Research Article DILI, Autoimmune, Cholestatic and Genetic Diseases Comprehensive analysis and insights gained from long- term experience of the Spanish DILI Registry Graphical abstract Authors Camilla Stephens, Mercedes Robles- Diaz, Inmaculada Medina-Caliz, ., Neil Kaplowitz, M. Isabel Lucena, Younger age Fatal/liver transplant outcome Hepatocellular injury Female Raúl J. Andrade Drug-induced liver injury Hepatocellular injury (nR) DILI Platelet count AST values Total bilirubin values Correspondence Liver-related mortality lucena@uma.es (M.I. Lucena). + Hepatocellular injury Underlying hepatic disease Lay summary DIAIH vs. DILI DILI therapy (N = 128; 20%) Clinical information on drug- Female 52% corticosteroids 39% ursodeoxycholic acid induced liver injury (DILI) Hepatocellular injury ALT and AST values 9.4% MARS In patients with collected from enrolled patients in Duration of causative drug Hy’s Law prognostic performance varies with individual drugs Bilirubin values the Spanish DILI Registry can guide Need for immunosupresants Fatal/liver transplantation physicians in the decision-making process. We have found that older patients with hepatocellular type liver injury and patients with additional liver conditions are at a Highlights higher risk of mortality. The type of  Clinical parameters can help predict DILI phenotype and outcome. liver injury, patient sex and  Older patients with cytolitic DILI and those with liver disease have analytical values of aspartate aminotransferase and total bili- worse outcome. rubin can also help predict clinical  Serum AST at DILI onset should be assessed as it strongly predicts outcomes. poor outcome.  Prognostic potential of Hy’s law in DILI varies between causative agents. https://doi.org/10.1016/j.jhep.2021.01.029 © 2021 European Association for the Study of the Liver. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). J. Hepatol. 2021, 75, 86–97 Research Article DILI, Autoimmune, Cholestatic and Genetic Diseases Comprehensive analysis and insights gained from long-term experience of the Spanish DILI Registry Camilla Stephens1,14,†, Mercedes Robles-Diaz1,14,†, Inmaculada Medina-Caliz1, Miren Garcia-Cortes1,14, Aida Ortega-Alonso1,14, Judith Sanabria-Cabrera1,15, Andres Gonzalez-Jimenez1, Ismael Alvarez-Alvarez1, Mahmoud Slim1, Miguel Jimenez-Perez2, Rocio Gonzalez-Grande2, M. Carmen Fernández3, Marta Casado3, German Soriano4,14, Eva Román4,5,14, Hacibe Hallal6, Manuel Romero-Gomez7,14, Agustin Castiella8, Isabel Conde9,14, Martin Prieto9,14, Jose Maria Moreno-Planas10, Alvaro Giraldez11, J. Miguel Moreno-Sanfiel12, Neil Kaplowitz13, M. Isabel Lucena1,14,15,*, Raúl J. Andrade 1,14, Participating clinical centres 1 UGC Aparato Digestivo and Servicio de Farmacología Clínica, Instituto de Investigación Biomédica de Málaga-IBIMA, Hospital Universitario Virgen de la Victoria, Universidad de Málaga, Málaga, Spain; 2UGC Aparato Digestivo, Complejo Hospitalario Regional de Málaga, IBIMA, Málaga, Spain; 3UGC Farmacia, Hospital Universitario Torrecárdenas, Almeria, Spain; 4Servicio de Gastroenterología, Hospital de la Santa Creu i Sant Pau, Universitat Autònoma de Barcelona, Spain; 5Escola Universitària d’Infermeria EUI-Sant Pau, Universitat Autònoma de Barcelona, Barcelona, Spain; 6Servicio Aparato Digestivo, Hospital General Universitario J.M. Morales Meseguer, Murcia, Spain; 7UGC Aparato Digestivo, SeLiver Group IBIS, Hospital Universitario Virgen del Rocío, Universidad de Sevilla, Sevilla, Spain; 8Servicio de Aparato Digestivo, Hospital Universitario de Donostia, San Sebastian. Spain; 9Unidad de Hepatología, Servicio de Aparato Digestivo, Hospital Universitari i Politècnic La Fe, Valencia, Spain; 10Servicio de Digestivo, Complejo Hospitalario Universitario de Albacete, Albacete, Spain; 11Servicio de Aparato Digestivo. Hospital Virgen del Rocío, Sevilla, Spain; 12Servicio de Aparato Digestivo. Hospital Universitario de Canarias, Tenerife, Spain; 13University of Southern California Research Center for Liver Diseases, Keck School of Medicine, Los Angeles, California, USA; 14Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas (CIBERehd), Madrid, Spain; 15Platform ISCiii for Clinical Research and Clinical Trials UICEC- IBIMA, Málaga, Spain Background & Aims: Prospective drug-induced liver injury 6-month overall mortality, for which comorbidity burden played (DILI) registries are important sources of information on idio- a more important role. The prognostic capacity of Hy’s law varied syncratic DILI. We aimed to present a comprehensive analysis of between causative agents. Empirical therapy (corticosteroids, 843 patients with DILI enrolled into the Spanish DILI Registry ursodeoxycholic acid and MARS) was prescribed to 20% of pa- over a 20-year time period. tients. Drug-induced autoimmune hepatitis patients (26 cases) Methods: Cases were identified, diagnosed and followed pro- were mainly females (62%) with hepatocellular damage (92%), spectively. Clinical features, drug information and outcome data who more frequently received immunosuppressive therapy were collected. (58%). Results: A total of 843 patients, with a mean age of 54 years (48% Conclusions: AST elevation at onset is a strong predictor of poor females), were enrolled up to 2018. Hepatocellular injury was outcome and should be routinely assessed in DILI evaluation. associated with younger age (adjusted odds ratio [aOR] per year Mortality is higher in older patients with hepatocellular damage 0.983; 95% CI 0.974–0.991) and lower platelet count (aOR per and patients with underlying hepatic conditions. The Spanish unit 0.996; 95% CI 0.994–0.998). Anti-infectives were the most DILI Registry is a valuable tool in the identification of causative common causative drug class (40%). Liver-related mortality was drugs, clinical signatures and prognostic risk factors in DILI and more frequent in patients with hepatocellular damage aged > −65 can aid physicians in DILI characterisation and management. years (p = 0.0083) and in patients with underlying liver disease Lay summary: Clinical information on drug-induced liver injury (p = 0.0221). Independent predictors of liver-related death/ (DILI) collected from enrolled patients in the Spanish DILI Reg- transplantation included nR-based hepatocellular injury, female istry can guide physicians in the decision-making process. We sex, higher onset aspartate aminotransferase (AST) and bilirubin have found that older patients with hepatocellular type liver values. nR-based hepatocellular injury was not associated with injury and patients with additional liver conditions are at a higher risk of mortality. The type of liver injury, patient sex and Keywords: Hepatotoxicity; DILI; epidemiology; liver-related death; causative agents; analytical values of aspartate aminotransferase and total bili- outcome; risk factors; therapy in DILI; drug-induced autoimmune hepatitis. rubin can also help predict clinical outcomes. Received 16 September 2020; received in revised form 13 January 2021; accepted 15 © 2021 European Association for the Study of the Liver. Published by January 2021; available online 1 February 2021 Elsevier B.V. This is an open access article under the CC BY-NC-ND * Corresponding author. Address: Departamento de Farmacología y Pediatría, Facultad de Medicina, Boulevard Louis Pasteur 32, Universidad de Málaga, 29071 Málaga, license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Spain; Tel.: +34-952-131572, fax: +34-952-131568. E-mail address: lucena@uma.es (M.I. Lucena). † Co-first authors. https://doi.org/10.1016/j.jhep.2021.01.029 Journal of Hepatology 2021 vol. 75 j 86–97 Introduction structure and procedures of the registry, data collection and case Drug-induced liver injury (DILI), particularly unpredictable enrolment have been published elsewhere.8 Clinical data corre- idiosyncratic DILI, can have a significant impact, which is not sponding to each patient with DILI is collected using a stand- limited to patient safety but also affects healthcare costs, drug ardised protocol to ensure that information necessary to development and the range of marketed drug treatments. Hence, adjudicate DILI is collected: (1) detailed medication history there has been a growing interest in enhancing the under- including herbal and dietary supplements (HDS) and over-the- standing of DILI both from a clinical, epidemiological and mo- counter medications; (2) biochemistry, detailed viral serology lecular point of view over the last decades. work-up (including viral hepatitis E on a routine basis since Improved medication safety analyses and toxicological 2016), imaging and, if available, histological data to exclude