Understanding the outcomes of patients with cirrhosis and hepatocellular carcinoma (HCC) in the England ( ODR1718_432 )

Cirrhosis is a major risk factor for the development of hepatocellular carcinoma (HCC). Diagnosis and treatment of HCC in the setting of cirrhosis is complex and depends not only of the stage of cancer but also on the severity of underlying cirrhosis and other patient factors

Objectives:

1. Which baseline factors determine treatment allocation and overall survival in HCC?

2. How is HCC treatment allocation associated with overall survival and decompensation of cirrhosis?

3. What is the nature of regional variation in baseline factors and treatment allocation in HCC?

Project aim – include hypotheses if relevant

To exploit routine health data to investigate how variation in baseline characteristics at HCC diagnosis and geographical location influence treatment allocation and clinical outcomes. In addition to overall survival, the rate of progression of underlying liver disease will also be investigated, to understand the competing mortality of HCC and cirrhosis.

HCC will be defined using ICD-10 code C22.0. The cohort of patients will be defined by those patients with a new diagnosis of HCC between 01/01/2007 and 31/12/2016. This dataset will be linked to the Hospital Episode Statistics (HES) database. A HES extract will be obtained containing information on Finished Consultant Episodes (FCEs) for those individuals in the HCC cohort. This extract period will start from 5 years prior to the HCC diagnosis date and continue until death or the end of the study period.

A focussed HES extract will be retrieved and analysed in order to achieve the specific aims of the study, as described in the following sections:

Which baseline factors determine treatment allocation and overall survival in HCC?

In order to assess the baseline factors, the following non-identifiable data will be extracted:

Age at diagnosis in 5 year age bands (from Cancer Registry)

Sex (from Cancer Registry)

Ethnic category and broad ethnic group (from Inpatient HES).

This is required because it may be associated with some aetiologies of liver disease, including viral hepatitis which is more common in migrant populations.

Cancer stage (using “Best ‘registry’ stage at diagnosis of the tumour” from Cancer Registry)

There is an expectation that the data quality of HCC cancer stage is poor due missing values, but this will be included to comparison purposes.

Medical co-morbidities

Diagnostic codes related to medical comorbidities contained within Inpatient HES will be extracted, along with the time interval from HCC diagnosis date to the start of the associated episode. In addition to Inpatient HES Charlson Index, codes for individual co-morbidities will be extracted because they impact the progression of liver disease.

Aetiology of underlying liver disease

Specific diagnostic codes relating to different liver disease aetiologies will be extracted from episodes contained within the total study period.

Presence of cirrhosis at HCC diagnosis

Specific diagnostic and procedure codes relating to cirrhosis or its complications (ascites, oesophageal or gastric varices and hepatic encephalopathy) will be extracted from episodes within the total study period. The time interval from HCC diagnosis to the start of the associated episode will be extracted. If these codes appear at any point during the study, it will be assumed that the HCC occurred in the background of cirrhosis.

Cirrhosis stage at HCC diagnosis

Diagnosis and procedure codes specific to cirrhosis-related complications (along with the time interval from HCC diagnosis to the start of the associated episode) will be extracted from Inpatient HES. Analysing a time interval from 5 years before HCC diagnosis to 3 months after will enable the calculation of the baseline Baveno stage. We expect that this algorithm will generate a new SOP for the classification of liver disease severity from Inpatient HES for use by others:

Stage 1: No varices, no ascites

Stage 2: Varices, no ascites

Stage 3: Ascites +/- varices

Stage 4: Bleeding +/- ascites

Overall survival in relation to these baseline factors will be established by extracting the time interval from HCC diagnosis date to death from the Cancer Registry. The certified cause of death and ‘underlying cause of death’ will be extracted from the Cancer Registry.

How is HCC treatment allocation associated with overall survival and decompensation of cirrhosis?

An Inpatient HES extract spanning the total study period will be used to identify specific HCC-related treatments. The procedure codes and time interval from HCC diagnosis to the episode containing the treatment will be recorded. The site code of treatment will also be collected.

In order to assess the use of sorafenib, data held within the cancer registry (AV_treatment) and within the Systemic Anti-Cancer Therapy Data Set (SACT) will be analysed. Sorafenib will be searched within the ‘raw regimen’ data item in SACT, along with the time interval from HCC diagnosis until the start of the drug regimen. The ‘organisation code of provider’ will be recorded to identify the treating centre.

