About the BARD Score for Advanced Fibrosis in Steatotic Liver Disease
Three items and a maximum of four points make BARD the simplest of the serum fibrosis scores: BMI of 28 or more scores 1, an AST/ALT ratio of 0.8 or more scores 2, and diabetes scores 1. A total of 0 or 1 excludes advanced fibrosis with a 96% negative predictive value, while 2 to 4 carries an odds ratio of 17 (95% CI 9.2–31.9) for advanced fibrosis. The asymmetric weighting is deliberate — the AST/ALT ratio counts double because it was the strongest of the three predictors. Its trade-off is bluntness: with only five possible values it discriminates less well than FIB-4 or the NAFLD Fibrosis Score, and it spared the fewest biopsies of any score in head-to-head comparison.
Formula
BARD = (BMI ≥ 28 ? 1 : 0) + (AST/ALT ratio ≥ 0.8 ? 2 : 0) + (diabetes ? 1 : 0)- BMI ≥ 28
- 1 point. Note the threshold is 28, not the conventional obesity cut-off of 30.
- AST/ALT ratio ≥ 0.8
- 2 points — the only double-weighted item, and the one that most determines the total.
- diabetes
- 1 point if present.
- Because the AST/ALT ratio alone is worth 2 points, a patient with a ratio of 0.8 or above is already at the high-risk threshold regardless of BMI and diabetes status.
- Conversely a patient below the ratio threshold cannot exceed 2 even with both other points, so the ratio effectively gates the result.
- All three items are thresholds rather than continuous terms, so a BMI of 27.9 and 40 differ by a point at one boundary and not at all thereafter.
- The BMI cut-off is 28 kg/m², which is lower than the standard definition of obesity and is easy to misremember as 30.
Interpreting the result
A score of 0 or 1 excludes advanced fibrosis with a negative predictive value of 96%, and in a patient with no other concerning features that supports metabolic management in primary care without hepatology referral. A score of 2 to 4 carries an odds ratio of 17 for advanced fibrosis, but this must be read for what it is — a strong relative risk in a cohort where a quarter of patients had advanced fibrosis, not a high probability in an individual. Positive predictive value is the weak point across all simple serum scores and BARD is no exception, so a high score is an instruction to test further, typically with FIB-4, the NAFLD Fibrosis Score or elastography. One structural caveat dominates interpretation: in a patient with diabetes and a BMI of 28 or more, two points are already banked before the transaminases are considered, so the score's ability to separate risk in that very common group is limited.
| Score | Band | What it means | Action |
|---|---|---|---|
| 0–1 | Low risk of advanced fibrosis | Advanced fibrosis excluded with 96% negative predictive value | Manage metabolic risk factors; hepatology referral not indicated on this basis alone |
| 2–4 | High risk of advanced fibrosis | Odds ratio 17 (95% CI 9.2–31.9) for advanced fibrosis, but only modest positive predictive value | Advanced fibrosis cannot be excluded — proceed to FIB-4, the NAFLD Fibrosis Score or transient elastography |
What the BARD Score needs (3 inputs)
- BMI (kg/m²)
- Scores 1 point at 28 or above, 0 below. A single step, not a gradient.
- AST and ALT (U/L)
- Entered as the AST/ALT ratio, which scores 2 points at 0.8 or above — double-weighted because it was the strongest single predictor.
- Diabetes
- Scores 1 point when present.
Units. AST and ALT are used only as a ratio, so their units cancel provided both come from the same assay. BMI is in kg/m² with a threshold of 28. No albumin, platelet count or unit conversion is required — which is the score's main practical advantage.
What it returns
- BARD score
- An integer from 0 to 4. Only five values are possible, which is the source of both its simplicity and its bluntness.
- Risk category
- Low risk (0–1) or high risk (2–4) of advanced fibrosis.
How it is calculated
The score came out of an analysis of 827 patients with NAFLD across two geographically separate tertiary centres, a cohort that was 51% female with a median BMI of 33, 35% diabetic, 91% insulin-resistant, and 24% with advanced fibrosis. Univariate analysis identified several features each carrying an odds ratio of at least 2.4 for advanced fibrosis, including BMI of 28 or more, age over 50, an AST/ALT ratio of 0.8 or more, insulin resistance and diabetes. Forced-entry logistic regression then reduced these to three that could be combined in a weighted sum, with the AST/ALT ratio receiving double weight. Notably, age did not make the final score despite being a univariate predictor — a difference from FIB-4 and the NAFLD Fibrosis Score, both of which include it, and one that spares BARD their tendency to over-call in older patients.
