Biochemical tests of liver function are used to investigate a wide range of symptoms and for various purposes in asymptomatic people. This raises the potential to discover significant disease at an early stage in its natural history, but also produces abnormal results in apparently healthy people. This is particularly so in the “statin prescribing” era.
The usual liver function tests include two separate types of tests. Serum albumin and bilirubin provide a true measure of hepatic function, although both can be abnormal for several non‐hepatic causes. The remaining tests are of liver enzyme levels, usually alanine aminotransferase (ALT) or aspartate transaminase (AST) in combination with alkaline phosphatase (ALP) +/− γ‐glutamyl transpeptidase (GGT).
These answers attempt to allow the user to identify situations in which values of one or more of these tests outside the 95% population reference range (mean (2 standard deviations (SDs)) for the population) may indicate the presence of clinically significant disease.
In view of the statistical phenomenon of regression to mean,1
the principle of interval retesting to establish one‐off from stable and progressive increases seems sound. There is guidance, but limited evidence, on retesting intervals, although in the asymptomatic context and in view of the slow time course of the potential underlying diseases, the retesting interval of months rather than weeks seems reasonable for minor increases, commonly described as up to 3–5 times the upper limit of normal (ULN). Indications for further investigation of persistent rises commonly refer to 1.5 ULN. The limitations of reference ranges are further compounded for liver tests by the fact that population results are not normally distributed and exhibit a right skew effect.
How should I investigate an isolated slightly raised alkaline phosphatase in an asymptomatic adult?
- Checking of liver and bone profiles if not already carried out, to exclude other raised indices.
- Isolated raised values up to approximately 20% above the ULN (typically 120 IU/l) are more likely to reflect a statistical rather than clinical finding.
- If <1.5 ULN: recheck in 1–3 months.
- If >1.5 ULN (2 measurements, 6 months apart): further investigation indicated.
- If >3 ULN (single measurement): further investigation indicated.
Reference ranges of ALP are age and sex‐related, being lower in women than men, but rising in women beyond middle age. In addition, bone‐specific ALP is higher in black women than in white.2
Laboratory methods and reference ranges unfortunately vary by method, and users should acquaint themselves with local ranges. The laboratory reference range contains 95% of the population values. Notwithstanding exceptions described earlier, 2.5% of people would therefore be expected to have activities above the ULN, and, in view of the low prevalence of disease in patients with minor rises just over the mean (up to approximately 20%), this is more likely to represent a statistical rather than a clinically relevant finding. Using a typical local reference range of 35–104 IU/l, approximately 2.5% of a normal population will be expected to have values >104 IU/l (mean (2 SDs)), compared with approximately 0.1% >121 IU/l (mean (3 SDs)).
ALP has two main sources in the non‐pregnant adult: liver and bone.
Higher ALP activities are also seen as a normal variant3
and are associated with a range of medical conditions (congestive heart failure, hyperthyroidism, pregnancy) and certain drugs (ibuprofen, acetaminophen, cefotaxime) (reviewed in Aranda‐Michel and Sherman2
ALP is an intrahepatic cannalicular enzyme, which when raised because of liver disease reflects intrahepatic or extrahepatic obstruction rather than hepatocellular damage, although both may coexist.
In addition to clinical assessment of possible risk factors for, or indicators of, liver disease to distinguish potential hepatic risk factors from bone disease, GGT4,5,6,7,8
(or 5′ nucleotidase when used as an alternative9
) offers the best means of identifying a hepatic origin.
Recommendations distinguish isolated from paired or multiple‐enzyme abnormalities as an initial indicator of potentially significant disease. Although not evidence‐based, a threshold of 1.5 times ULN is cited by several authors as an indication for further investigation at 1–3 months if raised <1.5 ULN4
with further investigation if >1.5 ULN on two occasions at least 6 months apart.5
This is consistent with a review in the Journal of Insurance Medicine6
describing the threshold as a means of distinguishing potentially significant disease, although the same review highlights the paucity of direct clinical data relating liver enzyme abnormalities to mortality. On the basis of mathematical calculation of mortality associated with biopsy diagnoses, this review concluded that, if alcohol and hepatitis were excluded, background mortality was “fairly low” for patients with mild to moderate liver profile abnormalities.
