Colorectal Cancer Risk Following Adenoma Removal: A Large Prospective Population-Based Cohort Study

In the United Kingdom, colorectal cancer is the fourth most common incident cancer and second most common cause of cancer-related death (1). The United Kingdom has inferior colorectal cancer survival rates to comparable countries, commonly attributed to later clinical presentation (2, 3). Colorectal cancer typically originates from precancerous colorectal polyps, providing an opportunity for prevention if detected early and successfully excised.

The vast majority of colorectal polyps are either adenomas or hyperplastic polyps (4). Adenomas are dysplastic polyps, which can progress via the adenoma–carcinoma sequence to invasive cancer. Hyperplastic polyps are classified within the serrated group of colorectal polyps, and are traditionally considered benign. A related subtype of serrated polyps, namely sessile serrated polyps (also known as sessile serrated adenomas), may progress to carcinoma via the serrated carcinogenesis pathway, (5, 6) but these are relatively rare (7). Individuals who have only hyperplastic polyps removed at endoscopy are not typically entered into surveillance (8). Individuals with adenomas detected will usually be entered into a surveillance regime at an interval of 1, 3, or 5 years primarily depending on the number and size of adenomas detected or genetic predisposition to colorectal cancer (8, 9).

The ultimate aim of clearing the bowel of adenomas, and entering patients into surveillance, is to reduce colorectal cancer risk. Building upon the historic findings of Winawer and colleagues (10), several recent randomized controlled trials (11–14) and population-based case–control studies (15, 16) have demonstrated the success of polypectomy in reducing colorectal cancer risk. The U.S. Prostate, Lung, Colorectal, and Ovarian cancer screening trial has reported a significant 21% reduced risk of colorectal cancer among the intervention arm who underwent flexible sigmoidoscopy at baseline plus 3 or 5 years later, compared with usual care (11). A ‘one-off’ flexible sigmoidoscopy was also effective at reducing colorectal cancer risk by 23% in a UK trial (12), whereas similar reductions were observed in Italian, but not Norwegian, trials (13, 14).

However, little is known about whether polypectomy is effective at reducing colorectal cancer risk in routine clinical practice. The aforementioned trials (11–14) and case–control studies (15, 16) may incur selection bias due to modest response rates limiting the likelihood of participants being representative of the general population. Few studies have investigated colorectal cancer risk following polypectomy in the general population. A French study observed a 2.2-fold greater incidence of colorectal cancer among advanced adenoma patients compared with the general population (17), whereas a large Dutch study concluded that excess colorectal cancer risk in adenoma patients was limited to the first few years of follow-up (18). In addition, given the enhanced opportunity for polyp detection within bowel cancer screening programs (19, 20), the appropriate clinical management of patients following polypectomy is even more pertinent.

The aim of this investigation was to quantify colorectal cancer risk following polypectomy of adenomas in a large population-based study.

The Northern Ireland colorectal polyp register (NICPR) was derived from electronic pathology reports relating to all colorectal biopsies within Northern Ireland between 2000 and 2005. The NICPR has ethical approval from ORECNI: 10/NIR02/53. This timeframe precedes bowel cancer screening in this region, and therefore reflects clinical investigation of a symptomatic population. As outlined in Fig. 1, relevant SNOMED morphology codes for adenomas and hyperplastic polyps were extracted. Few polyps were diagnosed or coded as sessile serrated polyps during this time period, and therefore these were excluded from analysis. Individuals ages <16 years were also removed. We applied two strategies for identifying and excluding patients who may be genetically predisposed to colorectal cancer due to polyposis syndromes. First, patients were removed from the NICPR if they attended a genetic follow-up service at the Belfast Health and Social Care Trust (the sole referral center in Northern Ireland). Second, we conducted a medical note review as part of a nested case–control study (described in detail below) to identify participants with a personal or family history of colorectal polyps or polyposis syndromes.

