Crohn and Rosenberg first described rectal cancer complicating UC more than 80 years ago. In their manuscript, they suggested that the malignancy was a complication of the disease [16]. Three years after Crohn and Rosenberg, Bargen, at the Mayo Clinic, reported a series of 17 patients with both chronic colitis and colorectal cancer [17]. Other cases and series followed, and the crude frequency calculations from these studies served as evidence supporting a link between UC to CRC. With the application of modern epidemiologic methods, true incidence calculations, cumulative incidence calculations, and standardized incidence rates confirmed the association between UC and CRC. Cumulative incidence rates have largely become the standard by which clinicians and public health experts assess the time-dependent risk of cancer develops in colitis. Similarly, standardized incidence rates describe the estimate of the relative risk for developing colon cancer for a segment of a colitis population (such as colitis patients with universal disease) as compared to the general population. While initial series using this more accurate epidemiologic terminology came from large referral centers in which “incident” cases were referred for evaluation and management due to a suspicion for or even the actual presence of CRC, the use of more appropriate terminology was an advance over the previously used crude rates [18–20]. Due to these now obvious referral biases, however, these first “modern” studies overestimated the true risk of CRC in UC. Subsequent studies from population-based data sources used more realistic calculations for determining the incidence of CRC in UC. Without referral and other selection biases, the cancer incidence calculated in these manuscripts was substantially lower than previously reported [21–25]. These studies, however, may have underestimated the true risk of cancer in long-standing UC, as they included many patients with UC who had undergone previous colectomy in the denominator of the incidence calculations. In a meta-analysis of the risk of CRC in ulcerative colitis in which 116 studies were included, Eaden and colleagues found the overall prevalence of CRC to be 3.7% and an overall incidence rate of 3 cases per 1000 person years duration (95% confidence interval ranging from 2 to 4 cases per 1000 person years duration). The rate increased with each decade of disease, leading to a calculated incidence of 12 per 1000 person years in the third decade of colitis [18]. These data corresponded to a cumulative incidence of CRC of 2% at 10 years, 8% at 20 years, and 18% at 30 years disease duration [26]. It is worth noting, however, that referral centers accounted for 64% of the studies included in Eaden’s study; only 13 population-based reports were located by the Medline search performed as part of the meta-analysis [26]. Based on these and older data, typical estimates of CRC incidence usually range between 0.5–1% per year after 10 years of colitis.

More recent studies, however, have raised the possibility that prior studies have overestimated the incidence and risk for CRC in this population. More recent publications from Denmark [27], Hungary [28], Canada [29], and Olmsted County, Minnesota (with its relatively small population) [30], have suggested a CRC in UC incidence of between 1 in 500 and 1 in 1600 per year, far lower than the 1 in 300 rate calculated in Eaden’s meta-analysis [26]. These have corresponded to relative risk calculations ranging from 1.1 to 2.7 times the general population. While some have argued that these more “modern” calculations support a declining incidence over calendar time, as seen by Rutter and colleagues [31], no definitive analysis has been performed to support this hypothesis. To what extent such reductions in incidence (if they exist) are a function of colonoscopic surveillance (see below), chemoprevention with mesalamine-based agents or other medicines (also below), or other factors remains unknown.

A number of clinical variables have been demonstrated to modify the risk for colorectal cancer in UC patients. These variables include duration of UC, anatomic extent of disease, age at UC diagnosis, concomitant primary sclerosing cholangitis (PSC), a family history of colorectal cancer, and inflammatory activity. The use of certain medications may lessen the risk of developing CRC, but the impact of these potentially chemopreventive agents is modest. Table 55.1 classifies these different risk modifiers.

