Cancer in inflammatory bowel disease

Abstract

Patients with long-standing inflammatory bowel disease (IBD) have an increased risk of developing colorectal cancer (CRC). Many of the molecular alterations responsible for sporadic colorectal cancer, namely chromosomal instability, microsatellite instability, and hypermethylation, also play a role in colitis-associated colon carcinogenesis. Colon cancer risk in inflammatory bowel disease increases with longer duration of colitis, greater anatomic extent of colitis, the presence of primary sclerosing cholangitis, family history of CRC and degree of inflammation of the bowel. Chemoprevention includes aminosalicylates, ursodeoxycholic acid, and possibly folic acid and statins. To reduce CRC mortality in IBD, colonoscopic surveillance with random biopsies remains the major way to detect early mucosal dysplasia. When dysplasia is confirmed, proctocolectomy is considered for these patients. Patients with small intestinal Crohn’s disease are at increased risk of small bowel adenocarcinoma. Ulcerative colitis patients with total proctocolectomy and ileal pouch anal-anastomosis have a rather low risk of dysplasia in the ileal pouch, but the anal transition zone should be monitored periodically. Other extra intestinal cancers, such as hepatobiliary and hematopoietic cancer, have shown variable incidence rates. New endoscopic and molecular screening approaches may further refine our current surveillance guidelines and our understanding of the natural history of dysplasia.

Key Words: Colon cancer, Inflammatory bowel disease, Dysplasia, Chemoprevention, Colonoscopy, Genomic instability

INTRODUCTION

The two forms of inflammatory bowel disease (IBD), Crohn’s disease (CD) and ulcerative colitis (UC), are characterized by chronic, relapsing inflammation of the intestines. First described in a report by Crohn and Rosenberg in 1925[1], colorectal cancer (CRC) in patients with IBD has long been recognized. Even years after their disease is controlled with medications, IBD patients still live with the fear of developing cancer. CRC is the most common site of cancer in IBD, although cancer in other organs can occur. Together with the hereditary syndromes of familial adenomatous polyposis and hereditary nonpolyposis colorectal cancer, IBD is among the top three high-risk conditions for CRC. To date, our understanding of CRC pathobiology has come from studies of patients with UC more so than Crohn’s colitis. This review will focus mainly on the problem of CRC, and then address other cancers such as small intestinal adenocarcinoma, cholangiocarcinoma, and hematologic malignancies.

COLORECTAL CANCER IN PATIENTS WITH IBD

Prevalence and Incidence of CRC

The exact magnitude of the risk of cancer has been difficult to quantify due to various biases and methodological errors in published studies. Early estimates of CRC complicating UC were based on crude percentages and all were from major medical institutions, predominantly tertiary referral centers. Studies from these centers often included a greater proportion of patients who had more severe disease and cancer had already complicated their colitis. These center-based studies often overestimate the risk. Later population-based studies tended to include more patients with limited disease or those who have undergone colectomy and may thereby underestimate the true risk. Based on a 2001 meta-analysis by Eaden et al, including 116 studies from around the world, the prevalence of CRC in patients with UC is approximately 3.7% overall and 5.4% for those with pancolitis[2]. In comparison, CRC in Crohn’s disease has been less well studied. Early studies showed no statistically significant increase in cancer risk among the Crohn’s disease patients. However, the lack of the risk of cancer in these studies was often due to inclusion of all patients with Crohn’s disease and failed to correct for the small subsets of those with extensive, longstanding and unresected colonic disease. Hence, when patients with longstanding, anatomically substantial Crohn’s colitis are considered, the risk of CRC is similar between Crohn’s colitis and UC[3]. Indeed, a population-based study from Manitoba, Canada found that the risk for colon cancer among patients with both UC and Crohn’s colitis is approximately 2-3 fold greater than the general population and that the risk of rectal cancer is increased 2-fold in UC but not Crohn’s colitis[4].

Clinical features of colitis-associated CRC

Compared with sporadic colorectal carcinoma (SCC), CRC arising in patients with IBD has several distinguishing clinical features. Colitis-associated colorectal cancer (CAC) affects individuals at a younger age than the general population. CAC progresses to invasive adenocarcinoma from flat and nonpolypoid dysplasia more frequently than SCC. CACs more often have a higher proportion of mucinous and signet ring cell histology. There is background of chronic inflammation in colitis and a higher rate of two or more synchronous primary CRCs. The multifocality of CAC relates to the broader field effect of mucosal inflammation that gives rise to the neoplasia. In some studies, patients were found to have cancer more proximal in the colon. Finally, the sequence of molecular events leading from dysplasia to invasive adenocarcinoma is different from that of SCC (discussed below).

Molecular features of sporadic colon cancer

To place the molecular pathogenesis of colitis-associated neoplasia in proper perspective, it is important to appreciate the molecular events involved in the development of SCC. SCC arises as a result of genomic instability. The two main types of genomic instability that contribute to colon carcinogenesis are chromosomal instability (CIN) and microsatellite instability (MSI), accounting for 85% and 15% of SCC, respectively. Chromosomal instability results in abnormal segregation of chromosomes and abnormal DNA content (aneuploidy). As a result, loss of chromosomal material (loss of heterozygosity) often occurs, such as APC and p53. These genes can also be rendered nonfunctional by mutation.

