Ten-year combination treatment with colchicine and ursodeoxycholic acid for primary biliary cirrhosis: a double-blind, placebo-controlled trial on symptomatic patients


Correspondence to: Dr Battezzati Divisione di Medicina Interna III, Dipartimento di Medicina, Chrurgia e Odontoiatria Ospedale San Paolo via Di Rudinì, 8, 20143 Milano, Italy. E-mail: battehsp@mailserver.unimi.it



Combined medical treatment may provide further benefit to primary biliary cirrhosis (PBC) patients administered ursodeoxycholic acid (UDCA).


To evaluate the long-term effects of colchicine and UDCA in symptomatic PBC patients.


We extended up to 10 years the double-blind treatment of 44 symptomatic PBC patients originally included in a 3-year multicentre study comparing UDCA and colchicine (U + C) versus UDCA and placebo (U + P). Outcome measures were death or liver transplantation; incidence of clinically relevant events; clinical and quantitative variables retaining prognostic information.


Mean follow-up was 7 ± 3 years. One patient was lost, three withdrew because of jaundice (U + P); two patients stopped colchicine but remained on UDCA. Eleven patients (two for liver-unrelated reasons, U + P) and six patients (U + C) died, three and two patients, respectively, were transplanted (incidence rate difference, five cases per 100 patient-years; 95% CI, −1 to 11). Hepatocellular carcinoma developed in one (U + P) and four (U + C) patients (difference, −2; CI, −5 to 1), portal hypertension complications in nine patients from each group (difference, 1; CI, −5 to 6). Trends of serum bilirubin, Mayo score, antipyrine clearance were similar among treatment groups.


In cirrhotic PBC patients, colchicine does not offer additional benefits to UDCA. In this population, UDCA does not obviate disease progression.


Administration of ursodeoxycholic acid (UDCA) to patients with primary biliary cirrhosis (PBC) is followed by marked improvement of serum liver biochemistry and, less consistently, of symptoms and liver histology.1 However, it is still unclear whether such favourable effects are reflected by improved survival or delay of liver transplantation, and whether they extend to all PBC patients.2, 3 Combination therapy has been advocated in PBC to achieve additional benefits, especially in terms of clinically relevant events.4

To evaluate the association of colchicine to UDCA for treatment of symptomatic PBC, in 1989 a group of Italian investigators, including the authors of the present study, designed a long-term, double-blind multicentre trial the results of which after 3 years of follow-up have been recently published in this journal.5 At that time, colchicine had emerged as a promising drug in view of a number of reports suggesting its possible effect on survival in liver cirrhosis of different origin and, to a lesser extent, in PBC itself.6–9 Additional merits of colchicine in PBC, including lack of any predicted significant toxicity, were highlighted and the additive effects of combined therapy with UDCA were envisaged.10, 11 Patients with more advanced disease were expected to take most benefit from the association between a drug showing antifibrotic properties, like colchicine, and UDCA, which had been reported at that time to be less helpful, and even unsafe, when administered at standard doses in this category of patients.12, 13

Some other studies in PBC failed to demonstrate a substantial or protracted effect of colchicine, either alone or in combination with UDCA.14–17 More favourable results emerged from the analysis of the Italian trial after 3 years of treatment.5 Indeed, when compared with the prolonged course of PBC, the time span of the studies published to date may well have been too limited to address the effects of colchicine on clinically relevant end-points. Accordingly, two of the centres participating in the Italian trial agreed to continue the study up to 10 years after patient enrolment, while maintaining the double-blind format.

We report here the results of the extension up to 10 years of the Italian randomized, double-blind clinical trial comparing efficacy of treatment with UDCA and colchicine with that of UDCA and a placebo. The primary end-points for the comparison were survival or need of liver transplantation and incidence of other clinically relevant outcome events. Further outcome measures were changes in some clinical variables and quantitative liver function tests retaining prognostic information.


The general features of this study design are outlined in the report of the 3-year results of a randomized, double-blind trial,5 of which the present study represents a continuation. That trial originally involved six centres that enrolled 90 patients and were individually responsible for their care and performing laboratory tests. According to the study protocol, data had to be analysed in blinded conditions after all patients had been treated for at least 3 years. Two of the participating centres agreed to extend the duration of the study to achieve sounder evidence regarding the effects of combination therapy on clinically relevant end-points. The present report concerns the 44 patients (42 with liver cirrhosis) who had been enrolled by the two centres, and were subsequently followed up to December 1999 without disclosing treatment assignments.

