A randomized controlled trial of colchicine plus ursodiol versus methotrexate plus ursodiol in primary biliary cirrhosis: Ten-year results

Authors

  • Marshall M. Kaplan,

    Corresponding author
    1. Division of Gastroenterology, Department of Medicine and the Tupper Research Institute, New England Medical Center and Tufts University School of Medicine, Boston, Massachusetts
    • 750 Washington Street, Box 233, Boston, MA 02111
    Search for more papers by this author
    • fax: 617-636-4207

  • Steven Cheng,

    1. Division of Gastroenterology, Department of Medicine and the Tupper Research Institute, New England Medical Center and Tufts University School of Medicine, Boston, Massachusetts
    Search for more papers by this author
  • Lori Lyn Price,

    1. Division of Clinical Care Research, Department of Medicine and the Tupper Research Institute, New England Medical Center and Tufts University School of Medicine, Boston, Massachusetts
    Search for more papers by this author
  • Peter A. L. Bonis

    1. Division of Gastroenterology, Department of Medicine and the Tupper Research Institute, New England Medical Center and Tufts University School of Medicine, Boston, Massachusetts
    Search for more papers by this author

Abstract

Primary biliary cirrhosis frequently progresses despite treatment with ursodeoxycholic acid (UDCA), the only approved therapy. Previous studies suggested that colchicine and methotrexate may improve biochemical tests of liver function, symptoms, and liver histology. The aim of the present study was to determine if the addition of colchicine or methotrexate to UDCA would improve survival free of liver transplantation. Eighty-five patients with histologically confirmed primary biliary cirrhosis whose serum alkaline phosphatase levels were at least twice the normal level and who were not yet candidates for liver transplantation were randomly assigned to receive colchicine or methotrexate in a double-blind study. UDCA was administered to all patients after 2 years. The primary end point was survival free of liver transplantation. Patients were followed up for a total of up to 10 years or until treatment failure. Data were analyzed on an intention-to-treat basis. Transplant-free survival was similar in both groups: 0.57 for colchicine plus UDCA and 0.44 for methotrexate plus UDCA, results that are similar to those predicted by the Mayo prognostic model. Significant improvement in liver biochemical tests and liver histology was observed in a subset of patients in both treatment groups who remained in the study for all 10 years. In conclusion, neither colchcine plus UDCA nor methotrexate plus UDCA improved survival beyond that predicted by the Mayo prognostic model. However, clinical, histologic, and biochemical improvement observed among those who remained in the study for 10 years suggests a possible benefit of these drugs in a subset of patients. (HEPATOLOGY 2004;39:915–923.)

Primary biliary cirrhosis (PBC), a chronic progressive cholestatic liver disease of presumed autoimmune etiology, is characterized by the destruction of small intrahepatic bile ducts and the eventual development of cirrhosis and liver failure.1 Although several drugs have been studied, there is still no totally effective treatment. Studies on the only approved therapy, ursodeoxycholic acid (UDCA), have consistently demonstrated improvement in biochemical tests of liver function, but its effect on transplant-free survival is unclear.2–7 There is less experience with colchicine and methotrexate, both of which have been evaluated in case series and small controlled trials given alone or in combination with UDCA. Although some studies have suggested that these drugs may improve biochemical tests of liver function and slow the rate of disease progression,8–9 others have produced discordant results.10–13

In 1988, we began a prospective double-blind trial to compare methotrexate with colchicine in the treatment of PBC. Several years after we began our trial, UDCA was reported to improve biochemical tests of liver function in PBC and to slow its rate of progression.14 As a result, we revised our protocol so that each patient who had been in our study for at least 2 years would receive UDCA in addition to colchicine or methotrexate and would be followed for a total of 10 years. We previously reported the results of the first 2 years of this study, the period in which patients received either colchicine or methotrexate alone.15 Colchicine and methotrexate were associated with a significant decrease in serum levels of alanine aminotransferase (ALT) and alkaline phosphatase, while bilirubin and albumin were unchanged. Improvement in pruritus was observed with both drugs, while methotrexate but not colchicine was associated with significant improvement in liver histology. There were not enough reportable events (i.e., treatment failures) during the first 2 years of the study to allow meaningful analysis of the primary outcome, referral for liver transplantation, or death.

