About the Age, Serum Bilirubin, INR, Serum Creatinine (ABIC) Score for Alcohol-Related Hepatitis
Four routine values — age, bilirubin, INR and creatinine — sort patients with alcohol-related hepatitis into three distinct prognostic groups rather than the single severe/not-severe split most other scores offer. ABIC = (age × 0.1) + (bilirubin × 0.08) + (INR × 0.8) + (creatinine × 0.3), with cut-offs at 6.71 and 9.0. In the derivation cohort those bands corresponded to 90-day survival of 100%, 70% and 25%, and to one-year survival of 97.1%, 64.3% and 33.3%, with an area under the curve of 0.82 for 90-day mortality. The middle band is what distinguishes it: identifying an intermediate-risk group is precisely what a binary threshold like Maddrey's 32 cannot do.
Formula
ABIC = (age × 0.1) + (bilirubin mg/dL × 0.08) + (INR × 0.8) + (creatinine mg/dL × 0.3)- age
- In years, weighted 0.1.
- bilirubin
- Total bilirubin in mg/dL, weighted 0.08.
- INR
- International normalised ratio, weighted 0.8.
- creatinine
- Serum creatinine in mg/dL, weighted 0.3.
- This is a plain linear sum with no logarithms, which makes ABIC unusually easy to compute at the bedside compared with the MELD family.
- Bilirubin and creatinine must be in mg/dL. The coefficients were published for those units, and SI values give a badly wrong score.
- The two cut-offs, 6.71 and 9.00, are the published Kaplan-Meier-derived boundaries — they are not rounded, and rounding them shifts patients between risk groups.
- Age contributes up to about 6–7 points on its own in an older patient, so it can dominate the total; a high score in an elderly patient does not necessarily indicate severe liver injury.
Interpreting the result
A score below 6.71 identified a group in which survival at 90 days was complete in the derivation cohort and 97.1% at one year — a genuinely low-risk group in a disease with a fearsome overall reputation. Between 6.71 and 9.00, roughly 70% survived to 90 days and 64.3% to one year, which is the band where management intensity and the balance of risks around immunosuppression matter most. At 9.00 or above, only about 25% survived 90 days and 33.3% reached one year, and that magnitude of risk should be reflected explicitly in discussions with the patient and family, in early consideration of transplant candidacy where appropriate, and in decisions about ceilings of care. The fact that the same thresholds stratify both 90-day and one-year mortality means a single admission calculation carries information well beyond the index illness — surviving the acute episode does not reset the risk.
| Score | Band | What it means | Action |
|---|---|---|---|
| < 6.71 | Low risk | 100% survival at 90 days and 97.1% at one year in the derivation cohort | Supportive management, nutrition, and abstinence intervention; the prognosis of the acute episode is good |
| 6.71 to < 9.00 | Intermediate risk | 70% survival at 90 days and 64.3% at one year | Active management with close monitoring; assess corticosteroid candidacy and reassess response at day 7 with the Lille model |
| ≥ 9.00 | High risk | 25% survival at 90 days and 33.3% at one year | Escalate care planning, discuss prognosis explicitly, consider transplant candidacy where appropriate, and address ceilings of care |
What the ABIC Score needs (4 inputs)
- Age (years)
- Multiplied by 0.1 — the only fixed, non-modifiable term in the score, and the reason older patients start from a worse baseline.
- Total bilirubin (mg/dL)
- Multiplied by 0.08. Because bilirubin values in alcohol-related hepatitis are often high, this small coefficient still contributes substantially.
- INR
- Multiplied by 0.8 — the largest coefficient, reflecting hepatic synthetic failure.
- Creatinine (mg/dL)
- Multiplied by 0.3, capturing the renal dysfunction that drives much of the mortality in this condition.
Units. ABIC's coefficients assume mg/dL. Convert first if your laboratory reports SI units: bilirubin µmol/L ÷ 17.1 = mg/dL; creatinine µmol/L ÷ 88.4 = mg/dL. This calculator takes mg/dL directly.
What it returns
- ABIC score
- A continuous value, typically between about 5 and 12.
- Risk group
- Low (< 6.71), intermediate (6.71 to < 9.00) or high (≥ 9.00) risk of death at 90 days and one year.
How it is calculated
Dominguez and colleagues prospectively enrolled patients with biopsy-proven alcohol-related hepatitis between 2000 and 2006 and evaluated biochemical, clinical, portal haemodynamic and histological parameters against 90-day mortality using Cox regression. Notably, the haemodynamic and histological variables did not survive into the final model: age, bilirubin, creatinine and INR independently predicted death, and the score was built from admission values alone. Kaplan-Meier analysis then identified two cut-offs, 6.71 and 9.0, that separated three groups with clearly different survival, and the same boundaries were found to stratify one-year mortality as well as 90-day. The model was then confirmed in an independent prospective cohort.
