About the NAFLD Fibrosis Score (NFS)
Two cut-offs, not one, are what make the NAFLD Fibrosis Score usable: below −1.455 advanced fibrosis is excluded with a negative predictive value of 93%, above 0.676 it is confirmed with a positive predictive value of 90%, and the wide indeterminate zone between them is an honest admission that a third of patients cannot be classified by blood tests alone. Six routine variables feed it — age, BMI, impaired fasting glucose or diabetes, the AST/ALT ratio, platelet count and albumin — giving an AUROC of 0.88 in the derivation cohort of 733 biopsy-proven patients. Applied across that cohort it would have spared 75% of biopsies while classifying 90% of those correctly.
Formula
NFS = −1.675 + (0.037 × age) + (0.094 × BMI) + (1.13 × impaired fasting glucose or diabetes) + (0.99 × AST/ALT ratio) − (0.013 × platelets) − (0.66 × albumin)- age
- In years.
- BMI
- In kg/m².
- impaired fasting glucose or diabetes
- 1 if either is present, otherwise 0.
- AST/ALT ratio
- AST divided by ALT, both in U/L — the ratio, not the individual values.
- platelets
- Platelet count in ×10⁹/L.
- albumin
- Serum albumin in g/dL.
- Albumin must be in g/dL. A value in g/L fed into a coefficient of −0.66 will drive the score deeply negative and falsely exclude advanced fibrosis — the most dangerous unit error on this page.
- The score is unbounded in both directions, so there is no maximum or minimum to sanity-check against.
- Dysglycaemia is scored as a single binary term: impaired fasting glucose and established diabetes contribute identically, with no gradation for glycaemic control.
- It is the AST/ALT ratio that enters the model, not AST and ALT separately, so proportionally similar transaminases give the same contribution whether both are mildly or markedly raised.
Interpreting the result
A score below −1.455 excludes advanced (F3–F4) fibrosis reliably, with a negative predictive value of 93% in the construction group and 88% on validation, and supports lifestyle-focused metabolic management with periodic reassessment rather than hepatology referral. Above 0.676 advanced fibrosis is present in the great majority — positive predictive value 90% and 82% respectively — and warrants hepatology assessment and surveillance planning. Between the two the score has genuinely not answered the question, and the correct response is another test, usually transient elastography or a second serum marker, not a repeat NFS. Two caveats shape everything above: the score over-calls in older patients because age sits in the numerator, and it under-calls when albumin is preserved despite significant fibrosis, which is common in compensated disease.
| Score | Band | What it means | Action |
|---|---|---|---|
| < −1.455 | Advanced fibrosis unlikely | Advanced (F3–F4) fibrosis excluded with 93% negative predictive value in derivation, 88% on validation | Manage metabolic risk factors and reassess periodically; hepatology referral not indicated on this result alone |
| −1.455 to 0.676 | Indeterminate | Neither ruled in nor ruled out — roughly a quarter of patients land here | Proceed to transient elastography or a second non-invasive marker; do not simply repeat the NFS |
| > 0.676 | Advanced fibrosis likely | Advanced fibrosis confirmed with 90% positive predictive value in derivation, 82% on validation | Refer for hepatology assessment; plan cirrhosis surveillance if confirmed |
What the NAFLD Fibrosis Score needs (6 inputs)
- Age (years)
- Weighted 0.037. Fibrosis risk rises with age, but so does the score's tendency to over-call in older patients.
- BMI (kg/m²)
- Weighted 0.094. Requires a measured height and weight, which is what limits the score's use in remote or registry settings.
- Impaired fasting glucose or diabetes
- Adds 1.13 when present — the single largest coefficient, reflecting how strongly dysglycaemia drives fibrosis in MASLD.
- AST and ALT (U/L)
- Entered as the AST/ALT ratio, weighted 0.99. A ratio approaching or exceeding 1 suggests more advanced disease.
- Platelet count (×10⁹/L)
- Weighted −0.013, so falling platelets raise the score — the portal-hypertension signal common to most fibrosis indices.