Although the reference range for women rises with age, the prevalence of primary biliary cirrhosis also rises. Measurement of antimitochondrial antibodies in cases of persistently increased ALP >1.5 times ULN of liver origin would therefore seem appropriate.
Raised ALP without a concomitant rise in GGT in a non‐pregnant adult is likely to be of bone origin.4,5,6,7,8
Formal distinction can often, but not always, be achieved by isoenzyme analysis if uncertainty remains in the presence of marked abnormalities.
Calcium studies combined with the clinical filter will help in the diagnosis of treatable bone disease due to osteomalacia and parathyroid disease, and therefore seem clinically justified. However, the most common reason for raised bone ALP in older people is (often subclinical) Paget's disease.
Paget's disease is estimated to have a prevalence of 2–6% in the UK,10
and to have autosomal dominant inheritance with variable penetrance in some families.11
ALP activity is reported to correlate with disease activity, and the relative risk of Paget's disease in patients with raised ALP is 10.9, although Paget's disease was also found in 2.3% of people with normal ALP.12
The currently established indications for treatment of Paget's disease are based on symptoms and preparation for orthopaedic surgery.13
Presently, in the absence of these situations, a further investigation on minor bone‐related ALP would seem unwarranted in an asymptomatic patient with normal calcium studies.
How should I investigate an isolated case of slightly raised bilirubin in an asymptomatic adult?
- Values up to 20% above ULN are likely to be statistical rather than clinical abnormals.
- Values <1.5 times ULN: interval retest in 1–3 months unless clinical suspicion of disease.
- Values >1.5 ULN: confirm proportion of indirect (unconjugated) bilirubin.
- If >70%, unconjugated, probable Gilbert syndrome: no further testing needed if non‐progressive on interval retesting, unless haemolysis is clinically suspected. If rising on restesting, consider haemolysis and test haptoglobin, lactate dehydrogenase and blood count with reticulocyte count.
- Values >3 times ULN: clinical disease probable, further investigation required. Consider ultrasound (conjugated >50%) or haemolysis (unconjugated >70%).
As a product of the degradation primarily of haem, and excreted in bile, circulating bilirubin concentrations are determined by the balance of production and excretion.
Production is increased principally in disorders of red blood cell metabolism and excretion in disorders of hepatic conjugation (producing unconjugated or indirect bilirubin) or removal (producing conjugated or direct bilirubin). Laboratory discrimination of direct from indirect bilirubin becomes less reliable at lower total bilirubin concentrations (typically < approximately 40–50 mmol/l).
Using the same principles as for ALP and a typical range of 5–21 mmol/l, 2.5% of a population would be expected to have a total bilirubin >21 mmol/l and 0.1% >25 mmol/l. Users should again check their local ranges.
The clinical assessment should logically seek to exclude drugs that interfere in the bilirubin reaction (propranolol, oxytetracycline, methotrexate and levodopa)2
and take this into account in the decision to investigate further for stable raised bilirubin.
As described earlier, a raised bilirubin in conjunction with other liver function test abnormalities will prompt further investigation directed by predominant pattern of hepatocellular (transaminases) or cannalicular (GGT and ALP) changes.
The most common cause of “slight” isolated rises in bilirubin is Gilbert syndrome, a benign untreated congenital defect of glucuronide conjugation, present in up to 5% of the population.14
This produces unconjugated raised bilirubin (particularly during intercurrent illness or fasting) usually not exceeding 68 mmol/l9,15
(or up to 85 mg/dl2,16
). Although diagnostic investigations may be used to diagnose this, reassurance will in most cases be sufficient.9,15
Identification of raised bilirubin in conjunction with raised cannalicular enzyme (ALP or GGT) should prompt consideration of potential intrahepatic or extrahepatic cholestasis and is not described further in this answer.
