Systematic review and individual-patient-data meta-analysis of non-invasive fibrosis markers for chronic hepatitis B in Africa

Abstract

In sub-Saharan Africa, simple biomarkers of liver fibrosis are needed to scale-up hepatitis B treatment. We conducted an individual participant data meta-analysis of 3,548 chronic hepatitis B patients living in eight sub-Saharan African countries to assess the World Health Organization-recommended aspartate aminotransferase-to-platelet ratio index and two other fibrosis biomarkers using a Bayesian bivariate model. Transient elastography was used as a reference test with liver stiffness measurement thresholds at 7.9 and 12.2kPa indicating significant fibrosis and cirrhosis, respectively. At the World Health Organization-recommended cirrhosis threshold (>2.0), aspartate aminotransferase-to-platelet ratio index had sensitivity (95% credible interval) of only 16.5% (12.5-20.5). We identified an optimised aspartate aminotransferase-to-platelet ratio index rule-in threshold (>0.65) for liver stiffness measurement >12.2kPa with sensitivity and specificity of 56.2% (50.5-62.2) and 90.0% (89.0-91.0), and an optimised rule-out threshold (<0.36) with sensitivity and specificity of 80.6% (76.1-85.1) and 64.3% (62.8-65.8). Here we show that the World Health Organization-recommended aspartate aminotransferase-to-platelet ratio index threshold is inappropriately high in sub-Saharan Africa; improved rule-in and rule-out thresholds can optimise treatment recommendations in this setting.

Fig. 1. Map of included studies.

n the number of hepatitis B patients included from each site in the current analysis. Area of circles is proportionate to cohort size. Attribution to iStockphoto (https://www.istockphoto.com/vector/africa-map-gm517581017-49316640).

Fig. 2. Flowchart of data sources.

HIV human immunodeficiency virus, ALT alanine aminotransferase, AST aspartate aminotransferase, ULN upper limit of normal, HCV hepatitis C virus, HDV hepatitis D virus, GGT gamma-glutamyl transferase,TE transient elastography.

Fig. 3. Performance of biomarkers of liver fibrosis (APRI, GPR, FIB-4) and ALT using transient elastography as a reference for the diagnosis.

The diagnosis of A LSM > 12.2 kPa (associated with cirrhosis) and B LSM > 7.9 kPa (associated with significant fibrosis): Bayesian random effects model^a. Point estimates are shown as red circles with 95% credible intervals shown as error bars. CrI credible interval, APRI aspartate aminotransferase-to-platelet ratio index, FIB-4 fibrosis-4 score, GPR gamma[1]glutamyl transferase-to-platelet ratio, ALT alanine aminotransferase, LSM liver stiffness measurement. ^aBayesian bivariate random effects model (n = 3548 biologically independent samples) adjusted for sex, study, hazardous alcohol consumption, reason for testing (suspected liver disease or asymptomatic screening) and categorical body mass index; summary statistics show population average. Youden’s J = maximisation of Youden’s J statistic, with equal weight to sensitivity and specificity (J = sensitivity + specificity − 1). Rule-in models were chosen where specificity exceeds 90%; rule-out models where sensitivity >80%.

Fig. 4. Relationship between sensitivity and specificity for APRI used to diagnose liver stiffness measurement >12.2 kPa (associated with cirrhosis).

Relationship among A all participants; B asymptomatic screening populations and C patients with suspected liver disease^a. Point estimates at different APRI thresholds are shown as blue (specificity) or orange (sensitivity) dots, connected by straight line segments with 95% credible intervals shown as transparent error bands. ^aBayesian bivariate random effects model fitted for different thresholds of APRI using 60 equally spaced quantiles. Specificity is defined as the probability of a negative test given the absence of LSM > 12.2; sensitivity is defined as the probability of a positive test, given the presence of LSM > 12.2. Source data are provided as a Source Data file.

Fig. 5. Illustration of using APRI to classify patients in asymptomatic screening and suspected liver disease populations.

Using A rule-in and rule-out thresholds, B the WHO recommended threshold and C Youden J derived threshold^a. ^aA single grey dot represents one person with liver stiffness measurement (LSM) < 7.9 kPa, one orange dot represents a person with LSM > 7.9 kPa associated with significant fibrosis, and one red dot represents a person with LSM > 12.2 kPa associated with cirrhosis (F4). Individual proportions rounded to nearest whole individual and represent the effect of applying APRI thresholds to the average populations included in the cohorts, stratified by reason for hepatitis B testing. Source data are provided as a Source Data file.