About the Albumin-Bilirubin (ALBI) Grade
Child-Pugh asks you to judge how much ascites there is and how encephalopathic the patient looks; ALBI asks for nothing but albumin and bilirubin. It grades liver function in hepatocellular carcinoma from those two laboratory values as (log₁₀ bilirubin in µmol/L × 0.66) + (albumin in g/L × −0.085), and splits the result at −2.60 and −1.39 into three grades. Its practical value is not that it beats Child-Pugh overall — it performs about as well — but that it divides Child-Pugh grade A, where most patients with hepatocellular carcinoma present, into two groups with clearly different survival.
Formula
ALBI = (log₁₀ bilirubin [µmol/L] × 0.66) + (albumin [g/L] × −0.085)- bilirubin [µmol/L]
- Total bilirubin in µmol/L. Multiply mg/dL by 17.1 to convert. Using mg/dL directly is the single most common error and produces a wrongly favourable score.
- albumin [g/L]
- Albumin in g/L, not g/dL. Multiply g/dL by 10. A value of 3.5 g/dL is 35 g/L, and entering 3.5 shifts the score by roughly 2.7 points — more than the entire width of grade 2.
- 0.66 and −0.085
- Coefficients fitted in the derivation cohort. Note the asymmetry of sign: bilirubin pushes the score up, albumin pulls it down, and the score is negative across the normal range as a result.
- Grade boundaries are −2.60 and −1.39. Both are exact published values, not rounded conveniences, and grade 2 is bounded on both sides.
- The log₁₀ term means the score compresses at high bilirubin. Beyond about 10 mg/dL, further rises move the score comparatively little, so ALBI discriminates poorly at the jaundiced end.
- No cap or floor is applied to either input in the published model.
- ALBI is calculated identically regardless of the presence of ascites or encephalopathy. That is deliberate — but it means a patient with refractory ascites and preserved laboratory values can be ALBI grade 1, which is a real limitation rather than a scoring quirk.
Interpreting the result
Read the grade against the treatment being considered. Grade 1 indicates the best-preserved liver function and the widest set of options, including resection and repeated locoregional therapy; grade 3 indicates poor hepatic reserve, where aggressive therapy is likely to precipitate decompensation and the discussion shifts toward best supportive care or, where the tumour permits, transplant assessment. Grade 2 is the working middle where most decisions actually get made, and the score's continuous value is worth reading inside it — a patient at −2.5 is not the same as one at −1.5. The published median survival figures span wide ranges across the derivation and validation cohorts, from 18.5 to 85.6 months in grade 1, 5.3 to 46.5 months in grade 2, and 2.3 to 15.5 months in grade 3, because those cohorts differed in tumour stage and treatment era. Read them as a rank ordering, not as a prognosis for an individual — the tumour, not the liver, is often what determines survival at any given grade.
| Score | Band | What it means | Action |
|---|---|---|---|
| ≤ −2.60 | ALBI Grade 1 | Best-preserved liver function. Median survival 18.5–85.6 months across the derivation and validation cohorts | Widest range of options — resection, ablation and repeated locoregional therapy are all reasonable on hepatic grounds |
| > −2.60 to ≤ −1.39 | ALBI Grade 2 | Intermediate liver function. Median survival 5.3–46.5 months | Treat, but weigh hepatic cost against tumour benefit; read the continuous score within the band rather than the grade alone |
| > −1.39 | ALBI Grade 3 | Poorest liver function. Median survival 2.3–15.5 months | Aggressive locoregional or systemic therapy is likely to precipitate decompensation; consider transplant assessment or supportive care |
What the ALBI Grade needs (2 inputs)
- Serum albumin
- Entered in g/dL and converted internally to g/L, which is what the published coefficient expects. Albumin carries a negative coefficient — higher albumin lowers the score, and a lower score is better liver function.
- Serum total bilirubin
- Entered in mg/dL and converted internally to µmol/L. It enters as a base-10 logarithm, so the score is far more sensitive to a rise from 1 to 3 mg/dL than to a rise from 10 to 12.
Units. This matters more here than in almost any other liver score. The published formula takes bilirubin in µmol/L and albumin in g/L — SI units — while most US and Japanese charts report mg/dL and g/dL. This calculator accepts the conventional units and converts internally (bilirubin × 17.1, albumin × 10), but a hand-calculation from a mg/dL and g/dL chart will produce a score roughly 2 to 3 points too low, which is wider than an entire grade. If you are checking a score computed elsewhere and it looks implausibly good, unit confusion is the first thing to test.
