Patients with primary sclerosing cholangitis (PSC) and inflammatory bowel disease (IBD) have a high risk of developing colorectal cancer and dysplasia. Ursodeoxycholic acid (UDCA) has been suggested to have chemopreventive effects on the development of colorectal cancer and dysplasia but long-term data and larger trials are lacking.
To evaluate the effect of high dose (17–23 mg/kg/day) UDCA on colorectal neoplasia in a cohort of patients with PSC and IBD.
From our previous 5-year randomised controlled trial of UDCA vs. placebo in PSC, we performed a follow-up of 98 patients with concomitant IBD from entry of the trial 1996–1997 until 2009 for development of colorectal cancer or dysplasia.
The total follow-up time was 760 person-years. Dysplasia/cancer-free survival was compared between placebo- (n = 50) and UDCA-treated (n = 48) patients. There was a similar frequency of dysplasia or cancer after 5 years between patients originally assigned to UDCA or placebo (13% vs. 16%) and no difference in dysplasia/cancer-free survival (P =0.46, log rank test). At the end of 2009 no difference in cancer-free survival was detected, 30% of the placebo patients compared with 27% of UDCA patients had developed colorectal cancer or dysplasia.
Long-term high dose ursodeoxycholic acid does not prevent colorectal cancer or dysplasia in patients with primary sclerosing cholangitis-associated inflammatory bowel disease.
In Scandinavia about 80% of the patients with primary sclerosing cholangitis (PSC) have an associated inflammatory bowel disease (IBD). IBD in PSC patients is characterised by a mild pancolonic inflammation, rectal sparing and backwash ileitis. Despite the usually mild course of the IBD in PSC, these patients carry almost five times the risk of colorectal dysplasia or cancer compared to patients with IBD alone. The reason for the increased risk of developing colorectal neoplasia in PSC patients is obscure. Patients with cholestatic liver disease have decreased bile acid secretion and a relatively high proportion of secondary bile acids. It has been speculated that the secondary bile acids play a role in the colorectal carcinogenesis in PSC. The fact that colorectal dysplasia and cancer in PSC are often right-sided where the bile acid concentration is highest supports this theory.[5, 6]
During the 1990s, ursodeoxycholic acid (UDCA) was introduced as a drug for the treatment of cholestatic liver diseases. It showed promising effects on survival and symptoms in primary biliary cirrhosis (PBC) and was later tried in PSC. However, in contrast to the effect in PBC no convincing treatment benefit has been shown in PSC. Although UDCA was shown to improve liver enzyme tests in some patients, it did not influence survival.[7, 8] UDCA was initially tested at a dose of 13–15 mg/kg/day and the drug was considered safe. The Scandinavian study with a higher drug dose of 17–23 mg/kg/day was not able to show effect on survival, but this study was underpowered. In a recently published randomised double blind controlled trial of very high dose UDCA (28–30 mg/kg/day) the drug was associated with higher rates of serious adverse events.
Theoretically, UDCA may exert its protective effect on colonic mucosa by reducing the amount of secondary bile acids. A few small trials have studied the chemopreventive properties of UDCA on colorectal dysplasia and cancer in PSC patients with IBD.[10-12] Based on these studies, the American Gastroenterological Association (AGA) and the European Crohn′s and Colitis Organization (ECCO) recommend treatment with UDCA for patients with PSC and longstanding IBD.[13, 14] Recently the American Association for the Study of Liver Diseases (AASLD) discourages the use of UDCA in PSC based on the negative outcomes in UDCA-treated patients in the study of very high dose UDCA. Long-term data and larger randomised trials regarding UDCA′s chemopreventive effect on the colorectal mucosa in PSC are still lacking, and UDCA remains controversial as a chemopreventive agent.
In this cohort study, based on a previous 5-year randomised controlled trial with UDCA treatment at a dose of 17–23 mg/kg/day, we investigate the long-term effects of UDCA on the development of colorectal cancer and dysplasia in PSC.
