Chemoprevention of Colorectal Neoplasia in Ulcerative Colitis: The Effect of 6-Mercaptopurine

Transcription

1 CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2005;3: Chemoprevention of Colorectal Neoplasia in Ulcerative Colitis: The Effect of 6-Mercaptopurine SIERRA MATULA,* VICTORIA CROOG,* STEVEN ITZKOWITZ,* NOAM HARPAZ, CAROL BODIAN, SABERA HOSSAIN, and THOMAS ULLMAN* *Dr. Henry D. Janowitz Division of Gastroenterology, Department of Medicine; Division of Gastrointestinal Pathology, Department of Pathology; and Department of Biomathematical Sciences, Mount Sinai School of Medicine, New York, New York Background & Aims: Evidence suggests that mesalamine-based anti-inflammatory medicines may prevent colorectal cancer (CRC) in ulcerative colitis (UC). If mesalamine exerts its chemopreventive effect by its anti-inflammatory activity, then other medications that reduce colitis activity also should possess chemopreventive properties. Our aim was to determine the effect of the immunomodulators 6-mercaptopurine (6MP) and azathioprine (AZA) in preventing the development of dysplasia or CRC in UC. Methods: Patients with UC who underwent a surveillance colonoscopy in were identified from a gastrointestinal pathology database. A proportional hazards analysis assessing 6MP/ AZA use as a time-changing covariate was performed to evaluate the effect of 6MP/AZA on: (1) progression to any neoplasia (low-grade dysplasia, high-grade dysplasia, or CRC), and (2) progression to advanced neoplasia (high-grade dysplasia or CRC). Results: A total of 315 subjects met inclusion criteria and were followed for an average of 8 years from their first surveillance examination. There were no significant differences in rates of progression to advanced neoplasia or to any neoplasia between 6MP/AZA users and never-users by log-rank testing. The proportional hazards analysis resulted in hazard ratios of 1.06 (95% confidence interval, ) and 1.30 (95% confidence interval, ) when considering the effect of exposure to 6MP/AZA on progression to any or to advanced neoplasia, respectively. The results were unaffected by known potential confounders. Conclusions: In UC patients with no initial history of dysplasia, 6MP/AZA use appears to have little or no effect on the rate of neoplastic transformation in the colon. Importantly, the use of 6MP/AZA did not increase malignant transformation in UC. Patients with ulcerative colitis (UC) have an increased risk for colorectal cancer (CRC) when compared with the general population. 1 3 Although prophylactic colectomy is arguably the most definitive method to reduce this risk, most patients and their physicians elect to undergo a program of surveillance. This entails regular office visits and periodic colonoscopies with multiple biopsies of colonic mucosa, using the histopathologic finding of dysplasia or cancer as an indication for colectomy, and the absence of dysplasia as an indication for continued surveillance. Although there are no controlled studies showing a mortality benefit for dysplasia surveillance, some studies have suggested a survival benefit 4,5 and the practice of surveillance is endorsed in several practice guidelines. 6 8 Surveillance colonoscopy has many known limitations, 9,10 not the least important of which is the development of colorectal neoplasia despite careful observation. 11 Because of these limitations, there is growing interest in whether chemoprevention can serve as an adjunct to dysplasia surveillance. Mesalamine-based medications are the most commonly prescribed anti-inflammatory drugs for the treatment of UC, and have been shown in some studies to prevent colorectal neoplasia in this patient population. 11 Although the mechanism(s) by which mesalamine might exert its chemopreventive effect has yet to be established clearly, it seems reasonable that the chemopreventive activity of mesalamine compounds relates to their ability to reverse inflammation. 12 If this is true, we hypothesized that other anti-inflammatory medications used to treat UC also might be chemopreventive. The purine analog immunomodulators, 6-mercaptopurine (6MP) and azathioprine (AZA), are among the most commonly prescribed drugs for the treatment of UC. Although their specific mode of action continues to be investigated, 13 they ultimately reduce colonic inflammation and are effective at maintaining remission in UC To date, only a few studies have examined the role of 6MP or AZA on the progression of neoplasia in UC. Of those that have addressed this question, no Abbreviations used in this paper: AZA, azathioprine; CRC, colorectal cancer; HGD, high-grade dysplasia; IBD, inflammatory bowel disease; LGD, low-grade dysplasia; 6MP, 6-mercaptopurine; NoD, no dysplasia; UC, ulcerative colitis by the American Gastroenterological Association /05/$30.00 PII: /S (05)00738-X

