Novel Approaches to Detect Significant Liver Disease in the General Population

Abbreviations

ALT

alanine aminotransferase

APRI

aspartate aminotransferase‐to‐platelet ratio index

ARFI

acoustic radiation force impulse

AST

aspartate aminotransferase

AUD

alcohol use disorder

AUROC

area under the receiver operating curve

BBV

bloodborne virus

CI

confidence interval

ELF

Enhanced Liver Fibrosis

FIB‐4

Fibrosis‐4

GP

general practitioner

LBT

liver blood test

LFT

liver function test

MR

magnetic resonance

NAFLD

nonalcoholic fatty liver disease

NFS

NAFLD Fibrosis Score

NPV

negative predictive value

PPV

positive predictive value

Sens

sensitivity

Spec

specificity

Liver disease remains a significant health burden worldwide and is increasingly driven by excess body weight and alcohol consumption.¹

Morbidity and mortality related to liver disease are largely dependent on the presence of advanced fibrosis; however, in its early stages, detection may be difficult, and consequently liver disease is frequently diagnosed only when hepatic decompensation occurs.

Therefore, the early detection of advanced fibrosis is of critical importance in minimizing the rising morbidity and mortality from liver disease.

However, the traditional approach to liver disease detection has relied largely on the recognition of abnormal liver function tests (LFTs) with subsequent follow‐up serological testing to diagnose specific diseases.

This is problematic for two reasons. First, abnormal LFTs are frequently overlooked in primary care settings, and a fibrosis risk assessment is typically undertaken only in specialist hepatology clinics; second, LFTs are normal in about 20% of patients with cirrhosis. Therefore, without an appreciation in primary care of risk factors for advanced fibrosis and the poor positive predictive value (PPV) of abnormal LFTs in its identification, opportunities to recognize serious liver disease will continue to be missed.

The Challenge of Detecting Liver Disease in The Population

Although 10% to 20% of all liver blood test (LBT) panels contain at least one abnormality,² just 3% to 5% translate into a specific or significant liver disease.³ This low “conversion rate” can lead to a false sense of reassurance that adverse outcomes are rare and explain why, in one UK study, just 50% of abnormal liver tests were ever followed up.² These observations contribute to late diagnosis and highlight the urgent need to support primary care to find “the needles in the haystack.”

Given around one in five patients with cirrhosis has normal LFTs, fibrosis testing also should be considered in scenarios with a high pretest probability, such as the finding of fatty liver on imaging (especially in the context of type 2 diabetes mellitus), excess alcohol consumption, or unexplained splenomegaly and/or thrombocytopenia.

Any fibrosis testing approach on a population basis requires the use of a test that is cheap, readily available in a range of health care settings, accurate, and reproducible.

It remains unlikely that whole population screening is feasible, either from a cost‐effectiveness or a clinical service standpoint. Hence a more targeted, but assertive, approach to fibrosis detection is required, focusing on those at greatest risk as already outlined. A proposed initial approach is outlined in Fig. 1.

FIG 1.

Target groups and methods of liver fibrosis screening in the population.

There remains a fine balance to be struck here between not relying on a specific liver diagnosis to be made before fibrosis assessment, yet still providing a framework to diagnose and treat specific conditions, such as viral or autoimmune hepatitis, regardless of whether advanced fibrosis is likely.

Methods of Fibrosis Assessment

When considering which fibrosis testing approach to take, the key is being able to identify these high‐risk populations and test at scale. This requires improved knowledge in primary care and robust pathways that do not require an existing liver diagnosis but help facilitate one. Information technology is also critical, in particular the ability to automate or “reflexively” undertake a fibrosis risk assessment. Building these tools into electronic clinical management systems will improve case capture and act as a positive feedback loop to further raise awareness and knowledge in primary care.

There are four major categories of fibrosis assessment, ranging from simple “scores” derived from commonly measured laboratory tests and/or clinical features to measurement of serum biomarkers of fibrosis through to measurement of liver stiffness via differing elastography methods and, finally, liver biopsy. These are summarized in Fig. 2.

FIG 2.

Methods of fibrosis assessment and their cost/access gradients.

Given the extremely low cost and near‐universal availability (certainly in comparison with other modalities), nonproprietary, simple fibrosis scores represent the best initial strategy.

The negative predictive value (NPV) of most of these fibrosis assessment tools in secondary care nonalcoholic fatty liver disease (NAFLD) cohorts is high at approximately 90%,⁴ and their performance characteristics are summarized in Table 1. In community settings, where disease prevalence is lower, the NPV is likely to be higher still. However, the major limitations of simple fibrosis scores are their poor PPV, and so they are more useful to exclude, rather than predict, advanced fibrosis.

TABLE 1.

Performance Characteristics of Different Fibrosis Tests for the Diagnosis of Advanced Fibrosis in NAFLD

Test	AUROC (95% CI)	Cutoff	Sens (%)	Spec (%)	PPV (%)	NPV (%)
AST:ALT ratio	0.83 (0.74‐0.91)	0.8	74	78	44	93
			52	90	55	89
APRI	0.67 (0.54‐0.8)	1	27	89	37	84
BARD score	0.77 (0.68‐0.87)	2	89	44	27	95
FIB‐4 score	0.86 (0.78‐0.94)	1.30	85	65	36	95
		3.25	26	98	75	85
NAFLD	0.81 (0.71‐0.91)	−1.455	78	58	30	92
Fibrosis score		0.676	33	98	79	86

Reproduced with permission from Gut.⁴ Copyright 2010, British Society of Gastroenterology.

