How this instrument works
The NAFLD Fibrosis Score is a linear equation built from six variables a clinician already has on hand: age in years, body mass index, whether impaired fasting glucose or diabetes is present, the ratio of AST to ALT on a liver panel, the platelet count, and serum albumin. Each variable carries its own weight in the sum — age, BMI, the AST/ALT ratio, and the diabetes flag push the score up, while platelets and albumin pull it down — reflecting how each one, independently, tracked with biopsy-confirmed fibrosis stage in the patients the formula was built from.
Paul Angulo and colleagues published the score in 2007 after analysing 733 patients with biopsy-proven nonalcoholic fatty liver disease across several centres, then validating the cutoffs in a separate group. Below −1.455, the score carries a strong enough negative predictive value that advanced fibrosis, stages F3–F4, can usually be excluded without a biopsy. Above 0.676, the positive predictive value is strong enough to flag a patient who should be prioritised for biopsy or referred to hepatology. Between those two lines sits an indeterminate zone, where the score alone does not settle the question.
That middle zone is not a small footnote — in the original validation, roughly a quarter to a third of patients landed there, neither clearly low-risk nor clearly high-risk. For those patients the score's job is done; what follows is further testing, commonly liver stiffness measurement by elastography, another blood-based index, or a biopsy after all. The score also assumes NAFLD is the correct diagnosis to begin with — it was built and tested in that population, and its numbers do not transfer cleanly to fibrosis arising from other causes of liver disease.
- Enter Age in years and BMI in kg/m² as recorded in the chart.
- Set the impaired fasting glucose / diabetes toggle to Yes if either is documented.
- Enter the AST/ALT ratio, platelet count (×10⁹/L), and albumin (g/dL) from the same lab panel.
- Read the NAFLD Fibrosis Score and compare it against the −1.455 and 0.676 cutoffs shown.
Worked example — age 50, BMI 28, no diabetes, ratio 1.0
Work each term from the age-50, BMI-28 profile: 0.037 × 50 = 1.85. 0.094 × 28 = 2.632. The diabetes term drops out entirely — 1.13 × 0 = 0. The AST/ALT term: 0.99 × 1.0 = 0.99. Platelets subtract 0.013 × 220 = 2.86. Albumin subtracts 0.66 × 4.0 = 2.64. Starting from the constant −1.675 and adding each term in turn: −1.675 + 1.85 + 2.632 + 0 + 0.99 − 2.86 − 2.64 ≈ −1.703 — below the −1.455 cutoff, a low probability of advanced fibrosis.
Change the profile to an older, heavier patient with diabetes — age 65, BMI 35, impaired fasting glucose present, AST/ALT of 1.5, platelets down to 140, albumin down to 3.2 — and the same six terms sum to about 2.70, more than three points higher and well past the 0.676 cutoff into the high-probability range. Every input moved in the direction that raises fibrosis risk at once: older, heavier, diabetic, a higher AST/ALT ratio, fewer platelets, lower albumin — the kind of clustering that shows up together as liver disease advances.
Questions
What does the NAFLD Fibrosis Score actually decide?
It sorts patients with nonalcoholic fatty liver disease into three groups by their likelihood of already having advanced fibrosis — bridging scar tissue or cirrhosis, stages F3 and F4 — without requiring a liver biopsy to find out. A low score lets biopsy be deferred with reasonable confidence; a high score flags who should be prioritised for it or referred onward; the middle group still needs another test before the question is settled.
Why do platelets and albumin subtract from the score?
Both fall as liver disease advances, for mechanical reasons: a scarred, hypertensive liver traps and destroys more platelets via an enlarged spleen, and a failing liver simply synthesises less albumin. Because falling values point toward more fibrosis, their coefficients are negative — a lower platelet count or lower albumin pushes the score up toward the high-risk range, even though the raw subtraction looks like it should pull the total down.
What happens if my score lands in the indeterminate zone?
It happens often — roughly a quarter to a third of patients score between −1.455 and 0.676 in the studies behind this tool — and it means the six inputs here were not enough to settle the question either way. The usual next step is an additional noninvasive test, most often liver stiffness measurement by elastography, or in some cases a biopsy after all.
Does this score work for liver disease that isn't NAFLD?
Not reliably. The equation was derived and validated exclusively in patients with biopsy-confirmed nonalcoholic fatty liver disease, and its coefficients reflect how those specific patients' fibrosis related to age, BMI, and the lab values used. Applying it to viral hepatitis, alcohol-related liver disease, or other causes of fibrosis sits outside what the original study tested.
How were the −1.455 and 0.676 cutoffs chosen?
Angulo and colleagues set the lower cutoff to maximise negative predictive value — catching almost everyone who truly does not have advanced fibrosis — and the upper cutoff to maximise positive predictive value, catching almost everyone who does, then validated both in a separate patient cohort. The two-cutoff design deliberately trades a wide indeterminate middle for confidence at both ends, rather than drawing one line and living with more errors near it.
Can the score replace a liver biopsy entirely?
No — it is a triage tool, not a diagnosis. Biopsy remains the reference standard for staging fibrosis directly; what this score does is reduce how often that invasive step is needed, by identifying patients confident enough to skip it and prioritising the ones who should have it. Clinical judgment and lab trends over time still guide the final decision.
References
- Angulo P et al. 2007, Hepatology — original NAFLD Fibrosis Score study (PubMed)
- AASLD — Noninvasive Liver Disease Assessment practice guidance
Read this first: This instrument computes a screening figure from population formulas — it is not a diagnosis, and it cannot see the whole picture a clinician can. Use it to inform a conversation, not to replace one.