PAHA model: An alternative non-invasive predictor of liver cirrhosis in patients with chronic hepatitis B infection


  • Oyekoya T Ayonrinde,

    Corresponding author
    1. Department of Gastroenterology, Fremantle Hospital, Fremantle, Western Australia, Australia;
    2. School of Medicine and Pharmacology, The University of Western Australia, Fremantle Hospital, Fremantle, Western Australia, Australia
    Search for more papers by this author
  • John K Olynyk

    1. Department of Gastroenterology, Fremantle Hospital, Fremantle, Western Australia, Australia;
    2. School of Medicine and Pharmacology, The University of Western Australia, Fremantle Hospital, Fremantle, Western Australia, Australia
    Search for more papers by this author

Dr Oyekoya Ayonrinde, School of Medicine and Pharmacology, University of Western Australia, Fremantle Hospital Campus, PO Box 480, Fremantle 6959, Western Australia. Email:

Liver fibrosis resulting from chronic liver injury is the harbinger of cirrhosis, with its inherent potential complications and associated morbidity and increased mortality. Hepatic fibrogenesis is a dynamic process incorporating hepatocellular injury associated with chronic inflammation and continuous extracellular matrix (ECM) protein remodeling choreographed by hepatic stellate cells.1 Knowledge of the stage of fibrosis in chronic liver disease guides clinical decisions about the timing and approach to interventions. Although this has traditionally relied on liver biopsy, a suite of non-invasive models for liver fibrosis relying on individual or various combinations of putative biomarkers has been developed. These have been principally assessed in chronic hepatitis C virus infection (CHC), which is a dominant cause of cirrhosis and hepatocellular carcinoma (HCC) in the ‘developed world’. CHC and chronic hepatitis B virus (HBV) infection (CHB) have different natural histories and often affect different populations, so data from CHC cannot be directly extrapolated to represent CHB. Unfortunately, most blood-based models for liver fibrosis exhibit a significant level of incompetence at lower stages of liver fibrosis.2 Further, few have been validated for use in CHB, which remains a global public health problem with over 350 million people chronically infected worldwide.3 The burden of CHB is most significant in the Asia-Pacific region and Sub-Saharan Africa and in migrants from these regions.

Historically, liver biopsy has been considered the gold standard reference diagnostic and prognostic test for assessing liver disease. Following initial reports of liver biopsy dating as far back as 1883,4 liver biopsy techniques and indications have been further refined; however, the risk of significant bleeding or death related to liver biopsy5 remains relatively unchanged over more than 50 years. As increasing pharmaco-therapeutic options for chronic liver disorders, particularly CHB and CHC, have become available, the role of liver biopsy in guiding treatment decisions has been highlighted. However, liver biopsy for histology is an imperfect gold standard for assessing liver fibrosis alone, as demonstrated by an increasing body of evidence.6–8 It has been plagued with concerns about the invasive nature of the procedure, hence complication risk and limited patient acceptance, sampling error, inter- and intra-observer variability and cost. Complications of liver biopsy include pain (up to 84%), bleeding (in up to 0.04%) and death (up to 0.01%).5 Also, histologic staging of liver fibrosis presents thedynamic process of extracellular matrix deposition and remodeling as a categorical, non-linear result. Together with the invasive nature of liver biopsy, it is clear that serial assessment of liver histology is not practical at a population level. Consequently, non-invasive models for determining the presence and severity of liver fibrosis in chronic liver disorders are appealing to patients and clinicians.

Predictive models for liver fibrosis are generally derived from panels of direct and indirect continuous peripheral blood variables.9 Direct markers reflect extracellular matrix (ECM) turnover. These include matrix metalloproteinases (MMPs), which are involved in degrading collagens, and tissue inhibitors of metalloproteinases (TIMPs), which regulate MMP activity.10 MMPs, TIMPs, collagen, hyaluronic acid and YKL-40 also mirror scarring in liver fibrosis10 but do not independently predict significant fibrosis. Indirect markers of hepatic synthetic dysfunction and portal hypertension include serum aminotransferase levels, coagulation profile, haptoglobin, albumin level and platelet count.9FibroTest11 and Hepascore12 are non-invasive tests of liver fibrosis that incorporate patient age, gender and both direct and indirect blood variables. FibroTest is a composite model based on serum alpha-2 macroglobulin, total bilirubin, gamma-glutamyltransferase (GGT), apolipoprotein A1 and haptoglobin. In contrast, Hepascore utilises total bilirubin, GGT, hyaluronic acid, and alpha-2 macroglobulin.