Facts & figures
| ABIC | Risk group | 90-day survival | One-year survival |
|---|---|---|---|
| < 6.71 | Low | 100% | 97.1% |
| 6.71 to < 9.00 | Intermediate | 70% | 64.3% |
| ≥ 9.00 | High | 25% | 33.3% |
Area under the curve for 90-day mortality was 0.82. The same cut-offs stratified one-year mortality, so the score's prognostic reach extends beyond the acute episode.
| Score | Timing | Output |
|---|---|---|
| Maddrey's DF | Admission | Binary: severe at ≥ 32 |
| GAHS | Admission | Steroid benefit at ≥ 9 |
| ABIC | Admission | Three risk tiers (6.71, 9.0) |
| Lille | Day 7 of steroids | Binary: non-responder at ≥ 0.45 |
ABIC and Lille are complementary rather than competing — one is static and taken at baseline, the other dynamic and taken after a week of treatment. Combining a static score with Lille predicted survival better than either alone.
Evidence
Derivation — Dominguez
2008 · n = 103103 patients with biopsy-proven alcohol-related hepatitis prospectively enrolled between 2000 and 2006, with biochemical, clinical, portal haemodynamic and histological parameters assessed by Cox regression against 90-day mortality. Age, bilirubin, creatinine and INR were the independent predictors retained.
Area under the curve 0.82 for 90-day mortality. Cut-offs of 6.71 and 9.0 identified low, intermediate and high-risk groups with 90-day survival of 100%, 70% and 25%, and the same thresholds also stratified one-year mortality (97.1%, 64.3% and 33.3%).
Confirmatory cohort — Dominguez
2008 · n = 80An independent prospective cohort of patients with alcohol-related hepatitis, used to confirm that the ABIC cut-offs reproduced the same three-tier stratification of 90-day and one-year mortality.
Reproduced the risk stratification found in the derivation cohort, supporting ABIC as a tool for prognostic stratification at both 90 days and one year.
Joint static-dynamic model — Louvet
2015 · n = 538Patients with severe alcohol-related hepatitis treated with corticosteroids, drawn from databases in France and the United Kingdom (derivation n = 538), with validation in trial cohorts from the United States, France, Korea and Belgium (n = 604). Three joint-effect models were compared: Maddrey+Lille, MELD+Lille and ABIC+Lille.
Combining a static score such as ABIC with the dynamic Lille model predicted survival at two and six months significantly better than either static or dynamic models alone, in both derivation and validation cohorts.
How it compares
ABIC Score vs Maddrey's Discriminant Function
Maddrey's DF answers whether a patient has severe disease at all; ABIC answers how severe, splitting that severe group into three tiers with 90-day survival of 100%, 70% and 25%.
Maddrey's discriminant function exists for one purpose — the threshold of 32 that opens the corticosteroid conversation — and it does that job with two laboratory values. ABIC uses four and produces a gradient instead of a line, which is what makes it useful once a patient is already known to be severe. They are best used in sequence rather than as alternatives: Maddrey's DF to establish severity and steroid candidacy, ABIC to describe the risk the patient actually faces.
ABIC Score vs Lille model
ABIC and Lille answer different questions at different times — ABIC stratifies baseline risk at admission, Lille measures corticosteroid response at day 7 — and combining a static score with Lille predicts survival better than either alone.
ABIC is static: it uses admission values and cannot register whether treatment is working. Lille is dynamic, requiring both a day-0 and a day-7 bilirubin, and exists specifically to identify non-responders early. Louvet and colleagues tested joint-effect models pairing Lille with Maddrey's DF, MELD and ABIC across a derivation cohort of 538 and a validation cohort of 604, and found the combinations predicted two- and six-month survival significantly better than static or dynamic models used in isolation.
ABIC Score vs Glasgow Alcoholic Hepatitis Score
GAHS is built to predict who benefits from corticosteroids; ABIC is built to predict who survives — overlapping inputs, but they are answering different clinical questions.
GAHS combines age, white cell count, urea, prothrombin time ratio and bilirubin, and its threshold of 9 identifies patients in whom steroids improved survival while those below it derived no benefit. ABIC shares age, bilirubin and a coagulation measure but substitutes creatinine for urea and white cell count, and its output is a three-tier mortality estimate rather than a treatment-benefit signal. Using both is reasonable: GAHS to inform whether to give steroids, ABIC to describe the prognosis irrespective of that decision.
Pearls & pitfalls
- Age is the term you cannot treat. In an older patient it can push the score toward the intermediate band on its own, so a raised ABIC is not automatically evidence of severe liver injury — inspect the components.
- Bilirubin and creatinine must be in mg/dL; SI units with the published coefficients produce a wildly incorrect score.
- The cut-offs are 6.71 and 9.00 exactly. Rounding 6.71 to 6.7 or 7 moves real patients between risk groups.
- ABIC is a baseline score using admission values. It does not measure treatment response, and recalculating it at day 7 is not a substitute for the Lille model.
- It presumes the diagnosis. Alcohol-related hepatitis is a clinical diagnosis with important mimics — biliary obstruction, sepsis, drug-induced liver injury — and ABIC will happily score any of them.
- The same thresholds predict one-year as well as 90-day mortality, so a patient discharged after surviving the acute illness remains in the risk group the admission score assigned.
