About the King's College Criteria for Acute Liver Failure
Two unrelated criteria sets share this name, and applying the wrong branch is the commonest error made with it. In paracetamol-induced acute liver failure the criteria are met by an arterial pH below 7.30 after fluid resuscitation, or by all three of INR above 6.5, creatinine above 300 µmol/L and grade III–IV encephalopathy together. In every other aetiology they are met by an INR above 6.5 alone, or by any three of five factors: age under 11 or over 40, an unfavourable aetiology, a jaundice-to-encephalopathy interval beyond 7 days, INR above 3.5, and bilirubin above 300 µmol/L. Meeting them is a trigger to contact a transplant centre; not meeting them excludes very little, because pooled sensitivity in paracetamol failure is only about 58%.
Formula
Paracetamol: arterial pH < 7.30 (post-resuscitation) OR (INR > 6.5 AND creatinine > 300 µmol/L AND encephalopathy grade III–IV)
Non-paracetamol: INR > 6.5 OR any 3 of {age < 11 or > 40; seronegative/drug/halothane aetiology; jaundice-to-encephalopathy > 7 days; INR > 3.5; bilirubin > 300 µmol/L}- Arterial pH
- Below 7.30, measured after adequate fluid resuscitation. Sufficient on its own in the paracetamol branch, regardless of encephalopathy grade.
- INR
- Above 6.5 corresponds to the original paper's prothrombin time above 100 seconds; above 3.5 corresponds to prothrombin time above 50 seconds. The paper was written in prothrombin-time seconds and the INR equivalents are conventional conversions, not values the authors published.
- Creatinine
- Above 300 µmol/L (≈ 3.4 mg/dL). Paracetamol branch only.
- Encephalopathy grade
- Grade III or IV on the West Haven scale. Paracetamol triad only; the pH route does not require encephalopathy at all.
- Bilirubin
- Above 300 µmol/L (≈ 17.5 mg/dL). Non-paracetamol branch only.
- Every threshold is a STRICT inequality. An INR of exactly 6.5, an age of exactly 40, an interval of exactly 7 days and a pH of exactly 7.30 all fail to meet their respective criteria.
- O'Grady 1989 is a UK paper and states its thresholds in SI units, so the conventional-unit equivalents are derived from the SI figures rather than the reverse. This matters for bilirubin: 17.5 mg/dL is 299 µmol/L and therefore falls just short of the published > 300 µmol/L. This calculator tests against the SI value.
- The paracetamol triad is conjunctive — all three components must be present. Two of the three is not a partial result; it does not meet the criteria at all.
- The pH criterion is worthless without the resuscitation qualifier. Hypovolaemic patients are frequently acidotic on arrival and correct with fluid alone.
- The lactate criterion (> 3.5 mmol/L at ~4 hours, > 3.0 at ~12 hours) is from Bernal 2002 and is not part of the original criteria; the calculator reports it separately.
Interpreting the result
Meeting the criteria means the patient's predicted mortality without transplantation is high enough that transplantation is indicated, and the correct response is an immediate conversation with a transplant centre rather than continued observation. Not meeting them means considerably less than clinicians often assume. Pooled sensitivity in paracetamol-induced failure is around 58% against a specificity of about 89%, so roughly two in five patients who go on to die will not have met the criteria at the point they were checked. The criteria are therefore a one-way test: a positive result is decisive, a negative result is provisional and must be repeated. Any patient with acute liver failure and encephalopathy warrants transplant-centre discussion regardless of what this calculator returns.
