FIB-4 and Other Liver Fibrosis Scores Without a Biopsy: How to Read the Result

The FIB-4 index was developed in 2006 in patients co-infected with HIV and hepatitis C virus. In the paper by Sterling et al., the formula took the form: age (in years) times AST (U/L), divided by the product of the platelet count (10^9/L, i.e. thousands per microlitre) and the square root of ALT (U/L). So you need four numbers from one blood draw plus one demographic fact, and the calculation fits on a phone calculator.

Why these four values? Age is included because fibrosis accumulates over years, so risk rises with age. The AST-to-ALT ratio tends to rise as fibrosis progresses (we wrote about this in our article on AST higher than ALT). The platelet count tends to fall with portal hypertension and worsening liver function. Combining these features discriminates advanced fibrosis better than single parameters. In the original paper (832 patients, Ishak scoring) the area under the ROC curve was 0.765, and at a cut-off below 1.45 the negative predictive value for advanced fibrosis was 90% with sensitivity of 70%; a cut-off above 3.25 gave a specificity of 97%. Those cut-offs, however, referred to HIV/HCV co-infection, not steatosis.

In fatty liver disease, different cut-offs are most often cited than in the original paper: 1.3 and 2.67. One source is the study by Shah et al. (2009), which compared FIB-4 with seven other non-invasive indices in 541 adults with NAFLD. The median FIB-4 was 1.11, and the area under the ROC curve for advanced fibrosis was 0.802 — higher than for the NAFLD fibrosis score (0.768), the AST/ALT ratio (0.742) or APRI (0.730). A FIB-4 ≤1.30 had a 90% negative predictive value and ≥2.67 an 80% positive predictive value.

Note the second part of the authors' conclusions: at a specificity fixed at 90% (FIB-4 = 1.93), the sensitivity for detecting advanced fibrosis was only 50%. In other words, FIB-4 is better than other simple indices but far from perfect — its value lies mainly in the fact that a low result rules out advanced disease well and a high one suggests it. This is why an intermediate result (and it is common) needs a second, different test.

Because age is part of the formula, FIB-4 behaves differently in different age groups. The index was designed and validated mainly in people aged 35–65. In older people age itself raises the result, producing many false positives. In young people age is so low that the result rarely crosses the threshold even in real disease.

The practical conclusion is twofold. First, if you are 65 or older and your FIB-4 is a little above 1.3, do not treat it as an alarm — the doctor may apply a higher cut-off (about 2.0) or order a second test. Second, if you are 35 or younger and have risk factors (obesity, diabetes, clearly raised ALT), a normal FIB-4 should not reassure you completely, because in this group the index detects disease poorly. In both situations the decision belongs to the doctor, not to your own interpretation of a number.

Because FIB-4 often gives an intermediate result, modern practice rests on two steps: a cheap index from routine tests and, if the result is uncertain, a second, more accurate test. The second test may be ELF (Enhanced Liver Fibrosis), a blood test measuring markers of extracellular matrix remodelling, or transient elastography (FibroScan), which measures liver stiffness in kilopascals. We write about elastography itself in our knowledge-base entry.

Of particular interest in outpatient practice are data from population samples and primary care. In the study by Srivastava et al. (2019), a two-step algorithm — FIB-4 followed by the ELF test — was evaluated in 3,012 patients with NAFLD; using the pathway detected five times more cases of advanced fibrosis and cirrhosis (OR 5.18) with 81% fewer unnecessary referrals to specialist clinics. In the study by Kjaergaard et al. (2023), involving 3,378 participants from the general population and risk groups, ELF alone (cut-off ≥9.8) gave significantly fewer false positives (11%) than FIB-4 (35%) and NFS (45%), and a strategy of FIB-4 followed by ELF in uncertain cases gave 8% false positives, 4% false negatives and 88% correct classifications.

Type 2 diabetes is a special group: advanced fibrosis is more common in it, so even a good screening test has a lower negative predictive value. In the meta-analysis by Han et al. (2024), covering 12 studies and 5,624 patients with NAFLD and type 2 diabetes, at low cut-offs (1.3–1.67) sensitivity was 0.74, specificity 0.62, negative predictive value 0.82 and positive predictive value 0.49. At high cut-offs (2.67–3.25) sensitivity fell to 0.33 but specificity rose to 0.92. The authors judged the accuracy of FIB-4 to be moderate and pointed to the limited number of studies and their heterogeneity; the negative predictive value at the low cut-off did not reach 0.90 when the prevalence of advanced fibrosis exceeded 30%.

Other non-invasive indices include APRI (AST-to-platelet ratio), the NAFLD fibrosis score (which also needs body weight, diabetes status and albumin) and simple indices such as the AST/ALT ratio. Compared with them, FIB-4 performed best in studies and is the simplest to use, which is why it became the first-step index in most modern pathways. Recommendations of scientific societies (among them EASL in 2021 and AASLD in 2023) base the first stage of fibrosis assessment on such non-invasive tests, and the detailed cut-offs and algorithms may differ — it is worth asking your doctor which rules your centre uses.

FIB-4 is an exceptionally valuable tool, because it turns routine, cheap tests into information about prognosis — and it is also easy to overrate or underrate. The most sensible approach is to treat it as the first filter in a two-step pathway: a low result lets you calmly focus on lifestyle, a high or intermediate one moves you on to more precise tests. If you belong to a risk group (obesity, diabetes, insulin resistance, persistently raised enzymes), it is worth asking your doctor to calculate FIB-4 at the next opportunity; it is a simple conversation that can detect disease before it gives symptoms.