Drug-Induced Liver Injury and Acute Liver Failure: Drug-Induced Liver Injury and Acetaminophen

0:00Welcome to Board Pearls. This is episode one of two of the Drug-Induced Liver Injury and Acute Liver Failure chapter, in the Liver Disease module. This episode is drug injury from mechanism through acetaminophen: metabolic activation to a reactive metabolite, Hy's Law as the severity threshold, the histology patterns that map to drug class, and acetaminophen toxicity treated with N-acetylcysteine guided by the Rumack-Matthew nomogram.

0:28Start with the unifying observation: drug-induced liver injury is the diagnosis the boards reward when a patient with new abnormal liver tests is taking almost anything, because essentially any drug, herbal, or supplement can produce essentially any pattern of injury. That sounds defeatingly broad, but underneath the breadth is a mechanism that organizes most of what looks idiosyncratic: metabolic activation. The hepatic cytochromes turn a parent drug into a reactive intermediate, and phase-two conjugation either quenches that intermediate or fails to, and injury appears where phase one outpaces phase two. Hold that frame and the offender lists become predictable, the histology readable, and the severity rules non-arbitrary.

1:13The first split is between intrinsic and idiosyncratic injury. Intrinsic hepatotoxicity is dose-dependent and predictable, the prototype being acetaminophen, appearing within days of a sufficient exposure, so anyone who takes enough gets hurt. Idiosyncratic hepatotoxicity is dose-independent and unpredictable, depending on genetic variation in cytochrome metabolism and on host immune recognition of drug-modified proteins, with a window from drug start to recognition of typically one to three months, peaking around six to eight weeks. A third category, indirect hepatotoxicity, is injury from the drug's effect on the immune system or on a coexisting viral infection rather than a hepatocyte-toxic metabolite, and that's where the checkpoint inhibitors and hepatitis B reactivation after rituximab live, measured in months and caused by immune unleashing rather than direct cell killing.

2:02The epidemiology is worth keeping qualitative: population incidence is rare, but drug injury is the leading category of acute liver failure in the US registry, combining acetaminophen and non-acetaminophen injury, so it's rare on the population side and dominant on the acute liver failure side. Risk factors split into environmental and genetic. Age is bimodal, with children carrying the unique risks of aspirin-related Reye syndrome and valproate-induced fulminant injury, and older patients carrying higher idiosyncratic risk from isoniazid and a higher cholestatic phenotype in general, with isoniazid hepatitis rising markedly with age. Female sex isn't a uniform risk for injury overall, but women are overrepresented in the autoimmune-like phenotype from minocycline, methyldopa, nitrofurantoin, and diclofenac. Polypharmacy raises risk through competition for shared cytochrome clearance. Chronic alcohol induces CYP2E1 and depletes glutathione, which is the central mechanism behind acetaminophen toxicity at a therapeutic dose in the alcoholic patient, and fasting and malnutrition similarly deplete glutathione. And a prior injury from any drug raises the risk from a different drug. One nuance the boards exploit: chronic liver disease and cirrhosis don't raise the risk of injury from most agents, because impaired metabolism of the parent actually reduces toxic-metabolite production, but when a cirrhotic does develop injury the consequences are worse because it lands on a smaller reserve, so baseline cirrhosis isn't a contraindication to starting a drug but is a reason to watch more closely. Genetic risk is dominated by HLA polymorphisms that determine how the immune system recognizes drug-modified peptides, the textbook example being the allele that confers sharply increased risk of hypersensitivity hepatitis from abacavir, which is the one currently screened by mandated genotyping, so the favored stem is the patient starting an antiretroviral who develops rash, eosinophilia, and hepatitis, the answer being the abacavir hypersensitivity the screen was meant to prevent, with other HLA associations existing but not routinely screened, which is why two patients on the same antibiotic can have different outcomes.