2 1016 MATULA ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 3, No. 10 chemopreventive role for 6MP was found However, there are several important limitations to these studies. First, in 3 of these studies, 6MP was not the primary variable under investigation, 17,18,21 and in the other 2 studies, colorectal cancer was not the primary end point. 19,20 Second, one of the reports only studied UC patients who had primary sclerosing cholangitis, 20 a group considered to be at particularly high risk for developing colorectal neoplasia. 11 Third, none of these studies accounted for dose changes over time. Finally, it is not clear whether patients studied were initially free of colorectal neoplasia. These methodologic limitations indicate that any potential chemopreventive role of 6MP/AZA remains an open question. To address this issue in a more comprehensive fashion, we performed a retrospective cohort study that takes into account the dose and duration of 6MP/AZA use to investigate whether these medications can prevent the development of dysplasia or CRC in patients with UC who were dysplasia-free at the time surveillance was initiated. Materials and Methods Cohort Identification After approval by the Mount Sinai School of Medicine institutional review board and in accordance with the Health Insurance Portability and Accountability Act of 1996 (HIPAA) guidelines, we ran a query of the Mount Sinai Hospital Gastrointestinal Pathology Registry and identified all UC patients who underwent at least 1 surveillance colonoscopy between January 1996 and December This time period was selected to allow sufficient follow-up time for patients to develop colorectal neoplasia. Records then were reviewed dating back to the time of their first surveillance examination by a gastroenterologist at our institution, whether in 1996, 1997, or earlier. All histopathologic diagnoses included in the study interval were rendered according to the criteria of the Inflammatory Bowel Disease (IBD) Morphology Study Group. 22 Patients were included only if they had UC for at least 7 years and had a finding of no dysplasia (NoD) at their initial surveillance colonoscopy. Recommendations from a number of different sources suggest initiating dysplasia surveillance at 8 years of UC. We chose 7 years of UC as a cut-off point so that examinations performed to rule out dysplasia in the 8th year of disease would be included. Patients were excluded if they had a finding of indefinite for dysplasia, polypoid or flat dysplasia, or CRC before or during their initial colonoscopy. Patients also were excluded for prior colonic surgery of a segment of ulcerative colitis, a diagnosis of Crohn s disease, inadequate or absent information on 6MP exposure, inaccessible medical records, or lack of endoscopic or surgical follow-up evaluation after their initial colonoscopy. Clinical Information Dating back to their first surveillance examination by a Mount Sinai gastroenterologist, a single reviewer (V.C.) abstracted demographic data for the creation of a UC Surveillance Database. All patients medical records were abstracted for demographic information, histologic and endoscopic findings at all colonoscopies and surgeries, and the history or presence of extraintestinal manifestations of UC. The anatomic extent of disease was classified based on histologic findings as follows: extensive colitis, histologic evidence of colitis proximal to the splenic flexure; left-sided colitis, disease proximal to the sigmoid colon but not proximal to the splenic flexure; and proctosigmoiditis, disease confined to the sigmoid colon and rectum. Pathology reports were reviewed for all surveillance colonoscopies from the time of initial surveillance at Mount Sinai until the end point (see later) was reached. The degree of dysplasia, if any, was recorded, as was the number of biopsy specimens reviewed. Endoscopy notes were reviewed, as were any surgical reports, and both were correlated with the corresponding pathology report. Exposure to medications (mesalamine-based agents, purine analogs, cyclosporine, corticosteroids, folate, and any experimental medications used to treat active UC or to maintain its remission) was recorded as positive if a patient had at least 3 months of consistent use of the medication beginning with the time of his or her first surveillance examination. A second reviewer (S.M.) revisited the medical records to abstract detailed information pertaining to 6MP/AZA use from the time of initial NoD to the time of end