Alkaline Phosphatase, Bilirubin Levels and Outcomes in PBC
This study reports a robust and uniquely powered, independent evaluation of the largest meta-analysis of individual data on PBC to date. We unequivocally show that both increased serum alkaline phosphatase and bilirubin levels are strongly associated with reduced transplant-free survival in patients with PBC and that a combination of both variables improves prognostic prediction for patients. These associations are independent of use of UDCA and follow-up time and held for multiple subgroups. These data support that both alkaline phosphatase and bilirubin provide meaningful surrogate end points in PBC that can reasonably be used in clinical trials.
Prior studies have shown an association between normalization, percentage decreases or absolute decreases of alkaline phosphatase levels and improved prognosis with UDCA therapy. The present study reports for the first time a near log-linear association between alkaline phosphatase levels and transplant-free survival and clearly shows that the lower the alkaline phosphatase value, the greater the transplant-free survival time. This applied not only to alkaline phosphatase levels during follow-up but also to baseline levels irrespective of subsequent UDCA therapy. The suitability of alkaline phosphatase as a surrogate end point for clinical benefit is further supported by the finding that the prognostic information provided by alkaline phosphatase levels was confirmed across a wide range of subgroups such as nontreated patients, relatively young patients, and patients with histologically early and late disease. This finding is of considerable clinical significance because alkaline phosphatase constitutes one of the 3 potential diagnostic criteria and is used routinely to assess disease activity.
Our study additionally confirms that as baseline bilirubin levels or bilirubin levels increase over time, the likelihood of survival correspondingly decreases. The predictive ability of alkaline phosphatase levels was shown in addition to bilirubin to discriminate high-risk and low-risk patients. This is an important observation because bilirubin on its own is unsuitable as a surrogate end point in clinical trials because it is typically elevated only when the disease has progressed to the stage at which liver function becomes impaired. Most patients likely to be included in such studies will have normal levels precluding the possibility of observing potential beneficial treatment effects based on bilirubin alone.
It has been suggested that the best way to evaluate the utility of a biomarker as a good surrogate end point may be a meta-analysis of clinical trials of one or more interventions. A 4-level hierarchy of evidence to consider the validation of surrogate end points has been proposed:
Level 1: a true clinical-efficacy measure; Level 2: a validated surrogate endpoint (for a specific disease setting and class of interventions); Level 3: a non-validated surrogate endpoint, yet one established to be "reasonably likely to predict clinical benefit" (for a specific disease setting and class of interventions); Level 4: a correlate that is a measure of biological activity but that has not been established to be a higher level.
The particular challenge of confirming biomarkers as surrogate end points in PBC is that there is only one approved treatment, and previous meta-analyses of published clinical trials that have been conducted in PBC have been interpreted in conflicting ways. Interpretation of the data is compromised due to design issues, such as a lack of consistent long-term follow-up. Our approach was therefore to conduct a more rigorous patient-level meta-analysis of existing cohorts of patients at centers across North America and Europe with long-term follow-up data of large numbers of patients with PBC. This design has sufficient power to intensively study alkaline phosphatase and bilirubin as potential surrogate end points in different settings, subpopulations, and time points. Based on these current results, we postulate that alkaline phosphatase and bilirubin levels are "reasonably likely to predict clinical benefit" in PBC. This is of relevance to future trial design for new therapeutic agents.
Alternative surrogates have been suggested, such as liver histology, which may provide key information on treatment effects in PBC. However, liver biopsy is not routinely conducted in patients with PBC. Given its invasive nature and small but well-recognized risks, liver histology, with its added inherent sampling variability, is not an ideal surrogate for widespread use in patients with PBC. Noninvasive elastography-based assessment of liver fibrosis may potentially be used as a reliable alternative in the prediction of fibrosis; however, further long-term evaluation is required in PBC. Similarly other biochemical surrogates have been suggested but as of yet are not widely studied. We focused on the routine biochemical measurements that have been used for many years in both the diagnosis and management of patients with PBC, because this approach is likely to be the most easily applied in practice.
There are some limitations to our study. The availability of some clinical data (such as ascites, edema, pruritus, fatigue, or use of diuretics) and laboratory data (including prothrombin time, immunoglobulin M, and immunoglobulin G levels) in the individual databases varied considerably. In many cases, in particular when databases contained data of patients entered more than 10 to 20 years ago, it was not possible to collect these data consistently in a reliable way. Further, no uniform or generally accepted or validated methods had been used in the contributing centers to quantify subjective signs and symptoms. As a consequence, within the context of this study, we were unfortunately unable to include this type of information in our analyses and, in particular, were not able to calculate the Mayo risk score and to compare the prognostic information provided by this established prediction tool with that provided by alkaline phosphatase and bilirubin.