studies have led to a notable reduction in hepatotoxicity-related alternative causes of liver injury; (3) outcome. The biochemical post-marketing drug withdrawals issued by the US Food and criteria for DILI used in this registry were initially those estab- Drug Administration since 1997, reflecting a greater emphasis on lished by the CIOMS and later adapted to those of Aithal et al. hepatotoxicity in drug development.1 The ability to predict who (alanine aminotransferase [ALT] > −5 x upper limit of normal is at risk of developing idiosyncratic DILI is in line with the [ULN], alkaline phosphatase [ALP] > −2 xULN or ALT > −3 xULN concept of personalized medicine and would enhance drug together with total bilirubin [TBL] >2 xULN).20,21 In the study safety by enabling patients without DILI risk to securely benefit cohort, 86% fulfilled the more stringent criteria of Aithal et al. at from an effective treatment (including treatments with black box detection. warnings), while providing alternative medications to those at Cases induced by acetaminophen and occupational exposure increased risk of DILI. A better understanding of DILI and its to toxins were excluded from the study cohort. In addition, cases implication for the individual patient is required to reach this of drug-induced autoimmune hepatitis (DIAIH) were analysed as goal.2 Considerable progress has been made in understanding a distinct cohort. Diagnosis of DIAIH was based on the following: the pathophysiological mechanisms of idiosyncratic hepatotox- a temporal relationship between drug intake and the appearance icity,3 however, translation to DILI prediction and treatment is of an autoimmune hepatitis (AIH) phenotype; no prior evidence still a work in progress.4 of AIH; and cases fulfilled the simplified AIH criteria.22 For pa- Because of the lack of reliable animal models that can tients with multiple episodes (re-exposure to the causative agent reproduce the complexity of idiosyncratic DILI, prospective or a second episode induced by a different drug, recurrent DILI) collection of phenotypic information and biological samples data pertaining to one of the episodes only were included in the from identified DILI cases is still the most valuable resource for current study to avoid duplication of demographic data. All pa- DILI research. Idiosyncratic DILI is relatively rare, with an esti- tients underwent follow-up until liver profile normalisation, mated yearly incidence rate of approximately 14-19 cases per when possible. Therapy for DILI, if any, was decided by the 100,000 (based on prospective population-based studies in physician in charge, recorded and analysed. Europe), or 23.8 per 100,000 (based on a more recent retro- The pattern of liver injury (hepatocellular, cholestatic and spective Chinese study).5–7 The low incidence rate makes it mixed) was determined by calculating the ratio (R) of ALT to ALP difficult for individual hospital units/research groups to obtain from the first available blood analysis after DILI recognition, sufficient numbers of cases to perform studies with high statis- using multiples of the ULN for both values.21 Severity was tical power. To circumvent this issue a number of prospective assessed using the severity index defined by Aithal et al.21 Death DILI registries have emerged. The establishment of national and and need for liver transplantation were assessed within a international DILI registries over the last 30 years has provided a maximum of 6 months from DILI onset. significant step forward in understanding DILI.8–12 The Spanish Eosinophilia was defined as serum eosinophils exceeding 4- DILI Registry, founded in 1994, was a pioneer initiative to 6% of total leukocyte count depending on the normal range of strengthen DILI epidemiological figures, phenotypic character- individual hospitals, and lymphopenia as serum lymphocytes isations, risk factor identification, and prognosis. Our group has <10%. Patients with hypersensitivity features were those who also contributed to facilitating standardised diagnostic and cau- presented with at least one of the following features: rash, fever, sality assessment procedures over more than 20 years.13–19 eosinophilia or lymphopenia. The Charlson comorbidity index An analysis of 461 DILI cases enrolled into the Spanish DILI (CCI) was calculated as the total of a patient’s weighted comorbid Registry over the first 10-year period was published in 2005,8 conditions according to Charlson et al.23 Heavy alcohol con- and many of the findings have since been replicated in other sumption was defined as > −60 g (men) or > −40 g (women) of large DILI cohorts.6,7,9–12 With a near doubling of enrolled cases alcohol per day. in the Spanish DILI Registry since 2005 it is timely to undertake a The study protocol was approved by the local Ethics Com- new analysis. In the present study, we aimed to provide an mittee at the Virgen de la Victoria University Hospital in Málaga, updated description of clinical features, outcomes of special Spain, and all subjects gave informed consent. populations, management and main therapeutic groups featured in 843 prospectively recruited individuals in the Spanish DILI Statistical analysis Registry. Variables were examined using descriptive statistics. Continuous variables were presented as mean ± SD or median (IQR), as appropriate. Qualitative variables were described using fre- Material and methods quency distributions. Inferential statistics were used to compare Design groups. Differences in continuous variables between groups The Spanish DILI Registry, established in 1994, is a prospective were assessed using Mann-Whitney U test or Kruskal-Wallis test multicentre study focusing on prospectively identifying bona fide (post hoc: Dunn’s test) or ANOVA, as appropriate. Qualitative DILI cases, mainly idiosyncratic DILI cases. The operational variables were compared using Pearson v2 or Fisher’s exact test. Journal of Hepatology 2021 vol. 75 j 86–97 87 Research Article DILI, Autoimmune, Cholestatic and Genetic Diseases Table 1. Comparison of demographics, clinical characteristics, laboratory parameters and outcome between different patterns of liver injury in 843 Spanish DILI cases. Total registry n = 843 Hep n = 482 (57%) Chol n = 173 (21%) Mix n = 188 (22%) p value* Age (yr), mean ±SD (range) 54±18 (11-91) 51±18 (11-88) 61±17 (16-90) 55±18 (14-91) <0.0001 Female, % 48 50 43 46 0.2140 BMI (kg/m2), mean ±SD 26 ±3.8 26 ±3.8 26 ±3.9 26 ±3.7 0.6294 Diabetes mellitus, % 12 11 17 11 0.1192 Hypertension, % 20 21 37 31 0.0007 Dyslipidaemia, % 14 11 17 16 0.0748 Underlying hepatic disease, % 6.3 5.6 6.4 7.9 0.5223 History of drug allergy, % 15 19 7.3 13 0.009 DILI episode characteristics Jaundice, % 69 66 77 70 0.0327 Rash, % 7.9 7.0 8.8 8.8 0.5411 Hospitalisation, % 60 57 69 59 0.0317 Total oral daily dose (mg), mean ±SD 925 ±1,056 786 ±995 1,222 ±1,257 1,068 ±1,127 0.0018 Duration of therapy (d), mean/median (IQR) 63/27 (8-64) 71/32 (9-82) 43/16 (9-41) 61/16 (8-56) <0.0001 Time to onset (d), mean/median (IQR) 58/25 (10-62) 66/30 (12-73) 40/22 (9-39) 56/20 (8-45) <0.0001 Concomitant drugs, % None 26 28 20 28 0.0667 1-2 drugs 40 41 40 36 3-4 drugs 21 20 22 22 > −5 drugs 13 10 18 13 Laboratory parameters at onset x ULN, mean ±SD TBL 7.0 ±6.9 6.4 ±6.8 8.9 ±7.3 6.8 ±6.4 <0.0001 AST 15 ±21 23 ±25 3.3 ±3.0 6.1 ±7.5 <0.0001 ALT 19 ±22 28 ±25 3.9 ±3.8 7.7 ±6.3 <0.0001 ALP 2.2 ±2.1 1.3 ±0.9 4.2 ±3.2 2.4 ±1.9 <0.0001 INR 1.3 ±0.7 1.4 ±0.7 1.2 ±0.6 1.1 ±0.4 0.0001 Glucose (mg/dl) 113 ±52 110 ±52 124 ±66 110 ±31 0.0025 eGFR (ml/min/1.73 m2) 96 ±65 100 ±76 94 ±63 90 ±31 0.1949 Haemoglobin (g/dl) 14 ±1.8 14 ±1.7 13 ±1.8 14 ±1.7 <0.0001 Platelets x103/ll 233 ±90 223 ±89 253 ±103 240 ±73 0.0003 Lymphopenia, % 24 19 32 26 0.0047 Peripheral eosinophilia, % 23 20 26 27 0.0800 Positive autoantibody titres, % 20 25 17 12 0.0021 Severity, % Mild 31 36 20 28 <0.001 Moderate 59 51 73 66 Severe 6.2 7.3 5.8 3.7 Fatal/transplantation 3.7 5.4 1.2 1.6 Outcome Time to resolution (d), median (IQR) 108 (56-218) 103 (50-192) 132 (68-272) 107 (59-199) 0.2275 Liver-related death, n (%) 18 (2.1) 14 (2.9) 1 (0.6) 3 (1.6) 0.1765 Liver transplantation, n (%) 13 (1.5) 12 (2.5) 1 (0.6) 0 0.0309 Death due to other causes§, n (%) 14 (1.7) 5 (1.0) 8 (4.6) 1 (0.5) 0.0058 Statistical tests: Pearson chi-squared test or Fisher’s exact test, as appropriate, for qualitative variables; Kruskal-Wallis test or ANOVA, as appropriate, for quantitative variables. ALP, alkaline phosphatase; ALT, alanine aminotransferase; AST, aspartate aminotransferase; Chol, Cholestatic; DILI, drug-induced liver injury; eGFR, estimation glomerular filtration rate according to