Introduction
The increase in consumption of drugs and dietary supplements – whose interactions with other xenobiotics has not been fully studied – leads to increased drug-induced liver injury (DILI). Adverse reactions related to drug-drug interactions (DDI) are becoming increasingly important, including those between xenobiotics and “natural medicine” substances. In patients’ opinion, herbal drugs cannot be harmful, and they do not inform doctors about their consumption. The clinical picture of adverse reactions is polymorphic. Some cases include liver damage of variable intensity: from mild, with rapid recovery after discontinuation of the drug, to life-threatening acute liver failure [1]. Many cases fall in the middle of this spectrum and are characterized by long-term jaundice with skin itching and significant weakness lasting for weeks. Most patients totally recover after discontinuing the drug causing the adverse effects, but some of them do not recover for more than 6-12 months [2, 3]. In some cases, the exposure to the xenobiotic may trigger a disease process in the mechanism of autoaggression, which will continue despite discontinuing the drug [2]. Sometimes liver damage leads to the development of liver cirrhosis and hepatocellular carcinoma after many years [4].
Chronic sequelae of DILI
One of the first comprehensive studies was “Drug-induced liver disease” (1978) by Hyman J. Zimmerman, who had been investigating the problem of toxic liver damage since the late 1960s [5]. The currently recommended classification of drug-induced liver damage published in the EASL Clinical Practice Guidelines draws significantly on these experiences [1]. Probably the first study of the natural history of DILI was performed by Aithal and Day, based on a histological database survey [6]. A total of 13 of 33 (39%) patients followed for an average of 5 years had significant persistent abnormalities in liver function tests and/or liver imaging. Most patients developed jaundice and 3 patients developed liver failure. The inclusion of patients with clinically significant DILI who underwent liver biopsy may have biased the study results. Subsequent studies have not found a similarly high percentage of patients with chronic liver injury following DILI, which was found in only 3.4-18% of the study participants, and the follow-up period was highly variable. It is well documented that recovery from cholestatic liver injury is longer than that from cytolytic hepatocyte injury [7].
Drug-induced autoimmune reactions in the liver
In the course of DILI, cases are observed with the presence of nonspecific autoantibodies (anti-nuclear antibodies – ANA, anti-smooth muscle antibodies – ASMA) and elevated immunoglobulin (Ig)G levels. Laboratory test results resemble classic autoimmune hepatitis (AIH), and the term drug-induced AIH (DIAIH) has been adopted for these cases. The immune response is targeted at the neoantigens, formed from the hepatic drug metabolites, binding to cytochrome P450 [8].
Among drugs, the association of hydralazine, minocycline, nitrofurantoin, and methyldopa with DIAIH is confirmed [9-11]. A group of American researchers associated with the Drug-Induced Liver Injury Network (DILIN) undertook studies aimed at determining the main autoimmune features of DILI caused by hydralazine, minocycline, nitrofurantoin, and methyldopa. In the study by de Boer et al., approximately 90% of DIAIH cases occurred in women, and in 74% of them signs of hepatocyte damage were observed. In the majority of patients with DILI induced by nitrofurantoin and minocycline, and in approximately 50% of people with liver damage after methyldopa and hydralazine, signs of autoimmunity resembling AIH were observed [10].
Interestingly, these markers of the autoimmune process have been found to decrease with regression of liver damage, and the patients do not carry the most common HLA alleles seen in patients with true AIH [10]. In a Mayo Clinic study characterizing DIAIH (primarily caused by nitrofurantoin and minocycline), among patients diagnosed with AIH, the clinical, biochemical, and histological features were very similar between DIAIH and true AIH. The only significant difference between them was the lack of relapse after discontinuation of corticosteroids in the DIAIH group compared with a relapse rate of 65% in the true AIH group [11].