Loss of APC function is typically an early event in SCC pathogenesis. The APC gene thus has been considered the “gatekeeper” of the colon (Figure 1). Some 85% of all sporadic and inherited colorectal tumors show loss of APC function, usually through protein truncation or allelic loss[5]. The APC gene is located on chromosome 5q21-q22. The key tumor suppressor function of the APC protein lies in its ability to destabilize free β-catenin[6]. Among the 15% of colon carcinomas that retain wild-type APC, point mutations have been found in β-catenin that change one of the four serine/threonine residues in the N-terminus, the putative targets of glycogen synthase kinase-3β (GSK-3β)[7]. These mutations thus render β-catenin refractory to phosphorylation by GSK-3β, increasing free β-catenin levels. Accumulation of stabilized free β-catenin is an early event and perhaps the initiating event in intestinal tumorigenesis. Inactivation of both APC alleles is found in a majority of small colorectal adenomas in humans and in the smallest detectable tumors in mice heterozygous for an inactivating mutation in APC[8]. Furthermore, intestine-specific expression of a dominant-negative form of β-catenin, which lacks the putative GSK-3β targets sites, leads to the development of adenomas[9]. How the loss of APC or stabilization of β-catenin leads to development of cancer is not yet fully understood. Once a sporadic adenoma forms, other changes in genetic regulation occur, such as induction of k-ras oncogene and loss of function of tumor suppressor genes on chromosome 18q in the region of the deletion in colon cancer (DCC) and in pancreatic cancer (DPC4) genes. Loss of p53 gene function occurs late and is believed to be the defining event that drives the adenoma to carcinoma.

Figure 1 Comparison of molecular alterations in sporadic colon cancer and colitis-associated colon cancer. Mut, mutation. Modified from Ref 71. With permission.

Tumors that arise via the CIN/tumor suppressor gene pathway are typically microsatellite stable (MSS). The remaining 15% of sporadic CRCs arise through the MSI pathway. The MSI pathway involves the primary loss of function of genes that usually repair DNA base-pair mismatches that occur during the normal process of DNA replication in dividing cells. In humans, at least six different proteins (hMSH2, hMLH1, hPMS1, hPMS2, hMSH6, and hMLH3) comprise the mismatch repair system[10]. These proteins form specific heterodimers to coordinate DNA repair and the recruitment of other proteins, such as polymerases and helicases, necessary for mismatch repair. Germline mutations of DNA mismatch repair genes, predominantly hMLH1 (33%) and hMSH2 (31%), are responsible for the familial syndrome of heriditary nonpolyposis colon cancer[11]. In addition, approximately 15% of sporadic tumors from the colon, rectum, and other organs demonstrate MSI[12]. Interestingly, the most common mechanism causing MSI in sporadic colon cancers is not genetic mutation, but rather transcriptional silencing of hMLH1 as a consequence of methylation of the hMLH1 promoter[13].

Epigenetic alterations can also contribute to altered gene expression in colon carcinogenesis. The CpG island methylator phenotype occurs when cytosines in the promoter region of genes become extensively methylated. A number of human cancer genes that contain hypermethylation of promoter CpG islands have been identified[14]. These include hMLH, p16^INK4a, and E-cadherin. The process of methylation is an area of intense investigation, and it is anticipated that this line of research should help to define further the molecular pathways involved in CRC in a variety of clinical settings.

Molecular features of CRC in IBD

The neoplastic transformation in IBD is thought to be similar to the adenoma-carcinoma sequence in sporadic CRC (Figure 1). However, unlike SCC, where dysplastic lesions arise in one or two focal areas of the colon, in colitic mucosa, it is not unusual for dysplasia or cancer to be multifocal, reflecting a broader “field effect”. Many of the molecular alterations responsible for sporadic CRC development also play a role in colitis-associated colon carcinogenesis. The emerging evidence suggests that the two major pathways of CIN and MSI also apply to CACs and with roughly the same frequency (85% CIN, 15% MSI). Distinguishing features of CAC, however, are differences in the timing and frequency of these alterations (Figure 1). For example, APC loss of function, considered to be a very common early event in SCC, is much less frequent and usually occurs late in the colitis-associated dysplasia-carcinoma sequence. Conversely, p53 mutations in sporadic neoplasia usually occur late in the adenoma-carcinoma sequence, whereas in patients with colitis, p53 mutations occur early and are often detected in mucosa that is non-dysplastic or indefinite for dysplasia.

Methylation is assuming increasing importance as a mechanism contributing to the genetic alterations in CAC (Figure 1). Methylation of CpG islands in several genes seems to precede dysplasia and is more widespread throughout the mucosa of UC patients[15].

CRC risk factors

Several factors have been identified which either increase or decrease CRC risk in the setting of IBD (Table 1). With respect to increasing CRC risk, the most important factor, reproducibly found across many studies, is the duration of colitis. CRC is rarely encountered before 7 years of colitis. The pooled estimate of cumulative CRC incidence in UC in Eaden’s meta-analysis was 2% at 10 years, 8% at 20 years, and 18% after 30 years of disease[2]. This may overestimate the risk since many studies in the meta-analysis were from the pre-surveillance era. Thus, a recent study by Rutter et al showed lower cumulative incidence of CRC in patients with UC even at the referral-based St. Mark’s Hospital in London: 2.5% after 20 years of disease, 7.6% after 30 years, and 10.8% after 40 years of follow-up[16].

Table 1 Factors associated with CRC risk.