4:11The other epidemiologic shift is that herbal and dietary supplements have become the fastest-growing category, second only to amoxicillin-clavulanate among single agents in the registry, and they split into two camps with very different prognoses. Anabolic androgenic steroids for bodybuilding produce a striking cholestatic phenotype with very high bilirubin and only mild transaminase elevation, and the cholestasis can persist for months but rarely kills or requires transplant. Non-bodybuilding herbal injury is the dangerous group, where concentrated green tea extract and kava produce hepatocellular and sometimes fulminant injury and multi-ingredient supplements produce idiosyncratic injury with an unidentifiable culprit, and this group leads to transplant far more often than prescription-drug injury, which is why the supplement history is non-negotiable. So the recognition trap on the epidemiology side is the missed history: the patient on no prescription drugs but on three supplements from a wellness store isn't low-risk, they're in the disproportionately morbid herbal subset, and the only way to know is to ask.

5:11That sets up the mechanism, because most of what looks idiosyncratic is mechanistic underneath. The hepatic two-step is the framework: phase-one reactions, dominated by the cytochrome family, oxidize the parent drug into a reactive intermediate, and phase-two reactions conjugate that intermediate to glutathione and other groups and render it water-soluble for excretion, so injury arises when phase one outpaces phase two, and the reactive intermediate then either binds covalently to hepatocyte macromolecules, the direct toxic mechanism, or modifies host proteins into neoantigens the immune system recognizes, the immunoallergic mechanism, with the same drug producing either pattern in different hosts depending on cytochrome induction, glutathione stores, and HLA type. Acetaminophen is the cleanest example. At therapeutic dose, more than ninety percent is conjugated to non-toxic metabolites, and about five to ten percent is oxidized by CYP2E1 into the reactive intermediate NAPQI, which at therapeutic dose is quietly detoxified by glutathione, so toxicity occurs at the dose where NAPQI production exceeds the rate at which the glutathione pool can quench it. So the toxic dose isn't a fixed number, it's the dose at which two host factors get out of balance: glutathione depletion lowers the threshold, and CYP2E1 induction raises NAPQI generation per gram, and chronic alcohol does both, while fasting, malnutrition, and advanced HIV deplete glutathione and isoniazid, rifampin, and the anticonvulsants induce the enzyme. That's the mechanism behind the most-tested vignette: a chronic alcohol user develops a viral illness, stops drinking because they feel awful, takes therapeutic-range acetaminophen for the fever, and presents with very high transaminases, a near-normal bilirubin, and synthetic dysfunction, never having taken an overdose, because the enzyme was induced from the alcohol, glutathione was depleted from the alcohol and the fasting, and therapeutic-range drug was enough to produce fulminant injury. The point now is that the toxic dose moved because the host changed.

7:11Isoniazid uses the same machinery for a different outcome, acetylated then oxidized to a hepatotoxic intermediate, with slow acetylators at higher risk and rifampin adding risk by inducing the enzyme, and many patients develop transaminase elevations of which most adapt while a small percentage develop clinically significant hepatitis with jaundice, amplified by older age, alcohol, and concurrent rifampin, mostly in the first three months, which is why monitoring concentrates there. Valproate is the mitochondrial example, its metabolites inhibiting beta-oxidation and depleting carnitine to produce microvesicular steatosis, with the signature being microvesicular steatosis and hyperammonemia even when transaminases are only modestly elevated, and children under two with mitochondrial enzyme deficiencies at particularly high risk, with carnitine supplementation reversing the hyperammonemic component. Methotrexate is the cumulative-fibrosis example, producing chronic fibrosis rather than acute hepatitis, dose-cumulative, monitored with periodic tests and noninvasive fibrosis assessment rather than chasing transient elevations, amplified by obesity, diabetes, alcohol, and fatty liver, with a rising transaminase an unreliable early signal so elastography has largely replaced surveillance biopsy. Statins anchor the other end of the spectrum, with a small percentage developing a transient mild transaminase elevation that adapts, severe injury rare and fulminant failure exceedingly rare, so they're safe to start in patients with stable elevated transaminases from fatty liver, discontinued only above five times normal or with jaundice. And amiodarone is the phospholipid example, accumulating in lysosomal membranes to produce a histology mimicking alcoholic liver disease in a patient who doesn't drink, with cumulative dose mattering more than the individual exposure and discontinuation not necessarily reversing it because of the long tissue half-life.