the Modification of Diet in Renal Disease study; Hep, hepatocellular; INR, international normalised ratio; Mix, mixed; TBL, total bilirubin; ULN, upper limit of normal. *comparison between Hep, Chol and Mix groups. § During time of follow-up. Multivariable logistic regression models were fitted to study comorbidity), while 52 patients (6%, mean age 64 years) had associations of clinical and demographic characteristics on DILI significant comorbidity with a CCI >2. The most frequent con- phenotypic expression. All statistical tests were 2-sided hy- ditions contributing to CCI were diabetes (12%), chronic pulmo- potheses performed at the 0.05 level of significance using STATA nary disease (6.4%) and congestive heart failure (5.8%). The v 13.0 (College Station, TX: StataCorp LP). majority of DILI cases (86%) were judged to have a single caus- ative agent, while 2 culprit drugs were attributed to 14% of the cases. The causative agents were mainly taken orally (94%; mean Results daily dose 925 mg over a median duration of 27 days). Of the Clinical and demographic characteristics cases caused by oral conventional drugs, 25% involved a daily A total of 843 DILI cases enrolled in the Spanish DILI Registry dose of <100 mg, 18% <50 mg and 6.4% < −10 mg (Table S1). Drugs between 1994 and 2018 fulfilled the inclusion criteria for the were given parenterally in 48 cases (60% intravenously, 23% current study. Using the CIOMS/RUCAM scale, 33% of the cases intramuscularly, 10% cutaneously/subcutaneously, 4.2% inhaled were classified as highly probable, 52% as probable and 15% as and 2.1% sublingually). The most common parenterally given possible. The mean age of the patients with DILI was 54 years causative agents were antibacterials 25%, immunomodulating with a similar distribution between males and females (Table 1). agents 17% and antineoplastics 15%. The median latency (time 791 patients (94%, mean age 54 years) had a CCI < −2 (none to mild 88 Journal of Hepatology 2021 vol. 75 j 86–97 from drug initiation to detection of symptoms/elevated liver (aOR 0.983 per year; 95% CI 0.974-0.991; p <0.001) and lower profile) was 25 days and the median time to resolution 108 days. platelet count at DILI recognition (aOR 0.996 per unit; 95% CI Of the 843 patients with DILI, 23% had peripheral eosinophilia, 0.994-0.998; p <0.001). The hepatocellular group had lower oral 69% presented with jaundice and 60% required hospitalisation. daily dose (p = 0.0018), but longer time to onset (p <0.0001) than Thirty-two cases with eosinophilia and/or lymphopenia also the other groups. Hospitalisation was more common in cases presented with rash and were consequently diagnosed as DRESS with a cholestatic pattern of injury (p = 0.0317) Nevertheless, (drug reaction with eosinophilia and systemic symptoms) cases fatal/transplantation outcomes were more frequent among he- and 3 as Stevens-Johnson syndrome. Eighty-four patients (15%) patocellular cases (p = 0.0082). Survival curves for the 3 different reported a history of drug allergy. Biopsy was performed in 141 types of liver injury likewise demonstrated a significant differ- (17%) patients. The most common features included cholestatic ence (p = 0.0118), with hepatocellular cases being 4 times more hepatitis (37%), hepatocellular necrosis (7.8%) and cholestasis likely to develop the worst outcome within 90 days than (7.1%). In total, 18 patients (2.1%) died from liver-related causes cholestatic cases (Fig. S1). and 13 (1.5%) underwent liver transplantation. These 31 patients had significantly higher comorbidity with 12.9% having a CCI >2 Influence of age and injury pattern on the clinical compared to 5.9% among those with a favourable outcome (p = presentation and outcome of DILI 0.016). A comparison between age groups (young: < −45 years old; middle- The distribution of liver injury pattern according to age and aged: 46-64 years old; and old: > −65 years old) according to injury sex is depicted in Fig. 1. The peak DILI age was 60-69 years for pattern (hepatocellular and cholestatic/mixed) was performed both males and females. Hepatocellular injury was the pre- (Table 2). Sex distribution was equal across all groups. As expected, dominant phenotype (57%) in all age groups except for patients increased age was associated with higher BMI (p <0.0004), dia- > −80 years in whom cholestatic injury predominated. Patients did betes mellitus (p <0.0001), hypertension (p <0.0001), dyslipidae- not differ in age, sex, BMI, type or severity of liver injury between mia (p <0.0001) and polypharmacy (p <0.0001) in both liver injury those enrolled in the first 10 years and thereafter, but prevalence groups. The frequency of jaundice was highest in the oldest age of hypertension, dyslipidaemia and the use of immunosuppres- category in both the hepatocellular (73% vs. 60% and 68%) and sants, immunostimulants and antineoplastic agents was signifi- cholestatic/mixed group (83 % vs. 65% and 67%). Younger age was cantly higher in the later period (data not shown). linked to shorter duration of therapy in hepatocellular cases, ranging from a median of 25 days in the youngest patients to 40 and 43 days in the older age groups (p = 0.0080), while the Comparison of clinical and demographic characteristics opposite was found for the cholestatic/mixed cases with a median between different patterns of liver injury of 31 days in the youngest patients vs. 16 and 15 days in the older Patients with a cholestatic pattern of injury were older (mean age groups (p = 0.0251). Hepatocellular patients aged > age of 61 years compared to 51 and 55 years for patients with −65 years had the highest proportion of liver-related fatalities (7.2%) hepatocellular and mixed pattern, respectively [p <0.0001]) compared with the younger age groups (1.2% and 1.6%; p = 0.0083). (Table 1). Similar sex distributions were seen across the groups. Hepatocellular injury based on the definition of the new Ratio However, men had a significantly higher risk of developing (nR) >5 (nR = AST or ALT in ULN (whichever highest)/ALP in amoxicillin-clavulanate-induced cholestatic type liver injury ULN)18 (aOR 4.914; 95% CI 1.316-18.350; p = 0.018), AST elevation (adjusted odds ratio [aOR] 2.249; 95% CI 1.342-3.769; p = 0.002). (aOR 1.015; 95% CI 1.002-1.028; p = 0.024), female sex (OR 2.744; Comorbid conditions such as diabetes mellitus, hypertension and 95% CI 1.180-6.380; p = 0.019) and TBL value at recognition (aOR dyslipidaemia, and correspondingly polypharmacy (> −5 drugs) 1.102; 95% CI 1.054-1.152, p <0.001) were independent predictors were more frequently seen in cholestatic patients. Platelet count of liver-related mortality and liver transplantation. differed significantly between the 3 groups (p = 0.0003). Hepa- Interestingly, an nR-based hepatocellular injury pattern was tocellular injury was independently associated with younger age not associated with 6-month overall mortality, while patients who died within 6 months had higher comorbidity burden compared to those who survived (mean CCI 2.25 vs. 0.6, p 100 Cholestatic Hepatocellular <0.001). This finding highlights the greater relevance of comor- 94 Mixed bidity burden over liver injury pattern when focusing on overall 85 80 79 mortality. Regarding the influence of alcohol consumption, there was no difference between the proportion of heavy and no/light Number of patients 71 60 62 drinkers with regards to liver-related death/liver transplantation (5.7% vs. 3.8%, p = 0.641) or 6-month overall mortality (8.6% vs. 53 50 2.9%, p = 0.099). 