Abbreviations:

UDCA, ursodeoxycholic acid; PBC, primary biliary cirrhosis; ALT, alanine aminotransferase; AST, aspartate aminotransferase; IgM, immunoglobulin M.

Methods

Patient Selection.

Inclusion and exclusion criteria have been described previously.15 Patients were included provided they had (1) a medical history and blood test results consistent with chronic cholestatic liver disease; (2) a serum alkaline phosphatase that was at least twice that of the upper limit of normal; (3) a serum bilirubin that was not greater than 10 mg/dL; (4) a liver biopsy performed within 12 months of entry that was consistent with primary biliary cirrhosis, and (5) imaging tests that demonstrated that bile ducts were patent.

Patients were excluded if they had end-stage liver disease, which was defined as: (1) a serum bilirubin level greater than 10 mg/dL or a serum albumin level less than 3.0 g/dL on two examinations two months apart; (2) hepatic encephalopathy; (3) hemorrhage from esophageal varices and/or portal gastropathy; (4) refractory ascites; or (5) signs of hypersplenism (i.e., a hematocrit less than 30, white blood cell count less than 2,500, and platelet count less than 100,000). Other reasons for exclusion were history of alcohol abuse, administration of drugs associated with chronic liver disease, contemplation of pregnancy, or any serious medical illness that might in itself cause liver dysfunction or shorten life expectancy.

Study Plan.

Each patient signed a consent form approved by the Human Investigation Review Committee of the New England Medical Center at Tufts University School of Medicine. Patients were admitted to the Clinical Study Unit of the New England Medical Center, where the following tests were obtained: a complete blood count, urinalysis, serum creatinine, blood urea nitrogen, serum bilirubin, alkaline phosphatase, aspartate aminotransferase (AST), ALT, albumin, immunoglobulin M (IgM), cholesterol, ceruloplasmin, antimitochondrial antibody, and prothrombin time. Percutaneous liver biopsy was performed unless it had been done within the previous 12 months. The liver biopsy slides were fixed in D-5 (mercuric chloride and formaldehyde) and stained with hematoxylin-eosin and Gomori trichrome stains. Upper gastrointestinal endoscopy was performed to evaluate the presence and severity of esophageal varices and/or portal gastropathy; endoscopic retrograde cholangiography, CT, or ultrasonography was performed if the patency of the bile ducts was in question.

Patients were randomly assigned, using a computer-generated list in blocks of four, to receive either methotrexate at 15 mg per week (5 mg q. 12 hours ×3) or colchicine at 0.6 mg twice daily. Colchicine and identical-appearing placebo were provided by Eli Lilly and Company (Indianapolis, IN), and methotrexate and identical-appearing placebo were provided by Lederle Laboratories (Pearl River, NY). To ensure that the two treatment groups were as similar as possible at baseline, patients were stratified into three groups based on their serum bilirubin level and whether or not they had been received colchicine prior to the study.15

Patients and investigators were blinded to the treatment assignments. Active drug and placebo were kept in the hospital pharmacy and dispensed by a research pharmacist who was the only caregiver with access to the randomization code. Each patient received one active study drug and placebo (given in the corresponding dosing intervals) for the other. Patient compliance was monitored by a research nurse who dispensed medications during each visit.

Initially, the study was designed such that each patient would remain in the double-blind phase of the study for 6 years or until clear evidence of progression of disease or drug toxicity was detected and the patient was judged a treatment failure. Evidence of progression included: (1) an increase in serum bilirubin of 3 mg/dL or more to a level exceeding 4 mg/dL that was maintained for at least 3 months; (2) a decrease in serum albumin of 0.8 g/dL or more to a level below 3 g/dL that was maintained for at least 3 months; and 3) appearance of a serious complication, such as hepatic encephalopathy, intractable ascites, or hemorrhage from esophageal varices. Patients who developed toxicity attributable to a drug were also classified as treatment failures. Drug toxicity attributable to methotrexate included interstitial pneumonitis and persistent cytopenias that were not due to hypersplenism (determined after patients had undergone bone marrow biopsy). Drug toxicity attributable to colchicine included intractable diarrhea and cytopenias not due to hypersplenism. Patients classified as treatment failures were censored and then referred for liver transplantation if clinically indicated or followed outside of the study.