Some authors suggest a relationship between the risk of DILI and predisposition to the development of autoimmune diseases. It was observed that in patients diagnosed with DILI with the presence of ANA autoantibodies, other concomitant autoimmune diseases were more frequently diagnosed, and in addition, women predominated in this group [12]. During long-term follow-up, 22% of patients with DILI developed AIH.
Autoimmune reactions during cancer immunotherapy
Liver damage associated with cancer immunotherapy clinically resembles autoimmune hepatitis. In contrast to DIAIH, it is not a consequence of an immune reaction against neoantigens generated as a result of introducing a xenobiotic or its metabolites into the body [1, 13]. Immunotherapeutic agents that act as immune checkpoint inhibitors (ICIs) enhance the T-cell response and restore a strong anti-tumor immune response that was suppressed by the growing neoplasm. Immune checkpoints are surface molecules found on both immune and tumor cells. They include cytotoxic T-lymphocyte associated protein 4 (CTLA-4, the target of ipilimumab), programmed death receptor 1 (PD-1, the target of pembrolizumab and nivolumab), and programmed death-ligand 1 (PD-L1, the target of atezolizumab, avelumab, and durvalumab), which are involved in downregulating the immune response [14].
Immune checkpoint inhibitors have changed the paradigm of cancer therapy from direct drug-mediated destruction of tumor cells to augmenting the host immune response against tumors. The histological pattern of the liver injury associated with immunotherapy is different for anti-CTLA-4 and anti-PD-1/PD-L1 agents. Anti-CTLA-4-induced hepatotoxicity showed a specific pattern of granulomatous hepatitis associated with severe necrotic and inflammatory activity of the lobule, fibrin deposits, and central vein endotheliitis. The histological picture of patients receiving anti-PD-1/PD-L1 drugs alone was more heterogeneous and was characterized by active hepatitis with focal or confluent necrosis and mild to moderate periportal inflammatory activity that was not associated with granulomatous reaction [14, 15]. Interestingly, in contrast to idiopathic AIH, ICI-related hepatitis is usually “seronegative” and resolves after immunosuppressive therapy without recurrence, after discontinuation of ICIs [13, 15].
Drug-induced sclerosing cholangitis
The first reports appeared in the 1980s and concerned patients treated for liver metastases and intraarterial infusion of cytotoxic agents such as floxuridine suggested rather local toxic drug effects on the bile ducts [16]. Single case reports also concerned systemic treatment with other drugs. Due to cholestatic damage during docetaxel chemotherapy, imaging studies were performed, which revealed diffuse intrahepatic dilatation of the bile ducts with multifocal stenosis and dilatation of the intrahepatic bile ducts. Docetaxel was discontinued, and liver biopsy showed subacute obstructive changes in the bile ducts, eosinophilic infiltrates, tubular cholestasis and biliary stenosis. Five months later, imaging studies still indicated secondary sclerosing cholangitis [17].
Several reports have shown that ketamine can also cause sclerosing cholangitis with diffuse biliary strictures, both intrahepatic and extrahepatic, which have been reported mainly in association with ketamine abuse [18]. Recently, sclerosing cholangitis induced by ketamine taken as a sedative in the intensive care unit was reported in a patient with COVID-19 [19]. Another study observed the development of multiple intrahepatic biliary strictures and hilar strictures in a 40-year-old woman shortly after surgery for a benign esophageal tumor. This was considered secondary to the sevoflurane or ceftriaxone used in therapy. Follow-up imaging studies after 3 months showed abnormal cholangiograms despite normal liver function tests [20].
Of the 25 patients with DILI who underwent magnetic resonance cholangiopancreatography (MRCP) as part of the liver injury diagnostics, 10/25 (40%) had cholangiographic changes [20]. Histological examination of liver biopsy specimens, available in 4 patients in the analysis by Gudnason et al., showed mainly tubular cholestasis and inflammatory infiltration in the portal spaces with features of acute and/or chronic hepatitis. All DILI patients who developed secondary PSC had cholestatic liver damage, significant hyperbilirubinemia, and a chronic disease course compared with patients who did not have biliary changes [20]. A prospective study conducted by DILIN showed the presence of PSC-like changes in 7% of DILI patients who underwent MRCP. However, the percentage of patients with PSC may have been underestimated, because only a small group of patients underwent MRCP testing during DILI diagnostics [21].