A number of investigators have demonstrated that the duration of ulcerative colitis correlates with the risk of cancer [18, 32–34]. Duration of disease, however, can be a rather subjective measurement. Most studies have used the date of UC diagnosis as the point at which the clock starts, but others have argued that the time of symptom onset is a better measure of disease duration. Whichever point is chosen, a number of distortions can be imagined that would impact the findings in any individual study. If date of diagnosis is used as a starting point, then patients with long-standing, subclinical disease would appear to have relatively shorter duration of disease, and such subjects would contribute less to any calculation of the effect of disease duration. Conversely, by using date of first symptoms, subjects who were without colitis might mistakenly contribute years of disease-free time to calculations of duration. This distinction in the definition of disease duration may be particularly problematic for patients with primary sclerosing cholangitis (PSC) who frequently have clinically quiescent colitis. Without unanimity in definition, there is variability in the estimate of this factor’s effect on subsequent CRC development. In Eaden’s meta-analysis, the effect of duration was made clear as the passage of each successive decade resulted in an increase in incidence. Incidence was calculated to be 2 per 1000 patient years (95% C.I. 1–4/1000) at 10 years and 11 per 1000 (95% C.I. 4–28/1000) at 30 years; the rate at 20 years was intermediate [26]. As the overall curve for cumulative CRC risk starts to meaningfully exceed that of the general population by 8–10 years, most clinicians will initiate surveillance colonoscopy once this threshold has been reached. Because many of the studies that were entered in to Eaden’s meta-analysis antedated the widespread application of colonoscopic dysplasia surveillance, it remains unclear whether duration of colitis exerts a seeming exponential effect, as Eaden found, or a linear effect, which might result if highest-risk patients are serially removed from the denominator via colectomy from surveillance-identified dysplasia.

The length of involved colon also correlates with cancer risk: the greater the surface area of colitis, the greater the cancer risk. Defining the anatomic extent of ulcerative colitis, as with duration of disease, can vary from study to study. In initial reports documenting this independent risk factor, anatomic extent was defined by a barium enema at diagnosis. Flexible endoscopy long ago replaced barium radiography for diagnosing colitis and its extent, but there is no consensus as to whether naked eye findings at colonoscopy or microscopic extent determined histologically should be the gold standard for measuring extent. Additionally, definitions of “pancolitis,” “universal colitis,” and “extensive colitis” vary within studies, although they are all typically used to describe disease proximal to the splenic flexure. Another feature that invites confusion into the definition of anatomic extent is the timing of the measurement. As extent can change over time [35], should we take the extent at diagnosis or at some point in follow-up? Like other questions surrounding the issue of extent, this question has been left unresolved, although the majority of studies have used the terms “extent” and “extent at diagnosis” interchangeably. Extent at follow-up has not been well studied as an independent risk factor.

A population-based investigation of a cohort of more than 3000 patients with UC defined extent of UC by barium enema exam at diagnosis and demonstrated an impressive gradient of risk as one moves from proctitis (standardized incidence ratio of 1.7, 95% confidence interval 0.8–3.2) to left-sided colitis (SIR=2.8, 95% C.I. 1.6–4.4) to pancolitis (SIR=14.8, 95% C.I. 11.4 to 18.9) [22]. Devroede [18], Greenstein [34], Gyde [33], Katzka [36], Mir-Madjlessi [37], and Gilat [23] all reported similar gradients in their studies. This finding was confirmed, though not directly studied, in Eaden’s meta-analysis [26]. In terms of “how” extent should be defined, it is worth noting that a group from University of California San Francisco found CRC in areas proximal to the endoscopically perceived margin of colitis that turned out to have microscopic disease in that region [38]. On this basis, clinicians should consider the most proximal extent of disease microscopically as the proximal extent of disease and plan any prevention strategy accordingly.

Important to pediatricians, age of colitis onset, as a variable independent of disease duration, has been implicated in some studies to modify the risk of IBD-related colon cancer. This hypothesis, however, remains in question. Reporting one of the highest published cumulative rates of CRC in colitis, Devroede and colleagues found that at 35 years of follow-up, 43% of subjects with documented UC prior to age 15 had developed CRC [18]. This study, however, reflected pediatric patients seen at a large referral center; additionally, the number of patients available to analyze after 35 years of follow-up was quite small, with the error surrounding this point estimate correspondingly quite wide. While some investigators have failed to demonstrate a link between age of colitis onset and the subsequent development of CRC [23, 36], others have confirmed the direction if not the magnitude of Devroede’s findings [22, 33]. In the previously mentioned study by Ekbom, for example, the authors found that the relative risk of cancer in colitis decreased with advancing age—younger patients have a higher risk [22]. This overall gradient was confirmed by Eaden, who found that cumulative rates of CRC were greater than the pooled estimates for CRC among adult colitis patients, though this difference did not meet conventional thresholds for statistical significance [26]. Although neither the precise nature nor the precise magnitude of CRC risk for younger patients with UC has been determined, extra caution should be applied to pediatric patients given both the suggestion of an increased risk from the medical literature and the obvious increased lifetime risk given a longer life expectancy.

