Isoniazid Toxicity

Complex MONDO:0027677 Pathograph 28 Show in embeddings browser Drug Toxicity

Isoniazid toxicity is two different problems wearing one drug's name, and the difference is a matter of direction. The hepatic arm is about what the liver does to isoniazid: it bioactivates the drug to something reactive, and how much it makes depends on the patient's NAT2 acetylator genotype. The neurological arms are about what isoniazid does to vitamin B6. Isoniazid is a hydrazide, it reacts with pyridoxal, and the resulting functional B6 deficiency starves any enzyme that needs pyridoxal 5-phosphate. Given slowly, over months of treatment, that presents as a peripheral neuropathy - the nerve is deprived rather than poisoned. Given all at once in overdose, GABA synthesis fails for want of the same cofactor, and the patient seizes in a way ordinary anticonvulsants do not fix. The antidote is the missing vitamin, matched gram for gram to the dose swallowed.

Ask OpenScientist

Ask a research question about Isoniazid Toxicity. OpenScientist will conduct autonomous deep research using the Disorder Mechanisms Knowledge Base and PubMed literature (typically 10-30 minutes).

Submitting...

Do not include personal health information in your question. Questions and results are cached in your browser's local storage.

11
Pathophys.
10
Phenotypes
2
Hypotheses
28
Pathograph
5
Genes
5
Medical Actions
1
Deep Research
◈

Mechanistic Hypotheses

2
Acetylhydrazine bioactivation without an immune response
acetylhydrazine_metabolic_idiosyncrasy ⚠ DEPRECATED
⚠ Overturned model — shown for reference, not as current mechanism

DisMech records superseded hypotheses explicitly rather than deleting them, so that claims still circulating in reviews, textbooks and older diagnostic criteria can be checked against an assessment. This model is not part of the disease mechanism DisMech asserts.

Citation volume does not decide standing here. A hypothesis may retain more supporting than refuting citations simply because the supporting literature accumulated for decades before the refutation landed; where the two conflict, DisMech follows the more recent and more direct evidence. Supporting citations below are retained for the historical record.

Evidence balance 1 support
The long-standing account: isoniazid is acetylated to acetylhydrazine, which is oxidised to a reactive species that injures the hepatocyte, and the injury is a metabolic idiosyncrasy with no immune involvement. This is the model most textbooks still carry and it is the reason slow acetylation was originally expected to matter. It is recorded here as deprecated rather than omitted, because a curator meeting the acetylhydrazine story in a secondary source should be able to find out from this entry that it has been superseded.
Pathograph links
Hepatic Bioactivation of Isoniazid -> Hepatocyte Injury and Death DIRECT The reactive metabolite injures the hepatocyte. This edge is shared by both hypotheses, which disagree about which metabolite it is and about whether an immune response follows, not about whether the hepatocyte is injured.
Show evidence (1 reference)
PMID:26773235 SUPPORT Other
"Previous mechanistic hypotheses have classified this type of drug-induced liver injury (DILI) as 'metabolic idiosyncrasy' which was thought not to involve an immune response and was mainly due to the bioactivation of the acetylhydrazine metabolite."
States the older model in the review's own words, and marks it as previous.
Direct bioactivation of isoniazid with an immune response
direct_bioactivation_immune CANONICAL
Evidence balance 2 support
The current account: isoniazid itself is bioactivated to a reactive metabolite, and in some patients that provokes an immune response. Two phenotypes follow - a common mild injury that resolves as immune tolerance develops, and an uncommon severe one carrying anti-drug and anti-CYP antibodies that can progress to liver failure. The clinically important consequence is that resolving transaminase elevation on continued treatment is an expected outcome rather than a paradox.
Pathograph links
Hepatic Bioactivation of Isoniazid -> Hepatocyte Injury and Death DIRECT The reactive metabolite injures the hepatocyte. This edge is shared by both hypotheses, which disagree about which metabolite it is and about whether an immune response follows, not about whether the hepatocyte is injured.
Hepatic Bioactivation of Isoniazid -> Immune-Mediated Amplification of Hepatic Injury INDIRECT KNOWN INTERMEDIATES Under the current model the reactive metabolite provokes an immune response in some patients. The intermediate is hapten formation on hepatic proteins.
Immune-Mediated Amplification of Hepatic Injury -> Isoniazid-Induced Liver Injury DIRECT The immune arm determines which patients progress rather than tolerate.
Show evidence (2 references)
PMID:26773235 SUPPORT Other
"more recent studies support an alternative hypothesis, specifically, that INH itself is directly bioactivated to a reactive metabolite, which in some patients leads to an immune response and liver injury"
States the superseding model: direct bioactivation of the parent drug, with an immune component.
PMID:26773235 SUPPORT Other
"Most cases involve mild liver injury, which resolves with immune tolerance, while other cases appear to have a more severe phenotype that is associated with the production of anti-drug/anti-CYP P450 antibodies and can progress to liver failure."
Gives the two phenotypes and the tolerance mechanism behind the common one.
⚙

Pathophysiology

11
Hepatic Bioactivation of Isoniazid
Conversion of isoniazid to a reactive metabolite in the liver. Which species does the damage is the point on which the field has changed its mind, so this node carries both models as hypothesis groups rather than asserting one chemistry.
⚠ Acetylhydrazine bioactivation without an immune response Direct bioactivation of isoniazid with an immune response
Downstream
Hepatocyte Injury and Death The reactive metabolite injures the hepatocyte. This edge is shared by both hypotheses, which disagree about which metabolite it is and about whether an immune response follows, not about whether the hepatocyte is injured. ⚠ Acetylhydrazine bioactivation without an immune response Direct bioactivation of isoniazid with an immune response
Immune-Mediated Amplification of Hepatic Injury Under the current model the reactive metabolite provokes an immune response in some patients. The intermediate is hapten formation on hepatic proteins. Direct bioactivation of isoniazid with an immune response
Show evidence (2 references)
PMID:26773235 SUPPORT Other
"Isoniazid (INH) remains a mainstay for the treatment of tuberculosis despite the fact that it can cause liver failure."
Establishes the hepatic toxicity of the drug this node bioactivates.
PMID:41793109 SUPPORT Other
"Isoniazid (INH) and rifampicin (RIF) are cornerstone first-line antituberculosis drugs, yet their clinical utility is often limited by drug-induced liver injury (DILI)."
Confirms hepatotoxicity as the limiting toxicity of this drug in practice. The source treats isoniazid together with rifampicin, whose co-administration it describes as potentiating the injury, so it is not evidence for isoniazid acting alone.
NAT2 Slow Acetylator Genotype
Reduced N-acetyltransferase 2 activity, which changes how much isoniazid is routed down the acetylation pathway and how long the parent drug persists. Slow acetylators carry roughly three times the risk of anti-tuberculosis drug-induced liver injury. This is one of the better-quantified pharmacogenetic effects in clinical toxicology.
Downstream
Hepatic Bioactivation of Isoniazid Acetylator status modifies how much reactive metabolite is generated. The outcome measured in the cited meta-analysis is anti-tuberculosis drug-induced liver injury, not metabolite concentration, so the placement of the effect at this node is mechanistic inference.
Show evidence (2 references)
PMID:41657030 SUPPORT Human Clinical
"Slow acetylator genotypes were significantly associated with an increased risk of AT-DILI (pooled OR = 3.02; 95% CI = 2.50-3.64; p < 0.001)."
Gives the pooled effect size and interval for the claim in this node's description.
PMID:41657030 SUPPORT Human Clinical
"A total of 48 studies comprising 11,035 patients were included."
Gives the size of the evidence base behind that estimate.
Hepatocyte Injury and Death
Hepatocellular injury, showing first as transaminase elevation. Most of it is mild and resolves even on continued treatment, which is what the tolerance arm of the current model predicts.
Downstream
Isoniazid-Induced Liver Injury Accumulated hepatocyte injury becomes clinically apparent liver injury.
Show evidence (1 reference)
PMID:26773235 SUPPORT Other
"Most cases involve mild liver injury, which resolves with immune tolerance, while other cases appear to have a more severe phenotype that is associated with the production of anti-drug/anti-CYP P450 antibodies and can progress to liver failure."
Gives both the common mild injury and the severe minority this node covers.
Immune-Mediated Amplification of Hepatic Injury
Innate and adaptive immune responses to drug-modified hepatic proteins. In most patients this resolves into tolerance. In a minority it does not, and those are the cases carrying anti-drug and anti-CYP antibodies that progress.
Direct bioactivation of isoniazid with an immune response
Downstream
Isoniazid-Induced Liver Injury The immune arm determines which patients progress rather than tolerate. Direct bioactivation of isoniazid with an immune response
Show evidence (1 reference)
PMID:41793109 SUPPORT Other
"Additionally, we examined immune-mediated mechanisms, including hapten formation, Tool-like receptor-4 (TLR4) driven innate immune responses, and human leukocyte antigen (HLA) associated adaptive recognition, which integrated with toxic insults under the dual-hit hypothesis."
Names the innate and adaptive components of this node.
Isoniazid-Induced Liver Injury
Clinically apparent hepatitis, occasionally progressing to acute liver failure. The management threshold is biochemical: treatment is interrupted at a transaminase elevation above three times normal with symptoms or jaundice, or five times normal without.
Downstream
Elevated circulating hepatic transaminase concentration The biochemical readout, and the one that is monitored.
Jaundice Appears in more advanced injury and lowers the interruption threshold.
Decreased liver function The severe minority phenotype, progressing toward failure.
Show evidence (2 references)
PMID:17021358 SUPPORT Other
"Treatment should be interrupted and, generally, a modified or alternative regimen used for those with ALT elevation more than three times the upper limit of normal (ULN) in the presence of hepatitis symptoms and/or jaundice, or five times the ULN in the absence of symptoms."
Gives the ATS interruption thresholds quoted in this node's description.
PMID:17021358 SUPPORT Other
"Knowledge of the metabolism of anti-TB medications and of the mechanisms of TB DILI is incomplete."
The official statement's own assessment of how well this mechanism is understood, which is why this entry carries two hypothesis groups rather than one settled chain.
Functional Pyridoxine Deficiency
Depletion of usable vitamin B6. Isoniazid is a hydrazide and reacts with pyridoxal, and the result is that the patient has a deficiency of the cofactor rather than a poison in the nerve. Everything downstream of this node is an enzyme failing for want of pyridoxal 5-phosphate, which is why one node supplies two clinically unrelated syndromes on two different timescales, and why one vitamin treats both.
Downstream
Peripheral Nerve Deprivation of Pyridoxal Phosphate The slow arm. Months of treatment at ordinary doses deplete the cofactor enough to affect peripheral nerve.
Impaired GABA Synthesis The fast arm. A single large ingestion depletes the cofactor acutely, and the enzyme that fails first is the one making the brain's main inhibitory transmitter.
Show evidence (5 references)
PMID:29897291 SUPPORT Human Clinical
"INH leads to peripheral neuritis by causing a deficiency in the serum level of pyridoxine which depends on the dose of INH, duration of treatment and the patient's nutritional and acetylator status."
States that the drug produces a pyridoxine deficiency and that this is how the neuritis arises, which is the whole basis of this node.
PMID:15756083 SUPPORT Other
"Pyridoxine (vitamin B6) is a co-factor in many enzymatic pathways involved in amino acid metabolism: the main biologically active form is pyridoxal 5-phosphate."
Establishes pyridoxal 5-phosphate as the active cofactor whose loss this node describes, and that it serves many pathways - which is why one deficiency has several consequences.
PMID:15756083 SUPPORT Other
"Pyridoxine has been used as an antidote in acute intoxications, including isoniazid overdose, Gyromitra mushroom or false morrel (monomethylhydrazine) poisoning and hydrazine exposure."
Places isoniazid alongside monomethylhydrazine and hydrazine as poisonings answered by the same vitamin, which is the shared hydrazide chemistry this node rests on. The source states the antidotal use, not the chemical reaction.
+ 2 more references
Peripheral Nerve Deprivation of Pyridoxal Phosphate
Peripheral nerve deprived of the cofactor. The insult that reaches the neuron here is a nutritional one, which is unlike most of the toxic neuropathies - nothing is accumulating in the nerve, something is missing from it. Risk tracks the things that lower B6 reserve to begin with: malnutrition, alcoholism, and slow acetylation.
Downstream
Isoniazid-Induced Peripheral Neuropathy Cofactor deprivation produces the clinical neuropathy. The intervening cellular steps are not established by the sources cited here.
Show evidence (1 reference)
PMID:29897291 SUPPORT Human Clinical
"Isoniazid (INH)-induced peripheral neuritis is not uncommonly reported in adults, especially those with malnutrition and alcoholism, but it is very rare in children."
Names the nutritional risk factors this node describes, and the age distribution.
Isoniazid-Induced Peripheral Neuropathy
Distal sensory symptoms appearing over months of treatment. It is preventable and reversible: pyridoxine prophylaxis is recommended for those at risk, and symptoms resolve when the deficiency is corrected.
Downstream
Peripheral neuropathy The clinical readout of the slow B6 arm.
Show evidence (1 reference)
PMID:29897291 SUPPORT Human Clinical
"the World Health Organization recommends pyridoxine prophylaxis for children on INH who are malnourished or have HIV infection"
Establishes the prophylaxis recommendation and the risk groups it targets. The recommendation quoted is for children specifically.
Impaired GABA Synthesis
Failure to make gamma-aminobutyric acid, because the enzyme that makes it requires pyridoxal 5-phosphate. This is what turns an overdose into a seizure that ordinary anticonvulsants cannot hold: the drugs that potentiate GABA signalling need GABA to potentiate, and there is not enough of it.
Downstream
Isoniazid Overdose Seizures Loss of inhibitory transmission generates the seizures.
Show evidence (1 reference)
PMID:17535059 SUPPORT Human Clinical
"The seizures are often refractory to traditional antiepileptics."
Records the refractoriness this node explains. The source states the clinical observation; it does not itself state the GABA-synthesis mechanism, and the inference from one to the other is this entry's.
Isoniazid Overdose Seizures
Generalised seizures in acute overdose, resistant to standard treatment because the deficit is upstream of the receptors those drugs act on. Clinically they arrive as part of a triad with coma and metabolic acidosis.
Downstream
Lactic Acidosis Sustained convulsive activity is followed by a lactic acidosis. The intermediate usually given is lactate generated by the seizing muscle, and the sources cited here report the acidosis without assigning its origin, so the edge is recorded as indirect rather than as a measured production step.
Seizure The clinical readout of the fast B6 arm.
Coma Part of the overdose triad, alongside the seizures.
Nausea Early gastrointestinal feature, preceding the seizures.
Vomiting Early gastrointestinal feature, preceding the seizures.
Ataxia Early cerebellar sign, preceding the seizures.
Show evidence (3 references)
PMID:17535059 SUPPORT Human Clinical
"Isoniazid toxicity produces a triad of coma, metabolic acidosis and seizures."
Gives the triad this node is named for.
PMID:8522667 SUPPORT Other
"Patients with INH overdose may present with nausea, vomiting, ataxia, symptoms reminiscent of atropine intoxication, coma and grand mal seizures."
Gives the fuller presentation, including the early gastrointestinal and cerebellar features that precede the seizures.
PMID:8522667 SUPPORT Other
"INH overdose should be suspected in any patient presenting with seizures and metabolic acidosis."
States the diagnostic pairing, which is what makes this syndrome recognisable before an ingestion history is available.
Lactic Acidosis
Lactic acidosis in acute overdose. It is separated from the seizure node because the entry treats it as a consequence of the seizing rather than as a parallel toxicity, and because that causal claim is weaker than the seizure claim and should be visible as its own step.
Downstream
Metabolic acidosis The acidosis as measured clinically.
Show evidence (1 reference)
PMID:8522667 SUPPORT Other
"INH overdose should be suspected in any patient presenting with seizures and metabolic acidosis."
Pairs the acidosis with the seizures as the recognisable presentation, which is the diagnostic value of this node.
⬡

Pathograph

Use the checkboxes to hide or show graph categories. Hover nodes for evidence and cross-linked metadata.
Pathograph: causal mechanism network for Isoniazid Toxicity Interactive directed graph showing how pathophysiology mechanisms, phenotypes, genetic factors and variants, experimental models, environmental triggers, and treatments relate through causal and linked edges.
●

Phenotypes

10
Digestive 4
Decreased liver function HP:0001410 Human Phenotype Ontology (HP) Relation: this clinical feature is this phenotype This clinical feature is Hepatitis progressing to liver failure, annotated with Decreased liver function (HP:0001410). HP:0001410 is a phenotype from the Human Phenotype Ontology.
Show evidence (1 reference)
PMID:26773235 SUPPORT Other
"Most cases involve mild liver injury, which resolves with immune tolerance, while other cases appear to have a more severe phenotype that is associated with the production of anti-drug/anti-CYP P450 antibodies and can progress to liver failure."
Establishes progression to liver failure in a minority. No frequency band is asserted; the source distinguishes most from other without quantifying.
Vomiting HP:0002013 Human Phenotype Ontology (HP) Relation: this clinical feature is this phenotype This clinical feature is Vomiting (HP:0002013), qualified as temporality acute. HP:0002013 is a phenotype from the Human Phenotype Ontology.
Temporal: ACUTE
Show evidence (1 reference)
PMID:8522667 SUPPORT Other
"Patients with INH overdose may present with nausea, vomiting, ataxia, symptoms reminiscent of atropine intoxication, coma and grand mal seizures."
Names vomiting among the early overdose features.
Nausea HP:0002018 Human Phenotype Ontology (HP) Relation: this clinical feature is this phenotype This clinical feature is Nausea (HP:0002018), qualified as temporality acute. HP:0002018 is a phenotype from the Human Phenotype Ontology.
Temporal: ACUTE
Show evidence (1 reference)
PMID:8522667 SUPPORT Other
"Patients with INH overdose may present with nausea, vomiting, ataxia, symptoms reminiscent of atropine intoxication, coma and grand mal seizures."
Names nausea among the early overdose features.
Jaundice HP:0000952 Human Phenotype Ontology (HP) Relation: this clinical feature is this phenotype This clinical feature is Jaundice (HP:0000952). HP:0000952 is a phenotype from the Human Phenotype Ontology.
Show evidence (1 reference)
PMID:17021358 SUPPORT Other
"Treatment should be interrupted and, generally, a modified or alternative regimen used for those with ALT elevation more than three times the upper limit of normal (ULN) in the presence of hepatitis symptoms and/or jaundice, or five times the ULN in the absence of symptoms."
Names jaundice as the finding that lowers the interruption threshold, which is the role this phenotype plays.
Metabolism 2
Elevated circulating hepatic transaminase concentration HP:0002910 Human Phenotype Ontology (HP) Relation: this clinical feature is this phenotype This clinical feature is Elevated transaminases, annotated with Elevated circulating hepatic transaminase concentration (HP:0002910). HP:0002910 is a phenotype from the Human Phenotype Ontology.
Show evidence (1 reference)
PMID:17021358 SUPPORT Other
"Treatment should be interrupted and, generally, a modified or alternative regimen used for those with ALT elevation more than three times the upper limit of normal (ULN) in the presence of hepatitis symptoms and/or jaundice, or five times the ULN in the absence of symptoms."
Establishes transaminase elevation as the monitored abnormality and gives the thresholds at which it becomes actionable.
Metabolic acidosis HP:0001942 Human Phenotype Ontology (HP) Relation: this clinical feature is this phenotype This clinical feature is Metabolic acidosis (HP:0001942), qualified as temporality acute. HP:0001942 is a phenotype from the Human Phenotype Ontology.
Temporal: ACUTE
Show evidence (1 reference)
PMID:8522667 SUPPORT Other
"INH overdose should be suspected in any patient presenting with seizures and metabolic acidosis."
Pairs the acidosis with the seizures as the recognisable overdose presentation.
Nervous System 4
Peripheral neuropathy HP:0009830 Human Phenotype Ontology (HP) Relation: this clinical feature is this phenotype This clinical feature is Peripheral neuropathy (HP:0009830), qualified as temporality chronic. HP:0009830 is a phenotype from the Human Phenotype Ontology.
Temporal: CHRONIC
Show evidence (1 reference)
PMID:29897291 SUPPORT Human Clinical
"Isoniazid (INH)-induced peripheral neuritis is not uncommonly reported in adults, especially those with malnutrition and alcoholism, but it is very rare in children."
Establishes the phenotype and the populations in which it is seen. The source's terms - not uncommon in adults, very rare in children - are qualitative and age-split, so no frequency band is asserted.
Seizure HP:0001250 Human Phenotype Ontology (HP) Relation: this clinical feature is this phenotype This clinical feature is Seizure (HP:0001250), qualified as temporality acute. HP:0001250 is a phenotype from the Human Phenotype Ontology.
Temporal: ACUTE
Show evidence (2 references)
PMID:17535059 SUPPORT Human Clinical
"Isoniazid toxicity produces a triad of coma, metabolic acidosis and seizures."
Names seizures as part of the overdose triad.
PMID:17535059 SUPPORT Human Clinical
"The seizures are often refractory to traditional antiepileptics."
Establishes the refractoriness named in the description.
Coma HP:0001259 Human Phenotype Ontology (HP) Relation: this clinical feature is this phenotype This clinical feature is Coma (HP:0001259), qualified as temporality acute. HP:0001259 is a phenotype from the Human Phenotype Ontology.
Temporal: ACUTE
Show evidence (1 reference)
PMID:17535059 SUPPORT Human Clinical
"Isoniazid toxicity produces a triad of coma, metabolic acidosis and seizures."
Names coma as part of the overdose triad.
Ataxia HP:0001251 Human Phenotype Ontology (HP) Relation: this clinical feature is this phenotype This clinical feature is Ataxia (HP:0001251), qualified as temporality acute. HP:0001251 is a phenotype from the Human Phenotype Ontology.
Temporal: ACUTE
Show evidence (1 reference)
PMID:8522667 SUPPORT Other
"Patients with INH overdose may present with nausea, vomiting, ataxia, symptoms reminiscent of atropine intoxication, coma and grand mal seizures."
Names ataxia among the presenting features of overdose.
🧬