9:05The unifying picture is the cofactor logic: alcohol induces the enzyme and depletes glutathione, which sensitizes to acetaminophen, isoniazid, methotrexate, and vitamin A; fasting and malnutrition deplete glutathione; enzyme induction by the anticonvulsants, rifampin, isoniazid, and alcohol accelerates reactive-metabolite generation; and enzyme inhibition by the macrolides, azoles, and cimetidine slows elimination of the parent. The candidate who can explain in mechanistic terms why a drug became toxic in a particular host, which cytochrome and which glutathione status, outperforms the one working from offender lists.

9:45That mechanism gets converted into severity by Hy's Law, which has four components: transaminases over three times normal, total bilirubin over two times normal, alkaline phosphatase under two times normal to exclude obstruction, and no other clear cause, and when all four are met, mortality is about ten percent. Why does that combination matter? Because hepatocellular injury severe enough to disrupt bilirubin excretion has crossed a threshold of functional liver-mass loss, so the patient is no longer in the safe zone where stopping the drug guarantees recovery, since the transaminase says the cells are being hurt and the bilirubin says the residual mass can't handle a normal load, which is why the bilirubin is the severity threshold, not the transaminase. When Hy's Law progresses to acute liver failure with encephalopathy, mortality without transplant rises to roughly eighty percent, so the rule is a way to act early, before fulminant failure, on the biochemical evidence that the patient has crossed the line where stopping the drug alone may not be enough, triggering discontinuation, hospitalization, and transplant-center awareness. The R value sits next to it doing a different job, classifying the injury by pattern, hepatocellular over five, cholestatic under two, mixed between, with hepatocellular injury more likely to evolve to acute failure and carry higher acute mortality and cholestatic and mixed injury more likely to be prolonged and evolve into chronic injury, so most cases are hepatocellular with the rest split between cholestatic and mixed, and R isn't a severity score, it tells you which trajectory the patient is on. And the causality assessment method scores points across the temporal relationship, the course after withdrawal, the risk factors, competing drugs, exclusion of viral, autoimmune, and biliary causes, prior published reports, and response to rechallenge, with timing, dechallenge response, exclusion of alternatives, and known prior reports mattering more than the arithmetic.

11:47A teaching trap embedded here is the drug injury that mimics autoimmune hepatitis, a small fraction of cases being histologically and serologically indistinguishable, with positive autoantibodies, hypergammaglobulinemia, plasma-cell-rich interface hepatitis, and a response to steroids, the classic offenders being nitrofurantoin, minocycline, hydralazine, and methyldopa, and the discriminator being that drug-induced autoimmune-like hepatitis usually doesn't need long-term immunosuppression, remitting when the drug stops, while idiopathic autoimmune hepatitis flares when steroids taper. So a patient with the autoimmune phenotype who's been on nitrofurantoin for years for recurrent urinary infections forces the question of whether the immune reaction is the disease or the response to the drug, and the test is what happens when the drug stops.

