|Home | About | Journals | Submit | Contact Us | Français|
Colonoscopy and sigmoidoscopy provide protection against colorectal cancer, but the magnitude and duration of protection, particularly against cancer of the proximal colon, remain uncertain.
We examined the association of the use of lower endoscopy (updated biennially from 1988 through 2008) with colorectal-cancer incidence (through June 2010) and colorectal-cancer mortality (through June 2012) among participants in the Nurses’ Health Study and the Health Professionals Follow-up Study.
Among 88,902 participants followed over a period of 22 years, we documented 1815 incident colorectal cancers and 474 deaths from colorectal cancer. With endoscopy as compared with no endoscopy, multivariate hazard ratios for colorectal cancer were 0.57 (95% confidence interval [CI], 0.45 to 0.72) after polypectomy, 0.60 (95% CI, 0.53 to 0.68) after negative sigmoidoscopy, and 0.44 (95% CI, 0.38 to 0.52) after negative colonoscopy. Negative colonoscopy was associated with a reduced incidence of proximal colon cancer (multivariate hazard ratio, 0.73; 95% CI, 0.57 to 0.92). Multivariate hazard ratios for death from colorectal cancer were 0.59 (95% CI, 0.45 to 0.76) after screening sigmoidoscopy and 0.32 (95% CI, 0.24 to 0.45) after screening colonoscopy. Reduced mortality from proximal colon cancer was observed after screening colonoscopy (multivariate hazard ratio, 0.47; 95% CI, 0.29 to 0.76) but not after sigmoidoscopy. As compared with colorectal cancers diagnosed in patients more than 5 years after colonoscopy or without any prior endoscopy, those diagnosed in patients within 5 years after colonoscopy were more likely to be characterized by the CpG island methylator phenotype (CIMP) (multivariate odds ratio, 2.19; 95% CI, 1.14 to 4.21) and microsatellite instability (multivariate odds ratio, 2.10; 95% CI, 1.10 to 4.02).
Colonoscopy and sigmoidoscopy were associated with a reduced incidence of cancer of the distal colorectum; colonoscopy was also associated with a modest reduction in the incidence of proximal colon cancer. Screening colonoscopy and sigmoidoscopy were associated with reduced colorectal-cancer mortality; only colonoscopy was associated with reduced mortality from proximal colon cancer. Colorectal cancer diagnosed within 5 years after colonoscopy was more likely than cancer diagnosed after that period or without prior endoscopy to have CIMP and microsatellite instability. (Funded by the National Institutes of Health and others.)
Randomized, controlled trials have shown that screening with flexible sigmoidoscopy reduces the incidence of colorectal cancer and associated mortality, albeit with diminished effectiveness for cancers of the proximal colon.1–3 Although comparable data from randomized, controlled trials of screening colonoscopy are not yet available,4 colonoscopy is also widely endorsed by expert bodies for population-based screening, largely on the basis of case–control studies that show associations with reduced colorectal-cancer incidence and mortality.5–9 However, as with flexible sigmoidoscopy, there is uncertainty about the effectiveness of colonoscopy in reducing the incidence of and mortality associated with proximal colon cancer10–19 and about the frequency and interval at which testing should be offered.5–9 Moreover, it remains unclear why a considerable proportion of colorectal cancers are diagnosed in persons who have recently undergone colonoscopy.5 Such cancers may result from missed lesions or from the rapid progression of new neoplasia,20–25 which may be associated with specific molecular characteristics.25
To address these uncertainties, we conducted a prospective analysis of the association between lower gastrointestinal endoscopy and the long-term risk of incident colorectal cancer in two large U.S. cohorts prospectively followed over a period of 22 years. We also comprehensively examined the molecular features in a subset of tumors.
We used data from two prospective cohort studies: the Nurses’ Health Study, which included 121,700 U.S. female nurses, 30 to 55 years of age at enrollment in 1976; and the Health Professionals Follow-up Study, which included 51,529 U.S. male health professionals, 40 to 75 years of age at enrollment in 1986.26,27 The return of mailed questionnaires was considered to constitute written informed consent.
The study protocol was approved by the institutional review boards of the Harvard School of Public Health and Brigham and Women’s Hospital. The authors assume full responsibility for the analyses and interpretation of these data.