Genetic Associations

5
NAT2
Gene: NAT2 hgnc:7646 HUGO Gene Nomenclature Committee (hgnc) Relation: this disease-associated gene is this gene This disease-associated gene is NAT2 (hgnc:7646). hgnc:7646 is a gene from the HUGO Gene Nomenclature Committee. relationship_type: SUSCEPTIBILITY
Show evidence (5 references)
PMID:41657030 SUPPORT Human Clinical
"Slow acetylator genotypes were significantly associated with an increased risk of AT-DILI (pooled OR = 3.02; 95% CI = 2.50-3.64; p < 0.001)."
Gives the pooled effect size for the hepatic arm.
PMID:29897291 SUPPORT Human Clinical
"INH leads to peripheral neuritis by causing a deficiency in the serum level of pyridoxine which depends on the dose of INH, duration of treatment and the patient's nutritional and acetylator status."
Names acetylator status among the determinants of the pyridoxine deficiency, which is the basis for saying this gene modifies the neuropathy arm too. The statement is qualitative and gives no effect size for that arm.
PMID:33135175 SUPPORT Human Clinical
"For N-acetyltransferase 2 (NAT2), NAT2*5 frequency was lower in cases (OR 0.69, 95% CI 0.57-0.83, P = 0.01)."
Reports NAT2*5 as less frequent in cases, which cuts against reading the pooled slow-acetylator odds ratio as applying uniformly across slow alleles.
+ 2 more references
CYP2E1
Gene: CYP2E1 hgnc:2631 HUGO Gene Nomenclature Committee (hgnc) Relation: this disease-associated gene is this gene This disease-associated gene is CYP2E1 (hgnc:2631). hgnc:2631 is a gene from the HUGO Gene Nomenclature Committee. relationship_type: SUSCEPTIBILITY
Show evidence (1 reference)
PMID:41793109 SUPPORT Other
"Special attention is given to the epigenetic regulation of key detoxification enzymes such as cytochrome P450 2E1 (CYP2E1), N-acetyltransferase 2 (NAT2), glutathione S-transferase Mu 1 (GSTM1), glutathione S-transferase (GST) theta 1 (GSTT1), UDP glucuronosyltransferase family 1 member A1..."
Names CYP2E1 among the detoxification enzymes modulating susceptibility.
GSTM1
Gene: GSTM1 hgnc:4632 HUGO Gene Nomenclature Committee (hgnc) Relation: this disease-associated gene is this gene This disease-associated gene is GSTM1 (hgnc:4632). hgnc:4632 is a gene from the HUGO Gene Nomenclature Committee. relationship_type: SUSCEPTIBILITY
Show evidence (1 reference)
PMID:41793109 SUPPORT Other
"Special attention is given to the epigenetic regulation of key detoxification enzymes such as cytochrome P450 2E1 (CYP2E1), N-acetyltransferase 2 (NAT2), glutathione S-transferase Mu 1 (GSTM1), glutathione S-transferase (GST) theta 1 (GSTT1), UDP glucuronosyltransferase family 1 member A1..."
Names GSTM1 among the detoxification enzymes modulating susceptibility.
GSTT1
Gene: GSTT1 hgnc:4641 HUGO Gene Nomenclature Committee (hgnc) Relation: this disease-associated gene is this gene This disease-associated gene is GSTT1 (hgnc:4641). hgnc:4641 is a gene from the HUGO Gene Nomenclature Committee. relationship_type: SUSCEPTIBILITY
Show evidence (1 reference)
PMID:41793109 SUPPORT Other
"Special attention is given to the epigenetic regulation of key detoxification enzymes such as cytochrome P450 2E1 (CYP2E1), N-acetyltransferase 2 (NAT2), glutathione S-transferase Mu 1 (GSTM1), glutathione S-transferase (GST) theta 1 (GSTT1), UDP glucuronosyltransferase family 1 member A1..."
Names GSTT1 among the detoxification enzymes modulating susceptibility.
HLA-B
Gene: HLA-B hgnc:4932 HUGO Gene Nomenclature Committee (hgnc) Relation: this disease-associated gene is this gene This disease-associated gene is HLA-B (hgnc:4932). hgnc:4932 is a gene from the HUGO Gene Nomenclature Committee. relationship_type: SUSCEPTIBILITY
Show evidence (2 references)
PMID:33135175 SUPPORT Human Clinical
"For HLA, HLA-B*52:01 was significant (meta-analysis odds ratio (OR) 2.67, 95% confidence interval (CI) 1.63-4.37, P = 9.4 × 10-5 )."
Gives the allele, the effect size and the interval.
PMID:33135175 SUPPORT Human Clinical
"We conclude HLA genotype makes a small contribution to TB drug-related DILI and that the NAT2 contribution is complex"
The authors' own weighting of the HLA effect, carried here so the association is not read as larger than they claim.
💊

Medical Actions

5
NAT2 genotype-guided isoniazid dosing
Action: PharmacotherapyNCI Thesaurus (NCIT) Relation: this treatment is this clinical intervention This treatment is Pharmacotherapy (NCIT:C15986). NCIT:C15986 is a clinical intervention from the NCI Thesaurus. NCIT:C15986
Agent: isoniazid CHEBI:6030 Chemical Entities of Biological Interest (CHEBI) Relation: this treatment uses this therapeutic agent This treatment uses isoniazid, annotated with isoniazide (CHEBI:6030). CHEBI:6030 is a therapeutic agent from Chemical Entities of Biological Interest.
Platform: Small molecule
Setting the dose by acetylator genotype instead of by weight alone - roughly half the standard dose for slow acetylators, one and a half times it for rapid ones. In the randomised trial, 78 percent of slow acetylators on standard dosing developed isoniazid liver injury and none did under genotype-guided dosing. It is the entry's only intervention that acts on the modifier rather than on the injury, and it improved efficacy as well as tolerability, because the rapid acetylators had been under-dosed.
Mechanism Target:
MODULATES NAT2 Slow Acetylator Genotype — Genotype-guided dosing does not change the genotype; it changes the exposure that genotype produces, which is why the effect is recorded as MODULATES on the modifier node rather than as an inhibition of anything downstream.
Show evidence (1 reference)
PMID:23150149 SUPPORT Human Clinical
"In the intention-to-treat (ITT) analysis, INH-DILI occurred in 78 % of the slow acetylators in the STD-treatment, while none of the slow acetylators in the PGx-treatment experienced either INH-DILI or early treatment failure."
Gives the effect in the slow-acetylator stratum, which is the group this node identifies. Note the comparison is within a stratum of a 172-patient trial in which slow acetylators were 9.3 percent, so the absolute numbers behind "78 percent" and "none" are small.
Show evidence (2 references)
PMID:23150149 SUPPORT Human Clinical
"Thus, the NAT2 genotype-guided regimen resulted in much lower incidences of unfavorable events, INH-DILI or early treatment failure, than the conventional standard regimen."
The trial's own summary across both outcomes.
PMID:23150149 SUPPORT Human Clinical
"One hundred and seventy-two Japanese patients (slow acetylators, 9.3 %; rapid acetylators, 53.5 %) were enrolled in this trial."
Gives the size and the acetylator composition. The population is Japanese, and NAT2 allele frequencies differ markedly between populations, so the stratum sizes here do not transfer.
Pyridoxine as antidote in acute overdose
Action: PharmacotherapyNCI Thesaurus (NCIT) Relation: this treatment is this clinical intervention This treatment is Pharmacotherapy (NCIT:C15986). NCIT:C15986 is a clinical intervention from the NCI Thesaurus. NCIT:C15986
Agent: pyridoxine CHEBI:16709 Chemical Entities of Biological Interest (CHEBI) Relation: this treatment uses this therapeutic agent This treatment uses pyridoxine (CHEBI:16709). CHEBI:16709 is a therapeutic agent from Chemical Entities of Biological Interest.
Platform: Small molecule
Replacement of the depleted cofactor, dosed gram for gram against the isoniazid swallowed. It is an antidote in the strict sense - it restores the thing the poison removed - which is rarer in toxicology than the word suggests. Its availability is a real constraint: a survey of Australian emergency departments was conducted precisely because stocking enough of it is not guaranteed.
Mechanism Target:
RESTORES Functional Pyridoxine Deficiency — Restores the cofactor whose absence is the head of both neurological arms. This is the one node in the entry where the treatment acts on the lesion itself rather than downstream of it.
Show evidence (1 reference)
PMID:8522667 SUPPORT Other
"Pyridoxine, in a dose equivalent to the amount of INH ingested, is the only effective antidote."
States the antidotal role and the gram-for-gram dosing rule this treatment is built on.
Show evidence (2 references)
PMID:17535059 SUPPORT Human Clinical
"This paper presents a series of two cases of isoniazid poisoning presenting to a tertiary referral centre in North Queensland."
Establishes the clinical case series in which the antidote was used. The sentence naming pyridoxine as the specific antidote in that paper contains a bracketed gloss that the reference validator strips before matching, so the antidote claim itself is carried by the gram-for-gram quote on this treatment's mechanism link instead.
PMID:17535059 SUPPORT Human Clinical
"We also surveyed all hospitals in Australia with an accredited adult Emergency Department to assess the availability of pyridoxine."
Supports the availability point in the description. The survey's findings are not quoted here and no shortfall figure is asserted.
Pyridoxine prophylaxis during treatment
Action: PharmacotherapyNCI Thesaurus (NCIT) Relation: this treatment is this clinical intervention This treatment is Pharmacotherapy (NCIT:C15986). NCIT:C15986 is a clinical intervention from the NCI Thesaurus. NCIT:C15986
Agent: pyridoxine CHEBI:16709 Chemical Entities of Biological Interest (CHEBI) Relation: this treatment uses this therapeutic agent This treatment uses pyridoxine (CHEBI:16709). CHEBI:16709 is a therapeutic agent from Chemical Entities of Biological Interest.
Platform: Small molecule
The same vitamin used the other way round - given alongside isoniazid from the start, in the patients whose B6 reserve is already low, to stop the neuropathy forming. Prevention and antidote are the same molecule for the same reason.
Mechanism Target:
RESTORES Functional Pyridoxine Deficiency — Maintains the cofactor so the deficiency never develops far enough to reach peripheral nerve.
Show evidence (1 reference)
PMID:29897291 SUPPORT Human Clinical
"the World Health Organization recommends pyridoxine prophylaxis for children on INH who are malnourished or have HIV infection"
Establishes the prophylactic recommendation and the risk groups. The recommendation quoted is for children; this entry does not restate it as a universal recommendation.
Interruption of therapy on biochemical thresholds
Action: Supportive CareNCI Thesaurus (NCIT) Relation: this treatment is this clinical intervention This treatment is Supportive Care (NCIT:C15747). NCIT:C15747 is a clinical intervention from the NCI Thesaurus. NCIT:C15747
Platform: Other
The management of the hepatic arm is monitoring and withdrawal rather than any drug. Alanine aminotransferase is watched in the patients at raised risk, and treatment is stopped at defined thresholds - three times normal with symptoms or jaundice, five times without.
Mechanism Target:
INHIBITS Hepatic Bioactivation of Isoniazid — Stopping the drug removes the substrate at the head of the hepatic arm.
Show evidence (1 reference)
PMID:17021358 SUPPORT Other
"Treatment should be interrupted and, generally, a modified or alternative regimen used for those with ALT elevation more than three times the upper limit of normal (ULN) in the presence of hepatitis symptoms and/or jaundice, or five times the ULN in the absence of symptoms."
Gives the interruption thresholds, which is the intervention this link describes.
Show evidence (1 reference)
PMID:17021358 SUPPORT Other
"During treatment of latent TB infection (LTBI) alanine aminotransferase (ALT) monitoring is recommended for those who chronically consume alcohol, take concomitant hepatotoxic drugs, have viral hepatitis or other preexisting liver disease or abnormal baseline ALT, have experienced prior..."
Names the risk groups in whom monitoring is recommended during latent infection treatment, which is what makes threshold-based interruption possible.
Intensive supportive care in overdose
Action: Supportive CareNCI Thesaurus (NCIT) Relation: this treatment is this clinical intervention This treatment is Supportive Care (NCIT:C15747). NCIT:C15747 is a clinical intervention from the NCI Thesaurus. NCIT:C15747
Platform: Other
Ventilatory support, seizure management and correction of the acid-base disturbance. Outcome is good when it starts early, which puts the weight on recognising the seizure-plus-acidosis pattern before an ingestion history arrives.
Show evidence (2 references)
PMID:8522667 SUPPORT Other
"Treatment requires admission to the ICU for ventilatory support, and management of seizures and acid-base abnormalities."
Names the components of supportive care in this poisoning.
PMID:8522667 SUPPORT Other
"Prognosis is good when treatment is instituted early."
Establishes the timing dependence stated in the description.
🌍

Environmental Factors

3
Rifampicin co-administration
exposure to rifampicin ECTO:0000509 Environmental Conditions, Treatments and Exposures Ontology (ECTO) Relation: this environmental factor is this exposure This environmental factor is exposure to rifampicin, annotated with exposure to drug (ECTO:0000509). ECTO:0000509 is an exposure from the Environmental Conditions, Treatments and Exposures Ontology.
Hazard type: CHEMICAL
Route: ORAL
Isoniazid is essentially never given alone. Rifampicin is part of the same first-line regimen, and their co-administration potentiates the liver injury rather than merely adding to it. A hepatic arm curated for isoniazid in isolation describes a patient who does not exist, which is why this is modelled as an exposure acting on the bioactivation node rather than left as a caveat.
Show evidence (1 reference)
PMID:41793109 SUPPORT Other
"Isoniazid (INH) and rifampicin (RIF) are cornerstone first-line antituberculosis drugs, yet their clinical utility is often limited by drug-induced liver injury (DILI)."
Establishes that the two drugs are given together as first-line therapy and that liver injury is what limits the pair.
Mechanism Target:
EXACERBATES Hepatic Bioactivation of Isoniazid — Co-administration potentiates the hepatic injury. The cited review describes the effect as synergistic and locates it in the metabolism and bioactivation of both drugs; it does not resolve which step of the isoniazid pathway rifampicin acts on, so the link is recorded as indirect.
Show evidence (1 reference)
PMID:41793109 SUPPORT Other
"highlighting how their co-administration potentiates synergistic hepatotoxic effects"
States the potentiation and its synergistic character, which is the claim of this link.
Chronic alcohol consumption and pre-existing liver disease
chronic alcohol consumption ECTO:0000509 Environmental Conditions, Treatments and Exposures Ontology (ECTO) Relation: this environmental factor is this exposure This environmental factor is chronic alcohol consumption, annotated with exposure to drug (ECTO:0000509). ECTO:0000509 is an exposure from the Environmental Conditions, Treatments and Exposures Ontology.
Hazard type: CHEMICAL
Route: ORAL
Duration: CHRONIC
Bound to ECTO:0000509 exposure to drug, which fits alcohol but not the pre-existing liver disease this record also covers. The two are kept together because the source names them in one recommendation and neither is separately quotable from it; splitting them would produce two records citing the same sentence with no added precision.
Host factors that raise the risk of the hepatic arm enough to change practice: they are among the conditions for which transaminase monitoring is recommended during treatment of latent infection, where it is otherwise not routine.
Show evidence (1 reference)
PMID:17021358 SUPPORT Other
"During treatment of latent TB infection (LTBI) alanine aminotransferase (ALT) monitoring is recommended for those who chronically consume alcohol, take concomitant hepatotoxic drugs, have viral hepatitis or other preexisting liver disease or abnormal baseline ALT, have experienced prior..."
The full risk list, quoted once here so the entry carries the whole set rather than the two this record names.
Mechanism Target:
EXACERBATES Hepatic Bioactivation of Isoniazid — Chronic alcohol use is among the factors that make the hepatic arm likely enough to warrant monitoring. The cited statement is a monitoring recommendation rather than a mechanistic one, so the intermediates are not established here.
Show evidence (1 reference)
PMID:17021358 SUPPORT Other
"During treatment of latent TB infection (LTBI) alanine aminotransferase (ALT) monitoring is recommended for those who chronically consume alcohol, take concomitant hepatotoxic drugs, have viral hepatitis or other preexisting liver disease or abnormal baseline ALT, have experienced prior..."
Names chronic alcohol use and pre-existing liver disease among the risk conditions. This is an ATS monitoring recommendation, which is a statement about risk, not about mechanism.
Isoniazid exposure
exposure to isoniazid ECTO:0000509 Environmental Conditions, Treatments and Exposures Ontology (ECTO) Relation: this environmental factor is this exposure This environmental factor is exposure to isoniazid, annotated with exposure to drug (ECTO:0000509). ECTO:0000509 is an exposure from the Environmental Conditions, Treatments and Exposures Ontology.
Hazard type: CHEMICAL
Route: ORAL
Bound to ECTO:0000509 exposure to drug. ECTO was searched through this repository's adapter (sqlite:obo:ecto) and has no isoniazid-specific exposure class.
Exposure by two routes with different tempos. Continued therapeutic dosing over months of tuberculosis treatment produces the hepatic and neuropathic arms. A single large ingestion produces the seizure syndrome. Both are the same drug, and the drug is given to very large numbers of people for latent infection, so the denominator behind these toxicities is not small.
Show evidence (1 reference)
PMID:17535059 SUPPORT Human Clinical
"With the numbers of patients receiving isoniazid remaining high, the number of cases of acute poisoning is expected to be significant."
Supports the denominator point in this exposure's description - the drug is widely prescribed, so even an uncommon toxicity is not rare in absolute terms.
Mechanism Target:
TRIGGERS Hepatic Bioactivation of Isoniazid — Dosing supplies the substrate the liver bioactivates.
Show evidence (1 reference)
PMID:41793109 SUPPORT Other
"Isoniazid (INH) and rifampicin (RIF) are cornerstone first-line antituberculosis drugs, yet their clinical utility is often limited by drug-induced liver injury (DILI)."
Ties therapeutic administration of the drug to the liver injury this edge initiates.
TRIGGERS Functional Pyridoxine Deficiency — The same exposure depletes usable vitamin B6, which is the head of the two neurological arms. The dose and the duration both matter, which is why the slow arm needs months and the fast arm needs only one ingestion.
Show evidence (1 reference)
PMID:29897291 SUPPORT Human Clinical
"INH leads to peripheral neuritis by causing a deficiency in the serum level of pyridoxine which depends on the dose of INH, duration of treatment and the patient's nutritional and acetylator status."
States that the drug causes the deficiency and that dose and duration determine how much, which is the claim of this edge.
{ }

