Value of liver biopsy in the diagnosis of drug-induced liver injury

Abstract

Background & Aims:

The utility of liver biopsy in diagnosing or staging idiosyncratic drug-induced liver injury (DILI) is unclear. The aim of this study was to determine whether liver histology impacted causality assessment in suspected DILI using a novel simulation model.

Methods:

Fifty patients enrolled in the DILI Network (DILIN) who had liver biopsies performed within 60 days of DILI onset were randomly selected. All had standard DILIN consensus causality scoring using a 5-point scale (1=definite, 2=highly likely, 3=probable, 4=possible, 5=unlikely) based on 6-month post-injury data. Three experienced hepatologists independently performed causality assessment using redacted case records with the biopsy and selected post-biopsy laboratory data removed. The 3 hepatologists also reviewed the liver histology with a hepatopathologist and then repeated causality assessment for each case.

Results:

Of the 50 cases, there were 42 high causality DILI Cases (1,2 or 3) and 8 low causality cases (4,5). The hepatologists judged that liver biopsy was indicated in 62% of patients; after histology review, biopsy was judged to have been helpful in 70% of patients. Histology review changed the causality score in 68% of patients, with an increase in DILI likelihood in 48% and a decrease in 20%. Biopsy results changed diagnostic certainty from less certain (3 or 4) to highly certain (1,2 or 5) in 38% of patients.

Conclusions:

Liver histologic findings may help clarify the diagnosis of DILI. Histology appears to be particularly helpful in cholestatic or equivocal cases of DILI (possible or probable), shifting assessment toward a greater or less certainty of a DILI diagnosis.

Lay Summary:

The results of a liver biopsy in patients with suspected drug-induced liver injury (DILI) can help physicians in being more certain that the liver injury is due to a drug or another cause.

Introduction

The diagnosis of drug-induced lived injury (DILI) remains challenging in the absence of specific biomarkers. DILI is suspected in cases of elevated liver biochemical tests with a compatible history of exposure to a drug, herbal product, or dietary supplement (HDS). As DILI has a wide spectrum of clinical presentations that can mimic virtually any liver disease including viral and autoimmune hepatitis, fatty liver, vascular liver disease, benign liver tumors, biliary tract disease and malignancy, it is important to exclude these possibilities when assessing causality. In addition, data on concomitant medications, latency (time from drug start to injury onset), rate of decline in liver biochemistries after drug stop (dechallenge) and rechallenge are also critical.

Expert consensus opinion is used to assess causality in the Drug-Induced Liver Injury Network (DILIN), a National Institutes of Health (NIH)-funded multicenter observational cohort study that enrolls patients with suspected DILI in the United States (1). Causality assessment on a 5 point categorical scale is undertaken only after a 6 month follow up visit has been completed. A clinical narrative summarizing the initial presentation and clinical and laboratory outcome are provided along with a summary of all available laboratory, radiological, and histological data.

Other diagnostic methods for causality assessment include the Roussel Uclaf Causality Assessment Method (RUCAM) (2). The RUCAM score uses 8 factors in 7 domains (age over 55 years; presence of alcohol; latency (time from drug start to liver injury); time to improvement of liver injury; exclusion of other causes; hepatoxicity published or in product label; concomitant medications and positive rechallenge) and assigns points in each to produce an overall score. It has 5 categories- highly probable, probable, possible, unlikely and excluded. Liver histology is not required for DILIN enrollment, nor is it one of the domains included in the RUCAM score. However, nearly 50% of the patients enrolled into DILIN do undergo a liver biopsy for either diagnostic, prognostic, or therapeutic reasons. In clinical practice, liver biopsy is usually undertaken in a situation where the diagnosis is uncertain or there is concern regarding the severity of the injury. Typically, this will occur early during the course of the injury and may guide subsequent management such as the initiation of steroids or referral for liver transplantation.

Performing a randomized clinical trial to determine if liver biopsy affects causality assessment in DILI is not feasible. Using patients that had already undergone a liver biopsy as part of the work up for suspected DILI introduces a selection bias as the initial provider or enrolling investigator felt a biopsy was necessary and may have influenced subsequent management. To try and determine whether liver histology findings impact causality assessment, this study assessed causality using a simulated dataset prior to the time when an actual liver biopsy was obtained and then repeated causality assessment after reviewing the liver biopsy in a well-characterized cohort of patients with suspected DILI.

