Diagnostic and Prognostic Assessment of Suspected Drug-induced Liver Injury in Clinical Practice

Raúl J. Andrade; Mercedes Robles-Díaz


Liver International. 2020;40(1):6-17. 

In This Article

Diagnostic Tools for Drug-induced Liver Injury Assessment

Liver Tests for Liver Injury Assessment

The absence of diagnostic DILI biomarkers has led to that serum ALT/aspartate aminotransferase (AST), ALP and TBL levels still remain the pillars for DILI case detection and qualification.[27] Minor and reversible increases in ALT/AST that occurs with some drugs such as statins or that may indicate pre-existing liver disease (ie fatty liver) should not be classified as DILI. Moreover, ALT lacks specificity as a rise in serum values can also be related to other organ damages, most often muscle injury, which can be drug induced (ie rhabdomyolysis) and is generally accompanied by a disproportionate increase in AST with regard to ALT. Testing for creatine phosphokinase (CPK) can assist in distinguishing between liver- and muscle-driven ALT elevations. An isolated elevation in TBL usually related to its unconjugated fraction does not qualify as DILI as it generally indicates haemolysis or Gilbert syndrome.[16] Nonetheless, a rise in serum ALT is highly sensitive for hepatocyte injury, and when accompanied by an elevation in TBL becomes a reliable biomarker of liver injury in DILI and liver dysfunction (the so-called Hy's law, see the next section).[28] Cholestatic damage is characterized by markedly elevated serum ALP in association with raised gamma glutamyl transferase (GGT). However, an isolated elevation of GGT is insufficient to qualify as DILI as it does not indicate liver damage.[16] The performance of AST and GGT in replacing ALT and ALP, respectively, when the latter are unavailable at DILI recognition was analysed in a study of 588 patients included in the Spanish DILI Registry. Whereas AST values can reliably substitute ALT in calculating the pattern of injury, the utility of GGT in replacing ALP is limited.[29]

Importantly, liver biochemical analyses should be performed when DILI is first recognized as these values more accurately reflect the actual liver injury. However, abnormal liver tests when first found do not represent the true onset time of liver cell injury, which may already be advanced, subsiding or past.[30] To clarify this issue, serial aminotransferase measurements are necessary. Liver biochemistry should also be tested in DILI patients until complete normalization for diagnostic reassurance. Steady decline of aminotransferases supports the diagnosis of DILI, whereas flare-ups and/or incomplete resolution of biochemical abnormalities suggest competing aetiologies. In addition, persistently elevated aminotransferases may indicate a chronic outcome. Importantly, elevation of liver enzymes upon re-exposure to the suspected agent provides strong evidence for causality although the required threshold for this elevation is still controversial.[31]

Clinicians should also bear in mind that elevated serum aminotransferases inaccurately reflect the extent to which the liver is damaged in insidious or atypical varieties of DILI such as indolent fibrosis (methotrexate), sinusoidal obstruction syndrome, cirrhosis or microvesicular steatosis secondary to mitochondrial toxicity. In such instances, the threshold values defined for case qualification may not be reached and DILI must be suspected and diagnosed, according to compatible histological/imaging findings in the context of exposure to specific drugs/toxicants.[31]

Laboratory assessment of a DILI suspicion should also include coagulation parameters and serum albumin to further scrutinize potential severity of the liver damage. Elevated international normalized ratio (INR) values (>1.5), which indicates impending liver failure, should prompt referral to a liver transplant unit.

Serology and Other Laboratory Tests for Excluding Alternative Causes

Because of the current absence of specific biomarkers, the diagnosis of DILI still relies on the exclusion of alternative causes of liver injury. Classification of injury pattern can assist in the initial diagnostic approach guiding the necessary work-up to exclude the most common causes of hepatitis and cholestasis (Figure 1). Patient age and a detailed medical history to exclude alcohol abuse, comorbidities (such as sepsis, congestive heart failure, recent episodes of syncope or hypotension, which would indicate ischaemic hepatitis), should be retrieved. In addition, ascertainment of risk factors for viral hepatitis and the local burden of infectious diseases potentially affecting the liver are paramount to correctly assess the case.

