What is drug-induced liver injury (DILI)?

Drug-induced liver injury (DILI) is hepatotoxicity caused by pharmaceutical agents, herbal medicines, dietary supplements, or other xenobiotics. It is the leading cause of acute liver failure in Western countries and the most common reason for post-market drug withdrawal. DILI is classified as intrinsic (predictable, dose-dependent — e.g. paracetamol/acetaminophen in overdose) or idiosyncratic (unpredictable, dose-independent, immune or metabolic mechanism — e.g. flucloxacillin, amoxicillin-clavulanate, isoniazid). Most clinically significant DILI is idiosyncratic.

What is the R-ratio and how is it used to classify DILI?

The R-ratio classifies DILI by injury pattern using liver biochemistry. R = (ALT/ULN) ÷ (ALP/ULN). R ≥5 = hepatocellular injury pattern (predominantly ALT elevation); R ≤2 = cholestatic injury pattern (predominantly ALP elevation); 2 < R < 5 = mixed pattern. This classification informs differential diagnosis and drug causality assessment — some drugs preferentially cause hepatocellular injury (isoniazid, paracetamol, statins, NSAIDs) while others cause cholestatic injury (erythromycin, amoxicillin-clavulanate, flucloxacillin, chlorpromazine).

What is Hy's Law in DILI?

Hy's Law identifies serious drug-induced hepatocellular injury likely to result in fatal liver failure or transplant. The criteria are: ALT or AST ≥3× upper limit of normal (ULN) PLUS total bilirubin ≥2× ULN, in the absence of other causes of elevated bilirubin (haemolysis, biliary obstruction). Hy's Law signals hepatocellular jaundice — approximately 1 in 10 patients meeting Hy's Law criteria will develop fatal DILI or require liver transplant (the Rule of 2's). It is a regulatory signal threshold required to be tracked in all clinical trials.

What is the RUCAM causality assessment method?

RUCAM (Roussel Uclaf Causality Assessment Method), updated as the CIOMS/RUCAM scale, is the standard structured tool for assessing the likelihood that a specific drug caused a DILI episode. It scores seven domains on a points-based scale: time to onset (compatible with drug exposure), course of reaction (improvement after drug withdrawal — dechallenge), risk factors (age, alcohol, pregnancy), concomitant medications, non-drug causes excluded, published literature, and rechallenge. Total score: >8 = highly probable; 6–8 = probable; 3–5 = possible; 1–2 = unlikely; ≤0 = excluded. RUCAM scores 1–5 (possible) are the most common in clinical practice, reflecting the inherent difficulty of excluding other causes of liver disease.

Guide

Drug-Induced Liver Injury (DILI)

A clinical guide to drug-induced hepatotoxicity — classification by injury pattern (R-ratio), causality assessment (RUCAM), Hy's Law severity criteria, high-risk drug classes, novel biomarkers (miR-122, GLDH, CK-18), and clinical management of DILI episodes.

Classification: intrinsic vs idiosyncratic DILI

DILI is classified into two mechanistic categories with distinct clinical implications:

Intrinsic (predictable) DILI

Dose-dependent hepatotoxicity that occurs in essentially all individuals given a sufficiently high dose. The classic example is paracetamol (acetaminophen) overdose — toxic reactive metabolite NAPQI accumulates when hepatic glutathione is depleted at doses above ~150 mg/kg. Intrinsic DILI has a predictable onset (hours to days), is dose-proportional, and can be anticipated from preclinical toxicology. Management: N-acetylcysteine (NAC) to replenish glutathione; dose guided by Rumack-Matthew nomogram. Carbon tetrachloride, methotrexate (chronic), and amiodarone are other intrinsic hepatotoxins.

Idiosyncratic DILI

Unpredictable, dose-independent hepatotoxicity affecting a small fraction of exposed individuals (1:10,000 to 1:100,000). Mechanisms are immune-mediated (hapten formation → T-cell response, HLA-dependent — e.g. flucloxacillin and HLA-B*57:01; amoxicillin-clavulanate and HLA-A*02:01) or metabolic (idiosyncratic mitochondrial dysfunction, reactive metabolite formation in susceptible individuals). Onset is variable (days to months after starting drug) and not predictable from dose. The Drug-Induced Liver Injury Network (DILIN) ↗ prospective registry is the leading source of clinical data on idiosyncratic DILI.