Liver damage during ICI therapy usually manifests as inflammatory changes in liver biopsies, but there have been recent reports of cases similar to PSC [22]. In almost 80% of cases, bile duct dilatation is observed without biliary obstruction [23]. In addition, multiple intrahepatic strictures were observed in 30% of patients, and in the vast majority of cases (> 90%), diffuse hypertrophy of the wall in the extrahepatic bile ducts was observed. The long-term consequences of secondary sclerosing cholangitis on the progression of the disease to cirrhosis or the development of cholangiocellular carcinoma (CCC) have not yet been determined [23].
Vanishing bile duct syndrome
In cholestatic liver damage, abnormal liver function tests are observed for a longer period than in mixed or hepatocytic liver damage [13]. Most patients recover within a year of disease onset, but in some the recovery may be very slow. This is particularly true in cases of severe destruction, loss, and atrophy of the bile ducts, which is called vanishing bile duct syndrome (VBDS). A definition of ductopenia has been proposed as the loss of intralobular bile ducts in less than 50% of portal spaces [24]. VBDS has been reported mainly in patients with jaundice and cholestasis persisting for months or years. In some of these patients, secondary biliary cirrhosis developed, which required liver transplantation [24]. However, the reversibility of VBDS has also been described [25].
In a study conducted by the DILIN group in the USA, Bonkovsky et al. bile duct atrophy in a group of patients with DILI. During the first 10 years of observation, approximately 34% of 1050 patients underwent liver biopsy; of these, 7% (n = 26) had bile duct atrophy. In total, 14/26 (54%) had severe ductopenia (< 50% of portal spaces with detectable bile ducts) meeting the VBDS criteria, while the remaining patients had a mild form. Jaundice was present in 96% of patients with VBDS, but only in 70% of the entire DILIN group. Pruritus was reported in almost 80% of patients with VBDS vs. 54% of the remaining group of patients with DILI. The level of aminotransferase activity was similar in both groups, and the most common histological picture of liver biopsy samples showed a relatively moderate inflammatory process. The only prognostic factor for an unfavorable course of the disease was the severity of bile duct atrophy in the examined liver biopsies [26].
Overall, the prognosis in patients with observed bile duct atrophy was unfavorable, and mortality related to hepatopathy in the VBDS group was 19% vs. 6.2% in the entire group of 899 patients with DILI [25]. Overall, 2/26 (8%) required liver transplantation compared with only 4% in the entire group of patients analyzed by DILIN. Elevated liver function test values were found in 94% of patients with ductopenia after 6 months of follow-up compared with 47% of patients with other histological patterns of liver damage.
DILI-associated liver cirrhosis
Liver fibrosis and development of liver cirrhosis are well-documented adverse effects of chronic methotrexate therapy [13, 27]. In the past, regular liver biopsies were performed in these patients during chronic methotrexate therapy, but it is not clear in what percentage of patients severe fibrosis or cirrhosis may develop. The range of study results was wide (0-26%), most likely due to sampling error resulting from patient selection for biopsy [28]. Decompensated liver cirrhosis as a consequence of methotrexate use has been reported rarely. Of the approximately 159,000 patients reported for liver transplantation in the USA between 1987 and 2011, only 117 (0.07%) developed methotrexate-induced liver damage. Compared with patients with other causes of end-stage liver disease, patients with methotrexate-induced liver damage are at significant risk of developing metabolic syndrome, as is the case in patients with non-alcoholic steatohepatitis (NASH) [28].