Details of the endoscopy assessment are provided in the Supplementary Appendix, available with the full text of this article at NEJM.org. In both cohorts, beginning in 1988 and continuing through 2008, as part of a questionnaire administered every 2 years, participants were asked whether they had undergone either sigmoidoscopy or colonoscopy and, if so, the reason for the investigation. In 2004, we additionally collected comprehensive information on whether previously reported lower endoscopies were colonoscopies or sigmoidoscopies.26,27 Every cycle thereafter, responses for sigmoidoscopy and colonoscopy were recorded separately.
When participants reported a diagnosis of colorectal polyps, consent was obtained to review medical records and pathology reports.26,27 Study physicians, who were unaware of all the data obtained from the questionnaires, confirmed adenomatous polyps. Persons with polyps that met one or more of the criteria for advanced adenoma (≥10 mm in diameter, tubulovillous or villous histologic features, or high-grade dysplasia) and persons with three or more adenomatous polyps were classified as having high-risk adenoma.5 Colonoscopic polypectomy was defined as the excision of one or more confirmed adenomatous polyps, excluding hyperplastic polyps. A negative endoscopy was defined as a procedure that did not result in the diagnosis of adenomas or colorectal cancer.
Detailed descriptions of cancer ascertainment and molecular analyses are provided in the Supplementary Appendix. A diagnosis of colorectal cancer was confirmed with the use of the National Death Index, medical records, and pathology reports. We extracted DNA from paraffin-embedded tumor specimens and normal tissue specimens. Microsatellite instability status and mutation status for BRAF (codon 600), KRAS (codons 12 and 13), and PIK3CA (exons 9 and 20) were determined as previously described.28,29 DNA methylation was quantified at eight CpG island methylator phenotype (CIMP)–specific promoters (CACNA1G, CDKN2A [p16], CRABP1, IGF2, MLH1, NEUROG1, RUNX3, and SOCS1) and in long interspersed nucleotide element 1 (LINE-1), with the use of the MethyLight technique or pyrose-quencing.28,30
A detailed description of the statistical analysis is provided in the Supplementary Appendix. We followed participants from the month of return of the 1988 baseline questionnaire through June 2010 for the incidence analysis and through June 2012 for the mortality analysis. We excluded participants with a baseline history of cancer (except for nonmelanoma skin cancer), ulcerative colitis, colorectal polyps, familial polyposis syndromes, or previous lower endoscopy (Fig. S1 in the Supplementary Appendix). We used Cox proportional-hazards models to calculate hazard ratios and 95% confidence intervals. All analyses were stratified according to age (in months), sex (in the combined cohort analysis), and calendar year of the questionnaire cycle. Multivariate models were adjusted for known or suspected risk factors for colorectal cancer, listed in Table 1.
For the incidence analysis, to minimize the influence of endoscopies performed for the diagnostic evaluation of colorectal cancer, we examined the association of endoscopy status reported on the biennial questionnaire before the diagnosis of colorectal cancer, death from any cause, or the end of follow-up, whichever came first. We used the most recently updated information for all variables before each 2-year follow-up and treated all variables as time-varying to account for changes during follow-up. For the mortality analysis, we evaluated the association of screening sigmoidoscopy or screening colonoscopy with mortality on the basis of the endoscopy status reported up to and including the date of diagnosis of colorectal cancer, death from any cause, or the last follow-up cycle, whichever came first.
We calculated the population-attributable risk, estimated as the proportion of incident colorectal cancers that would have been prevented in our population if all participants had undergone colonoscopy (with negative results or polypectomy) at least once and risk factors had not changed.31 We also conducted a case–case analysis using a logistic-regression model to examine whether specific molecular features were associated with cancer occurring within 5 years after colonoscopy. All statistical analyses were two-sided, and a P value of less than 0.05 was considered to indicate statistical significance.
Among 88,902 participants (31,736 men and 57,166 women), we documented a total of 1815 incident cases of colorectal cancer (in 714 men and 1101 women) during 22 years of follow-up, encompassing a total of 1,738,396 person-years. Age-adjusted demographic characteristics at the midpoint of follow-up (1998), according to endoscopy status, are described in Table 1.
In the combined cohorts, the multivariate hazard ratios for colorectal cancer among participants who had undergone endoscopy, as compared with those who had not, were 0.57 (95% confidence interval [CI], 0.45 to 0.72) after removal of adenomatous polyps, 0.60 (95% CI, 0.53 to 0.68) after negative sigmoidoscopy, and 0.44 (95% CI, 0.38 to 0.52) after negative colonoscopy (Table 2). These associations were consistent among men and women and were evident for all disease stages at presentation. A reduced incidence of distal colorectal cancer was observed with polypectomy (multivariate hazard ratio, 0.40; 95% CI, 0.27 to 0.59), negative sigmoidoscopy (multivariate hazard ratio, 0.44; 95% CI, 0.36 to 0.53), and negative colonoscopy (multivariate hazard ratio, 0.24; 95% CI, 0.18 to 0.32). However, only negative colonoscopy was associated with a significantly reduced risk of proximal colon cancer (multivariate hazard ratio, 0.73; 95% CI, 0.57 to 0.92).
