PBC Predict: Prediction models and predictive biomarkers for primary biliary cholangitis

Primary biliary cholangitis (PBC) is an autoimmune liver disease in which the body's own immune system causes damage to the small bile ducts inside the liver, eventually leading to cirrhosis (scarring of the liver). Current guidelines on the management of PBC recommend a 'step-up' approach to treatment, meaning that all patients start treatment with a medication called ursodeoxycholic acid (UDCA), which is only moderately effective but has few side effects. If the disease does not settle on UDCA alone, medication is added that is more effective - but also more likely to cause side effects. The obvious problem with this approach is that patients who need more effective treatment (i.e. those with more aggressive disease) end up waiting longer to receive it. On the other hand, 'step-down' treatment, in which everyone starts with more effective medication, would cause many patients to have unnecessary exposure to side effects. In many patients with PBC, the disease is not controlled by any of the medications that are currently available. These patients have increased risk of developing cirrhosis and needing a liver transplant. PBC is therefore a major cause of death related to liver disease, and one of the main reasons for having a liver transplant in the UK. Thus, new medications are needed for patients who do not respond to currently available medications - and accurate methods (e.g. new blood tests) are needed to predict, at the earliest possible stage, which patient will need more effective medication, and which medication will work best in that patient. This is the essence of precision medicine: to give the right medication to the right patient at the right time.

UK-PBC is an ambitious collaboration between Universities, Drug Companies and the NHS to develop precision medicine for PBC. It was started with funding from the Medical Research Council (MRC). As part of UK-PBC, my team established the UK-PBC Research Cohort ('Cohort'), which already consists of more than 6,000 patients with PBC, each providing a DNA sample and consenting for clinical information to be collected from any NHS organisation. We have previously used clinical information about patients in the Cohort to develop statistical models that predict response to first-line treatment with UDCA, or future need for liver transplantation. We have used genetic information about the Cohort for large-scale genetic studies that have highlighted drugs, currently used for treatment of other autoimmune conditions, that might also work in PBC.

In the proposed project, building on our previous success, we will expand the Cohort to more than 10,000 patients with PBC. We will collect large amounts of clinical information about these patients from a range of NHS organisations. We will use this clinical information to improve the accuracy of statistical models that predict response to treatment or future need for liver transplantation, which may then be used in clinical practice to tailor the treatment of PBC patients. We will combine clinical and genetic information to identify the genetic factors that determine response to treatment or future need for liver transplantation, and develop statistical models incorporating clinical and genetic information. We will measure gene regulation and gene expression in blood-based immune cells from a carefully selected group of newly diagnosed patients, and use these measurements to develop new blood tests that predict, at the earliest possible stage, which patients will need more effective medication. Finally, we will combine clinical and genetic information with measurements of gene regulation and gene expression to pinpoint the genes involved in causing PBC, which will help us to identify new medications that might work in this condition.

Primary biliary cholangitis (PBC) is an autoimmune liver disease that eventually leads to cirrhosis. Current guidelines recommend a step-up approach to treatment, whereby first-line therapy must fail before second-line agents are considered. Patients with more aggressive disease therefore wait longer to receive effective therapy. Many patients have inadequate response to all available medications, leaving them at risk of progressive liver disease. PBC is therefore a major cause of liver-related morbidity and mortality. There is a critical need in PBC for novel agents, and for simple, robust approaches (especially biomarkers) for early risk and treatment stratification, so that medications may be allocated at diagnosis based on the patient's predicted need for treatment and his or her likelihood of response. This is the essence of precision medicine.

UK-PBC is an ambitious collaboration between Academia, Industry and the NHS to develop precision medicine for PBC. In the proposed project, building on previous success, we will expand the existing, unique UK-PBC Research Cohort to more than 10,000 participants; collect maximal clinical data about participants from diverse NHS sources; and use these data to develop accurate clinical prediction models of treatment response and disease progression, suitable for risk and treatment stratification in clinical practice. We will undertake genotype-phenotype analysis to delineate the genetic basis of this phenotypic variation, and develop genetic risk scores and predictive models incorporating clinical and genetic variables. Finally, we will undertake epigenetic and transcriptional profiling of PBMC subsets from an inception cohort of treatment naïve patients to discover epigenetic and transcriptional biomarkers for risk and treatment stratification. Via integrated analysis of genetic, epigenetic and transcriptional datasets, we will pinpoint causal variants, genes and pathways to guide re-purposing of drugs.

1. Accurate risk prediction and risk stratification

In the proposed project, we will develop accurate clinical prediction models enabling precise phenotypic, risk and treatment stratification in PBC. This will have impact, as follows:

(a) We previously developed the UK-PBC Risk Score, a multivariable Cox model that predicts future need for liver transplantation. Current guidelines recommend use of the Risk Score for risk stratification in PBC - but it has important limitations. For example, it is not applicable to disease sub-phenotypes, and it predicts low risk more accurately than high risk. In the proposed project, to overcome these limitations, we will expand the existing, unique UK-PBC Research Cohort; obtain maximal clinical data about participants, and use statistical approaches that take account of disease sub-phenotype; the trajectory of treatment response, and intermediate disease states to improve the accuracy of clinical prediction models. These will have major impact by enabling more precise risk stratification, so that patients at higher risk may be prioritised for second-line treatment and closer monitoring, and those at lower risk may be reassured and spared unnecessary investigations and treatment;

(b) Current guidelines recommend a step-up approach to treatment, whereby first-line therapy must fail before second-line agents are considered. Patients with more aggressive disease therefore wait longer to receive effective therapy. The obvious disadvantage of this approach is that patients in greatest need of more effective treatment end up waiting longest to receive it. In the proposed project, building on our previous work, we will develop accurate clinical prediction models and identify predictive biomarkers to improve prediction of treatment response. This will have major impact, enabling baseline risk stratification and a change in approach from step-up treatment to stratified treatment from diagnosis.

2. Statistical models of disease to guide biomarker discovery and clinical trials

Statistical modelling of disease, as described in this proposal, will have impact on biomarker discovery by enabling the identification of disease sub-groups for cellular and molecular characterisation. Similarly, it will have impact on clinical trial design by, for example, influencing the choice of eligibility criteria and outcome measures.

3. Health economics

The proposed project will generate a large, prospective cohort that is well-characterized in terms of disease severity, healthcare utilization, symptoms and health utility. These invaluable data will be available for health economic analysis by Prof Luke Vale and his team at Newcastle University. Health economic analysis based on the UK-PBC Research Cohort will have major impact by determining the cost-effectiveness of different treatment strategies.

4. A re-callable cohort

The proposed project will establish a large, well-characterised cohort consented for recall to other studies, which will have major impact by facilitating recruitment to future observational or interventional studies.

5. Real-world data capture for second-line treatment

Obeticholic acid (OCA), the only drug licenced for treatment of PBC in more than 20 years, received conditional approval from the FDA based on improvement in the serum alkaline phosphatase (ALP) and bilirubin, accepted to be surrogate markers of long-term outcome. NHS England has mandated enrolment in the UK-PBC Research Cohort of patients starting second-line treatment with obeticholic acid or fibrates, with prospective follow-up to determine the impact of these drugs on long-term outcomes. The proposed project will therefore have impact by confirming that improvement in the ALP and bilirubin translates in the real-world setting to improvement in transplant-free survival. This is applicable to future agents with conditional approval based on improvement in liver biochemistry.