Liver cirrhosis has been associated with numerous drugs that cause DILI, probably with the exception of methotrexate. The best documented observations are those with nitrofurantoin and amiodarone [29, 30]. The clinical, biochemical, and histological phenotype of nitrofurantoin-induced DILI is highly variable and includes acute idiosyncratic DILI, DIAIH, and liver cirrhosis [31]. In patients with nitrofurantoin and minocycline related DIAIH, advanced liver fibrosis was detected only in nitrofurantoin-treated patients [31]. Most patients who developed liver cirrhosis due to amiodarone use had been treated for several years. In these patients, cases of decompensated liver cirrhosis that did not resolve after discontinuation of this drug were reported [32].
In a prospective study of the Spanish DILI registry of chronic liver diseases, 7 of 16 (63%) patients who underwent liver biopsy had liver cirrhosis. The drugs responsible for the development of liver cirrhosis identified in this study were very diverse, e.g. atorvastatin, bentazepam, ebrotidine, clopidogrel/atorvastatin, amoxicillin with clavulanic acid/ibuprofen and ranitidine, and the liver damage had an initially very variable clinical picture [7, 12].
DILI-associated vascular lesions
Toxic reactions after xenobiotics may also affect the endothelial cells of the hepatic veins and sinusoids. Drug-induced vascular damage can range from sinusoids and small venous vessels to large veins of the portal system – sometimes all lesions together in the same patient. These include sinusoidal obstruction syndrome (SOS), nodular regenerative hyperplasia (NRH), and peliosis hepatis [33]. Sinusoidal damage is sometimes difficult to recognize in liver biopsies. It is most often due to secondary hemorrhage due to vascular damage, which can affect the function of adjacent hepatocytes and lead to their necrosis, masking the cause of the disease [1].
Hepatic sinusoidal obstruction syndrome (SOS) encompasses a variety of toxic vascular lesions of the liver. They most often result in irreversible chronic liver disease, portal hypertension with its complications [34]. This syndrome was described a hundred years ago in association with the ingestion of herbs containing pyrrolizidine alkaloids (e.g. Crotalaria, Heliotropium, Senecio, and Symphytum). This equates to an estimated 8000+ cases per year, making herbal hepatotoxicity one of the most common causes of SOS [35]. SOS has also been reported as a complication of myeloablative regimens containing cyclophosphamide. A number of other medications have been associated with the development of SOS, including dacarbazine, gemtuzumab, and long-term use of thiopurines such as azathioprine, mercaptopurine, and thioguanine [36].
Currently, oxaliplatin-based chemotherapy – used as adjuvant therapy in patients with advanced colorectal cancer and liver metastases – is the main cause of SOS, NRH, and hepatic peliosis. SOS associated with oxaliplatin was first described in 2004 in the histopathological analysis of 87 liver resection specimens from patients who had undergone chemotherapy. It showed that 44 (51%) had sinusoidal dilatation and hemorrhage into the sinusoidal space. Sinusoidal damage associated with oxaliplatin occurs in most cases without significant increases in serum enzyme activity or clinically visible liver damage [34, 37].
One of the effects of chronic damage to the liver vessels may be NRH. As a consequence, extensive vascular changes leads to the formation of diffuse nodules in the liver parenchyma. The hepatocytes in the nodule are arranged in plates greater than 1 cell thick, whereas the hepatocytes between the nodules are compressed and atrophied into thin, parallel plates [38]. Characteristically, the nodules are not separated by fibrosis, although perisinusoidal fibrosis and incomplete fibrotic bands may be present. Magnetic resonance imaging can demonstrate these characteristic radiological patterns with a sensitivity and specificity of 75-80%. The cumulative rate of development of nodular regenerative hyperplasia in patients treated with azathioprine is estimated to be 0.5% over 5 years and 1.5% over 10 years. Early diagnosis and discontinuation of the drug have been shown to lead to resolution of the histological changes within 5 years [39].