Patient selection

Patients with an established diagnosis of PBC and symptomatic disease, as defined by presence of pruritus necessitating treatment during the 6 months prior to evaluation, or serum bilirubin exceeding 2 mg/dL at two separate determinations during the previous 6 months, or histological diagnosis of cirrhosis, were considered for admission into the study. Patients were excluded if they had severe liver disease, as defined by ascites, previous episodes of gastrointestinal bleeding or encephalopathy, serum bilirubin levels exceeding 10 mg/dL; evidence of malignant conditions or of other major diseases unrelated to PBC; alcohol abuse; previous treatment with colchicine or immunosuppressant agents; or if low compliance was anticipated. Patient enrolment started in March 1989 and lasted 1 year.

Study design

The study was carried out according to the principles of the Declaration of Helsinki and, because at that time approval by an ethical committee was not mandatory, it was approved in the centres where it had been established. Patients were thoroughly informed of the nature, purpose and requirements of the study before giving informed consent.

The study assessed in double-blind conditions the effects of the association of colchicine or a placebo to a standard therapeutic regimen of UDCA, which was administered to all patients at a dose of 250 mg twice daily starting from at least 3 months before entry into the trial. Colchicine, 1 mg daily, or an indistinguishable placebo were randomly assigned to patients according to a computer-generated list developed separately for each centre. Compliance was checked by count of unused capsules returned at each visit. Patients were allowed to continue taking supportive medication, such as calcium or lipid-soluble vitamins, or colestyramine, which had to be taken at least 4 h before or after the study drugs.

Clinical examination, routine haematological tests and blood chemistries were performed every 3 months during the first year and twice yearly thereafter. A liver biopsy was available for each patient before randomization.18 After the first 3 years of follow-up, the 44 patients enrolled in the two centres that agreed to continue the study maintaining the double-blind format continued to be followed with the same modalities as before. Antipyrine clearance was assessed at baseline and after 1, 3 and 5 years. Patients underwent ultrasonography at least yearly; upper gastrointestinal endoscopy was performed in all patients with cirrhosis and repeated at least every 2 years. Histological assessment was not included as a surrogate end-point because in the large majority of our PBC patients liver cirrhosis was present, a condition for which performing repeat liver biopsy seemed unjustified.

Laboratory investigations

Haematological tests and blood chemistries were performed in the clinical laboratory of the participating hospitals by routine automated techniques. Liver enzyme concentrations were standardized according to the reference values for each laboratory and were expressed as multiples of the upper limit of the reference values. Antipyrine clearance was determined centrally by a spectrophotometric assay19 on plasma samples obtained before oral administration of the drug (SIGMA Chemical Co., St. Luis, MO, USA) at the dosage of 18 mg/kg and after 3 h and 24 h.

Analysis of data

In the present study, interest of the analyses was focused on the following response variables: death (liver-related deaths or all deaths) or need of liver transplantation, which was the main outcome measure; and incidence of complications of portal hypertension (defined as ascites, gastrointestinal bleeding or portal-systemic encephalopathy) and of hepatocellular carcinoma. To afford an overall evaluation of treatment efficacy, criteria defining treatment failure were also set, comprising the occurrence of death, need of liver transplantation, development of complications of portal hypertension or hepatocellular carcinoma, doubling of serum bilirubin to levels higher than 4 mg/dL, treatment discontinuation for any cause, or loss to follow-up.

Secondary end-points were represented by the changes of serum bilirubin, the Mayo score20 and antipyrine clearance. As a summary measure of the rate at which each variable changed over time, the slope of the line fitted to serial values obtained from each subject was measured by making a least square estimate of the regression coefficient.21 Use of a summary measure was chosen to avoid the need for repeated statistical testing at various time points. For serum bilirubin and antipyrine clearance, the coefficients were calculated after logarithmic transformation of values in order to achieve a normal distribution of the residuals around the regression line. Mean values of the regression coefficients from each treatment group were compared using a two-sample t-test. Intensity of pruritus was graded on a scale from 0, denoting absence of symptom, to 2, denoting severe intensity with skin lesions.

