Huet et al. [49] have used a different approach, looking at portal hypertension in a study unlikely to be repeated because of its invasive nature. A total of 132 patients had portohepatic gradient and biochemical values measured at inclusion and every two years. After two years of treatment, a decreased or stable portohepatic gradient (hazard ratio, 4.64; 95% CI: 2.01–10.72) and normalization of AST level (hazard ratio, 2.89; 95% CI: 1.03–8.05) were predictive of better survival on multivariate analysis. Responders (defined as either stable or improved portohepatic gradient and normalized AST level at two years) had a 15-year survival similar to that of a matched local Canadian population.
Histology can be a hard endpoint, but only in the context of clinical trials where large numbers of biopsy specimens are available to offset the effect of sampling error. Transition rates between Ludwig histologic stages have been calculated using Markov modelling (namely that, given the present state, the future and past states are independent). In a large randomized trial of D-penicillamine, which was found to have no effect on histologic progression, the mean increase in histologic stage for PBC can be estimated to be one stage every 1.5 years. After four years, 31% of patients whose initial biopsy showed stage 1 had progressed to cirrhosis, and 50% of patients who started in stage 2 had progressed to cirrhosis [50]. An earlier study (on a relatively small number of patients) did indicate that patients who have liver fibrosis or cirrhosis on biopsy had a worse survival than those without fibrosis or cirrhosis [51]. On its own, presence or absence of cirrhosis is not a highly predictive surrogate marker for final outcome, that is, death, presumably because there are other features which factor into progression of disease. A detailed review of liver histology in PBC suggests that the presence of lymphoplasmacytic interface hepatitis is a marker of more rapidly progressive disease [52], and in another report of four cases, rapidly progressive bile duct loss, even in the absence of cirrhosis led to liver failure [53]. Corpechot described the progression toward cirrhosis in 183 UDCA-treated patients with PBC [52]. A total of 254 pairs of liver biopsy specimens collected during 655 patient-years were studied and the incidence of cirrhosis after five years of UDCA treatment was 4%, 12% and 59% among patients followed up from stages I, II and III respectively. At ten years, the incidence was 17%, 27% and 76% respectively. The median time for developing cirrhosis from stages I, II and III was 25 years, 20 years and 4 years respectively. The independent predictive factors of cirrhosis development were serum bilirubin greater than 17μmol/l, serum albumin less than 38g/l, and moderate to severe lymphocytic piecemeal necrosis. Validation of the importance of interface hepatitis is important, as are other potential factors (vascular supply to the liver, the presence of various inflammatory mediators and markers of tissue fibrosis). In the latest paper from France already discussed, independent predictive factors of death or transplant were baseline serum bilirubin level >17 μmol/l (RR, 1.7), histologic stage ≥ 3 (RR, 1.5), interface hepatitis (RR, 1.9), and the absence of biochemical response (ALP > 3 ULN or AST > 2 ULN, or bilirubin > 17μmol/l) (RR, 2.3) [48]. There is no good evidence that the AMA titer correlates with the course of PBC [54] (although it can fall with treatment [55]), but some groups have suggested specific anti-nuclear antibodies may delineate subtypes of PBC [27]. This, however, remains to be validated widely [56].
Aside from the increasing rarity of liver deaths from PBC, it has also become very difficult to use death as an endpoint, since most patients with decompensated PBC, if they have no contraindication, are referred for liver transplantation. As liver transplantation has become established for end-stage liver disease in particular, it has been used as a final outcome measure by many. Even transplant has some variability; however, certainly as some patients in the past have been transplanted for symptoms, different centers have longer waiting times, whilst factors such as blood group and body size can alter the duration of waiting. With the advent of scoring systems such as MELD, these issues will be largely addressed. Furthermore, perhaps because of therapy the rates of those patients with PBC going to transplant continues to fall markedly [24]. Over a 12-year period (1995 to 2006) PBC and PSC transplant data of first-time liver recipients from the United Network for Organ Sharing database were collected. Although the absolute number of liver transplantations in the USA increased by an average of 249 transplants per year the absolute number of transplants performed for PBC decreased by an average of 5.4 cases per year. The absolute number of transplantations for PSC showed no statistical change.
