About the Fatty Liver Index (FLI)
Every other calculator in this family stages fibrosis; the Fatty Liver Index answers the prior question of whether there is fat in the liver at all. Four measurements produce a 0–100 probability — triglycerides, BMI, GGT and waist circumference — with a score below 30 ruling steatosis out (negative likelihood ratio 0.2) and 60 or above ruling it in (positive likelihood ratio 4.3). Accuracy in the derivation cohort was 0.84 (95% CI 0.81–0.87), against ultrasound as the reference standard. Two of its four inputs are anthropometric rather than laboratory values, which makes it the rare liver score requiring a tape measure, and the reason it has been used so widely in population epidemiology.
Formula
y = 0.953 × ln(triglycerides mg/dL) + 0.139 × BMI + 0.718 × ln(GGT U/L) + 0.053 × waist cm − 15.745; FLI = (e^y / (1 + e^y)) × 100- triglycerides
- In mg/dL, natural-log transformed.
- BMI
- In kg/m², entered linearly.
- GGT
- In U/L, natural-log transformed.
- waist
- Waist circumference in cm, entered linearly.
- Triglycerides must be in mg/dL, not mmol/L. Multiply mmol/L by 88.57 to convert — using mmol/L directly understates the log term and pushes the score down.
- The logistic transformation constrains the output to 0–100, so extreme inputs saturate rather than producing runaway values.
- BMI and waist circumference both appear, and they are correlated; the score is therefore sensitive to body habitus twice over, which is part of why it performs well in obesity and less well at the lean end.
- GGT rises with alcohol intake independently of steatosis, so the score cannot separate metabolic from alcohol-related fatty liver.
Interpreting the result
A score below 30 makes hepatic steatosis unlikely, with a negative likelihood ratio of 0.2, and in a patient without other concerns that is sufficient to defer imaging. At 60 or above steatosis is likely, with a positive likelihood ratio of 4.3, and the appropriate response is to confirm as clinically indicated and — far more importantly — to move on to fibrosis staging, because the presence of fat says nothing about prognosis while the presence of fibrosis says almost everything. The 30 to 60 band is genuinely uninformative and calls for ultrasound if the answer matters. Because likelihood ratios rather than predictive values are what the score reports, its output must be combined with the pre-test probability: an FLI of 65 means something quite different in a lean 25-year-old than in a patient with type 2 diabetes and central obesity.
| Score | Band | What it means | Action |
|---|---|---|---|
| < 30 | Fatty liver ruled out | Hepatic steatosis unlikely (negative likelihood ratio 0.2) | Imaging can reasonably be deferred; continue metabolic risk management |
| 30 to < 60 | Indeterminate | Steatosis neither ruled in nor ruled out | Arrange ultrasound if establishing steatosis will change management |
| ≥ 60 | Fatty liver ruled in | Hepatic steatosis likely (positive likelihood ratio 4.3) | Assess metabolic risk factors and proceed to fibrosis staging with FIB-4 or the NAFLD Fibrosis Score — steatosis alone does not establish prognosis |
What the Fatty Liver Index needs (4 inputs)
- Triglycerides (mg/dL)
- Enters on a natural-log scale weighted 0.953. Should ideally be a fasting sample, as postprandial triglycerides are substantially higher.
- BMI (kg/m²)
- Weighted 0.139, entered linearly rather than logged.
- GGT (U/L)
- Gamma-glutamyl transferase, log-transformed and weighted 0.718. Sensitive to alcohol as well as to steatosis, which is the score's main confounder.
- Waist circumference (cm)
- Weighted 0.053. Measured at the level your service standardises on — the score is sensitive to technique, and it is the input most often estimated rather than measured.
Units. Triglycerides must be in mg/dL — multiply mmol/L by 88.57. GGT is in U/L, BMI in kg/m², and waist circumference in cm. The triglyceride conversion is the error most likely to distort the result.
What it returns
- Fatty Liver Index
- A probability rescaled to 0–100. It is not a percentage chance of steatosis for an individual, but a rank on a calibrated scale.
- Steatosis category
- Ruled out (< 30), indeterminate (30 to < 60), or ruled in (≥ 60).
How it is calculated
Bedogni and colleagues used data from the Dionysos Nutrition and Liver Study, comprising 216 subjects with and 280 without suspected liver disease, in whom fatty liver was diagnosed by ultrasonography and alcohol intake was recorded prospectively with a seven-day diary. Thirteen candidate predictors were screened by bootstrapped stepwise logistic regression — including sex, age, ethanol intake, ALT, AST, GGT, BMI, waist circumference, skinfold thickness, glucose, insulin, triglycerides and cholesterol — with the explicit aim of finding the simplest accurate algorithm rather than the best-fitting one. Four variables survived. That transaminases were tested and dropped while GGT and the two anthropometric measures were retained is the study's most instructive result, and it explains why normal ALT offers no reassurance about hepatic fat.