The chi-squared or Fisher’s exact test was used for the analysis of categorical variables, and the Mann–Whitney test for the analysis of continuous variables. The latter were summarized as means and standard deviations. For each outcome measure, incidence rate has been reported as the number of events for 100 patient-years, and difference in incidence rates between the two treatment groups with its 95% confidence interval (CI) were calculated. Cumulative proportions of patients developing an event were estimated by means of the Kaplan–Meier approach using the date of randomization as the starting point. The log-rank test was used to compare survival curves. Analyses were done according to the intention-to-treat principle using as the starting point the date of first treatment administration. All of the statistical analyses were two-sided and were made using Stata Statistical Software (Stata Corporation, College Station, TX).


Main patient characteristics at time of enrolment into the study are reported in Table 1. The 15 patients who had been included in an earlier trial on UDCA22 did not differ in any respect from the patients who received UDCA in the present study for the first time. Baseline values of the main laboratory variables and the Mayo score are reported in Table 2.

Table 1.   Clinical and laboratory characteristics of patients at entry into the study, according to treatment group Thumbnail image of
Table 2.   Changes in the main laboratory values and the Mayo score during the study Thumbnail image of

Time from randomization to last visit or death was 6.4 ± 2.7 years (range, 1–10 years) in the UDCA and placebo group and 7.3 ± 2.9 years (range, 8 months to 10 years) in the UDCA and colchicine group. The mean UDCA dose expressed on a weight basis was 8.8 ± 1.2 mg/kg in both treatment groups. During the study one patient, who was receiving UDCA and placebo, was lost to follow-up after 8 years of treatment; treatment was withdrawn in three patients from the same group after development of jaundice. In the UDCA and colchicine group, one patient developed severe diarrhoea that disappeared after treatment discontinuation, and another patient refused to continue blinded treatment: both of them remained on open-label UDCA. With the only exception of the patient who was lost, all of the other patients continued regular follow-up.

Incidence of clinically relevant events during the study period is reported in Table 3. No significant difference emerged from statistical comparison. There were more events in the UDCA and placebo group, which however, included the two deaths from causes other than liver disease. Cumulative proportion of deaths from all causes or need of liver transplantation in patients receiving the two treatment regimens is shown in Figure 1.

Table 3.   Major clinical outcomes of the study Thumbnail image of
Figure 1.

 Kaplan–Meier estimates of the probability of surviving without liver transplantation according to treatment group. P=0.11 8 (log-rank test).

Changes that occurred in the main variables considered during the study period are summarized in Table 2. Serial values of serum bilirubin, the Mayo score and antipyrine clearance from individual patients are shown in Figure 2. Analysis of the time trends of these variables did not disclose any significant difference between the two treatment regimens. The patients who died from liver-related causes or underwent liver transplantation showed clear trends towards worsening of serum bilirubin and Mayo score values over time. One patient from each treatment group, who died of complications of portal hypertension, maintained normal values of serum bilirubin throughout the study, while showing comparatively higher values of the Mayo score. Progressive lowering of serum albumin and development of intractable ascites before bleeding mainly contributed to raised Mayo score values in these two patients despite persistently normal levels of serum bilirubin.

Figure 2.

 Changes of serum bilirubin (Panel A), the Mayo score (Panel B), and antipyrine clearance (Panel C) in the two study groups. Patients who died of liver-related causes or underwent transplantation during the study period (not necessarily at time of last determination) are denoted by the + symbol. Dotted lines are defined by mean values of the regression coefficients and the intercepts of the lines fitted to data from each individual patient. Mean value of individual regression coefficients measuring the slope of the line fitted to the logarithm of serum bilirubin values (Panel A) was 0.133 (95% CI, 0.044 to 0.223) in the UDCA and placebo group and 0.096 (CI, 0.012 to 0.180) in the UDCA and colchicine group (P=0.981 for the difference between the two regression coefficients). The corresponding figures for the Mayo score (Panel B) were, respectively, 0.365 (CI, 0.126 to 0.604) and 0.333 (CI, 0.120 to 0.545) (P=0.880); those for the logarithm of antipyrine clearance values (Panel C) were, respectively, −0.107 (CI, −0.149 to −0.066) and −0.066 (CI, −0.112 to −0.019) (P=0.2 66). A positive regression coefficient indicates that values increased over time, whereas a negative one indicates a decrease, with the more extreme values denoting greater changes. A 95% CI that does not cross the 0-value points to a progressive change in the same direction within the treatment group. A significant difference between regression coefficients would indicate that different trends occurred in the two treatment groups.