Clinical outcomes will be determined including:

Overall survival

Survival post-treatment will be inferred from the time interval from HCC diagnosis date to death and the interval from HCC diagnosis to each HCC treatment

Rate of decompensation of cirrhosis following different treatments

Diagnosis and procedure codes specific to complications of cirrhosis which occur after the HCC diagnosis will be extracted, along with the time interval from HCC diagnosis date to the start of the associated episode. An updated cirrhosis stage will be calculated using the new Baveno stage SOP.

What is the nature of regional variation in baseline factors and treatment allocation in HCC?

Baseline factors at HCC diagnosis (in collaboration with the University of Liverpool)

The ‘broader geographical area’ and index of multiple deprivation (IMD) quintiles will be extracted from Inpatient HES. The baseline factors for each area will be assessed using the data items extracted in Section 1.

HCC Treatment allocation at different centres

The code of the NHS Trust in which the HCC treatment took place (‘site code of treatment’) will be extracted from the Inpatient HES episode associated with the specific HCC treatment. This is a necessary additional data item to a patient’s broader geographical area because it will also demonstrate where patients actually receive their HCC treatment. The variation in treatment allocation at different specialist centres can then be established and this will be correlated with the baseline factors.

c. Apply the described variation in baseline factors and treatment allocation to understand regional differences in survival (in collaboration with the University of Liverpool)

The factors described above will be used to identify predictors of regional variation in survival that may relate for instance to stage of liver disease at presentation or differences in treatment allocation.

Statistical Analysis

For each of the study sections, the following statistical analysis will be performed:

This analysis will be informed by the Cohort Overview, undertaken by the HCC-UK partners in Bristol (AB), which will describe the patient characteristics. Descriptive statistics will be used to characterise the baseline factors of the HCC cohort, assessing for significant associations between variables.

A two-state disease model (dead or alive) will be used for standard survival analysis: univariate analysis will be performed using the Kaplan-Meier method, with patients stratified by the baseline factors of interest such as cirrhosis stage. A Cox proportional hazards regression will be used to calculate a hazard ratio for the baseline determinants of overall survival.

a. Survival analysis will be performed using the Kaplan-Meier method, stratified by treatment allocation. Survival between the groups will be compared using the log-rank test. Patient numbers are expected to be large enough to allow further stratification by baseline factors of interest, such as liver disease aetiology.

b. The rate of decompensation of cirrhosis following different HCC treatments will be determined by calculating the risk of hospital readmission with complications of decompensated cirrhosis within 30 days. The baseline characteristics of those patients who have decompensation events will be compared using multivariate logistic regression in order to identify predictive factors.

A competing risk analysis will be performed using a cumulative incidence function (CIF) to describe the rate of cirrhosis decompensation events admission following different HCC treatments. The clinical states used in a multi-state disease model will be: compensated cirrhosis, decompensated cirrhosis, death related to HCC and death related to liver disease (from death certification).

a. The baseline factors will be cross-tabulated with broader geographical area and tested for significant associations. Analysis performed by HCC-UK partners in Liverpool (VK, TC, DP) will lead to an estimate of the median overall survival for each broader geographical area and linked IMD and travel time to destination. Significant differences in median survival between areas and the association with IMD quintile will be tested and adjusted for the variation in baseline factors.

b. For each centre treating patients with HCC, the proportion of patients receiving different HCC modality will be calculated. The statistical significance of any variation in treatment allocation between centres will be tested and adjusted for baseline characteristics.

c. Baseline factors and treatment allocation will be used in Cox proportional hazard models to investigate the potential differences in survival.

The study team expect that the development of the Baveno stage SOP will be applicable to a wide number of future population-based studies that require an assessment of cirrhosis severity.

We expect to publish these findings in peer reviewed scientific journals and present at scientific conferences. The two main publications include the description of regional variation in HCC treatment and survival (adjusted for cirrhosis severity) and the competing risk analysis of HCC treatment outcomes.

We hope that the project will inform future research and ongoing work within the HCC-UK/ NCRAS partnership on understanding variation in access to treatments and clinical trials, as well as cost-effectiveness analyses. An understanding of the natural history of cirrhosis in the setting of HCC provided by our analysis will inform the planning of surveillance programmes for HCC in cirrhosis. These analyses have the potential to identify variations in clinical practice and therefore to improve future resource allocation and clinical outcomes

Yielded Benefits is not a requirement for new applications.