Facts & figures
| Item | Threshold | Points |
|---|---|---|
| BMI | ≥ 28 kg/m² | 1 |
| AST/ALT ratio | ≥ 0.8 | 2 |
| Diabetes | Present | 1 |
Maximum 4. The double-weighted AST/ALT ratio effectively gates the result: at or above 0.8 the patient already meets the high-risk threshold; below it, they cannot exceed 2.
| Feature | Value |
|---|---|
| Patients | 827 across two tertiary centres |
| Female | 51% |
| Median BMI | 33 kg/m² (only 3% had a normal BMI) |
| Hypertension | 60% |
| Diabetes | 35% |
| Insulin resistance | 91% |
| Advanced fibrosis | 24% |
The high baseline prevalence of advanced fibrosis is why the odds ratio of 17 does not translate into a high positive predictive value in lower-prevalence community populations.
Evidence
Derivation and validation — Harrison
2008 · n = 827827 patients with NAFLD analysed at two geographically separate tertiary medical centres. Univariate analysis identified BMI ≥ 28, age over 50, an AST/ALT ratio ≥ 0.8, a low insulin-sensitivity index and diabetes as individually associated with advanced fibrosis (each odds ratio ≥ 2.4); forced-entry logistic regression reduced these to three, combined as a weighted sum.
A BARD score of 2–4 was associated with an odds ratio of 17 for advanced fibrosis (95% CI 9.2–31.9) and a negative predictive value of 96%. The authors concluded the score can reliably exclude advanced fibrosis from readily available clinical data, particularly among non-diabetics.
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 diabetic, of whom 27 (19%) had advanced fibrosis. BARD was compared directly against the AST/ALT ratio, APRI, FIB-4 and the NAFLD Fibrosis Score on blood taken at the time of biopsy.
BARD achieved an AUROC of 0.77 — fourth of the five scores, ahead only of APRI (0.67) and behind FIB-4 (0.86), the AST/ALT ratio (0.83) and the NAFLD Fibrosis Score (0.81). Its negative predictive value of 95% was among the highest, but it would have avoided biopsy in only 38% of patients, the lowest of the four scores capable of reliable exclusion.
How it compares
BARD Score vs FIB-4
FIB-4 is clearly the stronger test — AUROC 0.86 against BARD's 0.77 in direct comparison, and it spares 62% of biopsies against BARD's 38% — so BARD is justified only when a platelet count is unavailable.
Both are simple serum scores, but FIB-4's continuous output extracts far more information than BARD's five integer values. Head to head in 145 biopsy-proven patients, FIB-4 achieved the best discrimination of any score tested and BARD the second worst. Their negative predictive values were comparable (95% each), so BARD's rule-out claim holds up; what it loses is the proportion of patients it can confidently rule out at all. BARD's only practical advantage is that it needs no platelet count and no albumin.
BARD Score vs NAFLD Fibrosis Score
The NAFLD Fibrosis Score discriminates better and rules out advanced fibrosis in more patients, but BARD needs neither albumin nor a platelet count and gives a binary answer rather than a wide indeterminate zone.
In the same 145-patient comparison the NFS reached an AUROC of 0.81 against BARD's 0.77 and would have avoided biopsy in 52% of patients versus 38%. The NFS also uses six variables including albumin and platelets, so it is not always computable. There is a genuine structural trade-off: the NFS leaves roughly a quarter of patients formally indeterminate, whereas BARD always returns a low- or high-risk answer — at the cost of that answer carrying less information.
BARD Score vs APRI
BARD outperformed APRI for advanced fibrosis in steatotic liver disease (AUROC 0.77 versus 0.67) — APRI was derived in hepatitis C and does not transfer well to MASLD.
APRI combines AST with platelet count and was developed for chronic hepatitis C, where the relationship between transaminases and fibrosis differs from that in MASLD. In the head-to-head comparison of five simple scores in biopsy-proven NAFLD, APRI ranked last, and it was the only score whose negative predictive value did not exceed 90%. BARD, though crude, was built on a NAFLD cohort and reflects the metabolic phenotype directly.
Pearls & pitfalls
- The BMI threshold is 28 kg/m², not 30. Using the conventional obesity cut-off will misclassify patients between 28 and 30.
- The AST/ALT ratio is double-weighted, so it dominates: at 0.8 or above the patient is already high-risk, and below it they cannot reach 3 or 4.
- In a patient who is both diabetic and has a BMI of 28 or more, two of the four points are automatic — the score has little discriminating power left in that very common group, which is why the authors highlighted its performance in non-diabetics.
- BARD has no age term, unlike FIB-4 and the NAFLD Fibrosis Score. That spares it their over-calling in the elderly but also discards a genuine risk factor.
