Comedication with interacting drugs predisposes amiodarone users in cardiac and surgical intensive care units to acute liver injury: A retrospective analysis.

Risk factors and underlying mechanisms for liver injury associated with amiodarone remain elusive. This study aimed to investigate the drug-related covariates for acute liver injury by amiodarone-an intriguing compound of high lipophilicity, with a long half-life and notable efficacy.The medical, pharmacy, and laboratory records of new amiodarone users admitted to the cardiac or surgical intensive care units of a medical center were examined retrospectively. A Cox regression model with time-varying dose-related variables of amiodarone was utilized to estimate the hazard ratio (HR) of amiodarone-associated liver injury while adjusting for concomitant therapy and relevant covariates.Of the 131 eligible patients among 6,572 amiodarone users (46,402 prescriptions), 6 were identified as amiodarone-associated liver injury cases. In comparison to controls (n = 125), this liver injury cohort (n = 6) had significantly higher numbers of amiodarone-interacting (2.7 ± 2.0 vs 0.9 ± 0.9 drugs, P = .02) and hepatotoxic (3.8 ± 0.8 vs 2.5 ± 1.7 drugs, P = .03) comedications. The number of comedications with amiodarone-interacting potential (HR 2.07, 95% confidence interval [CI] 1.02-4.22, P = .04) and amiodarone cumulative doses standardized by body surface area (HR 6.82, 95% CI 1.72-27.04, P = .01) were independent risk factors for liver injury associated with amiodarone.Drug-related (amiodarone cumulative dose, interacting drugs) factors were significant predictors of amiodarone-associated acute liver injury. A prudent evaluation of each medication profile is warranted to attain precision medicine at the level of patient care, especially for those treated by medications with complex physicochemical and pharmacokinetic properties, such as amiodarone.

Introduction

The health care burden among the critically ill is substantial. Up to 78.0% of patients requiring intensive care had cardiac arrhythmias, including ventricular tachycardia (VT) and atrial fibrillation (AF).[ Amiodarone, an iodinated benzofuran derivative with distinct physicochemical (high lipophilicity) and pharmacokinetic properties (long half-life, t1/2), is an effective and widely used antiarrhythmic drug (AAD).[ In fact, it is the most commonly prescribed rhythm controller as documented previously.[

Nonetheless, amiodarone is infamous for its risk of end organ toxicity and undesired drug–drug interactions (DDIs). Fatal adverse reactions include pulmonary fibrosis, liver cirrhosis, and cardiac arrest subsequent to bradycardia; while the clinically relevant DDIs between amiodarone and many other pharmaceuticals lead to subtherapeutic (treatment failure) or supratherapeutic (toxicity) effects.[ To minimize risks, regular monitoring of organ functions and vigilant detection of clinically relevant DDIs during amiodarone therapy are recommended.[

Liver injury associated with amiodarone is not uncommon. The incidences of elevated liver enzymes and hepatitis or cirrhosis were observed to be 15% to 30% and < 3%, respectively.[ Although amiodarone per se,[ hypoxia (e.g., hepatic hypoperfusion),[ or polysorbate 80 (a co-solvent in parenteral formulations),[ have been postulated to be responsible for these effects, the exact underlying mechanisms or risk factors for hepatic damage remain elusive. Frail individuals under intensive care are usually treated with complex drug regimens. This intricate polypharmacy might elicit substantial interactions with amiodarone and, hence, contribute to the occurrence of drug-induced liver injury (DILI). Alternatively, DILI could solely be the consequence of comedication with hepatotoxic pharmaceuticals in the complex polypharmacy treatment.

A full assessment is more feasible among patients under critical care due to the availability of complete and detailed medical documents. This study, with particular emphasis on incorporating drug-related covariates such as comedications (DDI and DILI potential) and amiodarone doses (daily and cumulative), aims to evaluate the incidence of amiodarone-related acute liver injury, the frequencies of DILI and DDI coprescriptions, and possible risk factors for amiodarone-associated acute liver injury in the settings of cardiac and surgical intensive care units (ICUs) using medical records at a medical center in Taiwan.

Patients and methods

Study design and population

We performed a retrospective cohort study of patients > 18 years of age who began amiodarone therapy from February 1, 2011 to January 31, 2013 at the cardiac and surgical ICUs of a medical center (Fig. 1). The study was approved by the Research Ethics Committee of the National Taiwan University Hospital.

Figure 1

Patient selection and study flow. ALP = alkaline phosphatase, ALT = alanine aminotransferase, ICUs = intensive care units, LFTs = liver function tests, T-bil = total bilirubin, ULN = upper limit of normal.