Primary sclerosing cholangitis is a chronic cholestatic liver disease in which there is progressive inflammatory fibrosis of the biliary tree. It is an infrequent complication of IBD, affecting 2–8% of patients with ulcerative colitis. However, among patients with PSC, 62–72% have underlying IBD. Since the intersection of CRC and PSC would be expected to occur in small absolute numbers in patients with UC, it is largely through case-control studies and referral center-based cohort studies that the majority of data have been generated to support an association between PSC and CRC in UC. Although a positive association has not always been noted [39–41], most studies do support such an association, with derived odds ratios from these “positive” studies ranging from 9 to 16 [42–46]. In a population-based study from Sweden, Kornfeld and colleagues found a substantially elevated cumulative incidence of CRC in UC/PSC patients: 33% at 20 years [47]. As noted above, since colitis activity in PSC is often mild or even subclinical, PSC patients in these studies might well have had a longer duration of disease than was appreciated, making it difficult to tease out the precise, independent contribution of PSC to the development of CRC.

Family history of CRC has long been recognized as a risk factor for the development of sporadic colorectal cancer. This risk increases according to the number of relatives affected with CRC [48]. In UC, only a few clinical studies have been performed to investigate the independent contribution of a positive family history for colorectal cancer. An early study from Lashner’s group at the University of Chicago supported family history of CRC as a potential risk factor for CRC in colitis, although the association did not reach statistical significance [49]. A second report from the Cleveland Clinic documented a lower rate of positive family history of CRC among UC patients with cancer or dysplasia compared to UC controls without colonic neoplasia, though this finding, too, failed to exclude the null hypothesis [50]. Both of these studies, however, were designed to test hypotheses concerning the association between folic acid supplementation and colorectal cancer in colitis. Testing for family history as a risk factor was performed as part of a secondary analysis, and these studies did not specify the rigor with which a family history was obtained.

More recently, a handful of studies have suggested an increased risk for CRC in UC when a positive family history of CRC was documented. Nuako and colleagues at the Mayo Clinic were the first to clearly demonstrate this increased risk, calculating an odds ratio of 2.3 (95% CI 1.1–5.1) in their case-control study [51]. In a population-based study from Scandinavia, Askling and colleagues found a similar elevated risk of 2.5 (95% C.I. 1.4–4.4) [52], while Eaden (in the UK) found an even greater risk (OR 5.0, 95% CI 1.1–22.8) in a multivariable model using case-control derived data [53]. Whatever the absolute magnitude, it appears quite likely that a positive family history confers an increased risk of CRC in UC.

Curiously, although inflammation has been assumed to be a key factor contributing to higher risk of colonic neoplasia in UC, few studies have examined this issue. One well-conducted retrospective case-control study recently reported that histologic inflammation was indeed associated with an increased risk of neoplastic progression based on a retrospective case-control analysis of patients followed at a specialized center [54]. A retrospective cohort study from Mount Sinai, New York, has also demonstrated a link between histologic inflammation on dysplasia and cancer risk, with a twofold risk increase for each unit of inflammation derived from a 4-point scale [54, 55].