What it returns
- ALBI score
- A continuous value, typically between about −4 and 0. Negative throughout the clinically relevant range, which catches people out — a more negative score means better liver function.
- ALBI grade (1, 2 or 3)
- Grade 1 at −2.60 or below, grade 2 above −2.60 up to and including −1.39, grade 3 above −1.39. The grade is what published survival figures attach to.
How it is calculated
The model was built by taking the variables Child-Pugh uses, discarding the two that require judgement, and asking which of the remainder independently predicted survival in patients with hepatocellular carcinoma. Only albumin and bilirubin survived, and the coefficients were fitted to a Japanese cohort of 1,313 patients spanning all tumour stages. Both variables are markers of hepatic synthetic and excretory function rather than of tumour biology, which is why the resulting index behaves as a liver-function measure and was shown to retain its meaning in patients with chronic liver disease and no cancer at all. The logarithmic transformation of bilirubin reflects the shape of its relationship with survival — the prognostic difference between a bilirubin of 1 and 2 is much greater than between 8 and 9.
Facts & figures
| Variable | Child-Pugh | ALBI |
|---|---|---|
| Bilirubin | Yes, in three bands | Yes, as log₁₀, continuous |
| Albumin | Yes, in three bands | Yes, continuous |
| INR / prothrombin time | Yes | No |
| Ascites | Yes — graded subjectively | No |
| Encephalopathy | Yes — graded subjectively | No |
| Output | Grade A, B or C | Continuous score plus grade 1, 2 or 3 |
ALBI keeps two of Child-Pugh's five variables and uses both as continuous measures rather than banding them. The gain is reproducibility and finer resolution; the loss is that clinically obvious decompensation contributes nothing to the score.
Evidence
Derivation — Japanese cohort, all HCC stages
2015 · n = 1,313Patients with hepatocellular carcinoma of all stages from Japan, used to fit a model restricted to objective laboratory measures of liver function.
The two-variable model performed at least as well as Child-Pugh grade, and within Child-Pugh grade A it identified two classes with clearly different prognoses.
International validation cohorts
2015 · n = 5,097Comparable hepatocellular carcinoma cohorts from other geographical regions, reported in the same publication.
ALBI performed at least as well as Child-Pugh grade in all geographical regions tested.
Specific clinical settings — resection and sorafenib
2015 · n = 1,657Patients undergoing hepatic resection (n = 525) and patients receiving sorafenib for advanced hepatocellular carcinoma (n = 1,132), tested separately to confirm the grade held across treatment contexts.
The grade retained prognostic discrimination in both the surgical and the systemic-therapy setting.
Chronic liver disease without HCC
2015 · n = 501Patients with chronic liver disease and no hepatocellular carcinoma, included specifically to test whether the index measures liver dysfunction rather than tumour burden.
Utility in this group supported the authors' contention that ALBI is genuinely an index of liver function.
ALBI-T and further validation — Hiraoka 2018
2018Japanese cohort study proposing a modified ALBI grade and combining it with tumour burden as the ALBI-T score, assessing hepatic function and prognosis in hepatocellular carcinoma.
Supported subdividing ALBI further and combining it with tumour staging, on the basis that liver function alone does not determine outcome.
How it compares
ALBI Grade vs Child-Pugh score
Comparable overall, but ALBI is objective and subdivides Child-Pugh A into two prognostically distinct groups — which is where most patients with hepatocellular carcinoma actually sit.
In the derivation paper ALBI performed at least as well as Child-Pugh grade in every geographical region tested, using two variables instead of five and none requiring judgement. The meaningful difference is resolution rather than accuracy: the majority of patients with hepatocellular carcinoma present as Child-Pugh A, and ALBI splits that class in two with clearly different prognoses. The trade-off runs the other way for clinical decompensation — Child-Pugh counts ascites and encephalopathy, ALBI ignores them, so a patient with refractory ascites and preserved laboratory values will be graded far more favourably by ALBI than by Child-Pugh. Use ALBI to stratify hepatic reserve within a class; do not use it to decide whether a patient is decompensated.
ALBI Grade vs MELD 3.0
Different purposes entirely — MELD 3.0 ranks transplant priority by short-term mortality, ALBI grades hepatic reserve to decide whether a liver will tolerate cancer treatment.
MELD 3.0 is fitted to three-month waiting-list mortality and includes creatinine, INR, sodium, sex and albumin; it is the allocation instrument and ALBI has no role there. ALBI answers a question MELD was never asked: given that this patient has hepatocellular carcinoma and we are choosing between resection, embolisation and systemic therapy, how much liver function is there to spend. The two are routinely calculated on the same patient for different decisions, and a low MELD does not imply ALBI grade 1.