Methods

This study was performed in patients already enrolled in DILIN, and overarching methods of the DILIN network, as well as methods unique to the present study, are outlined below. A more detailed description of the DILIN prospective study design has previously been published (1).

DILIN methods

Enrollment criteria:

The DILIN prospective study is a multicenter observational cohort study that has been enrolling patients with suspected DILI in the US since 2004. Eligible patients must meet predefined laboratory criteria including elevated serum aspartate aminotransferase (AST) or alanine aminotransferase (ALT) or serum alkaline phosphatase (Alk P) levels with/without jaundice or coagulopathy. Liver biopsy is done at the discretion of the local health care provider as part of standard care, but biopsy reports and slides are obtained for central review whenever possible. Subjects are required to be enrolled within 6 months of onset.

DILIN study design:

At the time of enrollment, patients undergo a thorough medical history with details of drug or HDS exposure. Additional laboratory testing is performed for other causes of liver injury that were not adequately excluded at the time of the initial evaluation. Patients are asked to return for a follow-up visit at 6 months after enrollment, and those with evidence of persistent liver injury to return again annually for up to 4 years. The DILIN prospective study was approved by the Institutional Review Boards at each clinical site and data coordinating center and by a central Data Safety and Monitoring Board appointed by the NIDDK. All enrolled subjects provided written informed consent.

Causality assessment:

Enrolled subjects undergo a formal causality assessment process after their 6 month follow up visit using two methods: consensus expert opinion by 3 hepatologists (3) and the RUCAM (calculated by the enrolling principal investigator). The hepatologists are provided a clinical narrative and a comprehensive data worksheet. Each investigator scores the case independently, and then the 3 reviewers come to a consensus during a causality committee conference call, with a majority vote of the full committee if there are persistent disagreements over the final causality score. The DILIN causality system employs a 5-point likelihood score: 1 (definite: ≥ 95% likelihood), 2 (highly likely: 75%−94% likelihood), 3 (probable: 50%−74% likelihood), 4 (possible: 25%−49% likelihood) or 5 (unlikely: <25% likelihood). Scores of 1, 2 and 3 are considered high confidence DILI while scores of 4 and 5 are considered non-DILI and an alternative diagnosis is recorded wherever feasible (or classified as unknown). In subjects in whom more than 1 agent is implicated, an overall causality score is assigned, and separate causality scores are assigned to each suspected drug or HDS.

Re-assessment of causality with liver biopsy information:

For the purposes of this study, we randomly selected 50 cases who had had a liver biopsy that was available to the central DILIN pathologist (DEK) and was adequate for interpretation (with at least 10 portal tracts) (Figure 1). A sample size calculation was not possible as there was no prior data on the effects of biopsy on DILI causality. The 50 cases were selected out of a total of 413 cases that met the criteria for the study. To mirror the early diagnostic practice situation, biopsies had to have been performed within 60 days of the liver injury onset and the dates of all the other investigations carefully noted. A new clinical narrative and data worksheet were prepared by a single investigator with the liver biopsy report and all the clinical information after the date of the liver biopsy redacted. Three DILIN investigators, who had not previously reviewed the case, scored these 50 simulated cases independently and came to consensus causality score (pre-biopsy score). The hepatologists were asked if they would have requested a liver biopsy based on the information available from the redacted clinical narrative and data worksheet. The central DILIN pathologist reviewed digitalized liver biopsy slides during a video conference call with the reviewers to recreate a real time consultative experience and discussion. Immediately after each biopsy was discussed, the reviewers rescored the case by consensus producing a post-biopsy causality score, without the benefit of any additional laboratory, imaging or outcome data. They also assessed whether the biopsy was helpful or not even if the score did not change.

Figure 1.

Flow diagram of study design

Statistical Considerations:

Selected characteristics such as age, latency, peak enzyme values, causality score and severity score were compared between those with and without biopsy where two sample non-parametric Wilcoxon test was used for continuous variables and Chi-square test was used for categorical variables. The marginal distributions of the causality likelihood were compared between those with and without biopsy by weighted least squares approach via SAS procedure CATMOD to take into account of correlation of causality scores within the same subject. If the overall comparison is statistically significant, comparisons of probabilities between those with and without biopsy at each causality likelihood were carried out to determine which causality category had significant difference. All P values reported are two-sided and a level of 0.05 or less is considered statistically significant.