Saturation of the conjugation process by excess production of bilirubin will lead to a predominance of unconjugated bilirubin. One review describes raised bilirubin >50% conjugated and raised bilirubin >70% unconjugated as representing direct and indirect hyperbilirubinaemia, respectively.2
If the rise is isolated, and predominantly unconjugated, exclusion of increased haem turnover leaves a highly probable diagnosis of Gilbert disease. No further investigation is recommended if bilirubin concentrations fall or do not continue to rise on interval retesting. The British Columbian guideline recommended interval retesting of 1–3 months for rises in any liver test <1.5 ULN. This is consistent when there is a small bilirubin rise in an asymptomatic patient, particularly in light of the bimodal serum decrease time, which has a long terminal phase (due to δ‐bilirubin formation17
). Common sense would suggest that, in this period, retesting be carried out in the earlier part for larger rises and later for smaller rises. It would be unlikely to see large rises in bilirubin—for example, > 3 ULN—in the absence of other changes, and a laboratory opinion would seem reasonable in this instance (author interpretation).
Haemolysis is conventionally diagnosed by the laboratory appearances on a blood film, combined with reduced haptoglobin, reticulocytosis and raised lactate dehydrogenase. There is no clear evidence as to whether one or all of these tests are required.
If evidence of haemolysis is found, further investigation will be determined by the clinical context, usually in conjunction with secondary‐care advice or referral.
Resorption of large haematomas can produce a similar pattern of hyperbilirubinaemia; this should be clinically apparent provided requesters of tests are aware of this as a possible cause.
Several recommendations for liver function testing are focused on patients with risk factors for, or clinical evidence of, suspected liver disease.4,5,8
None recommend general population screening, an approach that is also not considered to be cost effective.18
How should I investigate abnormal transaminases in asymptomatic patients without risk factors for, or clinical features of, liver disease?
- If raised <3 ULN: recheck in 1–3 months
- If still raised (two measurements, 6 months apart): further investigation indicated as shown in table 1.
Table 1Investigations in asymptomatic patients with increased liver enzyme level
- If >3 ULN (single measurement); further investigation indicated as shown in table 1.
AST and ALT are sensitive markers of hepatocyte injury but both, particularly AST, are also found in decreasing concentration in cardiac muscle, skeletal muscle, kidney, brain, pancreas, lung and blood cells. As transaminases are present in highest concentrations in the liver, large increases are usually due to hepatocyte injury.
Serum ALT activity varies with age, sex, race, body mass index, acute illness and exercise (reviewed in Gianni et al8
), and compelling evidence from diseases such as hepatitis C, non‐alcoholic fatty liver disease (NAFLD) and haemochromatosis suggest that considerable liver injury may be present with normal transaminase levels.20,21
Perhaps the most impressive evidence that ALT normal ranges are misleading comes from trials of treatment of hepatitis C virus infection (HCV) in patients with normal transaminase levels. These showed that if HCV was eradicated, the previously normal levels fell further.22
The normal population reference range is therefore very approximate for clinical purposes. Certain chromogenic drugs may also interfere analytically (acetaminophen, tetracycline and aminosalicylic acid), independently of liver injury.23
The American Gastroenterology Association24
defines a mild transaminase abnormality as up to 5 times ULN but does not define a lower limit for investigation. Several guidance documents refer to persistently raised transaminases of >1.5 ULN as warranting investigation. It has been suggested that increased levels of transaminases should either be investigated after 1–3 months4
or 3 months,25
or that there should be a staged approach to investigation over 0–6 months.24,26
Some expert reviews suggest treating the most likely cause of the abnormal transaminases—that is, alcohol (by abstinence), hepatotoxic drugs (by drug cessation) and NAFLD (by weight reduction and diabetes or triglyceride control, where applicable) before more detailed investigation if transaminases remain abnormal after a period of observation24,26
with a referral (biopsy) threshold of 2.5 ULN.5
All the guidance found emphasised the importance of clinical history and examination to identify possible causes (particularly NAFLD, which is described in one study as prevalent in 23% of American adults27
) before proceeding to further investigation in these cases.
The prevalence of abnormal transaminases will therefore vary with the type of population studied, the threshold used to define “abnormal” and the time interval over which the results are studied. For example, when transaminases are measured as part of routine preoperative assessment for elective surgery, 0.1–0.3% of patients are found to have abnormalities if abnormal is defined as transaminases >2 ULN but 3.5% if any degree of transaminase elevation outside the population “reference” range is considered to be abnormal.28
Most data that are available deal with abnormal AST or ALT levels in subsections of American populations, based on single transaminase measurements.
