Application of the fibrosis-4 score in a generalised population

Abstract

BACKGROUND

To pick up liver fibrosis, prior to presenting with cirrhosis, several different risk scores exist. One such, the fibrosis-4 (FIB-4) score, was originally validated in a population with hepatitis C and human immunodeficiency virus infection and subsequently expanded to metabolic dysfunction-associated steatotic liver disease. Local and national (British) guidelines outline when deranged liver function tests warrant calculation of a FIB-4 score.

AIM

To investigate whether the population tested with FIB-4 locally resembled the validation group/if FIB-4 was applied according to guidelines.

METHODS

All FIB-4 requests in one month from two hospitals were collected from the local laboratory system, along with laboratory and demographic data. The data collected were narrowed down to those that matched the initial validation population’s laboratory values from 2006. The group was then further narrowed down to exclude patients dissimilar to the validation population (using electronic medical records to exclude based on age, obesity, and other factors as outlined by the original study from which the FIB-4 score originates).

RESULTS

It was found that the vast majority of patients having a FIB-4 calculated were grossly dissimilar to the initial population on which the score was originally validated, and the patients who did resemble it often had a FIB-4 calculated even if there was no apparent indication to do so.

CONCLUSION

There are multiple compounding factors of error in indiscriminate use of FIB-4. Probability of liver steatosis would likely be better assessed by using adaptations of the score on appropriate populations.

Key Words: Risk assessment; Liver fibrosis; Fatty liver; Non-alcoholic fatty liver disease; Metabolic dysfunction-associated steatotic liver disease

Core Tip: The fibrosis-4 (FIB-4) score is used indiscriminately and when applied inappropriately may lead to misleading results and diagnostic noise. This study shows how different a generalised population is from the population on which the FIB-4 score was validated. Other adaptations of the FIB-4 score exist that have shown diagnostic superiority on specific populations which could be applied, even reflexively using artificial intelligence.

INTRODUCTION

Several different scores exist to quantify the risk and severity of liver fibrosis in patients with metabolic dysfunction-associated steatotic liver disease (MASLD)[1]. The fibrosis-4 (FIB-4) score[2], developed in 2006, using age, platelet count, aspartate aminotransferase (AST) and alanine aminotransferase (ALT) serum activities, was initially conceived as a tool to estimate the risk of liver fibrosis in patients with known concomitant human immunodeficiency virus (HIV) and hepatitis C virus (HCV), but has been widely expanded to include patients with MASLD[3-6].

According to the British Society of Gastroenterology (BSG) guidelines[7], which have been adopted in local practise, patients with deranged liver functions tests (LFTs) in a hepatitic pattern should have a battery of investigations [referred to as a non-invasive liver screen (NILS)] including a liver ultrasound (US), before using the calculations of FIB-4 or non-alcoholic fatty liver score (NFS) to determine fibrosis risk (Figure 1). If no other cause is identified, and steatosis is seen on the US, the FIB-4 (or NFS) score is recommended.

Figure 1 Pathway for investigating deranged liver functions tests as per the British Society of Gastroenterology guidelines. ELF: Enhanced Liver Fibrosis score; LFT: Liver functions tests; BMI: Body mass index; FIB-4: Fibrosis-4; US: Ultrasound.

FIB-4 is searchable and requestable directly on local electronic order communications (ICE, Clinisys) and is also included in the test profile described as being for first line investigations of abnormal LFTs. FIB-4 is calculated by the Laboratory Information Management System (LIMS) using the formula[8] and then may be reported against the diagnostic thresholds as set by the BSG guidelines[7].