Results

Reviewing the entire DILIN prospective cohort enrolled between 2004 and 2020, almost half of all patients underwent a liver biopsy (Table 1). Biopsied patients had a longer latency and more severe liver injury but the distribution of causality scores were similar. The median age of the 50 simulated cases was 52 years (range 19–85) with an equal sex distribution, 41 (82%) were jaundiced and the median time from presentation to liver biopsy was 14 days (range 2–59 days) (Table 2). The majority of patients had hepatocellular injury (56%). Of the 50 cases, 14 (28%) involved implicated drugs that are considered well known hepatotoxins (e.g. amoxicillin-clavulanate, sulfamethoxazole/trimethoprim, minocycline and nitrofurantoin) (6). As expected, the majority of patients had viral hepatitis testing, autoimmune serology and abdominal imaging available at the time of the biopsy. Overall, 42 (84%) patients were considered to have high confidence DILI (17 definite, 21 highly likely and 4 probable) and 8 (16%) were attributed to other etiologies (5 possible and 3 unlikely) (Table 3).

Table 1.

Characteristics of 2099 adjudicated cases in the DILIN prospective cohort from 2004–2020.

Total N=2099	Liver Biopsy N= 1022	No Liver Biopsy N=1077	P value
Median Age (years)	52.3	50.5	0.03
Median Latency (days)	60.0	35.5	<0.001
Median peak values
ALT (U/L)	1077	963	<0.001
Alk P (U/L)	297	252	<0.001
Tbili (mg/dL)	14.0	5.6	<0.001
Causality Score			0.19
Definite	158 (15.5%)	190 (17.6%)
Highly Likely	410 (40.1%)	453 (42.1%)
Probable	220 (21.5%)	230 (21.4%)
Possible	164 (16.0%)	147 (13.6%)
Unlikely	70 (6.8%)	57 (5.3%)
Severity score			<0.001
Mild	170 (16.6%)	351 (32.6%)
Moderate	128 (12.5%)	251 (23.3%)
Moderate-hospitalized	369 (36.1%)	293 (27.2%)
Severe	254 (24.9%)	138 (12.8%)
Fatal	101 (9.9%)	44 (4.1%)

Table 2.

Clinical features of 50 patients with liver biopsy performed within 2 months of suspected DILI onset enrolled in the DILIN prospective study

Feature	Value (range or per cent)
Median age (years)	52 (19–85)
Female sex	25 (50%)
Peak median ALT (U/L)	526 (66–2894)
Peak median Alk P (U/L)	214 (48–1406)
Peak median total bilirubin (mg/dL)	10.3 (0.5–32.6)
Jaundice (total bilirubin ≥3 mg/dL)	41 (82%)
Median time after DILI onset for liver biopsy (days)	14 (2–59)
Hospitalized at biopsy	28 (56%)
Median ALT (U/L) at biopsy (within 7 days) (N=43)	288 (57–2447)
Median Alk P (U/L) at biopsy (within 7 days) (N=44)	195 (39–1310)
Median total bilirubin (mg/dL) at biopsy (within 7 days) (N=44)	8.9 (0.5–32.6)
Corticosteroids given after biopsy	11 (22%)
Injury type
Hepatocellular	28 (56%)
Cholestatic	9 (18%)
Mixed	13 (26%)
Single implicated drug or HDS	38 (76%)
Implicated drug class
Antibiotics	19 (38%)
Anti-lipid	7 (14%)
HDS	5 (10%)
Chemotherapeutic	3 (6%)
Others	16 (32%)
Some viral hepatitis serology available at time of the biopsy	45 (90%)
Negative anti-HAV IgM	41/41 (100%)
Negative HBsAg	42/43 (98%)
Negative anti-HBc IgM	32/32 (100%)
Negative anti-HCV	37/42 (88%)
Negative HCV RNA	12/15 (80%)
Negative HEV RNA	6/6 (100%)
Negative anti-HEV IgG	38/50 (76%)
Negative anti-HEV IgM	49/50 (98%)
ANA/SMA available at time of the biopsy	41 (83%)
Negative ANA	24/40 (60%)
Negative SMA	30/33 (91%)
Abdominal imaging available at time of the biopsy	47 (48%)

Abbreviations: HAV, hepatitis A virus; HBsAg, hepatitis B surface antigen; HBc, hepatitis B core; HCV, hepatitis C virus; ANA, anti-nuclear antibody; SMA, smooth muscle antibody

Table 3.