Patients were categorized into an “adenoma register” if their incident polyp was recorded as an adenomatous polyp(s) with or without a concurrent hyperplastic polyp(s) diagnosis. Patients who only received a hyperplastic polyp diagnosis were grouped into a “hyperplastic polyp register.” SNOMED coding was used to identify tubular, tubulovillous, or villous adenoma subtypes. Information on polyp size was not readily available, and so patients with “advanced risk” adenomas were considered to be those with multiple adenomas of any histological type or at least one adenoma including a villous component, that is, villous or tubulovillous. In order to minimize misclassification of polyps detected but not removed at first investigation, all polyps detected within 6 months of the baseline polyp were reclassified as “incident” polyps. Subsequent polyps were those diagnosed at least 6 months after incident polyps, and up to December 31, 2005.

The Northern Ireland Cancer Registry collects information on all cancer diagnoses since 1993. All individuals diagnosed with colorectal cancer (ICD-10 codes C18-C20) up to December 31, 2010, were matched on the basis of name, date of birth, and/or address to the NICPR. Individuals with a previous or concurrent colorectal cancer diagnosis were excluded from the NICPR. Colorectal cancer cases diagnosed at least 6 months after their incident polyp were considered to be incident colorectal cancer, and were considered for further case note review as part of a nested case–control study, as detailed below. The NICPR was also matched to death records supplied by the Northern Ireland Registrar General's Office to the Northern Ireland Cancer Registry to detect patients who died by December 31, 2010.

Patients who developed colorectal cancer at least 6 months post-incident polyp diagnosis were included in a nested case–control study. Controls were matched 1:1 to colorectal cancer cases by age (±1 year), sex, year of incident polyp diagnosis, and were alive at the time of their matched cases' colorectal cancer diagnosis. Tumor verification officers reviewed hospital case notes for information on all colorectal cancer and polyp diagnoses, polyp characteristics (including size, number, and dysplasia grade), personal and family history of chronic bowel diseases, and all lower gastrointestinal investigations conducted (including endoscopies) from 1 year prior to incident polyp diagnosis up to end of follow-up. A small number of individuals who were found to have a personal history of polyposis syndromes (or a family history of colorectal polyps as a proxy for an undiagnosed polyposis syndrome) were excluded from further nested case–control study analysis due to the likelihood that they are genetically predisposed to colorectal cancer. We did not exclude individuals reporting a family history of colorectal cancer, because only a small proportion (5%–10%; 21) of these cases are likely to be due to genetic conditions and a known family history may simply reflect shared lifestyle or other nonhereditary colorectal cancer risk factors between family members. Cases for whom we were able to retrieve notes to enable inclusion within the case–control study did not differ from those not included with respect to age, sex, polyp location, morphology, or advanced status, and are therefore representative of the entire colorectal cancer case set.

Patients were followed up from the date of their incident polyp diagnosis (excluding the first 6 months of follow-up) until their date of colorectal cancer diagnosis, date of death or December 31, 2010. Descriptive characteristics were compared using independent t-tests and χ² tests for continuous and categorical variables, respectively. Individuals with less than 6 months follow-up were excluded from analysis. Colorectal cancer incidence was calculated per 100 person-years of follow-up, comparing observed incidence with that of expected incidence in the Northern Ireland population between 2000 and 2010. Cox proportional hazards models were applied to investigate the association between colorectal cancer risk and demographics or incident polyp characteristics. Stratified analysis was conducted by these variables to explore effect modification. Assumptions for Cox proportional hazards models were checked by visual inspection of Kaplan–Meier plots. Standardized colorectal cancer incidence ratios (SIR) and corresponding 95% confidence intervals (CI) were analyzed per 100,000 population, and separately for males and females. Sensitivity analysis was performed after excluding those with less than 1 year of follow-up. For the nested case–control study, colorectal cancer risk was assessed between comparative groups by applying conditional logistic regression analysis to generate odds ratios (OR) and 95% CI. Statistical analysis was conducted using Intercooled Stata version 11.0 (StataCorp).

Over a 6-year time period, n = 6,972 individuals had a least one adenoma removed at their index colonoscopic investigation (Fig. 1). There was a slight male predominance for adenoma patients, and the mean age at adenoma removal was 62.3 years. The majority of adenomas were detected and removed from the colon, and 22% of individuals had multiple adenomas removed at their first investigation in this time period (Table 1).