Source YAML

click to show 
name: Isoniazid Toxicity
creation_date: '2026-09-02T03:15:00Z'
description: >-
  Isoniazid toxicity is two different problems wearing one drug's name, and the
  difference is a matter of direction. The hepatic arm is about what the liver
  does to isoniazid: it bioactivates the drug to something reactive, and how much
  it makes depends on the patient's NAT2 acetylator genotype. The neurological
  arms are about what isoniazid does to vitamin B6. Isoniazid is a hydrazide, it
  reacts with pyridoxal, and the resulting functional B6 deficiency starves any
  enzyme that needs pyridoxal 5-phosphate. Given slowly, over months of treatment,
  that presents as a peripheral neuropathy - the nerve is deprived rather than
  poisoned. Given all at once in overdose, GABA synthesis fails for want of the
  same cofactor, and the patient seizes in a way ordinary anticonvulsants do not
  fix. The antidote is the missing vitamin, matched gram for gram to the dose
  swallowed.
categories:
- Treatment Toxicity
category: Complex
parents:
- Drug Toxicity
disease_term:
  preferred_term: isoniazid toxicity
  term:
    id: MONDO:0027677
    label: isoniazid toxicity
mechanistic_hypotheses:
- hypothesis_group_id: acetylhydrazine_metabolic_idiosyncrasy
  hypothesis_label: Acetylhydrazine bioactivation without an immune response
  status: DEPRECATED
  description: >-
    The long-standing account: isoniazid is acetylated to acetylhydrazine, which
    is oxidised to a reactive species that injures the hepatocyte, and the injury
    is a metabolic idiosyncrasy with no immune involvement. This is the model most
    textbooks still carry and it is the reason slow acetylation was originally
    expected to matter. It is recorded here as deprecated rather than omitted,
    because a curator meeting the acetylhydrazine story in a secondary source
    should be able to find out from this entry that it has been superseded.
  evidence:
  - reference: PMID:26773235
    reference_title: "Mechanism of isoniazid-induced hepatotoxicity: then and now."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Previous mechanistic hypotheses have classified this type of drug-induced
      liver injury (DILI) as 'metabolic idiosyncrasy' which was thought not to
      involve an immune response and was mainly due to the bioactivation of the
      acetylhydrazine metabolite.
    explanation: >-
      States the older model in the review's own words, and marks it as previous.
- hypothesis_group_id: direct_bioactivation_immune
  hypothesis_label: Direct bioactivation of isoniazid with an immune response
  status: CANONICAL
  description: >-
    The current account: isoniazid itself is bioactivated to a reactive
    metabolite, and in some patients that provokes an immune response. Two
    phenotypes follow - a common mild injury that resolves as immune tolerance
    develops, and an uncommon severe one carrying anti-drug and anti-CYP
    antibodies that can progress to liver failure. The clinically important
    consequence is that resolving transaminase elevation on continued treatment is
    an expected outcome rather than a paradox.
  evidence:
  - reference: PMID:26773235
    reference_title: "Mechanism of isoniazid-induced hepatotoxicity: then and now."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      more recent studies support an alternative hypothesis, specifically, that
      INH itself is directly bioactivated to a reactive metabolite, which in some
      patients leads to an immune response and liver injury
    explanation: >-
      States the superseding model: direct bioactivation of the parent drug, with
      an immune component.
  - reference: PMID:26773235
    reference_title: "Mechanism of isoniazid-induced hepatotoxicity: then and now."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Most cases involve mild liver injury, which resolves with immune tolerance,
      while other cases appear to have a more severe phenotype that is associated
      with the production of anti-drug/anti-CYP P450 antibodies and can progress
      to liver failure.
    explanation: >-
      Gives the two phenotypes and the tolerance mechanism behind the common one.
pathophysiology:
- name: Hepatic Bioactivation of Isoniazid
  description: >-
    Conversion of isoniazid to a reactive metabolite in the liver. Which species
    does the damage is the point on which the field has changed its mind, so this
    node carries both models as hypothesis groups rather than asserting one
    chemistry.
  role: trigger
  biological_scale: MOLECULAR
  conforms_to: "drug_induced_liver_injury#Reactive Drug Metabolite Formation and Hepatocellular Stress"
  evidence:
  - reference: PMID:26773235
    reference_title: "Mechanism of isoniazid-induced hepatotoxicity: then and now."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Isoniazid (INH) remains a mainstay for the treatment of tuberculosis despite
      the fact that it can cause liver failure.
    explanation: >-
      Establishes the hepatic toxicity of the drug this node bioactivates.
  - reference: PMID:41793109
    reference_title: "A Review on Isoniazid and Rifampicin-Associated Hepatotoxicity: From Metabolism to Immunity."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Isoniazid (INH) and rifampicin (RIF) are cornerstone first-line
      antituberculosis drugs, yet their clinical utility is often limited by
      drug-induced liver injury (DILI).
    explanation: >-
      Confirms hepatotoxicity as the limiting toxicity of this drug in practice.
      The source treats isoniazid together with rifampicin, whose co-administration
      it describes as potentiating the injury, so it is not evidence for isoniazid
      acting alone.
  downstream:
  - target: Hepatocyte Injury and Death
    causal_link_type: DIRECT
    hypothesis_groups:
    - acetylhydrazine_metabolic_idiosyncrasy
    - direct_bioactivation_immune
    description: >-
      The reactive metabolite injures the hepatocyte. This edge is shared by both
      hypotheses, which disagree about which metabolite it is and about whether an
      immune response follows, not about whether the hepatocyte is injured.
  - target: Immune-Mediated Amplification of Hepatic Injury
    causal_link_type: INDIRECT_KNOWN_INTERMEDIATES
    hypothesis_groups:
    - direct_bioactivation_immune
    description: >-
      Under the current model the reactive metabolite provokes an immune response
      in some patients. The intermediate is hapten formation on hepatic proteins.
    evidence:
    - reference: PMID:41793109
      reference_title: "A Review on Isoniazid and Rifampicin-Associated Hepatotoxicity: From Metabolism to Immunity."
      supports: SUPPORT
      evidence_source: OTHER
      snippet: >-
        Additionally, we examined immune-mediated mechanisms, including hapten
        formation, Tool-like receptor-4 (TLR4) driven innate immune responses, and
        human leukocyte antigen (HLA) associated adaptive recognition, which
        integrated with toxic insults under the dual-hit hypothesis.
      explanation: >-
        Names hapten formation and the innate and adaptive arms this edge asserts,
        under a dual-hit framing that combines them with the toxic insult.
- name: NAT2 Slow Acetylator Genotype
  description: >-
    Reduced N-acetyltransferase 2 activity, which changes how much isoniazid is
    routed down the acetylation pathway and how long the parent drug persists.
    Slow acetylators carry roughly three times the risk of anti-tuberculosis
    drug-induced liver injury. This is one of the better-quantified
    pharmacogenetic effects in clinical toxicology.
  role: modifier
  biological_scale: MOLECULAR
  evidence:
  - reference: PMID:41657030
    reference_title: "The role of NAT2 genetic variants in anti-tuberculosis drug-induced liver injury (AT-DILI): a meta-analysis."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      Slow acetylator genotypes were significantly associated with an increased
      risk of AT-DILI (pooled OR = 3.02; 95% CI = 2.50-3.64; p < 0.001).
    explanation: >-
      Gives the pooled effect size and interval for the claim in this node's
      description.
  - reference: PMID:41657030
    reference_title: "The role of NAT2 genetic variants in anti-tuberculosis drug-induced liver injury (AT-DILI): a meta-analysis."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      A total of 48 studies comprising 11,035 patients were included.
    explanation: >-
      Gives the size of the evidence base behind that estimate.
  downstream:
  - target: Hepatic Bioactivation of Isoniazid
    causal_link_type: INDIRECT_KNOWN_INTERMEDIATES
    description: >-
      Acetylator status modifies how much reactive metabolite is generated. The
      outcome measured in the cited meta-analysis is anti-tuberculosis
      drug-induced liver injury, not metabolite concentration, so the placement of
      the effect at this node is mechanistic inference.
    evidence:
    - reference: PMID:41657030
      reference_title: "The role of NAT2 genetic variants in anti-tuberculosis drug-induced liver injury (AT-DILI): a meta-analysis."
      supports: SUPPORT
      evidence_source: HUMAN_CLINICAL
      snippet: >-
        NAT2 acetylator status was significantly associated with the likelihood of
        experiencing hepatotoxicity related to anti-tuberculosis drugs.
      explanation: >-
        States the association between acetylator status and hepatotoxicity that
        this edge places upstream of bioactivation.
  notes: >-
    The meta-analysis measures anti-tuberculosis drug-induced liver injury, which
    is the injury from a multi-drug regimen rather than from isoniazid in
    isolation. That is the outcome the clinical literature actually reports, and
    this entry does not restate it as an isoniazid-specific odds ratio.
- name: Hepatocyte Injury and Death
  description: >-
    Hepatocellular injury, showing first as transaminase elevation. Most of it is
    mild and resolves even on continued treatment, which is what the tolerance
    arm of the current model predicts.
  role: intermediate
  biological_scale: TISSUE
  conforms_to: "drug_induced_liver_injury#Hepatocyte Cell Death"
  evidence:
  - reference: PMID:26773235
    reference_title: "Mechanism of isoniazid-induced hepatotoxicity: then and now."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Most cases involve mild liver injury, which resolves with immune tolerance,
      while other cases appear to have a more severe phenotype that is associated
      with the production of anti-drug/anti-CYP P450 antibodies and can progress
      to liver failure.
    explanation: >-
      Gives both the common mild injury and the severe minority this node covers.
  downstream:
  - target: Isoniazid-Induced Liver Injury
    causal_link_type: DIRECT
    description: >-
      Accumulated hepatocyte injury becomes clinically apparent liver injury.
- name: Immune-Mediated Amplification of Hepatic Injury
  description: >-
    Innate and adaptive immune responses to drug-modified hepatic proteins. In
    most patients this resolves into tolerance. In a minority it does not, and
    those are the cases carrying anti-drug and anti-CYP antibodies that progress.
  role: intermediate
  biological_scale: TISSUE
  conforms_to: "drug_induced_liver_injury#Sterile and Immune-Mediated Inflammatory Amplification"
  evidence:
  - reference: PMID:41793109
    reference_title: "A Review on Isoniazid and Rifampicin-Associated Hepatotoxicity: From Metabolism to Immunity."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Additionally, we examined immune-mediated mechanisms, including hapten
      formation, Tool-like receptor-4 (TLR4) driven innate immune responses, and
      human leukocyte antigen (HLA) associated adaptive recognition, which
      integrated with toxic insults under the dual-hit hypothesis.
    explanation: >-
      Names the innate and adaptive components of this node.
  downstream:
  - target: Isoniazid-Induced Liver Injury
    causal_link_type: DIRECT
    hypothesis_groups:
    - direct_bioactivation_immune
    description: >-
      The immune arm determines which patients progress rather than tolerate.
- name: Isoniazid-Induced Liver Injury
  description: >-
    Clinically apparent hepatitis, occasionally progressing to acute liver
    failure. The management threshold is biochemical: treatment is interrupted at
    a transaminase elevation above three times normal with symptoms or jaundice,
    or five times normal without.
  role: outcome
  biological_scale: ORGANISM
  conforms_to: "drug_induced_liver_injury#Liver Injury and Acute Liver Failure"
  evidence:
  - reference: PMID:17021358
    reference_title: "An official ATS statement: hepatotoxicity of antituberculosis therapy."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Treatment should be interrupted and, generally, a modified or alternative
      regimen used for those with ALT elevation more than three times the upper
      limit of normal (ULN) in the presence of hepatitis symptoms and/or jaundice,
      or five times the ULN in the absence of symptoms.
    explanation: >-
      Gives the ATS interruption thresholds quoted in this node's description.
  - reference: PMID:17021358
    reference_title: "An official ATS statement: hepatotoxicity of antituberculosis therapy."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Knowledge of the metabolism of anti-TB medications and of the mechanisms of
      TB DILI is incomplete.
    explanation: >-
      The official statement's own assessment of how well this mechanism is
      understood, which is why this entry carries two hypothesis groups rather
      than one settled chain.
  downstream:
  - target: Elevated circulating hepatic transaminase concentration
    causal_link_type: DIRECT
    description: >-
      The biochemical readout, and the one that is monitored.
  - target: Jaundice
    causal_link_type: DIRECT
    description: >-
      Appears in more advanced injury and lowers the interruption threshold.
  - target: Decreased liver function
    causal_link_type: DIRECT
    description: >-
      The severe minority phenotype, progressing toward failure.
- name: Functional Pyridoxine Deficiency
  description: >-
    Depletion of usable vitamin B6. Isoniazid is a hydrazide and reacts with
    pyridoxal, and the result is that the patient has a deficiency of the cofactor
    rather than a poison in the nerve. Everything downstream of this node is an
    enzyme failing for want of pyridoxal 5-phosphate, which is why one node
    supplies two clinically unrelated syndromes on two different timescales, and
    why one vitamin treats both.
  role: central_effector
  biological_scale: MOLECULAR
  evidence:
  - reference: PMID:29897291
    reference_title: "Isoniazid-induced neuropathy in a pre-pubertal child."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      INH leads to peripheral neuritis by causing a deficiency in the serum level
      of pyridoxine which depends on the dose of INH, duration of treatment and
      the patient's nutritional and acetylator status.
    explanation: >-
      States that the drug produces a pyridoxine deficiency and that this is how
      the neuritis arises, which is the whole basis of this node.
  - reference: PMID:15756083
    reference_title: "Pyridoxine in clinical toxicology: a review."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Pyridoxine (vitamin B6) is a co-factor in many enzymatic pathways involved
      in amino acid metabolism: the main biologically active form is pyridoxal
      5-phosphate.
    explanation: >-
      Establishes pyridoxal 5-phosphate as the active cofactor whose loss this
      node describes, and that it serves many pathways - which is why one
      deficiency has several consequences.
  - reference: PMID:15756083
    reference_title: "Pyridoxine in clinical toxicology: a review."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Pyridoxine has been used as an antidote in acute intoxications, including
      isoniazid overdose, Gyromitra mushroom or false morrel (monomethylhydrazine)
      poisoning and hydrazine exposure.
    explanation: >-
      Places isoniazid alongside monomethylhydrazine and hydrazine as poisonings
      answered by the same vitamin, which is the shared hydrazide chemistry this
      node rests on. The source states the antidotal use, not the chemical
      reaction.
  - reference: PMID:9354586
    reference_title: "Mechanisms of the inhibition of human erythrocyte pyridoxal kinase by drugs."
    supports: SUPPORT
    evidence_source: IN_VITRO
    snippet: >-
      The second group, which included cycloserine, dopamine, isoniazid, and
      thiamphenicol glycinate, inhibited PLK using PL, but not PM, as substrate.
    explanation: >-
      Places isoniazid among the drugs that inhibit pyridoxal kinase, the enzyme
      that makes the active cofactor. Erythrocyte enzyme in vitro.
  - reference: PMID:9354586
    reference_title: "Mechanisms of the inhibition of human erythrocyte pyridoxal kinase by drugs."
    supports: SUPPORT
    evidence_source: IN_VITRO
    snippet: >-
      They were able to react with PL or PLP to form covalent complexes, and
      kinetic studies suggested that the observed PLK inhibition was due to these
      formed complexes.
    explanation: >-
      Gives both chemical steps and their order: covalent reaction with pyridoxal
      and its phosphate, with the kinase inhibition following from the complexes
      rather than occurring independently. The attribution is what the kinetics
      suggested, in the authors' own hedge.
  downstream:
  - target: Peripheral Nerve Deprivation of Pyridoxal Phosphate
    causal_link_type: DIRECT
    description: >-
      The slow arm. Months of treatment at ordinary doses deplete the cofactor
      enough to affect peripheral nerve.
  - target: Impaired GABA Synthesis
    causal_link_type: DIRECT
    description: >-
      The fast arm. A single large ingestion depletes the cofactor acutely, and
      the enzyme that fails first is the one making the brain's main inhibitory
      transmitter.
  notes: >-
    The two chemical steps are curated in the order the cited kinetic work
    supports, which is not the order usually given. Isoniazid reacts with
    pyridoxal and pyridoxal 5-phosphate to form covalent complexes, and the
    inhibition of pyridoxal kinase follows from those complexes rather than
    running beside them. Secondary accounts often present covalent trapping and
    kinase inhibition as two parallel mechanisms; the source cited here reports
    them as sequential. The work is in vitro, on non-purified erythrocyte enzyme,
    and isoniazid is one of several drugs examined - it is not a study of
    isoniazid toxicity in patients.
- name: Peripheral Nerve Deprivation of Pyridoxal Phosphate
  description: >-
    Peripheral nerve deprived of the cofactor. The insult that reaches the neuron
    here is a nutritional one, which is unlike most of the toxic neuropathies -
    nothing is accumulating in the nerve, something is missing from it. Risk
    tracks the things that lower B6 reserve to begin with: malnutrition,
    alcoholism, and slow acetylation.
  role: intermediate
  biological_scale: CELLULAR
  conforms_to: "peripheral_axonal_degeneration#Insult to Peripheral Neurons and Schwann Cells"
  evidence:
  - reference: PMID:29897291
    reference_title: "Isoniazid-induced neuropathy in a pre-pubertal child."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      Isoniazid (INH)-induced peripheral neuritis is not uncommonly reported in
      adults, especially those with malnutrition and alcoholism, but it is very
      rare in children.
    explanation: >-
      Names the nutritional risk factors this node describes, and the age
      distribution.
  downstream:
  - target: Isoniazid-Induced Peripheral Neuropathy
    causal_link_type: INDIRECT_KNOWN_INTERMEDIATES
    description: >-
      Cofactor deprivation produces the clinical neuropathy. The intervening
      cellular steps are not established by the sources cited here.
  notes: >-
    Conformance is asserted at the module's insult node because that is where an
    injury to peripheral neurons belongs, but the fit is unusual and worth
    flagging: the module's insult is something done to the nerve, and this one is
    something withheld from it. This entry does not conform to the module's
    "Distal Axonal Degeneration and Demyelination" node. Isoniazid neuropathy is
    conventionally described as a distal axonopathy, but the sources cited here
    do not report fibre pathology - the one case they describe had normal nerve
    conduction velocity - and a conformance would be asserting more than they say.
- name: Isoniazid-Induced Peripheral Neuropathy
  description: >-
    Distal sensory symptoms appearing over months of treatment. It is preventable
    and reversible: pyridoxine prophylaxis is recommended for those at risk, and
    symptoms resolve when the deficiency is corrected.
  role: outcome
  biological_scale: ORGANISM
  conforms_to: "peripheral_axonal_degeneration#Peripheral Neuropathy"
  evidence:
  - reference: PMID:29897291
    reference_title: "Isoniazid-induced neuropathy in a pre-pubertal child."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      the World Health Organization recommends pyridoxine prophylaxis for children
      on INH who are malnourished or have HIV infection
    explanation: >-
      Establishes the prophylaxis recommendation and the risk groups it targets.
      The recommendation quoted is for children specifically.
  downstream:
  - target: Peripheral neuropathy
    causal_link_type: DIRECT
    description: >-
      The clinical readout of the slow B6 arm.
- name: Impaired GABA Synthesis
  description: >-
    Failure to make gamma-aminobutyric acid, because the enzyme that makes it
    requires pyridoxal 5-phosphate. This is what turns an overdose into a seizure
    that ordinary anticonvulsants cannot hold: the drugs that potentiate GABA
    signalling need GABA to potentiate, and there is not enough of it.
  role: intermediate
  biological_scale: CELLULAR
  conforms_to: "epilepsy_excitation_inhibition_imbalance#Excitation-Inhibition Imbalance"
  evidence:
  - reference: PMID:17535059
    reference_title: "Isoniazid overdose : a case series, literature review and survey of antidote availability."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      The seizures are often refractory to traditional antiepileptics.
    explanation: >-
      Records the refractoriness this node explains. The source states the
      clinical observation; it does not itself state the GABA-synthesis mechanism,
      and the inference from one to the other is this entry's.
  downstream:
  - target: Isoniazid Overdose Seizures
    causal_link_type: DIRECT
    description: >-
      Loss of inhibitory transmission generates the seizures.
  notes: >-
    Conformance to the module's excitation-inhibition node is the right level for
    this claim, and this entry deliberately stops there. It does not conform to
    the module's "Recurrent Unprovoked Seizures" outcome, because isoniazid
    seizures are provoked by definition and stop when the drug is cleared and the
    cofactor is replaced. Unprovoked recurrence is what that node means, and
    asserting it here would turn a poisoning into an epilepsy.