Therapeutic trial design: assessment of credibility
Understanding the limitations of the available evidence is essential for those interested in the evidence base for treatments of PBC. Numerous trials form the basis of the current treatment offered to patients. In evaluating such work a number of concerns arise for a disease such as PBC which is relatively uncommon and has a slow and varied natural history. To establish appropriate study size requires not only a large sample of historical data to chart the natural history of untreated disease but also pilot data, from which an estimate of the expected benefit can be calculated (sample size must also assess the frequency of adverse events). Historical data may not, though, always remain valid, for example the exclusion of chronic hepatitis C was only possible after approximately 1990, and patterns of presentation may change with time. The pivotal patient descriptions in the New England Journal of Medicine by Sherlock in 1973 clearly do not reflect those seen now [23, 57]. Recognized improvements in trial management have helped produce more robust data, with much thought given to careful trial design (e.g. double blinded, prior sample size calculation, stopping rules, crossover arms, and intention-to-treat analysis). The intention-to-treat principle is particularly relevant when carrying out statistical analysis of outcome to avoid as much bias as possible when interpreting the data. This means that all patients who are randomized need to be included in the final analysis, even though they may have been censored at a very early point in the trial because of untimely death, need to withdraw from the study, non-compliance, etc. Using the intention to treat principle allows for a very conservative analysis of the effectiveness of therapy, but permits assessment of the “real life” situation, that is, the generalizability of the study.
Meta-analysis can also overcome some concerns, particularly sample size, but many assumptions are made about inclusion and exclusion of studies, and original raw data is not always accessed. Additionally, meta-analysis is a tool to define the gaps in knowledge (and thus to help the design of better studies), rather than a true grading system for an intervention, meaning conclusions drawn mayreflect the quality and quantity of the evidence available as much as anything else. Variability in clinical practice, referral biases and varied ethnicity also generate difficulties as clinicians pool their patient resources. Finally, a major challenge relates to early detection of disease such that significant events, for example death or liver transplantation are unlikely to occur during any single trial. The use of surrogate markers (biochemical response, histological change, disease risk scores) as endpoints are by necessity employed, but their own origin and limitations need to be recognized as already discussed [58].
Randomized controlled trials of treatment for primary biliary cirrhosis
Immunosuppression
As the archetypal presumed autoimmune disease, inevitably immunosuppression has been evaluated as a therapy. Because the majority of patients with PBC are women and osteoporosis is prevalent (but not necessarily causally related to PBC [59–61]), corticosteroid therapy has for the most part been avoided. Trials of immunosuppressive therapy in PBC have employed azathioprine, cyclosporin, methotrexate, prednisolone, chlorambucil, thalidomide and, more recently, budesonide and mycophenolate mofetil. Neither of the two trials of azathioprine showed a beneficial effect of this drug on survival [62, 63]. The first trial had an inadequate sample size, lacked a placebo control group and predetermined stopping rules, and the results were not analyzed according to intention to treat. The second trial of azathioprine although much larger (248 patients), did not include a sample size calculation to assure that it had adequate power, and the withdrawal rate was greater than 20% in both the azathioprine and placebo groups. Despite randomization, the two groups were not stratified for factors known to influence survival and were not comparable at baseline. Only after employment of the Cox multiple regression analysis to adjust for these baseline differences was a benefit of treatment on survival observed. This difference in survival between the treatment groups could only be measured in months, which may not be clinically meaningful. Patients were followed for up to ten years, but the number of patients still being followed at that period was only nine in the azathioprine and none in the placebo group. It appears that the intention to treat principle was not used in the analysis, since 32 patients were excluded from the analysis because of incomplete data. Thus, the validity of the small benefit in survival can be further questioned. A1d Mycophenolate mofetil similarly has been disappointing, with no important clinical benefits, albeit in data from a single, small pilot study of 25 patients with PBC lasting one year [64].
Several small trials and one large trial (349 patients) of cyclosporin therapy have been published [65–68]. This trial ran into the same problems as had been encountered with the large azathioprine trial, that is, lack of comparability of the two treatment groups at baseline. Even though there were a similar number of deaths, 30 in the cyclosporin group and 31 in the placebo group, the authors concluded that survival was improved in the cyclosporin-treated patients. Renal impairment was observed in 9% and systemic hypertension in 11% of the cyclosporin-treated patients (1.7% and 1% in placebo-treated respectively). These two serious adverse effects in patients whose disease is relatively slowly progressive precludes the use of this drug in the long-term treatment of PBC. A1d Tacrolimus, a similar agent, has not been studied to date formally in PBC, neither has the other increasingly popular immuno-suppressant, rapamycin.
