Biomarkers to diagnose liver fibrosis
A cross-sectional study to optimise the diagnosis of liver disease in patients with chronic hepatitis B in African countries.
This is the first project from HEPSANET. We shared data from 12 sites in 8 countries to understand how best to diagnose liver cirrhosis, and to determine who needs hepatitis B treatment.
We collected data from patients with hepatitis B in 12 African countries. We shared and analysed these data to work out how best to diagnose cirrhosis.
The World Health Organisation produced guidelines in 2015 recommended starting treatment for patients who had persistently elevated liver enzymes (a blood test called alanine aminotransferase or ALT) together with high hepatitis B viral load, measured in the blood. Secondly, those who had clinical evidence of cirrhosis were recommended to start treatment. They also recommended using a ratio between the liver enzyme AST (asparate aminotransferase) and the platelet count- an equation called APRI.
Data from several centres in Africa emerged, showing that using these WHO guidelines could cause many patients with evidence of cirrhosis to miss the chance to have treatment, even in patients with advanced cirrhosis.
Patients with cirrhosis may have evidence such as bleeding from the upper part of the stomach or oesophagus, the development of jaundice or confusion due to the build up of toxins the liver normally processes (called hepatic encephalopathy), or fluid building up in the abdomen which can cause abdominal swelling (called ascites). In these cases, it is usually easy for the clinician to see the signs of liver disease and start treatment. But starting treatment at this stage is likely to be too late.
Treatment guidelines need to find a balance between saving treatment for those patients who need it (avoiding the cost of giving treatment to patients who are at low risk of liver disease) and avoiding under-treatment, where patients who could benefit from treatment do not meet the criteria in treatment guidelines.
Our study examined the role of routinely available and cheap blood tests in helping to diagnose cirrhosis or significant liver fibrosis, and examined the performance of APRI and other biomarkers including a similar test called the gamma glutamyl transferase to platelet ratio index (GPR).
You can read the paper published in Nature Communications. A blog post “behind the paper” is published here on Nature Portfolio.
A collaborative effort.
We contacted authors who had studied hepatitis B in observational studies from West, Central, East and Southern Africa. All authors who were contacted agreed to participate. In total 12 sites from 8 countries contributed data from hepatitis B patients to this study.
We uesd a tool called fibroscan to define who had cirrhosis. This is a machine that sends out a pulse of sound at a low frequency through the liver and watches the movement of the sound wave. The speed of the sound can be be used to estimate the stiffness of the liver. Stiffer harder livers reflect more advanced liver disease.
Current WHO guidelines fail patients in African countries
Combining data from these patients, we estimated the diagnostic performance of the simple blood test based biomarkers APRI (using AST and platelets), GPR (using GGT and platelets) and FIB-4 (AST, ALT age and platelets).
As you can see in the figure, we found that using the WHO recommended threshold for APRI (0f 2.0 to diagnose cirrhosis) resulted in a very low sensitivity of 16.5%. That means that for 100 patients with cirrhosis, only 17 would be positive using the APRI test at 2.0.
We derived new thresholds based on the individual patient data meta-analysis (combining patient data from many centres using a mathematical model).
We found that an APRI threshold of 0.65 could act as a rule-in threshold with better sensitivity of 56.2% (of those with cirrhosis, 56 out of 100 would test positive) and specificity of 90% (of those who did not have cirrhosis 90 out of 100 would test negative), while a rule-out threshold of 0.36 would result in sensitivity of 80.6% and specificity of 64.3%
Sensitivity
or specificity?
We found that the biomarkers APRI and GPR were the best (and performed very similarly).
We also found that using these biomarkers required making significant trade-offs between maximising sensitivity or specificity.
With a lower APRI threshold number, sensitivity is better (figure B, the orange line) most patients with cirrhosis will be diagnosed, however specificity (the blue line) falls and more patients who do not have cirrhosis will test positive, potentially resulting in more patients getting treatment who might not need it.
We found the tests are imperfect but that lowering the threshold would go a long way towards improving access to treatment for those who have cirrhosis and could immediately improve care for most patients.
Our study highlights the need to gather longitudinal data collected over time to better understand which patients develop problems from hepatitis B like liver decompensation or liver cancer, and to refine our treatment criteria. This is a current project for the HEPSANET team.