Adult patients (age ≥ 18 years) who initiated amiodarone therapy at cardiac and surgical ICUs with baseline and follow-up alanine aminotransferase (ALT) and total bilirubin (T-bil) levels were eligible for the study. To minimize potential confounding effects, the exclusion criteria included recent users of amiodarone (patients who took amiodarone within the previous 6 months prior to the ICU admission), amiodarone therapy initiated at settings other than the target ICUs, and patients who had baseline liver function abnormalities apparently caused by factors other than amiodarone (e.g., escalating trends of liver biochemistry values prior to amiodarone initiation). The latest laboratory data within 6 months before amiodarone therapy were recorded as the baseline level. Moreover, individuals with apparent underlying liver or biliary tract diseases were also excluded, including patients with hepatitis (i.e., viral, alcoholic, autoimmune, or nonalcoholic steatohepatitis), parenchymal liver diseases, liver abscess, liver cirrhosis, portal hypertension, hepatic coma, biliary tract disorders, hepatorenal syndrome, or hepato-biliary malignancy.

Recent users or those ineligible were excluded (as summarized in Fig. 1).

Data source

The study utilized data from relevant medical charts, electronic medical records, laboratory data files, and pharmacy dispensing datasets. Relevant patient-level outpatient visit, ER visit, and hospitalization data were collected. Patient demographic information, International Classification of Diseases, Ninth Revision—Clinical Modification (ICD-9-CM) diagnosis and procedure codes, prescriptions details (amiodarone and concurrent medications), and laboratory data [liver biochemistry values: serum ALT, T-bil and, if available, aspartate aminotransferase (AST), direct bilirubin, and alkaline phosphatase (ALP) levels during baseline and amiodarone use] were all recorded (as listed in Table 1). Prescription details including the drug names, formulations, strength or concentrations, dosages, frequencies or rates, routes of administration, and dates of initiation and discontinuation were also noted. The cumulative dose of amiodarone for each patient was also calculated. A factor of 0.5 was applied for the oral-to-intravenous (IV) amiodarone dosage conversion (equivalent dosage) to account for oral bioavailability (approximately 0.5).[

Table 1

Patient demographics and univariate analysis for amiodarone-associated liver injury.

Definition of amiodarone-associated liver injury

Patients who developed liver function abnormalities without other apparent causes during amiodarone therapy or within 7 days after discontinuation were regarded as amiodarone-related liver injury cases. Liver function abnormality was defined as ALT levels 3 or more times the upper limit of normal (ULN) (or > 3 × pretreatment baseline if the baseline level was slightly higher than reference range, i.e., 1 × ULN < baseline level ≤ 3 × ULN).[ During amiodarone therapy, the first date the patient presented with elevated ALT levels as defined above was deemed the event date. Cases of amiodarone-associated liver injury and nonliver injury controls were followed-up for 3 months from the event date or until the date of last amiodarone use, respectively.

Comedications of hepatotoxic or amiodarone-interacting potential

Comedications taken concurrently were also recorded for further assessment. The potentially hepatotoxic agents, described as comedications of hepatotoxic potential or medications with DILI potential, examined in the study included certain nonsteroidal anti-inflammatory drugs, anti-infectives, anticonvulsants, antipsychotics, antidepressants, angiotensin-converting enzyme inhibitors, furosemide, antilipemic agents, antidiabetic agents, antineoplastic and immunomodulatory agents, and acetaminophen.[

The DDI list of moderate or major severity of concern to amiodarone included anticoagulants (warfarin, dabigatran, and rivaroxaban), antiarrhythmics (diltiazem and quinidine), antihypertensives (atenolol, betaxolol, bisoprolol, carvedilol, esmolol, labetalol, and metoprolol), digoxin, statins (HMG-CoA reductase inhibitors: atorvastatin, fluvastatin, rosuvastatin, and simvastatin), anti-infectives (azithromycin, clarithromycin, erythromycin, levofloxacin, moxifloxacin, metronidazole, rifabutin, atazanavir, darunavir, lopinavir, and ritonavir), analgesics (fentanyl and lidocaine), clonazepam, ziprasidone, zolpidem, theophylline, cyclosporine, sirolimus, methotrexate, and iohexol.[

Statistical analysis

The data are described as mean ± standard deviation (SD) for continuous variables and as frequencies (percentage, %) for categorical variables; for their comparison, we used the Wilcoxon rank-sum test and Fisher's exact test, respectively.

Cox's proportional hazards (PH) model with time-dependent covariates (called the “Cox's model”) was used to estimate the hazard ratio (HR) and 95% confidence interval (CI) for each independent covariate in predicting the risk of amiodarone-related acute liver injury. The formulations (IV and oral) and dosages of amiodarone were taken into account in the PH models. Time-dependent covariates included daily doses (on the event day), the 3-day dosage (an aggregate dosage from the prior 72 hours till the DILI event), the cumulative dose (cDose, the total dosage from therapy initiation until the event date), and the body surface area (BSA)-standardized dose of either the daily, 3-day, or cDose.