Factors that increase CRC risk

Longer duration of colitis

Greater extent of colonic involvement

Family history of colorectal cancer

Primary sclerosing cholangitis

Young age of IBD onset (some studies)

Backwash ileitis (possibly)

Severity of histologic inflammation

History of dysplasia

Factors that decrease CRC risk

Prophylactic total proctocolectomy

Regular doctor visits

Surveillance colonoscopy

Chemoprevention

Extent of colitis is an independent risk factor for the development of CRC. The more colonic surface that is involved with colitis, the greater the CRC risk. However, different criteria exist regarding classification of the extent of colitis. Most early reports of surveillance programs used barium enema results at diagnosis as the standard for defining disease extent. A population-based study of over 3000 UC patients in Sweden examined by barium enema demonstrated that patients with proctitis had a standardized incidence ratio (SIR) of 1.7 [95% confidence interval (CI) 0.8-3.2] compared with age-matched population controls without UC, whereas those with left-sided colitis had a SIR 2.8 (95% CI 1.6-4.4), and those with pancolitis had SIR 14.8 (95% CI 11.4 to 18.9)[17]. Endoscopic and histologic evidence of inflammation are valid alternative criteria particularly in high risk patients. Very few studies have correlated CRC risk with histologic extent of disease, even though microscopic evidence of colitis is arguably a better indicator of disease extent than either endoscopic or radiographic changes. Mathy et al[18] reviewed 30 colectomy specimens and showed that dysplasia and CRC can arise in areas of microscopic colitis that are proximal to areas of gross colitis, suggesting that indeed histologic changes, even without colonoscopic alterations, might better define disease extent for the purposes of cancer risk. Backwash ileitis, defined as pancolitis with superficial involvement of the terminal ileum, has been suggested as an additional increased risk of CRC[19], but this requires additional confirmation.

IBD patients with a family history of CRC have at least a two-fold higher risk of CRC[20–22]. Both UC and Crohn’s disease patients with primary sclerosing cholangitis (PSC) are at particularly high risk for developing colorectal neoplasia[23]. In a population-based Swedish study, the cumulative incidence of CRC in UC patients with PSC was 33% at 20 years[24]. By comparing patients with UC with and without PSC, Shetty et al[25] reported an adjusted relative risk (RR) for dysplasia or cancer of 3.15 (95% CI, 1.37 to 7.27) in patients with PSC. Current practice guidelines recommend all patients with PSC not previously known to have IBD should undergo a colonoscopy to determine their status. For those patients with IBD, screening and subsequent surveillance should begin at the diagnosis of PSC, and if the patient progresses to the point at which liver transplantation is necessary, prophylactic colectomy should be considered. Young age at onset of colitis has been reported to be an independent risk factor for CRC in some[21726], but not all studies[2728]. There is insufficient evidence to support starting screening and surveillance before 8 years of disease in these patients. The severity of colitis has not been considered an independent risk factor for CRC when activity of disease is defined according to the frequency of clinical exacerbations[22]. However, a case-control study by Rutter et al showed that increased severity of inflammation, both endoscopically and histologically, correlates with increased frequency of dysplasia[29]. It is important to realize that the patient with quiescent IBD is also at increased CRC risk.

With regard to reducing CRC risk, there are two main choices: removing the colon versus conducting a lifelong program of surveillance (Table 1). Although prophylactic total proctocolectomy after 7-10 years of colitis would prevent most cancers, it would result in many colectomies that were not necessary and a substantially altered quality of life for patients. Thus cancer prevention in this patient population has focused on periodic surveillance colonoscopies. Surveillance should be viewed as a program that includes regular visits to the doctor, the use of medications to control inflammation (some of which may have chemopreventive effects, see below), and regular colonoscopies. The goal of surveillance colonoscopy is to detect neoplastic lesions before they become biologically dangerous. Thus, the detection and interpretation of dysplasia is crucial to successful surveillance.

DIAGNOSING DYSPLASIA

Macroscopic classification of dysplasia

By definition, dysplasia is unequivocal neoplasia. Despite considerable heterogeneity in appearance[30], dysplasia in IBD is often classified macroscopically as raised or flat, depending on whether it corresponds to an endoscopically visible lesion. Raised lesions, conventionally referred to by the term DALM (dysplasia associated lesion or mass)[13], can appear as polyps, bumps, plaques and velvety patches[3132]. Such lesions can blend easily with the gross inflammatory abnormalities commonly encountered in colons with IBD, making their endoscopic detection difficult even for experienced practitioners.

Unlike sporadic cancers arising from polypoid lesions, IBD-associated cancers can arise from flat dysplastic lesions. Flat dysplasia is detected microscopically in random biopsies from unremarkable mucosa. Its detection therefore depends on adequate sampling of the mucosa by the endoscopist, or more recently, by chromoendoscopy methods which highlight suspicious lesions and permit targeted biopsies (see below). If random biopsies are performed without dye spray enhancement, it has been estimated that to exclude dysplasia with a 90% certainty, 33 biopsy specimens are required, and to increase the accuracy to 95%, nearly twice the number of biopsy specimens are required[33]. Current surveillance strategies recommend annual colonoscopy with multiple biopsy specimens (4 circumferential) taken from every 10 cm of diseased colon, with additional biopsy specimens at sites of strictures or raised lesions[34]. However, questionnaire surveys have suggested that the number of biopsies taken by endoscopists in routine practice often falls short of recommended guidelines[3536].