Materials and Methods
Between 1996 and 2001 we conducted a 5-year randomised, double blind, placebo controlled multicentre trial that aimed to investigate the effect of UDCA on survival without liver transplantation in Scandinavian PSC patients. A total of 110 patients were randomised to UDCA and 109 to placebo. Twenty-one patients were excluded because they did not come to any follow-up appointments or never started taking the capsules. Thus, 97 treated and 101 placebo patients remained and were included in the trial.
In the present study we included all the patients from the original trial who had a diagnosis of IBD. Among the 198 patients, 168 patients had a diagnosis of IBD [ulcerative colitis (UC), Crohn′s disease (CD) or indeterminate colitis (IC)]. After exclusion of 50 patients who had undergone proctocolectomy before study entry (20 due to cancer or dysplasia), five patients who had colorectal dysplasia or cancer before or within 6 months of study entry and 15 patients who were not included in colonoscopic surveillance programmes, 98 patients remained who were suitable for analysis. Of these 98 patients, 48 had been randomised to UDCA and 50 to placebo. The database from the previous UDCA trial was used to collect clinical data. At the closure of the original trial the responsible investigators provided data on IBD, including family history of colorectal cancer, medical treatment, smoking, surveillance colonoscopies, development of colorectal dysplasia or cancer and colorectal surgery.
In 2009 we did a follow-up of all patients included in the original trial and collected information from the investigators regarding UDCA and other medical treatment after the end of the trial, survival data, prevalence of colorectal cancer or dysplasia and surgery performed during and after the randomised trial. All investigators were asked to fill in a specific protocol and to provide information on clinical characteristics including all colonoscopies with histological data. Follow-up was censored at time of first dysplasia or cancer, last colonoscopy, colectomy or death. Of the 98 patients included in our study, we were able to trace 77 patients for follow-up to 2009, (Figure 1).
Descriptive statistics was used to characterise the data. Cancer-free survival was assessed by using a Kaplan–Meier model with a log rank test using colorectal cancer (CRC), high-grade (HGD) or low-grade dysplasia (LGD), as primary endpoints. Lesions classified as indefinite dysplasia (IND) were not taken into account and were classified as nondysplasia. Covariates considered for prediction of cancer and dysplasia were assessed by Fisher's exact test for binomial data and 2-sample t-test for continuous variables. Colectomy and death were used as censoring variables. All statistical analyses were made using sas system for Windows version 9.1 and Stat view. P-values < 0.05 were considered significant.
The regional ethical review board in Stockholm, Sweden approved this study.
The baseline characteristics of the 98 patients are described in Table 1. The mean age at inclusion if the trial in the UDCA-treated group was 39 years (range, 18–67) and in the placebo group 43 years (range, 22–67). The two groups were well-balanced regarding, gender, age at onset of IBD, type of IBD and colonic surveillance. The intensity of colonoscopic surveillance during the 5-year trial was similar in the UDCA- and placebo-treated patients in terms of numbers of colonoscopies performed, with an average of 2.3 (range 1–7) colonoscopies in the UDCA group and of 1.8 (range 1–8) in the placebo group. The clinical routine at most centres was to take at least 20 biopsies at 10 different locations. In 83 patients the colitis was categorised as extensive, four patients had distal disease and two patients had proctitis only. In nine cases the extent of the colonic disease was not reported. The frequency and indication for colectomy were similar in the two groups. Information concerning family history of colorectal cancer was available in 73% (72/98). One patient in the UDCA group had a first-degree relative and one in the placebo group a second-degree relative with colorectal cancer.
Table 1. Clinical characteristics of study patients included in the Scandinavian UDCA trial 1996–2001
Five patients died during the first 5 years of the study, two in the UDCA group (cholangiocarcinoma in both patients) and three (two due to cholangiocarcinoma, one liver failure) in the placebo group. At the closure of the original trial 13% (n = 6) of the patients in the UDCA group compared to 16% (n = 8) of the patients in the placebo group reached the combined endpoint of dysplasia or CRC. In each group there were five patients with LGD and one patient with CRC. HGD was found in none of the patients in the UDCA group vs. two in the placebo group.
