VP, et al. Gut 2014;0:1–7. doi:10.1136/gutjnl-2014-307729
1
Copyright Article author (or their employer) 2014. Produced by BMJ Publishing Group Ltd (& BSG) under licence.
Downloaded from http://gut.bmj.com/ on November 14, 2015 - Published by group.bmj.com
GI cancer METHODS Study population, data sources and follow-up KPNC is a large, integrated healthcare delivery system in California, serving more than 3.3 million members from Santa Rosa to the north, Sacramento to the east and Fresno to the south. In 1994, KPNC initiated the Colorectal Cancer Prevention (CoCaP) programme, which offered a screening flexible sigmoidoscopy at least once every 10 years to all plan members aged 50 years or older. During this time, sigmoidoscopy screening was vigorously promoted within the KPNC health system through active inreach to patients at the time of an office visit for other reasons. Sigmoidoscopy exams were performed almost exclusively without sedation; bowel preparation was generally two enemas performed prior to the procedure. For the first 3 years of the CoCaP programme, sigmoidoscopy report data from all KPNC facilities were entered into a computerised database. These data included several items based on patient self-report: family history of CRC, history of prior sigmoidoscopy and colonoscopy, and personal history of CRC and colorectal polyps. Additionally, endoscopists reported the depth of insertion of the sigmoidoscope, whether the exam was limited by the presence of stool, and the number and size of any polyps identified. Both the patient and the endoscopist reported whether the sigmoidoscopy was performed due to symptoms or for screening. Other sources of data included KPNC membership databases, which provided information on patient demographics (age, sex and race/ethnicity) and enrolment dates, and various inpatient and outpatient databases which documented dates and types of medical procedures and diagnoses occurring during the follow-up period. Finally, the KPNC Tumour Registry identified all cases of CRC, collecting information on date of diagnosis, location within the colorectum, and stage. Using the CoCaP database, we previously defined a cohort of 72 483 KPNC members who had a screening sigmoidoscopy negative for both polyps and cancer between 1994 and 1996; approximately 80% of the total number of those screened were included based on their negative exam.21 Cohort inclusion and exclusion criteria, and cancer incidence during the first 5 years after negative sigmoidoscopy have previously been described.20 Briefly, cohort members were men and women aged 50 years and older, who were considered to be at average risk of developing CRC. Patients were excluded from the cohort if either the patient or the endoscopist indicated that the sigmoidoscopy was not for screening, if the endoscopist noted that the examination was being performed due to inflammatory bowel disease or if the patient reported a personal history of colorectal polyps/ cancer and/or a high-risk family history of CRC (defined as one first-degree relative diagnosed with CRC before age 55 years or CRC in multiple first-degree relatives). Patient follow-up concluded at the time of diagnosis of distal (rectum and sigmoid colon) or proximal (other colon sites excluding the appendix) CRC, death, health plan membership termination or subsequent colorectal procedures including sigmoidoscopy, colonoscopy or barium enema. Sigmoidoscopy resulted in censoring only in the analysis of distal CRC incidence, since the occurrence of sigmoidoscopy would have negligible impact on the subsequent incidence of proximal cancer. Colorectal examinations resulted in censoring because the goal of this type of analysis is to determine the amount of time required after a negative screen for CRC to be diagnosed, and subsequent exams could alter this time course due to additional interventions (ie, polypectomy). However, procedures occurring within 6 months prior to CRC diagnosis were considered to be part of the diagnosis process 2
and did not result in censoring. This updated analysis includes all CRC diagnoses and censoring events through 10.5 years of follow-up.