During a total of 44,724 person-years of follow-up, 193 patients were diagnosed with incident colorectal cancer. The rate of progression was relatively stable over time, and was consistently higher in males than females (Fig. 2). Over this period, 0.43% of adenoma patients developed colorectal cancer each year, as shown in Table 2. Colorectal cancer risk was significantly elevated in males and in individuals aged 60 years or older with adenomas. Rectal adenoma patients had an increased colorectal cancer risk (HR, 1.70; 95% CI, 1.27–2.28) compared with those who had only colonic adenomas excised. No colorectal cancer developed in patients who had adenomas removed from both the colon and rectum. Those with tubulovillous or villous adenomas were more likely to develop colorectal cancer compared with those with tubular adenomas (HR, 1.51; 95% CI, 1.02–2.23).

Several factors were associated with reduced risk of colorectal cancer among adenoma cases, including being diagnosed with concurrent hyperplastic polyps (HR, 0.45; 95% CI, 0.24–0.82) or subsequent polyps (Table 2). Individuals diagnosed with multiple incident adenomas also had a decreased risk of colorectal cancer (HR, 0.67; 95% CI, 0.45–0.97). Sensitivity analysis removing the first year of follow-up revealed largely similar results (Supplementary Table S1). As shown in Table 2, the risk of progression to colorectal cancer was lower in patients diagnosed with only hyperplastic polyps (0.17% per year), compared with adenoma patients. Colorectal cancer risk in hyperplastic polyp patients did not differ significantly by sex, polyp location, or multiplicity; however, individuals aged 60-<80 years were at significantly increased risk of colorectal cancer compared with younger patients (data not shown).

As outlined in Table 3, individuals diagnosed with adenomas were almost three times more likely to develop colorectal cancer than the general population (SIR, 2.85; 95% CI, 2.61–3.25). This heightened risk was observed in both males and females, for those with advanced adenomas, and after excluding cases in the first year of follow-up. Unexpectedly, hyperplastic polyp patients also had an increased colorectal cancer risk compared with the general population (SIR, 1.79; 95% CI, 1.45–2.33), which did not alter in sensitivity analysis excluding incident hyperplastic polyp patients who had a subsequent adenoma (SIR, 1.80; 95% CI, 1.46–2.36). Overall, an excess of 125 and 18 colorectal cancer cases developed among adenoma and hyperplastic polyp patients, respectively.

A post hoc nested case–control study was conducted amongst colorectal cancer cases arising in adenoma and hyperplastic polyp cases (n = 193, 82.5%) and matched controls whose medical notes were available for review, to determine further factors that may explain the excess colorectal cancer risk. A small number of colorectal cancer cases were identified to have a personal history of polyposis syndromes (n = 3) or family history of colorectal polyps (n = 7), or both (n = 1) and were excluded from further analysis since they are likely to be genetically predisposed to a greater colorectal cancer risk, leaving n = 148 case–control pairs in the remaining adenoma analysis and n = 34 pairs in the hyperplastic polyp analysis. As shown in Table 4, a small proportion of excess colorectal cancer risk could be attributed to concurrent inflammatory bowel disease (OR, 2.25; 95% CI, 0.98–5.17). Patients with large or right-sided incident polyps or adenomas with high-grade dysplasia had nonsignificantly increased risks of developing colorectal cancer.

Conversely, those with multiple adenomas, a colorectal cancer family history or an adenoma removed prior to the initiation of our study period (and therefore potentially undergoing surveillance) did not have an increased colorectal cancer risk. Adenoma patients who had a full colonoscopy during the follow-up period had a nonsignificant reduced risk of colorectal cancer, particularly if they had multiple (two or more) full colonoscopies performed (OR, 0.59; 95% CI, 0.32–1.09). If patients had attended for investigative procedures other than full colonoscopy (e.g., sigmoidoscopy or barium enema), they had a significant increased risk of colorectal cancer (OR, 2.18; 95% CI, 1.32–3.60). In addition, individuals who did not attend for a planned procedure were at an increased risk of colorectal cancer, which was exaggerated if multiple appointments were missed (Table 4).

Nonsignificant increased colorectal cancer risks were observed for individuals with multiple and/or right-sided hyperplastic polyps, and those who did not attend for a planned follow-up investigation (data not shown).

Overall, annual colorectal cancer risk among individuals following polypectomy of adenomas in the nonscreening setting is 0.43%. Despite undergoing polypectomy, colorectal cancer risk is significantly elevated for incident adenoma patients compared with the general population.