During the course of the study, Poupon and colleagues reported that UDCA improved biochemical tests of liver function and might prolong survival free of liver transplantation.14 The protocol was therefore revised so that each patient in the study could receive UDCA (12–15 mg/kg body weight per day), provided by Ciba-Geigy Corporation (Summit, NJ), after being in the study for at least 2 years; the primary end point would be survival free of liver transplantation. This revision was approved by the Human Investigation Committee of the New England Medical Center, and a separate consent form was signed by each patient prior to entry into this phase of the study.

Patients continued to be followed as described above for a total of up to 10 years. Repeat blood tests, endoscopy, and a liver biopsy (if cirrhosis was not present initially) were performed at 2-year intervals.

Statistical Analysis.

We planned to enroll 60 patients for a period of 2 years based on an assessment of the relative efficacy of colchicine and methotrexate in decreasing serum alkaline phosphatase and/or serum bilirubin levels after 2 years of drug therapy.8, 16 Alkaline phosphatase and bilirubin levels were chosen because there were few data available on survival, referral for liver transplantation, or effects on liver histology from which to base power calculations at the time the study was designed. We estimated that 60 patients would provide a comparison between the two treatments, having 90% power to detect a difference between a 25% chance of success on colchicine and a 65% chance of success on methotrexate at the 5% significance level based on the definition of success described previously.15

Enrollment was larger than planned, including 85 patients seen during the 2-year randomization period.15 We previously reported that there were no significant differences in biochemical and histologic results between the two treatment groups when data from all three of the strata described above were combined or analyzed separately (Table 1).15 Thus, we present data from all three strata in this report.

Table 1. Clinical and Laboratory Characteristics of Patients With PBC at Entry to the Study According to Treatment Group
 Normal ValuesColchicine (n = 43)Methotrexate (n = 42)P Value
  • NOTE. Sample size is 41 in methotrexate group for IgM and varices.

  • *

    Mean ± SE.

  • Geometric mean and relative SE.

Age (yr) 51 ± 1.4*51 ± 1.5*0.92
Female 42 (98%)40 (95%)0.54
Antimitochondrial antibody 39 (91%)38 (90%)0.97
Asymptomatic 7 (16%)7 (17%)0.96
Fatigue 1.0 ± 0.21.0 ± 0.10.91
Pruritus 1.6 ± 0.21.3 ± 0.20.23
Hepatomegaly 38 (88%)31 (74%)0.09
Splenomegaly 15 (88%)13 (31%)0.70
Spider nevi 0.2 ± 0.10.3 ± 0.10.29
Esophageal varices 7 (16%)12 (29%)0.16
Stage IV (cirrhosis) 23 (53%)16 (38%)0.15
Serum bilirubin0.1–1.0 mg/dl1.1 ± 1.11.3 ± 1.10.33
Serum albumin3.5–4.5 g/dl4.1 ± 0.044.0 ± 0.050.19
Serum alk. phosph.116 IU494 ± 1.1485 ± 1.10.86
Serum ALT25 U/L79 ± 1.196 ± 1.10.13
Hematocrit38–4536 ± 0.436 ± 0.40.89
White blood cell count5000–10,0006.0 ± 0.45.4 ± 0.40.26
IgM0.6–2.63 ng/ml4.4 ± 1.13.8 ± 1.10.30
Cholesterol≤220 mg/ml263 ± 1.0251 ± 1.10.54
Prothrombin time9.5–11.5 sec10.1 ± 0.110.1 ± 0.10.66
AST≤35 U/L76 ± 1.195 ± 1.10.07

Cumulative mortality and survival free of liver transplantation were calculated by the Kaplan-Meier method and compared with the logrank test. Patients were censored at the time they withdrew from the study for any reason: treatment failure, drug toxicity (also listed as a cause of treatment failure), intercurrent illness, or patient preference. For comparison, survival was also determined after including patients who were followed after withdrawal17 and by analyzing results only after UDCA was introduced. Observed survival was compared with survival predicted by the Mayo model.17–20 Mayo risk scores were calculated for all patients at the onset of the study and also for those patients at the time that UDCA was introduced.