40 43 41 29 32 Comparison of clinical and demographic characteristics 20 24 24 22 between DILI patients with and without pre-existing liver 20 19 16 13 16 13 conditions 10 3 9 8 As shown in Table 3, 53 patients suffered from underlying he- 0 10-19 20-29 30-39 40-49 50-59 60-69 70-79 ≥80 Years patic conditions before the DILI episode. These conditions 12/16 41/30 57/44 59/66 59/73 101/89 78/57 31/23 Male/ included chronic viral hepatitis (55%), alcohol-related liver dis- Female (n) ease (23%), fatty liver disease (11%), idiopathic AIH (5.7%), alpha- Fig. 1. Distribution of 836 DILI patients in the Spanish DILI Registry ac- 1 antitrypsin deficiency (1.9%), iron metabolism disorder (1.9%), cording to pattern of liver injury and age. primary biliary cholangitis (1.9%), primary sclerosing cholangitis Journal of Hepatology 2021 vol. 75 j 86–97 89 Table 2. Comparison of demographics, clinical characteristics, laboratory parameters and outcome between different age groups in hepatocellular and cholestatic/mixed DILI. 90 Research Article Hepatocellular Cholestatic/Mixed < −45 yr 46-64 yr > −65 yr < −45 yr 46-64 yr > −65 yr n = 182 n = 169 n = 125 p value n = 84 n = 124 n = 152 p value Female, % 50 52 46 0.6188 38 52 41 0.0774 BMI (kg/m2); mean ±SD 25 ±4.0 26 ±3.2 27 ±3.8 <0.0001 25 ±4.0 26 ±3.4 27 ±3.7 0.0004 Diabetes mellitus, % 2.8 14 18 <0.0001 3.6 9.7 23 <0.0001 Hypertension, % 5.2 23 38 <0.0001 3.5 33 50 <0.0001 Dyslipidaemia, % 2.2 15 19 <0.0001 3.6 22 20 0.0010 Underlying hepatic disease, % 6.0 5.3 5.6 0.9578 9.5 10 3.3 0.0464 History of drug allergy, % 17 19 21 0.732 3.9 14 11 0.154 DILI episode characteristics, % Jaundice 60 68 73 0.0568 67 65 83 0.0010 Rash 7.2 8.3 5.2 0.6086 5.6 14 7.0 0.0695 Hospitalisation 53 53 67 0.0334 64 52 73 0.0019 Total oral daily dose (mg), mean ±SD 832 ±975 724 ±897 730 ±940 0.6383 816 ±925 1,122 ±1,127 1,307 ±1,287 0.0126 Duration of therapy (d), mean/median (IQR) 53/25 (8-62) 89/40 (10-95) 72/43 (8-93) 0.0080 74/31 (9-76) 56/16 (8-47) 38/15 (8-33) 0.0251 Time to onset (d), mean/median (IQR) 52/25 (11-61) 79/34 (11-81) 70/33 (13-76) 0.1998 69/26 (9-62) 48/19 (8-38) 37/21 (8-37) 0.1734 Journal of Hepatology 2021 vol. 75 j 86–97 Concomitant dugs, % None 37 31 12 <0.0001 37 27 15 <0.0001 1-2 drugs 41 41 42 44 37 36 3-4 drugs 18 19 25 18 23 23 > −5 drugs 4.4 8.9 22 1.2 13 26 Laboratory parameters at onset x ULN, mean ±SD TBL 5.8 ±7.3 5.7 ±5.6 7.9 ±7.3 0.0057 7.4 ±7.9 7.1 ±7.1 8.5 ±6.1 0.0074 AST 25 ±28 23 ±22 22 ±23 0.9452 4.4 ±6.2 4.6 ±5.2 5.1 ±6.4 0.0826 ALT 30 ±27 30 ±26 24 ±20 0.3159 5.1 ±3.5 5.8 ±5.5 6.4 ±6.5 0.2428 ALP 1.2 ±0.7 1.4 ±1.0 1.4 ±0.9 0.0153 2.4 ±2.1 3.3 ±2.9 3.7 ±2.9 <0.0001 INR 1.4 ±0.7 1.3 ±0.7 1.5 ±0.8 0.1664 1.2 ±0.5 1.1 ±0.4 1.1 ±0.6 0.4104 DILI, Autoimmune, Cholestatic and Genetic Diseases Glucose (mg/dl) 98 ±39 116 ±57 119 ±55 <0.0001 96 ±18 119 ±51 126 ±60 <0.0001 eGFR (ml/min/1.73 m2) 117 ±61 92 ±35 86 ±118 <0.0001 105 ±31 96 ±55 83 ±50 <0.0001 Hemoglobin (g/dl) 14 ±1.7 14 ±1.6 13 ±1.8 0.0098 14 ±1.7 13 ±1.8 13 ±1.9 0.0030 Platelets x103/ll 243 ±92 218 ±98 202 ±68 0.0005 268 ±109 244 ±87 235 ±76 0.1322 Lymphopenia, % 15 15 32 0.0013 30 28 29 0.9383 Peripheral eosinophilia, (%) 19 23 17 0.4132 24 25 29 0.7090 Positive autoantibody titres, % 17 29 31 0.0156 4.2 21 15 0.0071 Severity, % Mild 45 34 25 0.0004 27 33 15 0.0073 Moderate 41 59 58 64 61 80 Severe 9.3 4.7 8.0 4.8 4.0 5.3 Fatal/transplantation 5.4 3.0 8.8 3.6 1.6 0 Outcome Time to resolution (d); median (IQR) 96 (48-172) 97 (52-182) 117 (61-327) 0.1881 124 (72-218) 106 (60-198) 122 (62-340) 0.4365 Liver-related death, n (%) 3 (1.6) 2 (1.2) 9 (7.2) 0.0083 3 (3.6) 1 ( 0.8) 0 0.0336 Liver transplantation, n (%) 7 (3.8) 3 (1.8) 2 (1.6) 0.5866 0 1 (0.8) 0 0.5778 Death due to other causes§, n (%) 1 (0.5) 1 (0.6) 3 (2.4) 0.3143 1 (1.2) 3 (2.4) 5 (3.3) 0.6345 Statistical tests: Pearson chi-squared test or Fisher's exact test, as appropriate, for qualitative variables; Kruskal-Wallis test or ANOVA, as appropriate, for quantitative variables. ALP, alkaline phosphatase; ALT, alanine amino- transferase; AST, aspartate aminotransferase; DILI, drug-induced liver injury; eGFR, estimation glomerular filtration rate according to the Modification of Diet in Renal Disease study; INR, international normalised ratio; TBL, total bilirubin; ULN, upper limit of normal. § During time of follow-up. Table 3. Comparison of demographics and clinical characteristics between subjects with and without underlying hepatic conditions. With underlying hepatic condition* n = 53 Without underlying hepatic condition n = 790 p value Age (yr), mean ±SD (range) 52 ±15 (26-83) 54 ±18 (11-91) 0.1706 Female, % 30 49 0.0087 BMI (kg/m2), mean ±SD 26 ±4.7 26 ±3.7 0.7206 Diabetes mellitus, % 7.6 13 0.3864 Hypertension, % 16 28 0.1079 Dyslipidaemia, % 11 14 0.6282 Pattern of DILI (Hep/Chol/Mix, %) 51/21/28 58/21/22 0.5223 DILI episode characteristics Jaundice, % 65 70 0.4709 Rash, % 4.3 8.1 0.3401 Hospitalisation, % 49 60 0.1270 Laboratory parameters at onset x ULN, mean ±SD TBL 6.9 ±6.8 7.0 ±6.9 0.9301 AST 15 ±27 15 ±21 0.5710 ALT 14 ±13 19 ±23 0.2689 ALP 2.0 ±1.8 2.2 ±2.2 0.4819 INR 1.6 ±0.9 1.3 ±0.6 0.0341 Platelets x 103/ll 209 ±100 235 ±90 0.0413 Albumin, mg/dl 3.7 ±1.0 3.98 ±1.7 0.2930 Outcome Liver-related death due to DILI 4 (7.5) 14 (1.8) 0.0221 Liver transplantation 0 13 (1.6) 0.6216 Statistical tests: Pearson chi-squared test or Fisher’s exact test, as appropriate; for qualitative variables; Mann–Whitney U test for quantitative variables. ALP, alkaline phosphatase; ALT, alanine aminotransferase; AST, aspartate aminotransferase; Chol, Cholestatic; DILI, drug-induced liver injury; Hep, hepatocellular; INR, in- ternational normalised ratio; TBL, total bilirubin; ULN, upper limit of normal. *Underlying hepatic diseases include alcoholic liver disease, alpha-1 antitrypsin deficiency, chronic viral hepatitis, fatty liver disease, idiopathic autoimmune hepatitis, iron metabolism disorder, primary biliary cholangitis, primary sclerosing cholangitis and prior liver transplantation. Table 4. Comparison of demographics, clinical characteristics and outcome between treated (corticosteroids, UDCA and MARS) and non-treated DILI patients. Corticosteroids (n = 66) UDCA (n = 50) MARS (n = 12) No treatment (n = 497) p value Age (yr), mean ±SD (range) 53 ±20 (16-88) 55 ±18 (17-91) 41 ±18 (20-73) 54 ±18 (11-90) 0.170 Female, % 55 46 33 48 0.628 Diabetes mellitus, % 12 10 8.3 11 0.983 Hypertension, % 17 26 17 20 0.844 Dyslipidaemia, % 9.1 10 17 16 0.669 Underlying hepatic disease, % 7.6 4.0 0 6.2 0.855 DILI episode characteristics Type of liver injury, % (Hep/Chol/Mix) 55/26/20 48/22/30 67/17/17 59/18/23 0.521 Jaundice, % 89 88 100 65a,b,c <0.001 Hospitalisation, % 91 67a,c 100 46a,b,c <0.001 Hypersensitivity features, % 48c 51c 83 40c 0.010 Rash, % 12 16 17 5.7b 0.021 Lymphopenia, % 26 12 33 18 0.111 Eosinophilia, % 29 26 17 22 0.602 Peak laboratory parameters xULN, mean ±SD TBL 14.7 ±11.4 16.6 ±12.0 29.8 ±16.9a 7.8 ±8.7a,b,c <0.001 AST 21.2 ±23.0 20.1 ±31.0 10.6 ±14.6 17.8 ±24.7 0.317 ALT 24.4 ±30 21.3 ±25.0 12.4 ±9.2 20.9 ±24.4 0.824 ALP 2.8 ±3.0 3.3 ±3.4 2.8 ±2.3 2.2 ±2.1a,b 0.001 INR 1.4 ±0.7 1.4 ±0.8 1.3 ±0.7 1.3 ±0.6 0.312 Severity, % <0.001 Mild 12 6.0 0 35a,b,c Moderate 62 76 67 57b Severe 17 10 17 5.6a Fatal/transplantation 9.1 8.0 17 2.0a,b,c Outcome Time to resolution (d), median (IQR) 112 (79-183) 142 (79-288) 147 (92-741) 96 (49-178) 0.193 Liver-related death, n (%) 4 (6.1) 3 (6.0) 1 (8.3) 9 (1.8)a 0.011 Liver transplantation, n (%) 2 (3.0) 1 (2.0) 1 (8.3) 1 (0.2)a,b,c 0.011 Chronicity†, n (%) 3 (4.6) 3 (6.0) 1 (8.3) 38 (7.7) 0.751 Statistical tests: Pearson chi-squared test or Fisher’s exact test, as appropriate, for qualitative variables; Kruskal-Wallis test and Dunn’s test (post hoc) for quantitative variables. ALP, alkaline phosphatase; ALT, alanine transaminase; AST, aspartate transaminase; DILI, drug-induced liver injury; INR, international normalised ratio; MARS, molecular adsorbents recirculation system; TBL, total bilirubin; UDCA, ursodeoxycholic acid. a: p <0.05 vs. corticosteroids; b: p <0.05 vs. ursodeoxycholic acid; c: p <0.05 vs. MARS. † Defined as cases who failed to resolve (normalisation of liver biochemistry, imaging test or histology) within 365 days. Journal of Hepatology 2021 vol. 75 j 86–97 91 Research Article DILI, Autoimmune, Cholestatic and Genetic Diseases Table 5. Comparison of demographics, clinical characteristics, laboratory parameters and outcome between DILI-autoimmune hepatitis (DIAIH) and DILI cases. DIAIH n = 26 DILI n = 843 p value Age (yr), mean ±SD (range) 57±17 (15-86) 54±18 (11-91) 0.5504 Female, % 62 48 0.162 BMI (kg/m2), mean ±SD 25 ±5.0 26 ±3.8 0.1592 Diabetes mellitus, % 15 12 0.550 Hypertension, % 28 20 0.318 DILI episode characteristics Type of liver injury, % 0.002 Hepatocellular 92 57 Cholestatic 4.0 21 Mixed 4.0 22 Jaundice, % 69 69 0.953 Rash, % 4.5 7.9 1.000 Hospitalisation, % 39 54 0.205 Duration of therapy* (d), median (IQR) 65 (27-274) 27 (8-64) 0.0044 Laboratory parameters at onset xULN, mean ±SD TBL 5.7 ±5.5 7.0 ±6.9 0.6656 AST 24 ±17 15 ±21 0.0001 ALT 28 ±19 19 ±22 0.0002 ALP 2.2 ±2.8 2.2 ±2.1 0.8643 Autoantibodies ANA, % 88 12 <0.001 ASMA, % 44 8.9 <0.001 AMA, % 4.0 1.9 0.397 Anti-LKM-1, % 0 1.1 1.000 IgG (g/L), mean (SD) 19.5 ±10.7 11.9 ±4.6 <0.001 Treatment, % <0.001 Corticosteroids/azathioprine 58 9.9 Ursodeoxycholic acid 3.8 7.5 MARS 0 1.8 Severity, % 0.784 Mild 35 31 Moderate 54 59 Severe 7.7 6.2 Fatal/transplantation 0/1 (3.8) 18/13 (3.7) Statistical tests: Pearson chi-squared test or Fisher’s exact test, as appropriate, for qualitative variables; Mann–Whitney U test for quantitative variables. ALP, alkaline phosphatase; ALT, alanine aminotransferase; ANA, anti-nuclear antibodies; anti-LKM-1, anti-liver kidney microsome