Until we discover or develop a meaningful chemopreventive agent and effective strategies to identify a minimal risk subgroup, only two acceptable forms of CRC prophylaxis exist: surgery and dysplasia surveillance. In dysplasia surveillance, high-risk patients are identified by the identification of neoplasia (either dysplasia or cancer) at colonoscopy and are subsequently referred to surgery, while cancer and dysplasia-free patients continue with periodic colonoscopy [66]. The presumption is that only the highest-risk patients will undergo a colectomy, and lower-risk patients will be able to maintain a higher quality of life with their colons intact. A third option, watch and wait, with colonoscopy performed only for symptoms, is available but due to the available evidence that symptomatic cancers are associated with a worse survival than asymptomatic ones [67, 68] never used in clinical practice.

Without question, the most effective method for minimizing CRC risk in UC patients is to perform a total proctocolectomy. This nearly eliminates the risk of colon or rectal cancer, and, while cancers have been reported in case reports and series in patients who have undergone either hand-sewn or stapled anastomoses, the risk of such an event is quite small. In the pre-endoscopic era, this strategy of cancer prevention was often advocated for patients with long-standing colitis, and should still be considered, particularly for patients with medically refractory or difficult disease. As surgery is not without its potential complications and change in quality of life, however, and as the absolute risk of developing a lethal colon cancer may not be sufficiently high to warrant such a radical approach in all colitis patients, surgical prophylaxis in asymptomatic patients with long-standing colitis is now viewed with skepticism by both patients and clinicians. At present, surgical options (for colorectal cancer prophylaxis or as primary treatment for colitis-related dysplasia or cancer) include total proctocolectomy with creation of an ileal pouch-anal anastomosis (often referred to as a restorative proctocolectomy) or total proctocolectomy with end ileostomy. Subtotal colectomy with ileorectal anastomosis is to be avoided, although there are no studies comparing this procedure to either of the other conventional choices. Pouch surgery is generally reserved for younger patients, as it requires sufficient anal sphincter tone. Following pouch surgery, patients may expect to have five or more bowel movements per day due to pouch size and ileal flow. Possible complications include sexual and bladder dysfunction, incontinence, pouchitis (which usually responds to short courses of antibiotics but may become chronic and refractory), fistula formation, stricture formation, anastomotic leakage, and pouch failure. The overall failure rate (the proportion of patients eventually converted to end ileostomy) is approximately 5% [69]. It should also be noted that the malignant potential of ileal pouch mucosa in colitis patients remains unknown. Initial reports of pouch dysplasia have been reported, and there have been reports of cancer in the cuff of rectal mucosa to which the pouch is anastomosed [70, 71]. While cancer risk following proctocolectomy with Brooke ileostomy is close to nil, the loss of anorectal function and attendant stoma make this option less appealing to most patients who would otherwise be candidates for pouch surgery. Potential complications of total proctocolectomy with end ileostomy include sexual and bladder dysfunction, stomal fistula, parastomal hernia, and small bowel obstruction [69].

As it results in too many colectomies in patients who would otherwise be unaffected by CRC, prophylactic total proctocolectomy is seldom performed. Even if limited to the high-risk groups of patients with long-standing and extensive UC, with or without PSC or a family history of CRC, a large number of colectomies would be performed in patients who would never develop CRC. What is needed is a tissue marker that better determines the highest-risk patients, those with an imminent risk of colorectal cancer. While imperfect on many levels, mucosal dysplasia serves as such a marker.

In 1967 Morson and Pang first reported the association between mucosal dysplasia and CRC in patients with UC [72]. In their seminal report, they noted that rectal dysplasia, then termed “precancer” and identified by blind rectal biopsy of colitis mucosa, heralded the presence of an invasive adenocarcinoma elsewhere in the colon. If appropriately discriminating, mucosal dysplasia, it was hypothesized, could be used as a diagnostic test to identify the highest-risk patients to whom surgery would be offered.