12:30That brings us to histology, where pattern recognition becomes the answer, because drugs cluster by their site of metabolic activation, their biliary-versus-hepatocellular processing, and the immune phenotype they elicit, so the same handful of patterns recur across hundreds of agents. Hepatocellular injury with centrilobular necrosis is the textbook acetaminophen lesion, because zone three has the highest CYP2E1 density so the NAPQI, and the necrosis, is produced where the enzyme is concentrated, shared by isoniazid, methyldopa, and halothane. Diffuse hepatocellular injury mimicking viral hepatitis, with portal lymphocytic infiltrate and apoptotic bodies, comes from isoniazid, nitrofurantoin, allopurinol, diclofenac, and statins. Cholestatic injury with bland canalicular cholestasis and minimal inflammation is the amoxicillin-clavulanate pattern, the most common single agent, and its recognition fingerprint is striking: the cholestasis appears a few weeks after the antibiotic course is finished, so the patient took a week of the antibiotic for sinusitis, finished it, felt better, and turned jaundiced a month later, and the delay is the trap, because the drug history is "I'm not on any medication" since the course ended a month ago, so the move is to ask about exposures in the last three months, not the last week, with erythromycin, anabolic and contraceptive steroids, the azoles, and chlorpromazine other reliable cholestatic offenders. Persistent cholestatic injury can progress to vanishing bile duct syndrome, from amoxicillin-clavulanate, the fluoroquinolones, azithromycin, and several anticonvulsants, with progressive ductopenia that can look like primary biliary cholangitis except for the drug exposure and the negative antibody. Mixed injury with both portal inflammation and bile duct injury comes from sulfonamides, phenytoin, and azathioprine, with phenytoin in particular producing the drug-rash-eosinophilia hypersensitivity syndrome. Granulomatous hepatitis with non-caseating granulomas comes from allopurinol, sulfonamides, hydralazine, quinidine, phenytoin, and isoniazid, overlapping with sarcoid and infections, discriminated by the temporal relationship and resolution after withdrawal. Steatosis comes two ways: macrovesicular with large droplets from glucocorticoids, methotrexate, and tamoxifen, and microvesicular with small droplets, the signature of mitochondrial toxicity from tetracycline, valproate, the older nucleoside antiretrovirals, and salicylates in Reye syndrome, and the microvesicular pattern is the more dangerous because it reflects acute mitochondrial failure, carrying hyperammonemia and acidosis even with only modest transaminase elevation. And vascular injury is the last pattern, often the answer when the vignette emphasizes portal hypertension or Budd-Chiari without much hepatocellular injury: sinusoidal obstruction syndrome from pyrrolizidine-alkaloid herbal teas and myeloablative chemotherapy, presenting with rapid weight gain, painful hepatomegaly, jaundice, and ascites; oral contraceptives producing peliosis, Budd-Chiari, and adenomas; and the thiopurines producing nodular regenerative hyperplasia and non-cirrhotic portal hypertension, the answer in a treated HIV patient with portal hypertension and minimal fibrosis on biopsy. The mechanistic logic underneath is worth holding: centrilobular necrosis tracks zone-three cytochrome density, cholestasis tracks canalicular-transporter interference, granulomas track delayed-type hypersensitivity, microvesicular steatosis tracks mitochondrial poisoning, and vascular injury tracks endothelial damage, so the biopsy is rarely diagnostic alone but combining the pattern with the drug history and timing narrows the answer.

16:00That brings us to the prototype, acetaminophen, the leading cause of acute liver failure in the US and the UK, whose mechanism, dose, nomogram, and antidote the boards expect cold. The mechanism is the one from the cytochrome section: ninety percent conjugated, five to ten percent to NAPQI detoxified by glutathione, toxicity when NAPQI exceeds glutathione capacity so it binds covalently to mitochondrial proteins and centrilobular necrosis follows, becoming biochemically apparent at a day to three days, peaking at three to four days, then recovery or progression to failure. The dose thresholds: an acute single ingestion of seven and a half grams or more in an adult, or over a hundred fifty milligrams per kilogram in a child, puts the patient in the toxic window, with suicidal overdoses typically higher, but chronic supratherapeutic ingestion is the more treacherous scenario because the time course is stretched and patients present late, so in a chronic alcohol user or malnourished or fasting patient, sustained ingestion of four grams or even less a day for several days can produce fulminant toxicity through enzyme induction and glutathione depletion, which is the most-tested vignette. And a diagnostic anchor: transaminases over five thousand are acetaminophen in about a quarter of cases and ischemic hepatitis in most of the rest, with few alternatives.