type 1 antibodies; ASMA, anti-smooth muscle antibodies; AMA, anti-mitochondrial antibodies; AST, aspartate aminotransferase; DILI, drug-induced liver injury; MARS, molecular adsorbents recirculation sys- tem; TBL, total bilirubin; ULN, upper limit of normal. *Duration of causative agent before DILI detection. (1.9%) and prior liver transplantation (1.9%). Fourteen patients (MARS, 9.4%) (Table 4). These patients did not differ from those (26%) had cirrhosis. The most frequently implicated drugs in without DILI treatment with regards to demographics and un- patients with DILI and underlying hepatic conditions were anti- derlying conditions. Hepatocellular injury predominated in all TB medications (11, 21%), amoxicillin-clavulanate (4, 7.5%) fol- groups, with MARS-treated patients presenting the highest lowed by atorvastatin, fluvastatin, norfloxacin, flutamide, proportion (67%) and UDCA-treated patients the lowest (48%). azathioprine and ibuprofen (2, 3.8% each). Jaundice and hospitalisation were both significantly more The patients with pre-existing liver conditions were pre- frequent in the treatment groups (p <0.001). This was also re- dominantly males (70%), with a similar age to those without pre- flected in higher values of TBL and a higher rate of death/liver existing conditions. No significant differences were found in the transplantation in patients treated with corticosteroids (9.1%), clinical presentation between the 2 groups, except for interna- UDCA (8.0%) and MARS (17%), compared to non-treated patients tional normalised ratio (INR), which was higher in patients with (2.0%). Although a large proportion of patients receiving MARS underlying hepatic diseases (1.6 vs. 1.3, p = 0.0341). Furthermore, treatment presented with moderate severity injury, this group liver-related death due to DILI was significantly more common in had the highest values of TBL and anabolic androgenic steroids patients with underlying hepatic diseases (4 [7.5%] vs. 14 [1.8%], (AAS) were the culprit agents in a large fraction of cases (58%). p = 0.0221). Amoxicillin-clavulanate (20%) and anti-TB treatments (11%) predominated as culprit agents among the corticosteroid- Analysis of DILI cases according to therapy received for the treated patients. Similarly, amoxicillin-clavulanate was highly episode represented among the UDCA-treated patients (34%). Recorded information on therapy received for the DILI episode was available for 625 cases. In total 128 patients received spe- Comparison of clinical and demographic characteristics cific therapy to ameliorate DILI at discretion of the physician in between DILI and DIAIH patients charge, which included corticosteroids (52%), ursodeoxycholic The 843 DILI cases were also assessed against an independent acid (UDCA, 39%), or molecular adsorbents recirculation system cohort of 26 DIAIH cases. These cases were mainly induced by 92 Journal of Hepatology 2021 vol. 75 j 86–97 Table 6. Number of DILI patients stratified by age attributed to specific causative agents. ATC Main pharmacological groups, n (%) Total registry n = 843 < −45 yr n = 266 46-64 yr n = 293 > −65 yr n = 277 A Alimentary tract and metabolic agents excluding anabolic agents 46 (5.5) 16 (6) 19 (6.5) 11 (4) Drugs for peptic ulcer drugs 28 9 12 7 B Antithrombotic agents 18 (2.1) 2 (0.7) 5 (1.7) 11 (4) C Cardiovascular agents 89 (11) 8 (3) 40 (14) 39 (14) ACE inhibitors+angiotensin II antagonists 20 1 8 10 Statins 44 3 20 20 Fibrates 7 1 5 1 D Dermatologicals 6 (0.7) 3 (1.1) 2 (0.7) 1 (0.4) G Genito-urinary system and sex hormones 20 (2.4) 10 (3.8) 8 (2.7) 2 (0.7) H Thyroid therapy 10 (1.2) 3 (1.1) 6 (2) 1 (0.4) J Anti-infectives 337 (40) 95 (36) 112 (38) 127 (46) Antibacterials for systemic use 266 64 90 110 Amoxicillin-clavulanate 193 42 66 84 Penicillins/cephalosporins excluding amoxicillin-clavulanate 16 8 2 5 Macrolides 18 6 6 6 Fluoroquinolones 31 5 11 15 Antimycobacterials 64 26 20 17 L Antineoplastic and immunomodulating agents 66 (7.8) 18 (6.8) 22 (7.5) 26 (9.4) Antineoplastic agents 17 5 8 4 Endocrine therapy 26 4 4 18 Immunosuppressants 17 4 9 4 M Musculoskeletal system 90 (11) 31 (12) 27 (9.2) 31 (11) Nonsteroidal anti-inflammatory drugs 78 26 25 26 N Central nervous system 99 (12) 43 (16) 31 (11) 24 (8.7) Antiepileptics 25 15 6 4 Antipsychotics 12 6 1 5 Antidepressants 28 12 10 5 – Herbal products and dietary supplements 29 (3.4) 10 (3.8) 15 (5.1) 4 (1.4) – Anabolic androgenic steroids 22 (2.6) 21 (7.9) 1 (0.3) 0 ACE, angiotensin converting enzyme; DILI, drug-induced liver injury. statins (31%) and antibacterials (23%). A comparison between the anti-TB, ibuprofen, flutamide, levofloxacin, nimesulide and car- 2 groups revealed a predominance of females (62% vs. 48%) and bamazepine (Table 7). hepatocellular injury (92% vs. 57%) in DIAIH. This was reflected in Cases caused by AAS, nimesulide and flutamide demonstrated significantly higher ALT (28 xULN vs. 19 xULN, p = 0.0002) and the highest mean TBL values at recognition, while paroxetine and AST (24 xULN vs. 15 xULN, p = 0.0001) in the same group. atorvastatin (hepatocellular) had highest mean ALT values. In Duration of therapy was found to be significantly longer in the addition, atorvastatin (cholestatic) and amoxicillin-clavulanate DIAIH cases (65 vs. 27 days, p = 0.0044). However, no significant (cholestatic) led to the highest mean ALP values (Table S2). differences were detected with regards to severity and outcome The most frequent causative agents associated with at least 8 (Table 5). DILI cases were examined with regards to Hy’s law and its prognostic value (Table 7). Of the 38 anti-TB cases, 16 (43%) met nR-based Hys law criteria and 4 of these cases (25%) led to death/ Therapeutic classes and individual drugs most commonly liver transplantation. Similarly, 12 of the 22 cases attributed to implicated in the Spanish DILI Registry flutamide (57%) met the same criteria and 4 patients (33%) died A total of 221 different causative drug treatments were impli- or required a liver transplantation. In contrast, cases attributed to cated in the 843 DILI cases in this study, with 791 cases caused by isoniazid, atorvastatin, diclofenac and AAS that fulfilled nR-based conventional medications, 51 by HDS (including AAS products Hy’s law criteria (57%, 44%, 50% and 47%, respectively) had for body building purposes), and 1 by a compound under favourable outcomes without fatalities/liver transplantation. investigation. Anti-infectives (337 cases) were the most common therapeutic class followed by central nervous system drugs (99 Discussion cases), musculoskeletal drugs (90 cases, of which 78 were caused In this study we have characterised 843 patients with DILI by non-steroidal anti-inflammatory drugs) and cardiovascular enrolled into the Spanish DILI Registry from the initiation in 1994 agents (89 cases) (Table 6). A comparison of clinical character- to 2018, with a particular focus on the influence of age and istics between individual causative agents associated with at presence of underlying liver conditions on phenotype and least 8 DILI cases in the overall Spanish DILI Registry is presented outcome of the DILI episode. Age and sex distributions in the in Table 7. Drugs with a predominant hepatocellular phenotype current study population were similar to those previously re- were isoniazid (95%, mainly as TB prophylaxis) and diclofenac ported from our registry, with a mean age of 54 years and 48% (93%). While no drugs had a strong predominance for a chole- females compared to 53 years and 49% in a previous analysis.8 static or mixed phenotype, azathioprine (64%) and fluvastatin Thirty-three percent of the total patient population were 65 (64%) were the drugs with the highest proportion of cholestatic years or older, which is a substantially higher proportion than in and mixed cases, respectively. Among the most frequent causa- a recent report from the US Drug-Induced Liver Injury Network tive drugs, fatal cases were reported for amoxicillin-clavulanate, (DILIN), in which only 17% of patients with DILI were > −65 years Journal of Hepatology 2021 vol. 75 j 86–97 93 Research Article DILI, Autoimmune, Cholestatic and Genetic Diseases old.10 Differences in patient age between DILI cohorts can lead to combinations; Chol, cholestatic; DILI, drug-induced liver injury; HDS, herbal and dietary supplements; Hep, hepatocellular; Mix, mixed; Tx, transplantation. Causative agents attributed to less than 8 cases in the Spanish DILI Fulfilment of nR-based Hy’s law criteria was based on blood analysis values at DILI recognition. AAS, anabolic androgenic steroids; Anti-TB, tuberculosis treatments consisting of