The significance of dysplasia in endoscopically visible lesions came from studies that reported high rates of cancer when patients with such lesions underwent colectomy[3132]. Blackstone et al reported cancers in 7 of 12 DALM-bearing colons, including 5 with only mild or moderate dysplasia in the preoperative biopsies[31]. A subsequent compilation of published results from ten surveillance programs reported cancers in 17 of 40 (43%) colectomies performed because of a DALM[37]. It was concluded that DALM is an indication for colectomy irrespective of the grade of dysplasia in preoperative biopsies. While not fully appreciated at the time, the original studies of DALMs dealt exclusively with lesions that could not be removed endoscopically for microscopic examination. Thus, the significance of such lesions as an indication for surgery is similar to that of endoscopically non-resectable sporadic adenomatous polyps, which frequently harbor invasive cancer at the polyp base despite the presence of low-grade dysplasia in their more biopsy-accessible upper portions.

More recently, we have come to realize that not all types of polypoid dysplasia in patients with IBD carry the same significance. Some polyps may be adenomatous polyps unrelated to colitis and can be managed by endoscopic polypectomy like polyps in the general population[38–40]. One example is the dysplastic polyp encountered in a bowel segment that is entirely free of disease (e.g., in the proximal colon of a patient with left-sided ulcerative colitis). In such cases, one would take the precaution to biopsy the mucosa surrounding the polyp to assure the absence of microscopic disease. Similarly, a dysplastic polyp with a well-defined stalk can be regarded as a sporadic adenoma, even when encountered in a colitic region, if the mucosa lining its stalk is non-dysplastic.

Conservative management is also reasonable for dysplastic polyps that are “adenoma-like”[4142]. These polyps are endoscopically indistinguishable from sporadic sessile adenomatous polyps, i.e., discrete and ovoid or round, are completely resectable by the endoscope, and are not surrounded by flat dysplasia. Such lesions have long posed a dilemma for endoscopists who were familiar with the DALM concept but reluctant to advocate colectomy for what appeared to be innocuous lesions and possibly nothing more than fortuitous adenomas. Histology has not provided a reliable means of making this distinction in individual cases[43], since the histological features of dysplasia in the setting of IBD and in true adenomas can be virtually identical. A 1999 study from The Mount Sinai Hospital in New York reported that conservative management of a cohort of 48 UC patients with a total of 70 such polyps, including 3 with high-grade dysplasia, resulted in no adverse outcomes during a mean follow-up period of 4.1 years[41]. Similar conclusions were reached in a concurrent study from Brigham and Women’s Hospital[42] that was confirmed upon longer follow-up[44]. As a result, the burden of deciding whether a polyp qualifies as adenoma-like rests with the endoscopist. Molecular markers may ultimately afford a more objective means of making these distinctions[4546], but to date, these analyses are not applicable to routine clinical practice.

Microscopic classification of dysplasia

Gastrointestinal dysplasia is defined microscopically as replacement of the native intestinal epithelium by an unequivocally neoplastic, but noninvasive, epithelium[47]. It is synonymous with the term “intraepithelial neoplasia” used in other organ systems. A standardized classification system of dysplasia in IBD was established and divided dysplasia into five categories: negative for dysplasia, indefinite for dysplasia, low-grade dysplasia (LGD), high-grade dysplasia (HGD) or invasive cancer[47]. A further subdivision of the indefinite category includes probably negative, probably positive, unknown, however many pathologists regard this as optional.

The cellular abnormalities that define dysplasia in IBD are analogous to those characterizing neoplastic tissue in general, namely nuclear abnormalities reflecting inappropriate cellular proliferation and cytoplasmic abnormalities reflecting clonality and aberrant differentiation. The distinction between low- and high-grade dysplasia depends upon the distribution of nuclei within the cells, low-grade dysplasia being characterized by nuclei that remain confined to the basal half of the cells and high-grade by nuclei that are stratified haphazardly between the basal and apical halves[47]. Not surprisingly, pathologists are frequently confronted with biopsies that lie in a gray zone between the two categories, and some degree of subjectivity is therefore unavoidable. The diagnostic category “indefinite for dysplasia” is an acknowledgement of the difficulty pathologists face in discriminating between dysplasia and reactive epithelial changes, although experienced pathologists are usually able to discriminate between the two.

There is inconsistency among pathologists in the diagnosis of dysplasia on biopsy. In one study[48], there was only 60% agreement for a diagnosis of LGD. Similarly Lim et al[49] found that the kappa coefficient for interobserver agreement between ten pairings of five specialist gastrointestinal (GI) pathologists ranged from 0.06 to 0.39. Other studies comparing diagnoses of dysplasia among different pathologists, both prospectively and retrospectively, have concluded that levels of interobserver agreement are fair at best even among specialists in gastrointestinal pathology[50–54]. The best agreement levels tend to occur at the two extremes of negative and high-grade dysplasia and the poorest levels in the two gray zones between low-grade and high-grade dysplasia and on either side of indefinite for dysplasia. From a practical standpoint, it has been recommended that diagnoses carrying serious management implications be reviewed by at least one additional pathologist with expertise in this area[55].

BIOLOGY OF DYSPLASIA

Natural history of dysplasia

The natural history of dysplasia is a key factor contributing to the outcome and success of surveillance. The model shown in Figure 1 suggests that colitic mucosa progresses in a systematic fashion: no dysplasia, indefinite dysplasia, LGD, HGD, and finally invasive cancer. Although this is a useful paradigm that facilitates the study of cancer risk markers in IBD, it remains unclear whether dysplasia of one grade may “progress” (or “regress”) to another grade. For example, patients undergoing regular colonoscopic surveillance have developed CRC without any prior dysplasia, and it is not necessary for LGD to progress to HGD before cancer arises in the colon[3056]. This highlights the need to develop markers that are complementary to dysplasia for predicting CRC risk in IBD patients-a subject of ongoing investigation.