The following categories of covariates were considered as candidates in the stepwise variable selection procedure for this regression model: patient characteristics (11 items; Table 1), comorbidity (10 items; Table 1), concomitant therapy (4 covariates; medications with DILI or amiodarone-interacting potential, Table 1), and amiodarone use (1 item) plus covariates for depicting daily/3-day/cDose amiodarone dosages (12 items; time-dependent variables, described above). Generalized additive models were used to detect the nonlinear effects of continuous covariates when appropriate.

All analyses were performed using the R statistical software (version 3.2.0, R Foundation for Statistical Computing, Vienna, Austria).[ Significant differences were assumed when the P was ≤ .05.

Results

Of the 6572 patients receiving a total of 46,402 amiodarone prescriptions, amiodarone therapy was initiated in 623 individuals (9.5%) in cardiac and surgical ICUs. The study cohort of 131 patients (21.0%), new amiodarone users and had baseline and follow-up liver biochemistry results, ultimately met the inclusion criteria for further analysis. The percentage of males, mean age, ALT level, T-bil level, and number of comorbidities were 67.9%, 67.3 ± 12.5 years, 20.8 ± 10.0 U/L, 0.65 ± 0.20 mg/dL, and 6.1 ± 3.4 diagnoses, respectively (Table 1). The proportions of comedicating with DILI (84.0%) or clinically significant amiodarone-interacting (61.8%) potential were high. The mean numbers (± SD) of DILI and DDI comedications were 2.5 ± 1.7 and 1.0 ± 1.1 drugs, respectively.

Six subjects exhibited ALT levels > 3 times the ULN (n = 4) or the pretreatment baseline (n = 2; 1 × ULN < baseline levels < 3 × ULN) during amiodarone therapy. Five of 6 hepatic injury cases received initial amiodarone administration intravenously. Each of the 6 cases of amiodarone-associated liver injury (100%) were co-prescribed 2 or more DILI potential medications, and 5 (83.3%) had amiodarone-interacting comedications. Ampicillin/sulbactam (IV) and acetaminophen were the most commonly observed DILI coprescriptions. The most frequently encountered DDI comedications were fentanyl (IV) and carvedilol.

The univariate analysis of covariates related to patient characteristics and comorbidity showed no significant difference between liver injury cases and non-injury controls (Table 1). The difference in the baseline ALT levels between the 2 groups was also insignificant (30.5 ± 26.3 vs 20.4 ± 8.7, P = .77; Table 1). However, the ALT levels of the liver injury cases ranged from 160 to 637 U/L on the event dates. The length of ICU stays (31.7 ± 37.5 vs 7.0 ± 6.8, P = .003) and the length of hospital stays (76.0 ± 66.3 vs 19.3 ± 14.4, P < .001) for DILI cases were also considerably prolonged compared to those of the controls (Table 2). Most interestingly, the number of comedications consisting of either DILI potential drugs (3.8 ± 0.8 vs 2.5 ± 1.7, P = .03) or amiodarone-interacting agents (2.7 ± 2.0 vs 0.9 ± 0.9, P = .02) was notably higher in the DILI case group.

Table 2

Clinical status and outcomes of the study cohort.

The latency period, from the start of therapy to the onset of laboratory evidence of liver injury, of the case cohort ranged from 1 to 10 days (mean, 4.3 ± 3.4 days). In contrast to this apparently acute hepatocellular injury effect, the time-to-recovery was relatively lengthy (up to 3 months; mean, 45.6 ± 32.1 days) (Table 2). Multivariate analyses using Cox's PH model demonstrated that the number of comedications with DDI potential (HR, 2.07; 95% CI, 1.02 to 4.22; P = .04) and the BSA-standardized cumulative dose (cDose/BSA, g/m2; HR, 6.82; 95% CI, 1.72 to 27.04; P = .01) were independently associated with risk of amiodarone-associated liver injury (Table 3).

Table 3

The multivariate analysis of factors associated with amiodarone-associated liver injury by Cox's proportional hazards model.

Discussion

This study incorporated comedications of DDI and DILI potentials and amiodarone doses (daily, 3-day, cumulative) as essential covariates to delineate the risks and clinical significance of amiodarone-associated liver injury in ICU settings. Within this critically ill, high comorbidity, and frequent DILI or DDI co-prescription cohort, the incidence of amiodarone-associated liver injury was prudently estimated to be 4.6%. All of the 6 liver-injury cases were detected while still on amiodarone therapy, mainly intravenously administered, with latency periods of ≤ 10 days. Patients developed liver injury during amiodarone use had significantly longer lengths of ICU and hospital stays even though their baseline characteristics were comparable to those of controls (Table 1).