The effect of treatment on biochemical tests of liver function over the course of the study was assessed by calculating the means of laboratory tests of interest at three time points: (1) at baseline, (2) at the last visit before UDCA was added, and (3) at the final visit. All patients had laboratory values for the first two time points. When 10-year follow-up data were unavailable (e.g., if a patient died, received a liver transplantation, or dropped out of the study for other reasons), we used measurements from the last available time point at which the tests were performed. Thus the same number of patients was used to determine mean biochemical values at each time point.

Previous work and preliminary analysis revealed that the distribution of values for six variables—bilirubin, cholesterol, alkaline phosphatase, ALT, AST, and IgM–tended to be skewed to the right. As a result, we calculated the logarithms of the values in an attempt to make the distributions closer to normal. We calculated the arithmetic means for untransformed variables (serum albumin and prothrombin time). The differences in means were compared with ANOVA. Bonferroni-adjusted P values are presented for pairwise comparisons in tables 2 and 4. Statistical significance was determined as a two-sided P value less than or equal to .05.

Table 2. Changes in Mean Laboratory Values During the Study*
  ColchicineMethotrexate
nMeanB-LB-UU-LnMeanB-LB-UU-L
  • *

    Baseline to last (B-L), baseline to Urso (B-U), Urso to last (U-L); see text for normal values.

Alkaline phosphataseBaseline43494<.0001<.0001<.000141476<.0001<.00010.02
 PreUrso43344   41240   
 Last43227   41188   
BilirubinBaseline431.10.990.040.006401.20.120.440.99
 PreUrso431.3   401.4   
 Last431.0   401.5   
ALTBaseline4379<.00010.0003<.00014196<.0001<.00010.41
 PreUrso4359   4152   
 Last4340   4144   
ASTBaseline4376<.00010.590.00024194<.0001<.00010.99
 PreUrso4369   4161   
 Last4351   4158   
AlbuminBaseline424.1<.0001<.00010.83404.0<.0001<.00010.99
 PreUrso423.6   403.6   
 Last423.7   403.5   
CholesterolBaseline39271<.00010.0080.0239244<.00010.0030.09
 PreUrso39240   39214   
 Last39215   39197   
IGMBaseline174.90.0060.590.15243.70.030.150.99
 PreUrso174.3   243.1   
 Last173.6   242.9   
Prothrombin timeBaseline3810.20.0020.990.023810.1<.00010.030.17
 PreUrso3810.4   3810.6   
 Last3810.9   3811.0   
Table 4. Changes in Laboratory Values for Patients Who Showed Improvement in Histology and/or Were not Transplanted/Died/Censored*
  Colchicine (n = 18)Methotrexate (n = 11)
MeanOverall P-ValueB-LB-UU-LMeanOverall P-ValueB-LB-UU-L
Alkaline phosphataseBaseline453<.0001<.0001<.0001<.0001504<.0001<.0001<.00010.08
 PreUrso295    143    
 Last148    96    
BilirubinBaseline0.840.00040.0030.990.00070.790.88
 PreUrso0.89    0.81    
 Last0.60    0.76    
ALTBaseline70<.0001<.00010.07<.0001100<.0001<.0001<.00010.20
 PreUrso54    40    
 Last28    29    
ASTBaseline65<.0001<.00010.500.000189<.0001<.00010.00020.31
 PreUrso56    43    
 Last34    34    
AlbuminBaseline4.10.00080.400.00060.044.00.14
 PreUrso3.7    3.8    
 Last3.9    3.9    
CholesterolBaseline2800.00020.00010.040.102090.06
 PreUrso241    202    
 Last212    177    
IGMBaseline5.10.0010.0010.500.022.50.14
 PreUrso4.2    1.9    
 Last2.7    1.6    
Prothrombin timeBaseline10.10.549.80.21
 PreUrso10.0    10.0    
 Last10.3    10.0    

Histologic Analysis.

Liver biopsy specimens obtained at entry, just before starting UDCA, and at 10 years were coded and grouped according to each patient. The treatment assignment and sequence of the biopsy (i.e., first, second, or third) were not known to the investigators (M.M.K., S.C.) who read the biopsies under code. The histologic stage of disease in each specimen was determined according to accepted criteria.21 In addition, each specimen was scored numerically by a modified Knodell score as described previously.9 Differences in mean values were analyzed using ANOVA, and appropriate pairwise comparisons were made using Student's t test.