rifampicin, isoniazid, pyrazinamide and ethambutol 2 (33) 4 (33) 0 0 0 0 0 0 0 0 0 1 0 0 0 1 (14) True nR-based Hy’s law (death/liver Tx), n (%) 3 (5.6) 4 (25) 2 (9.5) variations in causative agent frequency and comorbidities based on age-related differences in the use of many treatments. Registry, but with at least 1 case resulting in fatal/liver transplantation included: bicalutamide, calcium carbimide, clomethiazole, clopidogrel, disulfiram, ebrotidine, mianserin, nefazodone, retinol and sevoflurane. Consistent with previous reports, hepatocellular type of liver injury predominated in the current study population. Hepato- cellular damage was associated with lower platelet counts. Reduced platelet number has been associated with increased severity in DILI and acute liver failure (ALF) in general.24–26 A lower mean platelet count in the hepatocellular patients may therefore reflect the higher proportion of more severe cases in this subgroup, similar to INR with a higher mean value in he- patocellular cases. The role of platelets in acute liver injury is still debated, with evidence supporting both injury exacerbation and recuperation depending on the level of platelet count deviation (thrombocytopenia/thrombocythemia) and degree of platelet activation.27 Interestingly, platelets are a major source of circu- Table 7. Comparison of demographic and clinical DILI episode characteristics of individual causative agents registered in the Spanish DILI Registry. 0 0 12 (57) 12 (57) 7 (44) 4 (33) 1 (9.1) 3 (33) 9 (47) Fulfil nR-based Hy’s law criteria, n (%) 54 (29) 16 (43) 6 (25) 7 (50) 3 (30) 7 (78) 2 (29) 3 (38) 3 (50) 21 (75) lating extracellular vesicles, as demonstrated in, for example, patients with ALF,28,29 and could be of interest in the search for predictive biomarkers in liver conditions including DILI. Sixty percent of patients with DILI in the current study required hospitalisation, which highlights the fact that DILI, despite its rareness, is an economic burden on healthcare sys- tems. The proportion of hospitalisation was even higher, reaching up to 73%, in older patients with jaundice and chole- static/mixed type of liver injury, possible because this is a more vulnerable population.30 Nevertheless, the hepatocellular DILI cases were associated with a significantly higher proportion of death/liver transplantation, with older patients (> −65 years old) having the highest rate of fatal outcome. The reason for this is 33 33 33 Eosinophilia % 17 19 18 9.1 0 12 11 30 4.8 9.5 55 36 50 9.5 7.7 29 unclear. Aside from less frequent liver transplantation due to age restriction criteria, it could be associated with general age- related deficiency in recovery due to diminished capacity for tissue repair. The increased prevalence of comorbid conditions in older patients may also contribute to increased severity. This 33 Lymphopenia % 0 14 0 44 76 19 23 5.6 23 5.9 30 20 30 20 25 40 29 20 is supported by the higher fraction of patients with CCI >2 detected among patients with DILI who died or required liver transplantation. In contrast, patients aged 65 years and older in the DILIN population were not found to have a higher mortality and liver transplantation rate than the younger patients with DILI.10 Interestingly, the Spanish DILI registry demonstrates slightly better outcome data than the DILIN registry,10 with 2.1% Female % 14 44 17 100 0 46 32 43 40 54 54 45 80 89 75 38 62 48 64 vs. 3% liver-related death, 1.5% vs. 4% liver transplantation and 1.7% vs. 3.2% non-liver-related death, respectively. Reasons explaining these differences could be the lower proportion of 20 25 females (48% vs. 59%), diabetes mellitus (12% vs. 25%) and pre- Mix 50 23 31 10 44 27 31 0 27 9.1 9.1 0 0 0 4.8 8.3 64 Pattern of DILI, % existing liver diseases (6.3% vs. 10%) in the Spanish cohort. The AST value at DILI onset was independently associated 13 12 9.1 0 18 9.1 10 12 0 27 30 25 6.7 50 36 22 6.9 25 Chol 64 with the risk of ALF, as was hepatocellular injury, but only when calculated using nR.18 This underscores that AST has a higher 70 76 44 44 9.1 18 Hep 39 95 93 42 82 55 78 62 75 50 50 93 64 sensitivity than ALT in ALF prediction, and nR consequently performs better than the classical R for this purpose as previ- ously reported by the Spanish DILI Registry and validated by the n 21 16 15 12 11 11 11 11 10 9 8 8 8 193 38 25 22 22 29 US DILIN group.18,25 Thus, we suggest that AST should be included in the biochemical criteria for DILI assessment and be Amoxicillin-clavulanate performed routinely in DILI evaluation. Our findings may also support the use of nR-based Hy’s law instead the traditional Hy’s Carbamazepine law in drug development. Hepatocellular damage, however, was Erythromycin Valproic acid Azathioprine Levofloxacin Atorvastatin Simvastatin only found to be a predictor of liver-related death, but not of Nimesulide Ticlopidine Fluvastatin Paroxetine Diclofenac Flutamide Ibuprofen Isoniazid overall mortality. This confirms earlier findings in the DILIN Anti-TB cohort by Ghabril et al.31 Thus, prediction of liver-related death HDS AAS based on type of liver damage is useful in order to refer the 94 Journal of Hepatology 2021 vol. 75 j 86–97 patient to a liver transplant centre, while the comorbidity This confirms previous findings for DIAIH.33 The DIAIH group burden instead of type of liver injury would be useful when also had significantly longer latency. The reason for this is un- assessing the risk of overall mortality. The impact of pre-existing known and may be related to the attributed agents. In fact, liver disease on DILI susceptibility and outcome is not yet fully notably long latency has been reported previously for DIAIH, elucidated. Only 6.3% of the patients with DILI in the current particularly due to nitrofurantoin and infliximab.34,35 Patients study had underlying hepatic conditions, but DILI was more with DIAIH in the current study received immunosuppressive frequently associated with a fatal outcome in these patients. treatments more frequently but showed similar outcome with Similarly, 10% of North American patients with DILI have been regards to mortality and liver transplantation. In fact, DIAH reported to have pre-existing liver conditions and higher mor- outcome is generally good after withdrawal of the causative tality, although liver-related mortality did not differ significantly agent and immunosuppression, with low risk of relapses or between the North American patients with and without pre- progression to chronic liver injury.33 existing liver disease.10 The discrepancy between our data and In the current analysis, 5.7% of the causative agents were those of the DILIN cohort may come from differences in types of given parenterally, which in part accounted for a growing rep- underlying liver conditions. While viral hepatitis was highly resentation of immunomodulating agents. This is an emerging represented in both registries, alcohol-related liver conditions area where drug metabolism is not a consideration. We continue were more frequent in the Spanish registry and non-alcoholic to observe that most cases had a daily dose of the attributed drug fatty liver disease in the American registry. Anti-TB drugs were exceeding 50 mg. Anti-infectives, in particular antibacterials, found to be implicated in 21% of cases with underlying hepatic remain the main causative drug class in the Spanish DILI Regis- conditions, which is considerably higher compared to 7.6% in the try, independent of patient age. The proportion of antibiotics was entire study cohort. Most of these cases were positive for chronic highest among the cholestatic cases in the study population, viral hepatitis, which is often related to parenteral drug abuse with amoxicillin-clavulanate being responsible for a third of all and subsequently a population at higher risk of TB infection. cholestatic cases. This may explain why the cholestatic group had Anti-TB drugs are also among the drugs associated with the a significantly shorter time to onset as well as higher total daily highest rate of poor outcomes in our registry, which may also dose compared with the hepatocellular cases. contribute to increased severity among cases with underlying Antimycobacterials represented a large subgroup that mainly hepatic conditions. The reduced proportion of amoxicillin- included cases caused by combined anti-TB treatments and clavulanate cases, on the other hand, could be related to physi- isoniazid alone. These causative agents, in particular isoniazid, cians’ decisions to avoid this known hepatotoxic drug in patients have long been associated with increased DILI susceptibility in with chronic liver