Methotrexate was suggested to be of value in pilot studies [69, 70]. In one RCT of therapy sixty patients were recruited, 30 randomized to low-dose therapy (2.5mg three times per week). At the end of six years, the serum bilirubin and Mayo risk score were higher in those receiving methotrexate, suggesting that the drug may be toxic in patients with PBC [71]. In a large NIH effort (placebo-controlled, randomized, multicenter trial comparing the effects of methotrexate plus UDCA to UDCA alone on the course of PBC) methotrexate when added to UDCA for a median period of 7.6 years had no effect on the course of PBC treated with UDCA alone [72]. A1d This study illustrates some of the difficulties of trial design: at the time of planning, it was anticipated that the five-year, transplant-free survival would be 83%, whereas the experience in this trial corresponded instead to a 94.5% five-year and 84.3% eight-year transplant free survival probability. The lower event rate meant that patients had to be followed for a longer time period, and this in turn probably contributed to a higher rate of patients discontinuing methotrexate. The authors calculated, using the log rank statistic to compare the distributions of time to an event in a randomized trial, that they needed to observe 88 events in order to have 90% power to detect a two-fold improvement in the risk of an event. With their calculations in a study that followed subjects for an average of eight years (with an expected 7% rate of death and 16% rate of transplant or death), this implied they needed to randomize a total of 1257 subjects when using the endpoint of overall survival, or 550 subjects when using the endpoint of transplant-free survival. If one then appreciates their recruitment reality, the enormous challenges for investigators planning future studies is clear: in this case investigators at 12 centers in the USA screened 535 patients for possible entry into the trial. At screening 393 of the 535 patients were determined to meet the defined inclusion criteria, and 385 patients went on to meet the screening inclusion and exclusion criteria. Three hundred patients progressed to a pre-entry evaluation phase and ultimately 265 patients were randomized with equal probability in a double-blind fashion.
A small, three-year RCT of prednisolone has been done [73]. A1d A significant reduction in the serum bilirubin level (a valid surrogate outcome measure) was observed in treated patients, but osteoporosis in those who received corticosteroids worsened. However, a trial with a very small sample size of bisphosphonates in patients with PBC treated with corticosteroids indicates that etidronate significantly stabilizes bone mineral density in vertebrae of patients with PBC receiving corticosteroids [74]. Hence, it may be appropriate that corticosteroids be re-evaluated in the therapy of PBC now that patients can be given preventive therapies to reduce the complication of osteoporosis. Budesonide is an oral corticosteroid which is eliminated on first pass through the liver. Thus it was hoped that this agent could benefit patients with PBC without having a deleterious effect on bone mineral density. In the first small RCT by Leuschner et al., improvement in liver biochemistry, IgM values and liver histology was observed in the few patients studied, most of whom had very early disease [75]. In a second study by Angulo et al., no benefit was observed and the Mayo risk score increased significantly in those randomized to budesonide for a year and bone mineral density measurements deteriorated in the lumbar spine (p < 0.01); it is likely that many of these patients had advanced PBC so that the benefit of the first-pass effect was lost [76]. Subsequent work from Finland has alluded to a histological benefit when combined with UDCA and the observation that interface hepatitis on baseline biopsy is predictive of outcome in PBC (as well as a failure to normalize transaminases with UDCA) does suggest that a subgroup may benefit from immunosuppression [77]. A1d Interest therefore remains in using budesonide adjunc-tively in those with more florid hepatitic features. However, Wolfhagen et al. in another small RCT involving 50 patients with PBC with suboptimal response to UDCA, treated patients with additional prednisone (30mg/day tapered to 10mg/day) and azathioprine (50mg/day) [78]. There was no improvement in bilirubin. The trial was too small and of too short a duration to examine the effect on survival. A1d
A small study of 13 patients randomized to 0.5–4 mg daily of chlorambucil (mean 2mg daily) compared with placebo has been reported [79]. All treated patients developed some degree of bone marrow suppression and discontinuation of therapy was required in four. A 30% withdrawal rate due to drug toxicity indicates that this drug should not undergo further evaluation in patients with PBC. A1d A very small and short (six-month) RCT of 18 PBC patients taking thalidomide has been reported, showing little benefit of this treatment [80]. However, this study lacked adequate power to evaluate this form of therapy in any meaningful way. No benefit on serum bilirubin was observed during the six months of treatment.