In the meantime, we currently rely upon the histological identification of dysplasia to make management decisions. Refinements in interpreting dysplasia based on the 1983 standardized histological criteria[47] have enabled a more accurate prediction of which patients are more likely to progress to advanced neoplasia by excluding those whose biopsies only show reactive changes secondary to inflammation. This was amply illustrated by the St. Marks group who found that the 5-year cumulative rate of progression from LGD to HGD or cancer rose from 16% to 54% once biopsies were more precisely reclassified[5357].

Dysplasia (of any grade) is associated with a risk of concurrent CRC in IBD. An early study reported 12 cases, described as unresectable single polypoid masses, collections of polyps, or plaque-like lesions, 7 were found to have adenocarcinoma upon colectomies despite multiple biopsies that had not detected it[31]. In a review of ten prospective surveillance trials, 43% of patients who underwent colectomy because of DALM had coexistent CRC, 42% (10 of 24) patients with HGD, and 16% (3 of 19) patients with LGD who underwent immediate colectomy had synchronous CRC[37]. Ullman and colleagues at The New York Mount Sinai Hospital performed a retrospective cohort analysis of 46 patients with UC who had flat LGD but did not undergo immediate colectomy. They found that 27% (3 of 11) patients who underwent colectomy within 6 months of the initial detection of flat LGD had a surprise finding of cancer or HGD[56]. More recently, Rutter et al from the St. Mark’s Hospital reported 20% of patients with LGD who proceeded to colectomy had concurrent adenocarcinoma and 39.1% who had follow-up of the LGD progressed to subsequent HGD or CRC[16].

Assuming that early colectomy is not performed, what is the subsequent rate of progression? In the case of patients with HGD, 32% were found to have CRC after some follow-up period[37]. For those with LGD, the probability of eventually progressing to HGD or CRC was 16%-29%[37]. Data from St. Mark’s Hospital indicate that the 5-year cumulative probability of progressing from LGD to HGD or cancer is 54%[53]. Strikingly similar results were obtained from The Mount Sinai Hospital, with a 5-year progression rate of 53% among 46 patients with initial flat LGD[56]. A recent follow-up analysis from the St. Marks’ group indicates slightly lower, but still substantial rates of progression[16]. Likewise, a series of 18 patients with LGD followed at the Mayo Clinic demonstrated a 33% 5-year progression rate[58]. Despite these rather similar results from three different patient populations, some authors have reported a substantially lower rate of progression. Befrits et al followed 60 patients with flat LGD from the Karolinska Institute in Sweden and found that none developed cancer and only 2 cases of progression to HGD in DALMs over a mean follow-up period of 10 years[59]. In the series by Lim et al from Leeds in the U.K., they reported that only 3/29 (10%) patients with LGD progressed to HGD or cancer after 10 years[49]. It is worth noting that in the latter two studies, the designation of LGD included specimens that were interpreted prior to the 1983 consensus guidelines, so they might have included cases with indefinite dysplasia.

Because of uncertainty of flat LGD[5660], these studies have failed to achieve consensus on proper management of flat LGD. Hence, competing options should be discussed with each patient. A patient confirmed to have multifocal flat LGD (2 or more biopsies with LGD from a single screening or surveillance examination) or repetitive flat LGD (2 or more examinations with at least a single focus of LGD), should be strongly encouraged to undergo prophylactic total proctocolectomy. Furthermore, even for patients with confirmed unifocal LGD (only 1 biopsy positive for LGD in a screening or surveillance examination) should also be offered the option of undergoing prophylactic proctocolectomy, since evidence indicates that a 5-year rate of progression to HGD or CRC in this patient group seems to be similar to that of multifocal LGD[56].

THE MANAGEMENT OF DYSPLASIA

Once a decision is made to place a patient under surveillance, it is recommended that the patient formally agree to enter such a program and is willing to comply. Patients must be made to understand the limitations of surveillance and accept the concept that despite their own cooperation, dysplasia and cancer can still arise even in the hands of skilled endoscopists and pathologists[53].

The best proof that surveillance colonoscopy effectively reduces CRC mortality would be a prospective, randomized, controlled trial in which patients with longstanding IBD would undergo colonoscopic surveillance whereas controls matched for a similar risk profile would not. However, due to ethical, financial and practical limitations, this type of study will likely never come to pass. We must therefore rely on retrospective studies for insights as to the efficacy of surveillance colonoscopy. In a retrospective study by Choi et al, patients with chronic UC who developed cancer were divided into those who had surveillance colonoscopy and those who did not. Patients undergoing surveillance colonoscopy were found to have less advanced Dukes’ stage CRC than those who did not and correspondingly had an improved 5-year survival rate (77.2% vs 36.3%, P = 0.026)[61]. However, the best evidence that colonoscopy reduces mortality from CRC in UC comes from case-control studies. In one such study, Karlen and colleagues identified 2 of 40 patients with UC and 18 of 102 controls who had undergone at least one surveillance colonoscopy who died as a result of CRC (RR = 0.29), CRC mortality was reduced by as much as 78%, although this did not reach statistical significance[62]. Another study found a similar degree of protection[22]. In these studies, a protective effect was found for individuals who had even one or two surveillance exams. There is also good evidence from prospective, albeit uncontrolled, studies of surveillance colonoscopy that in general, patients who comply with surveillance have cancers detected at earlier stages compared to those who do not comply[5363]. Of course, cancers will still arise even within a surveillance program, but in balance, the practice of surveillance is beneficial[64].