17:17The Rumack-Matthew nomogram is the risk-stratification tool for acute single ingestion, plotting serum concentration against time on a log scale, with the treatment line drawn from a hundred fifty micrograms per milliliter at four hours descending to about five at twenty-four hours, so a level above the line within the four-to-twenty-four-hour window after a known single ingestion is an indication for N-acetylcysteine. The conditions matter: the nomogram doesn't apply to chronic supratherapeutic ingestion, to ingestions more than a day old, to extended-release formulations, or when the time of ingestion is unknown, and in all those the threshold is low and empiric treatment is the safe answer, so the nomogram is precise where it applies and silent everywhere else, and recognizing where it stops applying is most of the test. Activated charcoal is the first decontamination, given orally within a few hours in a cooperative patient with a protected airway, most effective in the first hour, and it doesn't interfere with subsequent antidote therapy. N-acetylcysteine is the antidote, working three ways, as a glutathione precursor replenishing the pool, by conjugating NAPQI directly, and through anti-inflammatory and microcirculatory benefits in the late phase that account for benefit even beyond the canonical window, so its indications are wider than the nomogram: start it when the level is above the treatment line in the window, when clinical suspicion is high regardless of level, when the ingestion was chronic supratherapeutic, when the time is unknown, when there's already transaminase elevation or coagulopathy, and empirically in any acute liver failure of indeterminate cause. The intravenous dosing is the tested set: a loading dose of a hundred fifty milligrams per kilogram over sixty minutes, given over an hour rather than faster to reduce anaphylactoid reactions, then fifty over the next four hours, then a hundred over the next sixteen, totaling three hundred over twenty-one hours, continued beyond that at a hundred per sixteen hours if there's persistent encephalopathy, coagulopathy, or transaminase elevation, until those resolve or the patient is transplanted. Time to the antidote drives mortality, which is why the empiric move beats waiting for the confirmatory level, because mortality is very low when it's started within twelve hours and climbs steadily thereafter, roughly doubling with each block of delay, so the right answer on the stem is to start it the moment the diagnosis is suspected. And the clinical phases are testable: the first day is nausea, vomiting, and malaise with normal transaminases, so patients can look entirely well, which is the danger; the next couple of days bring rising transaminases, right-upper-quadrant pain, and developing coagulopathy; days three to four are the peak, with transaminases often above ten thousand, peaking bilirubin and INR, jaundice, and encephalopathy, which in fulminant cases is acute liver failure with possible renal failure; and recovery from day four onward in survivors, with tests normalizing over a week or two and no chronic sequelae for those who don't progress.

20:03So the framework comes back together. Drug injury is the diagnosis to entertain when a patient with new abnormal liver tests is on essentially anything, including the herbal and supplement category the patient may not consider medication. The mechanism is metabolic activation, with cytochrome oxidation generating a reactive intermediate that phase two either quenches or fails to quench, and the cofactors that move the dose-injury threshold, especially alcohol and malnutrition, explain why the same drug is safe in one patient and toxic in another. Hy's Law is the severity threshold that converts biochemistry into a triage decision, because hepatocellular injury severe enough to disrupt bilirubin excretion has crossed the line where stopping the drug may not be enough. The R value tells you the biochemical pattern and predicts trajectory, and the histology patterns map to drug classes so you can read the biopsy backward to a suspect list. And acetaminophen is the prototype that runs the whole sequence at speed, with the nomogram covering acute single ingestion, empiric antidote covering everything else, and time to treatment roughly doubling mortality with every block of delay.

21:09Episode two picks up where the parenchymal injury becomes a clinical syndrome: acute liver failure defined by coagulopathy and encephalopathy under twenty-six weeks without prior cirrhosis, its etiology distribution led by acetaminophen, the two cause-specific entities with their own recognition cues and empiric therapies, Wilsonian and herpes acute liver failure, and the King's College Criteria that triage who needs to be listed for transplant.

21:37For the full chapter, the practice vignettes, and the topic-tagged question bank, head to board pearls dot com. You'll find the rest of the series on Apple Podcasts, Spotify, or wherever you listen to podcasts. That brings us to the end of episode one of two of chapter nineteen, and I'll see you in the next one.