Signatures in drug induced liver injury

Abstract

Purpose of review

Drug induced liver injury (DILI) can be induced by a myriad of drugs. Assessing whether the patient has DILI and assessing which drug is the most likely culprit are challenging. There has been too little attention paid to the concept that certain drugs appear to have unique clinical features or “phenotypes”.

Recent findings

Several case series of DILI due to various drugs have been published, and analysis of these case series points to the fact that individual drugs have characteristic DILI signatures. These clinical phenotypes can be characterized by latency, biochemical features (R-value), as well as clinical symptoms and signs. Several drugs including isoniazid, amoxicillin-clavulanic acid, anabolic steroids, beta-interferon, and others have highly unique clinical features. Such unique properties may be able to be used to improve adjudication processes.

Summary

Individual drugs have unique clinical DILI phenotypes or signatures. Furthermore, these may be able to be used to improve adjudication.

Introduction

Drug-induced liver injury (DILI) remains a relatively uncommon cause of liver disease, but its diagnosis remains challenging, as the diagnosis of DILI is largely one of exclusion. One of the more important aspects of making a diagnosis of DILI is the consideration of the likelihood that the implicated drug is responsible. For example, DILI appears to be relatively more likely to occur with INH, a drug with a high propensity to cause DILI, than with a statin, a drug with a lower probability to cause DILI. But reliable incidence data are unavailable as in most reported DILI series, the number of people at risk (i.e., taking the drug) is largely unknown.

The likelihood of a liver injury being liver related is currently assessed by causality tools such as RUCAM [1]. Such models take into consideration a number of variables, such as the drug latency (from start of the drug to onset of liver test abnormalities), biochemical pattern of injury caused by the drug (i.e. a cholestatic or mixed versus hepatocellular injury), and so on. [1] Standardized Expert Opinion Processes (SEOP), such as used in the North American Drug Induced Liver Injury Network (DILIN) are considered superior to models such as RUCAM.[2] One reason that such SEOP is likely superior it the fact that expert can take a drug’s DILI phenotype derived from previous reports into account when adjudicating cases.

It is being more and more recognized that specific drugs cause DILI in specific patterns [3–4]– these patterns include clinical features such as latency, biochemical patterns of liver test abnormality, the presence of autoimmune features, and even age and gender specificity. Here, we report on a number of these patterns, and further point to the implications that recognition of these patterns has for adjudication. In a case of liver injury occurring while on Bortezomib such approach was used to different Bortezomib DILI from Bortezomib associated HBV reactivation.[5]

Drug specific DILI phenotypes

The phenotype or clinical pattern of injury after DILI appears to vary for a wide array of different individual drugs. The variables important in this clinical variation include the following: latency (the number of days from drug start to presentation with abnormal liver tests), dechallenge (the number of days from the peak of abnormal liver tests to a value less than 50% of peak), R-value (the ratio between ALT in ULN/AlkPhos in ULN), de Ritis ratio (ratio between AST/ALT), extrahepatic manifestation such as rash, detection of autoantibodies, gender, and others. As highlighted below, genetic factors may also help define a clinical phenotype.

A review of recent case series that evaluated specific clinical features (gender, latency, dechallange, biochemical profile (R-value), de Ritis ratio, and other clinical features) reveal a number of notable clinical features.

Gender distribution in DILI

Although data are limited, it appears that certain drugs may be more likely to cause DILI in one gender or another (Figure 1a)[6–25]

Figure 1. Specific clinical features in DILI.

A.) Gender. The relative frequency of women (blue) and men (orange) reported to have DILI due to specific drugs in published case series is shown. The graph highlights drugs in which at least 10 cases were reported (for some drugs, several series have been combined). B.) Biochemical (R value) characteristics. The relative frequency of biochemical injury patterns in published case series is shown: hepatocellular injury in blue, mixed in orange, and cholestatic yellow; gray depicts mixed/cholestatic injury when the series did not separate these patterns. The graph highlights drugs in which at least 10 cases were reported (for some drugs, several series have been combined).

One of the more remarkable gender specific clinical phenotypes is with beta interferon: in a series of patients with DILI due to this agent, 90% of patients were women - and in the 2 cases reported in men, whether beta interferon truly caused DILI was questionable.[26] Several other drugs, such as celecoxib, asparginase, and OxyelitePro^® also seem to be more prevalent in women than in men. In contrast, DILI due to anabolic steroids has only been reported in men.[17] It should be noted that there is likely inherent bias involved in some of these assessments, since for example, it is unlikely that women take anabolic steroids than men. In contrast, it is possible that women are more likely to take weight loss supplements such as OxyelitePro^® than men.

