Patients and Methods
Study Design and Study Population
This study was a meta-analysis performed by the Global PBC Study Group, an international and multicenter collaboration between 15 liver centers in 8 North American and European countries, which combined individual patient data from major long-term follow-up cohorts. Most individual databases contained prospectively collected follow-up data on patients starting UDCA therapy.
This study was conducted in accordance with the protocol and the principles of the Declaration of Helsinki. The protocol was reviewed and approved by the institutional research board of the corresponding center and at each participating center in accordance with local regulations.
Both UDCA-treated and nontreated patients with an established diagnosis of PBC in accordance with European and American guidelines were eligible for inclusion in this study. Patients were excluded from analysis if follow-up data were insufficient or unavailable, the start date of treatment or the exact date of major clinical events was unknown, or they had concomitant liver disease.
Data Collection and Quality Assessment
Collected clinical and laboratory data included sex, age, diagnosis of PBC, liver histology, treatment (type of medication, dosage, and duration), duration and last date of follow-up, baseline antimitochondrial antibody status, baseline and yearly laboratory levels (serum alkaline phosphatase, total bilirubin, albumin, aspartate aminotransferase, alanine aminotransferase, γ-glutamyl transpeptidase, and platelets), and outcomes (death and cause of death, liver transplantation, hepatocellular carcinoma, ascites, and variceal bleeding).
Liver histology performed within 1 year of study entry or documented cirrhosis before study entry was classified as a baseline biopsy. Histological data was assessed for severity according to Ludwig and Scheuer's classification. Disease stage was classified histologically as early (stage I and II) or late (stage III or IV) and biochemically using serum albumin and bilirubin levels. According to this biochemical classification, early stage was defined by normal bilirubin and albumin levels, moderately advanced disease was defined by an abnormal bilirubin or albumin level, and advanced disease was defined by abnormal bilirubin and albumin levels.
Completeness, plausibility, and validity of the data were carefully verified. Extensive efforts, including site visits with review of medical charts, were undertaken to retrieve missing data. Data of the original cohorts were collected through the end of December 2012.
Statistical Analysis
Study entry was defined as the start date of UDCA therapy or the date of the first center visit for patients not treated with UDCA. The primary end point was defined as a composite of either liver transplantation or death. Patients without documented events during follow-up were censored at their last follow-up visit.
To study the association between the absolute alkaline phosphatase and bilirubin levels, the hazard ratios (HRs) of liver transplantation or death were estimated by applying a cubic spline function of alkaline phosphatase and bilirubin at baseline and yearly up to 5 years of follow-up.
To find an optimal threshold for each variable, alkaline phosphatase and bilirubin levels at 1 year of follow-up were categorized according to multiple thresholds ranging from 1.0 to 3.0 times the upper limit of normal (ULN) in steps of 0.1 (including 1.67 times the ULN for alkaline phosphatase levels). The C statistic was calculated for each of these thresholds to evaluate their ability to predict liver transplant-free survival. Accompanying HRs were calculated for each threshold by using the Cox proportional hazard regression model. The log-likelihood test was used to assess significance. Transplant-free survival was assessed for the peak thresholds of alkaline phosphatase and bilirubin levels and for a combination of both by Kaplan–Meier estimates. Log-rank test was used for comparisons between groups.
In addition to the predictive ability of absolute levels of alkaline phosphatase, the percentage change in alkaline phosphatase levels from baseline to 1-year follow-up was evaluated using the same approach.
All analyses were stratified by center to account for possible heterogeneity across center populations. The effects of alkaline phosphatase and bilirubin were adjusted for year of diagnosis, age at study entry, UDCA therapy, and sex.
To investigate if alkaline phosphatase and bilirubin levels are meaningful surrogate end points, the association with the clinical end point must hold true independent of time and specific patient subgroups. Therefore, the survival analyses were repeated at different time points and for multiple subgroups of patients. The time points analyzed were baseline and yearly up to 5 years of follow-up. Given the nature of this study, alkaline phosphatase or bilirubin levels were not always available for every patient at these time points. Accordingly, we aimed for the optimal use of the available data by assessing the association with hard clinical end points at baseline and several intervals thereafter up to 5 years. Subgroups were defined by treatment (UDCA-treated and nontreated patients), baseline alkaline phosphatase levels (>2.0 times the ULN and >4.0 times the ULN), baseline bilirubin levels (>1 times the ULN and >3 times the ULN), PBC disease state based on both histology and biochemistry, age at time of diagnosis (younger than 45 years and 45 years or older), sex, and date of diagnosis (before 1990, 1990–1999, and 2000–2009).
Normally distributed data are presented as mean ± SD and skewed distributed data as median and interquartile range. All analyses were 2 sided. P < .05 was considered statistically significant. Statistical analyses were performed with IBM SPSS Statistics 22.0 (SPSS Inc, Chicago, IL) and SAS 9.3 (SAS institute, Cary, NC).