The increased risk of colorectal cancer following adenoma removal observed in our study is in line with findings from a French study (17). In their analysis of 5,779 adenoma patients diagnosed between 1990 and 1999, 87 colorectal cancer were diagnosed, resulting in an SIR of 2.23 (95% CI, 1.67–2.92) for advanced adenomas (17). As a result of further case note review, a small proportion of colorectal cancer cases in this study were found to be associated with genetic predisposition (6.9% overall, and 8.5% in advanced adenoma cases). Applying these proportions to our overall SIR estimates would result in SIR of 2.65 and 2.53 for all and advanced adenomas, respectively, and are more comparable with the methods in the French study (17). The largest study to date, published from a Dutch population, only observed an elevated SIR (2.8; 95% CI, 2.5–3.1) in the second and third year of follow-up, which was then attenuated in later years (18). This contrasts with our study findings, which illustrate a consistently increased colorectal cancer risk over a prolonged time period. Earlier population-based studies of smaller sample size, tended to observe either no association or a reduced risk of colorectal cancer following adenoma removal. (15, 16, 22) Randomized controlled trials of flexible sigmoidoscopy have also mostly demonstrated a reduction in colorectal cancer incidence (11, 12, 14). Our findings indicate that the benefits seen for polyp removal in the tightly controlled environment of trials and screened populations may not be replicated in the general population who are more likely to be symptomatic when presenting for clinical investigation.

As expected, being male or ages 60 years or older was associated with an increased risk of colorectal cancer following adenoma removal (23, 24). Individuals whose incident adenomas contained a villous component also had increased colorectal cancer risk, which is consistent with previous reports (23). Other factors that would be expected to increase colorectal cancer risk, including a personal history of inflammatory bowel disease, presence of high-grade dysplasia in the adenomas, and large adenoma size (9, 25) were directly associated with colorectal cancer risk in the detailed case note review aspect of our study. All of these findings are reassuring for the robustness of our data, but collectively, these factors only account for a relatively small proportion of the excess colorectal cancer cases found in this population.

Several factors indirectly suggested that incomplete clearance of adenomas from the bowel may have led to the higher future colorectal cancer risk, which is consistent with previous study findings. (26–29) First, having only rectal adenomas diagnosed was associated with an increased risk of colorectal cancer, implying the colon may not have been examined. No colorectal cancer occurred among individuals who had adenomas diagnosed in both the colon and rectum. Similarly, having multiple adenomas removed was inversely associated with colorectal cancer risk. Finally, having concurrent adenoma and hyperplastic polyp removal at incident polypectomy (which suggests more thorough primary clearance of the bowel) or subsequent hyperplastic polyp detection (which indicates follow-up endoscopy) was associated with a reduced risk of colorectal cancer. This was also corroborated by results from the case note review, in which individuals undergoing a full colonoscopy had decreased risks of colorectal cancer, whereas patients who had other investigative procedures such as incomplete colonoscopy, flexible sigmoidoscopy, or barium enema had an increased risk of colorectal cancer. These findings have implications for current surveillance guidelines following polypectomy (9). Our results suggest that a significant proportion of patients diagnosed with colorectal adenoma do not undergo full colonoscopy and adenoma clearance, and this incurs an increased risk of colorectal cancer. The reasons for this are unclear but merit further consideration. Although patient choice is foremost in decision making, understanding potential barriers for colonoscopy compliance is an important area for future research. Having multiple full colonoscopies following incident adenoma removal was inversely associated with colorectal cancer risk, which argues for the benefit of continued surveillance (30). The increased yield of adenomas and colorectal cancer when performing full colonoscopy compared with other procedures is well established (31–33).

We also observed an unexpected increased colorectal cancer risk among individuals having only hyperplastic polyps removed. The potential for hyperplastic polyps to be a marker of adenomas has been previously debated (34, 35). Our results do not support this finding, since an increased colorectal cancer risk remained when individuals with subsequent adenomas were removed from analysis. However, we were unable to account for adenomas detected after 2005 and undetected concurrent or subsequent adenomas may still explain the excess colorectal cancer risk seen. Sessile serrated polyps represent another subset of serrated polyps, much rarer than hyperplastic polyps, and have only been well described in recent years (36). Sessile serrated polyps are more common in the proximal colon, are often multiple, are larger than typical hyperplastic polyps, and are now accepted as the precursors to serrated pathway colorectal cancers (37, 38). Sessile serrated polyps were not frequently reported by pathologists during this study period and it is likely many such polyps were misdiagnosed as hyperplastic polyps (36, 39). This is supported by our findings that multiple and right-sided “hyperplastic polyps” carried an increased risk of colorectal cancer progression.