| Score | Band | What it means | Action |
|---|---|---|---|
| Paracetamol — pH < 7.30 post-resuscitation | Criteria met | In the derivation cohort this identified a group with roughly 85–95% mortality without transplantation; it is the strongest single indicator in the paracetamol branch | Immediate transplant-centre referral; continue N-acetylcysteine and critical-care support |
| Paracetamol — INR > 6.5 AND creatinine > 300 µmol/L AND grade III–IV encephalopathy | Criteria met | Predicted mortality without transplantation of the same order as the pH route | Immediate transplant-centre referral; all three components must be present together |
| Non-paracetamol — INR > 6.5, or any 3 of the 5 factors | Criteria met | Predicted mortality without transplantation above 90% in the derivation cohort for the three-of-five group | Immediate transplant-centre referral alongside aetiology-specific treatment |
| Neither branch satisfied | Criteria not met | Does not indicate a good prognosis — pooled sensitivity is about 58% in paracetamol failure, so a large minority of eventual deaths are missed at any single assessment | Reassess at least twice daily; refer on clinical grounds anyway if encephalopathy is present or INR is rising |
What the King's College Criteria needs (10 inputs)
- Cause of acute liver failure
- Paracetamol (acetaminophen) or any other cause. This selects the entire criteria set — the two branches share no threshold except INR above 6.5, so choosing the wrong one applies the wrong rules rather than merely shifting a cut-off.
- Arterial pH (paracetamol branch)
- Measured after adequate fluid resuscitation. An acidosis that corrects with volume does not count; this is the single most important qualifier in the whole score and the one most often ignored. Optional here, because the triad can still be assessed without an arterial gas.
- INR
- Used in both branches, at different thresholds. Above 6.5 (prothrombin time above 100 seconds in the original paper) in either branch; above 3.5 (prothrombin time above 50 seconds) additionally counts as one of the five non-paracetamol factors.
- Serum creatinine (paracetamol branch)
- Threshold is above 300 µmol/L, about 3.4 mg/dL. Part of the three-part triad, which only counts when all three components are present together.
- Hepatic encephalopathy grade (paracetamol branch)
- Only grade III (somnolent but rousable, marked confusion) or grade IV (comatose) counts. Grade II does not, however unwell the patient appears.
- Arterial lactate (paracetamol branch, optional)
- Not part of O'Grady's 1989 criteria. Above 3.5 mmol/L after early resuscitation (~4 hours) or above 3.0 mmol/L after full resuscitation (~12 hours), from Bernal 2002. Reported separately rather than folded into the original set.
- Age (non-paracetamol branch)
- Counts as a factor if under 11 or over 40 years. The original paper says 'less than 11'; a great many secondary sources render this as 'under 10', which wrongly excludes a 10-year-old.
- Specific aetiology (non-paracetamol branch)
- The unfavourable group is seronegative (non-A non-B) hepatitis, idiosyncratic drug reactions and halothane hepatitis. Hepatitis A and hepatitis B carried better spontaneous survival in the derivation cohort and do not count as a factor.
- Jaundice-to-encephalopathy interval (non-paracetamol branch)
- Days from onset of jaundice to onset of encephalopathy; counts if longer than 7 days. A longer interval is worse, which is the opposite of most clinicians' intuition and worth pausing over.
- Serum bilirubin (non-paracetamol branch)
- Counts as a factor above 300 µmol/L, about 17.5 mg/dL.
Units. The original paper is British and states creatinine and bilirubin in µmol/L, so this calculator treats the SI figures as authoritative and derives the mg/dL equivalents from them. Creatinine above 300 µmol/L is about 3.4 mg/dL and the two conventions agree in practice. Bilirubin is the one to watch: 300 µmol/L is 17.54 mg/dL, so the commonly quoted '17.5 mg/dL' is fractionally below the published threshold. Coagulopathy is stated as prothrombin time in seconds in the source (> 100 s and > 50 s); the INR equivalents of 6.5 and 3.5 used here and everywhere else are conventional conversions.
What it returns
- Criteria met or not met
- A binary result, not a score. There is no gradation and no probability attached — the criteria either identify the patient as high-mortality-without-transplant or they do not.
- Which route was satisfied
- Whether the criteria were met by the pH route, the triad, INR alone, or three of five. This matters for documentation and for the transplant-centre conversation.
- Lactate criterion (paracetamol only)
- Reported separately, since it is a later addition rather than part of the original criteria and some units use it and some do not.