As a first step, serology tests for viral hepatitis A, B, C and E should be performed, particularly in patients with hepatocellular and mixed type of liver damage. Potentially challenging cases include, for example, patients who are hepatitis B surface antigen (HBsAg) carriers, in whom hepatitis B virus (HBV) reactivation as the cause of liver injury should be excluded by testing HBV-deoxyribonucleic acid (HBV-DNA). As there is no specific biomarker for acute hepatitis C (HCV), this variety of viral hepatitis can also be misdiagnosed as DILI. Indeed, in 1.3% of adjudicated DILI cases in the Drug-Induced Liver Injury Network (DILIN) prospective cohort, HCV-ribonucleic acid (HCV-RNA) tested positive making the diagnosis a challenge.[8] Besides, hepatitis E (HEV) is a common cause of viral hepatitis in Eastern countries but it is also an emerging cause in Western countries and can subsequently be a DILI confounder.[32,33] Anti-HEV IgM-positive cases ranged from 3% in the DILIN database[33] to 7% in the Spanish DILI Registry.[34] Drugs initially thought to be responsible for the anti-HEV IgM positive DILI cases in Spain actually had low hepatotoxicity potential, showed less compatible temporal sequences and/or presented with higher aminotransferase levels compared with anti-HEV IgM-negative cases.[34] However, anti-HEV serology has not yet reached consensus worldwide as a diagnostic test for active HEV infection.[35] Despite this limitation, a search for HEV infection as an alternative diagnosis is advisable in patients being assessed for DILI, particularly in cases in which the time to onset is less compatible with the drug signature of the suspected medication and in those with transaminase levels in the range of viral hepatitis.

When suspected DILI presents with a hepatocellular pattern, autoimmune hepatitis (AIH) is a potential alternative diagnosis that should be evaluated with autoantibodies (antinuclear antibodies (ANA); anti-smooth muscle antibodies (ASMA)) and serum IgG. However, DILI associated with drugs such as nitrofurantoin, minocycline, anti-tumour necrosis factor (TNF)-α and statins among many others[36–38] can exhibit an AIH-like phenotype indistinguishable from idiopathic AIH, making the differential diagnosis a challenge. In such instances, the history of exposure to the medication and a resolution of biochemical abnormalities with no relapse either spontaneously or upon corticosteroids tapering and withdrawal support the diagnosis of drug-induced AIH. Likewise, when suspected hepatotoxicity presents with a cholestatic pattern, primary biliary cholangitis needs to be excluded by anti-mitochondrial antibody (AMA) testing.[15] Alcoholic hepatitis rarely masquerades as DILI; a history of alcohol abuse with a predominance of AST over ALT elevation with ALT values usually below 300 IU/L and other biochemical features of chronic alcoholism such as high values of GGT and erythrocyte mean corpuscular volume make the diagnosis evident.

In younger patients with acute or chronic hepatitis, Wilson's disease should be ruled out by measuring ceruloplasmin levels. When ceruloplasmin—an acute phase reactant—is diminished or only slightly decreased, which may occur in Wilson's disease presenting as acute hepatitis, a 24-hour urine cooper excretion, ophthalmologic examination for Kayser-Fleischer rings and genetic testing of the ABCB7 gene are required.[39] Ischaemic hepatitis needs to be excluded in older patients and those with pre-existing cardiac disease, although prior hypotension or syncope was documented in only 53% of the cases in a systematic review of ischaemic hepatitis.[40] Very high aminotransferases values with a predominance of AST over ALT typically followed by a fast resolution is the biochemical hallmark of liver ischaemia.