Causality scores of 50 patients with liver biopsy performed within 60 days of suspected DILI onset enrolled in the DILIN prospective study

	Total Number	Definite	Highly Likely	Probable	Possible	Unlikely
Adjudicated DILIN Causality Score	50	17	21	4	5	3
Pre-biopsy Consensus Score	50	1	17	16	15	1
Post-biopsy Consensus Score	50	8	21	9	5	7
Biopsy requested (based on pre-score)	31	0	4	13	13	1
Biopsy helpful (based on pre-score)	35	1	6	16	12	0
No change in causality score (based on pre-score) after biopsy	16	1	10	1	3	1
Change in causality score (based on pre-score) after biopsy	34	0	7	15	12	0

After assessing the redacted clinical data prior to review of the liver biopsy (the pre-biopsy score), blinded review by 3 independent reviewers resulted in perfect agreement in 15 cases (30%, confidence interval (CI): 17%, 43%) and two out of three agreement in 24 cases (34%, 95% CI: 34%, 62%), implying an overall total or partial agreement of 78% (95% CI: 67%, 89%). The pairwise weighed Kappa statistics for assessing agreement between any two reviewers were (95% CI) were 0.50 (0.33, 0.66), 0.33 (0.14, 0.51), and 0.37 (0.10, 0.57). After reviewing the redacted clinical narrative and data worksheet that excluded liver histology of the 50 simulated cases, reviewers would have requested a liver biopsy in 31 (62%, 95% CI: 49%, 75%) patients. Of these 31 cases, there were 13 probable and 13 possible DILI cases based on the pre-biopsy consensus score. Reviewers were more likely to request a biopsy in cases with pre-biopsy causality scores of 3,4,5 (84%) vs those with scores of 1,2 (22%, p<0.001). After reviewing the liver histology, reviewers reported the biopsy as helpful in 35 (70%) of cases (Table 3). Histology review led to a change in the causality score in 34 (68%, 95% CI: 55%, 81%) cases, with an increase in consensus likelihood of DILI in 24 (48%) and a decrease in 10 (20%). Of the 16 probable cases on pre-biopsy score, only 1 case did not change score after biopsy review, with 11 cases moving to highly likely, 2 cases to definite and 2 cases to possible. Similarly, 12 of the 15 possible cases on pre-biopsy score changed after biopsy review with 6 cases moving to probable and 6 cases to unlikely. In 8 (16%) cases, the biopsy changed the causality score in a clinically meaningful way by crossing the 50% likelihood line from high confidence DILI (scores 1–3) to possible or unlikely DILI (scores 4,5) in 2 (4%) or from unlikely DILI to high confidence DILI in 6 (12%) (Table 4). There were no significant differences between these 8 cases and the other 42 cases in terms of implicated drug, age, time of biopsy, peak ALT and peak bilirubin but the peak Alk P was higher (p<0.006) suggesting an enrichment of cholestatic cases. A breakdown of the individual pre- and post-biopsy causality scores prior to reaching consensus scores is given in the supplementary material.

Table 4.

Characteristics of 8 patients with changes in causality scores from a 4 to a 3 or a 3 to a 4 after liver biopsy review.