ALBI Grade vs Milan criteria
Complementary halves of the same assessment — Milan describes the tumour, ALBI describes the liver it sits in, and transplant candidacy needs both.
The Milan criteria are entirely radiological: lesion size, lesion number, vascular invasion, extrahepatic spread. They say nothing about hepatic function, and a patient can be comfortably within Milan with a liver that will not tolerate resection or embolisation while waiting. ALBI supplies that missing axis. In practice the pair is what determines the route: within Milan with good hepatic reserve opens resection or bridging therapy toward transplant; within Milan with ALBI grade 3 points toward transplantation without an aggressive bridge, because the liver has no margin to lose.
Pearls & pitfalls
- The formula wants µmol/L and g/L. Feeding it mg/dL and g/dL produces a plausible-looking number that is badly wrong — this calculator converts for you, but a bedside hand-calculation from a US chart will not.
- A more negative score is better. The sign convention is the opposite of almost every other score in hepatology, and grade 1 is the most negative band.
- Ascites and encephalopathy contribute nothing. A patient with refractory ascites can be ALBI grade 1, and that is a real blind spot rather than a subtlety — do not use the grade as a summary of how well the patient is.
- The value of ALBI is inside Child-Pugh A, not as a replacement for it. Overall discrimination is comparable; the gain is separating a large class that Child-Pugh reports as uniform.
- Bilirubin enters as a logarithm, so the score saturates in jaundiced patients. Two patients at 15 and 25 mg/dL differ far less in ALBI than they do clinically.
- The published median survival ranges are wide because they pool cohorts of different stage and era. Quote the grade, not a specific month figure, when speaking to a patient.
- It is not an allocation score. MELD 3.0 governs transplant priority; ALBI has no role in it.
- Albumin is depressed by inflammation, sepsis and nutritional state independent of synthetic function, so an acutely unwell patient may grade worse than their liver warrants. Recheck when stable.
Critical actions
- Confirm the units before trusting a hand-calculated score, and convert bilirubin to µmol/L (mg/dL × 17.1) and albumin to g/L (g/dL × 10).
- Record the continuous score alongside the grade — within grade 2 especially, the position in the band carries information the grade discards.
- Assess ascites and encephalopathy separately and explicitly, since ALBI is blind to both and both change management.
- Pair the grade with tumour staging. ALBI describes the liver; BCLC stage and the Milan criteria describe the cancer, and treatment follows from both.
- Repeat after any acute illness has resolved before acting on a grade 3 result, because albumin falls with inflammation regardless of hepatic synthesis.
- In a Child-Pugh A patient facing resection or repeated embolisation, calculate ALBI specifically — that is the decision this score was built to inform.
Why this score exists
The authors' stated objection to Child-Pugh was methodological rather than statistical: it was never derived for hepatocellular carcinoma, its variables were selected pragmatically in 1964 for a study of portacaval shunt surgery, and two of the five are scored by subjective assessment. Their aim was an index built only from objective measures, fitted in patients who actually have hepatocellular carcinoma, and tested internationally rather than in one population. The paper is explicit that eliminating ascites and encephalopathy was the point of the exercise — and that inclusion of patients with chronic liver disease but no cancer was there to demonstrate that what the model captures is liver dysfunction. Whether removing clinical decompensation from a liver-function score is a gain or a loss remains genuinely contested, and the authors did not claim otherwise.
About the creator
First author, ALBI derivation
Led the international collaboration that derived and validated the ALBI grade.
Co-author; statistical development
Developed the statistical model underlying the ALBI grade and its subsequent refinements.
Co-author, Japanese derivation and validation cohorts
Contributed the Japanese cohorts in which the model was fitted and first tested.
Limitations
- Blind to ascites and encephalopathy by design, so it can substantially overstate hepatic reserve in a clinically decompensated patient.
- Albumin is not a pure marker of hepatic synthesis — it falls with inflammation, sepsis, nephrotic-range proteinuria and malnutrition, all common in this population.
- Albumin is also modifiable by infusion, which means the score can be improved without the liver changing at all.
- The logarithmic bilirubin term compresses differences at high bilirubin, limiting discrimination in jaundiced patients.
- Derived and largely validated in hepatocellular carcinoma; its use as a general liver-function index outside that setting rests on a single 501-patient cohort in the original paper.
- Published median survival ranges are wide and pooled across cohorts of different stage, era and treatment, so they support ranking rather than individual prognostication.