Although seemingly justified by instinct, explicit evidence of drug-specific factors for DILI is limited.[ Two drug-related risks, namely the number of DDI comedications and the BSA-standardized cumulative dose were identified for amiodarone-associated liver injury by a time-dependent multivariate analysis in our study. Amiodarone is unique in its physicochemical (octanol-water partition coefficient, log P = 7.2; amphiphilic) and pharmacokinetic (t1/2 ≒ 58 days) properties.[ With this composite nature of high lipophilicity and long half-life, it is of interest that the BSA-standardized cumulative dose of amiodarone can predict risk of liver injury; however, the daily dosage and the cumulative dose alone do not have such predictive power.

This study of amiodarone use in critical settings is the first to indicate that interacting comedications should not be overlooked in evaluating complex comedication regimens or complicated cases. Whether amiodarone per se, hypoxia, or co-solvent (parenteral formulation) is causative of acute hepatotoxicity remains undetermined.[ Nonetheless, the amiodarone-related risk factors disclosed herein may imply that the physiochemical nature of the compound itself plays a role in invoking DILI. Moreover, one of the 6 patients did not receive IV amiodarone, indicating that a co-solvent in parenteral formulations may not be a requisite. Our finding coincides with an early report that drug content (concentration) is also critical with regard to amiodarone-induced hepatic phospholipidosis.[

In an examination of a dataset including 343 medications, Chen et al[ recently demonstrated that oral drugs with a high lipophilicity (logP ≥ 3) and a high usual dosage (≥ 100 mg/day) were prone to DILI hazard. Our identification of the BSA-standardized cumulative dose as a predictor for amiodarone-associated liver injury, using the unique approach of analyzing patient medical data, not only underscores Chen et al.'s statement but also indicates the importance of the cumulative dose for drugs with unusually long half-lives and the necessity of taking the patient's stature (e.g., BSA) into consideration when making clinical judgments.

This study is limited by the retrospective nature, sample size, confined setting (critically ill), and single institution (tertiary care). Few liver injury cases were observed in the study. However, the goodness-of-fit measure concordance (0.90, se = 0.12; Table 3) of the final Cox's model in our study suggested an acceptable level of discrimination power. Although the enlisted medications with DILI potential were summarized from reliable reference resources and decade-long registries,[ a more restrictive definition might have been employed because certain medications were with little DILI evidence. Interestingly, the DILI-comedication covariate was not presented as a significant factor for predicting amiodarone-related liver injury despite a more inclusive definition was used in this study. Undoubtedly, further clinical and fundamental studies are required to delineate the exact mechanisms underlying amiodarone-associated acute hepatotoxicity.

Conclusions

Precision medicine, achieved by networking layers of information from scientific research, clinical practice, social or behavioral studies, and observations about optimal health care, is now on the horizon.[ This study found that drug-related covariates such as physicochemical properties, interaction profiles, and the BSA-standardized cumulative dose are all important factors when deliberating amiodarone-associated liver injury at the individual patient care level. Factors pertaining to medications are especially of concern in critical care settings where high comorbidity and frequent polypharmacy are common. We sincerely recommend that drug factors, in addition to well-recognized biomedical (molecular, genomic, cellular, physiological, and clinical), behavioral, and environmental factors, should be taken into consideration to make the utmost of precision medicine in the decades ahead.

Acknowledgments

The authors are grateful to Dr Fu-Chang Hu, MS, ScD, for his invaluable statistical discussions. We thank the NTU Hospital Information Technology Office and the Department of Pharmacy for providing relevant electronic medical records and dispensing data.

Author contributions

Conceptualization: Yunn-Fang Ho, Hsin-Ying Chou, Jan-Show Chu, Ping-Ing Lee.

Data curation: Yunn-Fang Ho, Hsin-Ying Chou.

Formal analysis: Yunn-Fang Ho, Hsin-Ying Chou, Ping-Ing Lee.

Funding acquisition: Yunn-Fang Ho.

Investigation: Yunn-Fang Ho, Hsin-Ying Chou, Ping-Ing Lee.

Methodology: Yunn-Fang Ho, Ping-Ing Lee.

Resources: Jan-Show Chu, Ping-Ing Lee.

Supervision: Ping-Ing Lee.

Validation: Yunn-Fang Ho.

Writing – original draft: Yunn-Fang Ho, Hsin-Ying Chou.

Writing – review & editing: Yunn-Fang Ho, Jan-Show Chu, Ping-Ing Lee.