As a second step, testing for cytomegalovirus, Epstein-Barr virus and herpes virus infection is usually performed. However, this should only be mandatory when liver damage is seen in association with extrahepatic manifestations such as rash, lymphadenopathy and atypical lymphocytes.


Liver imaging in DILI is used to exclude alternative aetiologies. An abdominal ultrasound is advisable in all DILI suspicions regardless of the biochemical pattern of damage to evaluate the biliary tract and to exclude parenchymal focal lesions. Additional imaging techniques would be justified in the clinical context of accompanying prominent abdominal pain and/or mixed or cholestatic injury. Thus, computerized tomography and magnetic resonance cholangiography (MRCP) are sometimes required to exclude gallstone disease and other competing aetiologies.[31] In rare instances, toxic damage to the biliary tract presenting as sclerosing cholangitis has been described with chemotherapeutic agents such as 5-fluorodeoxyuridine after hepatic intra-arterial infusions for treatment of hepatic metastases; these are consequences of ischaemic injury to the biliary tract rather than toxicity.[41,42] Likewise, dilatation of bile ducts has been attributed to ketamine abuse in some case reports. Secondary sclerosing cholangitis has also been reported in association with methimazole and docetaxel.[43] Recently, it was reported that a small proportion of unselected acute cholestatic/mixed DILI cases undergoing MRCP or endoscopic retrograde cholangiopancreatography (ERCP) had secondary sclerosing cholangitis-like changes. The implicated agents varied and included amoxicillin-clavulanate, amiodarone, atorvastatin, gabapentin, infliximab, 6-mercaptopurine, sevoflurane and venlafaxine. All the 10 patients were females and 70% presented with jaundice and had a longer time to resolution.[44] Thus, in a patient being assessed for suspected DILI, the identification of sclerosing cholangitis-like changes does not necessarily mean that the subject has primary sclerosing cholangitis as an alternative diagnosis.[31]

Liver Biopsy

Histological assessment in acute and chronic liver diseases is currently less frequently indicated than non-invasive tests, which are considered reliable particularly for staging fibrosis. Moreover, in acute liver injury, the degree of inflammation may lead to increased values of transient elastography, overestimating fibrosis.[45] However, as DILI lacks specific serum biomarkers, liver biopsy has long been considered a complementary diagnostic tool that can assist and reinforce the diagnostic process.[46] In a review of liver biopsies from 249 patients with DILI from a prospective observational cohort, the authors tried to establish correlations between pre-defined histological patterns and biochemical phenotypes. Although the hepatocellular and cholestatic biochemical patterns did not match perfectly with their histological counterparts, more severe inflammation and cell death were associated with hepatocellular pattern compared to higher frequency of bile plugs and ductal paucity in those with cholestasis.[47]

However, liver biopsy is not routinely performed in suspected DILI cases because it does not provide definite diagnostic information in most instances. Notable exceptions are when the de-challenge is incomplete or negative after drug discontinuation, which makes an alternative diagnosis more likely or the presentation suggest one of the phenotypes listed in Table 1, which require histology for a full characterization. One example that requires a liver biopsy is the appraisal of ductopenia caused by some agents that can lead to vanishing bile duct syndrome.[48] Also, in the case of AIH, whose diagnosis is established in patients with hepatocellular injury on the grounds of the detection of typical serum autoantibodies and elevated IgG, a compatible liver histology strongly reinforces the diagnosis.[31] Moreover, DILI can be indistinguishable from AIH even after detailed investigations, as described for 9% of instances.[49] Indeed, serum criteria used for the diagnosis of AIH are largely unspecific as a high prevalence of ANA (15%-24%), ASMA (up to 43%), anti-liver-kidney-microsomal antibody (anti-LKM, 1%) and raised immunoglobulin G levels (5%) can be found among asymptomatic individuals.[50] Hence, histological findings, albeit not pathognomonic, are included in the simplified diagnostic scale currently used for the diagnosis of AIH[51] making liver biopsy a necessary tool to properly assess AIH including cases suspected to be drug induced. Histological findings of AIH (n = 28) and DILI (n = 35) were blindly assessed by three expert pathologists in a study; hepatocellular cholestasis and portal neutrophils was indicative of DILI, while presence of fibrosis suggested the diagnosis of AIH.[52] Using dual immunohistochemistry staining of liver biopsies to characterize portal inflammatory infiltrates in another study, it was shown that inflammatory cells in DILI (that included cases of drug-induced AIH) were predominantly cytotoxic (CD8+) T cells, whereas mature B cells (CD20+) were more prominent in AIH.[53] Nevertheless, long-term follow-up of patients after drug discontinuation would be required to differentiate idiopathic from drug-induced AIH as the latter does not usually recur after withdrawal of the drug and resolution of liver damage.[54]