Biochemical (R-value) phenotype

The R-value is the ratio of ALT in ULN to ALP to ULN. This ratio is used to characterize the biochemical characteristics of the liver injury pattern; the pattern is considered “hepatocellular” when the R-value is >5, “mixed” when the R-value is between 2 and 5, and “cholestatic” when the R-value is <2. Current causality tools recommend different adjudication pathways for hepatocellular injury and mixed/cholestatic injury patterns.[1]

Although few series report actual R-value with median and interquartile ranges, the injury pattern is frequently reported and indicates clear differences between drugs (Figure 1b). One of the most notable patterns is with INH – which almost always causes a hepatocellular pattern and has yet to be reported as a cholestatic pattern. In contrast, several drugs have an almost pure cholestatic injury pattern (ie Bupivacaine [27,28] and Celecoxib[14]). Notably, the drug most frequently reported to cause DILI in Western registries [29], amoxicillin/clavulanic acid – rarely causes a hepatocellular injury pattern.

These data clearly demonstrate some degree of specificity in biochemical pattern for different drugs. This suggests that the R-value may be a valuable tool useful in adjudication of DILI causality, as there are differences in the characteristic R-values found among different drugs.

Latency

Latency is another important clinical variable that helps distinguish different drug clinical phenotype. Some drugs have very short latencies (ie. Bortezomib with <2 weeks [5]), while others, such as minocycline or nitrofurantoin, may take over a year ([15], Table 1a). It should be noted that in studies that have reported both biochemical features (hepatocellular and mixed/cholestatic) as well as latency, latency did not different between significantly as a function of whether the biochemical signature was cholestatic and hepatocellular (Table 1b).

Table 1a.

Latencies^* for different drugs

Drug	Latency	Reference
Amoxicillin/clavulanic acid	29 [IQR17–37]	De Lemos et al. [8]
Isoniazid	mean 103.4 (SD 59)	Hayashi et al. [12]
Statins	90 [30–180]	Björnsson et al. [14]
Azithromycin	19 [12.25–27] (6–29)	Martinez et al. [38]
Cephazolin	20 [18–26] (6–29)	Algahtani et al. [39]
QUINOLONES	5.5 [1–16.5] (1–41)	Orman ES, et al [40]
Cyproterone	150 [114.25–240.5] (33–4250	Bessone F, et al. [41]
Ceftriaxine	8 [4–15] (4–47)	Nakaharai et al.[42]
Methylprednisolon	28 [21–41] (3 to 60)	Bresteau et al. [43]
Celecoxib	15 (2–730) / 12 (1 to 42, plus 3: 152–730)	Mukthinuthalapati et al[16]
Minocycline	84 (mean 346, SD ±268)	Urban et al. [44]
Anabolic steroids	73	Stolz et al. [17]

^*Latency is measured in days from drug start to presentation with abnormal liver tests.

Data are given as median (except for INH whehre the mean was used in absence of a reported median), [] indicate interquartile range (IQR), “()” indicate range or standard deviation (SD)

Table 1b:

Latencies^* for different drugs

	Hepatocellular median [IQR] (range)	Mixed/Cholestatic median [IQR] (range)	Reference
Amoxicillin-Clavulanic acid	N=25 15 (9–21)	N=23 / 21 mixed: 21 (13–30) chol 14 (4–19)	Lucena et al. [7]
Statins	n=15 186 [84–538] (45–1866)	n=7 156 [66 to 774] (51–3600)	Russo et al. [15]
Azithromycin	n=10 19 [14.5–25.25] (13 to 65)	n=8 17 [7.75–30.75] (2 to 38)	Martinez et al. [31]
Quinolones	n=4 12.5 [4.75 to 27] (4–30)	n=8 1.5 [1 – 11] (1–39)	Orman ES, et al.[33]

^*Latency is measured in days from drug start to presentation with abnormal liver tests.

Square brackets [] indicate interquartile range (IQR),

Round brackets ()indicate range or standard deviation (SD)

Specific examples of drug specific phenotypes

Based on the available data, it is clear that there are drug specific phenotypes for different drugs (Table 2). For example, INH causes mostly a pure hepatocellular pattern, and has a more prolonged latency that drugs such as amoxacillin/clavulanic acid. In contrast, nitrofurantoin, and beta-interferon occurs mostly after 12 weeks of exposure and seems limited to females, as does methylprednisolone, which occurs around the first months of exposure. Anabolic steroids cause a remarkable clinical syndrome, termed “bland cholestasis” in which the bilirubin is elevated, typically far out of proportion to the elevation in aminotransferases.[17]

Table 2.