Nodular regenerative hyperplasia, as indicated by liver histology, was reported in 8% (8/97) of HIV-infected patients receiving antiretroviral therapy (ART). Another case series showed that all HIV-infected patients with non-cirrhotic portal hypertension who were exposed to long-term didanosine developed NRH [40]. Other drugs associated with this form of liver disease include 6-thioguanine, busulfan, bleomycin, cyclophosphamide, chlorambucil, cytosine arabinoside, carmustine, and doxorubicin. In recent literature, oxaliplatin remains the most commonly used drug associated with this pathology [37].
DILI-associated liver tumors
The annual incidence of hepatocellular adenoma (HA) among regular oral contraceptive users is 3-4 per 100,000, compared with an estimated incidence of 3 per million per year in the general population [41]. Hormone dose and duration of treatment have been associated with the risk of developing HA, and the risk was highest in women over 30 years of age who had used oral contraceptives for more than 24 months. However, the reported risk of hepatocellular adenoma applies to combined oral contraceptives and may be lower with newer progesterone-only drugs. A causal relationship between oral contraceptives and liver cancer has been accepted because of several reports of regression or resolution of HA after discontinuation of these drugs. Estrogen receptors have also been detected in a significant proportion of HA cells [41].
However, there are also reports of HA progression to hepatocellular carcinoma (HCC) 3-5 years after detection of a hepatic lesion, despite discontinuation of oral contraceptives. The morphology of HA with its extensive proliferation of blood-filled sinusoids supplied by high-pressure arterial flow causes 20-40% of them to bleed spontaneously, with resultant pain in the right upper quadrant. It is estimated that HA progression to HCC occurs in about 10% of cases [42]. The ultrasonographic features of HA are nonspecific, and multiphase computed tomography or magnetic resonance imaging allows them to be distinguished from hemangiomas, focal nodular hyperplasia (FNH), or HCC in the vast majority of patients.
For the first time, the association of primary liver tumors with androgen therapy in patients with Fanconi anemia taking androgenic anabolic steroids has been described. However, HA, HCC, and other tumors (CCC and angiosarcoma) occur not only in people taking androgens due to Fanconi anemia or other forms of aplastic anemia, but also in patients treated for other indications (hereditary angioedema and immune thrombocytopenia), as well as in those taking these drugs recreationally – bodybuilders. In a large series of 133 cases, HCC was associated with oxymetholone and methyltestosterone, and HA with danazol [43].
Monitoring drug hepatotoxicity
Biochemical monitoring of drug hepatotoxicity is not commonly used and should be limited to patients treated with high risk of hepatic drug toxicity. In the case of drugs such as isoniazid and statins, a transient and self-limiting increase in aminotransferase activity is observed. Their normalization during further use of the drug is caused by so-called metabolic adaptation [4, 44].
In some patients, the increase in aminotransferases is accompanied by subjective symptoms resulting in discontinuation of treatment. In the event of the need to continue therapy, if there is no alternative, such people must be closely monitored. Regular liver enzyme tests every 3-6 months and abdominal ultrasound are recommended.
The majority of reports concern hepatotoxicity of methotrexate, a drug used for many years in dermatology and rheumatology. According to the guidelines of rheumatology societies, aminotransferase activity should be assessed before starting treatment and then repeated every 2-4 weeks during the first three months of therapy, then every 8-12 weeks for the next 3-6 months, and every 12 weeks after 6 months [13]. In these patients, we assess pretreatment liver fibrosis using non-invasive tests (FIB-4, FibroTest, elastography) and then every year if the initial result is normal (or minimal fibrosis) and the patient does not have baseline hepatopathy. In the case of moderate initial fibrosis, elastography is recommended every 12 months if there are no concomitant liver diseases. In patients with risk factors, this test should be performed annually together with other non-invasive tests. Elastography and/or liver biopsy are recommended in all patients after exceeding a cumulative dose of methotrexate 3.5-4 g. Moreover, liver biopsy should be considered in patients with abnormal results of elastography and liver function tests [6, 13, 44].