point. Patients were considered to have used 6MP/AZA if they had at least 3 months of consistent use of the medication. In addition, to permit analysis of data using the time-changing covariate method (see later), dates of initiation of medication use, dose and/or formulation changes, and cessation of use were recorded. On entry into a Microsoft Access database (Microsoft, Inc, Redmond, WA), dosages were converted into 6MP equivalents for analysis by multiplying azathioprine and Imuran (Prometheus Laboratories, San Diego, CA) doses by.5, and keeping 6MP and Purinethol (GlaxoSmithKline, Inc, Research Triangle Park, NC) doses unchanged. Definitions and End Points Patients entered the study at the time of their first surveillance colonoscopy of record at Mount Sinai at which NoD was found. We defined the date of this initial examination as time 0 for each patient. The development of low-grade dysplasia (LGD), high-grade dysplasia (HGD), or CRC was defined collectively as any neoplasia. HGD or CRC was defined as advanced neoplasia. All patients were followed-up from time 0 until they reached one of the following end points: (1) advanced neoplasia, (2) colectomy for reasons other than HGD or CRC, or (3) their last or most recent colonoscopy on record. A finding of LGD was included for measuring any neoplasia as an outcome, but patients were not censored for a finding of LGD unless they also had 1 of the 3 end points just listed. Follow-up time was the duration of time between time 0 and the

3 October 2005 EFFECT OF 6MP 1017 Figure 1. Concept of time-changing covariates. At time 0, all patients begin surveillance and the daily dose of 6MP is recorded from this point forward for each patient. Each time a patient in the cohort develops an end point of interest (eg, advanced neoplasia), an event is registered by the model. At each such event, the mean daily dose of 6MP is calculated from time 0 to the time of the event for each patient in the cohort. In this figure, patient A develops HGD, and event A is registered. Each patient s mean daily dose of 6MP is calculated from time 0 to the time of event A. Patient A then drops out of the analysis. The next event in the cohort is the development of colon cancer in patient B. The mean daily dose of 6MP is calculated for each patient still under surveillance, including patient B, from time 0 to the time of event B. Patient B then drops out of the analysis, and the process continues for all remaining patients in the cohort. end point. All diagnoses were rendered by expert staff members of Mount Sinai s Division of Gastrointestinal Pathology and confirmed in intradepartmental conference by its director, a gastrointestinal pathologist with 2 decades of experience in the evaluation of IBD-associated dysplasia (N.H.). Data Analysis Means and SDs were calculated for all continuous variables and proportions were calculated for all binary variables. Crude proportions are shown for descriptive purposes. The 2 outcomes of interest in this study were progression to any neoplasia and progression to advanced neoplasia. Life-table analysis was used to describe the progression and to test for the influence of covariates known at entry, such as duration of disease, extent of disease, primary sclerosing cholangitis, and age at onset of disease. The influence of the primary variable of interest, purine analog use, was not always constant during follow-up evaluation. To account for this variable s changing status during follow-up evaluation and to allow for assessment of its cumulative effect, it was analyzed as a time-changing covariate in a proportional hazards model (Cox regression), with values updated as follow-up evaluation proceeded (Figure 1). The timechanging approach additionally was applied to test the influence of a known predictor of outcome: histologic status (NoD compared with LGD or indefinite for dysplasia). With the exception of the handling of purine analog exposure as a time-changing covariate, there are no differences between this type of proportional-hazards modeling and any other similar model that examines the influence of independent variables on a time-dependent outcome. The influence of purine analog use and histologic findings were considered separately and jointly. In separate analyses, purine analog use was characterized first by comparing those who ever used these medications with nonusers. Because some patients were on 6MP or AZA for a short period during their surveillance, usage also was assessed as a binary variable at the threshold of 25 mg/day of 6MP equivalents, thereby combining those with minimal exposure and those who never took these medications. Finally, the possibility