Subsequent studies revealed that, although by no means a perfect test, dysplasia was discriminating enough to be tested in clinical practice. Retrospective studies confirmed Morson and Pang’s findings, noting the presence of dysplasia either adjacent to or remote from cancer in colitis [73–75]. Additionally, cancer foci were discovered in colons resected for the indication of dysplasia [76]. These data, along with the advent of flexible fiber-optic instruments with their ability to deliver multiple mucosal samples to the pathologist’s microscope, led to the development of protocol-based surveillance programs. Unfortunately, no randomized, controlled trials of surveillance were performed. (This may have been a function of difficulty in defining suitable control patients: would patients allow themselves to be randomized to a “no surgery” or “no endoscopy” arm of a surveillance study? Or to a “prophylactic surgery” arm?) Nevertheless, based on the clinical characteristics of dysplasia and the results of numerous surveillance programs, as well as the very limited acceptability of other prevention strategies, namely, surgery for all long-standing colitis or waiting for cancer symptoms, periodic colonoscopy with biopsy for dysplasia became an accepted form of cancer prevention in UC. In addition to its widespread use in clinical practice, it has been advocated in guidelines statements for colon cancer prevention [77] and ulcerative colitis care [78].

Single-armed surveillance programs have demonstrated the feasibility, though not the efficacy, of conventional surveillance [24, 76, 79–89]. When “control” arms were used in these studies, they included patients in whom surveillance at another institution or referral to the institution for malignancy could be considered as “no surveillance.” Nevertheless, the finding that cancers found during surveillance were more often at earlier stages than cancers found in a “watch and wait” strategy contributed to the acceptance of dysplasia surveillance as a form of cancer prevention [67, 68]. Other key features about surveillance programs worth noting include the presence of advanced stage cancers despite inclusion in a surveillance program (some due to patient dropout and some due to progression while under surveillance) [86, 89, 79] the variable intervals used for surveillance, the variable rates of patient dropout, and the substantially varied rates of dysplasia incidence across studies. For surveillance to be effective, it should reduce CRC mortality in IBD patients. In the absence of prospective controlled studies, a well-designed population-based case-control study sheds light on this issue. Karlen and colleagues compared the exposure to colonoscopy among cases with CRC deaths and alive controls matched for age, gender, disease duration, and disease extent [90]. The point estimate of cancer mortality reduction from either one or only two previous colonoscopic exams was a threefold decrease. Although the odds ratio of 0.3 did not reach statistical significance, (95% CI: 0.1 to 1.3), this is certainly a clinically impressive result. A recent case-control study from the Mayo Clinic confirmed these findings and even crossed the threshold of statistical significance with an odds ratio of 0.4 for 1–2 surveillance examination (95% CI: 0.2 to 0.7) [91]. While these data were not population-based in their orientation, they nevertheless support the notion that surveillance is likely effective. Additional support comes from decision analysis models [92–94] that demonstrate improved outcomes for a population in surveillance compared to no surveillance. As with all such modeled data, there are many assumptions that lack real-world support, such as a lack of dropout while under surveillance and an orderly progression from no dysplasia to low-grade dysplasia to high-grade dysplasia to colorectal cancer [92, 93, 95]. Cost-effectiveness analyses have similarly predicted that surveillance was a superior strategy to no surveillance (although prophylactic colectomy, while unacceptable to patients, was the preferred strategy vis-à-vis life-years saved over time).

What then might limit the effectiveness of dysplasia surveillance in UC in practice? One factor may be difficulties in histologic interpretation among pathologists. Indeed this was thought to be so substantial a problem after the initial reports of surveillance studies that in 1983, an international group of experts convened to establish true definitions for the evaluation of dysplasia surveillance specimens: no dysplasia, indefinite for dysplasia (with three subtypes), low-grade dysplasia, high-grade dysplasia, and colorectal cancer [96]. Unfortunately, despite these codified definitions, substantial rates of disagreement, even among expert GI pathologists, have been noted [96–99]. Rates of disagreement among community pathologists, not surprisingly, have been substantial, too [97]. In these studies, crude rates of agreement have been as low as 40% and as high as 72%, with best agreement when no dysplasia was present; kappa values, which can account for chance agreement, were fair to good. Clearly, this system needs less subjectivity and overall improvement.