Age years	Sex	Injury	Implicated Drug	Time to biopsy [days]	Pre-Bx score	Post-Bx score	Peak ALT [U/L]	Peak Alk P [U/L]	Peak T bili [mg/dL]	Viral serology	AI serology	Imaging	Histological Findings and comment
85	F	HC	Levofloxacin	2	4	3	980	120	3.2	HCV RNA Pos	Pos	Neg	Chronic cholestasis and acute hepatitis superimposed on underlying chronic viral hepatitis and NASH. Despite two preexisting liver diseases, the pathology cannot be explained by either. Fluoroquinolones have been associated with both acute hepatitis and cholestatic injury patterns.
67	M	Chol	Voriconazole	7	4	3	606	1310	9.7	Neg	NA	Neg	Acute cholestasis and necrosis superimposed on HSCT-associated NRH and obliterative portal venopathy. Other cause of cholestasis excluded by history and azole antifungals are associated with both acute hepatitis and cholestatic injury patterns.
48	F	Chol	Azithromycin	12	4	3	239	1406	5.3	Neg	Neg	Neg	Changes of chronic cholestasis superimposed on HSCT or AML-associated NRH and obliterative portal venopathy. Although the cholestatic findings are subtle, they coincide with the clinical picture cholestatic injury and azithromycin is known to cause chronic cholestatic injury.
62	M	HC	Amiodarone	29	4	3	781	168	7.7	Neg	Pos	Bil dil	Cirrhosis with numerous Mallory-Denk bodies and mild steatosis. The injury is typical for cirrhosis related to amiodarone.
52	M	Chol	Cefuroxime	28	4	3	268	651	8.4	Neg	Neg	Neg	Chronic cholestasis with zone 3 necrosis, no changes to suggest PBC or PSC. Unusual combination injury favors DILI and cephalosporins are associated with cholestatic injury patterns.
26	F	HC	Infliximab	25	4	3	2894	200	4.4	Neg	Neg	Neg	Acute hepatitis with prominent plasma cell infiltrates. Infliximab is associated with AIH-like pathology and the autoimmune serologies were negative.
46	F	Chol	Amoxicillin	9	3	4	450	723	25.7	Neg	Pos	Neg	Acute hepatitis with diffuse parenchymal injury, venulitis, no cholestasis. Atypical for amoxicillin, which typically shows cholestatic hepatitis.

Abbreviations: F, female; M, male; HC, hepatocellular; Chol, cholestatic; Bx, biopsy; HCV, hepatitis C virus, NA, not available; Bil dil, biliary dilatation; AIH, autoimmune hepatitis; HSCT, hematopoietic stem cell transplant; NRH, nodular regenerative hyperplasia; AML, acute myelogenous leukemia; PBC, primary biliary cholangitis; PSC, primary biliary sclerosis.

Overall, liver biopsy review significantly changed the distribution of causality scores (P < 0.001) with significant changes in the definite (P=0.004), possible (P=004) and unlikely (P=0.009) categories. This was due to changes from the probable or possible categories to definite/highly likely or unlikely in 19 (38%, 95% CI: 25%, 51%) cases, making the diagnosis of DILI or excluding it with higher certainty (Figure 2). The post-biopsy review causality scores were very similar to the original DILIN causality scores (Table 3) in all causality categories except for the definite category where there is significantly higher percent of definite cases in the original adjudication than post-biopsy review without follow-up data (P=0.01).

Figure 2.

Distribution of 50 simulated causality scores pre- and post-liver biopsy review

Reviewing the 50 cases based on the suspected DILI agent, there were 13 cases that were associated with very commonly seen drugs in the DILIN (amoxicillin-clavulanate, minocycline, nitrofurantoin and trimethoprim-sulfamethoxazole). Of these 13 cases, 8 cases were among the 19 cases (42%) where a biopsy was not requested and 5 cases were amongst the 31 cases (16%) where a biopsy was requested. In the 8 cases where a biopsy was not requested in all 8 cases the reviewers felt the biopsy was unhelpful and not necessary as the clinical picture was very characteristic. Similarly, in the 5 cases where a biopsy was requested in all 5 cases the reviewers felt the biopsy was helpful.

In analyzing the discussion after reviewing all 50 cases, the biopsy suggested a therapeutic suggestion in 6 cases (to treat with corticosteroids) and in 3 cases the reviewers suggested additional studies (check Epstein-Barr virus serology in one case and check hepatitis E serology in 2 cases).

Representative examples of 2 the 8 cases in which the causality score changed are highlighted below:

Case Vignette 1:

A 52 year-old African-American male with a history of hypertension, diabetes, hyperlipidemia, remote heavy alcohol use and asthma was prescribed a 10 day course of cefuroxime 500 mg three times daily for folliculitis. He presented 5 weeks later with ALT 268 U/L, AST 506 U/L, Alk P 261 U/L and total bilirubin 8.4 mg/dL. Complete blood count was normal without eosinophilia, and serologies for hepatitis A, B and C and autoimmune hepatitis were negative. Abdominal imaging showed no biliary dilation but a fatty liver with morphologic changes consistent with cirrhosis. His other medications included stable doses of insulin, lisinopril, hydrochlorothiazide, gabapentin, cyclobenzaprine and albuterol and fluticasone/salmeterol inhalers. A skin biopsy showed suppurative folliculitis with nodular changes. However, his liver tests remained stably elevated without improvement for another week.