There is evidence that increasing transaminase levels (AST and ALT) are directly associated with increased liver‐related mortalities.29
In the UK, there is also evidence that the causes of abnormal transaminases are often not investigated. When a thorough investigation is carried out, however, most patients are found to have significant disease warranting treatment or follow‐up.30
No single study has compared the clinical or histological diagnoses carried out at a certain degree of transaminase abnormality with those at a different level of abnormality. There is no great difference between the percentage of abnormal liver biopsies found when investigating patients with transaminases any degree above the normal range, above 1.5 times or above twice the ULN. Investigating patients with a greater than threefold increase in transaminases may result in fewer normal liver biopsies—that is, fewer false positives—but will clearly result in an underdiagnosis of significant liver disease (table 2).31,32,33,34,35,36,37
The only reason to set a threshold for investigating abnormal transaminases seems to be one of expedience. Sherwood et al30
investigated patients with abnormal transaminases >2 times the reference range because lowering this threshold would have resulted in six times more patients to investigate, “well beyond our means of investigation”.
Table 2Findings on liver biopsy for abnormal transaminases
Table 1 shows an appropriate panel of initial investigations. There is no evidence to suggest that the level of enzyme abnormality in asymptomatic people has any influence on this risk of finding appreciable abnormalities on further investigation. In view of the poor correlation between the level of rise and biopsy findings, we do not recommend a lower threshold for carrying out basic investigations as shown in table 2 on persistently abnormal results.
As non‐alcoholic and alcoholic fatty liver diseases are the predominant causes, these can reasonably be sought from the clinical history before proceeding to more detailed (level 2) investigations.
How should I investigate abnormal transaminases in patients with risk factors for, or clinical features of, liver disease?
- Further investigations are advised for any enzyme rise associated with risk factors for, or clinical evidence of, liver disease except where an apparent cause is found (hypertriglyceridaemia, diabetes, alcohol and overweight) for which a 6‐month trial of appropriate intervention is recommended.
An asymptomatic person found to have abnormal liver enzymes should have a history taken for specific factors that may predispose to liver disease or suggest another cause. Alcohol consumption, prior medical interventions, such as surgery or transfusion, intravenous drug use or tattooing should be included, as should a detailed drug history, including non‐prescribed or herbal remedies.
Physical examination aims to detect stigmata of chronic liver disease, spider naevi, clubbing, hepatomegaly or splenomegaly. If present, these suggest significant liver disease, as these signs are rarely present in the absence of hepatic cirrhosis.
The history and examination should also aim to exclude non‐hepatic causes for liver enzyme increase. The most common causes that may be difficult to diagnose clinically are right‐sided heart failure or constrictive pericarditis, which can cause considerable hepatomegaly and abnormal liver tests with relatively modest symptoms, and endocrine disorders, notably obesity, diabetes and thyroid dysfunction.
There is good evidence that obesity38
cause abnormal liver enzymes via fatty liver, and the risk of death from liver disease was increased in the Verona Diabetes Study39
in patients with abnormal liver function tests. The condition is benign in those with simple steatosis40
although 10–30% of patients with considerable inflammation and particularly with fibrosis on index biopsy may progress to cirrhosis in 10 years.40,41
As yet, there is no specific medical treatment for NAFLD. Progressive weight loss can normalise liver enzyme levels and although there is no evidence that this improves long‐term outcome, it seems logical from a general health perspective. Note, however, that starvation dieting can also cause fatty liver.42
Risk factors for development of NAFLD, such as hypertriglyceridaemia and diabetes, should be dealt with, and a programme of exercise and weight reduction tried before liver biopsy is offered.
Further investigation is recommended if abnormal liver enzymes of any level are found in the presence of a major risk factor for liver disease or in particular if patients have stigmata of chronic liver disease.4,5,19,24
How should I investigate abnormal GGT in asymptomatic patients without risk factors for, or clinical features of, liver disease?
- GGT has limited use as a primary liver test. Its use is recommended in particular to establish the likely origin of a raised ALP.
- No clear consensus is found on thresholds for further investigation.
In the absence of guidance we recommend that:
- Mild rises (<3 ULN) justify interval retesting.
- Rises >5 ULN justify consideration of ultrasound if the cause is not apparent.