DILI is also classified by injury type (hepatocellular, cholestatic, mixed), severity (mild, moderate, severe, acute liver failure), and outcome (self-limited, chronic [>6 months], fatal). The LiverTox database ↗ (NIH/NCBI) is the definitive reference for individual drug DILI profiles and likelihood scoring.

R-ratio: classifying injury pattern

The R-ratio uses serum biochemistry to classify the dominant injury pattern, which informs differential diagnosis and causality:

R = (ALT / ALT_ULN) ÷ (ALP / ALP_ULN)

R value	Injury pattern	Typical drug causes	Differential diagnoses
R ≥ 5	Hepatocellular	Paracetamol, isoniazid, statins, NSAIDs, ketoconazole, nitrofurantoin (chronic)	Viral hepatitis (A, B, C, E), ischaemic hepatitis, autoimmune hepatitis
2 < R < 5	Mixed	Amoxicillin-clavulanate, carbamazepine, azathioprine, sulfonamides	Primary biliary cholangitis, overlap syndrome
R ≤ 2	Cholestatic	Flucloxacillin, erythromycin, chlorpromazine, anabolic steroids, OCP, droperidol	Primary sclerosing cholangitis, biliary obstruction (stone, tumour), PSC

The R-ratio should be calculated using the peak values of ALT and ALP during the episode, not the values at first presentation. A drug causing hepatocellular injury on initiation but cholestatic injury later in the course (or vice versa) may warrant re-calculation. ULN values for ALT and ALP should be laboratory-specific and sex-adjusted where applicable (ALT ULN differs between males and females).

Hy's Law: identifying serious DILI

Hy's Law is a regulatory benchmark for serious drug-induced hepatocellular injury, named after Dr Hyman Zimmerman who observed that hepatocellular jaundice caused by drugs carries approximately 10% mortality:

Hy's Law criteria (all three required):

ALT or AST ≥ 3× ULN
Total bilirubin ≥ 2× ULN
No other explanation for bilirubin elevation (haemolysis, Gilbert syndrome, biliary obstruction excluded)

Estimated mortality in Hy's Law cases: ~10% (the "Rule of 2's" — 1 in 10 will die or need transplant). FDA requires Hy's Law cases to be tracked and reported in all clinical trials as serious suspected adverse reactions regardless of causal attribution.

Temple's Corollary extends Hy's Law: if a drug causes ≥1 Hy's Law case in clinical development, the drug is likely capable of causing fatal DILI in a broader post-market population. The background rate of Hy's Law cases from any cause in a healthy trial population is approximately 1 per 10,000 person-years — any excess above this baseline is a signal. See the FDA DILI Guidance (2009) ↗ and FDA DILI biomarker guidance (2026) for current regulatory expectations.

RUCAM causality assessment

The RUCAM (Roussel Uclaf Causality Assessment Method) scale is the most widely used structured tool for assessing whether a specific drug caused a DILI episode. It scores seven domains:

Domain	Points	Key criteria
1. Time to onset	+1 to +2	Compatible with known DILI latency for this drug class; first exposure vs rechallenge
2. Course of reaction	+1 to +3	Improvement ≥50% within 30 days of stopping drug (dechallenge); positive rechallenge (+3)
3. Risk factors	0 to +2	Alcohol use (hepatocellular DILI); age ≥55 (cholestatic DILI); pregnancy
4. Concomitant drugs	−1 to −3	Another drug with known hepatotoxicity that has compatible time to onset
5. Non-drug causes excluded	−2 to +2	Exclusion of viral hepatitis, biliary disease, ischaemia, Wilson's disease, autoimmune hepatitis
6. Previous published reports	0 or +1	Known hepatotoxin in the literature or product information
7. Response to inadvertent rechallenge	0 or +3	ALT doubling on unintentional re-exposure; very high specificity for causality

Score interpretation: >8 = highly probable; 6–8 = probable; 3–5 = possible; 1–2 = unlikely; ≤0 = excluded. RUCAM is most reliable when applied prospectively with systematic data collection. A major limitation is that domain 5 (non-drug causes excluded) requires a clinical workup that may not be complete at the time of assessment. The updated CIOMS 2022 RUCAM revision improved inter-rater reliability with clearer scoring definitions. Online calculators are available at the RUCAM online calculator ↗.