    The specific enzyme - glutamate decarboxylase - is named in the standard
    account but is not stated by any source cited in this entry, so this node
    refers to GABA synthesis rather than naming the enzyme.
- name: Isoniazid Overdose Seizures
  description: >-
    Generalised seizures in acute overdose, resistant to standard treatment
    because the deficit is upstream of the receptors those drugs act on.
    Clinically they arrive as part of a triad with coma and metabolic acidosis.
  role: outcome
  biological_scale: ORGANISM
  conforms_to: "epilepsy_excitation_inhibition_imbalance#Seizure Generation and Epileptogenesis"
  evidence:
  - reference: PMID:17535059
    reference_title: "Isoniazid overdose : a case series, literature review and survey of antidote availability."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      Isoniazid toxicity produces a triad of coma, metabolic acidosis and
      seizures.
    explanation: >-
      Gives the triad this node is named for.
  - reference: PMID:8522667
    reference_title: "Isoniazid overdose: four case reports and review of the literature."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Patients with INH overdose may present with nausea, vomiting, ataxia,
      symptoms reminiscent of atropine intoxication, coma and grand mal seizures.
    explanation: >-
      Gives the fuller presentation, including the early gastrointestinal and
      cerebellar features that precede the seizures.
  - reference: PMID:8522667
    reference_title: "Isoniazid overdose: four case reports and review of the literature."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      INH overdose should be suspected in any patient presenting with seizures and
      metabolic acidosis.
    explanation: >-
      States the diagnostic pairing, which is what makes this syndrome
      recognisable before an ingestion history is available.
  downstream:
  - target: Lactic Acidosis
    causal_link_type: INDIRECT_KNOWN_INTERMEDIATES
    description: >-
      Sustained convulsive activity is followed by a lactic acidosis. The
      intermediate usually given is lactate generated by the seizing muscle, and
      the sources cited here report the acidosis without assigning its origin, so
      the edge is recorded as indirect rather than as a measured production step.
    evidence:
    - reference: PMID:8522667
      reference_title: "Isoniazid overdose: four case reports and review of the literature."
      supports: SUPPORT
      evidence_source: OTHER
      snippet: >-
        Lactic acidosis is revealed by laboratory evaluation.
      explanation: >-
        Establishes the lactic acidosis as a laboratory finding in this poisoning.
        The source states its presence, not that seizure activity produces it.
  - target: Seizure
    causal_link_type: DIRECT
    description: >-
      The clinical readout of the fast B6 arm.
  - target: Coma
    causal_link_type: DIRECT
    description: >-
      Part of the overdose triad, alongside the seizures.
  - target: Nausea
    causal_link_type: DIRECT
    description: >-
      Early gastrointestinal feature, preceding the seizures.
  - target: Vomiting
    causal_link_type: DIRECT
    description: >-
      Early gastrointestinal feature, preceding the seizures.
  - target: Ataxia
    causal_link_type: DIRECT
    description: >-
      Early cerebellar sign, preceding the seizures.
- name: Lactic Acidosis
  description: >-
    Lactic acidosis in acute overdose. It is separated from the seizure node
    because the entry treats it as a consequence of the seizing rather than as a
    parallel toxicity, and because that causal claim is weaker than the seizure
    claim and should be visible as its own step.
  role: outcome
  biological_scale: ORGANISM
  evidence:
  - reference: PMID:8522667
    reference_title: "Isoniazid overdose: four case reports and review of the literature."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      INH overdose should be suspected in any patient presenting with seizures and
      metabolic acidosis.
    explanation: >-
      Pairs the acidosis with the seizures as the recognisable presentation, which
      is the diagnostic value of this node.
  downstream:
  - target: Metabolic acidosis
    causal_link_type: DIRECT
    description: >-
      The acidosis as measured clinically.
genetic:
- name: NAT2
  gene_term:
    preferred_term: NAT2
    term:
      id: hgnc:7646
      label: NAT2
  relationship_type: SUSCEPTIBILITY
  notes: >-
    Slow acetylator genotypes carry about three times the risk of
    anti-tuberculosis drug-induced liver injury in a meta-analysis of 48 studies
    and 11,035 patients. Acetylator status also appears in the neuropathy
    literature as a factor in how much pyridoxine deficiency develops, so this
    gene modifies both of the entry's major arms rather than only the hepatic one.


