Introduction

Idiosyncratic drug-induced liver injury (DILI) is an unexpected reaction to conventional medications, herbal products, or dietary supplements (Andrade et al. 2019). From 14 to 25% of DILI cases courses with immunoalergic characteristics (Chalasani et al. 2015; Stephens et al. 2021) and, in this context, DILI can manifest with severe cutaneous adverse reactions (SCARs).

Drug Reaction with Eosinophilia and Systemic Symptoms (DRESS) is a challenging immune-mediated reaction caused by a variety of drugs that presents with dermatological manifestations, often as urticated maculopapular eruption (MPE), and systemic features, being the liver the most frequent internal organ involved. DRESS can progress from mild damage to fatal cases, with estimated mortality that ranges from 1.7 to 8% (Kardaun et al. 2013; Lee et al. 2022). Previous studies have identified limited drugs and drug classes as responsible agents for DRESS, such as antiepileptics, allopurinol, sulphonamides, antibiotics, and non-steroidal anti-inflammatory drugs (Chen et al. 2010; Bluestein et al. 2021; Skowron et al. 2015; Fang et al. 2018).

The coexistence of different criteria is one of the reasons that hinders the diagnosis and assessment of DRESS (Kim and Koh 2014). Furthermore, the characterization of DILI associated with DRESS is controversial since there is no agreement about the criteria used to define liver injury in the context of immunoalergic reactions (Sanabria-Cabrera et al. 2019). Liver injury is usually defined as a mild elevation of transaminases (> 2 times the upper limit of normal [ULN]) (Kardaun et al. 2013), which is less stringent criteria than the thresholds proposed by an international DILI working group to exclude transaminase elevations of uncertain significance and an adaptative phenomena (Aithal et al. 2011). DILI associated with DRESS has gained interest in recent years, and several studies of Asian patients with this particular phenotype of DILI have been published recently (Huang et al. 2021; Devarbhavi et al. 2022).

In the setting of two prospective DILI cohorts with long-term follow-up, the Spanish DILI Registry and the Latin American DILI (LATINDILI) Network, we aimed to comprehensively characterize the causative drugs, clinical characteristics, laboratory features and outcomes of patients with DILI associated with DRESS (henceforth, DILI-DRESS).

Methods

Study population

Information from well-vetted idiosyncratic DILI cases included in the Spanish DILI Registry and the LATINDILI Network since their establishment until 2022 was collected. Details of these registries have been described elsewhere (Bessone et al. 2016; Stephens et al. 2021). A structured case report form was used to record pharmacological and clinical data, the description of skin lesions, blood test results, imaging findings to rule out other causes of liver damage, and the outcome of liver injury. The study protocols were approved by local ethics committees. All subjects gave informed written consent.

The biochemical criteria for DILI were those proposed by the Council for International Organizations of Medical Sciences (CIOMS) (Danan and Benichou 1993), later adapted to those set in 2011 (Aithal et al. 2011), i.e., serum alanine aminotransferase (ALT) elevation ≥ 5 × ULN, serum alkaline phosphatase (ALP) ≥ 2 × ULN, or the combination of ALT ≥ 3 × ULN with a simultaneous elevation of total bilirubin (TBL) > 2 × ULN. The pattern of liver injury was defined by the nR value, i.e., (ALT/ULN or aspartate aminotransferase [AST]/ULN, whichever highest ÷ ALP/ULN) (Robles-Diaz et al. 2014). Cases were classified as hepatocellular (nR ≥ 5), cholestatic (nR ≤ 2), or mixed (nR > 2 and < 5). Severity was graded into mild (TBL < 2 × ULN), moderate (TBL ≥ 2 × ULN), severe (TBL ≥ 2 × ULN, and either International Normalized Ratio [INR] ≥ 1.5, ascites and/or encephalopathy, or another organ failure due to DILI), and fatal or transplantation (liver-related death or liver transplantation) (Aithal et al. 2011). Time to DILI recognition (latency) was defined as the time from the start of drug intake to the onset of DILI. The number of patients meeting the nR-based Hy’s law criteria was calculated (Robles-Diaz et al. 2014). DILI cases were followed-up until liver injury resolution, i.e., all liver parameters below the upper limit of normal.