Consequently, attention has turned to two‐stage approaches, whereby the majority without fibrosis can have this reliably excluded by simple fibrosis scores, leaving more costly and/or labor‐intensive methods, such as Enhanced Liver Fibrosis (ELF) and/or FibroScan, to help confirm an initial suspicion raised by the simple score.⁵

Which Population Fibrosis Detection Model to Apply?

Although there are retrospective head‐to‐head comparisons of simple fibrosis scores among secondary care cohorts in specific diseases, there are none for unselected, primary care populations. Even then, any difference in performance characteristics is marginal, and all have a high NPV.⁴ Similarly, there is no current evidence to support any combination of tests as part of a two‐stage approach to fibrosis detection.

An important question is whether fibrosis detection strategies should be specific to different underlying etiologies of liver disease, or whether a once‐size‐fits‐all strategy can be used. In support of the former, UK guidelines recommend a different (one‐stage) approach via FibroScan for assessing fibrosis in people with an alcohol use disorder (AUD) but a different (two‐stage) approach using Fibrosis‐4 (FIB‐4) and FibroScan in NAFLD. A Danish study demonstrated that an ELF test also can reliably exclude fibrosis in AUD, but that transient elastography remained the best noninvasive predictor of fibrosis.⁶ In primary care patients diagnosed with NAFLD, a recent two‐stage pathway using FIB‐4 and then ELF demonstrated increased cirrhosis detection and a significant reduction in specialist referrals.⁷

In contrast with this disease‐specific approach, three different community‐based fibrosis detection models have recently been reported from the United Kingdom.⁸, ⁹, ¹⁰ These models are summarized in Table 2. Although heterogenous regarding entry criteria, degree of automation, and methods of fibrosis detection, they all demonstrate an ability to evaluate large populations and improve the detection of specific liver diseases and advanced fibrosis/cirrhosis compared with traditional models of care. In addition, two studies⁸, ⁹ demonstrate cost‐effectiveness. Another common theme identified in these studies is a disappointingly low engagement or compliance with the developed pathway of between 40% and 50%.⁹, ¹⁰ Reasons for this probably include the low profile of liver disease in primary care and a subsequent lack of understanding regarding the risk for fibrosis. Consequently, further education is required to improve population liver disease detection. This is relevant because fibrosis assessment pathways reliant on a diagnosis being made⁷ could act as a barrier to fibrosis detection. In contrast, those reliant on abnormal LFTs⁹, ¹⁰ run the risk of missing significant liver disease as exemplified by the Nottingham model, which demonstrated that 30% of the cirrhosis detected would not have been so had it been solely reliant on abnormal LFTs.⁸

TABLE 2.

Summary of Community‐Based Fibrosis Detection Studies

Study	Target Group	Population Fibrosis Detection Method	Numbers Assessed/Compliance	Outcomes
Scarred Liver Project6 (Nottingham, England)	Population	Risk factors	25,018 screened	14% kPa 8‐15
		FibroScan	3688 at risk	6% kPa >15
			1239 FibroScans	39 new cirrhosis diagnoses
			GP compliance unknown	Cost‐effective
Camden & Islington7 (London, England)	NAFLD	FIB‐4	1452 screened	81% ↓ in referrals
		ELF	275 referred	29% advanced fibrosis
			GP compliance 55%	14.5% cirrhosis
Intelligent Liver Function8 (Dundee, Scotland)	Population	Abnormal LBT	229 intervention group	45% ↑ in diagnosis
		FIB‐4	64 with abnormal LBTs	Cost‐effective
		FibroScan	GP compliance 50%	Suggests a diagnosis
Gwent AST Project9 (Gwent, Wales)	Population	Abnormal ALT	17,770 abnormal ALT	28% kPa 8‐15
		Reflex AST:ALT ratio	2117 AST:ALT > 1 (12%)	29% kPa >15
		FibroScan	348 FibroScans	192 advanced fibrosis
			GP compliance 40%	81% ↑ in cirrhosis diagnosis

Summary

Assessment of fibrosis is critical in identifying those with significant liver disease in the population via clear pathways, those with abnormal liver tests, those with fatty liver, and/or those with excess alcohol consumption. Incorporation of these pathways, or at least the fibrosis aspects of them, into clinical management systems can facilitate recognition of specific diseases, as well as “reflex” calculation of simple fibrosis scores.

Engagement with primary care is also vital to bridge the primary care‐hospital provider gap and improve adherence to pathways, which is low in published series.⁷, ⁸, ⁹, ¹⁰

No one fibrosis test or combination thereof is perfect, or superior to another, and all have their merits and demerits. Although recent data suggest the aspartate aminotransferase/alanine aminotransferase (AST/ALT) ratio is not as accurate as FIB‐4, again this comes from a secondary care NAFLD cohort,¹¹ and comparative data of the effectiveness of differing components of population fibrosis detection strategies are unlikely to be available in the short term. In contrast, data from Gwent¹⁰ show that use of FIB‐4 on an unselected population leads to a 4‐fold increase in FibroScan requirements using a threshold of >1.3 and a 2‐fold increase if >3.25 is used.

Despite these challenges, consistent implementation, at scale, using any of the cheap and widely available simple fibrosis scores as the first step followed by FibroScan or ELF will improve the diagnosis of advanced liver disease in those with abnormal LFTs or NAFLD compared with traditional models of care. However, models should not be reliant just on those with abnormal LFTs, and risk factor analysis followed by FibroScan or ELF testing will further improve population detection rates.

The choice of fibrosis test (or combination thereof) also should be based on local factors: prevalence of differing liver diseases, clinical expertise, geographical and capacity constraints, availability of technology, and patient wishes. Finally, iteration of pathways is necessary to constantly refine approaches designed to meet the challenge of identifying significant liver disease in the general population.