- Albumin (g/dL)
- Weighted −0.66, so a falling albumin raises the score. Its inclusion is unusual among simple fibrosis indices.
Units. Albumin must be entered in g/dL (divide g/L by 10) and platelets in ×10⁹/L. AST and ALT are used only as their ratio, so their units cancel provided both come from the same assay.
What it returns
- NAFLD Fibrosis Score
- A continuous value, typically between about −4 and +2, reported to three decimal places.
- Fibrosis category
- Advanced fibrosis unlikely, indeterminate, or advanced fibrosis likely.
How it is calculated
Angulo and colleagues took 733 patients with biopsy-confirmed NAFLD and split them into a construction group of 480 and a validation group of 253, then ran multivariate modelling across routine demographic, clinical and laboratory variables to find which independently predicted advanced fibrosis. Six survived: age, hyperglycaemia, BMI, platelet count, albumin and the AST/ALT ratio. Rather than pick a single operating point, the authors then chose two — a low cut-off optimised for ruling out and a high one optimised for ruling in — accepting that patients falling between them would need something further. That decision is what distinguishes the NFS from single-threshold scores, and it is also why roughly a quarter of patients end up indeterminate.
Facts & figures
| Construction group (n = 480) | Validation group (n = 253) | |
|---|---|---|
| AUROC | 0.88 | 0.82 |
| NPV at low cut-off (−1.455) | 93% | 88% |
| PPV at high cut-off (0.676) | 90% | 82% |
Applied across all 733 patients, the two cut-offs would have avoided biopsy in 549 (75%) with a correct prediction in 496 (90%).
| Score | AUROC | NPV | Biopsies potentially avoided |
|---|---|---|---|
| FIB-4 | 0.86 | 95% | 62% |
| AST/ALT ratio | 0.83 | 93% | 69% |
| NAFLD Fibrosis Score | 0.81 | 92% | 52% |
| BARD | 0.77 | 95% | 38% |
| APRI | 0.67 | — | — |
145 patients with biopsy-proven NAFLD, 27 (19%) with advanced fibrosis (Kleiner stage 3–4). Positive predictive values were modest for every score tested — these are rule-out instruments.
Evidence
Derivation and validation — Angulo
2007 · n = 733733 patients with NAFLD confirmed by liver biopsy, split into a construction group of 480 and a validation group of 253. Routine demographic, clinical and laboratory variables were analysed by multivariate modelling to predict the presence or absence of advanced fibrosis; age, hyperglycaemia, BMI, platelet count, albumin and the AST/ALT ratio were the independent predictors retained.
AUROC 0.88 in the construction group and 0.82 on validation. The low cut-off of −1.455 excluded advanced fibrosis with negative predictive values of 93% and 88%; the high cut-off of 0.676 confirmed it with positive predictive values of 90% and 82%. Biopsy would have been avoided in 549 of 733 patients (75%) with a correct prediction in 496 (90%).
Comparative validation — McPherson
2010 · n = 145145 consecutive patients with biopsy-proven NAFLD from the Newcastle Hospitals Fatty Liver Clinic (2003–2009), mean age 51, mean BMI 35, half with diabetes, of whom 27 (19%) had advanced fibrosis. The AST/ALT ratio, APRI, BARD, FIB-4 and the NAFLD Fibrosis Score were calculated from blood taken at the time of biopsy.
The NFS achieved an AUROC of 0.81 with a negative predictive value of 92%, ranking third behind FIB-4 (0.86) and the AST/ALT ratio (0.83) and ahead of BARD (0.77) and APRI (0.67). Using it to exclude advanced fibrosis would have avoided biopsy in 52% of patients. Positive predictive values were modest across all scores.
How it compares
NAFLD Fibrosis Score vs FIB-4
FIB-4 is the better first-line test — it matched or beat the NFS on discrimination in head-to-head comparison and needs no BMI or albumin, making it usable on a routine blood panel alone.