The pre-biopsy causality score was possible DILI (rated “4” by the causality committee) as there was concern for a systemic process related to the skin rash and the possibility of underlying cirrhosis. Given the concern about underlying liver disease, after review of the redacted clinical data, all 3 hepatologists felt assessment of liver histology would be helpful in differentiating DILI from possible underlying cirrhosis with cholestasis.

A liver biopsy had been performed 28 days after the initial liver injury onset in this patient (Figure 3). Liver histology revealed marked chronic cholestatic change with both central necrosis and evidence of early bridging fibrosis (F3). In addition, plasma cell infiltration was identified in zone 3. This combination of findings could not be ascribed to chronic cholestatic liver diseases such as primary biliary cholangitis or autoimmune hepatitis. Although the pathology of cefuroxime injury is not well characterized, cephalosporins have been associated with biochemical cholestatic injury (4). After reviewing the histology findings, DILIN hepatologists felt this was probably a DILI event (changing the DILIN causality from 4 (possible) to 3 (probable)). During follow-up, the patient experienced gradual improvement in his liver biochemical tests without any corticosteroids over the following 3 months with normalization of his bilirubin.

Figure 3 A and B.

Representative liver pathology from a 52-year-old male with suspected DILI attributed to cefuroxime.

Figure 3A: Zone 3 necrosis with plasma cell infiltrates (H&E stain, 200x)

Figure 3B: Chronic cholestatic changes characterized by pseudoxanthomatous changes and ductular reaction (H&E stain, 400x)

Case Vignette 2:

A 46 year-old Caucasian female with a history of morbid obesity and gastric bypass surgery several years prior was prescribed amoxicillin for a tooth infection. She presented several days later with fatigue, pruritis and dark urine and then developed jaundice. Initial laboratory studies showed ALT 450 U/L, AST 960 U/L, Alk P 723 U/L and total bilirubin 20.4 mg/dL. Serologies for hepatitis A, B and C were negative but her ANA and smooth muscle antibody were mildly positive. Her other medications included esomeprazole at a stable dose. Her liver tests improved but her bilirubin remained elevated for several days and she developed some mental status changes and was transferred to a tertiary care center.

The pre-biopsy causality score was probable DILI (rated “3” by the causality committee) as there was adequate exclusion of other causes. However, given the short latency, positive autoimmune markers and failure to improve, after review of the redacted clinical data, all 3 hepatologists felt assessment of liver histology would be helpful in differentiating DILI from other causes of liver injury.

A liver biopsy had been performed 10 days after the initial liver injury onset in this patient (Figure 4). Liver histology revealed a diffuse parenchymal injury with lobular disarray and hepatocyte rosette formation with numerous acidophil bodies, venulitis and no cholestasis. There was mild portal inflammation with scattered eosinophils and occasional plasma cells. This pattern of injury is atypical for amoxicillin where cholestatic injury is more likely and hence was more suggestive of an acute hepatitis. After reviewing the histology findings, DILIN hepatologists felt this was less likely DILI (changing the DILIN causality from 3 (probable) to 4 (possible)) and were concerned about atypical viral hepatitis, particularly hepatitis E. During follow-up, the patient was treated with ursodiol and corticosteroids with gradual improvement in her liver biochemical tests. No hepatitis E testing was available at the time and the patient did not return for long-term follow up.

Figure 4 A-D.

Suspected amoxicillin injury with acute hepatitis on biopsy. The pattern of injury is atypical for amoxicillin. A. Diffuse parenchymal injury with lobular disarray and hepatocyte rosette formation (H&E, 200x). B. Numerous acidophil bodies are seen (arrows) (H&E, 600x). C. Portal areas showed mild inflammation with scattered eosinophils and occasional plasma cells (H&E, 400x). D. The central veins showed venulitis. (H&E, 400x).

Discussion

The role of liver histology in the management of DILI is uncertain but experts in this study believed that it was helpful in the majority (62%) of cases when performed within the first 60 days after liver injury onset. After reviewing the liver histology with an expert liver pathologist, the causality scores changed in 68% of cases, with an increased likelihood of DILI in 48% but also excluding DILI with more certainty in 20% of cases, suggesting liver histology helps clarify the diagnosis. However, in only a minority of cases (16%) did the causality score change in a clinically meaningful way by crossing from a likely DILI (1,2 and 3) to non–likely DILI (4,5).

