The impact of ursodeoxycholic acid on cancer, dysplasia and mortality in ulcerative colitis patients with primary sclerosing cholangitis

Authors


Dr B. A. Lashner, Cleveland Clinic Foundation/A30, 9500 Euclid Avenue, Cleveland, OH 44195, USA.
E-mail: lashneb@ccf.org

Summary

Background : Colorectal cancer in primary sclerosing cholangitis patients with ulcerative colitis is mostly right-sided where concentrations of carcinogenic secondary bile acids are highest.

Aim : To investigate whether ursodeoxycholic acid could be chemopreventive for colorectal cancer.

Methods : A historical cohort study was performed on primary sclerosing cholangitis patients with ulcerative colitis where the 28 patients (cases) who were treated with ursodeoxycholic acid for at least 6 months (mean 3.4 ± 2.7 years) were compared with the 92 patients (controls) who were not treated with ursodeoxycholic acid. The primary outcomes were colorectal cancer and dysplasia. The secondary outcome was overall mortality.

Results : The cumulative incidence of dysplasia or cancer was not significantly different between cases and controls (P = 0.17 by log-rank test). The adjusted relative risk for cases of developing dysplasia or cancer was 0.59 (95% CI 0.26–1.36). The cumulative mortality was significantly different between groups (P = 0.02 by log-rank test). The adjusted relative risk for cases of death was 0.44 (95% CI 0.22–0.90).

Conclusion : In ulcerative colitis patients with primary sclerosing cholangitis, ursodeoxycholic acid did not reduce the risk of developing cancer or dysplasia. However, ursodeoxycholic acid may reduce mortality.

Introduction

Ulcerative colitis (UC) predisposes patients to colorectal cancer (CRC). UC patients with primary sclerosing cholangitis (PSC) have a significantly increased risk for neoplasia, particularly on the right side of the colon.1 The specific factors that predispose the colon of patients with PSC to be more likely to become dysplastic have remained elusive, but the preponderance of right-sided cancers raises suspicion that the effects of carcinogenic secondary bile acids may be a factor.2, 3

Ursodeoxycholic acid (UDCA), a drug often used in PSC patients and patients with other cholestatic diseases of the liver, has several favourable characteristics. It has hepatoprotective effects against the possible cytotoxicity of other bile acids by increasing the proportion of primary to secondary bile acids.4 By reducing the concentrations of carcinogenic secondary bile acids, it also may be chemopreventive for CRC.5, 6 As orthotopic liver transplantation (OLT) eliminates intra-hepatic cholestasis, it should favourably affect bile acid kinetics and have a protective effect against CRC.

Ursodeoxycholic acid also has been studied for its effect in reducing disease activity in PSC patients. Several studies have shown that UDCA causes a significant improvement in liver biochemistry, liver fibrosis scores and progression in cholangiographic appearances.4, 7–9 An improvement in 4-year expected survival of PSC patients, after 1 year of UDCA by using the Mayo risk score as an indirect predictor, has been shown.7 UDCA has not been directly shown to have a mortality benefit in the PSC population.

The purpose of this study was to determine whether UDCA confers beneficial effects on the risk of developing cancer or dysplasia in UC patients with PSC. Overall survival benefits from UDCA was analysed as a secondary outcome.

Methods

This was a historical cohort study of 120 UC patients with PSC. One hundred thirty-two potential subject identification numbers were obtained from computerized records of all in-patients and out-patients seen at the Cleveland Clinic between 1976 and 1994. Information on 12 patients was incomplete, mostly due to the absence of a cholangiographic diagnosis of PSC, and these patients were excluded. Medical records were reviewed to confirm diagnoses and to collect demographic and disease-related variables. The common closing date for medical record information was December 31, 2000. The Institutional Review Board of the Cleveland Clinic Foundation approved this project.