Statistical analysis Descriptive statistics were examined for demographic and sigmoidoscopy exam characteristics of cohort members. χ2 Tests were used to determine whether the stage distribution of colorectal adenocarcinomas differed according to cancer site (distal or proximal) or time since the negative screen (years 1–5 vs 6+). The overall proportion of subjects diagnosed with CRC during follow-up was estimated using the Kaplan–Meier product limit estimator. To examine the temporal pattern of CRC incidence following a negative screening sigmoidoscopy, annual rates of distal and proximal colorectal adenocarcinomas were calculated by dividing the number of colorectal adenocarcinoma cases by the total person-years at risk for each year postscreening. This analysis uses a broader definition of adenocarcinoma than was used in the 5-year follow-up of this cohort (see online supplementary appendix 1).20 Further, updates to the KPNC Tumour Registry after the data were extracted for the 5-year follow-up also contributed to the net addition of more cases. As a result, the incidence rates we report here over the first 5 years of follow-up are higher than those previously reported.20 However, the general pattern of incidence (very low initially, increasing gradually over 5 years) was consistent regardless of adenocarcinoma definition. The duration of low CRC risk following a negative sigmoidoscopy was estimated in two separate ways using both external and internal comparisons. First, standardised incidence ratios (SIR) and 95% CIs22 were used to compare the observed annual incidence rates of distal and proximal colorectal adenocarcinomas with expected rates based on incidence in the San Francisco Bay area as a whole, as reported to the Surveillance, Epidemiology, and End Results (SEER) Tumour Registry for the years 1994–2006.23 Standardisation was performed on the basis of sex, age (in 5-year categories through age 84 years and 85 years and over), and calendar year. Second, we compared incidence rates of distal colorectal adenocarcinomas over all person-time accruing during the second half of follow-up (years 6 plus) to those in the first 5 years of follow-up; rate ratios (RR), rate differences (RD) and their associated 95% CI were calculated,22 providing an estimate of the amount of disease that could be prevented with a 5-year screening interval as compared with a 10-year screening interval. These RRs and RDs were calculated without adjusting for age, since examining patterns of cancer incidence to help inform the appropriate timing for rescreening should be based on the unadjusted, absolute risk of disease; for this purpose, it is not relevant whether any increase in risk over time is due to ageing of the cohort, increased time elapsed since the negative screening exam or a combination of the two. RRs and RDs were calculated both overall and for subgroups based on demographic and sigmoidoscopy exam characteristics. No subjects had missing data for any of the key variables used for the calculation of crude and adjusted CRC incidence rates; for subgroup analyses, when a substantial proportion of the data for any variable were missing (race/ethnicity, endoscopist specialty), results are reported separately for the missing/unknown category. Patterns of colorectal procedures were also examined. One concern for this analysis was that, as cohort members returned for rescreening later in follow-up, there would be an artefactual increase in cancer incidence reflecting prevalent cases diagnosed (earlier than they otherwise would have been) at rescreening. Doria-Rose VP, et al. Gut 2014;0:1–7. doi:10.1136/gutjnl-2014-307729
Downloaded from http://gut.bmj.com/ on November 14, 2015 - Published by group.bmj.com
GI cancer This was a particular concern at approximately 10 years postscreening, since KPNC recommends a 10-year screening interval for sigmoidoscopy. Though the indication for the initial negative sigmoidoscopy was available in the CoCaP database, indication for any subsequent censoring procedures was not. However, KPNC maintains a clinical database of all International Classification of Diseases (ICD)-9 diagnoses/V codes for all inpatient and outpatient visits that cover the entire study period. This enabled us to classify censoring procedures as ‘diagnostic’ or ‘not diagnostic’, based on whether there were symptoms that could lead to endoscopy/barium enema and V codes indicating screening (see online supplementary appendix 2). Any relevant symptom codes in the 6 months prior to the procedure (including day of exam) resulted in the classification as diagnostic, unless a V code for screening was also present during that time. We considered any procedures within the 6 months before or after the anniversary of the initial screening sigmoidoscopy to be conducted at approximately that anniversary date; for example, all exams occurring from 9.5 through 10.5 years of follow-up were classified as 10-year exams.
RESULTS Characteristics of the study cohort are shown in table 1. This analysis is based on 584 106 person-years of follow-up (mean 8.1 per person) that accrued during the 10.5-year follow-up period. During this time, the experience of 45 764 individuals was censored for reasons other than development of CRC or end of the follow-up period; reasons for censoring are shown in table 2. Over the entire follow-up period, 103 cases of distal and 333 cases of proximal colorectal adenocarcinomas were diagnosed. Approximately three-quarters of cancers were proximal in both the first and second halves of follow-up (table 2). For both distal and proximal cancers, stage distribution was also very similar in 1–5 versus 6+ years. However, distal cancers were more likely than proximal cancers to be diagnosed at an earlier stage (42% vs 31% localised, p=0.01) (table 3). During follow-up, 0.9% of subjects were diagnosed with CRC, including 0.6%, 1.5% and 2.7% of those aged