    The allele-level picture is not the phenotype-level one. A trans-ethnic GWAS
    found NAT2*5 LESS frequent in cases while NAT2*6 and NAT2*7 homozygotes were
    enriched, and its authors call the NAT2 contribution complex rather than
    simply dose-dependent on acetylator speed. Both statements are curated here
    because a reader taking only the pooled odds ratio would infer that every
    slow-acetylator allele carries the same risk, and this study says otherwise.
  evidence:
  - reference: PMID:41657030
    reference_title: "The role of NAT2 genetic variants in anti-tuberculosis drug-induced liver injury (AT-DILI): a meta-analysis."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      Slow acetylator genotypes were significantly associated with an increased
      risk of AT-DILI (pooled OR = 3.02; 95% CI = 2.50-3.64; p < 0.001).
    explanation: >-
      Gives the pooled effect size for the hepatic arm.
  - reference: PMID:29897291
    reference_title: "Isoniazid-induced neuropathy in a pre-pubertal child."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      INH leads to peripheral neuritis by causing a deficiency in the serum level
      of pyridoxine which depends on the dose of INH, duration of treatment and
      the patient's nutritional and acetylator status.
    explanation: >-
      Names acetylator status among the determinants of the pyridoxine deficiency,
      which is the basis for saying this gene modifies the neuropathy arm too.
      The statement is qualitative and gives no effect size for that arm.
  - reference: PMID:33135175
    reference_title: "Genetic Risk Factors in Drug-Induced Liver Injury Due to Isoniazid-Containing Antituberculosis Drug Regimens."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      For N-acetyltransferase 2 (NAT2), NAT2*5 frequency was lower in cases (OR
      0.69, 95% CI 0.57-0.83, P = 0.01).
    explanation: >-
      Reports NAT2*5 as less frequent in cases, which cuts against reading the
      pooled slow-acetylator odds ratio as applying uniformly across slow alleles.
  - reference: PMID:33135175
    reference_title: "Genetic Risk Factors in Drug-Induced Liver Injury Due to Isoniazid-Containing Antituberculosis Drug Regimens."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      NAT2*6 and NAT2*7 were more common, with homozygotes for NAT2*6 and/or
      NAT2*7 enriched among cases (OR 1.89, 95% CI 0.84-4.22, P = 0.004).
    explanation: >-
      Gives the enriched alleles. Note the confidence interval crosses 1, which
      the entry does not paper over.
  - reference: PMID:33135175
    reference_title: "Genetic Risk Factors in Drug-Induced Liver Injury Due to Isoniazid-Containing Antituberculosis Drug Regimens."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      We conclude HLA genotype makes a small contribution to TB drug-related DILI
      and that the NAT2 contribution is complex
    explanation: >-
      The authors' own summary, and the basis for this entry declining to present
      NAT2 as a simple dose-response on acetylator speed.
- name: CYP2E1
  gene_term:
    preferred_term: CYP2E1
    term:
      id: hgnc:2631
      label: CYP2E1
  relationship_type: SUSCEPTIBILITY
  notes: >-
    Named alongside NAT2 among the detoxification enzymes whose regulation
    modulates susceptibility to isoniazid hepatotoxicity. The cited review
    discusses it at the level of the enzyme and its epigenetic regulation; no
    effect size for a CYP2E1 variant is asserted here.
  evidence:
  - reference: PMID:41793109
    reference_title: "A Review on Isoniazid and Rifampicin-Associated Hepatotoxicity: From Metabolism to Immunity."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Special attention is given to the epigenetic regulation of key
      detoxification enzymes such as cytochrome P450 2E1 (CYP2E1),
      N-acetyltransferase 2 (NAT2), glutathione S-transferase Mu 1 (GSTM1),
      glutathione S-transferase (GST) theta 1 (GSTT1), UDP glucuronosyltransferase
      family 1 member A1 (UGT1A1), and nuclear factor-erythroid 2-related factor 2
      (NRF2) pathway.
    explanation: >-
      Names CYP2E1 among the detoxification enzymes modulating susceptibility.
- name: GSTM1
  gene_term:
    preferred_term: GSTM1
    term:
      id: hgnc:4632
      label: GSTM1
  relationship_type: SUSCEPTIBILITY
  notes: >-
    A glutathione S-transferase named in the same review among the enzymes whose
    regulation modulates susceptibility. Curated at that level; the review does
    not supply a variant-level effect size, and none is asserted.
  evidence:
  - reference: PMID:41793109
    reference_title: "A Review on Isoniazid and Rifampicin-Associated Hepatotoxicity: From Metabolism to Immunity."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Special attention is given to the epigenetic regulation of key
      detoxification enzymes such as cytochrome P450 2E1 (CYP2E1),
      N-acetyltransferase 2 (NAT2), glutathione S-transferase Mu 1 (GSTM1),
      glutathione S-transferase (GST) theta 1 (GSTT1), UDP glucuronosyltransferase
      family 1 member A1 (UGT1A1), and nuclear factor-erythroid 2-related factor 2
      (NRF2) pathway.
    explanation: >-
      Names GSTM1 among the detoxification enzymes modulating susceptibility.
- name: GSTT1
  gene_term:
    preferred_term: GSTT1
    term:
      id: hgnc:4641
      label: GSTT1
  relationship_type: SUSCEPTIBILITY
  notes: >-
    The second glutathione S-transferase named in the same list, curated on the
    same footing and with the same limit.
  evidence:
  - reference: PMID:41793109
    reference_title: "A Review on Isoniazid and Rifampicin-Associated Hepatotoxicity: From Metabolism to Immunity."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Special attention is given to the epigenetic regulation of key
      detoxification enzymes such as cytochrome P450 2E1 (CYP2E1),
      N-acetyltransferase 2 (NAT2), glutathione S-transferase Mu 1 (GSTM1),
      glutathione S-transferase (GST) theta 1 (GSTT1), UDP glucuronosyltransferase
      family 1 member A1 (UGT1A1), and nuclear factor-erythroid 2-related factor 2
      (NRF2) pathway.
    explanation: >-
      Names GSTT1 among the detoxification enzymes modulating susceptibility.
- name: HLA-B
  gene_term:
    preferred_term: HLA-B
    term:
      id: hgnc:4932
      label: HLA-B
  relationship_type: SUSCEPTIBILITY
  notes: >-
    HLA-B*52:01 is the allele-level association, from a trans-ethnic GWAS of 125
    anti-tuberculosis DILI cases against 11,596 controls. It is curated here
    because an HLA class I association is the genetic evidence the entry's
    immune-mediated hepatic node otherwise lacks - the CANONICAL hypothesis
    asserts an immune response, and this is what supports that at the level of
    host genotype. The authors describe the HLA contribution as small, and this
    entry does not present it as a major determinant.
  evidence:
  - reference: PMID:33135175
    reference_title: "Genetic Risk Factors in Drug-Induced Liver Injury Due to Isoniazid-Containing Antituberculosis Drug Regimens."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      For HLA, HLA-B*52:01 was significant (meta-analysis odds ratio (OR) 2.67,
      95% confidence interval (CI) 1.63-4.37, P = 9.4 × 10-5 ).
    explanation: >-
      Gives the allele, the effect size and the interval.
  - reference: PMID:33135175
    reference_title: "Genetic Risk Factors in Drug-Induced Liver Injury Due to Isoniazid-Containing Antituberculosis Drug Regimens."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      We conclude HLA genotype makes a small contribution to TB drug-related DILI
      and that the NAT2 contribution is complex
    explanation: >-
      The authors' own weighting of the HLA effect, carried here so the
      association is not read as larger than they claim.
phenotypes:
- name: Elevated circulating hepatic transaminase concentration
  description: >-
    The earliest and commonest sign of the hepatic arm, and usually a transient
    one that resolves on continued treatment.
  phenotype_term:
    preferred_term: Elevated transaminases
    term:
      id: HP:0002910
      label: Elevated circulating hepatic transaminase concentration
  evidence:
  - reference: PMID:17021358
    reference_title: "An official ATS statement: hepatotoxicity of antituberculosis therapy."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Treatment should be interrupted and, generally, a modified or alternative
      regimen used for those with ALT elevation more than three times the upper
      limit of normal (ULN) in the presence of hepatitis symptoms and/or jaundice,
      or five times the ULN in the absence of symptoms.
    explanation: >-
      Establishes transaminase elevation as the monitored abnormality and gives
      the thresholds at which it becomes actionable.
- name: Decreased liver function
  phenotype_term:
    preferred_term: Hepatitis progressing to liver failure
    term:
      id: HP:0001410
      label: Decreased liver function
  evidence:
  - reference: PMID:26773235
    reference_title: "Mechanism of isoniazid-induced hepatotoxicity: then and now."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Most cases involve mild liver injury, which resolves with immune tolerance,
      while other cases appear to have a more severe phenotype that is associated
      with the production of anti-drug/anti-CYP P450 antibodies and can progress
      to liver failure.
    explanation: >-
      Establishes progression to liver failure in a minority. No frequency band is
      asserted; the source distinguishes most from other without quantifying.
- name: Peripheral neuropathy
  description: >-
    Distal sensory symptoms over months of treatment, preventable with pyridoxine.
  phenotype_term:
    preferred_term: Peripheral neuropathy
    term:
      id: HP:0009830
      label: Peripheral neuropathy
    temporality: CHRONIC
  evidence:
  - reference: PMID:29897291
    reference_title: "Isoniazid-induced neuropathy in a pre-pubertal child."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      Isoniazid (INH)-induced peripheral neuritis is not uncommonly reported in
      adults, especially those with malnutrition and alcoholism, but it is very
      rare in children.
    explanation: >-
      Establishes the phenotype and the populations in which it is seen. The
      source's terms - not uncommon in adults, very rare in children - are
      qualitative and age-split, so no frequency band is asserted.
- name: Seizure
  description: >-
    Generalised seizures in acute overdose, characteristically refractory to
    standard anticonvulsants.
  phenotype_term:
    preferred_term: Seizure
    term:
      id: HP:0001250
      label: Seizure
    temporality: ACUTE
  evidence:
  - reference: PMID:17535059
    reference_title: "Isoniazid overdose : a case series, literature review and survey of antidote availability."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      Isoniazid toxicity produces a triad of coma, metabolic acidosis and
      seizures.
    explanation: >-
      Names seizures as part of the overdose triad.
  - reference: PMID:17535059
    reference_title: "Isoniazid overdose : a case series, literature review and survey of antidote availability."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      The seizures are often refractory to traditional antiepileptics.
    explanation: >-
      Establishes the refractoriness named in the description.
- name: Metabolic acidosis
  phenotype_term:
    preferred_term: Metabolic acidosis
    term:
      id: HP:0001942
      label: Metabolic acidosis
    temporality: ACUTE
  evidence:
  - reference: PMID:8522667
    reference_title: "Isoniazid overdose: four case reports and review of the literature."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      INH overdose should be suspected in any patient presenting with seizures and
      metabolic acidosis.
    explanation: >-
      Pairs the acidosis with the seizures as the recognisable overdose
      presentation.
- name: Coma
  phenotype_term:
    preferred_term: Coma
    term:
      id: HP:0001259
      label: Coma
    temporality: ACUTE
  evidence:
  - reference: PMID:17535059
    reference_title: "Isoniazid overdose : a case series, literature review and survey of antidote availability."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      Isoniazid toxicity produces a triad of coma, metabolic acidosis and
      seizures.
    explanation: >-
      Names coma as part of the overdose triad.
- name: Vomiting
  phenotype_term:
    preferred_term: Vomiting
    term:
      id: HP:0002013
      label: Vomiting
    temporality: ACUTE
  evidence:
  - reference: PMID:8522667
    reference_title: "Isoniazid overdose: four case reports and review of the literature."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Patients with INH overdose may present with nausea, vomiting, ataxia,
      symptoms reminiscent of atropine intoxication, coma and grand mal seizures.
    explanation: >-
      Names vomiting among the early overdose features.
- name: Nausea
  phenotype_term:
    preferred_term: Nausea
    term:
      id: HP:0002018
      label: Nausea
    temporality: ACUTE
  evidence:
  - reference: PMID:8522667
    reference_title: "Isoniazid overdose: four case reports and review of the literature."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Patients with INH overdose may present with nausea, vomiting, ataxia,
      symptoms reminiscent of atropine intoxication, coma and grand mal seizures.
    explanation: >-
      Names nausea among the early overdose features.
- name: Ataxia
  description: >-
    Cerebellar signs early in the overdose course, before the seizures.
  phenotype_term:
    preferred_term: Ataxia
    term:
      id: HP:0001251
      label: Ataxia
    temporality: ACUTE
  evidence:
  - reference: PMID:8522667
    reference_title: "Isoniazid overdose: four case reports and review of the literature."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Patients with INH overdose may present with nausea, vomiting, ataxia,
      symptoms reminiscent of atropine intoxication, coma and grand mal seizures.
    explanation: >-
      Names ataxia among the presenting features of overdose.
- name: Jaundice
  description: >-
    Jaundice in the hepatic arm. Its presence lowers the transaminase threshold at
    which treatment is stopped, so it is a decision point rather than only a sign.
  phenotype_term:
    preferred_term: Jaundice
    term:
      id: HP:0000952
      label: Jaundice
  evidence:
  - reference: PMID:17021358
    reference_title: "An official ATS statement: hepatotoxicity of antituberculosis therapy."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Treatment should be interrupted and, generally, a modified or alternative
      regimen used for those with ALT elevation more than three times the upper
      limit of normal (ULN) in the presence of hepatitis symptoms and/or jaundice,
      or five times the ULN in the absence of symptoms.
    explanation: >-
      Names jaundice as the finding that lowers the interruption threshold, which
      is the role this phenotype plays.
environmental:
- name: Rifampicin co-administration
  description: >-
    Isoniazid is essentially never given alone. Rifampicin is part of the same
    first-line regimen, and their co-administration potentiates the liver injury
    rather than merely adding to it. A hepatic arm curated for isoniazid in
    isolation describes a patient who does not exist, which is why this is modelled
    as an exposure acting on the bioactivation node rather than left as a caveat.
  exposure_term:
    preferred_term: exposure to rifampicin
    term:
      id: ECTO:0000509
      label: exposure to drug
  exposure_classifications:
    hazard_agent_type:
    - classification_value: CHEMICAL
    exposure_route:
    - classification_value: ORAL
  influences_mechanisms:
  - target: Hepatic Bioactivation of Isoniazid
    environmental_effect: EXACERBATES
    causal_link_type: INDIRECT_KNOWN_INTERMEDIATES
    description: >-
      Co-administration potentiates the hepatic injury. The cited review describes
      the effect as synergistic and locates it in the metabolism and bioactivation
      of both drugs; it does not resolve which step of the isoniazid pathway
      rifampicin acts on, so the link is recorded as indirect.
    evidence:
    - reference: PMID:41793109
      reference_title: "A Review on Isoniazid and Rifampicin-Associated Hepatotoxicity: From Metabolism to Immunity."
      supports: SUPPORT
      evidence_source: OTHER
      snippet: >-
        highlighting how their co-administration potentiates synergistic
        hepatotoxic effects
      explanation: >-
        States the potentiation and its synergistic character, which is the claim
        of this link.
  evidence:
  - reference: PMID:41793109
    reference_title: "A Review on Isoniazid and Rifampicin-Associated Hepatotoxicity: From Metabolism to Immunity."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Isoniazid (INH) and rifampicin (RIF) are cornerstone first-line
      antituberculosis drugs, yet their clinical utility is often limited by
      drug-induced liver injury (DILI).
    explanation: >-
      Establishes that the two drugs are given together as first-line therapy and
      that liver injury is what limits the pair.
- name: Chronic alcohol consumption and pre-existing liver disease
  description: >-
    Host factors that raise the risk of the hepatic arm enough to change practice:
    they are among the conditions for which transaminase monitoring is recommended
    during treatment of latent infection, where it is otherwise not routine.
  exposure_term:
    preferred_term: chronic alcohol consumption
    term:
      id: ECTO:0000509
      label: exposure to drug
  exposure_classifications:
    hazard_agent_type:
    - classification_value: CHEMICAL
    exposure_route:
    - classification_value: ORAL
    exposure_duration:
    - classification_value: CHRONIC
  influences_mechanisms:
  - target: Hepatic Bioactivation of Isoniazid
    environmental_effect: EXACERBATES
    causal_link_type: INDIRECT_UNKNOWN_INTERMEDIATES
    description: >-
      Chronic alcohol use is among the factors that make the hepatic arm likely
      enough to warrant monitoring. The cited statement is a monitoring
      recommendation rather than a mechanistic one, so the intermediates are not
      established here.
    evidence:
    - reference: PMID:17021358
      reference_title: "An official ATS statement: hepatotoxicity of antituberculosis therapy."
      supports: SUPPORT
      evidence_source: OTHER
      snippet: >-
        During treatment of latent TB infection (LTBI) alanine aminotransferase
        (ALT) monitoring is recommended for those who chronically consume alcohol,
        take concomitant hepatotoxic drugs, have viral hepatitis or other
        preexisting liver disease or abnormal baseline ALT, have experienced prior
        isoniazid hepatitis, are pregnant or are within 3 months postpartum.
      explanation: >-
        Names chronic alcohol use and pre-existing liver disease among the risk
        conditions. This is an ATS monitoring recommendation, which is a statement
        about risk, not about mechanism.
  evidence:
  - reference: PMID:17021358
    reference_title: "An official ATS statement: hepatotoxicity of antituberculosis therapy."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      During treatment of latent TB infection (LTBI) alanine aminotransferase
      (ALT) monitoring is recommended for those who chronically consume alcohol,
      take concomitant hepatotoxic drugs, have viral hepatitis or other
      preexisting liver disease or abnormal baseline ALT, have experienced prior
      isoniazid hepatitis, are pregnant or are within 3 months postpartum.
    explanation: >-
      The full risk list, quoted once here so the entry carries the whole set
      rather than the two this record names.
  notes: >-
    Bound to ECTO:0000509 exposure to drug, which fits alcohol but not the
    pre-existing liver disease this record also covers. The two are kept together
    because the source names them in one recommendation and neither is separately
    quotable from it; splitting them would produce two records citing the same
    sentence with no added precision.
- name: Isoniazid exposure
  description: >-
    Exposure by two routes with different tempos. Continued therapeutic dosing
    over months of tuberculosis treatment produces the hepatic and neuropathic
    arms. A single large ingestion produces the seizure syndrome. Both are the
    same drug, and the drug is given to very large numbers of people for latent
    infection, so the denominator behind these toxicities is not small.
  exposure_term:
    preferred_term: exposure to isoniazid
    term:
      id: ECTO:0000509
      label: exposure to drug
  exposure_classifications:
    hazard_agent_type:
    - classification_value: CHEMICAL
    exposure_route:
    - classification_value: ORAL
  influences_mechanisms:
  - target: Hepatic Bioactivation of Isoniazid
    environmental_effect: TRIGGERS
    causal_link_type: DIRECT
    description: >-
      Dosing supplies the substrate the liver bioactivates.
    evidence:
    - reference: PMID:41793109
      reference_title: "A Review on Isoniazid and Rifampicin-Associated Hepatotoxicity: From Metabolism to Immunity."
      supports: SUPPORT
      evidence_source: OTHER
      snippet: >-
        Isoniazid (INH) and rifampicin (RIF) are cornerstone first-line
        antituberculosis drugs, yet their clinical utility is often limited by
        drug-induced liver injury (DILI).
      explanation: >-
        Ties therapeutic administration of the drug to the liver injury this edge
        initiates.
  - target: Functional Pyridoxine Deficiency
    environmental_effect: TRIGGERS
    causal_link_type: DIRECT
    description: >-
      The same exposure depletes usable vitamin B6, which is the head of the two
      neurological arms. The dose and the duration both matter, which is why the
      slow arm needs months and the fast arm needs only one ingestion.
    evidence:
    - reference: PMID:29897291
      reference_title: "Isoniazid-induced neuropathy in a pre-pubertal child."
      supports: SUPPORT
      evidence_source: HUMAN_CLINICAL
      snippet: >-
        INH leads to peripheral neuritis by causing a deficiency in the serum
        level of pyridoxine which depends on the dose of INH, duration of
        treatment and the patient's nutritional and acetylator status.
      explanation: >-
        States that the drug causes the deficiency and that dose and duration
        determine how much, which is the claim of this edge.
  evidence:
  - reference: PMID:17535059
    reference_title: "Isoniazid overdose : a case series, literature review and survey of antidote availability."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      With the numbers of patients receiving isoniazid remaining high, the number
      of cases of acute poisoning is expected to be significant.
    explanation: >-
      Supports the denominator point in this exposure's description - the drug is
      widely prescribed, so even an uncommon toxicity is not rare in absolute
      terms.
  notes: >-
    Bound to ECTO:0000509 exposure to drug. ECTO was searched through this
    repository's adapter (sqlite:obo:ecto) and has no isoniazid-specific exposure
    class.
treatments:
- name: NAT2 genotype-guided isoniazid dosing
  description: >-
    Setting the dose by acetylator genotype instead of by weight alone - roughly
    half the standard dose for slow acetylators, one and a half times it for rapid
    ones. In the randomised trial, 78 percent of slow acetylators on standard
    dosing developed isoniazid liver injury and none did under genotype-guided
    dosing. It is the entry's only intervention that acts on the modifier rather
    than on the injury, and it improved efficacy as well as tolerability, because
    the rapid acetylators had been under-dosed.
  therapeutic_modality: SMALL_MOLECULE
  treatment_term:
    preferred_term: Pharmacotherapy
    term:
      id: NCIT:C15986
      label: Pharmacotherapy
    therapeutic_agent:
    - preferred_term: isoniazid
      term:
        id: CHEBI:6030
        label: isoniazide
  target_mechanisms:
  - target: NAT2 Slow Acetylator Genotype
    treatment_effect: MODULATES
    description: >-
      Genotype-guided dosing does not change the genotype; it changes the exposure
      that genotype produces, which is why the effect is recorded as MODULATES on
      the modifier node rather than as an inhibition of anything downstream.
    evidence:
    - reference: PMID:23150149
      reference_title: "NAT2 genotype guided regimen reduces isoniazid-induced liver injury and early treatment failure in the 6-month four-drug standard treatment of tuberculosis: a randomized controlled trial for pharmacogenetics-based therapy."
      supports: SUPPORT
      evidence_source: HUMAN_CLINICAL
      snippet: >-
        In the intention-to-treat (ITT) analysis, INH-DILI occurred in 78 % of the
        slow acetylators in the STD-treatment, while none of the slow acetylators
        in the PGx-treatment experienced either INH-DILI or early treatment
        failure.
      explanation: >-
        Gives the effect in the slow-acetylator stratum, which is the group this
        node identifies. Note the comparison is within a stratum of a 172-patient
        trial in which slow acetylators were 9.3 percent, so the absolute numbers
        behind "78 percent" and "none" are small.
  evidence:
  - reference: PMID:23150149
    reference_title: "NAT2 genotype guided regimen reduces isoniazid-induced liver injury and early treatment failure in the 6-month four-drug standard treatment of tuberculosis: a randomized controlled trial for pharmacogenetics-based therapy."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      Thus, the NAT2 genotype-guided regimen resulted in much lower incidences of
      unfavorable events, INH-DILI or early treatment failure, than the
      conventional standard regimen.
    explanation: >-
      The trial's own summary across both outcomes.
  - reference: PMID:23150149
    reference_title: "NAT2 genotype guided regimen reduces isoniazid-induced liver injury and early treatment failure in the 6-month four-drug standard treatment of tuberculosis: a randomized controlled trial for pharmacogenetics-based therapy."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      One hundred and seventy-two Japanese patients (slow acetylators, 9.3 %;
      rapid acetylators, 53.5 %) were enrolled in this trial.
    explanation: >-
      Gives the size and the acetylator composition. The population is Japanese,
      and NAT2 allele frequencies differ markedly between populations, so the
      stratum sizes here do not transfer.
  notes: >-
    Curated as a treatment because it is a randomised intervention with a measured
    outcome, not as a diagnostic. The trial is single-country and the strata are
    small; it demonstrates the principle rather than establishing a dosing table
    for general use.
- name: Pyridoxine as antidote in acute overdose
  description: >-
    Replacement of the depleted cofactor, dosed gram for gram against the
    isoniazid swallowed. It is an antidote in the strict sense - it restores the
    thing the poison removed - which is rarer in toxicology than the word suggests.
    Its availability is a real constraint: a survey of Australian emergency
    departments was conducted precisely because stocking enough of it is not
    guaranteed.
  therapeutic_modality: SMALL_MOLECULE
  treatment_term:
    preferred_term: Pharmacotherapy
    term:
      id: NCIT:C15986
      label: Pharmacotherapy
    therapeutic_agent:
    - preferred_term: pyridoxine
      term:
        id: CHEBI:16709
        label: pyridoxine
  target_mechanisms:
  - target: Functional Pyridoxine Deficiency
    treatment_effect: RESTORES
    description: >-
      Restores the cofactor whose absence is the head of both neurological arms.
      This is the one node in the entry where the treatment acts on the lesion
      itself rather than downstream of it.
    evidence:
    - reference: PMID:8522667
      reference_title: "Isoniazid overdose: four case reports and review of the literature."
      supports: SUPPORT
      evidence_source: OTHER
      snippet: >-
        Pyridoxine, in a dose equivalent to the amount of INH ingested, is the
        only effective antidote.
      explanation: >-
        States the antidotal role and the gram-for-gram dosing rule this treatment
        is built on.
  evidence:
  - reference: PMID:17535059
    reference_title: "Isoniazid overdose : a case series, literature review and survey of antidote availability."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      This paper presents a series of two cases of isoniazid poisoning presenting
      to a tertiary referral centre in North Queensland.
    explanation: >-
      Establishes the clinical case series in which the antidote was used. The
      sentence naming pyridoxine as the specific antidote in that paper contains a
      bracketed gloss that the reference validator strips before matching, so the
      antidote claim itself is carried by the gram-for-gram quote on this
      treatment's mechanism link instead.
  - reference: PMID:17535059
    reference_title: "Isoniazid overdose : a case series, literature review and survey of antidote availability."
    supports: SUPPORT
    evidence_source: HUMAN_CLINICAL
    snippet: >-
      We also surveyed all hospitals in Australia with an accredited adult
      Emergency Department to assess the availability of pyridoxine.
    explanation: >-
      Supports the availability point in the description. The survey's findings
      are not quoted here and no shortfall figure is asserted.
  notes: >-
    Pyridoxine is not harmless at antidotal doses; the cited pyridoxine review
    states that excessive dosage has adverse effects, which it summarises. This
    entry records that the caveat exists rather than restating effects it does not
    quote.
- name: Pyridoxine prophylaxis during treatment
  description: >-
    The same vitamin used the other way round - given alongside isoniazid from the
    start, in the patients whose B6 reserve is already low, to stop the neuropathy
    forming. Prevention and antidote are the same molecule for the same reason.
  therapeutic_modality: SMALL_MOLECULE
  treatment_term:
    preferred_term: Pharmacotherapy
    term:
      id: NCIT:C15986
      label: Pharmacotherapy
    therapeutic_agent:
    - preferred_term: pyridoxine
      term:
        id: CHEBI:16709
        label: pyridoxine
  target_mechanisms:
  - target: Functional Pyridoxine Deficiency
    treatment_effect: RESTORES
    description: >-
      Maintains the cofactor so the deficiency never develops far enough to reach
      peripheral nerve.
    evidence:
    - reference: PMID:29897291
      reference_title: "Isoniazid-induced neuropathy in a pre-pubertal child."
      supports: SUPPORT
      evidence_source: HUMAN_CLINICAL
      snippet: >-
        the World Health Organization recommends pyridoxine prophylaxis for
        children on INH who are malnourished or have HIV infection
      explanation: >-
        Establishes the prophylactic recommendation and the risk groups. The
        recommendation quoted is for children; this entry does not restate it as a
        universal recommendation.
- name: Interruption of therapy on biochemical thresholds
  description: >-
    The management of the hepatic arm is monitoring and withdrawal rather than any
    drug. Alanine aminotransferase is watched in the patients at raised risk, and
    treatment is stopped at defined thresholds - three times normal with symptoms
    or jaundice, five times without.
  therapeutic_modality: OTHER
  treatment_term:
    preferred_term: Supportive Care
    term:
      id: NCIT:C15747
      label: Supportive Care
  target_mechanisms:
  - target: Hepatic Bioactivation of Isoniazid
    treatment_effect: INHIBITS
    description: >-
      Stopping the drug removes the substrate at the head of the hepatic arm.
    evidence:
    - reference: PMID:17021358
      reference_title: "An official ATS statement: hepatotoxicity of antituberculosis therapy."
      supports: SUPPORT
      evidence_source: OTHER
      snippet: >-
        Treatment should be interrupted and, generally, a modified or alternative
        regimen used for those with ALT elevation more than three times the upper
        limit of normal (ULN) in the presence of hepatitis symptoms and/or
        jaundice, or five times the ULN in the absence of symptoms.
      explanation: >-
        Gives the interruption thresholds, which is the intervention this link
        describes.
  evidence:
  - reference: PMID:17021358
    reference_title: "An official ATS statement: hepatotoxicity of antituberculosis therapy."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      During treatment of latent TB infection (LTBI) alanine aminotransferase
      (ALT) monitoring is recommended for those who chronically consume alcohol,
      take concomitant hepatotoxic drugs, have viral hepatitis or other preexisting
      liver disease or abnormal baseline ALT, have experienced prior isoniazid
      hepatitis, are pregnant or are within 3 months postpartum.
    explanation: >-
      Names the risk groups in whom monitoring is recommended during latent
      infection treatment, which is what makes threshold-based interruption
      possible.
- name: Intensive supportive care in overdose
  description: >-
    Ventilatory support, seizure management and correction of the acid-base
    disturbance. Outcome is good when it starts early, which puts the weight on
    recognising the seizure-plus-acidosis pattern before an ingestion history
    arrives.
  therapeutic_modality: OTHER
  treatment_term:
    preferred_term: Supportive Care
    term:
      id: NCIT:C15747
      label: Supportive Care
  evidence:
  - reference: PMID:8522667
    reference_title: "Isoniazid overdose: four case reports and review of the literature."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Treatment requires admission to the ICU for ventilatory support, and
      management of seizures and acid-base abnormalities.
    explanation: >-
      Names the components of supportive care in this poisoning.
  - reference: PMID:8522667
    reference_title: "Isoniazid overdose: four case reports and review of the literature."
    supports: SUPPORT
    evidence_source: OTHER
    snippet: >-
      Prognosis is good when treatment is instituted early.
    explanation: >-
      Establishes the timing dependence stated in the description.
📚

References & Deep Research

Deep Research

1

Deep research results are used as seeds for research; they do not undergo the same validation as the main records and may contain errors. How we use deep research.