The diagnosis of UC was confirmed with both a characteristic endoscopic examination of inflammation extending proximally from the rectum as well as from compatible histological examination, with no positive stool cultures.10 The date of the first characteristic symptoms of UC was used to designate the onset of UC. The diagnosis of PSC was identified from cholangiography showing biliary strictures compatible with PSC. These findings included multifocal, diffuse biliary strictures with alternating normal or dilated segments, involving both the intra- and extra-hepatic bile ducts.11, 12 Patients were excluded from the study for having: (i) absent or nondiagnostic cholangiographic evidence of PSC; (ii) concomitant chronic active hepatitis determined by serology and/or biopsy; (iii) acquired immunodeficiency syndrome cholangiopathy;13 (iv) cholangiocarcinoma diagnosed within 2 years of developing PSC; (v) choledocholithiasis; (vi) previous biliary surgery; or (vii) PSC developing after colectomy. The date of the first abnormal cholangiogram was used to designate the onset of PSC. Patients were not excluded if PSC preceded the diagnosis of UC. All laboratory values were extracted from the medical record. Clinical variables were obtained from both the medical record and from direct communication with the patient.

Patients who had ever taken daily UDCA for PSC for at least 6 months were compared with those who had never taken UDCA. In addition, a comparison was made between patients who had undergone OLT vs. those who had not. The primary outcomes were the occurrence of CRC and colorectal dysplasia. The secondary outcomes were CRC-related mortality and overall mortality. Dysplasia was defined by the criteria of the Inflammatory Bowel Disease/Dysplasia Morphology Study Group14 and classified as low-grade or high-grade dysplasia. Patients diagnosed with indefinite dysplasia were classified as having no definite dysplasia. Biopsies taken prior to the publication of the dysplasia definitions were re-read by an experienced gastrointestinal pathologist to conform to standard terminology. Malignancies proximal to the splenic flexure were considered to be proximal cancers, the remainder were distal cancers. Pancolitis was defined as endoscopic involvement of colitis from the rectum to a point proximal to the splenic flexure. Left-sided colitis was defined as disease from the rectum to a point at or distal to the splenic flexure, and proctitis was defined as rectal disease only.

Variables that were examined included UDCA intake and dose, which was determined both from the medical record and from direct communication with the patient. The average dose and duration of UDCA was not counted if it was started after a principal outcome event such as dysplasia or CRC or if it was started following colectomy. Mortality was determined from medical records, communication with family or social security death records.

Categorical variables were summarized as frequencies and percentages, and compared according to UDCA use with the chi-squared test. Continuous variables were summarized as means and standard deviations and compared according to UDCA use with the t-test. The Kaplan–Meier method was used to obtain cumulative incidence rates and the log-rank test was used to compare curves. The onset of UC was considered the origin of Kaplan–Meier curves. Patients were censored at colectomy. Proportional hazards analyses were used to identify both univariable and multivariable predictors of outcome. Statistical significance was defined as P ≤ 0.05.

Results

This analysis was based on data from 120 patients, 28 of whom used UDCA before the common closing date, and 92 of whom did not. The mean number of years that the UDCA group was taking UDCA before the study end points of CRC or dysplasia was 3.4 ± 2.7 years. The UDCA and control groups were analysed for a median of 21 and 22 years, respectively, from the time of UC symptom onset. Using a univariable proportional hazards analysis, three variables were identified that were associated with an increased risk of CRC or dysplasia in this population of UC patients with PSC; advanced age, male gender and extensive disease (disease proximal to the splenic flexure) at the time of the first colonoscopy.

The clinical variables between the two groups were relatively similar (Table 1). The only variable that was significantly different was the higher incidence of OLT in the UDCA group. In the UCDA group, five patients (17.9%) developed dysplasia and three (10.7%) developed CRC while 12 controls (13.0%) developed dysplasia and 15 (16.3%) developed cancer. As shown in Figure 1, differences in the Kaplan–Meier curves were not significant (P = 0.17 by log-rank test). Adjusting for age at symptom onset, extensive disease and sex, the relative risk among cases of developing dysplasia or cancer was 0.65 [95% confidence interval (CI) 0.27–1.55]. Controlling for liver transplantation as a time-dependent variable, the relative risk among cases of developing dysplasia or cancer was 0.59 (95% CI 0.26–1.36). No significant dose–response effect could be identified (Table 2).

Table 1.  Patient characteristics
DemographicUC patients with PSCP-value
UDCA (n = 28)No UDCA (n = 92)
  1. Values given in parentheses are percentages. UC, ulcerative colitis; PSC, primary sclerosing cholangitis; UDCA, ursodeoxycholic acid; IBD, inflammatory bowel disease; CRC, colorectal cancer; N.S., not significant.