In all cases, other non-related drug causes of liver injury, such as viral hepatitis, biliary obstruction, alcoholism, autoimmune hepatitis, and according to the clinical context, metabolic disorders, ischemic hepatitis, septic shock, Epstein Barr, or cytomegalovirus infection, were excluded. A panel of experts evaluated the causal relationship between the suspected drug and liver damage. Case likelihood categorization was based on the categories of the Roussel Uclaf Causality Assessment Method (RUCAM) (Bessone et al. 2016; Stephens et al. 2021). Only cases that scored at least “possible” (≥ 3 points) were included.

DILI-DRESS cases were defined as those who had DILI and fulfilled at least three of the following criteria: acute skin rash, fever above 38 ºC, enlarged lymph nodes, internal organ involvement, or haematological abnormalities (lymphocytosis, lymphocytopenia, eosinophilia or thrombocytopenia) (Peyrière et al. 2006). Eosinophilia was defined, based on blood work at DILI recognition as serum eosinophils exceeding 4–6% of total leukocyte count depending on the normal range of individual hospitals, and lymphopenia as serum lymphocytes < 10%. The case ascertainment of DILI-DRESS was done based on the scoring system proposed by the RegiSCAR group (Peyrière et al. 2006). Only patients that scored at least “possible” were included. Stevens-Johnson syndrome (SJS), toxic epidermal necrolysis (TEN), and acute generalized exanthematous pustulosis (AGEP) cases were excluded.

For comparison purposes with the DILI-DRESS group, we defined a group of patients with DILI without DRESS (from now on, DILI). Cases with missing or incomplete information on any hypersensitivity features (fever, rash, eosinophilia, lymphopenia, and arthralgia) were excluded.

Statistical analysis

Categorical data were expressed using frequency distributions, and differences were tested with the chi-square test or Fisher's exact test, as appropriate. For quantitative data, mean and standard deviation (SD), or median and interquartile range (IQR), were computed, and the Student's t-test or Mann–Whitney U test, as appropriate, were used to test differences between groups. A backward stepwise logit model was fitted to identify prognostic factors in DILI-DRESS cases. At each step, variables were chosen based on a p-value lower than a specified threshold of 0.05. All analyses were performed using STATA 17 (College Station, TX: StataCorp LLC), and a p-value lower than 0.05 was deemed statistically significant.

Results

Demographics, clinical characteristics, and outcome of DILI-DRESS and DILI cases

Out of 1,437 patients, we identified 53 DILI-DRESS cases (29 in the Spanish DILI Registry and 24 in the LATINDILI Network). Thus, the prevalence of DRESS (with confirmed DILI) in these registries was 3.7%. Two SJS/TEN/AGEP cases were excluded from this study. In addition, 881 DILI cases were included. No differences in demographic and clinical characteristics were found between DILI-DRESS cases from the Spanish DILI Registry and LATINDILI Network, except that Latin American patients were younger (mean age 54 years in the Spanish cases vs. 39 years in the Latin American cases; p = 0.006).

Demographics, clinical characteristics, and outcomes of DILI-DRESS and DILI cases were compared in Table 1. DILI-DRESS patients were younger than DILI (mean age 47 vs. 53 years, respectively; p = 0.042). However, when only Spanish cases were analysed, no differences in age were found (mean age 54 years in both groups; p = 0.927). Furthermore, there were significant differences in the pattern of liver damage (p = 0.018). Hepatocellular injury was predominant in DILI (63%), while 56% of DILI-DRESS patients presented with cholestatic/mixed damage. Indeed, the median elevation of ALP was significantly higher in DILI-DRESS cases (median 2.1 × ULN) than in DILI cases (median 1.6 × ULN; p = 0.003). Likewise, gamma-glutamyl transferase (GGT) levels were increased in DILI-DRESS cases compared to DILI (median 7.2 vs. 5.5 times ULN; p = 0.039).

Table 1 Comparison of demographics, clinical characteristics, laboratory parameters and outcome between DILI-DRESS and DILI cases
Full size table

All DILI-DRESS cases had rash, and 85% presented with eosinophilia. Moreover, 55% of these patients had fever, 34% had lymphopenia, and only 14% suffered from arthralgia. When compared to DILI, DILI-DRESS cases had a lower prevalence of positive autoantibody titres (21% and 5.1%, respectively; p = 0.013), and the hospitalization rate was higher (50% vs. 79%, respectively; p < 0.001).

Eleven DILI-DRESS patients underwent liver biopsy. Histological findings showed cholestasis with hepatitis (n = 4), zonal necrosis (n = 3), chronic hepatitis (n = 2, active in one of them), steatohepatitis (n = 1), and unspecific changes in the liver (n = 1).