Among the patients who did not show oesophageal varices on initial endoscopic evaluation, incidence of oesophageal varices was 20% in the UDCA and placebo group and 33% in the colchicine and UDCA group. Pruritus worsened in two patients, did not change in 13 and improved in seven patients in the UDCA and placebo group; one patient experienced worsening of pruritus, 12 no change, and nine an improvement in the UDCA and colchicine group.


The present study indicates that administration of colchicine protracted for up to 10 years in patients with PBC in the cirrhotic stage has no additional effect to treatment with UDCA. Use of an active drug such as UDCA allowed us to maintain in this trial a well-motivated group of patients and to extend double-blind treatment with colchicine or a placebo well beyond the time span of the available studies on this drug.

When administered alone for 4 years, colchicine did not prevent complications of cirrhosis, death, or need of liver transplantation over an average follow-up time of 8 years.14 Our study also suggests that colchicine given in association with UDCA for a longer period does not affect the ultimate outcome of PBC. Biochemical parameters of disease activity, a quantitative liver function test,23 and a validated prognostic indicator, the Mayo risk score,20 were similarly unaffected by colchicine administration. Marginal but transient effects of colchicine, either alone or in combination with UDCA, have been reported by some other studies.7, 14–16, 24 It cannot be excluded that such divergent findings are explained by differences in the study designs. The present study was specifically addressed to PBC patients with symptoms or liver cirrhosis, while more heterogeneous populations were enrolled in the other studies. We used a summary measure to compare liver function indices or prognostic variables between the two treatment regimens thus avoiding the need for repeated statistical testing at a series of time points.21 Comparison of data at various time points may also be misleading in long-term studies, because such data generally pertain to different subpopulations as a result of progressive removal of patients with the worst outcome.

The long-term data from this study also invite consideration of the efficacy of UDCA treatment, which was administered to all patients. Neither death, nor need of liver transplantation, nor development of major complications of cirrhosis were obviated by any of the treatment regimens. Disregarding assignment to colchicine, incidence of hepatocellular carcinoma was one case for every 42 patients-years of follow-up among male patients and one case for every 63 patients-years among female patients. Compared to a large English cohort of untreated patients,25 incidence of the tumour was remarkably similar among male patients and even higher in our female patients (one case for every 41 patients-years and one case every 159 patients-years, respectively, in the English study). Both PBC populations had histologically confirmed stage III or IV and similar length of follow-up (82 ± 34 months compared with 88 ± 55 months in the English study). Patients included in a trial undergo closer and more systematic observation, although higher accuracy of the estimated incidence of hepatocellular carcinoma is expected from the much larger English study. Anyway, the observed similarity of incidence figures strongly suggests that long-term UDCA administration does not prevent tumour occurrence in PBC once cirrhosis has developed.

Serial determinations of bilirubin values or of the Mayo score are consistent with the notion that progression of the disease may not be halted by medical treatment, especially in patients with more advanced disease. Assessment of antipyrine clearance also suggests that progressive deterioration of quantitatively assessed liver function occurs despite treatment. On the basis of preliminary data from our group, a dose of about 9 mg/kg of body weight, corresponding to the lower end of the range generally used in clinical trials, was chosen for long-term administration of UDCA when the present study was planned. Data on biliary enrichment with UDCA in PBC patients who were administered this bile acid have shown that such a dosage may be considered suitable for a population with more advanced disease.26, 27 A large controlled trial on PBC patients treated with higher UDCA doses also reported less favourable results in patients initially showing bilirubin levels greater than 2 mg/dL.13 It is therefore likely that UDCA administration is less effective in PBC when the disease has progressed to cirrhosis and that such treatment should be initiated earlier in the clinical course.

In conclusion, data from the present study indicate that long-term colchicine administration does not offer additional benefit to UDCA treatment in PBC patients in whom the disease has progressed to the cirrhotic stage. In the same patients, further progression of the disease does not seem to be obviated by UDCA treatment. Although limited in size, this study is unique for length of treatment starting from an advanced stage and does discourage any further study with such combination in PBC.