Male16 (57)59 (64)N.S.
Caucasian26 (93)88 (96)N.S.
Cigarette smoking at symptom onset
 Current 3 (11) 8 (9)N.S.
 Former 2 (7)10 (11) 
Orthotopic liver transplantation 9 (32)13 (14)0.03
Age at onset of UC (years)31 ± 1326 ± 120.06
Duration of UC (years)22 ± 1321 ± 10N.S.
Duration of PSC (years) 9 ± 5 8 ± 6N.S.
Pancolitis15 (54)46 (50)N.S.
Family history of IBD 2 (7) 5 (5)N.S.
Family history of CRC 2 (7) 4 (4)N.S.
5-Aminosalisylic acid use27 (96)73 (79)N.S.
Figure 1.

Colorectal cancer or dysplasia in UDCA and control groups.

Table 2.  Univariable proportional hazards analysis to predict LGD/HGD/cancer
DemographicHazard ratio95% CIP-value
  1. LGD, low-grade dysplasia; HGD, high-grade dysplasia; CI, confidence interval.

Ursodeoxycholic acid
 Any dose0.560.25–1.300.18
 Per 100 mg increase0.900.79–1.020.10
 Per 1 year increase0.880.73–1.060.17
 Per 100 mg-year increase0.980.95–1.010.14

All three CRCs in UDCA patients were detected endoscopically, but two of the 15 cancers in the control patients were determined from the cause of death column noted on their death certificate records and, in these two patients, the site of the malignancy was not known. Table 3 shows that one cancer (33.3%) in the UDCA group was proximal to the splenic flexure while 69.2% of the cancer among controls were classified as proximal (P = 0.25).

Table 3.  Cancer location for 16 patients with colorectal cancer at endoscopy or colectomy
 UDCA (n = 3)No UDCA (n = 13)*
  1. Values are expressed as n (%). UDCA, ursodeoxycholic acid.

  2. * The site of colorectal cancer was unknown in two patients.

Ascending colon1 (33.3)5 (38.5)
Transverse colon0 (0)4 (30.8)
Descending1 (33.3)1 (7.7)
Sigmoid1 (33.3)1 (7.7)
Rectum0 (0)2 (15.4)

There were significantly fewer deaths among the UDCA-treated patients and the differences in the Kaplan–Meier plots were statistically significant (P = 0.023 by log-rank test) (Figure 2). Adjusting for confounding variables and controlling for liver transplantation as a time-dependent variable, the relative risk of death among patients taking UDCA was 0.44 (95% CI 0.22–0.90). The cause of death, as determined by medical records and social security death records, in the two groups of patients is shown in Table 4. The largest differences in cause of death were seen with CRC and hepatic failure.

Figure 2.

All causes of mortality in UDCA and control groups.

Table 4.  Causes of death
Cause of deathUC patients with PSC: deaths
UDCA (n = 28)No UDCA (n = 92)P-value
  1. Values given in parentheses are percentages. UDCA, ursodeoxycholic acid.

Colorectal cancer1 (4)9 (10)
Cholangiocarcinoma03 (3)
Hepatic failure2 (7)14 (15)
Rejection03 (3)
Sepsis03 (3)
Other2 (7)5 (5)
Unknown01 (1)
Total5 (18)38 (41)0.04

Orthotopic liver transplantation did not have a protective effect for CRC or dysplasia. Among the 22 patients who underwent an OLT, 9% developed CRC or dysplasia prior to or in the same year as OLT while 27% developed CRC or dysplasia following OLT. Proportional hazards modelling did not identify OLT as a ‘protective’ variable for the development of CRC or dysplasia.

Discussion

It is now established that the risk of CRC among PSC patients with UC is elevated, but the reason for this increased risk is not clear. Demonstrating a preponderance of right-sided cancers would support the hypothesis that the increased risk of CRC is due to a high concentration of carcinogenic secondary bile acids delivered to the colon. In our earlier study, more than half of the cancers were proximal to the splenic flexure.1 UDCA, an attractive chemopreventive agent for CRC in UC patients as it diminishes the concentration of secondary bile acids, was associated with fewer CRCs, especially right-sided CRCs, but these differences were not significant. Our earlier study was designed to demonstrate the increased risk of CRC among UC patients with PSC. It was not designed to carefully examine UDCA as a possible chemopreventive agent.