Even though the damage was moderate in most patients in both groups (55%), those in the DILI-DRESS group developed a severe liver injury more frequently than DILI cases (13% vs. 5.9%, respectively), albeit these differences did not reach statistical significance. Of note, 13 out of 53 DILI-DRESS cases (25%) were treated with corticosteroids. Furthermore, there were no differences in liver-related death. Only one DILI-DRESS patient, due to anti-tuberculosis (TB) drugs died (after liver transplantation). Conversely, death due to non-liver-related causes was higher among the DILI-DRESS patients (5.7%; n = 3), compared to DILI (1.0%; n = 9) (p = 0.026). Information of the 53 DILI-DRESS patients is further detailed in Table 2.

Table 2 Demographics, clinical and biochemical characteristics of the 53 patients with DILI-DRESS
Full size table

Prognostic factors in DILI-DRESS cases

DILI-DRESS patients who developed a severe-fatal injury had predominantly a hepatocellular injury (88%) and marked elevations of serum transaminases and TBL levels at DILI recognition compared to those with mild-to-moderate liver damage. In addition, 63% of cases with a more severe injury presented with eosinophilia, compared to 89% of patients with milder damage.

An exploratory backward stepwise regression analysis was performed to identify factors associated with the development of a worse outcome in DILI-DRESS cases out of the following variables (p-value < 0.1 in univariate analysis): type of liver injury (hepatocellular vs. cholestatic/mixed), eosinophilia, ALT, AST, ALP, TBL, and nR-based Hy's law. Of these, higher TBL levels at DILI recognition (odds ratio [OR] 1.23; 95% confidence interval [CI] 1.04–1.45) and absence of eosinophilia (OR 8.77; 95% CI 1.11–69.20) were found as prognostic factors of worse outcome in DILI-DRESS patients.

Causative agents

The most common agents implicated in DILI-DRESS were carbamazepine (13%), anti- TB medications (isoniazid, rifampicin, and pyrazinamide, either alone or in combination, 13%), amoxicillin-clavulanate (11%), and allopurinol and lamotrigine (7.6% each). On the other hand, amoxicillin-clavulanate (15%), anti-TB (6.6%), ibuprofen (3.5%), and diclofenac (3.1%) were the most frequent causative drugs in DILI cases. Notably, among all cases of carbamazepine-induced liver injury in the two registries (n = 15), 47% of them were DILI-DRESS, as well as four out of six cases due to lamotrigine (67%). However, in cases due to other antiepileptics such as valproic acid or phenytoin (n = 9 each), few of them developed DILI-DRESS (11% and 22%, respectively). Conversely, among the six cases of DILI due to allopurinol, 67% presented with a DILI-DRESS phenotype.

According to the Anatomic Therapeutic Classification groups, anti-infectives for systemic use were the most common drugs in both groups (40% in DILI-DRESS and 34% in DILI). Furthermore, drugs for the nervous system were overrepresented in DILI-DRESS compared to DILI (32% vs. 8.6%, respectively). In contrast, drugs for the cardiovascular system and antineoplastic and immunomodulating agents were more common among DILI patients (Online Resource 1).

An ancillary analysis was performed to compare the clinical characteristics of DILI-DRESS patients according to the most frequent culprit drugs and drug classes, i.e., antiepileptic drugs (carbamazepine, lamotrigine, phenytoin, valproic acid and levetiracetam), anti-TB drugs, amoxicillin-clavulanate, and allopurinol (Table 3). Young women were more prone to present DILI-DRESS caused by antiepileptic or anti-TB medications. DILI-DRESS was induced by amoxicillin-clavulanate after a shorter duration of therapy. Hepatocellular injury was distinctive in DILI-DRESS caused by anti-TB drugs, while the cholestatic/mixed pattern of liver injury was the predominant damage caused by the other drugs. Moreover, it is worth noting that eosinophilia was less prevalent in those DILI-DRESS patients who had taken anti-TB medications and had a poorer outcome.

Table 3 Characteristics of DILI-DRESS cases caused by the most frequent culprit drug and drug classes
Full size table

Discussion

In the present study, including cases of two prospective DILI registries from Western countries, we have comprehensively characterized a cohort of DILI-DRESS cases. Compared to DILI, DILI-DRESS presented more frequently with a cholestatic or mixed injury. Furthermore, these cases were more frequently hospitalized and tended to course with a more severe injury, albeit their prognosis was comparable to DILI cases. Causative agents responsible for DILI-DRESS showed a distinctive pattern, with antiepileptics as the leading drug class.