In analyses restricted to endoscopy for screening, the results were similar to those obtained in our analyses of endoscopy for any indication (Table S1 in the Supplementary Appendix). In addition, we observed consistent results in the analysis that used propensity-score adjustment and in the subanalyses excluding cases of colorectal cancer diagnosed within 2 years after a previously reported initial endoscopy and excluding those for which the participant or medical record indicated that the diagnosis had been made at the initial screening endoscopy (Table S2 in the Supplementary Appendix). We estimated that the population-attributable risk of colorectal cancer (the proportion of incident cancers that would have been prevented with colonoscopy) was 40% (95% CI, 32 to 46) for all colorectal cancers, 22% (95% CI, 10 to 34) for proximal colon cancers, and 61% (95% CI, 52 to 69) for distal colorectal cancers.
To gain insight into the recommended screening interval for low-risk persons, we evaluated colorectal-cancer incidence according to the time since the last negative colonoscopy (Table 3). The multivariate hazard ratios for colorectal cancer were 0.35 (95% CI, 0.28 to 0.45) for an interval of 3.0 years or less after a negative colonoscopy as compared with no endoscopy, 0.40 (95% CI, 0.31 to 0.52) for 3.1 to 5.0 years, 0.52 (95% CI, 0.38 to 0.70) for 5.1 to 10.0 years, and 0.26 (95% CI, 0.12 to 0.59) for 10.1 to 15.0 years. In addition, reduced risks were observed up to 15.0 years after the last negative colonoscopy for both proximal colon cancer (multivariate hazard ratio for 5.1 to 15.0 years, 0.60; 95% CI, 0.38 to 0.94) and distal colorectal cancer (multivariate hazard ratio for 5.1 to 15.0 years, 0.35; 95% CI, 0.22 to 0.54).
Among participants who had undergone endoscopy with removal of adenomatous polyps, as compared with those who had not undergone endoscopy, a lower incidence of colorectal cancer was observed with a surveillance interval of 3.0 years or less (multivariate hazard ratio, 0.48; 95% CI, 0.33 to 0.69) and with an interval of 3.1 to 5.0 years (multivariate hazard ratio, 0.49; 95% CI, 0.33 to 0.73) (Table S3 in the Supplementary Appendix). Similar risks across time intervals were observed among participants with a history of adenoma in the proximal colon or distal colorectum. For participants with high-risk adenoma, the association was attenuated and of shorter duration, with a multivariate hazard ratio of 0.70 (95% CI, 0.43 to 1.14) for colonoscopy performed within 3.1 to 5.0 years after the last colonoscopy.
The inverse association of colonoscopy with colorectal cancer appeared to be similar across subgroups defined according to age, body-mass index, smoking status, and status with respect to regular use of aspirin (Table S4 in the Supplementary Appendix). Among participants with a family history of colorectal cancer, a significant association was no longer observed beyond 5 years after colonoscopy (multivariate hazard ratio, 0.91; 95% CI, 0.55 to 1.52). By contrast, there was a sustained association beyond 5 years among persons without a family history of colorectal cancer (multivariate hazard ratio, 0.43; 95% CI, 0.32 to 0.58) (P = 0.04 for interaction).
We considered only negative colonoscopies that occurred at least 4 years apart to account for repeat examinations performed within a shorter interval owing to inadequate bowel preparation. As compared with no endoscopy, the multivariate hazard ratios for colorectal cancer were 0.43 (95% CI, 0.35 to 0.51) after one negative colonoscopy, 0.32 (95% CI, 0.22 to 0.48) after two negative colonoscopies, and 0.23 (95% CI, 0.08 to 0.67) after three negative colonoscopies (Table S5 in the Supplementary Appendix).
We identified 62 cancers diagnosed within 5 years after colonoscopy for which molecular data were available (Table S6 in the Supplementary Appendix). As compared with cancers diagnosed in patients more than 5 years after colonoscopy or without any prior endoscopy, those diagnosed in patients within 5 years after colonoscopy were more likely to be characterized by CIMP (multivariate odds ratio, 2.19; 95% CI, 1.14 to 4.21), microsatellite instability (multivariate odds ratio, 2.10; 95% CI, 1.10 to 4.02), and an increased LINE-1 methylation level (multivariate odds ratio for each 30% increment, 3.21; 95% CI, 1.29 to 8.00). BRAF, KRAS, and PIK3CA mutations were not significantly associated with cancer diagnosed within 5 years after colonoscopy.