Specific drug phenotypes^*

Drug	Gender ratio F/M	Biochemical Pattern	Median Latency (d)	Comments
Anabolic Steroids	0 / 44	HC>M=Chol	73	Bilirubin is typically elevated far out of proportion to aminotransferases
INH	1 / 2	HC>>>M	>60 days	Rarely if ever associated with a biochemical pattern that is not HC
A/C	1 / 1.5	HC<M<Chol	29	Heterogeneous presentations
Statins	1 / 1	HC>M/Chol	90	Heterogeneous presentations
Minocycline	4 / 1	HC>>Chol	84	May mimic AIH
Clozapine	1 / 1	HC	34	Rarely if ever associated with a biochemical pattern that is not HC
Celecoxib	2 / 1	HC=M=Chol	15	Heterogeneous presentations
OxyElitePro	3 / 1	HC	28	The active injurious ingredient is believed to be green tea extract
b-Interferon	8 / 1	HC	102	Prominent female predominance
Methylprednisolone	28 / 1	HC	28	Prominent female predominance
Nitrofurantoin	42 / 0	HC>>M>>C	>84	May have very long latency, may mimic AIH, female predominance

^*Data have been gathered from the published literature – see text.

Abbreviations: A/C - amoxicillin-clavulanic acid; AIH-– autoimmune hepatitis; Chol - cholestatic injury pattern; F – female; HC - hepatocellular injury pattern; INH - isoniazid; M – male; M - mixed injury pattern

Genetic risk genes in DILI

The understanding of a drug’s “genetic risk profile” and the propensity to cause DILI is evolving. [30–32] Available data suggest that certain risk genes predispose to DILI associated with specific drugs. Therefore, in the future, it may be possible to use genetic information to help adjudicate causality in cases of suspected DILI.[8]

Genetics can be helpful in ascertaining the diagnosis of a variety of diseases, such as autoimmune hepatitis hemochromatosis or suspected Wilson’s disease.[33–35] Similarly, genetics may be helpful in causality adjudication for DILI. Several drugs have been demonstrated to have risk alleles with very high odds ratios for DILI (up to 80 for HLA-B57:01 which was found in 43/51 and 20/23 flucloxacillin DILI cases versus 4/60 in controls, respectively.[36] Furthermore, in patients who do not have HLA-B57:01, HLA-B57:03 was shown to be an additional risk allele.[37].

Although flucloxacillin has an impressive genetic risk association, significant, albeit impressive, genetic risk profiles (Table 3). However, most risk alleles identified for association with DILI have lower odds ratio and less dramatic effects, or are just infrequent in DILI cases despite high odds ratio (Table 2).

Table 3.

Risk alleles in DILI

	Risk Allele	present in DILI cases	present in controls	Odds ratio	RR	Reference
Flucloxacillin	B57:01	43/51 & 20/23	4/64	80.6 100	7.8 17.5	[43]
Amoxicillin - clavulanic acid	DRB1*1501	20/35	9/60	7.6	3.0	[45]
Amoxicillin - clavulanic acid	DRB1*1501	14/20	27/134	9.27.6	6.4	[46]
Minocycline	HLA-B35:02	4/25	0.6%	29.6	27	[47]
Ticlopidine	HLA-A*33:01	4/5		163.1		[48]
Methyldopa	HLA-A*33:01	2/4		97.8		[48]
Fenofibrate	HLA-A*33:01	3 / 4		58.7		[48]
Terbinafine	HLA-A*33:01	6/14		40.5		[48]
Enalapril	HLA-A*33:01	2/4		34.8		[48]
Sertraline	HLA-A*33:01	2/5		29		[48]
Erythromycin	HLA-A*33:01	2/10		10.2		[48]

Conclusion

The data suggest that specific drugs are associated with specific clinical characteristics when they cause DILI. Some of the more remarkable clinical features can be seen in biochemical and latency signatures. Gender may also be important - all cases of DILI reported to date from anabolic steroids have been limited to men, while interferon-beta associated severe DILI has been limited to women, and DILI due to nitrofurantoin and methylprednisolone occurs almost universally in women. Importantly, these data suggest that it may be possible to generate rigorous and specific clinical phenotypes for different drugs.

Bullet Points:

• Different drugs have different clinical feature, also known as “signatures”.

• Differences in signatures are most prominent for gender predilection, latency, and biochemical features are most prominent.

• Genetic risk alleles associated with DILI are evolving and it is possible that in the future, these may be able to be utilized to assist in DILI diagnosis.

• Unique signatures may be useful to improve causality assessment approaches.