In 145 biopsy-proven patients assessed with both, FIB-4 achieved an AUROC of 0.86 against 0.81 for the NFS, with negative predictive values of 95% and 92% and potential biopsy avoidance of 62% versus 52%. FIB-4 needs only age, AST, ALT and platelets, so it can be calculated retrospectively from existing results, whereas the NFS additionally requires a measured BMI and an albumin. The practical pathway most services adopt reflects this: FIB-4 first, then the NFS or elastography as the second test when FIB-4 is indeterminate.
NAFLD Fibrosis Score vs BARD score
BARD is simpler and matches the NFS on ruling advanced fibrosis out, but it discriminates less well overall and spares far fewer biopsies — the NFS is the more informative of the two.
BARD collapses three items into a 0–4 integer, so it is faster to compute but far coarser than the NFS's continuous output. Compared directly, BARD scored an AUROC of 0.77 against the NFS's 0.81, and although its negative predictive value was actually marginally higher (95% versus 92%), it would have avoided biopsy in only 38% of patients against the NFS's 52%. BARD's real advantage is that it needs no albumin and no platelet count.
NAFLD Fibrosis Score vs Fatty Liver Index
They answer sequential questions, not the same one — the Fatty Liver Index predicts whether fat is present in the liver, while the NAFLD Fibrosis Score predicts whether that fat has caused advanced scarring.
A patient can have a Fatty Liver Index of 90, confirming steatosis, and an NFS below −1.455 excluding advanced fibrosis; that combination is both common and reassuring, because steatosis without fibrosis carries a far better prognosis. Confusing the two leads to the error of treating a high Fatty Liver Index as evidence of serious liver disease. The correct sequence is to establish steatosis first, then stage fibrosis.
Pearls & pitfalls
- Albumin is in g/dL, not g/L. A value of 40 instead of 4.0 subtracts about 24 from the score and will falsely exclude advanced fibrosis in almost anyone.
- The indeterminate zone is wide by design and catches roughly a quarter of patients. Landing there is a valid result meaning 'do another test', not a failed calculation.
- Age is in the numerator, so the score drifts upward with age and over-calls advanced fibrosis in older patients — the same weakness FIB-4 has.
- Impaired fasting glucose and established diabetes score identically. A well-controlled diabetic and a newly dysglycaemic patient get the same 1.13.
- Preserved albumin can pull the score down in compensated disease with real fibrosis, which is one route to a false-negative result.
- It requires a measured BMI, which is why FIB-4 rather than the NFS became the primary-care and registry screening tool despite the NFS performing comparably.
- Positive predictive value is only modest in independent cohorts. Treat a high score as a strong indication to investigate, not as a diagnosis of cirrhosis.
Critical actions
- Confirm the albumin unit before trusting a strongly negative score — this is the single error most likely to cause harm here.
- Move to elastography or a second serum marker when the result is indeterminate, rather than repeating the same test.
- Interpret a high score in an older patient with added caution, and corroborate before committing them to a cirrhosis pathway.
- Establish that steatotic liver disease is actually present first; the NFS assumes the diagnosis and was not validated in viral or alcohol-related liver disease.
- Address the metabolic drivers — weight, glycaemic control, alcohol, cardiovascular risk — in every band, including the rule-out band, since MASLD progression is driven by factors the score only partially captures.
Why this score exists
The authors' stated ambition was to render liver biopsy unnecessary in a substantial proportion of patients, and the paper reports that arithmetic explicitly rather than leaving it as an inference: 549 of 733 biopsies avoided, 496 of those calls correct. That framing explains the two-threshold design. A single cut-off maximising overall accuracy would have produced a score that was wrong in both directions for the patients closest to the line, whereas separating a rule-out threshold from a rule-in threshold lets each be tuned for the predictive value that matters in its direction — at the price of formally conceding that the patients in between have not been answered. Naming that indeterminate group, rather than disguising it, is the design's honesty and its main practical limitation at once.