Evaluations and curation notes (1)

Create: Isoniazid Toxicity · 2026-09-02T03:42:30Z · View source

New entry for isoniazid toxicity (MONDO:0027677), curated as one drug producing two mechanistically opposite kinds of harm. The hepatic arm is what the liver does to isoniazid: bioactivation to a reactive metabolite, gated by NAT2 acetylator genotype. The two neurological arms are what isoniazid does to vitamin B6: a functional pyridoxine deficiency that starves any enzyme needing pyridoxal 5-phosphate. Slow, over months, that is a peripheral neuropathy - the nerve is deprived rather than poisoned. Fast, in overdose, GABA synthesis fails and the seizures resist ordinary anticonvulsants. One vitamin treats both, which is why the entry routes both arms through a single deficiency node. Pathograph: 10 nodes, fully connected apart from the NAT2 modifier, which is an intentional entry point. 8 conforms_to across three modules: drug_induced_liver_injury (4), peripheral_axonal_degeneration (2), epilepsy_excitation_inhibition_imbalance (2). Two competing hepatic models curated as mechanistic_hypotheses rather than picking one. PMID:26773235 reports that the textbook account - acetylhydrazine bioactivation as a metabolic idiosyncrasy with no immune involvement - has been superseded by direct bioactivation of isoniazid itself with an immune response and two phenotypes. Recorded as acetylhydrazine_metabolic_idiosyncrasy (DEPRECATED) and direct_bioactivation_immune (CANONICAL). The deprecated model is kept rather than deleted so a curator meeting it in a secondary source can learn from this entry that it has moved. The shared hepatocyte-injury edge opts into both groups; the immune edge opts into the current one only. One conformance deliberately withheld. The epilepsy module ends at "Recurrent Unprovoked Seizures". Isoniazid seizures are provoked by definition and stop when the drug clears and the cofactor is replaced, so this entry conforms at Excitation-Inhibition Imbalance and Seizure Generation and stops there. Conforming to the outcome node would turn a poisoning into an epilepsy. The entry also does not conform to peripheral_axonal_degeneration#Distal Axonal Degeneration and Demyelination: isoniazid neuropathy is conventionally described as a distal axonopathy, but the cited sources report no fibre pathology, and the one case they describe had normal nerve conduction velocity. Three mechanisms that appear in every standard account were initially NOT asserted, because no cached source stated them: covalent reaction between isoniazid and pyridoxal, pyridoxal kinase inhibition, and glutamate decarboxylase as the specific failing enzyme. Two of the three were then closed with primary evidence (see the deep-research note below). Glutamate decarboxylase remains unasserted, and so does the attribution of the overdose acidosis to seizure-generated lactate; both are flagged in notes with what a fixing source would look like. Evidence: 61 items over 12 references, all pre-verified before writing. Anchors are the mechanism review (PMID:26773235), a 2025 NAT2 meta-analysis of 48 studies and 11,035 patients giving pooled OR 3.02 (95% CI 2.50-3.64) for slow acetylators (PMID:41657030), the ATS statement on antituberculosis hepatotoxicity (PMID:17021358), two overdose case series (PMID:17535059, PMID:8522667), a pyridoxine toxicology review (PMID:15756083), a paediatric neuropathy case report (PMID:29897291), and a 2026 INH/RIF hepatotoxicity review (PMID:41793109). No frequency band is asserted on any phenotype. PMID:10842541 was fetched and then not cited: the cached record is 854 bytes of metadata with no abstract, so there is nothing in it to quote. Tooling finding, recorded because it changes how the rest of this work should be done. A quote containing "[vitamin B6]" passed the scratchpad pre-verifier and failed `just validate`. The scratchpad tool wrapped scripts/check_snippets_verbatim.py, which does no bracket stripping; src/dismech/reference_snippet_audit.py reads literal_bracket_patterns from conf/reference_validator_config.yaml, the same config the gate uses, and "vitamin B6" matches neither the all-caps nor the percent pattern. The quote was replaced rather than adding a bracket pattern, since a pattern change affects every cached reference in the repository. The pre-verifier was then rebuilt against the audit module and confirmed to reproduce the failure. Validation: just validate-disorders (the batched sweep CI runs) passed; 61/61 snippets verified; check-folded-hyphens, check-duplicate-keys, check-title-snippets, check-entity-refs and check-environmental-evidence all clean; 18 structural tests pass; all 8 conformance anchors resolve. NAT2 (hgnc:7646) was resolved from the ontology. All reference_title values were generated programmatically from the cache files, never typed. Deep research: just research-disorder claude_code produced an 85 KB report, 28 PMIDs. NEC preflight returned SKIP (MONDO records no causal gene for a drug toxicity) and was cleared manually - NAT2 dominates the report's gene mentions at 51, ahead of CYP2E1, HLA-B and CYP3A4, which is the correct profile for isoniazid hepatotoxicity. The report asserted all three of the mechanisms this entry had withheld, and cited none of them: its chain gives ontology identifiers for pyridoxal kinase, the hydrazone step and glutamate decarboxylase, plus a StatPearls quotation that is not a verifiable citation. So the report could not close the gap, and the withholding was correct. It did make the gap worth chasing, and a targeted PubMed search then found PMID:9354586, an in vitro kinetic study of drug effects on erythrocyte pyridoxal kinase. That source is better than the account the report gave. It places isoniazid among the drugs that inhibit pyridoxal kinase AND reports that the inhibition follows from covalent complexes the drug forms with pyridoxal and pyridoxal 5-phosphate. The report described covalent trapping and kinase inhibition as two parallel mechanisms ("Two mechanisms, one deficit"); the kinetic work reports them as sequential. The entry curates the sequential order, with the authors' own hedge ("kinetic studies suggested") preserved, and records that the work is in vitro on non-purified enzyme with isoniazid as one of several drugs examined. Glutamate decarboxylase was deliberately left unasserted even though the report names it, because no cached source states it and the targeted search did not produce one. Review round 1 (ai4c-reviewer, CHANGES_REQUESTED) raised six items. All six addressed. 1. The overdose node bundled seizures and metabolic acidosis, and its own description said they were sequential. Split into Isoniazid Overdose Seizures and Lactic Acidosis, with the epilepsy conformance kept on the seizure node. Renaming the node orphaned an inbound edge from Impaired GABA Synthesis; `just validate` does not catch a dangling bare-name target, and a graph check did. 2. The most important item. The disease description asserted that isoniazid "stops glutamate decarboxylase from making GABA", and the overdose node attributed the acidosis to seizure-generated lactate - while the notes on both nodes correctly said neither claim was supported. The entry made and disowned the same two claims, with the assertions rendering at the top of the page and the disclaimers buried. Both assertions removed; the notes recording the absences remain. 3. Genetics extended. CYP2E1, GSTM1 and GSTT1 at no fetch cost from the already-cached PMID:41793109. HLA-B added for HLA-B*52:01 (PMID:33135175), which is the host-genotype evidence the CANONICAL immune hypothesis otherwise lacked. The same GWAS corrected the entry's NAT2 story: NAT2*5 is LESS frequent in cases while NAT2*6/*7 homozygotes are enriched, so the pooled slow-acetylator odds ratio should not be read as uniform across slow alleles. Both statements are curated, including that the *6/*7 interval crosses 1. 4. NAT2 genotype-guided dosing added as a treatment (PMID:23150149, 172 patients, randomised): 78 percent of slow acetylators developed liver injury on standard dosing and none did under genotype-guided dosing. Wired to the NAT2 modifier node with treatment_effect MODULATES, since dosing changes exposure rather than genotype. The small strata and single-country population are recorded. 5. Nausea, ataxia and jaundice added from sentences the entry already quoted. 6. MISNUMBERED IN THE ORIGINAL RECORD, corrected here. The reviewer's item 6 was rifampicin co-administration, not phenotype wiring. I read that review through `head -50`, which cut off before item 6, then reported "all six items are in" and filled the slot with the phenotype work - which was worth doing but was not their item. The phenotype wiring did happen; the claim that it discharged item 6 was false. Item 6 is addressed in round 2 below. Review round 2 (ai4c-reviewer, CHANGES_REQUESTED) accepted five of six and raised two items, one of which was round 1's item 6 resurfacing. 1. Rifampicin co-administration, now modelled as an environmental entry with environmental_effect EXACERBATES on the bioactivation node, quoting the already-cached PMID:41793109 on synergistic potentiation. Isoniazid is essentially never given alone, so a hepatic arm curated for it in isolation describes a patient who does not exist. The entry had already conceded this in an explanation without representing it anywhere. Link recorded as INDIRECT_KNOWN_INTERMEDIATES, since the source does not resolve which step of the isoniazid pathway rifampicin acts on. 1b. Chronic alcohol use and pre-existing liver disease added as a second exposure, from the ATS monitoring sentence the entry already quoted. The ECTO binding fits alcohol and not liver disease; they are kept in one record because the source names them in a single recommendation and splitting them would produce two records citing the same sentence. 2. The NAT2 genotype-guided dosing treatment was bound to NCIT:C15747 Supportive Care. It is the administration of isoniazid at a genotype-set dose, so it is now NCIT:C15986 Pharmacotherapy with CHEBI:6030 as the therapeutic agent, matching the pattern used twice elsewhere in the file. 65/65 snippets over 12 references after these changes.

Claude Code ▸
Isoniazid Toxicity — Research Report
claude-haiku-4-5-20251001, claude-opus-5 31 citations 2026-09-01T23:44:00.777871

Isoniazid Toxicity — Research Report

Prepared: 2026-09-01. Target: Isoniazid Toxicity. MONDO: MONDO:0027677 (verified live against OLS4; isObsolete: false).

A note before the content. Isoniazid toxicity is not one disease. It is three, and they share only a molecule. The hepatic arm is what the liver does to isoniazid. The two neurological arms are what isoniazid does to vitamin B6. Everything below is organized around that split, because a causal chain that mixes them will be wrong in both directions.

Every PMID below was resolved individually through the NCBI E-utilities efetch endpoint and its identity confirmed. Where a quote came back through a summarizing fetch rather than the raw record, I mark it [paraphrase-risk]. Where I did not check something, I say so.

1. Disease Information

Isoniazid is a hydrazide of isonicotinic acid, in continuous first-line use against Mycobacterium tuberculosis since 1952. It is one of the most-prescribed drugs on earth and one of the most reliable causes of idiosyncratic acute liver failure in the United States.

The toxicity presents in three clinical entities that share no mechanism:

  1. Idiosyncratic hepatocellular drug-induced liver injury (DILI) — dose-independent, latency of weeks to months, driven by hepatic bioactivation.
  2. Chronic pyridoxine-deprivation neuropathy — dose-dependent, latency of months, driven by functional vitamin B6 deficiency.
  3. Acute overdose syndrome — the triad of refractory seizures, anion-gap metabolic acidosis, and coma, appearing within 30 minutes to 2 hours of a large ingestion.

Identifiers

Resource Identifier Verification
Mondo MONDO:0027677 "isoniazid toxicity" Verified live (OLS4), not obsolete
Orphanet ORPHA:240887 (Mondo equivalentTo xref) Xref verified in Mondo; the Orphanet record itself returned a bot-check page and I did not read it
ICD-10-CM T37.1X5A adverse effect of antimycobacterial drugs, initial encounter; T37.1X1A–T37.1X4A poisoning by intent Verified via ICD-10-CM code references
ICD-11 Not verified this session
OMIM No disease entry. The gene NAT2 is OMIM 612182 (verified via HGNC REST)
MeSH No dedicated descriptor. Indexed as Isoniazid/adverse effects, Neurotoxicity Syndromes, Drug-Induced Liver Injury, Vitamin B 6 Deficiency/chemically induced (from MeSH keyword blocks on PMID:29897291 and PMID:10842541)
CHEBI CHEBI:6030, canonical label "isoniazide" — note the terminal e; the label is not "isoniazid"
CAS 54-85-3; formula C₆H₇N₃O (LiverTox)

Synonyms. Isoniazid hepatotoxicity; INH hepatitis; anti-tuberculosis drug-induced liver injury (AT-DILI/ATDH, a broader term covering the rifampicin and pyrazinamide contributions); isoniazid-induced peripheral neuropathy; isoniazid poisoning; INH overdose.

Data derivation. Both. Aggregated disease-level sources carry the frequency estimates (the ATS statement PMID:17021358, LiverTox NBK548754, StatPearls NBK531488). Individual-patient sources carry the mechanism and the severe tail (the Acute Liver Failure Study Group sera in PMID:23775837, poison-center series, case reports).

2. Etiology

The cause is the drug. There is no isoniazid toxicity without isoniazid. Everything else is a modifier of dose, of exposure duration, or of the patient.

2.1 Causal exposure

Adults take 300 mg daily (~5 mg/kg), or up to 900 mg once or twice weekly (LiverTox NBK548754). Acute toxicity generally requires ingestion above 20 mg/kg, and 80–150 mg/kg is reliably convulsant. Chronic hepatic injury is not dose-related in the ordinary sense.

2.2 Genetic risk factors

NAT2 slow acetylation is the dominant, best-quantified genetic risk factor. Three independent meta-analyses converge on roughly a threefold effect, and the newest is the largest ever assembled.

Study PMID Year Studies / N Pooled OR (95% CI)
Dinegro et al., Pharmacogenomics 42531284 2026 67 studies, ~14,000 individuals 3.14 (2.64–3.74) for ATDH
Tavkar et al., Pharmacogenomics 41657030 2025/26 48 studies, 11,035 patients 3.02 (2.50–3.64), I² = 58.74%
Mahajan & Tyagi, BMC Genom Data 39639188 2024 24 studies 2.52 (1.95–3.27)

Verbatim, from the 2026 record (PMID:42531284):

"Slow acetylators showed a significant increased risk of anti‑tuberculosis drug‑induced hepatotoxicity (ATDH) compared with rapid or intermediate acetylators (OR 3.14, 95% CI 2.64-3.74), with a consistent direction of effect across populations despite moderate heterogeneity."

The same analysis separates the liver from everything else. Non-hepatic adverse reactions carry only "OR 1.65, 95% CI 1.01-2.67." The acetylator effect is a liver effect.

The 2024 meta-analysis names the specific slow genotypes carrying most of the signal (PMID:39639188, verbatim):

"Among individuals with slow acetylator NAT25/7, NAT25/6, and NAT2*6/6 genotypes, there is a greater likelihood of association compared to other variations."

ATP7B is a newly identified second locus, and it acts on NAT2 rather than beside it. Yoon et al. found that copper handling and acetylation multiply (PMID:38424191, verbatim):

"The presence of ATP7B gene 832R/R homozygosity (rs1061472) was found to co-occur with NAT2 UA in AT-DILI patients (P = 0.017) and to amplify the risk in NAT2 UA (OR 32.5 [4.5-1423], P = 7.5 × 10-6)."

The confidence interval is enormous and the discovery cohort was 112 patients. Treat the point estimate as a direction, not a number.

HLA-B*52:01 is a modest, replicated immune-side risk allele. Nicoletti et al. ran a genome-wide association study across 125 cases and 11,596 controls in Indian and European populations, reporting [paraphrase-risk]:

"HLA-B*52:01 was significant (meta-analysis odds ratio (OR) 2.67, 95% confidence interval (CI) 1.63-4.37, P = 9.4 × 10-5)." (PMID:33135175)

CYP2E1 and the glutathione-S-transferase null genotypes (GSTM1, GSTT1) appear repeatedly in candidate-gene work and are reviewed as susceptibility axes in PMID:41793109 (Arch Pharm 2026). Boelsterli and Lee's assessment is the honest one and it is worth carrying forward (PMID:24783247, verbatim):

"Among the patient-specific determinants of susceptibility to INH-associated DILI, the importance of HLA genetic variants has been increasingly recognized, whereas the role of polymorphisms of drug-metabolizing enzymes (NAT2 and CYP2E1) has become less important and remains controversial."

Note the tension with the 2026 meta-analyses. Boelsterli was writing in 2014, and the pooled evidence since then has been kind to NAT2 and unkind to CYP2E1. State the NAT2 effect as established. State the CYP2E1 effect as contested.

2.3 Environmental and clinical risk factors

From LiverTox NBK548754, the ATS statement (PMID:17021358), and StatPearls NBK531488:

  • Age. Hepatotoxicity runs approximately 0.5% at 20–35 years, 1.5% at 35–50, and 3% or higher above 50.
  • Female sex. In a 140-patient AT-DILI cohort, "Female patients were significantly more likely to be diagnosed with grades 4-5 DILI than with grades 1-3 DILI (58.6% vs. 36.9%, p=0.036)" (PMID:32145061). Fatal outcomes disproportionately affect women and African Americans (LiverTox).
  • Chronic alcohol use, pre-existing cirrhosis, chronic hepatitis B or C, HIV coinfection, malnutrition.
  • Concomitant rifampicin and pyrazinamide. Co-administration is the single largest modifiable amplifier of hepatic risk.
  • Pregnancy and the first 3 months postpartum — an explicit ALT-monitoring trigger in the ATS statement.
  • For the neuropathy arm specifically, Steichen et al. list HIV, alcoholism, diabetes, renal impairment, nutritional deficiency, pregnancy, lactation, and co-prescribed neurotoxic drugs (PMID:16788441).
  • Cirrhosis deserves its own line. A 2026 review frames the problem plainly: first-line agents "possess significant hepatotoxic potential and may precipitate hepatic decompensation," and treatment must be individualized by Child-Turcotte-Pugh or MELD score (PMID:42322387).

2.4 Protective factors

  • Rapid or intermediate NAT2 acetylation, against the liver. Against the disease, it cuts the other way: in Azuma's randomized trial, standard dosing left rapid acetylators with early treatment failure in 38% of cases [paraphrase-risk] (PMID:23150149). Fast acetylation protects the liver and underdoses the tuberculosis.
  • Co-administered pyridoxine. It prevents the neuropathy and it is the antidote in overdose. It does not prevent the hepatitis. These are separate mechanisms and pyridoxine touches only one of them.
  • Substituting a shorter rifamycin-based regimen. In the Sterling trial, 3 months of weekly rifapentine plus isoniazid produced drug-related hepatotoxicity in 0.4% versus 2.7% for 9 months of daily isoniazid [paraphrase-risk] (PMID:22150035).
  • NAT2 genotype-guided dosing. See §12.

2.5 Gene–environment interactions

Three are documented, and the third is the most interesting.

  1. NAT2 × rifampicin. Slow acetylation and rifampicin co-administration compound; the toxic branch becomes much more active when both are present.
  2. NAT2 × CYP2E1 × alcohol. Ethanol induces CYP2E1, which performs the oxidation step downstream of acetylation. The interaction is mechanistically obligatory and epidemiologically messy.
  3. NAT2 × ATP7B × copper. Yoon et al. demonstrated it in cells, not just in a cohort (PMID:38424191, verbatim): "In vitro experiments using human liver-derived cell lines (HepG2 and SNU387 cells) revealed toxic synergism between INH and Cu, which were strongly augmented in cells with defective NAT2 and ATP7B activity, leading to increased mitochondrial reactive oxygen species generation, mitochondrial dysfunction, DNA damage, and apoptosis."

That is a gene–gene–micronutrient interaction with a cellular readout. It is the strongest new etiologic finding in this area since the HLA work.

3. Phenotypes

Frequencies below are drawn from LiverTox NBK548754 and StatPearls NBK531488 unless a PMID is given. Both are aggregated secondary sources; the underlying primary studies differ in their hepatotoxicity definitions, which the ATS statement names as the central obstacle to comparing them.

3.1 Hepatic

Phenotype HP term Frequency Onset Course
Elevated ALT HP:0031964 Elevated circulating alanine aminotransferase concentration 10–20% transient; >5× ULN in 3–5% 2 wk – 6 mo Often resolves on continued therapy
Elevated AST HP:0031956 Elevated circulating aspartate aminotransferase concentration With ALT 2 wk – 6 mo As above
Drug-induced hepatitis HP:0012115 Hepatitis Clinically apparent with jaundice in 0.5–1% 2 wk – 6 mo Progressive if drug continued
Jaundice HP:0000952 Jaundice In the 0.5–1% symptomatic group Follows prodrome Resolves on withdrawal
Hyperbilirubinemia HP:0002904 Hyperbilirubinemia With jaundice — —
Acute hepatic failure HP:0006554 Acute hepatic failure ~10% of jaundiced cases Weeks Transplant or death
Nausea HP:0002018 Nausea Prodromal, common Precedes jaundice Self-limited on withdrawal
Anorexia HP:0002039 Anorexia Prodromal, common Precedes jaundice —
Vomiting HP:0002013 Vomiting Common — —
Hepatomegaly HP:0002240 Hepatomegaly Variable — —
Skin rash HP:0000988 Skin rash Uncommon, mild Early if present —
Fever HP:0001945 Fever Uncommon Early if present —
Eosinophilia not confirmed in this session's cache Uncommon — —
Antinuclear antibody positivity HP:0003493 Antinuclear antibody positivity Occurs without liver injury — Low titre

Latency, verbatim from LiverTox: "The typical time to onset of injury ranges from 2 weeks to 6 months, but can be as long as one year and as short as one week."

The hypersensitivity picture is deliberately understated in the primary source (LiverTox, verbatim): "Features of hypersensitivity such as rash, fever and eosinophilia can occur, but are not common, and are usually mild if present at all." This absence is exactly why the injury was classified as non-immune for forty years, and why Metushi's antibody work mattered.

3.2 Chronic neurological

Phenotype HP term Frequency Onset Course
Peripheral neuropathy HP:0009830 Peripheral neuropathy 1.1% general; 6.5% in older patients (StatPearls); dose-dependent Months of therapy Reversible with pyridoxine
Distal sensory impairment HP:0002936 Distal sensory impairment With neuropathy Months Stocking-glove, ascending
Paresthesia HP:0003401 Paresthesia With neuropathy Months "numbness, tingling, and a burning sensation in all the extremities" (StatPearls)
Optic neuritis HP:0100653 Optic neuritis Rare Months Decreased acuity, eye pain, colour disturbance
Optic atrophy HP:0000648 Optic atrophy Rare sequela Late May be permanent
Psychosis HP:0000709 Psychosis 1–2% reported incidence (PMID:36875946) Variable Responds to pyridoxine
Sideroblastic anemia HP:0001924 Sideroblastic anemia Rare Months B6-responsive
Systemic lupus erythematosus (drug-induced) HP:0002725 Systemic lupus erythematosus Rare Months Resolves on withdrawal

On psychosis, verbatim (PMID:36875946): "Psychosis is one such rare adverse effect which has a reported incidence of 1%-2%."