Although the gastroenterology community has put its faith in surveillance colonoscopy to prevent CRC in IBD patients, surveillance has its limitations. Previous studies have shown low rates of observer agreement for the histopathologic interpretation of biopsy specimens between general pathologists and GI pathologists, or even among expert GI pathologists[4748]. Endoscopists fail to take a sufficient number of biopsies to exclude the presence of dysplasia or cancer[33]. Unlike the dysplastic sporadic adenoma which typically assumes a discrete, polypoid shape surrounded by normal mucosa, dysplasia in the colitic colon can be flat or polypoid and is often difficult to discern. This may be particularly troublesome in a colon that is replete with inflammatory pseudopolyps. Furthermore, there is poor understanding of dysplasia amongst trained gastroenterologists. A survey of practicing gastroenterologists and senior GI fellows in the U.S. found that only 19% of respondents correctly identified dysplasia as neoplastic change[35]. Patient drop-out or non-compliance with surveillance contributes importantly to CRC mortality in IBD[4953] and this must be considered when embarking upon, or continuing, a course of surveillance in individual patients.

RECOMMENDED SURVEILLANCE STRATEGY

Despite the limitations of surveillance colonoscopy, dysplasia remains the best marker for managing cancer risk in IBD. After approximately 7-8 years of colitis, patients should undergo an “initial” surveillance colonoscopy to determine the extent of colitis and check for neoplasia[55]. The entire colon should be examined, with approximately 4 biopsies taken every 10 cm. Some experts suggest taking more biopsies in the distal rectosigmoid (e.g., approximately every 5 cm) since the distribution of neoplasia in UC still shows a distal predominance. Biopsies should be taken from flat mucosa, but if any raised or suspicious lesions are encountered, these should be removed if possible and processed in separate specimen containers (with additional biopsies taken near the base of the polyp). If a patient is experiencing moderate-severe colitis symptoms, one option is to control the inflammation medically prior to performing the examination in order to minimize difficulties with the histological interpretation of dysplasia. However, one should not defer the colonoscopy too long, since many expert pathologists now feel that they can readily interpret dysplasia even in the presence of active inflammation.

Figure 2 depicts a recommended surveillance strategy. If no dysplasia is detected, the examination should be repeated in 1-2 years. This interval derives in part from studies reporting that interval cancers can develop within 2 years after a surveillance examination[53]. If indefinite dysplasia is reported, the nature of the uncertainty should be discussed with the pathologist. If the suspicion of dysplasia is high (i.e., probably positive), repeat biopsy within 3-6 mo or less may be indicated; if low, the interval can be lengthened to every 6-12 mo. If LGD is detected in a discrete polyp that can be readily resected endoscopically and there is no flat dysplasia immediately adjacent to the polyp or elsewhere in the colon, surveillance can be continued, although the frequency of examinations can be temporarily reduced to every 3-6 mo, particularly to re-evaluate the area of polypectomy. Tattoo of the polypectomy site is advised to permit relocalization of the area on subsequent exams. If LGD is detected in flat mucosa (whether unifocal or multifocal), and is confirmed by a second expert GI pathologist, colectomy should be strongly considered. If the patient refuses, repeat surveillance exams should be undertaken within 3-6 mo or less. However, the patient should be advised that a negative subsequent examination is no assurance of safety, and that temporizing until there is histological progression to HGD or cancer as the indication for colectomy is risky. A patient in whom flat HGD or adenocarcinoma is found and confirmed by two expert pathologists should undergo colectomy unless serious co-morbidities dictate otherwise. If HGD is diagnosed in an adenoma-like polyp but it is completely removed without evidence of flat dysplasia in the adjacent mucosa or elsewhere in the colon, continued surveillance can be entertained. As with any set of recommendations, decisions should be individualized according to the situation of the patient. Hopefully, strategies for surveillance will become more refined as more knowledge of the natural history of dysplasia is obtained.

Figure 2 Suggested management scheme for dysplasia. Modified from Ref 71. With permission.

Patients who have only had small intestinal Crohn’s disease without colonic involvement are not considered to be at high risk for CRC. For patients with Crohn’s colitis, much less is known. To date, only one practice-based retrospective surveillance study has been reported in patients with Crohn’s colitis[65]. Of 259 patients with Crohn’s colitis affecting at least one-third of the colon for at least 8 years, 16% were found to have dysplasia or cancer over a 16 year period in which 663 examinations were performed, and there were no cancer deaths. While we await additional data on the subject, it seems wise to follow a UC-based surveillance strategy for patients with at least 8 years of substantial Crohn’s colitis. An important question that remains to be clarified is whether patients with dysplasia or cancer in the setting of segmental Crohn’s colitis can undergo segmental resection of the involved area or should proceed to a more extensive UC-like surgical approach. Another dilemma that is encountered with Crohn’s colitis is the management of strictures. This occurs more so than with UC. Finding a stricture in a colon of a patient with UC usually means an underlying malignancy, especially if the stricture is causing symptoms, and is located in the proximal colon[66]. However, since most strictures in Crohn’s colitis are benign, the patient can often be managed conservatively. Surveillance of such patients often requires using a narrower colonoscope or sometimes dilating a stricture to visualize the proximal mucosa[65]. Consideration should be given to adding brush cytology of strictures to regular forceps biopsies, and performing a barium enema to evaluate for colonic wall irregularity.