About the creator
First author, 2007 derivation study
Derived the score in 733 biopsy-proven patients and chose two thresholds rather than one, explicitly conceding an indeterminate middle group.
Senior author
Co-authored the derivation and led much of the early work defining NAFLD as a progressive disease.
Limitations
- Roughly a quarter of patients fall in the indeterminate zone, where the score provides no actionable answer and a further test is unavoidable.
- Positive predictive value was only modest in independent validation, so a high score justifies investigation rather than a diagnosis.
- Age in the numerator causes systematic over-calling in older patients, mirroring the well-documented problem with FIB-4 above 65.
- It requires a measured BMI and albumin, which excludes it from the retrospective, blood-panel-only screening niche FIB-4 occupies.
- Dysglycaemia is binary, so it cannot distinguish severity or duration of diabetes despite that being the dominant driver of fibrosis progression in MASLD.
- It was derived exclusively in biopsy-proven NAFLD and does not transfer to viral hepatitis, alcohol-related liver disease, or mixed aetiology.
- Both cut-offs were derived in tertiary-centre biopsy cohorts with a high prevalence of advanced fibrosis, so predictive values will be less favourable in lower-prevalence community populations.
If you are the patient
The NAFLD Fibrosis Score estimates how likely it is that fatty liver disease has caused significant scarring, called fibrosis, in the liver. It uses your age, your body mass index, whether you have diabetes or raised blood sugar, and three blood results — two liver enzymes (AST and ALT), your platelet count, and albumin, a protein made by the liver. The score gives one of three answers. A low score means significant scarring is very unlikely and no specialist referral is needed for that reason. A high score means scarring is likely and you should see a liver specialist. Many people fall in the middle, and that is a normal result, not a mistake — it simply means the blood tests cannot settle the question and another test, usually a liver scan called elastography or FibroScan, is needed. Having fat in the liver is not the same as having scarring, and most people with fatty liver never develop it.
Frequently asked questions
What are the NAFLD Fibrosis Score cut-offs?#
−1.455 and 0.676. Below −1.455 advanced fibrosis is excluded (93% negative predictive value); above 0.676 it is likely (90% positive predictive value); between the two the result is indeterminate.
What does an indeterminate NAFLD Fibrosis Score mean?#
That the blood tests cannot classify the patient, which happens to roughly a quarter of people and is a designed feature of the two-threshold approach. The next step is transient elastography or a second non-invasive marker, not a repeat of the same score.
How accurate is the NAFLD Fibrosis Score?#
AUROC was 0.88 in the derivation group and 0.82 on validation across 733 biopsy-proven patients. In independent comparison against other simple scores it achieved 0.81, behind FIB-4 at 0.86.
Is FIB-4 or the NAFLD Fibrosis Score better?#
FIB-4 performed slightly better head-to-head (AUROC 0.86 versus 0.81) and requires no BMI or albumin, which is why it is usually the first-line test. The NFS is commonly used as the second test when FIB-4 is indeterminate.
Which units does the score need for albumin?#
g/dL. Entering albumin in g/L — for example 40 instead of 4.0 — subtracts roughly 24 points and will wrongly exclude advanced fibrosis. Divide g/L by 10 to get g/dL.
Can the NAFLD Fibrosis Score diagnose cirrhosis?#
No. It estimates the probability of advanced fibrosis (F3–F4) and its positive predictive value in independent cohorts is only modest, so a high score indicates the need for hepatology assessment and confirmatory testing rather than establishing a diagnosis.
References
Original / primary reference
Validation and comparison
Guidelines
- Rinella ME, Neuschwander-Tetri BA, Siddiqui MS, et al. AASLD Practice Guidance on the clinical assessment and management of nonalcoholic fatty liver disease. Hepatology. 2023;77(5):1797-1835.
- European Association for the Study of the Liver. EASL–EASD–EASO Clinical Practice Guidelines on the management of metabolic dysfunction-associated steatotic liver disease (MASLD). J Hepatol. 2024;81(3):492-542.