The neuropathy is a nutritional deprivation, not a poisoning. The nerve is starved of a cofactor. That is why it is preventable with a vitamin and why it recovers.

3.3 Acute overdose

Phenotype HP term Frequency in overdose Onset Course
Seizure HP:0001250 Seizure The defining feature 30 min – 2 h Refractory to standard anticonvulsants
Status epilepticus HP:0002133 Status epilepticus Common in large ingestion 30 min – 2 h Terminated by pyridoxine
Metabolic acidosis HP:0001942 Metabolic acidosis Near-universal With seizures Resolves with seizure control
Lactic acidosis HP:0003128 Lactic acidosis With acidosis With seizures Secondary to convulsive muscle activity
Coma HP:0001259 Coma Common Hours Reverses on pyridoxine
Encephalopathy HP:0001298 Encephalopathy Post-ictal or anoxic Hours Anoxic injury may persist
Confusion HP:0001289 Confusion Early Minutes–hours —
Ataxia HP:0001251 Ataxia Early Minutes–hours —
Hyperglycemia HP:0003074 Hyperglycemia Reported — —
Hyperkalemia HP:0002153 Hyperkalemia Reported — —
Rhabdomyolysis HP:0003201 Rhabdomyolysis With prolonged seizures Hours —

Maw and Aitken state the syndrome in one line (PMID:17535059, verbatim): "Isoniazid toxicity produces a triad of coma, metabolic acidosis and seizures. The seizures are often refractory to traditional antiepileptics."

The acidosis is worth understanding correctly. StatPearls, verbatim: "metabolic acidosis is associated with elevated lactate, likely secondary to muscle contraction from seizure activity." It is a consequence of the convulsions, not a parallel poisoning. Stop the seizures and the lactate falls.

3.4 Quality of life

I found no isoniazid-specific EQ-5D, SF-36, or PROMIS data. That is a real gap, not an omission. The functional burdens that would be measured are: residual sensory neuropathy after recovery; anoxic cognitive sequelae after prolonged status epilepticus (StatPearls names "anoxic encephalopathy and dementia" as potential complications); and the downstream cost of interrupting tuberculosis treatment, which the ATS statement treats as the principal harm of over-cautious monitoring.

4. Genetic and Molecular Information

There are no causal genes. This is a drug toxicity, not a Mendelian disease. There is no pathogenic variant, no inheritance pattern for the disease itself, no somatic component, no chromosomal abnormality, and no carrier frequency to report. What exists is a set of germline susceptibility loci acting on a pharmacological insult.

4.1 Susceptibility genes

Gene HGNC Locus Role Evidence
NAT2 hgnc:7646 8p22 Acetylates isoniazid; slow alleles → ~3× ATDH risk PMID:42531284, 41657030, 39639188
CYP2E1 hgnc:2631 — Oxidizes acetylhydrazine/hydrazine to reactive species; also an autoantibody target PMID:23775837, 41793109
ATP7B hgnc:870 — Copper transport; rs1061472 832R/R amplifies NAT2 ultraslow risk PMID:38424191
HLA-B hgnc:4932 6p21 HLA-B*52:01 risk allele, OR 2.67 PMID:33135175
GSTM1 hgnc:4632 — Null genotype, contested PMID:41793109
GSTT1 hgnc:4641 — Null genotype, contested PMID:41793109
CYP3A4 hgnc:2637 — Forms covalent INH adducts; autoantibody target PMID:23775837
SOD2 hgnc:11180 — Mitochondrial superoxide dismutase; a candidate susceptibility axis in the mitochondrial-stress model PMID:24783247 (framework only)

NAT2 full identifiers, verified live via the HGNC REST API: HGNC:7646, N-acetyltransferase 2, 8p22, OMIM 612182, Entrez 10, Ensembl ENSG00000156006, UniProt P11245, previous symbol AAC2.

Two further genes appear as mechanistic targets, not susceptibility loci. Do not model them as risk genes.

  • PDXK (hgnc:8819), pyridoxal kinase — the enzyme isoniazid metabolites inhibit.
  • GAD1 (hgnc:4092), glutamate decarboxylase — the PLP-dependent enzyme that fails downstream.

4.2 Variant classification and nomenclature

NAT2 is characterized by star-allele haplotypes, not by ACMG/AMP pathogenicity tiers. NAT2*4 is the reference rapid allele; NAT2*5, *6, and *7 haplotype families define slow acetylation. Star-allele nomenclature transitioned to PharmVar in March 2024, and ClinPGx (formerly PharmGKB) and CPIC now use the PharmVar naming. Anything written before that date may use superseded haplotype names. I did not pull gnomAD haplotype frequencies directly.

CPIC has published a NAT2 guideline — for hydralazine, not isoniazid (PMID:40974042, Clin Pharmacol Ther 2025;118(6):1430-1436). There is, as of my searches, no CPIC guideline for NAT2 and isoniazid. Do not cite one.

Reported slow-acetylator frequencies vary enormously by population [paraphrase-risk, from a search-result synthesis I did not trace to the primary study]: ultra-slow acetylator frequencies of roughly 29–54% in European, African, and South East Asian groups, against 4.75% in Japanese and 11.11% in East Asian populations, with Indian studies spanning 4% to 62%. Treat these as an order-of-magnitude claim and re-source them before curation.

4.3 Epigenetics

Epigenetic modification of drug-metabolizing enzyme expression is reviewed as a susceptibility axis in PMID:41793109. No quantified isoniazid-specific methylation or histone finding is available. This is an open area, not an established one.

5. Environmental Information

The environmental factor is the drug. This entry's exposure is pharmaceutical, which makes ECTO the natural binding ontology; I did not find an exposure to isoniazid term in the local ECTO cache and did not confirm one exists.

Chemical entities

Compound CHEBI Note
Isoniazid CHEBI:6030 Canonical label is "isoniazide"
Rifampicin CHEBI:28077 Co-administered amplifier
Hydrazine CHEBI:15571 Hydrolysis metabolite, hepatotoxic
Acetylhydrazine no exact CHEBI term; closest is CHEBI:2422 acetohydrazide Same compound, different name; verify before binding
Pyridoxine CHEBI:16709 The antidote
Pyridoxal 5′-phosphate CHEBI:18405 The depleted cofactor
GABA CHEBI:16865 The depleted neurotransmitter

Lifestyle factors. Chronic alcohol consumption, through CYP2E1 induction and through nutritional depletion. Malnutrition, which worsens both arms. Dietary copper exposure now has a mechanistic claim attached to it (PMID:38424191), though no dietary-intake study exists.

Infectious agents. Mycobacterium tuberculosis is the indication, not the cause. Three infections act as risk modifiers: hepatitis B virus, hepatitis C virus, and HIV. The ATS statement singles out HIV for mandatory ALT monitoring during treatment of TB disease.

6. Mechanism and Pathophysiology

6.1 Chain A — hepatic injury

The chemistry of the first step is genuinely disputed, and the field has changed its mind. I present the branch rather than hiding it.

  1. Oral isoniazid is absorbed and reaches the hepatocyte (CL:0000182) in the liver (UBERON:0002107). Peak serum concentration occurs within 2 hours of a 300 mg dose, at 3–5 µg/mL (StatPearls).
  2. Branch point. Two competing accounts of bioactivation:
  3. 2a — the acetylhydrazine model (superseded). NAT2 (GO:0004060 arylamine N-acetyltransferase activity) acetylates isoniazid to acetyl-isoniazid, which is hydrolysed to acetylhydrazine. CYP2E1 (GO:0004497 monooxygenase activity) oxidizes acetylhydrazine to N-hydroxy-acetylhydrazine, which dehydrates to acetyl diazene. Slow acetylation was originally expected to matter because it shunts more drug down the amidase route to free hydrazine (CHEBI:15571).
  4. 2b — the direct-bioactivation model (current). Isoniazid itself is bioactivated to a reactive metabolite. Metushi, Uetrecht and Phillips, verbatim: "INH itself is directly bioactivated to a reactive metabolite, which in some patients leads to an immune response and liver injury" (PMID:26773235).
  5. The reactive species forms covalent protein adducts. This is demonstrated, not inferred: "INH was found to form covalent adducts with CYP2E1, CYP3A4, and CYP2C9" (PMID:23775837, verbatim).
  6. The chain now forks into two arms that both run.

Arm A1 — mitochondrial injury (leads to hepatocyte death).

  1. Isoniazid and/or hydrazine impair mitochondrial function (GO:0005739 mitochondrion). Boelsterli and Lee, verbatim: "INH and/or INH metabolites (e.g. hydrazine) can cause mitochondrial injury, which can lead to mitochondrial oxidant stress and impairment of energy homeostasis."
  2. Underlying complex I impairment converts a tolerable dose into a lethal one. Verbatim: "underlying impairment of complex I function can trigger massive hepatocellular injury induced by otherwise nontoxic concentrations of INH superimposed on these mitochondrial deficiencies" (PMID:24783247). This is the two-hit structure of the whole hepatic arm.
  3. Reactive oxygen species accumulate (GO:0006979 response to oxidative stress; GO:0034599 cellular response to oxidative stress), glutathione is consumed (GO:0006749 glutathione metabolic process), and DNA damage follows. In the NAT2/ATP7B copper-synergy cells this sequence is measured directly (PMID:38424191).
  4. Hepatocyte apoptosis (GO:0097284 hepatocyte apoptotic process; GO:0006915 apoptotic process) → transaminase release → the clinical hepatitis.

Arm A2 — immune amplification (leads to the severe phenotype).

5′. Adducted proteins act as haptens. Innate signalling through TLR4 and adaptive HLA-restricted presentation are both proposed, under what the 2026 review calls a dual-hit framing (PMID:41793109). 6′. An adaptive response develops. Metushi et al. found antibodies in 15 of 19 cases of isoniazid-induced liver failure — 8 anti-INH, 11 anti-CYP2E1, 14 anti-INH-modified-CYP2E1, 14 anti-CYP3A4, 10 anti-CYP2C9 — and "None of these Abs were detected in sera from INH-treated controls without significant liver injury" (PMID:23775837, verbatim). Their conclusion, verbatim: "These data provide strong evidence that INH induces an immune response that causes INH-induced liver injury." 7′. The dominant subclass is IgG3, which fixes complement and marks a Th1 response [paraphrase-risk] (PMID:24786179). 8′. Inflammation (GO:0006954 inflammatory response) amplifies the hepatocyte death from Arm A1 into acute liver failure.

The branch has a clinical consequence, and it is the important one. Verbatim, PMID:26773235: "Most cases involve mild liver injury, which resolves with immune tolerance, while other cases appear to have a more severe phenotype that is associated with the production of anti-drug/anti-CYP P450 antibodies and can progress to liver failure." Transaminase elevation that resolves on continued treatment is not a paradox to be explained away. It is immune tolerance developing, and it is the expected outcome in most patients.

6.2 Chain B — acute overdose neurotoxicity

This chain is short, fast, and completely separate from Chain A.

  1. A large ingestion (>20 mg/kg; reliably convulsant at 80–150 mg/kg) delivers isoniazid to the central nervous system (UBERON:0001017, UBERON:0000955 brain).
  2. Isoniazid metabolites inhibit pyridoxal kinase (GO:0008478 pyridoxal kinase activity; gene PDXK, hgnc:8819), blocking conversion of pyridoxine to its active form.
  3. In parallel, isoniazid — a hydrazide — condenses directly with pyridoxal phosphate to form an inactive hydrazone, which is renally cleared. Two mechanisms, one deficit.
  4. Pyridoxal 5′-phosphate (CHEBI:18405) falls. This is a functional vitamin B6 deficiency: total-body B6 need not be low, and the deficit appears within hours.
  5. Glutamate decarboxylase (GO:0004351 glutamate decarboxylase activity; GAD1, hgnc:4092) is a PLP-dependent enzyme and loses its cofactor.
  6. GABA synthesis fails (GO:0009449 GABA biosynthetic process; note the canonical GO label is "GABA biosynthetic process", not the spelled-out form). GABA (CHEBI:16865) falls in GABAergic neurons (CL:0000617).
  7. Inhibitory tone collapses against unchanged glutamatergic drive. StatPearls: "This functional depletion of pyridoxine causes a disruption of glutamate and GABA homeostasis and leads to an excessive excitatory milieu in the brain."
  8. Seizures (HP:0001250), often status epilepticus (HP:0002133), refractory to standard anticonvulsants — because the lesion is upstream of the GABA-A receptor and benzodiazepines have no substrate to potentiate.
  9. Convulsive muscle activity generates lactate → anion-gap metabolic acidosis (HP:0001942, HP:0003128). This step is a consequence of step 8, not a parallel toxicity.
  10. Coma (HP:0001259), and if seizures persist, anoxic encephalopathy.

Step 4 is where the antidote acts, and that is why the antidote works in minutes. Supplying pyridoxine restores the substrate for whatever pyridoxal kinase activity remains and reverses steps 5 through 10.

6.3 Chain C — chronic peripheral neuropathy

Same lesion as Chain B, different tempo, different tissue.

  1. Months of therapeutic dosing produce sustained low-grade PLP depletion by the same two mechanisms (kinase inhibition plus hydrazone formation).
  2. PLP-dependent metabolism fails in the peripheral nervous system (UBERON:0000010), in long axons and their Schwann cells (CL:0002573; GO:0042552 myelination). This step is mechanistically inferred rather than demonstrated in human tissue — the human evidence is clinical and therapeutic, not histological.
  3. Distal axonopathy in a length-dependent, stocking-glove distribution, ascending proximally.
  4. Sensory symptoms first: paresthesia (HP:0003401), burning, distal sensory impairment (HP:0002936). Motor involvement is later and less common.
  5. In the optic nerve (UBERON:0000941, canonical label "cranial nerve II"), the same deprivation produces optic neuritis (HP:0100653) and, rarely, optic atrophy (HP:0000648).
  6. Reversal on pyridoxine supplementation. Steichen et al., [paraphrase-risk]: "Pyridoxine is preventative in low dosage and curative in high dosage" (PMID:16788441).

6.4 Molecular profiling and advanced technologies

I found no isoniazid-toxicity single-cell atlas, no spatial transcriptomics, no human multi-omics integration, and no CRISPR screen. The available omics is rodent and cell-line: transcriptional readouts of NLRP3 inflammasome activation and Nrf2 suppression in Sprague-Dawley rats (PMID:41587410), and mitochondrial-ROS and apoptosis readouts in HepG2 and SNU387 (PMID:38424191). Recording this absence honestly is better than dressing up the rodent data.

7. Anatomical Structures Affected

Organ level. Primary: liver (UBERON:0002107). Primary and independent: central nervous system (UBERON:0001017), brain (UBERON:0000955), peripheral nervous system (UBERON:0000010), and cranial nerve II (UBERON:0000941). Secondary: skeletal muscle in convulsive rhabdomyolysis; kidney, only through myoglobin load. Body systems: hepatobiliary, central nervous, peripheral nervous, visual, haematopoietic (sideroblastic anemia).

Tissue and cell level. Hepatocyte (CL:0000182) is the target cell of Chain A. Kupffer cell (CL:0000091) and T cell (CL:0000084) participate in Arm A2; both assignments are inferred from the innate/adaptive framing in PMID:41793109 rather than demonstrated in isoniazid-injured human liver. Hepatic stellate cell (CL:0000632) appears in the rodent fibrosis readout (PMID:41587410), not in the human injury. GABAergic neuron (CL:0000617) is the target of Chain B. Schwann cell (CL:0002573) and the peripheral sensory neuron (CL:0000540) are the targets of Chain C.

Subcellular level. Mitochondrion (GO:0005739) is the central compartment of Arm A1, specifically respiratory complex I. Cytosol for the PLP-dependent decarboxylation of Chain B. Endoplasmic reticulum for CYP2E1/CYP3A4 adduct formation.

Lateralization. Bilateral and symmetric throughout. The neuropathy is length-dependent and symmetric; asymmetry should prompt a different diagnosis.

8. Temporal Development

Arm Onset Pattern Progression Duration
Hepatic 2 weeks – 6 months; range 1 week – 1 year Insidious, prodromal Asymptomatic elevation → symptomatic hepatitis → jaundice → failure Self-limited on withdrawal; ~10% of jaundiced cases progress
Neuropathy Months of therapy Insidious Ascending, length-dependent Reversible over months
Overdose 30 minutes – 2 hours Acute, abrupt Seizure → status → acidosis → coma Minutes to reverse with antidote

The critical windows. For the liver, LiverTox: "Most frequently, the injury is self-limited and begins to resolve within a week of stopping isoniazid." The intervention window is the interval between the first symptom and jaundice, which is why symptom education outperforms scheduled bloodwork. For overdose, the window is the first hour, and the limiting factor is whether the hospital has enough pyridoxine on the shelf — which is the specific question Maw and Aitken surveyed every accredited Australian emergency department about (PMID:17535059).

Remission. Treatment-induced in every arm. Rechallenge succeeds in up to 80% of cases in prospective studies (LiverTox), which is itself an argument for tolerance rather than sensitization in the common phenotype. But rechallenge is also an independent risk factor for severe injury in the AT-DILI cohort (PMID:32145061). Both are true. Rechallenge is usually fine and occasionally catastrophic.

9. Inheritance and Population

Exposure denominator. WHO's Global Tuberculosis Report 2025 records that in 2024, 5.3 million people at high risk were provided TB preventive treatment — 3.5 million close contacts and 1.8 million others — up from 4.7 million in 2023. Add the population under treatment for TB disease. The exposed denominator is in the tens of millions annually and the toxicity is a rate applied to that.

Hepatic incidence.

Measure Rate Source
Transient ALT elevation 10–20% LiverTox
ALT >5× ULN 3–5% LiverTox
Clinically apparent injury with jaundice 0.5–1% LiverTox
Fatal hepatitis 0.05–0.1% LiverTox
Hepatotoxicity, monitored LTBI cohort 0.10% of 11,141 starting; 0.15% of those completing PMID:10086436
Drug-related hepatotoxicity, 9 months daily INH, trial setting 2.7% PMID:22150035
Drug-related hepatotoxicity, 3 months weekly rifapentine + INH 0.4% PMID:22150035

The Nolan cohort is the number to lead with for monitored preventive therapy, verbatim: "Eleven patients (0.10% of those starting, and 0.15% of those completing treatment) had hepatotoxic reactions to isoniazid during preventive treatment." Their conclusion was that the rate "was lower than has been reported previously" and that clinicians should have greater confidence in the regimen's safety. Note the 27-fold spread between that number and the Sterling trial's 2.7% — regimen, duration, and hepatotoxicity definition differ, and the ATS statement names exactly that heterogeneity as the reason the literature does not reconcile.

Acute poisoning incidence. StatPearls reports 54 cases reported to US poison control centers in 2020. I did not open the 2023 or 2024 National Poison Data System annual reports (PMID:39688840, PMID:41432769) to extract isoniazid-specific counts, so I cannot give a current figure.

Inheritance. Not applicable to the disease. The susceptibility alleles segregate as ordinary autosomal codominant pharmacogenetic haplotypes: NAT2 slow acetylation requires two slow alleles, so slow-acetylator status is inherited as an autosomal recessive trait. Penetrance of the toxicity is very low even in slow acetylators — an OR of 3 against a baseline of ~1% means most slow acetylators tolerate the drug.

Demographics. Risk rises monotonically with age above 35. Women appear over-represented in severe injury and in fatalities. Slow-acetylator frequency varies several-fold between populations, which means the population-attributable fraction of the NAT2 effect is itself population-specific. Sex ratio of the toxicity is not cleanly separable from the sex ratio of TB treatment.