of a dose effect over time was assessed by considering the cumulative average dose of purine analogs as a continuous variable. Histologic findings were considered as negative or positive, and analyzed for association with progression to advanced neoplasia. A patient was classified as having a positive histologic history at the time of the first diagnosis of indefinite dysplasia or flat LGD. All analyses were performed using SAS software (SAS, Cary, NC) on a personal computer. Results Of the 543 patients identified who underwent surveillance colonoscopy in the calendar year , 315 patients ultimately met our inclusion criteria. Of these, 96 (30.5%) were exposed to 6MP or AZA (Table 1). The 6MP/AZA-exposed and -unexposed groups were very similar with respect to sex, age, duration and extent of disease, and the presence of primary sclerosing cholangitis. With respect to follow-up evaluation, the 6MP/AZA-exposed and -unexposed groups had comparable surveillance patterns, with approximately 8 years of follow-up evaluation, an average of 6 colonoscopies with examinations performed every 21 months, and a similar number of biopsy specimens per examination available for review. Of those patients who had taken 6MP or AZA, the average duration of use was 7.4 years and the average dose while on the medication was mg/day of 6MP equivalents. When analyzed as simple proportions of progression to the given end points, before accounting for dose and duration of medication use, the 2 groups had similar outcomes (Table 2). Approximately 16% of those exposed to 6MP/AZA and 18% of those not exposed progressed to any neoplasia; 5% of each group ultimately developed advanced neoplasia. There was a somewhat higher rate of colectomy for reasons other than HGD or CRC in the 6MP/AZA-exposed group (11.5% vs 5.5%, P.06). By univariate analysis, none of the variables tested reached a P value of.15 or less threshold for inclusion into the multivariable model. The variables tested were extent of disease, age at diagnosis, primary sclerosing cholangitis, male sex, and history of use of the following: mesalamine, cyclosporine, corticosteroids, or folate during the patient s disease history. Subsequent analysis by the proportional hazards model assessed 6MP/AZA use in 3 ways. First, we tracked

4 1018 MATULA ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 3, No. 10 Table 1. Cohort and Surveillance Characteristics Total 6MP/AZA 6MP/AZA a Patient characteristics N(%) (30.5) 219 (69.5) Male, N (%) 169 (54) 53 (55) 116 (53) UC duration, y b Age at diagnosis, y b Extensive colitis, N (%) 251 (80) 77 (79) 174 (80) Primary sclerosing cholangitis, N (%) 4 (1) 0 (0) 4 (2) Surveillance patterns Total number of examinations Examinations per patient b Examination interval, mo b Biopsy specimens per examination b MP use Average duration, y b Average dose, mg b Mesalamine, n (%) 248 (79) 83 (86) 165 (75) Glucocorticoids, n (%) 183 (58) 81 (84) 102 (47) Cyclosporine, n (%) 13 (4) 11 (11) 2 (1) Folate, n (%) 101 (32) 36 (38) 65 (30) a 6MP/AZA status by end of follow-up period. b Mean (SD). patients, switching them from nonuser to user status at the time of their first exposure to 6MP. We found neither a beneficial nor harmful effect for progression to any neoplasia (hazard ratio, 1.06; 95% confidence interval [CI], ) or to advanced neoplasia (hazard ratio, 1.30; 95% CI, ) (Table 3). Knowing that some patients were only on 6MP/AZA for part of their surveillance, we calculated an average use of 6MP/AZA use over time for each patient and assessed the impact of exposure at a threshold of 25 mg/day mean exposure (using the time-changing covariate model) as a cut-off between users and a group composed of rare users and nonusers. When analyzed at this threshold in the proportional-hazards model, we again found no appreciable protection or harm for 6MP/AZA use. Finally, we analyzed the mean 6MP/AZA dose as a continuous variable to allow for the possibility of a cumulative dose effect. Again, there was no appreciable difference, with a calculated hazard ratio for progression to advanced neoplasia of 1.95 (95% CI, ) for each 25 mg/day increase in 6MP/AZA use (Table 3). As a positive control for the time-changing covariate method of statistical modeling, we analyzed the effect of finding flat early grade dysplasia (indefinite for dysplasia or LGD) at surveillance on subsequent progression to advanced neoplasia. The model correctly identified these higher-risk patients as being more likely to progress to more advanced stages of neoplasia (Table 3). When controlling for a finding of early grade dysplasia in multivariate analysis, the estimate of the effect of 6MP was not altered