How it is calculated
O'Grady and colleagues analysed 588 patients with acute liver failure managed medically at King's College Hospital between 1973 and 1985, then tested the resulting indicators on a further 175 patients admitted in 1986–87. The analysis was done separately for paracetamol and non-paracetamol aetiologies because the two behave differently: paracetamol failure is a hyperacute injury in which the liver either regenerates quickly or does not, so the variables that predict death are the ones reflecting acute physiological collapse — acidosis, renal failure and profound coagulopathy. Non-paracetamol failure evolves over a longer course, so its predictors include static features known at presentation (age, aetiology) alongside the dynamic laboratory values, and the pace of the illness itself enters as the jaundice-to-encephalopathy interval. A slower onset predicts worse survival there because the subacute presentations, particularly seronegative hepatitis, regenerate poorly. The models were explicitly constructed for selecting patients for transplantation, which is why they were tuned to be specific rather than sensitive.
Facts & figures
| Paracetamol | Non-paracetamol | |
|---|---|---|
| Sufficient alone | Arterial pH < 7.30 after resuscitation | INR > 6.5 (PT > 100 s) |
| Combination route | ALL of: INR > 6.5, creatinine > 300 µmol/L, encephalopathy grade III–IV | ANY 3 of 5 factors |
| Age | Not used | < 11 or > 40 years |
| Aetiology | Not used (defines the branch) | Seronegative, drug reaction or halothane |
| Timing | Not used | Jaundice to encephalopathy > 7 days |
| Bilirubin | Not used | > 300 µmol/L (≈ 17.5 mg/dL) |
| Encephalopathy required? | Only for the triad, not for the pH route | Not a listed factor |
The only threshold common to both branches is INR > 6.5. Everything else differs, which is why the aetiology question has to be answered before any other input is meaningful.
| Population | Instrument | Sensitivity | Specificity |
|---|---|---|---|
| Paracetamol-associated ALF | King's College Criteria | 58% (95% CI 51–65) | 89% (95% CI 85–93) |
| Paracetamol-associated ALF | MELD | 80% (95% CI 74–86) | 53% (95% CI 47–59) |
The trade-off is the point rather than a flaw: for a decision that commits a potentially self-recovering patient to lifelong immunosuppression, the high-specificity instrument is the appropriate one. The meta-analysis concluded that MELD should not replace the King's criteria in paracetamol failure, while MELD may have a role in non-paracetamol failure.
Evidence
Derivation — O'Grady et al., King's College Hospital
1989 · n = 588588 patients with acute liver failure managed medically at a single tertiary liver unit between 1973 and 1985, analysed separately by aetiology. For paracetamol-induced failure, arterial pH, peak prothrombin time and serum creatinine were the variables associated with survival. For viral hepatitis and drug reactions, five variables emerged: aetiology (non-A non-B hepatitis or drug reaction), age under 11 or over 40, jaundice-to-encephalopathy interval beyond 7 days, serum bilirubin above 300 µmol/L, and prothrombin time above 50 seconds.
Indicators were tested retrospectively in a further 175 patients admitted during 1986–1987, and models were then constructed explicitly for the selection of patients for liver transplantation.
Meta-analysis — McPhail et al.
2016 · n = 2,153Systematic review and meta-analysis of 23 studies comprising 2,153 patients with acute liver failure, comparing the King's College Criteria against MELD for the prediction of hospital mortality.
In paracetamol-associated ALF the criteria showed 58% sensitivity (95% CI 51–65%) and 89% specificity (95% CI 85–93%), with a diagnostic odds ratio of 5.3; MELD showed 80% sensitivity and 53% specificity. The authors concluded the criteria more accurately predict mortality in paracetamol failure and MELD in non-paracetamol failure, while cautioning that neither is optimal for all patients.
Blood lactate as an early predictor — Bernal et al.
2002Cohort study at the same unit examining whether arterial lactate identifies poor-prognosis paracetamol-induced acute liver failure earlier than the original criteria, measured both after early (~4 h) and after full (~12 h) fluid resuscitation.
Lactate above 3.5 mmol/L after early resuscitation, or above 3.0 mmol/L after full resuscitation, identified patients with a poor prognosis and did so earlier than the original criteria; this became the 'modified' King's criteria used by some units.