New endoscopic techniques and molecular screening approaches

Recent publications on chromoendoscopy have demonstrated a greater yield for dye-spray targeted biopsies compared with numerous non-targeted biopsies and thus enhance the endoscopic detection of dysplastic lesions in colitic colons. In a randomized trial by Kiesslich et al[67], intraepithelial neoplasia was more than three times as likely to be detected using chromoendoscopy compared with surveillance using nontargeted biopsies (32/84 compared with 10/81; P = 0.003). Rutter et al detected no dysplasia in 2904 non-targeted biopsies in 100 patients, but in targeted biopsies, nine dysplastic lesions were detected, seven of which were visible only with dye spraying by using indigo carmine instead of methylene blue[68]. A recent report by Ochsenkuhn et al from Munich, Germany showed a low frequency of colorectal dysplasia in patients with long-standing IBD by fluorescence colonoscopy with 5-aminolevulinic acid[69].

Other approaches worth mentioning are to examine the biopsy tissue of patients with IBD for molecular alterations. The best tested of these are aneuploidy, mutations in p53 and ras, and glycosylation abnormalities, particularly increased expression of sialyl Tn antigen (sialyl 2,6 N-acetylgalactosamine)[70]. Because the DNA shed into stool should theoretically provide a more comprehensive sampling of abnormal cells than random pinch biopsies, stool DNA testing could potentially contribute to the management of patients with long-standing IBD who are at risk for developing CRC.

Chemoprevention

Despite the relative protection afforded by surveillance colonoscopies in IBD, there are still patients who develop CRC despite seemingly optimal surveillance. This raises the issue of whether chemoprevention in the form of either medications or dietary supplements might help reduce the risk of CRC in IBD[71].

Aspirin and other NSAIDs markedly reduce the incidence of, and mortality from, sporadic CRC. Since many patients with IBD take NSAIDs in the form of 5-aminosalicylates (5-ASA), investigators have asked whether 5-ASA compounds might also be protective. Although no study to date has been performed in a prospective manner specifically to address this question, the available data suggest that this may be so[72]. If 5-ASA compounds prevent colonic neoplasia by suppressing inflammation, it follows that other anti-inflammatory medications used in IBD patients should also be protective against CRC. Although one study reported that the use of systemic steroids, and even topical steroids resulted in a significant CRC risk reduction[22], and others confirm this observation[73], steroids cannot be used long-term for chemoprevention. There appears to be no chemopreventive activity of 6-mercaptopurine or azathioprine[74].

In the setting of sporadic CRC, low folate intake has been associated with an increased risk for developing colorectal adenomas and carcinomas[7576]. Patients with chronic IBD are predisposed towards folate deficiency because of inadequate nutritional intake, excessive intestinal losses with active disease, and reduced intestinal absorption from competitive inhibition from sulfasalazine use. Results of two studies suggest a trend towards protection against CRC in folate users, although neither study demonstrated statistical significance[7778]. Nonetheless, since it is rather safe and inexpensive, folate supplementation should be considered for CRC risk reduction in patients with longstanding IBD.

In animal models of colon carcinogenesis, ursodiol inhibits carcinogenesis-an effect that may be due to the ursodiol reducing the colonic concentration of the secondary bile acid deoxycholic acid. Ursodiol also has anti-oxidant activity. A study of UC patients with PSC demonstrated that ursodiol use was strongly associated with decreased prevalence of colonic dysplasia[79]. This protective effect remained after adjusting for duration of colitis, age at onset of colitis, and sulfasalazine use. In a follow-up to the randomized, placebo-controlled trial by Pardi et al at the Mayo Clinic, 52 patients with chronic PSC and chronic UC (mean 13 years) were followed for a total of 335 person-years. Ursodiol use was associated with a significant protection against the development of dysplasia and cancer (RR = 0.26, P = 0.034)[80]. At the present time, however, we do not know whether ursodiol can prevent neoplastic progression in UC patients without PSC.

Recently, there has been interest in the role statins may play as chemopreventive agents in a variety of cancers. In a population-based case-control study of patients who had diagnosis of CRC in northern Israel between 1998-2004, statin therapy was associated with a modest reduction in CRC in the non-IBD population, but a substantial 94% risk reduction in patients with IBD was observed in a subset analysis of a small number of patients[81]. Further studies will need to verify this benefit.

Although a single retrospective cohort study suggests that therapy with 6-mercaptopurine is not chemopreventive (or carcinogenic), there remain insufficient data regarding the chemopreventive role of immunomodulators in order to make recommendations and likewise, whether patients who require immunomodulator therapy should continue their 5-ASA therapies[74].

Small bowel cancer in IBD

Most cancers of the small bowel in Crohn’s disease are adenocarcinoma, usually in the terminal ileum or jejunum. The most common clinical presentation of small bowel cancer is intestinal obstruction[82]. Other important symptoms include diarrhea, weight loss, and abdominal fistulae. These symptoms are also found in Crohn’s disease. Risk factors for developing carcinoma in small bowel segments of involved mucosa in patients with Crohn’s disease are poorly defined, although case reports document them in strictured mucosa and fistulae[83–85]. Surgery should be considered if the fistulae or stricture cannot be adequately examined, or symptoms substantially worsen. There is some suggestion that 5-ASA compounds might lower the risk of small intestinal adenocarcinoma[86].