- The derivation cohort was Japanese, where hepatitis-related hepatocellular carcinoma predominates; the aetiological mix in metabolic and alcohol-related disease is different, and albumin behaves differently in those populations.
- Not validated or intended for transplant allocation, and no jurisdiction uses it for that purpose.
If you are the patient
The ALBI grade is a way of measuring how well your liver is working, using just two numbers from a routine blood test: albumin, a protein your liver makes, and bilirubin, a waste product your liver clears. Those two are put into a formula, and the answer places you in grade 1, 2 or 3 — grade 1 meaning the liver is working best, grade 3 meaning it has the least reserve. Doctors use it mainly in liver cancer, because most treatments for liver cancer put some strain on the liver itself, and this helps them judge how much strain yours can safely take. Two things are worth knowing. The score is a negative number, and a more negative number is better, which is confusing but normal. And the grade only looks at those two blood tests — it does not account for fluid in the abdomen or confusion, so your doctor will assess those separately. The grade is about your liver's capacity, not about the cancer itself, so it is only ever one half of the picture when a treatment is being chosen.
Frequently asked questions
How is the ALBI grade calculated?#
ALBI = (log₁₀ of bilirubin in µmol/L × 0.66) + (albumin in g/L × −0.085). The result is graded at two fixed boundaries: −2.60 or below is grade 1, above −2.60 up to and including −1.39 is grade 2, and above −1.39 is grade 3. Note the units — µmol/L and g/L, not mg/dL and g/dL.
What are the ALBI grade cut-offs?#
Grade 1 is a score of −2.60 or lower, grade 2 is above −2.60 and up to −1.39, and grade 3 is above −1.39. Because the score is negative throughout the clinically relevant range, grade 1 is the most negative band and represents the best liver function.
Is ALBI better than Child-Pugh?#
Not better overall — it performed at least as well as Child-Pugh in every region tested in the derivation study, using two objective variables instead of five. Its advantage is resolution and reproducibility: it removes the two variables Child-Pugh grades subjectively, and it splits Child-Pugh grade A, where most patients with hepatocellular carcinoma present, into two groups with clearly different survival. Its disadvantage is that it ignores ascites and encephalopathy entirely.
What survival does each ALBI grade predict?#
Across the derivation and validation cohorts, reported median survival was 18.5 to 85.6 months for grade 1, 5.3 to 46.5 months for grade 2, and 2.3 to 15.5 months for grade 3. Those ranges are wide because the cohorts differed in tumour stage, treatment and era, so the figures are best read as a rank ordering rather than as an individual prognosis.
Why is my ALBI score negative?#
Because albumin carries a negative coefficient and contributes far more in magnitude than the bilirubin term across normal values. A patient with an albumin of 40 g/L contributes −3.4 from albumin alone, which the bilirubin term rarely offsets. It is a scale convention, not an error, and a more negative score means better liver function.
Can ALBI be used in patients without liver cancer?#
It was tested in 501 patients with chronic liver disease and no hepatocellular carcinoma, and its performance there was what convinced the authors it measures liver dysfunction rather than tumour burden. That supports its use as a general index of hepatic reserve, but the survival figures and the great majority of the validation evidence come from hepatocellular carcinoma cohorts, so the grade carries less established meaning outside that setting.
Does ALBI account for ascites or encephalopathy?#
No, and that is deliberate — removing them was the design goal, because both are graded subjectively. The consequence is a genuine blind spot: a patient with refractory ascites and preserved albumin and bilirubin can score ALBI grade 1. Assess decompensation separately; do not read the grade as a statement of how well the patient is.
Is ALBI used for liver transplant allocation?#
No. Allocation runs on MELD 3.0 in the United States and comparable models elsewhere, and ALBI was neither derived nor validated for that purpose. Its role is in deciding whether a liver has the reserve to tolerate a cancer treatment — resection, embolisation or systemic therapy — which is a different question from ranking waiting-list urgency.
References
Validation and evidence
- Hiraoka A, Kumada T, Tsuji K, et al. Newly Proposed ALBI Grade and ALBI-T Score as Tools for Assessment of Hepatic Function and Prognosis in Hepatocellular Carcinoma Patients. Liver Cancer. 2018;8(5):312-325.
- Zhang ZY, Guan J, Wang XP, Hao DS, Zhou ZQ. Modified Child-Pugh grade vs albumin-bilirubin grade for predicting prognosis of hepatocellular carcinoma patients after hepatectomy. World J Gastroenterol. 2020;26(7):749-758.