meaningfully. Discussion The purine analog immunodulators are effective and of acceptable safety in the induction and maintenance of remission in UC As such, these agents successfully reduce inflammation in the colon. Unlike mesalamine, the question of whether these medications might act as chemopreventive agents has never been explored rigorously. A handful of studies have reported negative findings with respect to a possible role for 6MP/AZA in CRC chemoprevention in IBD. However, for the most part, 6MP/AZA use was not the primary variable studied, or CRC was not the primary end point analyzed. Connell et al 17 studied 755 patients with IBD who had taken 2 mg/kg/day of AZA between 1962 and Exposure ranged from 2 days to 15 years. CRC rate was one of several end points measured. Paradoxically, in an overall analysis of the data, they found significantly more observed cases of cancer than expected. However, when they analyzed only patients with UC in a nested casecontrol study, the observed-expected ratio was not statistically significant. Fraser et al 18 examined the records Table 2. Crude End Point Rates Total 6MP/AZA 6MP/AZA N Time to end point, y Any neoplasia, N (%) 54 (17) 15 (16) 39 (18) Advanced neoplasia, N (%) 16 (5) 5 (5) 11 (5) Colectomy, no HGD/CRC, N (%) 23 (7.3) 11 (11.5) 12 (5.5)

5 October 2005 EFFECT OF 6MP 1019 Table 3. Proportional Hazards Analysis 6MP/AZA exposure Any exposure 25 mg/day Average daily dose Control for model find/flgd Any neoplasia, HR (CI) 1.06 ( ) 1.00 ( ) 1.21 ( ) Advanced neoplasia, HR (CI) 1.3 ( ) 1.49 ( ) 1.95 ( ) 3.5 ( ) find/flgd, flat indefinite for dysplasia or flat low-grade dysplasia; HR (CI), hazard ratio and 95% confidence interval. of 2204 patients with IBD, 626 of whom had a history of AZA use. Duration was reported in ranges from 0 to more than 60 months, but the dose of medication was not noted. AZA did not show any effect on the rates of all cancers measured, including CRC and HGD. This held true even when UC patients were analyzed separately. In a study designed to determine the effect of folate deficiency on the risk for dysplasia or CRC in 98 patients with UC, Lashner et al 19 reported that the use of AZA for at least 6 months had no effect (relative risk, 1.12; 95% CI, ). Again, dose and duration of use were not assessed. In a study of UC patients who also had primary sclerosing cholangitis, Tung et al 20 showed that ursodiol use prevented the development of colorectal dysplasia and cancer. As a secondary variable, they looked at the effect of other medications, such as AZA, although neither the dose nor the duration of AZA use was described. When comparing patients who developed dysplasia with those who did not, the use of AZA had no significant protection or harm in this higher-risk group of UC patients, with a calculated odds ratio of.68 (95% CI, ). In a more recent study by Rutter et al 21 in which the primary variable tested was the effect of active inflammation on CRC risk in UC, AZA/6MP exposure was analyzed as a secondary variable. This case-control study involved 68 cases of colitis-related neoplasia and 136 matched controls. AZA was analyzed with regard to ranges of duration of use with categories of less than 1 year, 1 5 years, and more than 5 years. Univariate analysis showed a nonsignificant trend toward a protective effect of AZA use for 1 5 years, or greater than 5 years, with odds ratios of.34 (95% CI, ) and.73 (95% CI, ), respectively. This study looked at 6MP as the primary variable and used advanced neoplasia as a primary end point. It started with a population of UC patients who were documented to be free of dysplasia, to account for changes in dose and duration of 6MP use over time, and also to consider the intensity of colonoscopic surveillance. In our cohort of 315 patients with an initial finding of no dysplasia, the purine analog immunomodulators showed no chemopreventive effect. Although often there is fear that immunomodulator drugs might enhance the development of cancer in IBD (especially lymphoma), we found that the use of these agents did not appear to promote carcinogenesis of colonic epithelium. The retrospective nature of this study is a potential limitation. Although it would be ideal to study this question prospectively, such a venture becomes unfeasible once consideration is given to the fact that the neoplastic progression rate in a population initially without dysplasia is quite low, with events occurring after many years of disease. It might be worth investigating the chemopreventive effect of immunomodulators using a study population of patients with early grades of dysplasia (indefinite for dysplasia or LGD) who would be