Systematic review of prognostic tests — Craig et al.
2010Systematic review of prognostic tests in paracetamol-induced acute liver failure, assessing the King's College Criteria alongside the alternatives proposed to improve on them.
Confirmed the criteria's high specificity and limited sensitivity, and found that no proposed alternative had accumulated enough evidence to displace them in routine practice.
How it compares
King's College Criteria vs MELD / MELD-Na
Use the King's criteria for paracetamol-induced failure and consider MELD for non-paracetamol failure — the meta-analysis is explicit that MELD should not replace the criteria in paracetamol cases.
The two instruments make opposite trade-offs. In paracetamol-associated acute liver failure the King's criteria are 58% sensitive and 89% specific, while MELD is 80% sensitive and 53% specific. For a decision that commits a potentially self-recovering young patient to lifelong immunosuppression, specificity is what matters, which is why the criteria remain the standard there. In non-paracetamol failure the pooled figures favour MELD, and MELD is also the instrument used for chronic liver disease listing, so the two coexist rather than compete.
King's College Criteria vs CLIF-SOFA
Different patients entirely — the King's criteria are for acute liver failure in a previously healthy liver, CLIF-SOFA for organ failure in established cirrhosis.
Acute liver failure and acute-on-chronic liver failure are separate syndromes with separate prognostic literatures, and the distinction is the presence of pre-existing chronic liver disease rather than the severity of the presentation. Applying the King's criteria to a decompensated cirrhotic is a category error: the criteria were derived exclusively in patients without known chronic liver disease, and nothing in them was validated in cirrhosis. CLIF-SOFA and CLIF-C ACLF occupy that space.
King's College Criteria vs Clichy criteria
The main European alternative, built around factor V and age rather than pH and creatinine; neither has been shown clearly superior, and the King's criteria are more widely used.
The Clichy (or Villejuif) criteria were derived in French patients with fulminant viral hepatitis and use hepatic encephalopathy combined with a factor V level below 20% under age 30, or below 30% at 30 and over. They require a factor V assay that many laboratories do not run routinely, which is a practical reason the King's criteria travelled further. Comparative studies have not established a decisive advantage for either, and some units check both.
King's College Criteria vs Child-Pugh score
Not comparable — Child-Pugh grades chronic liver disease severity and has no role in acute liver failure.
Child-Pugh includes ascites and encephalopathy alongside bilirubin, albumin and INR, and was derived in cirrhotic patients undergoing portosystemic shunt surgery. Albumin has a half-life of around 20 days, which makes it close to meaningless in an illness that evolves over hours to days, and the score has no validated application in acute liver failure.
Pearls & pitfalls
- Answer the aetiology question first and get it right. The two branches share only one threshold (INR > 6.5); applying the paracetamol triad to a patient with seronegative hepatitis is not a minor approximation, it is the wrong test.
- The pH criterion requires adequate fluid resuscitation beforehand. An acidotic, hypovolaemic patient on arrival who corrects with volume does not meet it, and treating the admission gas as the answer will over-call the criteria.
- Grade II encephalopathy does not count towards the triad. Only grade III or IV does.
- The paracetamol triad is all-or-nothing. INR above 6.5 plus a creatinine above 300 µmol/L, with grade II encephalopathy, does not meet the criteria at all — there is no partial credit.
- The age factor is 'under 11', not 'under 10'. The original paper says less than 11, and the widely reproduced 'under 10' version silently excludes 10-year-olds.
- A longer jaundice-to-encephalopathy interval is worse, not better. Beyond 7 days counts as an adverse factor, because subacute presentations regenerate poorly.
- Bilirubin's threshold is 300 µmol/L. Because 17.5 mg/dL is only 299 µmol/L, the conventional-unit rounding sits fractionally below the published cut-off — this calculator tests the SI value.
- A negative result is not reassurance. Sensitivity in paracetamol failure is about 58%; refer anyone with encephalopathy regardless, and repeat the assessment at least twice daily.
- Do not correct the INR with fresh frozen plasma unless there is bleeding or a planned procedure. It removes the single most informative trend from the assessment without improving outcome.
