Should Patients with Abnormal Liver Function Tests in Primary Care be Tested for Chronic Viral Hepatitis

David T Arnold; Louise M Bentham; Ruth P Jacob; Richard J Lilford; Alan J Girling


BMC Fam Pract 

In This Article



The objective of the BALLETS study was to correlate the pattern of abnormal LFTs in primary care with the 'final' diagnosis. This was accomplished by fully investigating a cohort of patients with abnormal LFTs and no known liver disease by means of a panel of tests for the specific viral, auto-immune and genetic diseases shown in Table 1. This comprehensive investigation by means of the standard "liver panel" acts like a concertina, bringing forward the diagnosis of conditions, such as viral hepatitis and PBC, that would take years or decades to manifest if followed-up in the usual way. The BALLETS study was funded by the UK National Institute for Health Research through its Health Technology Assessment (HTA) programme. The full protocol is available online.[14]

Patients were recruited from eight Birmingham and three Lambeth practices from November 2005 to November 2008. The cohort was formed from patients with at least one abnormal analyte (from a panel of eight - alanine aminotransferase (ALT), aspartate aminotransferase (AST), ALP, bilirubin, gamma-glutamyltransferase (GGT), globulin (derived), total protein, and albumin) and no known liver disease. The biochemical measurements were carried out in three Clinical Pathology Accreditation UK (CPA UK) laboratories (conforming to International Quality Control Standards) in Birmingham and Lambeth. Laboratory specific ranges (incorporating age and sex adjustments for ALT, ALP and GGT) were used to define abnormality. Eligible patients from the participating GP practices were invited to attend special study clinics where the following information was obtained:

· Clinical data, including: country of birth, alcohol history, Body Mass Index (BMI), drug and medicine use, and reason for LFT testing.
· A repeat LFT.
· Blood tests for specific diseases (Table 1). Investigations for viral hepatitis tests included hepatitis B viral markers (HBV Surface Ag) and hepatitis C virus antibody (HCV Ab).[10,11]
· Ultrasound of upper abdomen.

Patients were followed up for two years, at which point LFTs were repeated along with a further ultrasound of the upper abdomen (results of the follow-on study will be reported at a later date).

Testing Strategies

A simple decision tree was constructed in Microsoft Excel (Microsoft Corp., Redmond Washington) to enable costs per case detected to be calculated for seven strategies.[15] The strategies were developed in consultation with a GP and hepatologist (P. Gill and J. Neuberger respectively) who were aware of the relevant literature and guidelines.

The root decision (or starting point) of the tree is the discovery of an abnormal LFT in primary care where the patient does not have known or self-evident liver disease. From the root node we identified seven decisions that may be considered by a GP under such a scenario:

▪ Strategy A: repeat the LFT panel and then perform a specific test for viral hepatitis if an abnormality is still present on re-testing. This could be considered the intuitive response by a GP on receiving an abnormal LFT in a patient without the indictors of a specific disease, and is the strategy recommended in the literature.[1–5]
▪ Strategy B: perform a viral test in all patients with an abnormal ALT. The rationale for this strategy is that ALT is the most specific indicator of viral hepatitis,[4] and has been recommended as the testing criterion by other authors.[16–18]
▪ Strategy C: select ALT as the trigger for viral testing, but nominate a higher threshold at twice the upper limit of normal, as recommended by Jamali.[19] This is also the threshold for instigating viral therapy for HBV in certain treatment guidelines.[20–22]
▪ Strategy D: perform a test for viral infection in all patients who originate from a country with an intermediate or high prevalence of viral hepatitis according to World Health Organisation (WHO) criteria.[23–25] Screening has been shown to be cost-effective for people who were born in prevalent countries and it is likely that testing would be more cost-effective still in a population with abnormal LFTs.[21,26]
▪ Strategy E: combine the two previous strategies by testing those who have an ALT exceeding twice the upper limit of normal and who also originate from a prevalent country.
▪ Strategy F: test all patients from prevalent countries as well as those with an ALT exceeding twice the upper-limit of normal.
▪ Strategy G: test all patients for viral hepatitis irrespective of the type or extent of abnormal LFT results.

There is also an option to take no action with respect to viral hepatitis, and while this may be a sound decision in some cases, for example when a LFT is ordered in the hope that a positive result will prompt a reduction in alcohol intake, this was not considered here.

In this study the hepatitis status for all patients was known. Moreover most had an ALT test result and the results of a repeat LFT panel. Thus it was possible to evaluate the performance of each of the above strategies.

Populating the Decision Tree with Probabilities/Statistical Model

All 1,236 patients were used in the evaluation of strategy G, but for all other strategies the effective sample size was reduced because of missing data in some of the patient records. Estimates of the proportion of patients undergoing viral tests, and the proportion of actual cases detected (sensitivity) were obtained using the sample of patients available for evaluating each strategy. The positive predictive value of a strategy was defined as the proportion of hepatitis cases among those selected for viral testing. Confidence limits for this quantity were calculated using Wilson's method for Binomial data.[27]

Estimation of Costs

The direct costs incurred at the time of the test were the laboratory costs of the liver function and viral hepatitis tests (personal correspondence with Pathology Lab Manager); the GP costs for scheduling each test; and following up on results. Administrative costs were estimated by estimating the time implications for a secretary to add patients to appointment slots and a receptionist to check the patient in for an appointment (personal correspondence with MidReC: West Midlands Research Consortium. Figures correct as of February 2009). The costs are presented in pounds sterling (£) (and were correct for the year 2009). Non-health service costs (patient travel cost and lost earnings) were not measured, but are considered in the discussion.


The number of cases detected per 100 patients was estimated as the sensitivity of the strategy (cases detected divided by cases present) multiplied by the prevalence (per 100 patients) of viral hepatitis in the whole sample of 1,236 patients. For each strategy the cost per case detected was then computed as the ratio of the cost per patient to the number of cases detected per patient. The strategy which minimised this quantity was taken as the base case. For each alternative strategy the incremental cost-effectiveness ratio (ICER) was computed, defined as the incremental cost per additional case detected compared to the base case. The analysis is deterministic and does not consider the impact of sampling variability. The results of these analyses were compared with published results of cost-effectiveness analysis of screening for chronic viral hepatitis, bearing in mind likely differences between a screening and a diagnostic population. We used this analysis to develop a "fast and frugal" heuristic[28] which we offer to readers for their consideration.