High-risk drug classes for DILI

The DILIrank database ↗ classifies FDA-approved drugs into four DILI concern levels (most concern, less concern, no concern, ambiguous) based on postmarketing case reports and clinical trial data. Drug classes with the highest DILI signal burden include:

Antibiotics

Amoxicillin-clavulanate — most common cause of cholestatic DILI in Western countries; HLA-A*02:01 association; delayed onset (1–6 weeks post-course)
Flucloxacillin — cholestatic DILI, HLA-B*57:01 association; high spontaneous resolution rate but can persist for months
Isoniazid — hepatocellular DILI in 0.1–1% of treated patients; risk increases with age and slow acetylator CYP2E1/NAT2 genotype
Nitrofurantoin (chronic) — chronic hepatitis/cirrhosis with long-term prophylactic use; distinct from acute DILI

Statins

Symptomatic hepatotoxicity rare (<0.001%); ALT elevation common but usually adaptive
Atorvastatin and rosuvastatin most frequently implicated in DILI reports (highest market share)
ALT monitoring not routinely recommended (2013 FDA label update) unless hepatic disease or symptoms
See our statins guide

NSAIDs

Sulindac and diclofenac have highest DILI rates among NSAIDs
Diclofenac: hepatocellular injury, reactive acyl glucuronide metabolite
Ibuprofen: low DILI rate; cholestatic pattern when it occurs
Nimesulide withdrawn in several countries due to hepatotoxicity

Immunosuppressants & biologics

Methotrexate: dose-cumulative hepatic fibrosis/cirrhosis (intrinsic); liver biopsy or FibroScan monitoring in rheumatology patients
Azathioprine: hepatocellular or cholestatic; TPMT/NUDT15 pharmacogenomics relevant (see PGx guide)
Checkpoint inhibitors (pembrolizumab, nivolumab): immune-mediated hepatitis; Grade 3–4 requiring corticosteroid immunosuppression in 1–5%

Anticonvulsants

Valproate: mitochondrial hepatotoxicity in children <2 years on polypharmacy; fatal in rare cases
Phenytoin: DRESS-associated hepatitis (HLA-B*15:02 in Southeast Asian patients — see drug hypersensitivity guide)
Carbamazepine: mixed hepatocellular/cholestatic; also DRESS association

Herbal/dietary supplements

Second leading cause of DILI in the US after conventional drugs
Kava kava (Piper methysticum): hepatocellular injury; withdrawn in several countries
Green tea extract (EGCG): hepatocellular DILI in weight-loss supplements
Pyrrolizidine alkaloids (comfrey, borage oil): veno-occlusive disease
FDA MedWatch accepts DILI reports for supplements

Novel DILI biomarkers: miR-122, GLDH, and CK-18

Traditional DILI monitoring uses ALT, AST, ALP, GGT, and total bilirubin. These tests have significant limitations: ALT is non-specific (elevated by muscle disease, haemolysis, intense exercise), lags hepatocellular injury by 24–48 hours, and cannot distinguish hepatocellular necrosis from adaptive CYP induction. The 2026 FDA guidance recommends three novel biomarkers alongside ALT for clinical trials involving hepatotoxic drug candidates:

miR-122 (microRNA-122)

Mechanism: Liver-specific microRNA (~70% of total liver miRNA); released within 2–6 hours of hepatocellular damage.

Advantage over ALT: Earlier detection (before ALT rises), faster normalisation on recovery, not elevated by muscle or haemolysis. Highly liver-specific.

Assay: RT-qPCR or NGS on plasma/serum; requires careful pre-analytical handling (haemolysis elevates miR-122 falsely — centrifuge promptly, avoid freeze-thaw).

Clinical evidence: DILIN cohort, acetaminophen overdose studies — miR-122 outperforms ALT for early injury detection and has superior AUC-ROC for ALF prediction.

GLDH (Glutamate dehydrogenase)

Mechanism: Mitochondrial matrix enzyme; released from hepatocytes with severe mitochondrial injury or necrosis. Not present in muscle or erythrocytes.

Advantage over ALT: Differentiates true hepatocellular necrosis (GLDH elevated) from adaptive CYP induction (GLDH normal, ALT mildly elevated). Specific for liver and kidney tubular cells.

Clinical utility: Statin trials with adaptive ALT elevation — GLDH remains normal, supporting adaptive rather than toxic mechanism. Especially useful in trials where subjects exercise.