- The odds ratio of 17 is a relative measure from a cohort with 24% advanced fibrosis. It is not a probability, and positive predictive value in community populations is considerably lower.
- Only five score values exist, so BARD cannot track change over time in any meaningful way — it is a one-off triage tool, not a monitoring instrument.
Critical actions
- Use the correct BMI threshold of 28 kg/m² when scoring.
- Treat a score of 2–4 as 'not excluded' and proceed to a better test — FIB-4, the NAFLD Fibrosis Score or elastography — rather than as a positive diagnosis.
- Prefer FIB-4 or the NAFLD Fibrosis Score whenever a platelet count and albumin are available, since both discriminate better.
- Interpret the score cautiously in diabetic patients, where two points are frequently automatic.
- Address the metabolic drivers regardless of score — weight, glycaemic control, cardiovascular risk and alcohol — since these determine progression more than any single measurement.
Why this score exists
The paper's title states the purpose plainly — identifying patients *without* advanced disease — and BARD should be read as a rule-out instrument built to that specification rather than a general-purpose fibrosis estimate. The authors' wider finding was arguably as important as the score: in their 827 patients, insulin resistance was present in 91% and only 3% had a normal BMI, which reframed NAFLD as a metabolic disease with liver consequences rather than a primarily hepatic one. Against that background the choice of BMI, transaminase ratio and diabetes is not an arbitrary convenience selection but a deliberate use of the metabolic phenotype itself as the predictor. Their specific caveat that the score works best in non-diabetics follows directly from the design, since diabetes is one of the three inputs.
About the creator
First author, 2008 derivation study
Derived BARD across two tertiary centres, and later became senior author of the FAST score — the same investigator on both ends of the simple-to-elastography spectrum.
Senior author
Co-authored the derivation and has shaped much of the wider NAFLD/MASLD natural-history literature.
Limitations
- With only five possible values it is the bluntest of the serum fibrosis scores, and it cannot track change over time.
- It ruled out advanced fibrosis in the smallest proportion of patients (38%) among the scores capable of reliable exclusion, so most patients still need a further test.
- Two of the four points are frequently automatic in diabetic, obese patients, precisely the group most at risk, which erodes discrimination where it is most needed.
- It contains no age term, discarding a well-established fibrosis risk factor.
- Positive predictive value is modest; the headline odds ratio of 17 comes from a tertiary cohort with 24% advanced fibrosis and does not transfer to community prevalence.
- It was derived in tertiary-centre NAFLD populations with a median BMI of 33, so its behaviour in lean steatotic liver disease — where only 3% of the derivation cohort sat — is poorly characterised.
- It does not apply to viral hepatitis or alcohol-related liver disease.
If you are the patient
The BARD score is a quick way to check whether fatty liver disease is likely to have caused significant scarring of the liver. It uses just three things: your body mass index, the ratio between two liver enzymes in a routine blood test (AST and ALT), and whether you have diabetes. The score runs from 0 to 4. A score of 0 or 1 is reassuring — significant scarring is very unlikely, and in the original study this was correct 96% of the time. A score of 2 or more does not mean you definitely have scarring; it means it cannot be ruled out from these three pieces of information alone, so your doctor will arrange a better test, such as a different blood score or a liver scan. Because having diabetes and a raised body mass index each add a point automatically, people with both will often score 2 or more without that necessarily indicating liver scarring.
Frequently asked questions
What does BARD stand for?#
BMI, AST/ALT Ratio, and Diabetes — the three components. BMI of 28 or more scores 1, an AST/ALT ratio of 0.8 or more scores 2, and diabetes scores 1, for a maximum of 4.
What BARD score means low risk?#
0 or 1. That excludes advanced fibrosis with a negative predictive value of 96%. A score of 2 to 4 carries an odds ratio of 17 for advanced fibrosis and means further testing is needed.
Why is the AST/ALT ratio worth two points?#
Because it was the strongest of the three predictors in the derivation analysis. The practical effect is that it gates the result — at 0.8 or above a patient already meets the high-risk threshold, and below it they cannot score more than 2.
Is the BARD BMI threshold 28 or 30?#
28 kg/m². It is deliberately lower than the conventional definition of obesity, and using 30 instead will misclassify patients in that range.
Is BARD better than FIB-4?#
No. In direct comparison FIB-4 achieved an AUROC of 0.86 against BARD's 0.77 and ruled out advanced fibrosis in 62% of patients versus 38%. BARD's advantage is only that it needs no platelet count or albumin.
Does BARD work in patients with diabetes?#
Less well. Diabetes is itself one of the three inputs, so a diabetic patient with a BMI of 28 or more has two points automatically. The original authors specifically noted the score performs best in non-diabetics.