Facts & figures
| Metric | Value |
|---|---|
| Accuracy (AUROC) | 0.84 (95% CI 0.81–0.87) |
| FLI < 30 | Rules out fatty liver — negative likelihood ratio 0.2 |
| FLI ≥ 60 | Rules in fatty liver — positive likelihood ratio 4.3 |
| Reference standard | Ultrasonography |
Derived in the Dionysos Nutrition and Liver Study: 216 subjects with and 280 without suspected liver disease, with alcohol intake assessed by seven-day diary.
| Question | Answered by FLI? | Use instead |
|---|---|---|
| Is there fat in the liver? | Yes | — |
| Is there advanced fibrosis? | No | FIB-4, NAFLD Fibrosis Score |
| Is there steatohepatitis? | No | FAST score, biopsy |
| Is the cause metabolic or alcohol? | No | History; GGT is raised by both |
Evidence
Derivation — Bedogni (Dionysos study)
2006 · n = 496496 subjects from the Dionysos Nutrition and Liver Study — 216 with and 280 without suspected liver disease — in whom fatty liver was diagnosed by ultrasonography and alcohol intake assessed using a seven-day diary. Thirteen candidate variables were screened by bootstrapped stepwise logistic regression to find the simplest accurate algorithm.
An algorithm based on BMI, waist circumference, triglycerides and GGT achieved an accuracy of 0.84 (95% CI 0.81–0.87) for detecting fatty liver. A score below 30 ruled it out with a negative likelihood ratio of 0.2, and 60 or above ruled it in with a positive likelihood ratio of 4.3. Transaminases were among the variables tested and were not retained.
External validation — Koehler (Rotterdam Study)
2013 · n = 2,6522,652 participants of the population-based Rotterdam Study, mean age 76.3 ± 6.0 years, assessed between February 2009 and February 2012 with abdominal ultrasound, fasting bloods and anthropometry, to test whether the FLI identifies fatty liver from any cause and NAFLD specifically in a large white elderly population.
FLI score was independently associated with NAFLD in multivariable analysis (odds ratio 1.05 per unit, 95% CI 1.04–1.05). The authors concluded the FLI accurately identifies ultrasonography-confirmed NAFLD in a large elderly population, providing the external validation the original paper had called for.
How it compares
Fatty Liver Index vs FIB-4
They are sequential, not alternative — the Fatty Liver Index establishes whether steatosis is present, and FIB-4 then establishes whether it has caused advanced fibrosis, which is the question that carries the prognosis.
The two share no inputs and answer different questions. A patient with an FLI of 85 and a FIB-4 of 0.9 has fatty liver without advanced fibrosis, which is the commonest and most benign combination in MASLD. The dangerous misreading is to treat a high FLI as evidence of serious liver disease, or a low FIB-4 as evidence that there is no steatosis. In practice the FLI selects who is worth staging and FIB-4 does the staging.
Fatty Liver Index vs NAFLD Fibrosis Score
The Fatty Liver Index is a steatosis-detection tool validated against ultrasound; the NAFLD Fibrosis Score is a fibrosis tool validated against liver biopsy — different targets, different reference standards, and neither substitutes for the other.
Both use BMI, which invites confusion, but that is where the overlap ends. The FLI adds triglycerides, GGT and waist circumference to predict hepatic fat; the NFS adds transaminase ratio, platelets, albumin and dysglycaemia to predict advanced scarring. Their reference standards differ accordingly — ultrasound for one, biopsy for the other — which also means their published performance figures are not comparable, since detecting fat is an easier target than staging fibrosis.
Fatty Liver Index vs BARD score
Both are three- or four-item scores in steatotic liver disease, but the Fatty Liver Index finds fat while BARD rules out fibrosis — using either for the other's purpose is a category error.
The pairing is worth stating explicitly because both are marketed as simple MASLD scores and both include BMI. BARD's other inputs are the AST/ALT ratio and diabetes, aimed at scarring; the FLI's are triglycerides, GGT and waist, aimed at fat. A sensible pathway runs FLI first to establish steatosis, then a fibrosis score — BARD if platelets and albumin are unavailable, otherwise FIB-4 or the NAFLD Fibrosis Score, both of which discriminate better than BARD.
Pearls & pitfalls
- Triglycerides must be in mg/dL. Entering mmol/L (typically a value under 3 instead of around 150) collapses the log term and produces a falsely low score.
- The FLI measures fat, not fibrosis. A score of 100 tells you nothing about prognosis — that requires FIB-4, the NAFLD Fibrosis Score or elastography.
- GGT rises with alcohol as readily as with metabolic steatosis, so the FLI cannot distinguish MASLD from alcohol-related fatty liver; the drinking history does that.
- Waist circumference is technique-dependent and is the input most often estimated rather than measured. An inaccurate waist directly biases the result.