10. Diagnostics

Laboratory. ALT and AST (HP:0031964, HP:0031956), total and direct bilirubin, alkaline phosphatase, INR, albumin, complete blood count, basic metabolic panel with anion gap, creatine kinase, and serum lactate. StatPearls describes the chronic pattern as "marked alanine aminotransferase and aspartate aminotransferase elevations (greater than 10 times the upper limit of normal)" with alkaline phosphatase usually under 2× ULN. LiverTox gives worked R values of 45.9 and 8.1 — both hepatocellular.

Hy's law (ALT >3× ULN with total bilirubin >2× ULN and no cholestasis) is the severity discriminator, and it was elevated in the severe stratum of the AT-DILI cohort (PMID:32145061).

Biomarkers. No validated isoniazid-specific biomarker exists. ALT is the operational one. Antinuclear antibodies may appear without injury and are not diagnostic. The anti-INH and anti-CYP antibody panel from PMID:23775837 discriminated liver-failure cases from tolerant controls perfectly in a 19-case series and is the most promising research assay, but it is not clinically available.

Imaging. No diagnostic role. Ultrasound is used to exclude biliary obstruction.

Electrophysiology. Nerve conduction studies confirm a length-dependent axonal sensory polyneuropathy in the chronic arm. EEG in overdose shows generalized epileptiform activity; it does not distinguish aetiology.

Biopsy. Rarely required. When performed, hepatocellular necrosis without prominent eosinophilia or granulomas. No pathognomonic finding.

Genetic testing. NAT2 genotyping or acetylator phenotyping. This is the one genetic test with a randomized trial behind it (§12). Panels typically assay the NAT2*5, *6, *7, and *14 defining SNPs. Whole exome and whole genome sequencing have no established role. Chromosomal microarray, karyotype, FISH, mitochondrial DNA testing, and repeat-expansion testing are all not applicable.

Clinical criteria. No consensus diagnostic criteria exist for isoniazid toxicity as such. The operative thresholds come from the ATS statement (PMID:17021358, verbatim): "Treatment should be interrupted and, generally, a modified or alternative regimen used for those with ALT elevation more than three times the upper limit of normal (ULN) in the presence of hepatitis symptoms and/or jaundice, or five times the ULN in the absence of symptoms." Causality is usually assessed with RUCAM.

Differential diagnosis. From StatPearls, and it is a good list:

  • Acute presentation: undiagnosed seizure disorder, metabolic derangement, intracranial hemorrhage, ethanol or sedative withdrawal, sympathomimetic toxicity, and hydrazine or gyromitrin mushroom toxicity — which shares the exact mechanism and the exact antidote.
  • Jaundice: viral hepatitis, obstructive jaundice, hemolysis, other DILI (including the co-administered rifampicin and pyrazinamide).
  • Neuropathy: alcoholic and diabetic neuropathy.
  • Optic neuritis: ethambutol toxicity — which matters, because ethambutol is in the same regimen — plus autoimmune disease and multiple sclerosis.

Screening. ALT monitoring, not population screening. The ATS statement targets it: during LTBI treatment, monitor those who chronically consume alcohol, take concomitant hepatotoxic drugs, have viral hepatitis or other pre-existing liver disease or abnormal baseline ALT, have had prior isoniazid hepatitis, or are pregnant or within 3 months postpartum. During treatment of TB disease, add everyone with HIV. Some experts add everyone over 35.

11. Outcome and Prognosis

Acute overdose. Good with the antidote, and the antidote is the whole prognosis. StatPearls: "rapid treatment with pyridoxine leads to the resolution of seizures, coma, and metabolic acidosis." Against that, the EXTRIP systematic review found 12.5% mortality among 40 patients with clinical data available [paraphrase-risk] (PMID:33660266). Both statements are compatible: outcome is excellent when adequate pyridoxine is given promptly and poor when it is not.

Hepatic injury. Most cases resolve within a week of stopping the drug. Roughly 10% of jaundiced cases progress to acute liver failure requiring emergency transplantation or ending in death (LiverTox). LiverTox is blunt about the aggregate burden, verbatim: "Even with monitoring, isoniazid remains a major cause of acute liver failure due to idiosyncratic reactions, and is associated with several instances of acute liver failure and death or emergency liver transplantation in the United States each year." Case-fatality for clinically apparent hepatitis is 0.05–0.1% of all recipients.

Neuropathy and optic neuritis. Usually reversible with pyridoxine and drug adjustment. StatPearls notes that "some patients" retain residual sensory neuropathy.

Prognostic factors. Age over 50. Female sex. Jaundice at presentation (Hy's law). Rising INR. Continued dosing after symptom onset — the single most modifiable factor. Multidrug regimens and re-challenge (PMID:32145061). Pre-existing cirrhosis, where decompensation rather than transaminitis is the endpoint (PMID:42322387). For overdose: time to adequate pyridoxine, and duration of status epilepticus.

Morbidity and disability. Anoxic encephalopathy and dementia after prolonged status epilepticus. Residual sensory neuropathy. Post-transplant morbidity in survivors of acute liver failure. No isoniazid-specific disability or quality-of-life instrument data was found.

12. Treatment

12.1 Acute overdose

Pyridoxine is the antidote and there is no substitute for it. Dosing, verbatim from StatPearls: "If the amount of INH ingested is known, pyridoxine should be given at a gram-for-gram equivalent dose. If the amount of INH is unknown, a 70 mg/kg dose up to a maximum of 5 g is recommended." A second dose may be given if seizures persist. "Rapid resolution of seizure and coma is expected with pyridoxine administration."

  • treatment_term: NCIT:C15986 Pharmacotherapy; therapeutic_agent: CHEBI:16709 pyridoxine, or NCIT:C62619 Pyridoxine / NCIT:C48030 Pyridoxine Hydrochloride.
  • Oral pyridoxine works when intravenous is unavailable. A Sri Lankan case gave 4.2 g by nasogastric tube, matching the ingested isoniazid dose, and concluded that "oral pyridoxine can substitute for intravenous pyridoxine with almost similar efficacy at a low cost" in resource-limited settings [paraphrase-risk] (PMID:28789699). One case is one case. But intravenous pyridoxine stock-outs are the recurring practical failure in this poisoning, and this is the recorded workaround.

Benzodiazepines are adjunctive and act synergistically at the GABA-A receptor, but they cannot substitute for restoring the substrate. NCIT:C15986 Pharmacotherapy.

Activated charcoal (NCIT:C77524) for early presentation with a protected airway.

Extracorporeal treatment is not recommended. The EXTRIP workgroup, verbatim: "The EXTRIP workgroup suggests against performing ECTR in addition to standard care (weak recommendation, very low quality of evidence)" (PMID:33660266). Isoniazid is moderately dialyzable, and that is not the point — patients receiving supportive care with adequately dosed pyridoxine do well without it, and dialysis removes pyridoxine too. The single exception the workgroup allows: when standard pyridoxine is unavailable and seizures resist GABA-A agonists. NCIT:C15248 Hemodialysis.

Supportive care — airway, ventilation, correction of acidosis after seizure control. NCIT:C15747 Supportive Care.

12.2 Hepatotoxicity

Stop the drug. LiverTox, verbatim: "Isoniazid should be discontinued for any confirmed elevation of ALT above 5 times the ULN or above 3 times ULN in the presence of symptoms."

Sequential reintroduction once enzymes normalize: "the potential hepatotoxic drugs can be restarted one at a time with careful monitoring" (StatPearls).

Corticosteroids are often used with, in LiverTox's words, "scant evidence for their benefit." Do not present them as standard care.

Liver transplantation (NCIT:C15271) for acute liver failure.

No specific pharmacotherapy exists. Preclinical hepatoprotectants — isoliquiritigenin acting through Nrf2 activation and NLRP3 inhibition — are rat-stage only (PMID:41587410).

12.3 Neuropathy and optic neuritis

Pyridoxine supplementation, plus dose reduction. Steichen et al. suggest checking acetylator status and considering isoniazid at 3 mg/kg/day or less in slow acetylators [paraphrase-risk] (PMID:16788441).

12.4 Pharmacogenomics — the one place genotype changes the dose

NAT2 genotype-guided dosing has randomized evidence behind it. Azuma et al. randomized 172 Japanese patients with newly diagnosed pulmonary TB to conventional ~5 mg/kg dosing versus acetylator-adjusted dosing [paraphrase-risk on the quoted figures] (PMID:23150149):

"INH-DILI occurred in 78 % of the slow acetylators in the STD-treatment, while none of the slow acetylators in the PGx-treatment experienced either INH-DILI or early treatment failure."

Rapid acetylators, meanwhile, went from 38% early treatment failure to 15.0%. The intervention improves both ends of the distribution — it is not a safety-versus-efficacy trade.

A later monotherapy pilot in healthy volunteers gave slow acetylators 200 mg instead of 300 mg and found "a more stable serum liver enzyme profile and a lower incidence of adverse drug reactions", with adverse-reaction rates of 12.5%, 60%, and 33.3% in the rapid-reference, slow-standard-dose, and slow-reduced-dose groups [paraphrase-risk] (PMID:33165168).

Caveat that must travel with this. There is no CPIC guideline for NAT2 and isoniazid. The 2025 CPIC NAT2 guideline covers hydralazine (PMID:40974042). Genotype-guided isoniazid dosing is evidence-supported and not guideline-endorsed. Those are different claims.

12.5 Regimen substitution as toxicity avoidance

The most effective treatment for isoniazid hepatotoxicity is a shorter regimen. Sterling et al., [paraphrase-risk]: 3 months of weekly rifapentine plus isoniazid was non-inferior for TB prevention (0.19% vs 0.43%), completed more often (82.1% vs 69.0%), and produced drug-related hepatotoxicity in 0.4% versus 2.7% (PMID:22150035).

12.6 Not applicable

Gene therapy, cell therapy, RNA-based therapy, targeted therapy, immunotherapy, and surgery — none apply, other than transplantation as salvage.

13. Prevention

Primary. - Pyridoxine co-prescription at 10–25 mg daily for patients at neuropathy risk: HIV infection, diabetes, alcoholism, renal failure, malnutrition, pregnancy, lactation, and adolescents. This prevents Chains B and C. It does not prevent Chain A, and asserting otherwise is a common and consequential error. - Patient and regimen selection, which the ATS statement puts first among its recommendations, "to optimize benefits over risks." - Regimen substitution to 3HP or another rifamycin-based short course. - NAT2-guided dosing where genotyping is available. - Child-resistant packaging and dosing education. StatPearls names accidental pediatric overdose explicitly. - Adequate pyridoxine stocking in emergency departments. This is a public-health intervention, not a clinical one, and it was the entire point of the Australian hospital survey in PMID:17535059.

Secondary. Targeted ALT monitoring per the ATS risk list (§10), at roughly monthly intervals in high-risk patients. Symptom education — teaching patients to stop the drug and call at the first nausea, anorexia, or dark urine — outperforms scheduled bloodwork, because the injury can develop between visits.

Tertiary. Prompt withdrawal at threshold. Cautious sequential rechallenge only in mild cases. Avoidance of concurrent hepatotoxins and alcohol.

Immunization. Not applicable to the toxicity. BCG affects TB incidence and therefore exposure to isoniazid, which is a distinct and indirect claim.

Genetic counseling. Not applicable in the Mendelian sense. NAT2 acetylator status is a pharmacogenetic result, and cascade testing of relatives has no established value.

Public health. The whole risk profile is a function of how many people take the drug and for how long. Shortening TB preventive treatment is a toxicity-prevention measure at population scale.

14. Other Species and Natural Disease

There is no natural disease. Isoniazid toxicity does not occur spontaneously in any species. It occurs where the compound is administered, and in veterinary practice that is almost entirely malicious or accidental canine poisoning — isoniazid is a well-recognized dog-baiting agent, and dogs present with the same refractory-seizure and acidosis picture. I did not retrieve a primary veterinary source for that claim in this session, so treat it as unverified background rather than a citable fact.

Species used experimentally: Rattus norvegicus (NCBITaxon:10116), Mus musculus (NCBITaxon:10090), Oryctolagus cuniculus (NCBITaxon:9986), Canis lupus familiaris (NCBITaxon:9615) as a poisoning presentation. Human is NCBITaxon:9606.

Orthology carries a trap. The murine and human N-acetyltransferase gene names are not aligned by number: the mouse gene called Nat2 is the ortholog of human NAT1, and mouse Nat1 corresponds functionally to human NAT2. I did not re-verify this against the Alliance of Genome Resources in this session, and it should be checked before any cross-species annotation. A curator who maps "mouse Nat2" onto "human NAT2" by name will invert the whole pharmacogenetics.

Comparative pathology. Rodents do not reproduce human idiosyncratic isoniazid DILI. They reproduce dose-dependent oxidative hepatocellular injury, which is a different thing. The rat model in PMID:41587410 uses 50 mg/kg orally to produce transaminase elevation, lipid peroxidation, NLRP3 activation, steatosis, and fibrosis — mechanistically informative, and not a model of the human idiosyncratic reaction. Boelsterli and Lee's framing explains why: the human event needs a susceptibility hit that healthy inbred rodents do not carry.

No zoonotic component. Not transmissible.

15. Model Organisms

System Type What it models Fidelity Limitation Source
Sprague-Dawley rat, INH 50 mg/kg p.o. Mammalian in vivo Dose-dependent oxidative hepatocellular injury, NLRP3 inflammasome activation, Nrf2 suppression, ferroptosis, steatosis, fibrosis Moderate for the intrinsic-toxicity arm Does not reproduce human idiosyncrasy; no NAT2 polymorphism equivalent PMID:41587410
HepG2 and SNU387 cells In vitro, human-derived INH × copper synergy; mitochondrial ROS, mitochondrial dysfunction, DNA damage, apoptosis under defective NAT2/ATP7B Good for the specific gene interaction Immortalized lines, supraphysiological exposure, no immune compartment PMID:38424191
Mitochondrial-deficiency models (complex I impairment as a sensitizing background) Mammalian in vivo, conceptual framework The two-hit structure: underlying mitochondrial deficit converts a nontoxic INH concentration into massive hepatocellular injury The most explanatory framework available I did not confirm the specific published model (e.g. an Sod2 heterozygous mouse) in this session; the framework is from the review PMID:24783247
Human patient sera, Acute Liver Failure Study Group Human, ex vivo Anti-INH and anti-CYP2E1/3A4/2C9 antibody response in liver failure Highest — this is the human disease 19 cases; no animal correlate exists PMID:23775837

What no model captures. The idiosyncrasy itself. Every animal model produces injury by dosing hard enough; the human disease appears in 0.5–1% of people taking an ordinary dose. Boelsterli and Lee close their review on exactly this point, verbatim: "points to the existing large gaps in our understanding of the pathogenesis."

Resources. MGI, RGD, Alliance of Genome Resources, Cellosaurus for HepG2 and SNU387. I did not query any of them directly.

Ontology Term Register

Every identifier below was resolved either against this repository's committed term caches or live against OLS4 / the HGNC REST API during this session. Labels are the canonical ontology labels, which in four cases differ from the name a curator would guess.

Watch these four. CHEBI:6030 is labelled "isoniazide". GO:0009449 is labelled "GABA biosynthetic process". UBERON:0000941 is labelled "cranial nerve II". There is no CHEBI term labelled "acetylhydrazine"; the compound is CHEBI:2422 "acetohydrazide", and I did not confirm CHEBI treats it as the same entity.

Domain Terms
Disease MONDO:0027677 isoniazid toxicity
Phenotype HP:0031964, HP:0031956, HP:0012115, HP:0000952, HP:0002904, HP:0006554, HP:0001399, HP:0002240, HP:0002018, HP:0002013, HP:0002039, HP:0000988, HP:0001945, HP:0003493, HP:0009830, HP:0002936, HP:0003401, HP:0100653, HP:0000648, HP:0000709, HP:0001924, HP:0002725, HP:0001250, HP:0002133, HP:0001942, HP:0003128, HP:0001259, HP:0001298, HP:0001289, HP:0001251, HP:0003074, HP:0002153, HP:0003201, HP:0001873, HP:0001903
Biological process / function GO:0004060, GO:0004497, GO:0006805, GO:0008478, GO:0004351, GO:0009449, GO:0042816, GO:0006979, GO:0034599, GO:0006749, GO:0097284, GO:0006915, GO:0006954, GO:0042552, GO:0005739
Cell type CL:0000182, CL:0000091, CL:0000632, CL:0000084, CL:0000617, CL:0002573, CL:0000540
Anatomy UBERON:0002107, UBERON:0000955, UBERON:0001017, UBERON:0000010, UBERON:0000941
Chemical CHEBI:6030, CHEBI:15571, CHEBI:2422, CHEBI:16709, CHEBI:18405, CHEBI:16865, CHEBI:28077
Gene hgnc:7646 NAT2, hgnc:2631 CYP2E1, hgnc:2637 CYP3A4, hgnc:870 ATP7B, hgnc:4932 HLA-B, hgnc:4632 GSTM1, hgnc:4641 GSTT1, hgnc:11180 SOD2, hgnc:8819 PDXK, hgnc:4092 GAD1
Treatment NCIT:C15986 Pharmacotherapy, NCIT:C62619 Pyridoxine, NCIT:C48030 Pyridoxine Hydrochloride, NCIT:C77524 Activated Charcoal, NCIT:C15248 Hemodialysis, NCIT:C15271 Liver Transplantation, NCIT:C15747 Supportive Care, NCIT:C49236 Therapeutic Procedure

What I Did Not Verify

Stated as fact, because it is fact.

  • I did not open the Orphanet record for ORPHA:240887. The site returned a bot-check page. The cross-reference from Mondo is verified; the Orphanet label and content are not.
  • I did not read OMIM directly. It returned HTTP 403. The NAT2 OMIM number 612182 comes from the HGNC REST record, not from OMIM.
  • I did not extract isoniazid-specific counts from the 2023 or 2024 National Poison Data System annual reports.
  • I did not pull gnomAD allele frequencies for NAT2 haplotypes. The population frequency figures in §4.2 came through a search synthesis and should be re-sourced.
  • I did not confirm the specific published mitochondrial-sensitization mouse model behind the complex I two-hit finding.
  • I did not confirm the mouse/human Nat1/Nat2 nomenclature inversion against the Alliance of Genome Resources.
  • I did not find any isoniazid-specific quality-of-life instrument data, single-cell dataset, spatial transcriptomics dataset, or CRISPR screen. I believe none exist; I did not exhaustively establish that.
  • Quotes marked [paraphrase-risk] came back through a summarizing fetch rather than the raw PubMed record. Verify them against the source before putting them in an evidence snippet.

Sources

Reference Validation

Checked with linkml-reference-validator 0.2.1.

Outcome Count
References checked 30
Resolved 30
Unresolved (possible confabulation) 0
Unverifiable 0
Quoted claims checked 2
Quoted claims found in source 2
Quoted claims not found in source 0
References weighed for topical relevance 30
On topic 21
Off topic 2

References that may not be about this subject

These identifiers resolve, so they are not fabrications, but the records they resolve to share almost none of this report's vocabulary. That is a clue and not a verdict - a paper can be relevant in ways its title and abstract do not spell out - so read them before deciding:

  • PMID:39688840 (1 mention) - 2023 Annual Report of the National Poison Data System® (NPDS) from America's Poison Centers®: 41st Annual Report.
  • shared terms: drug
  • PMID:41432769 (1 mention) - 2024 Annual report of the National Poison Data System® (NPDS) from America's Poison Centers(®): 42nd annual report.
  • shared terms: drug

Weighed against this report's own most characteristic terms: isoniazid, liver, toxicity, treatment, injury, drug, pyridoxine, nat2, neuropathy, acute, verbatim, failure, paraphrase-risk, month, livertox, patient, disease, gene, hepatitis, hepatic.

All extracted references resolved successfully. Resolving is not the same as being relevant, though - see the references listed above as possibly off topic.