An alternative approach has been to use drugs that may not interfere with the primary cause, that is, immune-mediated bile duct destruction, but interfere with the progression of the disease, either by reducing fibrogenesis or by reducing cholestasis. Two potential antifibrotic drugs have been assessed: colchicine in four good studies [81–84] and D-penicillamine in many more [85–92]. The studies of colchicine therapy are interesting, but unfortunately all the size of the populations studied are small. A Cochrane Hepato-Biliary Group systematic review of randomized clinical trials for this agent included ten trials involving 631 patients, only four of which were high-quality [93]. No significant differences were detected between colchicine and placebo/no intervention regarding mortality (relative risk (RR), 1.21; 95% CI: 0.71–2.06), mortality or liver transplantation (RR = 1.00; 95% CI: 0.67–1.49), liver complications, liver biochemical variables, liver histology, or adverse events. A1c While the numerous trials of D-penicillamine demonstrated great interest, the results are also disappointing. Unlike patients with Wilson’s disease, adverse effects of therapy were particularly common, resulting in a high withdrawal rate, similar to the experience with rheumatoid arthritis. This drug is not recommended for the treatment of PBC and a meta-analysis is consistent with this (including seven trials randomizing 706 patients) [94]. A1c
Reduction of cholestasis: ursodeoxycholic acid
The prevailing view remains that UDCA improves the natural history of PBC [95], although the evidence for efficacy has been cumulative, is imperfect, and has been appropriately challenged as dogma [96, 97]. A1a Japanese researchers first isolated UDCA from bear bile in 1927; bear bile itself having a track record in traditional medicine as a choleretic [98]. UDCA is the natural intestinal metabolite of chenodeoxycholic acid (7-ß epimer). In the 1960s patients with gallstones were found to have lower biliary levels of cholic acid and chenodeoxycholic acid, which led to trials of oral bile salt supplementation in hepatobiliary disease. UDCA is less hydrophobic and more hydrophilic than other bile acids, with apparently less toxicity towards the hepatobiliary epithelium. Its therapeutic mechanism of action includes expansion of the hydrophilic bile acid pool, a direct choleretic effect (administration results in a bicarbonate-rich choleresis and modulation of biliary transporter proteins), anti-inflammatory effects (UDCA interacts with the glucocorticoid receptor in vitro) and anti-apoptotic effects on hepatic epithelia [99]. Cessation of therapy with UDCA leads to a return of abnormal liver biochemistry in those with PBC who respond, demonstrating that the initiating and perpetuating factors of PBC remain unresolved. UDCA normally accounts for about 4% of bile acids but with pharmacotherapy UDCA becomes the predominant bile acid. Leuschner showed that when treating gallstones with UDCA in those with histologically confirmed HBsAg-negative chronic active hepatitis a fall in transaminases occurred [100]. In further work to investigate the dose-response relationship for UDCA, three different doses were given to patients with PBC, PSC and chronic hepatitis. This allowed Podda et al. to show that significant falls in liver biochemistry occurred at dosages of 4–5mg/kg daily, but higher doses (8–10mg/kg and 12–15mg/kg) induced further improvements, with the changes correlated approximately with the enrichment of bile by UDCA [101]. Subsequent studies confirmed a correlation between the degree of bile enrichment and improvement in liver biochemistries and PBC Mayo risk score, and overall in PBC the data suggest that the optimum dose of UDCA if used should be 13 to 15mg/kg per day, given as a single daily dose [40, 55,102,103]. UDCA at this dose is very safe, with minimal side effects: weight gain (approx. 3 kg in the first 12 months [104]), hair loss and, rarely, diarrhea and flatulence. Additionally, nothing suggests UDCA is tera-togenic but it is usually stopped before and during the first trimester, and restarted subsequently, with there being a good safety profile from its use in intrahepatic cholestasis of pregnancy [105].
Justification for the widespread use of UDCA
Many studies have attempted to demonstrate clinical efficacy for UDCA, particularly following initial studies showing a fall in bilirubin levels in patients with PBC, with most trials showing beneficial effects of UDCA on biochemical parameters in particular. With such a slow natural history any individual trial will inevitably lack the power to address endpoints such as death or liver transplantation. Additional criticism can be made for assuming that every patient benefits equally, that is, identifying and treating more patients with mild disease may be self-fulfilling if those patients were never destined to progress. Moreover whilst appealing to assume that the natural history of the disease is unchanged, the absence of a clear cut patho-physiologic understanding means that such an assumption should not necessarily be considered correct.
The cumulative evidence that convinces most to prescribe UDCA, has nevertheless come from individual and pooled analysis. Of note the Mayo Clinic followed up patients with PBC treated with UDCA and have demonstrated a treatment-related prolongation in transplant free survival [106–108]. French cohorts with long-term follow-up similarly demonstrate survival benefit [109], and this was reiterated and furthered by a Spanish cohort including nearly 200 patients treated for a mean of 6.7 years [47]. Pathophysiologically early disease is most likely to respond to intervention before duct loss is established and fibrosis present. In those with early stage disease, a biochemical response to UDCA (normalization or a 40% reduction from baseline) after one year was associated with a similar survival to the matched control population. In the Netherlands ter Borg and colleagues were able to show a survival benefit without transplant for low-risk patients compared to survival predicted by the Mayo risk model (which, as discussed, may not be a perfect risk score to employ in this case if the comparison group has predominantly early stage disease, as survival may be underestimated); moreover overall survival was nearly the same as for a control population [110]. B4 A UK study that examined a large cohort not receiving UDCA did not, however, demonstrate differences in outcomes between treated and untreated patients [111]. A recent publication, as discussed, looked at defining the biochemical response a year after treatment, to identify UDCA-treated patients at risk of death or transplant and showed how UDCA responders had a ten-year, transplant-free survival rate of 90%, compared to 51% for those who did not [48]. B4
Related posts:
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