- BMI and waist both appear despite being correlated, so the score responds strongly to adiposity and is least reliable in lean steatotic liver disease.
- The output is a likelihood-ratio-based rank, not an individual probability. Interpret it against the patient's pre-test probability rather than reading 70 as 'a 70% chance'.
- Ultrasound was the reference standard, and ultrasound itself is insensitive to mild steatosis — so the FLI is calibrated against an imperfect comparator.
Critical actions
- Convert triglycerides to mg/dL before calculating, and use a fasting sample where possible.
- Follow a high FLI with fibrosis staging rather than stopping at the steatosis diagnosis — fibrosis is what determines outcome.
- Take an explicit alcohol history, since GGT and therefore the FLI cannot separate metabolic from alcohol-related fat.
- Measure waist circumference properly rather than estimating it, and use a consistent anatomical landmark.
- Interpret the score alongside pre-test probability, particularly in lean patients where the anthropometric terms carry less signal.
Why this score exists
The authors were unusually explicit about the score's intended users and its unfinished state. They positioned it as a tool to help physicians select subjects for ultrasonography and intensified lifestyle counselling, and to help researchers select patients for epidemiological studies — a triage and cohort-definition instrument, not a diagnostic test. They then closed by stating plainly that validation in external populations was needed before it could be used for those purposes, a caveat later satisfied by the Rotterdam Study. The methodological choice worth noting is that they optimised for simplicity rather than maximal fit: thirteen variables were available, including insulin and skinfold thickness, and they deliberately stopped at four.
About the creator
First author, 2006 derivation study
Derived the index from the Dionysos Nutrition and Liver Study, deliberately stopping at four variables when thirteen were available.
Senior author
Led the Italian liver-epidemiology programme from which the index came.
Limitations
- It detects steatosis only and carries no information about fibrosis, steatohepatitis, or prognosis.
- It cannot distinguish metabolic from alcohol-related fatty liver, because GGT is elevated by both.
- Ultrasonography was the reference standard, and ultrasound has limited sensitivity for mild steatosis, so the score is calibrated against an imperfect benchmark rather than histology or MRI-PDFF.
- Both BMI and waist circumference are included, making the score heavily adiposity-driven and least reliable in lean steatotic liver disease.
- Waist circumference measurement is operator-dependent and not standardised across services, introducing variability the formula cannot correct for.
- The derivation cohort was Italian and the principal external validation was in a white elderly Dutch population, so performance across other ethnicities and younger age groups is less well characterised.
- It reports likelihood ratios rather than predictive values, so it cannot be read as an individual probability without accounting for pre-test risk.
If you are the patient
The Fatty Liver Index estimates how likely it is that there is extra fat in your liver. It uses four measurements: your body mass index, your waist measurement, and two blood tests — triglycerides, a type of fat in the blood, and GGT, a liver enzyme. The result is a number from 0 to 100. Below 30 means fat in the liver is unlikely. Above 60 means it is likely, and your doctor may arrange a scan to confirm it. Between the two, the test cannot say either way. One important point: this score only looks for fat, not for scarring. Fat in the liver is very common and on its own is usually not dangerous — it is scarring, called fibrosis, that causes serious liver problems, and that needs different tests. So a high score is a reason for further checks rather than a cause for alarm. The GGT blood test also rises with alcohol, so your doctor will ask about your drinking to work out the likely cause.
Frequently asked questions
What is a normal Fatty Liver Index score?#
Below 30 makes hepatic steatosis unlikely, with a negative likelihood ratio of 0.2. Between 30 and 60 is indeterminate, and 60 or above makes steatosis likely (positive likelihood ratio 4.3).
Does the Fatty Liver Index measure liver scarring?#
No. It detects fat only. Fibrosis — the feature that determines prognosis — requires a different test such as FIB-4, the NAFLD Fibrosis Score, or transient elastography. This is the most important limitation of the score.
What is the Fatty Liver Index formula?#
y = 0.953 × ln(triglycerides in mg/dL) + 0.139 × BMI + 0.718 × ln(GGT) + 0.053 × waist in cm − 15.745, then FLI = (e^y / (1 + e^y)) × 100, giving a score from 0 to 100.
Which units does the Fatty Liver Index need for triglycerides?#
mg/dL. If your laboratory reports mmol/L, multiply by 88.57 first — using mmol/L directly will produce a falsely low score.
Can the Fatty Liver Index tell alcohol-related from metabolic fatty liver?#
No. GGT is raised by alcohol and by metabolic steatosis alike, so the score identifies fat without identifying its cause. A drinking history is required to distinguish them.
Has the Fatty Liver Index been externally validated?#
Yes. The original authors called for external validation, and the Rotterdam Study subsequently confirmed in 2,652 elderly participants that the FLI accurately identifies ultrasonography-confirmed NAFLD, with an odds ratio of 1.05 per unit increase.