Kaplan–Meier curves for cancer/dysplasia-free survival during the original study are displayed in Figure 2. There was no significant difference between the two groups during the randomised study period of 5 years, P =0.46 (log rank test).
Follow-up in 2009
After the closure of the original trial, 21 of the 37 UDCA patients available for follow-up (57%) continued the treatment. In the placebo group 23 of 40 patients (58%) had received treatment with UDCA after 2001, (P =0.94). The reasons for continuing or ending treatment with UDCA are unknown and were decided by the local investigators. In the UDCA group 19 of the 21 patients continued on high dose and two patients received a lower dose than 17 mg/kg/day. In the placebo group 22 of 23 patients were put on high dose UDCA, in one patient the dose was unknown.
Frequency of cancer or dysplasia in patients available for follow-up in 2009 (total follow-up time 760 person-years) is described in Table 2. At the end of 2009, additional eight patients had died, three in the placebo and five in the UDCA group. In total, 28 (29%) of all the patients developed dysplasia or cancer, 13 (27%) in the UDCA group and 15 (30%) in the placebo group.
Table 2. Cancer and dysplasia at follow-up 2009, n (%)
Number and (percentages) if not stated otherwise. Patients are counted once, highest grade of dysplasia.
11 patients lost to follow-up.
10 patients lost to follow-up.
Cancer or dysplasia
Cancer or HGD
Indication for colectomy
No of colonoscopies, median (range)
There was no difference in dysplasia and cancer-free survival between patients initially assigned to UDCA or placebo (P =0.73), Figure 3.
Since low-grade dysplasia does not always progress to HGD or cancer in IBD patients, we performed an additional analysis where only cancer and HGD were considered events. In this analysis we found a lower prevalence of colorectal cancer and dysplasia in patients initially assigned to UDCA but the difference did not reach statistical significance, (P =0.07), (log rank).
Table 3 shows possible covariates associated with the development of colorectal neoplasia in PSC-associated IBD. No differences were detected regarding age at onset of IBD or IBD duration; the only parameter associated to dysplasia or cancer was the number of colonoscopies.
Table 3. Covariates considered for prediction of cancer and dysplasia. Age and duration represents medians and ranges; binomial data represents % and numbers. P values are based on Fisher exact test for binomial data and 2-sample t-test for continuous variables
An analysis of patients that had never been treated with UDCA (n = 17) compared with those who were treated with UDCA until the end of follow-up (n = 21) was performed. No difference was seen between these two groups regarding cancer and dysplasia-free survival, (Figure 4).
In this long-term follow-up study on high-dose UDCA in patients with PSC-associated IBD no chemopreventive effect on colorectal cancer/dysplasia could be shown. Our results are in accordance with a cohort study by Wolf et al. where no difference in colorectal cancer/dysplasia was found between 28 patients who were treated with UDCA for at least 6 months and 92 patients who were not treated.