Critical actions
- Contact the regional liver transplant centre as soon as the criteria are met — this is a same-hour action, not something to review at the next ward round.
- Discuss early even when the criteria are not met, if encephalopathy is present, INR is climbing, or there is acidosis. The criteria lag the clinical picture and miss a substantial minority.
- Continue N-acetylcysteine in paracetamol-induced failure regardless of the time since ingestion and regardless of whether the criteria are met.
- Measure arterial pH only after adequate fluid resuscitation, and document the resuscitation status alongside the value.
- Reassess at least twice daily. In non-paracetamol failure the jaundice-to-encephalopathy interval alone can convert a negative result to a positive one with no change in any laboratory value.
- Pursue and treat the aetiology in parallel — antivirals for hepatitis B, steroids for autoimmune hepatitis, delivery in pregnancy-related failure — since transplant assessment does not replace specific treatment.
- Manage the complications that determine outcome independently of the score: raised intracranial pressure, hypoglycaemia, and infection.
Why this score exists
The criteria were built for a specific institutional problem rather than as a general prognostic exercise: King's College Hospital was receiving acute liver failure referrals from across the UK and had to decide, quickly and with limited organs, which patients would die without a transplant. That framing explains the design choice that most often frustrates users — the deliberate preference for specificity over sensitivity. The authors were more willing to miss a patient who would go on to die than to transplant one who would have recovered spontaneously, because in paracetamol overdose the recovering patient is frequently young and otherwise well, and an unnecessary graft costs them a lifetime of immunosuppression. Reading a negative result as reassurance inverts the intent of the instrument.
About the creator
First author, 1989 derivation study, Institute of Liver Studies, King's College Hospital
Led the analysis of 588 medically managed acute liver failure patients from which both criteria sets are drawn.
Co-author of the 1989 derivation study.
Co-author of the 1989 derivation study; performed the statistical analysis.
Senior author, Institute of Liver Studies, King's College Hospital
Directed the liver unit whose cohort produced the criteria.
Limitations
- Sensitivity is limited by design. Around 58% in paracetamol-associated failure, meaning a large minority of patients who die never meet the criteria at the moment they are assessed.
- Derived at a single tertiary centre between 1973 and 1985, before modern intensive care, routine N-acetylcysteine use and current transplant practice — all of which have improved spontaneous survival and may have shifted the criteria's calibration.
- The original thresholds are prothrombin-time seconds, not INR. The INR equivalents in universal use (6.5 and 3.5) are conventional conversions rather than values the authors published or validated, and prothrombin time reagents vary between laboratories.
- The non-paracetamol branch bundles heterogeneous aetiologies — seronegative hepatitis, autoimmune hepatitis, Wilson disease, Budd-Chiari, pregnancy-related failure — that have markedly different natural histories, into one rule set.
- Not validated in acute-on-chronic liver failure, in paediatric acute liver failure (where dedicated criteria and the paediatric listing framework apply), or in Wilson disease, for which a specific prognostic index exists.
- Says nothing about transplant candidacy. A patient can meet the criteria and still be unsuitable for transplantation on grounds of sepsis, irreversible neurological injury, or comorbidity — the criteria identify need, not suitability.
- The lactate addition improves early identification but is not part of the original criteria, and practice varies on whether to act on it.
If you are the patient
The King's College Criteria are a checklist that liver doctors use in sudden, severe liver failure to judge whether someone is likely to survive with medical treatment alone or is likely to need a liver transplant to survive. Which checklist is used depends on the cause: there is one version for liver failure caused by paracetamol overdose and a different version for every other cause. If the criteria are met, it means the medical team believes the chance of recovering without a transplant is low, and they will contact a transplant centre urgently. If the criteria are not met, that is genuinely reassuring only in part — the checklist was deliberately designed to be cautious about recommending a transplant, because a transplant means taking anti-rejection medication for life, and it misses some people who turn out to need one. That is why the team will keep repeating the assessment, often several times a day, and may refer to a transplant centre anyway based on how someone looks rather than purely on the checklist.