expected to show higher rates of neoplastic progression, thereby allowing for a more clear elucidation of differences between exposed and unexposed groups. Another potential limitation to this study is the possibility of selection bias. The clinical decision to initiate purine analog therapy usually is made in response to disease severity and the desire to minimize corticosteroid use. Users and nonusers of 6MP/AZA likely constitute patients with different disease severity, as evidenced by the different colectomy rates in our study. Although Rutter et al 21 showed that histologic and endoscopic evidence of active inflammation may increase an individual s risk for developing CRC, previous work has shown that more frequent colitis exacerbations may lessen the risk for developing CRC. 23 Whether there is an inherent difference in risk among more severely affected UC patients, and whether this may have biased our results, remains unknown. The study of inflammation as a risk factor is complicated additionally by the fact that inflammation often is regional or patchy, it can change over time, and there have been no standardized scoring systems for assigning histologic or endoscopic degrees of inflammation. We are in the process of pursuing the issue of histologic inflammation as a risk factor for CRC in UC. Although the low numbers of biopsies per examination that were performed in our patients may call in to question the effectiveness of nonprotocol-based surveil-

6 1020 MATULA ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 3, No. 10 lance, equivalence in biopsies per examination between users and nonusers of purine analogs virtually eliminates the possibility that this may have biased our results, and makes the examination of both any neoplasia and advanced neoplasia worthy end points. The strengths of a cohort study design are that all patients begin with a similar risk profile, being free of dysplasia at initial surveillance examination, and medication exposure can be assessed as the primary distinguishing trait between these cohorts. As their histories are reviewed over time, independent risk factors for progression or independent protective variables ideally could be identified with this study design. However, the low rate of progression to neoplasia in this lower-risk group may have limited our ability to find a clinically important association between purine analog use and the risk for colorectal neoplasia. Indeed, although the hazard ratios for all measurements hovered close to 1, the confidence intervals were too wide for any definitive determination to be made. Assuming a similar event rate and a similar duration of follow-up evaluation, we would have needed approximately 2000 patients without dysplasia at onset to find significant protection from advanced neoplasia for the hazard ratios we found comparing any exposure with no exposure. Given that 6MP showed no chemopreventive effect in this small study, we conclude for the time being that not all anti-inflammatory medications are chemopreventive in UC. Therefore, any chemopreventive activity of the mesalamine-based medications may work by a unique mechanism, such as the induction of apoptosis in dysplastic cells, the accumulation of cells in specific stages of mitosis, 26 or improved DNA replication fidelity. It also is possible that the mesalamines are chemopreventive via drug-specific inhibition of proinflammatory cytokines, 27 drug-specific scavenging activity of reactive oxygen species, 28 or activation of peroxisome proliferator-activated receptor. 29 By contrast, 6MP/ AZA appear to exert their anti-inflammatory effect by inducing apoptosis of activated T cells. 30 Regardless of the specific mechanism of mesalaminebased medications, the end result still is the suppression of inflammation, a result shared by 6MP. It is hoped that as more is learned about the role of inflammation in colon carcinogenesis, new hypotheses will emerge regarding the role of chemoprevention in UC. References 1. Eaden JA, Abrams KR, Mayberry JF. The risk of colorectal cancer in ulcerative colitis: a meta-analysis. Gut 2001;48: Ekbom A, Helmick C, Zack M, et al. Ulcerative colitis and colorectal cancer. A population-based study. N Engl J Med 1990;323: Itzkowitz SH. Cancer prevention in patients with inflammatory bowel disease. Gastroenterol Clin North Am 2002;31: Karlen P, Kornfeld D, Brostrom O, et al. Is colonoscopic surveillance reducing colorectal cancer mortality in ulcerative colitis? A population based case control study. Gut 1998;42: Choi PM, Nugent FW, Schoetz DJ Jr, et al. Colonoscopic surveillance reduces mortality from colorectal cancer in ulcerative colitis. Gastroenterology 