Due to the nature of our study, biochemical data were not always available at the fixed time points during follow-up. This was mainly encountered when the original data had been collected more than 20 years ago. Data on dose changes or interruption of UDCA therapy was also not uniformly available. However, we believe that these limitations had no major impact on the reliability of the results, considering the unique large size of the study population, the prospective nature of most of the data, the inclusion of both UDCA-treated and nontreated patients, the substantial incidence of clinical end points, and the duration of follow-up. Additionally, adjusting for missing data by multiple imputations of the data, the results did not change (Supplementary Table 6).
Based on our present results, any decrease in alkaline phosphatase or bilirubin levels translates into improved prognosis; lower levels are clearly associated with better transplant-free survival. In our population, the most discriminative alkaline phosphatase threshold after 1 year of follow-up was 2.0 times the ULN, which is an earlier proposed threshold, although an alkaline phosphatase threshold of 1.5 times the ULN, 1.67 times the ULN, or 3.0 times ULN would all work well as a surrogate end point in a clinical trial setting. For bilirubin, the choice of threshold is even clearer; the spline plots (Figure 1) suggest that a choice of bilirubin <1.0 times the ULN is reasonable. However, designing clinical trials implies the a priori requirement to estimate the quantitative effect of an experimental intervention on a given end point. Based on the current study, we are not able to translate these data into a specific threshold for a clinical trial in general.
In conclusion, our study shows that alkaline phosphatase and bilirubin levels strongly correlate with the ultimate outcomes of death and liver transplantation in patients with PBC; the lower the alkaline phosphatase and bilirubin levels the better the transplant-free survival times. This robust analysis suggests that these variables can reasonably be regarded as useful surrogate end points in clinical trials. There is a high unmet medical need for new therapies for this rare autoimmune liver disease, and this study provides an important impetus for the selection of appropriate end points and to facilitate the conduct of meaningful therapeutic intervention trials in the absence of long-term outcome studies.
Discussion
This study reports a robust and uniquely powered, independent evaluation of the largest meta-analysis of individual data on PBC to date. We unequivocally show that both increased serum alkaline phosphatase and bilirubin levels are strongly associated with reduced transplant-free survival in patients with PBC and that a combination of both variables improves prognostic prediction for patients. These associations are independent of use of UDCA and follow-up time and held for multiple subgroups. These data support that both alkaline phosphatase and bilirubin provide meaningful surrogate end points in PBC that can reasonably be used in clinical trials.
Prior studies have shown an association between normalization, percentage decreases or absolute decreases of alkaline phosphatase levels and improved prognosis with UDCA therapy. The present study reports for the first time a near log-linear association between alkaline phosphatase levels and transplant-free survival and clearly shows that the lower the alkaline phosphatase value, the greater the transplant-free survival time. This applied not only to alkaline phosphatase levels during follow-up but also to baseline levels irrespective of subsequent UDCA therapy. The suitability of alkaline phosphatase as a surrogate end point for clinical benefit is further supported by the finding that the prognostic information provided by alkaline phosphatase levels was confirmed across a wide range of subgroups such as nontreated patients, relatively young patients, and patients with histologically early and late disease. This finding is of considerable clinical significance because alkaline phosphatase constitutes one of the 3 potential diagnostic criteria and is used routinely to assess disease activity.
Our study additionally confirms that as baseline bilirubin levels or bilirubin levels increase over time, the likelihood of survival correspondingly decreases. The predictive ability of alkaline phosphatase levels was shown in addition to bilirubin to discriminate high-risk and low-risk patients. This is an important observation because bilirubin on its own is unsuitable as a surrogate end point in clinical trials because it is typically elevated only when the disease has progressed to the stage at which liver function becomes impaired. Most patients likely to be included in such studies will have normal levels precluding the possibility of observing potential beneficial treatment effects based on bilirubin alone.