CK-18 (Cytokeratin-18 M30/M65)

Mechanism: Cytokeratin-18 is the major intermediate filament protein in hepatocytes. M30 fragment = caspase-cleaved CK-18 (apoptosis marker); M65 = total CK-18 (apoptosis + necrosis). M30/M65 ratio distinguishes apoptotic vs necrotic cell death.

Advantage: Mechanistic insight into DILI pathogenesis; NASH/steatohepatitis biomarker as well (M30 elevated in NASH progression). Assay: ELISA (M30-Apoptosense, Diapharma).

Clinical evidence: CK-18 M65 outperforms ALT for distinguishing severe (transplant-listed) from non-severe DILI in the DILIN cohort.

These biomarkers are not yet used in routine clinical monitoring but are increasingly incorporated into Phase 2–3 clinical trial protocols. Their qualification as surrogate endpoints for regulatory decision-making is an active area of FDA/EMA biomarker qualification work. The Pharmacovigilance World Summit 2027 includes dedicated sessions on DILI biomarker implementation.

Clinical management of DILI

The primary management of DILI is drug withdrawal. Additional considerations:

Stop the offending drug — complete dechallenge; monitor LFTs every 1–2 weeks until downtrend confirmed. Do not rechallenge idiosyncratic DILI (risk of more severe reaction).
Severity-guided response:
- Mild (ALT <5× ULN, asymptomatic, no jaundice): close monitoring, usually self-limited
- Moderate (ALT 5–10× ULN or symptoms): hepatology referral, consider RUCAM and thorough workup
- Severe (Hy's Law criteria, ALT >10× ULN): hospitalise, urgent hepatology, consider liver transplant evaluation
Paracetamol overdose specifically: N-acetylcysteine (NAC) IV or oral within 24 hours of ingestion (most effective within 8 hours); dose per Poisons Information Centre guidance ↗. NAC may have benefit in non-paracetamol DILI acute liver failure as well (evidence limited).
Immune-mediated DILI (checkpoint inhibitor hepatitis): Corticosteroids (prednisolone 0.5–1 mg/kg/day) for Grade 3–4 elevations; mycophenolate mofetil for steroid-refractory cases.
Cholestatic DILI with persistent elevation beyond 6 months: ursodeoxycholic acid (UDCA) may accelerate resolution (limited evidence); exclude primary biliary cholangitis.
Report to pharmacovigilance: All suspected DILI should be reported via FDA MedWatch ↗ (US) or the TGA Adverse Drug Reactions reporting system (Australia) and enrolled in the DILIN registry ↗ if available.

Related guides: hepatic dosing adjustments, drug metabolism in liver disease, pharmacogenomics (HLA-B*57:01, TPMT/NUDT15 testing for DILI risk reduction).

FAQ

What is drug-induced liver injury (DILI)?: Hepatotoxicity caused by pharmaceutical agents, herbal medicines, or dietary supplements. DILI is the leading cause of acute liver failure in Western countries. It is classified as intrinsic (dose-dependent, predictable — e.g. paracetamol overdose) or idiosyncratic (dose-independent, unpredictable, affecting 1:10,000 to 1:100,000 exposed individuals via immune or metabolic mechanisms).
What is the R-ratio and how does it classify DILI?: R = (ALT/ULN) ÷ (ALP/ULN). R ≥5 = hepatocellular (isoniazid, paracetamol, statins); R ≤2 = cholestatic (amoxicillin-clavulanate, flucloxacillin); 2<R<5 = mixed. Calculated using peak values. Informs differential diagnosis and drug causality likelihood.
What is Hy's Law in DILI?: ALT or AST ≥3× ULN plus total bilirubin ≥2× ULN, with other causes of bilirubin elevation excluded. Signals serious hepatocellular jaundice with ~10% mortality/transplant rate. FDA requires tracking in all clinical trials. Temple's Corollary: a single Hy's Law case in development indicates the drug can cause fatal DILI post-market.
What is the RUCAM causality assessment method?: A seven-domain points-based scale (time to onset, dechallenge, risk factors, concomitant drugs, non-drug causes excluded, published reports, rechallenge) scoring DILI causality: >8 = highly probable; 6–8 = probable; 3–5 = possible; 1–2 = unlikely. Updated CIOMS 2022 revision improves inter-rater reliability.