Lack of perfection from practicing pathologists is not the only reason for surveillance not to reach its potential. Gastroenterologists also fall short of ideal practices. One variable that contributes to lack of uniform clinician practices stems from the uncertainty that surrounds the predictive value of dysplasia. While there is near-universal agreement that patients found to have high-grade dysplasia should undergo colectomy due to rates of concurrent adenocarcinoma near 50% [74], considerable controversy surrounds the management of low-grade dysplasia. Adding to the controversy is the fact that LGD can be flat or polypoid, unifocal or multifocal, or not repeatedly found on sequential colonoscopic exams. Few studies have directly addressed these variables in patients with LGD.

How to best manage LGD depends in large part on how likely patients with this finding are to either already harbor or progress to more advanced neoplasia (HGD or cancer). More specifically, the essential unanswered question is whether failure to perform a colectomy in patients with LGD results in a poor outcome. In a landmark study from St. Marks Hospital in which the Inflammatory Bowel Disease Morphology Study Group’s 1983 definitions were used [96], the rate of progression to advanced dysplasia from LGD was 54% at 5 years [100]. In the same year as the St. Mark’s publication, a systematic review of surveillance programs by Bernstein and colleagues noted a 19% rate of cancer at “immediate colectomy” following the discovery of LGD. These results were confirmed by studies from the Mayo Clinic [101] and Mount Sinai in New York [102], in which the rates of progression for flat LGD were 33% (95% CI 9 to 56%) and 53% (95% CI 29 to 77%), respectively. Furthermore, in the Mount Sinai study, 19% of patients who underwent colectomy within 6 months of their initial flat LGD finding were found to have CRC in their resection specimens. Of those who progressed, cases of node-positive cancer without intervening HGD were found. Neither the number of biopsies positive for LGD nor any other clinical variable were found to be predictive of subsequent progression, with unifocal flat LGD carrying a 5-year rate of progression of 53% [102]. Investigators from the University of Washington where an aggressive biopsy protocol is followed [103] and from the Karolinska Institute, Sweden [104], however, discovered less frequent progression and no cancers. Not all investigators have discovered the same high risk for LGD as that noted by St. Mark’s, Mayo Clinic, and Mount Sinai in New York. A group from Karolinska in Sweden noticed a near-total lack of progression following discovery of LGD, but this group’s pathologists did not use the full panoply of IBD Morphology Study Group definitions, as readings of “indefinite” for dysplasia were not allowed. Additionally, a number of patients were included whose discovery of LGD occurred prior to establishment of the Riddell criteria [105]. Additionally, the Leeds, UK, group led by Lim found little progression from LGD, leading him and his coauthors to conclude that continued surveillance with satisfactory biopsy practices was a safe alternative to surgery [106]. Finally, the University of Chicago group found a low actuarial rate of progression among patients with both flat and polypoid low-grade dysplasia [107]. While the variable rates of progression (perhaps secondary to variable biopsy practices, observer variation in the interpretation of dysplasia, or imperfect follow-up) make it difficult to draw absolute conclusions for the management of patients with flat LGD, early colectomy for LGD that is histologically confirmed by two expert pathologists should be strongly considered at the least. For patients who defer or refuse colectomy for LGD, gastroenterologists must make certain that patients return for follow-up examinations and that surveillance is appropriately performed with an adequate number of biopsies taken to exclude dysplasia.

It should be noted that a negative exam following LGD can occur for a number of reasons: (1) the previous examination was a false positive due to pathologic interpretation error; (2) the present examination is a false negative due to sampling or interpretation error; or (3) both exams were accurate. Not finding dysplasia on a repeat colonoscopy following one that detected LGD is no reassurance that dysplasia has regressed or will not “recur” [89]. It was estimated that to exclude dysplasia with 95% confidence, 56 biopsies must be performed, and to exclude 90% confidence, 33 biopsies should be taken [108]. This number of biopsies is rarely performed even in academic centers [109, 110]. Eaden noted that 57% of UK gastroenterologists take fewer than 10 biopsies in a surveillance exam based on their response to a questionnaire [109]. In a study examining actual gastroenterologists’ practices, Ullman and colleagues found that the mean number of evaluable biopsies in patients with LGD was only 17.5 [102]. Such undersampling represents another limitation for dysplasia surveillance among gastroenterologists. Whether such practices truly limit the effectiveness of surveillance remains unknown.