Ursodeoxycholic acid is being increasingly used by hepatologists to treat patients with PSC to improve liver function tests, but the purported effect of slowing the rate of progression of disease has not been demonstrated.4 Only 23% of patients in our study were taking UDCA, and now a vast majority takes the medication. UDCA has not been shown to improve mortality in PSC. While our study did not have mortality as a primary outcome measure, our results indicate that mortality (mostly liver-related and CRC-related mortality) is diminished by UDCA to the level of statistical significance. For patients with UC and PSC, UDCA could have several beneficial effects.

Two studies have examined the effectiveness of UDCA as a chemopreventive agent in UC patients with PSC. Tung et al. 5 from the University of Washington studied 59 patients with UC and PSC and found that UDCA was associated with a lower frequency of colorectal dysplasia (adjusted odds ratio 0.14, 95% CI 0.03–0.64). Dysplasia was found in 32% of the 41 patients treated with UDCA and 72% of the 18 patients not treated with UDCA. The main difference in results between our two studies is the extremely high rate of dysplasia among controls in the Tung study, which may be explained by their classification of patients who were indefinite for dysplasia as having dysplasia.

Pardi et al.6 from the Mayo Clinic found UDCA to be chemopreventive for CRC or dysplasia. The data for their study were obtained from a prior randomized, placebo-controlled trial on 52 patients with PSC and UC designed to evaluate the effect of UDCA on liver function. Three of 29 patients (10%) assigned to UDCA developed dysplasia compared with eight of 23 patients (35%) treated with placebo. The relative risk of UDCA-treated patients for developing CRC or dysplasia was 0.26 (95% CI 0.06–0.92). The difference between our studies is the very low rate of dysplasia seen in the treatment group, possibly because of the short follow-up time (mean follow-up 6.8 years) in the Pardi study.

There are some important limitations to our study. Although our study was more than twice as large as the other studies, a type II error is still possible. Also, some potential confounding variables, such as 5-aminosalicylic acid use, nonsteroidal anti-inflammatory drug use and co-morbidities were not studied. Furthermore, patients took UDCA for a variable amount of time during their illness, possibly obscuring a chemopreventive effect.

Ideally, the onset of UDCA use should be used as the origin of the Kaplan–Meier curves. Such an approach is impractical, although, as the controls were not ‘exposed’ to UDCA. In cohort studies on UC patients, UC onset has been successfully used as the origin of the Kaplan–Meier curves and provides for the fairest comparison. Such an approach shifts both curves to the right and should not bias the results. The onset of PSC is too indefinite in this population to be used as a starting point.

Ursodeoxycholic acid was given to PSC patients with UC at the discretion of their physician to only a minority of patients. There is no information as to why these patients were selected, but it could have been related to physician or patient preferences, cost of medication, patient adherence or co-morbid illnesses. At the time, there was no ‘guideline’ in place for treating PSC patients with UDCA. UDCA was used by all hepatologists to a varying degree. We adjusted for factors we could measure and thought were important, but it is possible that additional factors could have resulted in selection bias.

There was a statistically significant overall mortality benefit seen among patients treated with UDCA. However, UDCA was not associated with a significant decrease in the number of deaths from hepatic failure. Even combining hepatic failure, rejection and sepsis, the difference in liver-related mortality is not significant. However, by combining liver and CRC-related deaths, there was a significant mortality benefit. Only one other study has shown a potential mortality benefit from UDCA in a UC population with PSC.7 Harnois et al. from the Mayo Clinic showed that high-dose UDCA improved liver function tests and the Mayo risk score to theoretically increase life expectancy by 4 years. UDCA appears to be beneficial in this population.

In summary, UDCA did not significantly reduce the risk of developing cancer or dysplasia in UC patients with PSC. However, UDCA did have benefits for decreasing mortality. Patients with PSC and UC should be considered for UDCA therapy.

Acknowledgement

The authors would like to acknowledge the Ann and E. Bradley Jones Endowment Fund for the study of IBD.

Ancillary