conditions. Hence, our data show that chronic older patients.36,37 Interestingly, our findings do not corroborate liver disease increases the likelihood of mortality related to liver this. A relatively young median age (49 years) was also found for dysfunction in patients suffering from hepatotoxicity. Presum- 60 North American patients with hepatotoxicity due to isoniazid ably, underlying hepatic conditions diminish the liver’s capacity in a report from the DILIN registry.38 Some of the causative to recuperate from a DILI episode and subsequently increase the agents were unequally represented with regards to patient sex. It probability of a more severe outcome. Interestingly, we did not is unknown if this is due to biological differences increasing find evidence to support that heavy alcohol consumption susceptibility or in some instances is simply due to the nature of significantly enhances the risk of a poor DILI outcome. This the condition that the treatment is prescribed for. corroborates earlier findings from the DILIN cohort, but also The prognostic value of Hy’s law has been validated in various comes with the same limitation of self-reported alcohol intake large DILI cohorts, but little is known about its applicability to that may be underestimated.32 specific causative drugs. We found considerable differences Therapeutic approaches for DILI are used in some cases, when comparing the proportion of cases attributed to a specific although limited evidence is available to determine efficacy. Due causative agent that fulfils nR-based Hy’s law criteria and the to the lack of uniform clinical guidelines for the use of DILI proportion of these with a fatal/liver transplantation outcome. treatments, the decision to initiate such treatments is often left Almost half of the anti-TB cases fulfilled nR-based Hy’s law and to the physician in charge. The effect of pharmaceutical treat- 25% of these had the worst outcome, which is in line with Hyman ments in the current study is not assessable as the criteria for Zimmerman’s observations. Several other causative agents for treatment initiation varied between the corresponding physi- which up to 57% of cases met nR-based Hy’s law did not have any cians. As expected, the treated patients presented a more severe fatal/liver transplantation outcomes. Overall, these findings episode reflected in higher degree of jaundice and hospital- show that Hy’s law performance differs between different isation. This, however, increases the probability of a fatal/liver causative agents. transplantation outcome in these patients compared to the non- While providing important information on DILI, our study treated group, independent of the treatment effect. Despite the also has limitations. These include the lack of information on wide use of treatments for managing DILI, limited evidence is paediatric DILI because paediatric units do not participate in the available to demonstrate effectiveness. Our analysis, while Spanish DILI Registry. Secondly, DILI caused by immune check- providing a picture of what clinicians do in real practice, also point inhibitors, an emerging issue over recent years, is not highlights the necessity to undertake well-designed clinical trials covered in this study because the use of these drugs was still to determine the true effect of DILI treatments. limited in Spain at the time of this analysis. DIAIH is a distinct form of DILI that has been increasingly In conclusion, we have presented hypothesis-generating recognised in the last decade. Twenty-six DIAIH cases are findings on phenotypic variations, outcome and causative currently enrolled in the Spanish DILI registry. Compared with agents in DILI based on cases enrolled in the Spanish DILI Reg- the 843 conventional DILI cases, a higher proportion of females istry over 20 years. The ever-expanding Spanish DILI Registry, and hepatocellular liver injury were noted in the former group. while confirming many findings of the prior smaller cohort from Journal of Hepatology 2021 vol. 75 j 86–97 95 Research Article DILI, Autoimmune, Cholestatic and Genetic Diseases the first decade, has identified new findings and insights not Hospital Regional Universitario de Málaga: M Jiménez Pérez, previously appreciated, which can aid physicians in DILI case R González Grande, S López Ortega, I Santaella, A Ocaña, P characterisation and management, and provide a basis for deci- Palomino; sion making by regulatory authorities. Hospital Torrecárdenas, Almería: MC Fernández, G Peláez, A Porcel, M Casado, M González Sánchez; Abbreviations Hospital Universitario Virgen del Rocío, Sevilla: M Romero- AAS, anabolic androgenic steroids; AIH, autoimmune hepatitis; Gómez, R Millán-Domínguez, B Fombuena, R Gallego, J Ampuero, ALF, acute liver failure; ALP, alkaline phosphatase; ALT, alanine JA del Campo, R Calle-Sanz, L Rojas, A Rojas, A Gil Gómez, E Vilar; aminotransferase; AST, aspartate aminotransferase; CCI, Charlson Hospital Sant Pau, Barcelona: G Soriano, C Guarner, EM Ro- comorbidity index; DIAIH, DILI-autoimmune hepatitis; DILI, mán, MA Quijada Manuitt, RM Antonijoan Arbos; drug-induced liver injury; DILIN, drug-induced liver injury Hospital Parc Tauli, Barcelona: J Sánchez Delgado, M Vergara network; HDS, herbal and dietary supplements; INR, interna- Gómez; tional normalised ratio; MARS, molecular adsorbents recircula- Hospital Morales Meseguer, Murcia: H Hallal, E García Oltra, tion system; UDCA, ursodeoxycholic acid; ULN, upper limit of JC Titos Arcos, A Pérez Martínez, C Sánchez Cobarro, JM Egea normal. Caparrós; Hospital Universitario de Donostia, San Sebastián: A Castiella, E Zapata, J Arenas, A Gómez García, FJ Esandi; Financial support Hospital de Basurto, Bilbao: S Blanco, P Martínez Odriozola; The present study has been supported by grants of Instituto de Hospital Marqués de Valdecilla, Santander: J Crespo, P Iru- Salud Carlos III cofounded by Fondo Europeo de Desarrollo zubieta, J Cabezas; Regional – FEDER (contract numbers: PI19/00883, PI16/01748, Hospital Virgen del Rocío, Sevilla: A Giráldez Gallego, E del P PI18/00901, PI18/01804, PI-0285-2016, PI-0274-2016, PI-0310- Rodríguez Seguel, M Cuaresma; 2018, PT17/0017/0020) and Agencia Española del Medicamento. Hospital de León, León: J González Gallego, F Jorquera, S CIBERehd and Plataforma ISCIII Ensayos Clinicos are funded by Sánchez Campos; Instituto de Salud Carlos III. MRD holds a Joan Rodes (JR16/ Hospital Alto Deba Mondragón, Guipúzcoa: P Otazua, A de 00015)/Acción B clinicos investigadores (B-0002-2019) and JSC a Juan Gómez; Rio Hortega (CM17/00243) research contract from ISCIII and Hospital Universitario San Cecilio, Granada: J Salmerón, A Consejería de Salud de Andalucía. The funding sources had no Gila, R Quiles; involvement in the study design; in the collection, analysis, and Hospital Clínico Valladolid: JM González, S Lorenzo; interpretation of data; in the writing of the report or in the de- Hospital La Fe, Valencia: M Prieto, I Conde Amiel, M Beren- cision to submit the manuscript for publication. guer, M García-Eliz, Hospital de Sagunto, Valencia: J Primo, JR Molés, A Garayoa; Conflict of interest Hospital de Laredo, Cantabria: M Carrascosa; The authors have no conflict of interest to disclose. Hospital 12 de Octubre, Madrid: E Gómez Domínguez, L Please refer to the accompanying ICMJE disclosure forms for Cuevas; further details. Hospital Germans Trias i Pujol, Badalona, Barcelona: M Farré, E Montané, AM Barriocanal, AL Arellano, Y Sanz, RM Morillas, M Authors’ contributions Sala, H Masnou Ridaura; Study concept and design: CS, MRD, MIL, RJA; Case recruitments: Hospital Clínic, Barcelona: M Bruguera, P Gines, S Lens, JC MRD, MGC, AOA, JSC, MJP, RGG, MCF, MC, GS, ER, HH, MRG, AC, García, Z Mariño; EM, AMB, IC, MP, JMMP, AG; Case diagnosis: MRD, MCG, AOA, Hospital Universitario de Canarias, La Laguna, Tenerife: M MIL, RJA; Data acquisition: IMC; Statistical analyses: AGJ, IAA, Hernández Guerra, JM Moreno Sanfiel, C Boada Fernández del MS; Analysis and interpretation of data: CS, MRD, IMC, JSC, AGJ, Campo; IAA, MIL, RJA; Drafting of the manuscript: CS, MRD, MIL, RJA; Hospital Infanta Elena, Valdemoro, Madrid: M Tejedor, R Critical revision of the manuscript: NK, MIL, RJA González Ferrer; Hospital de Alcorcón, Alcorcón, Madrid: C Fernández, M Fer- nández Gil; Data availability statement Hospital Reina Sofía, Córdoba: JL Montero, M de la Mata; All data analysed in this article were obtained from the Spanish Hospital Miguel Servet, Zaragoza: J Fuentes Olmo, EM Fer- DILI Registry database, a privately owned database maintained nández Bonilla; by the registry’s coordinating center in Malaga, Spain. The