Among the drugs requiring especially careful monitoring with liver function tests are ICIs, a group of drugs with a varied profile of liver toxicity. The clinical picture of hepatotoxicity is variable, ranging from transient increases in aminotransferase activity through hepatitis to liver failure. The latency time is from 6 to 14 weeks after the start of treatment (median time 52 days; median doses – 3 cycles of immunotherapy), but it can also occur after longer periods of treatment, sometimes several months after drug discontinuation [15]. Monitoring of immune-mediated hepatitis (IMH) involves clinical evaluation and laboratory tests before each drug administration. In patients with grade 1 liver damage, therapy can be continued with more frequent laboratory monitoring. In patients with grade 2 liver damage, ICI administration should be temporarily suspended, and if the liver enzymes/bilirubin remain elevated for 3-5 days after drug discontinuation, corticosteroids should be considered. Laboratory monitoring should be performed every 3 days. Permanent discontinuation of ICIs and treatment with high-dose steroids is recommended for grade 3 or 4 hepatotoxicity until enzyme elevation improves to grade 1/grade 2 (Table 1).
Table 1
Monitoring and treatment of immune-mediated hepatitis
Generally, corticosteroid dosing and route of administration (oral or intravenous) are based on grade of liver injury, with suggested strategies for dose escalation/tapering according to patient response. A slow taper over several weeks is suggested because of the relatively long half-lives of ICIs. The choice of second-line immune suppressive therapy in patients who are resistant (an initial response to steroids, but with a rebound flare in liver enzymes during steroid taper) or refractory (failure of liver enzymes to improve after 48-72 hours of steroid therapy) to corticosteroids is not established. Mycophenolate mofetil (MMF) is frequently used, although successful outcomes have been reported with other treatments including tacrolimus, azathioprine, cyclosporine, and infliximab [13, 44-46].
The significant latency of symptoms, which can appear many months after discontinuation of ICI therapy, requires monitoring of these patients for the development of delayed adverse effects with consequences similar to untreated AIH [5, 15].
Tuberculostatic therapy requires close monitoring of aminotransferase activity due to the risk of liver damage. Among these drugs, isoniazid (INH) used for the prophylaxis (isoniazid preventative therapy – IPT) and treatment of latent (LTBI) or active tuberculosis (TB) is considered the most hepatotoxic. In a recently published meta-analysis study, the frequency of INH-induced liver injury (INH-ILI), defined as ALT > 5 times above the upper limit of normal (ULN), in patients with LTBI receiving IPT was low (2.6%). No significant differences were found in the frequency of INH-ILI in patients older or younger than 50 years, children, patients living with HIV, candidates for liver, kidney, or lung transplant, treatment duration (6, 9, or 12 months), type of study design, or study provenance [47]. The exact frequency of INH-ILI is difficult to determine in patients with active TB when combinations of antituberculosis drugs are used and other drugs such as rifampicin and pyrazinamide are also potentially hepatotoxic. The incidence of severe INH-ILI appears to be lower than previously appreciated [48]. The recommendations for laboratory monitoring while receiving INH have been changed over the last decades. It seems that monthly liver test monitoring is reserved for patients with baseline liver test abnormalities due to many causes or HIV-infected patients on antiretroviral therapy. Periodic liver tests can also be performed in those older than 35 years [49].
Conclusions
The persistence of abnormal liver function tests after acute DILI is highly variable and may be due to different study eligibility criteria. In addition, it may be influenced by the heterogeneity of the DILI patient population or differences in the chemical composition of the preparations. In most cases, DILI has an acute clinical course that fully remits after discontinuation of the toxic preparation. The natural history of most chronic phenotypes after acute DILI has not been well characterized, and further studies are needed to examine the natural history of patients with persistent abnormal liver function tests lasting longer than 6 or 12 months.