Results

Eighty-five out of 130 patients referred with a diagnosis of primary biliary cirrhosis were admitted to the clinical study unit for randomization.15 Of the 130 patients who were referred for the study, 28 did not meet the eligibility criteria. Ten of these patients had end stage PBC as defined above, 16 had serum alkaline phosphatase levels that were less than twice the upper limit of normal, and two did not have PBC. Seventeen additional patients met the eligibility criteria but did not wish to participate in a double-blind study. Overall, 43 patients were randomized to receive colchicine and 42 were randomized to receive methotrexate.

Twenty-nine percent of the methotrexate group had esophageal varices at baseline compared with 16% of the colchicine group (P = .16) (Table 1). Fifty-three percent of those in the colchicine group had cirrhosis (stage IV) at randomization compared with 38% who received methotrexate (P = .15). There were only three males in the study; one received colchicine, and two received methotrexate.

Figure 1 summarizes the disposition of all 85 patients during the 10 years of the trial. Of the 43 patients who received colchicine plus UDCA, 18 remained in the study until its completion at 10 years. Twenty-six dropped out of the study before 10 years: 15 were treatment failures, three dropped out because of drug toxicity (two with diarrhea and one with weakness) and were classified as treatment failures, one developed another illness, and six left the study for personal reasons. Of the 18 treatment failures, 11 had liver transplantation, three died while awaiting transplantation, and four were alive and well at 10 years.

Figure 1.

Flow diagram of subject progress through the phases of the randomized trial.

Of the 42 patients randomized to methotrexate, 11 remained in the study until its completion after 10 years. Thirty-two dropped out of the study before 10 years: 16 were treatment failures, seven dropped out because of drug toxicity (six with pneumonitis and one with thrombocytopenia) and were also classified as treatment failures, four developed other illnesses, and four left the study for personal reasons. Of the 23 treatment failures, 11 had liver transplantation, five died (including four elderly patients who declined liver transplantation), two await liver transplantation, and five are alive and well. One of the 11 patients who had a liver transplantation was randomized to receive methotrexate but never took any. She had a liver transplantation shortly after randomization because of severe osteoporosis. This patient was included in the intention-to-treat analysis of the survival data.

We were able to obtain accurate follow-up data at 10 years on 84 of the 85 patients in the study by means of telephone calls or e-mail messages to referring physicians and/or patients. Of those who were randomized to colchicine, 9 of 11 who had liver transplants and 29 of 33 who did who did not were alive after 10 years. Of those who were randomized to methotrexate, 9 of 11 who had liver transplants and 25 of 31 who did not were alive after 10 years. One patient who withdrew after 6 years was lost to follow-up.

Figure 2 displays survival free of liver transplantation. Estimated survival after 10 years in the colchicine plus UDCA group was 0.57 (95% confidence interval 0.41–0.73), while that of the methotrexate plus UDCA group was 0.44 (95% confidence interval 0.26–0.62). The results were not statistically different (P = .12). Figure 2 also displays predicted 7-year survival using the Mayo model where the risk score at baseline was −1.01.18, 19 Survival in both groups was similar to the predicted survival. There was no beneficial effect from either drug regimen, both of which included the concomitant use of UDCA for up to 8 years. There was also no improvement in survival compared to that predicted by the Mayo model when patients were included who had withdrawn from the study but continued to be followed17 and when survival was analyzed only after UDCA was introduced (Mayo risk score −5.07; survival data not shown). There was also no difference in survival between the two treatment groups when only those in groups 1 and 2 were analyzed, excluding patients in group 3 who had clinically advanced disease at the time of randomization (data not shown).

Figure 2.

Survival free of liver transplantation and expected survival according to the Mayo model.

Table 2 displays the means of the laboratory tests, comparing values at baseline with those after treatment with colchicine or methotrexate alone (just before UDCA was added) and after UDCA was added. In the colchicine plus UDCA group, there were significant improvements in IgM, AST, ALT, alkaline phosphatase, and cholesterol levels when baseline values were compared with final values. There were slight but significant changes in albumin levels and prothrombin times and no change in bilirubin levels.

In the methotrexate plus UDCA group, most of the improvement in laboratory tests occurred while patients were on methotrexate alone. There were significant improvements in IgM, AST, ALT, alkaline phosphatase, and cholesterol levels when baseline was compared with final values. There were slight but significant changes of albumin levels and prothrombin times and no significant change in bilirubin levels.