The lack of standardized diagnostic criteria for DILI makes this diagnosis difficult, even among experts. In the DILIN prospective study, 18% of the enrolled cases of suspected DILI were attributed to a non-DILI cause of liver injury with a 6-month follow-up period (5). The conclusion that the liver injury was not due to a medication in 18% of cases was due to a number of factors, as the decision to enroll a patient in DILIN is made based on the available clinical data to exclude other causes of liver disease such as viral and autoimmune serology and imaging. The decision to perform a liver biopsy in patients enrolled in DILIN is at the discretion of the referring or principal investigator and is not required for enrollment. The reasons to perform a liver biopsy are variable and include to provide diagnostic information that makes a diagnosis more likely compared to prior cases; assessment of disease severity; failure of liver biochemical tests to improve after drug discontinuation; and differentiation from treatable conditions such as autoimmune hepatitis. In DILIN, the 50% of patients who underwent a liver biopsy had a longer latency and more severe injury although the final causality scores were similar to those who did not undergo liver biopsy (Table 1).

The role of the pathologist in DILI cases is to provide an interpretation of the histological findings in light of the clinical history. Since DILI remains largely a diagnosis of exclusion, the initial focus is on identification of other alternative etiologies of liver injury. To that avail, a diagnosis of acute viral hepatitis (A, B, C) can reliably be determined through blood testing while pancreaticobiliary and malignant parenchymal disease can usually be excluded via contrast enhanced imaging. In addition, medical history can help exclude a diagnosis of ischemic hepatitis and most instances of alcoholic liver disease but reliably identifying patients with sporadic autoimmune hepatitis, non-alcoholic steatohepatitis, and cholestasis frequently requires a liver biopsy. When the patient has a known underlying liver disease, allowances need to be made for that disease. A few drugs, such as amiodarone, have characteristic injury patterns, but most cases of DILI require more detailed evaluation and comparison with literature reports of biopsy findings for the suspect agents. Clues to the presence of DILI include unusual patterns of liver injury, such as significant eosinophil infiltration, granulomas, cholestatic hepatitis, necrosis, steatosis, vascular changes and bile duct loss. Despite the wide variety of injury seen in DILI, histologic pattern classification is possible with the DILIN reporting that 83% of 249 cases systematically reviewed could be classified into 5 distinct patterns- acute hepatitis, chronic hepatitis, acute cholestatic, chronic cholestatic and cholestatic-hepatitic (6). Certain histologic findings such as lack of eosinophils or granulomas, degree of necrosis, type of steatosis and cholangiolar cholestasis were associated with severe or fatal outcome. In addition, liver biopsy can be helpful in distinguishing DILI from autoimmune hepatitis (7). In the case example the combination of zonal necrosis and chronic cholestatic changes would be difficult to explain by a non-DILI cause or acute on chronic liver injury. However, some histological findings, like microvesicular steatosis or veno-occlusive disease, are more frequently due to DILI or toxic injury than other non-drug causes. The pathologist’s report should reflect on the evidence for and against DILI. Evidence against DILI might include features that can be explained by known co-morbidities, such as diabetes or chronic viral hepatitis. Evidence for DILI might include features not adequately explained by co-morbidities or by clinical testing already performed. Weighing the histological evidence against the clinical history and testing would permit an experienced pathologist to provide a likelihood that the injury was due to DILI.

The 50 patients included in this study were required to have an adequate liver biopsy (at least 10 portal tracts) performed within 2 months of onset of liver injury. Since the aim of the study was to determine the impact of liver histologic examination on causality assessment in patients with suspected DILI, biopsies performed early in the course of liver injury were felt to be more likely to be performed for diagnostic purposes rather than those performed later wherein ductopenia or progressive fibrosis may be of concern (8, 9). Despite this selection bias, these 50 patients were similar to the overall DILIN cohort in terms of age, pattern of liver injury, jaundice and suspected agent.

In most of the cases under review (N=31), hepatologists felt that a histologic assessment would be helpful which was confirmed after the histology was reviewed, with a change in the simulated causality score in 25 cases. However, even in the 19 cases where reviewers felt a biopsy was not necessary, 9 cases had a change in causality score, and in retrospect, the reviewers felt that the biopsy was helpful in 7 cases. This suggests that incremental objective data found in a liver biopsy can help confirm a clinical suspicion of DILI in the absence of a validated, diagnostic biomarker. In fact, initiation of corticosteroids was suggested in 6 cases after review of the biopsy and in 3 cases further serological or infectious studies were suggested. (Table 2 and Supplementary Tables 1 and 2).