Our study has several strengths, including its large size, length of follow-up and relevance to current practice, given the timeframe of adenoma patients studied. The population-based nature of the study is important to emphasize, demonstrating colorectal cancer risk following polypectomy in a “real world” setting. This is the first insight into this association in a UK population, among whom health care is free at the point of access. Northern Ireland has a relatively stable population with limited migration (40), therefore the completeness of passive follow-up is excellent. Furthermore, our results were not diluted with bowel cancer screening-detected adenomas/cancers because a screening programme only commenced in mid-2010. This study does have some limitations that should be considered. In the NICPR, we did not have information on polyps diagnosed prior to 2000 or after 2005, or information on polyp size, or detailed polyp topography. However, we overcame these limitations by conducting a detailed note review in a nested case–control study. The number of polyps in patients with more than one polyp was also underestimated due to the potential for SNOMED codes to have been entered only once by the reporting pathologist, if a patient had multiple polyps diagnosed of the same histological type at one episode. Comparison with the case note review data suggests that approximately 10% of patients with multiple adenomas were misclassified as singular in the overall register.

In conclusion, this large population-based study demonstrates that colorectal cancer risk remains significantly higher than that of the general population following adenoma removal, outside of screening programs. Our findings suggest the increased colorectal cancer risk may be partially due to incomplete examination and/or adherence to follow-up colonoscopies. There is a need for full colonoscopy and adenoma clearance after a diagnosis of colorectal adenoma to reduce risk of subsequent colorectal cancer. Further research is required to identify optimal surveillance protocols, and adherence to these, following adenoma removal.

No potential conflicts of interest were disclosed.

Conception and design: M.B. Loughrey, L.J. Murray, B.T. Johnston, P.B. Allen, M.M. Cantwell

Development of methodology: H.G. Coleman, L.J. Murray, B.T. Johnston, P.B. Allen, M.M. Cantwell

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.G. Coleman, L.J. Murray, A.T. Gavin, M.M. Cantwell

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.G. Coleman, M.B. Loughrey, L.J. Murray, B.T. Johnston, A.T. Gavin, M.J. Shrubsole, S.K. Bhat, M.M. Cantwell

Writing, review, and/or revision of the manuscript: H.G. Coleman, M.B. Loughrey, L.J. Murray, B.T. Johnston, A.T. Gavin, M.J. Shrubsole, S.K. Bhat, V. McConnell, M.M. Cantwell

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.G. Coleman, A.T. Gavin, M.M. Cantwell

Study supervision: H.G. Coleman, B.T. Johnston, M.M. Cantwell

Other (the genetic aspect of this manuscript and patients known to the regional genetic service whose pathology of adenoma, etc., was genetic and was integral to the manuscript interpretation as opposed to sporadic adenomas): V. McConnell

The authors acknowledge the contribution of IT and data management staff in the Northern Ireland Cancer Registry, namely Dr. Richard Middleton, Mr. Giulio Napolitano, and Mr. Colin Fox, for their help in collecting and extracting data that contributed to the Northern Ireland Colorectal Polyp Register. Special gratitude is extended to Mrs. Donna Floyd and Ms. Michelle McGaughey for their assistance in piloting and designing the data extraction form used in the case note review for the nested case–control study, and to Mrs Donna Floyd, Mrs. Rosemary Ward, and Mrs. Kate Donnelly (all Northern Ireland Cancer Registry), for conducting the case note review. We would also like to acknowledge the input of Dr. Khaled Kasim in initiating cleaning and analysis of the Northern Ireland colorectal polyp register database.

This work was supported by funding from Cancer Focus Northern Ireland (formerly the Ulster Cancer Foundation) awarded to Marie M. Cantwell, Anna T. Gavin, and Liam J. Murray. The Northern Ireland Cancer Registry (for which Anna T. Gavin is Director) is funded by the Public Health Agency for Northern Ireland.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.