Issues Surrounding Identification of Drug-Induced Liver Injury

Issues Surrounding Identification of Drug-Induced Liver Injury January 21, 2003 Mark Avigan, MD CM Deputy Director, Division of Drug Risk Evaluation Office of Drug Safety CDER, FDA Playbill ‘Much ado about nothing’ or ‘Macbeth’ ...not ‘As you like it’ Idiosyncratic Serious Liver Injury (ISLI) Characteristics of ‘Signals’ • Often detected in clinical trial subjects prior to drug approval; (Rxed individuals, n ~ 3,000) • Often generated by individuals with clinically mild/reversible drug-induced liver injury • Detected ‘signal’ may suggest but does not prove an association with ISLI (frequency > 1/50,000). Probability of true association is linked to – signal ‘strength’ (a function of both severity and frequency of drug-induced liver abnormalities or injury in the tested population) – level of ‘noise’ (similar abnormalities due to unrelated causes) Elements Which Influence ‘Signal’ Strength • Profile of abnormalities (after common etiologies of newly apparent liver injury have been ruled out) –combinations and levels of abnormalities (e.g. serum ALT, bilirubin, Alk-P, etc.) –frequency/distribution of abnormalities –presence/absence of subjects at risk for DILI –treatment with drugs that may interact •Test subject profile •Protocol of drug administration –dose –duration/frequency of treatment Relationship of Signal to ISLI in Postapproval Treatment Population –Levels of appropriate/inappropriate drug usage –Range of duration of drug exposure –Proportion of treated subjects at risk for DILI –Access to health care resources –Frequency/intensity of medical follow-up Signal Strength Hierarchy • ISLI + ALF • Clinical Drug-induced Hepatitis • Direct bilirubin > 1.5 X ULN; ALT > 3 X ULN • ALT > 8 X ULN • ALT > 3 X ULN • ALT > 1.5 X ULN Conundrums • There are no currently validated ‘biomarkers’ 100% specific for ISLI (that specifically tag bad drugs, toxic mechanisms or vunerable patients) • ISLI association with ‘weak’ signals can only be ruled out empirically by large-scale exposure to the drug in question with adequate surveillance • Absence of a ‘strong’ signal prior to drug approval does not guarantee that ISLI will not become apparent during large-scale exposure Responding to Uncertainty Whether Drug Causes ISLI/ALF Issues to consider: • Probability that there is no association with ISLI (based on signal profile) •Treatment benefit & alternative Rx options • Worst case scenario (range of ISLI incidence/severity that may occur) • Practicality/effectiveness of monitoring large numbers of patients Factors which Influence Interpretation of a Signal • Mechanism(s) of injury – hepatocellular necrosis – hepatocellular degeneration/apoptosis – injury of hepatocyte canalicular membranes or biliary epithelium – selective toxicity of specific organelles (e.g. mitochondria) – induction of collagen synthesis and fibrosis •Range of timing, tempo and reversibility of toxicity How have signals ‘played out’? Examples Tacrine Simvastatin Bromfenac Troglitazone Tacrine Hydrochloride • Cholinesterase inhibitor for Rx of Alzheimer’s • Clinical Studies revealed: –50% subjects developed ALT increases; Median time to onset - 6 wks –25% ALT > 3 X ULN; 7% ALT > 10 X ULN •ALT increases managed by protocol of LFT monitoring and dose reduction; permanent drug withdrawal if ALT > 5 X ULN •Despite biopsy evidence of hepatocellular necrosis, jaundice or clinically serious liver injury is very rare Simvastatin • Statin for Rx of hypercholesterolemia linked to increased risk for atherosclerosis-related events •5% treated subjects - AST > 2 X ULN •Long-term Rx studies –Include the ‘4S’ Study - Median F/U 5.4 yrs; Simvastatin 20 - 40 mg daily: n = 2,221 –ALT > 3X ULN - 0.7% vs 0.6% placebo (more than once); No ALT changes after 6 months of constant Rx –No association of ALT elevations with jaundice/hepatitis –Recommended monitoring up to 1 yr (if normal ALT) Bromfenac Sodium • Approved 7/97 for short term use but withdrawn 6/98 due to reports of liver failure • Single dose (n=1000) and 1 week (n=500) Rx studies –no safety problem observed; 0.4% ALT > 3 X ULN • Chronic OA and RA studies (n = 830) –2.8% ALT > 3 X ULN; 2.3% 3 X - 8 X ULN; 0.5% > 8 X ULN; dose-related effects, majority within 90 days after RX began & reversible upon D/C; jaundice not noted. • Rx recommended < 10 days and LFT monitoring if > 4 weeks • PM liver failure cases - most Rxed 2-8 months Troglitazone •Approved 1/97 for Rx of Type II diabetes; Withdrawn 3/00 in US due to cases of liver failure •US Clinical studies of Troglitazone Rx (n=2510) – 1.9% ALT > 3 X ULN vs 0.6% placebo; – 1.7% ALT > 5 X ULN; 0.2% ALT > 30 X (n=5; 2/5 also developed jaundice) – Median onset of ALT rise: 4 months; range 1-8 months • In the NIH Diabetes Prevention Trial (n = 585) –3.0% ALT > 3 X ULN; 1 patient developed ALF 94 days after Rx began (Day 57: ALT ~ 2.2 X