During follow-up, we identified a total of 474 deaths attributable to colorectal cancer. We observed lower mortality from colorectal cancer among participants who had undergone screening sigmoidoscopy (multivariate hazard ratio, 0.59; 95% CI, 0.45 to 0.76) and among those who had undergone screening colonoscopy (multivariate hazard ratio, 0.32; 95% CI, 0.24 to 0.45) than among those who had never undergone screening endoscopy (Table 4). Screening colonoscopy was associated with reduced mortality from both distal colorectal cancer (multivariate hazard ratio, 0.18; 95% CI, 0.10 to 0.31) and proximal colon cancer (multivariate hazard ratio, 0.47; 95% CI, 0.29 to 0.76), whereas screening sigmoidoscopy was associated only with reduced mortality from distal colorectal cancer (multivariate hazard ratio, 0.31; 95% CI, 0.20 to 0.49).
In two large, U.S. prospective cohort studies, we found that the long-term incidence of colorectal cancer was lower among men and women who had a history of negative sigmoidoscopy, negative colonoscopy, or polypectomy for adenoma than among those who had no history of endoscopy. Negative colonoscopy was associated with a lower incidence of both distal colorectal cancer and proximal colon cancer, whereas negative sigmoidoscopy and colonoscopy with polypectomy were associated primarily with a lower incidence of distal colorectal cancer. We estimated that 40% of colorectal cancers (including 61% of distal colorectal cancers and 22% of proximal colon cancers) that developed during follow-up would have been prevented if all the participants in our study had undergone colonoscopy. Moreover, screening sigmoidoscopy and screening colonoscopy were associated with lower mortality from colorectal cancer, as compared with no endoscopy, although only screening colonoscopy was associated with lower mortality from proximal colon cancer.
Previous randomized, controlled trials have had inconsistent findings regarding the influence of sigmoidoscopy on the incidence of proximal colon cancer,1–3,32 probably owing to differences in subsequent exposure to colonoscopy. In the U.K. Flexible Sigmoidoscopy Screening Trial, no reduction in the incidence of proximal cancer was detected; however, only 5% of participants underwent follow-up colonoscopy on the basis of sigmoidoscopic findings.2 By contrast, the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial showed a 14% reduction in the incidence of proximal colon cancer, potentially owing to the 21.9% of participants who underwent colonoscopy for follow-up of sigmoidoscopic findings or outside the study protocol.1 As is consistent with the findings in previous randomized, controlled trials,2,3 our results suggest that screening sigmoidoscopy alone is probably insufficient for reducing the incidence of proximal colon cancer and associated mortality.
Our results are consistent with the findings of the National Polyp Study, which showed a lower incidence of colorectal cancer among persons after colonoscopic polypectomy, as compared with population-based estimates of expected rates.16,33 Our study expands on these results, since we were able to directly compare actual incidences of cancer among persons after polypectomy with the incidences among persons from the same background population who did not undergo endoscopy, while adjusting for potential confounders. We did not observe a significantly reduced incidence of proximal colon cancer in association with polypectomy. This result might be due, in part, to limited statistical power. Alternatively, the presence of an adenoma may be a marker of an increased risk of subsequent proximal colon cancer that is not completely mitigated by polypectomy. A recent case–control study also showed a smaller reduction in the incidence of proximal colon cancer, as compared with distal colorectal cancer, after polyp-ectomy.14
In our analysis, negative colonoscopy was associated with a significantly reduced incidence of distal colorectal cancer or proximal colon cancer up to 15 years after the procedure. Previous estimates of the duration of protection associated with a negative colonoscopy have varied widely, ranging from 5 to 20 years.11–13,34 These inconsistent results may be due to relatively short follow-up12,13,34 or the limitations of a case–control design,11 including biases related to selection of controls. Our findings support the 10-year examination interval recommended by existing guidelines for persons at average risk who have a negative colonoscopy.5–9 Our study suggests that even a single negative colonoscopy is associated with a very low long-term risk of colorectal cancer.2,3 However, our data support screening at more frequent intervals for persons with a family history of colorectal cancer.