A number of studies have demonstrated an increased risk of developing adenocarcinoma of the small intestine with small intestinal Crohn’s disease. Although the absolute number of cases of small bowel adenocarcinoma is low, because of the rarity of this cancer in the general population, the risk is approximately 10-12-fold greater than the general population[486]. In the Uppsala study[87], the investigators reported only one observed case compared with 0.3 expected cases, but the confidence interval was wide. In the Copenhagen study[88], two cases were observed vs 0.04 expected cases, a 50-fold increased occurrence. In the Tel Aviv study[89], none of the patients developed small cancer. The material was probably far too small to expect any small bowel cancer cases. In Oxford, a 10-fold increased relative risk was observed for cancer of the small intestine[90]. A population-based Swedish study revealed a significantly increased number of cancer of the small intestine (standardized morbidity ratio, 15.64; 95% CI, 4.26-40.06), however, the occurrence of colorectal cancer was not increased[91]. Another population-based Canadian study encompassing the years 1984-1997 demonstrated an increased incidence rate ratio of carcinoma of the small intestine (17.4; 95% CI, 4.16-72.9)[4].

Patients with UC who have undergone total proctocolectomy with ileal pouch anal-anastomosis (IPAA) have a very small risk of dysplasia arising within the ileal mucosa of the pouch itself. The risk is thought to be higher in patients with chronic pouchitis and associated severe villous atrophy[92], but this has not been shown in all series[93]. Indeed, a study of 160 patients who underwent biopsy a total of 222 times with an average surveillance time of 8.4 years after surgery showed that in 1800 pouch-years of surveillance, only one patient had focal LGD of the pouch[94]. The risk of neoplasia is greater in the anal transitional mucosa between the pouch and the anal canal, particularly if a cuff of rectal mucosa has been left, and if the indication for the IPAA was rectal dysplasia or cancer[95]. While there are currently no guidelines for endoscopic surveillance after an IPAA procedure, in those patients who have chronic pouchitis and severe villous atrophy or whose original indication for IPAA was dysplasia or cancer, a program of periodic endoscopy with biopsies, paying particular attention to any anal transition zone, is reasonable.

Other cancers in patients with IBD

Squamous cell carcinoma of the anus has been reported in patients with longstanding, complicated perianal Crohn’s disease[85]. Worsening perianal symptoms in such patients should warrant heightened vigilance for this tumor which often requires examination under anesthesia for adequate tissue diagnosis.

An increased risk for hepatobiliary cancers in patients with UC has been found in several[496–98] but not all[87] studies. For many of these patients, primary sclerosing cholangitis was the predisposing factor.

The risk of hematopoietic cancer in patients with IBD has been a growing concern. Early case series from The Cleveland Clinic[96] and The Mount Sinai Hospital[99], as well as other centers[100] reported an increase in leukemia in patients with UC. A recent large cohort study from Sweden[101], which included nearly 50 000 IBD patients concluded this population has a marginally increased risk of hematopoietic cancer, and in UC, lymphoma occurred as expected (SIR 1.0) but myeloid leukemia occurred significantly more often than expected (SIR 1.8). In Crohn’s disease, there was a borderline significant increased lymphoma risk (SIR 1.3), essentially confined to the first years of follow up. However, population-based studies from Denmark[97], Sweden[87] and Canada[4] have failed to substantiate any increased risk of leukemia. Likewise, although an increased number of lymphomas have been reported in some case series[99102], other series[96] and several population-based studies[87919798] do not support the notion that patients with UC or Crohn’s disease are at increased risk of lymphoma. However, a population-based study from Canada reported an increased rate of lymphoma among male patients with Crohn’s disease[4].

The risk of lymphoma or leukemia in IBD has raised concerns regarding the lymphogenic potential of immunomodulatory therapy. Following the introduction of tumor necrosis factor inhibitors in the treatment of Crohn’s disease, subsequent reports indicated an excess of malignant lymphoma among treated patients[103104] and raised fears of an iatrogenic lymphoma risk. However, these reports have also highlighted the lack of robust data on the expected occurrence of malignant lymphomas in TNF naïve (but otherwise treated) patients with IBD[105–107]. Studies examining the risk of lymphoma associated with azathioprine (AZA) and 6-mercaptopurine (6-MP) have yielded variable results. Heterogeneity in the type, dose, and duration of immunomodulatory therapy may be responsible for this discrepancy. A few studies with suboptimal dosing failed to demonstrate an increased risk of lymphoma[4108–112]. In contrast to these reports, other studies have demonstrated an increased risk of lymphoma after purine analog therapy[113–115]. In one such study, Kandiel et al reported a 4-fold-higher risk of lymphoma inpatients treated with AZA or 6-MP compared with the general population[114]. Another recent study showed a statistically significant increase in the development of malignancies among IBD patients treated with 6-MP who developed sustained leucopenia[115].

FUTURE DIRECTIONS

The future looks promising with respect to new developments in the management of cancer risk in IBD. Chromoendoscopy is likely to be used more for management, but whether the predictable increase in the yield of dysplasia will alter the overall natural history remains to be studied. In the modern era of molecular diagnostics, tissue and even stool samples of patients with IBD can be investigated for molecular alterations. For example, University of Washington investigators have demonstrated that because there is often widespread genomic instability throughout the colon of IBD patients, it may be possible to analyze rectal biopsies by DNA fingerprinting or fluorescence in situ hybridization methods[116] to identify patients at particularly high risk. The advent of technology to extract human DNA from stool and look for specific DNA mutations associated with sporadic colon carcinogenesis[117118] implies that a similar approach may also be worthwhile in IBD patients. It is anticipated that refinements in our knowledge of cancer biology, clinical practice, and molecular discovery will bring a new level of sophistication to the management of patients with longstanding IBD and lower the incidence of CRC in this high-risk population.