ULN) •Thru 6/98 - 24 cases of liver failure What Can We Learn (1)? • ‘Hy’s law is strongly supported – index cases of ALT/AST elevations (> 3X ULN) with jaundice portend cases of ISLI/ALF (caveats include absence of biliary obstruction, other etiologies of injury) – among cases with jaundice, progression to liver failure and/or death is approximately 10% - 15%. • Because ISLI/ALF is rare, prior to approval when ALT/AST elevations with no increases of bilirubin occur, drug association with severe clinical outcomes often cannot be ruled in/out. • The range of levels or incidence of ALT/AST elevations, alone, do not always predict association with ISLI or ALF. What Can We Learn (2)? • The relationship of onset of liver injury with duration of drug exposure is not predictable. Over time of exposure, the hazard may be ‘front-loaded’, constant or only develop after a threshhold period. •There are no measures that predict when injury will resolve or progress if hepatotoxic drugs are not withdrawn. •The range of tempos of injury is a characteristic both of individual drugs and patients. Rapid acceleration of liver injury in some individuals may preclude an absolute protective value of standardized periodic LFT monitoring. Increased Incidence of Elevated ALT/AST! What next? • Determination if drug causes/does not cause liver injuries with elevated bilirubin, clinical hepatitis or ISLI is important. • Exclusion of ‘signal’ with both elevated ALT/AST and bilirubin requires a robust safety database – adequate powering of data sets (empirically ~1/10 cases may lead to ALF or death) – adequate clinical assessment of injury cases to enhance attribution of causality – duration of treatment and LFT monitoring should be sufficient to enable reasonable time-dependent hazard analysis • Adequately powered ‘net’ to directly measure ISLI, ALF, death (rate < 1/10,000) often requires PM surveillance or epidemiologic studies Post-marketing Surveillance Spontaneous Reporting (AERS) • Voluntary / Passive surveillance (large net) • No preconceived hypothesis of toxicity • Provides resource for detection of ALF/ISLI signal during drug’s marketing phase with potential for F/U by FDA • Does not provide denominator of precise universal usage • Denominator of drug usage can be derived from separate longitudinal drug utilization databases and enables an exposure-time hazard analysis • Limitations include: – Poor consistency in reporting quality (correct dx?, incomplete info) – Under-reporting may be profound and fluctuates during life of drug (cannot be generalized to different phases of drug or to other drugs) – Absence of disease/demographic matched control group Attribution of Causality to Drug(s) in Reports of Hepatotoxicty(AERS) • Rules of differential dx are no different than in premarketing studies or at bed-side •Analysis of causality requires informative reports –Accuracy of attribution is enhanced by • use of consistent criteria (e.g. CIOMS, CDS or M&V scales) • proactively pursuing patient info including medical records • Absence of adequate info/patient histories is major stumbling block. Lack of critical info does not imply lack of causality! • Presence of underlying liver disease may cause confusion. Epidemiological Studies Resources include: – Administrative claims databases – Prescription Event Monitoring (PEM) – Registries based on exposure or clinical events Administrative Claims Databases • Access to large blocks of patients with linkages between prescription dispensing & coded medical interventions / dxes • Hepatotoxicity event rates/Rxed patients; (comparison with event rates in similar non-exposed patients possible). • Significant limitations include: – recording of events may not be reliable – non-randomized cohorts – information not obtained in ‘real time’ – heterogeneous duration of Rx times; hepatotoxicity risk may vary – possible concomitant Rx with other hepatotoxic drugs – databases may be too small to detect very rare events – separate analysis for each drug ALF Study Group (Active Surveillance Project; PI -W. M. Lee) • Began 1998 with data collected from over 300 ALF cases • Recently, 17 sites • 13% cases drug-induced ALF (not acetaminophen) • Potential for web based real-time reporting • Enriched population for all ALF-inducing drugs • Potential useful monitor for ‘bad actors’ –Bromfenac and Troglitazone were linked to 4/28 druginduced ALF; 1998-2000 (higher n for both than any other drug) How Can Drug-Induced ISLI/ALF Monitoring be Improved? • Improvement of reporting procedures – solicit support of medical societies, academia, the pharmaceutical industry, health care organizations and patient groups – physician education – incentives for reporting – web-based modules that seek critical info • Expansion of active surveillance sites • Support research that is targeted to identify basic mechanisms of liver injury and patient susceptibility