There were no significant differences between the methotrexate plus UDCA group compared with the colchicine plus UDCA group for any variable when baseline values were compared with the final values. However, the decrease in alkaline phosphatase, ALT, and AST levels from baseline to the time that UDCA was begun was significantly greater in the methotrexate group than in the colchicine group (226 IU vs. 150 IU, P = .004; 44 U/L vs. 20 U/L, P = .02; and 33 U/L vs. 7 U/L, P = .008, respectively). Conversely, the decrease in ALT and AST levels were significantly greater in the colchicine plus UDCA group from the time that ursodiol was started to the final values (19 U/L vs. 8 U/L, P = .03 and 18 U/L vs. 3 U/L, P = .004). There were no other significant differences at any time point for any of the other values between the two treatment groups.

Twenty-nine patients remained in the study for all 10 years, 18 on colchicine and UDCA and 11 on methotrexate and UDCA. Of this group, 26 patients (16 on colchicine plus UDCA and 10 on methotrexate plus UDCA) had liver biopsies at baseline, before adding UDCA and at 10 years (Table 3). Biopsies were of adequate size for interpretation. The mean number of portal triads was 19.9 ± 11.0 S.D. (range: 5–51; median: 16).

Table 3. Histologic Activity Index (HAI) and Stage at Three Time Points
 Colchicine (n = 16)Methotrexate (n = 10)
Histologic Activity IndexStageHistologic Activity IndexStage
Baseline12.52.512.12.5
PreUrso8.71.97.72.0
Last6.01.94.91.8
 HAIStage*
B-LB-UU-LB-LB-UU-L
  • *

    Global P-value greater than 0.05, no pairwise comparisons performed; N/A = not applicable.

Colchicine<0.00010.0060.07N/AN/AN/A
Methotrexate0.00040.020.21N/AN/AN/A

There was significant histologic improvement in both groups when the baseline was compared with the final biopsy but no significant differences between the two treatment groups. No patient in either group whose baseline histologic stage was 1 to 3 progressed to cirrhosis or extensive fibrosis. Significant improvement in the Knodell score9 occurred while patients were receiving colchicine or methotrexate alone, while additional (but not statistically significant) improvement occurred with the addition of UDCA. Overall, mean necroinflammatory scores fell from 12.4 to 6.0 on colchicine plus UDCA and from 12.1 to 4.9 on methotrexate plus UDCA. The histologic stage also decreased in both groups, although the change was not statistically significant.

There was a corresponding improvement in biochemical tests of liver function for patients who remained in the study for 10 years (Table 4). For the 11 patients in the methotrexate plus UDCA group, mean alkaline phosphatase levels fell from 504 IU to 96 IU, while serum bilirubin and albumin levels remained stable. Of the 18 patients on colchicine plus UDCA, mean alkaline phosphatase levels decreased from 453 IU to 148 IU after 10 years, while serum bilirubin and albumin levels worsened slightly. The decreases in alkaline phosphatase, ALT, and AST levels were significantly greater in the methotrexate plus UDCA group compared with the colchicine plus UDCA group from baseline to the time that UDCA was begun (361 IU vs. 158 IU, P = .002; 60 U/L vs. 16 U/L, P = .001; and 46 U/L vs. 9 U/L, P = .003, respectively). The only other significant difference between these two groups was in the alkaline phosphatase level from baseline to final where the decrease in the methotrexate plus UDCA group was significantly greater (408 IU vs. 305 IU, P = .0007).

There were significant differences in baseline variables between the patients who showed histologic improvement and/or were in the study after 10 years (responders) compared with those who were treatment failures. Nonresponders were significantly more likely to have stage 4 disease (24 of 31 vs. 5 of 29, P < .001), esophageal varices (12 of 31 vs. 2 of 29, P = .002), higher serum bilirubin levels (1.92 vs. 0.83 mg/dL, P < .0001) and higher mean ALT levels (111 vs. 90 U/L, P < .01). Age, serum albumin, and alkaline phosphatase levels were not significantly different.

Discussion

The results of this 10-year study demonstrated that there was no difference in survival free of liver transplantation in patients treated with colchicine plus UDCA versus methotrexate plus UDCA. In addition, neither drug combination improved survival as predicted by the Mayo model.17–20 These observations are disappointing and suggest that the encouraging results observed after 2 years of the study (improvement in symptoms, biochemical tests of liver function, and liver histology) did not translate into improved survival, even with the addition of UDCA to each treatment group.