Two other studies have suggested a chemopreventive effect of UDCA on colorectal mucosa in IBD. The study by Pardi et al. also used a previously randomised cohort, but compared to our study, a lower dose of UDCA was studied and the randomised study period was 2 years compared to 5 years in our study. Despite the shorter exposure of UDCA and fewer patients (n = 52) included they were able to demonstrate a significant protective effect on development of colorectal cancer/dysplasia in their group of PSC-IBD patients. Tung et al. reported data from a surveillance programme including patients with ulcerative colitis and primary sclerosing cholangitis showing that the use of UDCA strongly reduced the risk for development of colorectal dysplasia. The reason for the contradictory results in comparison to our study is unclear. We showed an overall frequency of colorectal cancer or dysplasia of 29%, which is in line with previous studies speaking against detection bias.[5, 17, 18]
One reason for the disparity of the results may be partly explained by different UDCA doses. The dose of UDCA in our study (17–23 mg/kg/day) is higher than doses previously studied and appear to have less chemopreventive effect than the lower dose of 13–15 mg/kg/day. In addition, Alberts et al. have shown that a low dose of UDCA 8–10 mg/kg prevented recurrence of adenomas. Another recent study demonstrated that the use of very high dose UDCA 28–30 mg/kg/day was associated with increased risk of colorectal dysplasia and cancer in patients with UC and PSC. This suggests that the chemopreventive effect of UDCA on development of colorectal dysplasia or cancer decreases with increased doses. It can be speculated that the increased serum levels of hepatotoxic bile acids such as lithocholic acid (LCA), which may increase the risk of CRC, have been found in PSC patients on treatment with high dose UDCA. High LCA levels are presumably produced by bacterial 7-dehydroxylation of unabsorbed bile acids that pass into the colon.
The fact that UDCA-treated patients were older at diagnosis and had a shorter duration of colitis compared to placebo treated patients in the study by Tung et al., may have influenced their results although the importance of IBD duration as a risk factor in PSC-associated colorectal neoplasia has been questioned. We found no differences regarding these risk factors for development of colorectal neoplasia between our two groups. Fifty-five patients were excluded from this study because of previous colectomy, colorectal dysplasia or cancer before or within 6 months of study entry supporting a high risk of dysplasia early in the time course of the disease. A recent study has specifically evaluated the importance of the duration of the IBD in PSC for the risk of dysplasia or cancer. The yearly risk for dysplasia or cancer was shown to be is high soon after the coexistence of the two diseases. The only covariate associated to a higher incidence of colorectal neoplasia in our study was the number of colonoscopies performed indicating a surveillance bias since the finding of dysplasia leads to a need for additional colonoscopies.
We found a tendency towards less colonic neoplasia in UDCA-treated patients at the first follow-up in 2001, which encouraged us to extend the follow-up with the hypothesis that the difference would increase with time. After 12 years, however, there was no difference in cancer/dysplasia-free survival between the two groups. The follow-up in 2009 is more difficult to interpret, as the analysis is based on the originally randomised groups and not the true exposure of UDCA. Patients with early stages of dysplasia may have been more likely treated with UDCA to prevent further development of colorectal dysplasia or cancer based on the current knowledge at the time. Therefore, there is a large risk for introducing treatment bias when an evaluation of the true exposure of UDCA is made. Nevertheless, we performed an analysis of ‘true exposure’ that showed no significant differences between placebo, UDCA or patients treated with both UDCA and placebo regarding development of colorectal neoplasia. This result should be cautiously interpreted.
There were no cases of HGD among UDCA users compared to five in the placebo group. This is an interesting finding and may indicate that treatment with UDCA still might have a chemopreventive effect for development of cancer. This is highly speculative due to the few numbers of events.
There are limitations of the present study including its retrospective design. The two randomised groups were, however, similar regarding factors of importance for development of dysplasia or cancer in IBD. Therefore, in spite of the fact that the original study was initially not designed to evaluate the risk for colorectal cancer or dysplasia we believe that our results well mirror the effect (or rather absence of effect) of UDCA. Another drawback is that we did not perform re-evaluation of presence of dysplasia in the colon biopsies. Due to the multicentre nature of the study and to the number of investigators involved collecting all the biopsies for dysplasia, reviewing was considered to be unrealistic.
In conclusion, treatment with UDCA 17–23 mg/kg/day does not reduce the risk of colorectal cancer and dysplasia in patients with PSC. The use of UDCA to prevent colorectal malignancy in PSC-associated IBD should be questioned.
Declaration of personal interests: None. Declaration of funding interests: The preparation of this paper was funded in part by the Swedish Cancer Society and the Wallenberg Foundation.