Genetic Testing

Recent genome-wide association studies have identified a number of human leucocyte antigens (HLA) genotypes and haplotypes associated with DILI related to a selected group of drugs. Nowadays, HLA genotyping is widely accessible, affordable and can assist diagnosis in selected clinical contexts.[55] As with most polygenic disorders, the pre-treatment value of genetic testing (to prevent DILI in carriers of a specific allele) is very low, but the high negative predictive values (>95%) of some of these alleles can be used to exclude DILI when the subject is not a carrier and the clinical picture could be ascribed to an alternative aetiology. Besides this, when the subject is receiving a combination of medications, genetic testing may help to clarify the role of a particular drug if the patient carries a specific HLA allele associated with hepatotoxicity for one of the agents. An additional value of HLA typing could be in the differential diagnosis of DILI vs AIH as carriage of an HLA risk allele associated with a specific drug[56] (eg HLA-A* 33:01 in a subject taking terbinafine that experience acute liver injury with autoimmunity features) would favour the diagnosis of DILI, whereas its absence and the presence of the typical HLA allele associated with AIH, such as HLA-DRB1*03:01 and DRB1*04:01 would support the diagnosis of AIH. Although HLA typing is applicable to a very limited group of drugs, genome-wide association studies (GWAS) have also recently identified non-HLA genetic variants associated with DILI in general, which could also be useful for clarifying ambiguous cases.[57]

Scales Used in DILI Causality Assessment

The lack of specific tests or biomarkers to confirm a DILI diagnosis makes it very important to include a systematic evaluation to confidently attribute a liver injury episode to a drug. Several causality assessment methods specific for DILI have been developed over the past decades, which provide a framework for a more objective evaluation in suspected cases of DILI. Of the several diagnostic scales in place, the Council for International Organizations of Medical Sciences (CIOMS) scale, also called Roussel Uclaf Causality Assessment Method (RUCAM), is still considered the most reliable and reproducible method that correlates better with expert review.[58] RUCAM gives points to seven distinct domains: (a) temporal relationship between exposure to a particular drug and liver injury (both its onset and course), (b) exclusion of alternative non-drug-related aetiologies, (c) exposure to other medications that could explain DILI, (d) risk factors for the adverse hepatic reaction, (e) evidence in the literature regarding DILI from the drug in question and (f) response to re-exposure to the medication. The total score ranges from −9 to +10 and classifies the event as highly probable (>8), probable (6–8), possible (3–5), unlikely (1–2) or excluded (≤0) according to its likelihood of being DILI.[59] However, the RUCAM scale has some limitations: (a) when there is missing information (frequently when reviewing retrospective cases) or (b) no data on de-challenge (cases of acute liver failure), (c) when the patient takes multiple drugs during the same time period or (d) when a drug typically produces delayed DILI (eg amoxicillin-clavulanate). In addition, (e) the questions posed by the scale require some degree of subjectivity by the user to interpret as well as answer. Also, (f) the value added by the risk factors is controversial. Despite its limitations, the RUCAM is the most commonly used diagnostic tool for DILI, and its use increases consistently and objectivity in causality assessment.[60]