FibroTest11 (and its offshoots) have been validated for staging fibrosis in various populations, including CHB but with variable area under the receiver-operating characteristic curve (AUROC). Also, emerging data indicate that Hepascore is useful for identifying cirrhosis in a predominantly Asian CHB population.Transient elastography is increasingly reported to be a rapid, relatively accurate test for liver fibrosis and recently was compared against a new model called the APGA index (aspartate aminotransferase [AST], platelet count, GGT and alphafetoprotein [AFP]).13

In the current edition of the Journal, Lee and colleagues14 describe an alternative non-invasive predictor of liver cirrhosis in CHB, called the PAHA (platelet count, AST, Haptoglobin, and Apolipoprotein-A1) model, which compares favorably with existing noninvasive liver fibrosis models. Remarkably, the predictive accuracy of the PAHA model was superior to the previously described AST/ALT ratio, PGA (prothrombin time, GGT, Apo-A1), PGAA (prothrombin time, GGT, apolipoprotein A1, α2-macroglobulin), age platelet index (API), Forns fibrosis index (FFI), and AST to platelet ratio index (APRI). The PAHA model was derived in a prospectively selected, predominantly male, non-obese, treatment-naive, adult Korean population with CHB who had undergone percutaneous liver biopsy and had no demonstrable decompensation of cirrhosis. Liver histology on biopsy specimens of at least 15 mm length was used as the reference test forcomparison with the PAHA model. The PAHA model is derived from platelet count, serum AST, haptoglobin and apolipoprotein-A1, which were independent predictors of liver cirrhosis when considered as categorical variables in a multivariate logistic regression model. PAHA had a robust AUROC for cirrhosis (Metavir F4) of 0.92 and sensitivity, specificity, and positive and negative predictive values of 88%, 81%, 64%, and 94%, respectively, when a threshold of 5.5 was applied. This results in a likelihood ratio of a positive test result (LR+) of 4.6, likelihood ratio of a negative test result (LR−) of 0.15 and a reasonable diagnostic odds ratio of 30.9. Consistent with other fibrosis biomarker models PAHA was less discriminatory (AUROC 0.78) for advanced fibrosis (Metavir F3-F4).

The strength of the PAHA model is the potential of this as a non-invasive liver fibrosis test in high HBV-prevalence societies (primarily developing countries), where histologic assessment of HBV severity is restricted by availability, cost and potentially limited therapeutic consequence. It also has the attraction of having been developed in the highly HBV-endemic Asia-Pacific region, where HBV infection is associated with up to 80–90% of HCC cases in Korea, China, Singapore, India, Vietnam, Taiwan and Papua New Guinea.3 Unfortunately, the authors have not proffered a cost for the PAHA model, as this may ultimately limit the utility of the test. Notably, details of the prevalence of excessive alcohol intake have not been provided. Also, in univariate analysis there was a significant difference in platelet count between the cirrhosis and non-cirrhosis groups, with thrombocytopenia already identifying cirrhosis in 50% of patients using the relatively cheap and available platelet count. The platelet count could predict the presence of advanced fibrosis in CHB, with AUROC of 0.68, negative predictive value 78% and specificity 87% in a study from Taiwan,15 thus potentially reducing the cost in relation to the proportion of patients requiring either liver biopsy or assessment with models based on panels of biomarkers. The study by Lee and colleagues has not compared the PAHA model with models incorporating direct markers of ECM turnover; hence it is uncertain if it would be superior to these. The ultimate test for PAHA lies in external validation in a different population, validation in different chronic liver disorders and comparison against other noninvasive models that incorporate direct markers of ECM turnover. Nevertheless, since the more complex models incorporating direct markers are not readily available in large parts of the Asia-Pacific region, PAHA would clearly have a role if it demonstrates improved accuracy for distinguishing significant fibrosis from non-significant or absent fibrosis in diagnosis and longitudinal assessment of treated and untreated patients with chronic liver disorders.

In summary, PAHA is a refreshing addition to the armamentarium of clinicians managing CHB in the Asia-Pacific region and beyond. Such combinations of clinicopathological markers may eventually replace the need for liver biopsy in many patients with CHB.