GGT is found in hepatocytes and in biliary epithelial cells. Its main use is in the diagnosis of biliary tract disease, in combination with ALP. It is widely and incorrectly regarded as a sensitive and specific marker of alcohol misuse.
The same statistical considerations apply to transaminases as to ALP, bilirubin and transaminases. Likewise, the actions to take in patients with risk factors for, or clinical evidence of, liver disease are the same, from the same documents. Some evidence, however, suggests that the yield from investigating raised GGT may be different from that for ALT or AST. The myth that a raised GGT is sensitive and specific for alcohol excess persists despite considerable evidence to the contrary. Studies have shown that the sensitivity of raised GGT for alcohol varies between 52% and 94%,43,44
depending on the population under study. Evidence also shows that the reference range for GGT is misleading; if a true reference range is established in non‐drinkers, ULN values of 45 for men and 35 for women are obtained.45
In a group with moderate alcohol intake, the levels rise to 66 and 40, respectively.
Hepatic fibrosis is present in only 11% of cases of asymptomatic patients with a raised GGT alone who do not drink excessively.46
By contrast, patients drinking excess alcohol who have an isolated raised GGT rarely had considerable liver damage.47
The same reviews that handled transaminases, ALP and bilirubin cite the primary purpose of GGT assessment as the clarification of the likely origin of an isolated ALP,2,4,5,8,24
used as a second‐line test. The British Columbian guidelines also stated that an isolated non‐progressive rise in GGT seldom reflects considerable liver disease, and a second review2
referred to raised GGT as a common reason for mistaken diagnosis of viral hepatitis or inappropriate referral for liver biopsy, which often shows no abnormality in cases of an isolated minor GGT rise.46
This point also appears in a Journal of Insurance Medicine
review, drawing the same conclusions as above and offering little support for the utility of isolated GGT levels for actuarial predictive purposes.6
Conversely, however, GGT is raised in 50% of cases of NAFLD.48
Therefore, there seems to be limited enthusiasm in the same reviews, which recommend the investigation of other slightly raised liver enzymes for the use of GGT levels in isolation, and consensus emerges that its main use lies in the clarification of a raised ALP. Differences in practice exist between laboratories: some do not include GGT in a liver profile, reserving it for second‐line use, whereas others do. As highlighted by Aranda‐Michel and Sherman,2
it is important that requesters recognise the limitations of this test.
When should I refer for secondary assessment and possible liver biopsy?
In the absence of clear guidance, we recommend:
- Laboratory results >2 ULN (unexplained by laboratory tests);
- Or following disease‐specific protocols if laboratory results show an underlying cause, to stage disease where applicable.
For most patients with abnormal liver enzyme levels, the pattern is non‐specific, commonly raised ALT, ALP and GGT. The only investigation likely to produce a diagnosis if the above investigations are normal is liver biopsy.
The pathological effects of alcohol, drugs and the non‐alcoholic fatty liver variants can only be diagnosed on liver biopsy. Percutaneous liver biopsy does, however, entail a finite risk, primarily of haemorrhage, and the decision to offer liver biopsy is therefore a balance between the potential benefits of a treatable diagnosis to prevent hepatic fibrosis and risk. This benefit is clear in conditions such as autoimmune hepatitis but less so in conditions such as alcoholic liver injury. Conditions with a specific treatment that can change the natural history of the liver injury are found in between 7% and 18% of patients undergoing biopsy for abnormal liver enzymes.
Transaminase of twice the ULN are only associated with a histologically normal liver in around 6% of cases.37
The most common diagnosis made on liver biopsy in this setting is either NAFLD or non‐alcoholic steatohepatitis37
(table 3). This diagnostic group accounts for two thirds of unexplained increase in liver enzymes. As above, however, clinical intervention data lacking the merits of liver biopsy in these situations are uncertain. Other important diagnoses include conditions such as antimitochondrial antibody‐negative primary biliary cirrhosis and autoimmune hepatitis. A marked fibrosis on biopsy is common (table 2). An isolated mild, isolated, non‐progressive rise in GGT is often not linked to liver disease and is rarely a reason for liver biopsy.
Table 3Findings on liver biopsy for unexplained abnormal liver enzyme increase in 354 patients37
GMS contract indicator: None.