It has been suggested that the best way to evaluate the utility of a biomarker as a good surrogate end point may be a meta-analysis of clinical trials of one or more interventions. A 4-level hierarchy of evidence to consider the validation of surrogate end points has been proposed:
Level 1: a true clinical-efficacy measure; Level 2: a validated surrogate endpoint (for a specific disease setting and class of interventions); Level 3: a non-validated surrogate endpoint, yet one established to be "reasonably likely to predict clinical benefit" (for a specific disease setting and class of interventions); Level 4: a correlate that is a measure of biological activity but that has not been established to be a higher level.
The particular challenge of confirming biomarkers as surrogate end points in PBC is that there is only one approved treatment, and previous meta-analyses of published clinical trials that have been conducted in PBC have been interpreted in conflicting ways. Interpretation of the data is compromised due to design issues, such as a lack of consistent long-term follow-up. Our approach was therefore to conduct a more rigorous patient-level meta-analysis of existing cohorts of patients at centers across North America and Europe with long-term follow-up data of large numbers of patients with PBC. This design has sufficient power to intensively study alkaline phosphatase and bilirubin as potential surrogate end points in different settings, subpopulations, and time points. Based on these current results, we postulate that alkaline phosphatase and bilirubin levels are "reasonably likely to predict clinical benefit" in PBC. This is of relevance to future trial design for new therapeutic agents.
Alternative surrogates have been suggested, such as liver histology, which may provide key information on treatment effects in PBC. However, liver biopsy is not routinely conducted in patients with PBC. Given its invasive nature and small but well-recognized risks, liver histology, with its added inherent sampling variability, is not an ideal surrogate for widespread use in patients with PBC. Noninvasive elastography-based assessment of liver fibrosis may potentially be used as a reliable alternative in the prediction of fibrosis; however, further long-term evaluation is required in PBC. Similarly other biochemical surrogates have been suggested but as of yet are not widely studied. We focused on the routine biochemical measurements that have been used for many years in both the diagnosis and management of patients with PBC, because this approach is likely to be the most easily applied in practice.
There are some limitations to our study. The availability of some clinical data (such as ascites, edema, pruritus, fatigue, or use of diuretics) and laboratory data (including prothrombin time, immunoglobulin M, and immunoglobulin G levels) in the individual databases varied considerably. In many cases, in particular when databases contained data of patients entered more than 10 to 20 years ago, it was not possible to collect these data consistently in a reliable way. Further, no uniform or generally accepted or validated methods had been used in the contributing centers to quantify subjective signs and symptoms. As a consequence, within the context of this study, we were unfortunately unable to include this type of information in our analyses and, in particular, were not able to calculate the Mayo risk score and to compare the prognostic information provided by this established prediction tool with that provided by alkaline phosphatase and bilirubin.
Due to the nature of our study, biochemical data were not always available at the fixed time points during follow-up. This was mainly encountered when the original data had been collected more than 20 years ago. Data on dose changes or interruption of UDCA therapy was also not uniformly available. However, we believe that these limitations had no major impact on the reliability of the results, considering the unique large size of the study population, the prospective nature of most of the data, the inclusion of both UDCA-treated and nontreated patients, the substantial incidence of clinical end points, and the duration of follow-up. Additionally, adjusting for missing data by multiple imputations of the data, the results did not change (Supplementary Table 6).
Based on our present results, any decrease in alkaline phosphatase or bilirubin levels translates into improved prognosis; lower levels are clearly associated with better transplant-free survival. In our population, the most discriminative alkaline phosphatase threshold after 1 year of follow-up was 2.0 times the ULN, which is an earlier proposed threshold, although an alkaline phosphatase threshold of 1.5 times the ULN, 1.67 times the ULN, or 3.0 times ULN would all work well as a surrogate end point in a clinical trial setting. For bilirubin, the choice of threshold is even clearer; the spline plots (Figure 1) suggest that a choice of bilirubin <1.0 times the ULN is reasonable. However, designing clinical trials implies the a priori requirement to estimate the quantitative effect of an experimental intervention on a given end point. Based on the current study, we are not able to translate these data into a specific threshold for a clinical trial in general.
In conclusion, our study shows that alkaline phosphatase and bilirubin levels strongly correlate with the ultimate outcomes of death and liver transplantation in patients with PBC; the lower the alkaline phosphatase and bilirubin levels the better the transplant-free survival times. This robust analysis suggests that these variables can reasonably be regarded as useful surrogate end points in clinical trials. There is a high unmet medical need for new therapies for this rare autoimmune liver disease, and this study provides an important impetus for the selection of appropriate end points and to facilitate the conduct of meaningful therapeutic intervention trials in the absence of long-term outcome studies.
SHARE