Given the small number of cases where the simulated causality score after the liver biopsy changed in a clinically meaningful way in this study, it is unclear whether there are any factors that would indicate a liver biopsy should be considered as part of the work up for suspected DILI. The only significant finding was a higher peak alkaline phosphatase in the 8 cases where the score changed (Table 3). Amongst the 50 cases there were 28 hepatocellular, 13 mixed and 9 cholestatic cases that mirrors what is typically seen in DILIN. Of the 8 cases where there was a significant change in the causality score (table 4), 4 were hepatocellular and 4 were cholestatic. In addition, 2 cases had a change in score from probable to definite after the biopsy and both of these cases were mixed. Although the numbers are small it would suggest that biopsy was more helpful in cholestatic rather than mixed or hepatocellular DILI cases.

In addition, if the liver injury is associated with drugs that are well known to cause DILI such as amoxicillin-clavulanate, minocycline, nitrofurantoin and trimethoprim-sulfamethoxazole, a biopsy is likely not necessary. In all 8 cases where these drugs were the implicated agent and the reviewers felt a biopsy was not necessary, the biopsy proved to be unhelpful as the diagnosis was made on the clinical presentation. In the other 5 cases where these drugs were implicated and a biopsy was requested, there were other factors that influenced the decision, including severe hepatocellular injury in 2 cases and 2 cases where hepatitis C clouded the clinical presentation.

There are several limitations to this study. First, there is likely an element of selection bias, since we focused only on patients who had already undergone a liver biopsy for clinical purposes. In particular, we speculate that there was a reason for the responsible physician to perform liver biopsy- perhaps to assist with diagnosis or exclusion of other causes of liver injury. Alternatively, the degree of inflammation or fibrosis may have been of interest to the bedside clinician to help stage the patient’s disease process. However, almost half of all patients in the entire DILIN prospective cohort underwent a liver biopsy as part of the work up and causality scores for patients with and without a biopsy were very comparable (Table 1), which argues against substantial selection bias in patients selected for biopsy.

Secondly, the current study was performed under optimal circumstances; in particular, the ability to have expert hepatopathologist input is not possible for the general medical community. It is possible that expert hepatopathology insight provided clues to the diagnosis of DILI that may not otherwise be available to all providers. Although the numbers are small, biopsy findings particularly changed the causality in cholestatic suspected DILI cases suggesting these cases may benefit from expert hepatopathology review. Lastly, the retrospective nature of this study with an attempt to simulate clinical thinking is likely not the same as how a clinician may utilize hepatic histology in a real world setting.

In conclusion, expert hepatologists felt that the information provided by a liver biopsy was helpful in most suspected cases of DILI. Currently, excluding a diagnosis of viral hepatitis (A, B, C, E) is relatively straightforward, as is exclusion of benign and malignant pancreaticobiliary disease and hepatic malignancy via imaging. However, reliably differentiating acute DILI from other disorders such as autoimmune hepatitis, NASH, and alcoholic liver disease is more difficult and liver histology is likely to be helpful in such patients. In patients with liver injury from drugs well known to cause liver injury a biopsy is likely not necessary. Finally, liver histology appears to be most useful in cases where there is cholestatic liver injury and in equivocal cases of DILI (possible or probable), shifting assessment toward a greater certainty of a DILI or another diagnosis.

HIGHLIGHTS.

Value of Liver Biopsy in the Diagnosis of Drug-Induced Liver Injury: A Retrospective Simulation Study

• The value of liver biopsy in drug-induced liver injury (DILI) is unclear

• Review of 50 patients with early liver biopsy performed as work up for DILI

• Causality was performed before and after expert hepatopathology review

• Liver histology review changed the causality score in 68% of patients

• Biopsy clarified the diagnosis of DILI in cholestatic or equivocal cases of DILI

Abbreviations

ALT

Alanine aminotransferase

Alk P

Alkaline phosphatase

DILI

Drug induced liver injury

DILIN

Drug Induced Liver Injury Network

HDS

Herbal and dietary supplements

Data availability statement:

Research data including liver biopsy material were collected as part of participants standard clinical care and are privacy protected for patient confidentiality.