Among participants with a history of adenoma, we observed a reduced incidence of cancer up to 5 years after colonoscopy, which supports current surveillance guidelines.5,6 However, we found that the apparent reduction in risk was attenuated among participants with high-risk adenomas, a finding that is consistent with the results of other studies.15 This observation may reflect a persistently elevated incidence of cancer associated with predisposing host or lifestyle risk factors, the biologic characteristics of high-risk adenomas, or the uncertain quality of colonoscopic detection and clearance of neoplasia in persons with high-risk lesions.20–25
Our finding that cancer diagnosed within 5 years after colonoscopy was associated with specific molecular features (CIMP, microsatellite instability, and high-level LINE-1 methylation) complements the existing literature.20–25,35 Serrated lesions, particularly sessile serrated adenomas, are widely considered to be probable precursors of colorectal cancers characterized by CIMP, and these lesions may be particularly difficult to detect endoscopically or remove adequately.36–38 It remains unclear whether any of the challenges posed by these biologic differences can be addressed by improvements in colonoscopic technique, including more meticulous inspection or improved bowel cleansing.
Our study has several strengths. First, because we collected information biennially for a period of 22 years, we were able to update endoscopy status in order to accurately assess associations with the subsequent risk of colorectal cancer or death. Second, our detailed exposure information, including lifestyle factors, enabled us to finely adjust for potential confounders. Third, our prospective design minimized biases inherent in case–control studies, including recall and selection biases. Fourth, we were able to directly compare the incidence of colorectal cancer and mortality associated with colorectal cancer among persons who underwent endoscopy with the incidence and mortality among persons from the same background population who did not undergo endoscopy. By contrast, previous cohort studies have used comparisons with population-based estimates.12,16,33 Fifth, since all study participants were health care professionals, the accuracy of our classification according to endoscopy status was high. Finally, our comprehensive molecular profiling of tumors allowed us to elucidate molecular features of cancer occurring within 5 years after colonoscopy, adjusting for other potential confounding factors.
There are limitations to our study. As with all observational studies, we cannot rule out unmeasured confounding, including potential bias introduced by the pooling of data from two separate cohorts. Second, our participants were health care professionals, and our findings may not be generalizable to other populations. However, previous studies have shown that the prevalences of risk factors for colorectal cancer, including smoking and body-mass index, among our participants are consistent with those of the broader population,39,40 and the incidence and stage distribution of colorectal cancers in our cohorts are similar to those in other population-based registries. Moreover, there is little evidence to suggest that the putative mechanisms by which endoscopy is associated with a reduced incidence of colorectal cancer would differ according to occupation or educational background.
In conclusion, as compared with no endoscopy, colonoscopy and sigmoidoscopy were associated with a lower incidence of distal colorectal cancer, whereas only colonoscopy was associated with a reduced incidence of proximal colon cancer, and that reduction was modest. As compared with no screening endoscopy, screening colonoscopy and sigmoidoscopy were associated with lower mortality from colorectal cancer, whereas only colonoscopy was associated with lower mortality from proximal colon cancer. Tumor molecular features of the serrated pathway might be involved in the development of cancer within 5 years after colonoscopy.
Supported by grants from the NIH (P01 CA87969; P01 CA55075 and 1UM1 CA167552, to Dr. Willett; P50 CA127003, to Dr. Fuchs; R01 CA151993, to Dr. Ogino; R01 CA137178, to Dr. Chan; and K24 DK098311, to Dr. Chan) and from the Bennett Family Foundation and the Entertainment Industry Foundation through the National Colorectal Cancer Research Alliance. Dr. Chan is a Damon Runyon Clinical Investigator, Dr. Lochhead is a Scottish Government Clinical Academic Fellow and was supported by a Harvard University Frank Knox Memorial Fellowship, Dr. Inamura was supported by a Japan Society for the Promotion of Science Postdoctoral Fellowship for Research Abroad, and Dr. Kim was supported by an Early Exchange Postdoctoral Fellowship Grant from Asan Medical Center, Seoul, Korea.
We thank the participants and staff of the Nurses’ Health Study and the Health Professionals Follow-up Study for their valuable contributions; and Scott G. Smith of the Department of Nutrition, Harvard School of Public Health, for assistance with data management.
The views expressed in this article are those of the authors and do not necessarily represent the official views of the National Institutes of Health (NIH).
Dr. Chan reports receiving consulting fees from Bayer Health-care, Pfizer, Millennium Pharmaceuticals, and Pozen. No other potential conflict of interest relevant to this article was reported.
Disclosure forms provided by the authors are available with the full text of this article at NEJM.org.