A summary of the clinical characteristics and outcomes of DILI-DRESS patients in prospective and retrospective studies published from Asian countries is shown in Table 4. The prevalence of DILI-DRESS in this study (3.7%) is similar to other long-term population-based studies (5.3%) (Huang et al. 2021), but significantly lower compared to a recent single-centre Indian study (19%) (Devarbhavi et al. 2022). Geographic, ethnic, and demographic differences and divergences in drug prescription patterns could explain these discrepancies in DILI-DRESS incidence. For instance, sulphonamides, mainly dapsone, rarely used in Western countries, was the 2nd drug class involved in DILI-DRESS in one of these studies (Devarbhavi et al. 2022). Furthermore, the increased mortality rates observed in these studies could be due to several factors, such a higher prevalence of pre-existing hepatic diseases, comorbidities, or a more frequent involvement of extrahepatic organs, which might have determined the poorer outcome (Chalasani et al. 2015; Ghabril et al. 2019).

Table 4 Demographics, clinical information, and outcome of DILI-DRESS cases in prospective and retrospective studies
Full size table

In the present study cholestatic-mixed injury was reported as the most common liver injury pattern in patients with DILI-DRESS. This finding is consistent with other studies that reported a significantly higher frequency of cholestatic injury in DILI-DRESS cases (Lin et al. 2015; Huang et al. 2021; Devarbhavi et al. 2022).

Antiepileptics were the most common drug class in our study. These drugs are metabolized by the cytochrome P450 to arene oxide metabolites, which are usually detoxified by the epoxide hydrolase or glutathione transferase to inactive metabolites (Spielberg et al. 1981). However, some investigations have reported that patients with DRESS caused by antiepileptics had a reduced detoxification capability due to defects in the epoxide hydrolase enzyme, resulting in the accumulation of reactive metabolites that may act as haptens that trigger an immune response (Shear and Spielberg 1988; Chung et al. 2014). This circumstance might explain the overrepresented incidence due to antiepileptic drugs despite DILI-DRESS and DILI share common pathogenic pathways (Cueto-Sanchez et al. 2021; Stirton et al. 2022).

One-third of DILI-DRESS caused by antiepileptics or anti-TB medications fulfilled the nR-based Hy's law. However, none of those cases caused by antiepileptics evolved into fatal outcomes. Conversely, one of the two cases treated with anti-TB died of acute liver failure. These findings underscore that Hy’s Law prognostic performance may vary with individual drugs (Stephens et al. 2021). Interestingly, DILI-DRESS cases due to anti-TB, who exhibited the highest mortality rate, had a lower prevalence of eosinophilia. Indeed, in our exploratory analysis, higher TBL levels and lack of eosinophilia were associated with a worse outcome in DILI-DRESS patients, underlining the role of eosinophilia in the prognosis of DILI-DRESS. Consistently, prior investigations have described an association between eosinophilia and a more favourable outcome of liver injury in DILI (Björnsson et al. 2007; Pachkoria et al. 2008). These findings indicate that liver-related mortality in DILI-DRESS depends on the culprit agent.

The main strength of this study is the large sample of patients recruited in these two prospective DILI registries with a standardized methodology. In addition, only DILI-DRESS cases that met international criteria were included, ensuring the internal validity of our findings. However, some limitations should be acknowledged. Information about skin lesions was reported in the standardized case report form, but no skin biopsies were available. Nonetheless, only cases that fulfilled the RegiSCAR criteria and were considered in the case ascertainment as possible or probable (Peyrière et al. 2006) were included as DILI-DRESS. Moreover, data from some biochemical parameters to further characterize these DILI-DRESS patients were not available, such as metabolomic profiles. Therefore, future studies should be planned to investigate if severity in DILI-DRESS patients is associated with a distinctive bile acid profile as occurs in DILI cases (Ma et al. 2019; Xie et al. 2021).

In conclusion, in this well-characterized cohort from two long-term prospective DILI registries, DILI-DRESS cases presented more frequently with a cholestatic/mixed injury pattern. They exhibited greater severity but similar rates of liver-related death and liver transplantation than DILI. Antiepileptic drug class was found as a distinctive causative drug group in DILI-DRESS. Lamotrigine and allopurinol have a greater chance of presenting with this phenotype. Exploratory analyses identified higher TBL levels and lack of eosinophilia as prognostic factors of poor outcomes. These findings represent a step forward to deepen the understanding of the distinctive clinical features and prognosis of DILI associated with DRESS.