Critical actions
- Confirm the diagnosis of alcohol-related hepatitis before acting on the score, and exclude biliary obstruction, sepsis and drug-induced liver injury.
- Use ABIC to stratify baseline risk, then reassess corticosteroid response at day 7 with the Lille model rather than repeating ABIC.
- Treat a high-risk score as a trigger for explicit prognostic conversation with the patient and family, and for early consideration of transplant candidacy where the patient may be eligible.
- Address alcohol use directly in every risk group — abstinence is the single strongest determinant of long-term survival and is not captured anywhere in the score.
- Attend to nutrition, infection screening and thiamine replacement, none of which the score measures but all of which alter outcome.
Why this score exists
The authors stated two aims — improving clinical management and facilitating clinical trials — and the second explains the design as much as the first. A binary severity threshold makes trial recruitment crude: every patient above Maddrey's 32 is enrolled as though equivalent, when their actual mortality risk spans from moderate to overwhelming, which dilutes any treatment signal. A score that separates three tiers allows stratified randomisation and more honest subgroup analysis. It is also telling what the derivation discarded: portal haemodynamic and histological variables were measured and assessed, and did not earn a place, meaning a liver biopsy and hepatic venous pressure measurement added nothing to four numbers available from a routine admission blood panel.
About the creator
First author, 2008 derivation study
Derived the ABIC score in a prospective cohort of biopsy-proven alcohol-related hepatitis, identifying age, bilirubin, INR and creatinine as the independent predictors of 90-day mortality.
Senior author
Led the Barcelona group's programme of work on alcohol-related liver disease from which the score emerged.
Limitations
- The derivation cohort was small — 103 biopsy-proven patients from a single prospective series — so the precise survival figures attached to each band should be read as estimates rather than exact probabilities.
- Age contributes a fixed, unmodifiable share of the score, which means the bands blend liver-disease severity with chronological age and can misrepresent a young patient with fulminant disease or an older patient with milder injury.
- It uses admission values only and carries no information about treatment response, so it cannot guide the day-7 decision that most changes management.
- It contains no measure of infection, encephalopathy or gastrointestinal bleeding, all of which drive mortality in this population.
- Abstinence — the dominant determinant of survival beyond the acute episode — is not in the model, despite the score being used to predict one-year mortality.
- The published survival figures come from a cohort recruited between 2000 and 2006, before current practice in early transplant assessment for alcohol-related hepatitis.
If you are the patient
The ABIC score helps doctors estimate how serious an episode of alcohol-related hepatitis is, and what the outlook is over the next three months and the next year. It uses your age together with three blood results: bilirubin (which reflects jaundice), INR (a clotting test) and creatinine (which reflects kidney function). The score puts people into one of three groups — lower, middle or higher risk — rather than simply 'mild' or 'severe', which helps the team judge how intensive treatment needs to be and what to expect. Because your age counts toward the score, an older person can have a higher number without their liver being more damaged. The score is worked out when you are admitted; if you are given steroid treatment, a different score called the Lille model is used after a week to check whether it is helping. Whatever the score, stopping drinking completely is the single most important thing that improves long-term survival, and that is not something the score can measure.
Frequently asked questions
What is the ABIC score?#
A prognostic score for alcohol-related hepatitis calculated as (age × 0.1) + (bilirubin in mg/dL × 0.08) + (INR × 0.8) + (creatinine in mg/dL × 0.3). It stratifies risk of death at both 90 days and one year into three groups.
What are the ABIC score cut-offs?#
6.71 and 9.00. Below 6.71 is low risk, 6.71 to under 9.00 intermediate, and 9.00 or above high risk. The values are not rounded, and rounding them reassigns patients between groups.
What survival do the ABIC bands predict?#
In the derivation cohort, 90-day survival was 100% in the low-risk group, 70% in the intermediate group and 25% in the high-risk group. One-year survival was 97.1%, 64.3% and 33.3% respectively.
How does ABIC differ from Maddrey's discriminant function?#
Maddrey's DF is binary — a value of 32 or above defines severe disease and opens the corticosteroid conversation. ABIC produces three risk tiers, so it grades severity within the group Maddrey's DF has already flagged as severe.
Can ABIC tell whether corticosteroids are working?#
No. It uses admission values only and is a static baseline score. The Lille model, calculated after seven days of corticosteroids, is the instrument for assessing treatment response — and combining a static score such as ABIC with Lille predicts survival better than either alone.
Which units does ABIC need?#
Bilirubin and creatinine must be in mg/dL, matching the units the coefficients were published for. Using µmol/L values will produce a grossly inflated score.
References
Original / primary reference
Validation and comparison
- Louvet A, Labreuche J, Artru F, et al. Combining data from liver disease scoring systems better predicts outcomes of patients with alcoholic hepatitis. Gastroenterology. 2015;149(2):398-406.
- Thursz MR, Richardson P, Allison M, et al. Prednisolone or pentoxifylline for alcoholic hepatitis. N Engl J Med. 2015;372(17):1619-1628.