Our results are consistent with other studies that cast doubt on the efficacy of any of the three drugs that we used (UDCA, colchicine, and methotrexate) and are also consistent with the results of a recent multicenter study that found no benefit of the combination of methotrexate plus UDCA compared with UDCA alone.6, 7, 10, 12, 22, 23 However, some studies have found that UDCA does improve survival free of liver transplantation,17, 24 and closer inspection of our results suggests that treatment may be beneficial for some patients.

Although survival was similar to the natural history of PBC predicted by the Mayo model, we observed clinical improvement (rather than survival with deteriorating liver function) in 18 patients on colchicine plus UDCA and 11 on methotrexate plus UDCA. This was reflected by a decrease in liver biochemical tests and an improvement in symptoms (e.g., pruritus) and liver histology. Necroinflammatory scores decreased by 6.4 points in the colchicine group and by 7.2 points in the methotrexate group. A decrease of 2 or more points in the necroinflammatory score is considered to represent significant improvement in liver histology.25 In both groups, the statistically significant decrease in necroinflammatory score occurred while patients were receiving methotrexate or colchicine alone. Histologic stage also decreased from 2.5 to 1.9 on colchicine plus UDCA and from 2.5 to 1.8 on methotrexate plus UDCA but the results were not statistically significant. Nevertheless, they are noteworthy because the mean histologic stage generally increases by one stage every 1.5 to 2 years in untreated patients.26, 27

The results must also be considered in light of the characteristics of the patients, many of whom had advanced disease at enrollment; 46% had cirrhosis, 23% had esophageal varices, and 20% had serum bilirubin levels greater than 2 mg/dL. These features are significant negative predictors of long-term survival, and such patients are unlikely to respond to any medication. Excluding patients who received liver transplantation (almost all of whom had clinically advanced disease at baseline), 25 of 31 patients randomized to methotrexate and 29 of 33 randomized to colchicine were alive after 10 years.

Our study may not have been adequately powered to detect a difference in survival, making it possible that the negative findings were the result of a type 2 error. The power analysis was based on expected changes in alkaline phosphatase and bilirubin levels because there were few data available on survival, referral for liver transplantation, or effects on liver histology from which to base power calculations at the time the study was designed. Nevertheless, comparing our results with a validated survival model (the Mayo model) suggests that the magnitude of benefit from these interventions (if any) may not be detected even with a larger study involving a similar population of patients.

The data presented here (and the results of other investigators) do not support the routine use of combination therapy with methotrexate and/or colchicine plus UDCA. However, the favorable clinical and histologic responses described above suggest that further studies of these drugs in patients with early-stage disease are warranted, particularly those who respond incompletely to UDCA alone. Approximately 75% of patients treated with UDCA alone (the only approved treatment) have progressive disease despite treatment.28, 29

Methotrexate has been associated with numerous side effects (including pulmonary toxicity and hepatic fibrosis), creating reluctance for its use in the treatment of liver disease. However, serious adverse effects were uncommon in our patients. Furthermore, pulmonary toxicity was not seen in a group of 265 PBC patients followed for up to 4 years in a prospective trial comparing UDCA to UDCA plus methotrexate.30 Hepatic fibrosis or cirrhosis is uncommon in patients treated with long-term low–dose pulse methotrexate unless they drink alcohol or are morbidly obese.31, 32 Bone marrow suppression is also uncommon with low-dose methotrexate.

We observed slightly lower survival in the methotrexate group. While not statistically significant (and not clearly attributable to methotrexate), a possible negative impact of methotrexate on survival in a subset of patients cannot be excluded. Thus, patients receiving methotrexate should receive the appropriate counseling regarding potential toxicity and should undergo careful monitoring.

In conclusion, combination therapy with methotrexate or colchicine plus UDCA does not improve survival in patients with PBC. A possible benefit in patients with early-stage disease warrants further study.

Acknowledgements

We thank Christopher Schmid for his statistical assistance and many former gastroenterology trainees, particularly Drs. Dawn Provenzale, Archna Sharma, and George Dickstein, who helped care for these patients.

Ancillary