Frequently asked questions
What are the King's College Criteria?#
Prognostic criteria that identify patients with acute liver failure whose mortality without a liver transplant is high enough to justify urgent transplant referral. There are two separate sets: one for paracetamol (acetaminophen) poisoning and one for all other causes.
What are the King's College Criteria for paracetamol overdose?#
An arterial pH below 7.30 after adequate fluid resuscitation, on its own; or all three of INR above 6.5, serum creatinine above 300 µmol/L (about 3.4 mg/dL), and grade III or IV hepatic encephalopathy occurring together. Two out of the three does not count.
What are the non-paracetamol King's College Criteria?#
An INR above 6.5 on its own, or any three of these five: age under 11 or over 40; an aetiology of seronegative (non-A non-B) hepatitis, idiosyncratic drug reaction or halothane hepatitis; more than 7 days from jaundice to encephalopathy; INR above 3.5; and bilirubin above 300 µmol/L (about 17.5 mg/dL).
Does not meeting the criteria mean the patient will survive?#
No, and this is the most important caveat. Pooled sensitivity in paracetamol-associated failure is about 58%, so roughly two in five patients who go on to die do not meet the criteria when assessed. The criteria were deliberately built to be specific rather than sensitive. Refer any patient with acute liver failure and encephalopathy regardless, and repeat the assessment at least twice daily.
Is it age under 10 or under 11?#
Under 11. O'Grady's 1989 paper states 'age less than 11 years', though a large number of secondary sources and calculators reproduce it as 'under 10', which wrongly excludes a 10-year-old. This calculator uses the published threshold.
Where does arterial lactate fit in?#
Lactate is not part of the original 1989 criteria. Bernal and colleagues showed in 2002 that a lactate above 3.5 mmol/L after early fluid resuscitation, or above 3.0 mmol/L after full resuscitation, identifies poor-prognosis paracetamol patients earlier than the original criteria do. Some units use this as a 'modified' criterion; this calculator reports it separately rather than folding it in.
Why is a longer jaundice-to-encephalopathy interval worse?#
Because it marks a subacute rather than hyperacute course. Hyperacute failure — typically paracetamol — has a high rate of spontaneous liver regeneration despite looking dramatic. Subacute presentations, particularly seronegative hepatitis, evolve slowly and regenerate poorly, so a slower illness carries a worse prognosis even though it appears less alarming.
Can the King's College Criteria be used in cirrhosis?#
No. They were derived exclusively in patients without known chronic liver disease. A decompensated cirrhotic patient has acute-on-chronic liver failure, a different syndrome with its own prognostic tools such as CLIF-SOFA and CLIF-C ACLF, and MELD-based listing.
Should INR be corrected with fresh frozen plasma before applying the criteria?#
Not unless there is active bleeding or a procedure is planned. Correcting the INR removes the most informative trend available for monitoring these patients and has not been shown to improve outcome, while making the criteria uninterpretable.
References
Original / primary reference
Validation and later modification
- McPhail MJW, Farne H, Senvar N, Wendon JA, Bernal W. Ability of King's College Criteria and Model for End-Stage Liver Disease Scores to Predict Mortality of Patients With Acute Liver Failure: A Meta-analysis. Clin Gastroenterol Hepatol. 2016;14(4):516-525.e5.
- Bernal W, Donaldson N, Wyncoll D, Wendon J. Blood lactate as an early predictor of outcome in paracetamol-induced acute liver failure: a cohort study. Lancet. 2002;359(9306):558-563.
- Craig DGN, Ford AC, Hayes PC, Simpson KJ. Systematic review: prognostic tests of paracetamol-induced acute liver failure. Aliment Pharmacol Ther. 2010;31(10):1064-1076.
Guidelines
- Shingina A, Mukhtar N, Wakim-Fleming J, et al. Acute Liver Failure Guidelines. Am J Gastroenterol. 2023;118(7):1128-1153.
- Lee WM, Stravitz RT, Larson AM. Introduction to the revised American Association for the Study of Liver Diseases position paper on acute liver failure 2011. Hepatology. 2012;55(3):965-967.