1993;105: Kornbluth A, Sachar DB. Ulcerative colitis practice guidelines in adults (update): American College of Gastroenterology, Practice Parameters Committee. Am J Gastroenterol 2004;99: Eaden JA, Mayberry JF. Guidelines for screening and surveillance of asymptomatic colorectal cancer in patients with inflammatory bowel disease. Gut 2002;51(Suppl 5):V10 V Winawer S, Fletcher R, Rex D, et al. Colorectal cancer screening and surveillance: clinical guidelines and rationale update based on new evidence. Gastroenterology 2003;124: Shanahan F, Quera R. CON: surveillance for ulcerative colitisassociated cancer: time to change the endoscopy and the microscopy. Am J Gastroenterol 2004;99: Eaden J, Abrams K, McKay H, et al. Inter-observer variation between general and specialist gastrointestinal pathologists when grading dysplasia in ulcerative colitis. J Pathol 2001;194: Itzkowitz SH, Harpaz N. Diagnosis and management of dysplasia in patients with inflammatory bowel diseases. Gastroenterology 2004;126: Itzkowitz SH, Yio X. Inflammation and cancer IV. Colorectal cancer in inflammatory bowel disease: the role of inflammation. Am J Physiol 2004;287:G7 G Dubinsky MC. Azathioprine, 6-mercaptopurine in inflammatory bowel disease: pharmacology, efficacy, and safety. Clin Gastroenterol Hepatol 2004;2: George J, Present DH, Pou R, et al. The long-term outcome of ulcerative colitis treated with 6-mercaptopurine. Am J Gastroenterol 1996;91: Hawthorne AB, Logan RF, Hawkey CJ, et al. Randomised controlled trial of azathioprine withdrawal in ulcerative colitis. BMJ 1992;305: Rosenberg JL, Wall AJ, Levin B, et al. A controlled trial of azathioprine in the management of chronic ulcerative colitis. Gastroenterology 1975;69: Connell WR, Kamm MA, Dickson M, et al. Long-term neoplasia risk after azathioprine treatment in inflammatory bowel disease. Lancet 1994;343: Fraser AG, Orchard TR, Robinson EM, et al. Long-term risk of malignancy after treatment of inflammatory bowel disease with azathioprine. Aliment Pharmacol Ther 2002;16: Lashner BA, Heidenreich PA, Su GL, et al. Effect of folate supplementation on the incidence of dysplasia and cancer in chronic ulcerative colitis. A case-control study. Gastroenterology 1989; 97: Tung BY, Emond MJ, Haggitt RC, et al. Ursodiol use is associated with lower prevalence of colonic neoplasia in patients with ulcerative colitis and primary sclerosing cholangitis. Ann Intern Med 2001;134: Rutter M, Saunders B, Wilkinson K, et al. Severity of inflammation is a risk factor for colorectal neoplasia in ulcerative colitis. Gastroenterology 2004;126: Riddell RH, Goldman H, Ransohoff DF, et al. Dysplasia in inflammatory bowel disease: standardized classification with provisional clinical applications. Hum Pathol 1983;14: Pinczowski D, Ekbom A, Baron J, et al. Risk factors for colorectal cancer in patients with ulcerative colitis: a case-control study. Gastroenterology 1994;107:

7 October 2005 EFFECT OF 6MP Bus PJ, Nagtegaal ID, Verspaget HW, et al. Mesalazine-induced apoptosis of colorectal cancer: on the verge of a new chemopreventive era? Aliment Pharmacol Ther 1999;13: Allgayer H. Review article: mechanisms of action of mesalazine in preventing colorectal carcinoma in inflammatory bowel disease. Aliment Pharmacol Ther 2003;18(Suppl 2): Reinacher-Schick A, Schoeneck A, Graeven U, et al. Mesalazine causes a mitotic arrest and induces caspase-dependent apoptosis in colon carcinoma cells. Carcinogenesis 2003;24: Bantel H, Berg C, Vieth M, et al. Mesalazine inhibits activation of transcription factor NF-kappaB in inflamed mucosa of patients with ulcerative colitis. Am J Gastroenterol 2000;95: McKenzie SM, Doe WF, Buffinton GD. 5-aminosalicylic acid prevents oxidant mediated damage of glyceraldehyde-3-phosphate dehydrogenase in colon epithelial cells. Gut 1999;44: Rousseaux C, Lefebvre B, Dubuquoy L, et al. Intestinal antiinflammatory effect of 5-aminosalicylic acid is dependent on peroxisome proliferator-activated receptor-(gamma). J Exp Med 2005;201: Tiede I, Fritz G, Strand S, et al. CD28-dependent Rac1 activation is the molecular target of azathioprine in primary human CD4 T lymphocytes. J Clin Invest 2003;111: Address requests for reprints to: Thomas A. Ullman, MD, GI Division, Box 1069, Mount Sinai School of Medicine, One Gustave Levy Place, New York, New York thomas.ullman@ msnyuhealth.org; fax: (212) Supported by grants from the Burrill B. Crohn Research Foundation, the American Society for Gastrointestinal Endoscopy, the American College of Gastroenterology, and the Doris Duke Foundation. The authors thank the many Mount Sinai gastroenterologists and surgeons who permitted access to patient medical records.