Complejo Hospitalario Universitario de Albacete, Albacete: JM database is not publicly available. Moreno, P Martínez-Rodenas, M Garrido, C Oliva; Hospital Puerta del Mar, Cádiz: P Rendón; Participating clinical centres Hospital Carlos III, Madrid: J García Samaniego, A Madejón; Hospital Universitario Virgen de la Victoria, Málaga (coordi- Hospital Puerta de Hierro, Madrid: JL Calleja, JL Martínez nating centre): RJ Andrade, MI Lucena, C Stephens, M García Porras; Cortés, M Robles-Díaz, A Ortega-Alonso, J Pinazo, B García Hospital de Galdakao, Bizkaia: JL Cabriada; Muñoz, R Alcántara, A Hernández, MD García Escaño, E del Hospital Universitario Puerto Real, Cádiz: JM Pérez-Moreno, C Campo, I Medina-Cáliz, J Sanabria-Cabrera, A González-Jiménez, Lara R Sanjuán-Jiménez, A Cueto, I Álvarez-Álvarez, E Bonilla, D Di Zeo, H Niu, M Villanueva, A Papineau; 96 Journal of Hepatology 2021 vol. 75 j 86–97 Supplementary data to predict acute liver failure in patients with drug-induced liver injury. Gastroenterology 2014;147:109–118. Supplementary data to this article can be found online at https:// [19] Medina-Caliz I, Robles-Diaz M, Garcia-Muñoz B, Stephens C, Ortega- doi.org/10.1016/j.jhep.2021.01.029 Alonso A, Garcia-Cortes M, et al. Definition and risk factors for chronicity following acute idiosyncratic drug-induced liver injury. J Hepatol References 2016;65:532–542. Author names in bold designate shared co-first authorship [20] Benichou C. Criteria of drug-induced liver disorders. Report of an inter- national consensus meeting. J Hepatol 1990;11:272–276. [21] Aithal GP, Watkins PB, Andrade RJ, Larrey D, Molokhia M, Takikawa H, [1] Senior JR. Evolution of the Food and Drug Administration approach to et al. Case definition and phenotype standardization in drug-induced liver liver safety assessment for new drugs: current status and challenges. Drug injury. Clin Pharmacol Ther 2011;89:806–815. Saf 2014;37(Suppl 1):S9–S17. [22] Hennes EM, Zeniya M, Czaja AJ, Parés A, Dalekos GN, Krawitt EL, et al. [2] Andrade RJ, Chalasani N, Björnsson ES, Suzuki A, Kullak-Ublick GA, Simplified criteria for the diagnosis of autoimmune hepatitis. Hepatology Watkins PB, et al. Drug-induced liver injury. Nat Rev Primers 2019;5:58. 2008;48:169–176. [3] McGill MR, Jaeschke H. Animal models of drug-induced liver injury. Bio- [23] Charlson M, Szatrowski TP, Peterson J, Gold J. Validation of a combined chim Biophys Acta Mol Basis Dis 2019;1865:1031–1039. comorbidity index. J Clin Epidemiol 1994;47:1245–1251. [4] European Association for the Study of the Liver. EASL Clinical Practice [24] Lo Re 3rd V, Haynes K, Forde KA, Goldberg DS, Lewis JD, Carbonari DM, Guidelines: drug-induced liver injury. J Hepatol 2019;70:1222–1261. et al. Risk of acute liver failure in patients with drug-induced liver injury: [5] Sgro C, Clinard F, Ouazir K, Chanay H, Allard C, Guilleminet C, et al. Inci- evaluation of Hy’s law and a new prognostic model. Clin Gastroenterol dence of drug-induced hepatic injuries: a French population-based study. Hepatol 2015;13:2360–2368. Hepatology 2002;36:451–455. [25] Hayashi PH, Rockey DS, Fontana RJ, Tillmann HL, Kaplowitz N, [6] Björnsson ES, Bergmann OM, Björnsson HK, Kvaran RB, Olafsson S. Inci- Barnhart HX, et al. Death and liver transplantation within 2 years of onset dence, presentation, and outcomes in patients with drug-induced liver of drug-induced liver injury. Hepatology 2017;66:1275–1285. injury in the general population of Iceland. Gastroenterology [26] Stravitz RT, Ellerbe C, Durkalski V, Reuben A, Lisman T, Lee WM, et al. 2013;144:1419–1425. Thrombocytopenia is associated with multi-organ system failure in pa- [7] Shen T, Liu Y, Shang J, Xie Q, Li J, Yan M, et al. Incidence and etiology of tients with acute liver failure. Clin Gastroenterol Hepatol 2016;14:613– drug-induced liver injury in mainland China. Gastroenterolgoy 620. 2019;156:2230–2241. [27] Lisman T, Luyendyk JP. Platelets as modulators of liver disease. Semin [8] Andrade RJ, Lucena MI, Fernández MC, Pelaez G, Pachkoria K, García- Thromb Hemost 2018;44:114–125. Ruiz E, et al. Drug-induced liver injury: an analysis of 461 incidences [28] Balaphas A, Meyer J, Sadoul K, Fontana P, Morel P, Gonelle-Gispert C, et al. submitted to the Spanish registry over a 10-year period. Gastroenterology Platelets and platelet-derived extracellular vesicles in liver physiology 2005;129:512–521. and disease. Hepatol Commun 2019;3:855–866. [9] Suk KT, Kim DJ, Kim CH, Park SH, Yoon JH, Kim YS, et al. A prospective [29] Stravitz RT, Bowling R, Bradford RL, Key NS, Glover S, Thacker LR, et al. nationwide study of drug-induced liver injury in Korea. Am J Gastro- Role of procoagulant microparticles in mediating complications and enterol 2012;107:1380–1387. outcome of acute liver injury/acute liver failure. Hepatology [10] Chalasani N, Bonkovsky HL, Fontana R, Lee W, Stolz A, Talwalker J, et al. 2013;58:304–313. Features and outcomes of 899 patients with drug-induced liver injury: [30] Lucena MI, Sanabria J, García-Cortes M, Stephens C, Andrade RJ. Drug- the DILIN prospective study. Gastroenterology 2015;148:1340–1352. induced liver injury in older people. Lancet Gastroenterol Hepatol [11] Bessone F, Hernandez N, Mendizabal M, Sanchez A, Paraná R, Arrese M, 2020;5:862–874. et al. When the creation of a consortium provides useful answers: [31] Ghabril M, Gu J, Yoder L, Corbito L, Ringel A, Beyer CD, et al. Development experience of the Latin American DILI Network (LATINDILIN). Clin Liver and validation of a model consisting of comorbidity burden to calculate Dis (Hoboken) 2019;13:51–57. risk of death within 6 months for patients with suspected drug-induced [12] Aiso M, Takikawa H, Tsuji K, Kagawa T, Watanabe M, Tanaka A, et al. liver injury. Gastroenterology 2019;157:1245–1252. Analysis of 307 cases with drug-induced liver injury between 2010 and [32] Dakhoul L, Ghabril M, Gu J, Navarro V, Chalasani N, Serrano J, et al. Heavy 2018 in Japan. Hepatol Res 2019;49:105–110. consumption of alcohol is not associated with worse outcomes in patients [13] Lucena MI, Camargo R, Andrade RJ, Perez-Sanchez CJ, Sanchez de la with idiosyncratic drug-induced liver injury compared to non-drinkers. Cuesta F. Comparison of two clinical scales for causality assessment in Clin Gastroenterol Hepatol 2018;16:722–729. hepatotoxicity. Hepatology 2001;33:123–130. [33] Czaja AJ. Drug-induced autoimmune-like hepatitis. Dig Dis Sci [14] Lucena MI, Kaplowitz N, Hallal H, Castiella A, García-Bengoechea M, 2011;56:958–976. Otazua P, et al. Recurrent Drug-Induced Liver Injury (DILI) with different [34] de Boer YS, Kosinski AS, Urban TJ, Zhao Z, Long N, Chalasani N, et al. drugs in the Spanish Registry: the dilemma of the relationship to auto- Features of autoimmune hepatitis in patients with drug-induced liver immune hepatitis. J Hepatol 2011;55:820–827. injury. Clin Gastroenterol Hepatol 2017;15:103–112. [15] Andrade RJ, Lucena MI, Kaplowitz N, García-Munoz B, Borraz Y, [35] Björnsson ES, Bergmann O, Jonasson JG, Grondal G, Gudbjornsson B, Pachkoria K, et al. Outcome of acute idiosyncratic drug-induced liver Olafsson S. Drug-induced autoimmune hepatitis: response to corticoste- injury: long-term follow-up in a hepatotoxicity registry. Hepatology roids and lack of relapse after cessation of steroids. Clin Gastroenterol 2006;44:1581–1588. Hepatol 2017;15:135–1636. [16] Lucena MI, Andrade RJ, Fernández MC, Pachkoria K, Pelaez G, Durán JA, [36] Fountain FF, Tolley E, Chrisman CR, Self TH. Isoniazid hepatotoxicity asso- et al. Determinants of the clinical expression of amoxicillin-clavulanate ciated with treatment of latent tuberculosis infection: a 7-year evaluation hepatotoxicity: a prospective series from Spain. Hepatology from a public health tuberculosis clinic. Chest 2005;128:116–123. 2006;44:850–856. [37] Bright-Thomas RJ, Gondker AR, Morris J, Ormerod LP. Drug-related hep- [17] Lucena MI, Andrade RJ, Kaplowitz N, García-Cortes M, Fernández MC, atitis in patients treated with standard anti-tuberculosis chemotherapy Romero-Gomez M, et al. Phenotypic characterization of idiosyncratic over a 30-year period. Int J Tuberc Lung Dis 2016;20:1621–1624. drug-induced liver injury: the influence of age and sex. Hepatology [38] Hayashi PH, Fontana RJ, Chalasani NP, Stolz AA, Talwalkar JA, Navarro VJ, 2009;49:2001. et al. Under-reporting and poor adherence to monitoring guidelines for [18] Robles-Diaz M, Lucena MI, Kaplowitz N, Stephens C, Medina-Cáliz I, severe cases of isoniazid hepatotoxicity. Clin Gastroenterol Hepatol González-Jimenez A, et al. Use of Hy’s law and a new composite algorithm 2015;13:1676–1682. Journal of Hepatology 2021 vol. 75 j 86–97 97