Using a scoring system like the FIB-4 score to quantify risk of liver fibrosis in patients is a cost-effective method of avoiding referral for further evaluation or biopsy in patients with a low risk of cirrhosis, reducing pressure on specialists and risks of complications of biopsy[9] whilst identifying patients who are more likely to benefit from further workup[9]. Guidelines[7] recommend use of the FIB-4 score to determine which patients are at high risk of liver fibrosis and require further investigation on presenting with deranged LFTs. However, guidelines, if not followed properly can lead to spurious results. One example would be applying guidance to a patient cohort not represented by the evidence, a non-validated population, such as a population without risk factors for liver fibrosis. The FIB-4 score is a powerful tool in the estimation of risk of liver fibrosis in that it is a simple calculation requiring standard laboratory tests as input that is valuable for triage and risk stratification, however, its usefulness is limited to its application across appropriate patient populations and it has been shown to have reduced accuracy in specific patient populations [like those at the extremes of age and body mass index (BMI)].

Since the pilot FIB-4 population was composed of patients positive for HIV and HCV, it differs from the population on which it has been expanded to today, which may limit its relevance in application. Nonetheless, the FIB-4 score has been adopted as a scoring system for patients with MASLD[3] after further validation studies have taken place[3-6]. The thresholds used in the BSG guidelines for FIB-4 classification are derived from an adaptation of the initial FIB-4 cutoff values that have been shown to be safer in ruling out risk of fibrosis[10].

Besides presence of HIV and HCV, the original study was validated using further exclusion criteria that are often not considered on its widespread application. These other criteria for exclusion may to a greater extent limit its applicability to a broader patient group-namely based on their age, BMI, and laboratory results. This means that although some of these values are within the calculation, the validation group was comprised of a certain range of values, and thus values outside of that range were not validated. The initial study also excluded patients who had a known diagnosis of cancer or HIV-related illness, as well as anaemia and neutropenic patients. The FIB-4 is included in the standard NILS panel, meaning that all patients undergoing a NILS would have a FIB-4 calculated, regardless of whether the patient falls within a validated population or not.

The aim of this study was to review patient data from one month of requested FIB-4 calculations in a United Kingdom National Health Service laboratory service (covering a population of 450000) to assess the diagnostic significance of the FIB-4s being calculated. Diagnostic relevance was assessed using the following parameters: How similar the patient population on which FIB-4s were calculated were to the population on which the FIB-4 was originally validated. If those people having FIB-4 calculated were ‘appropriate’ as per local guidelines i.e. NILS confirmed a MASLD diagnosis.

MATERIALS AND METHODS

A PathManager search of the LIMS WinPath Enterprise (Clinisys) was run for a months’ worth (December 2025) of the test library code for FIB-4 at St Richard’s Hospital in Chichester and Worthing General Hospital. Patient demographics, laboratory and clinical data (if available on BMI and diabetes status, via Plexus Shared Care Record) were collected. Biochemical tests were run on Alinity platforms using Abbott reagent (Abbott, ALT and AST run without pyridoxal-5’-phosphate) and haematology on Alinity hq analyzer (Abbott). These patients were compared to the original population that the FIB-4 score was validated on, and their similarity was determined. The characteristics of the initial group that the FIB-4 score was validated on in 2007[2] are shown in Table 1; specific analytical methods were not stipulated.

Table 1 Range of values from initial validation population of the fibrosis-4 score, range.

	Range in initial validation set	Reference ranges
ALT (U/L)	14-156	5-40
AST (U/L)	13-97	5-40
Platelets	130-258 × 109/L	150-410 × 109/L
Age (years)	33-47	NA
BMI (kg/m2)	20-28	18.5-24.9
Albumin (U/L)	38-46	35-50

ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; NA: Not applicable; BMI: Body mass index.

The patients collected from December 2025 who had FIB-4s calculated were compared to the above validation set and were narrowed down to leave only the patients that fit the same range of laboratory values. Furthermore, those remaining patients were also filtered by those who would have been excluded from the validation data set for other criteria, as follows: Active HIV-related opportunistic infection or cancer, neutrophil count below 1500 cells/mm³, platelet count below 70000/mm³, haemoglobin below 110 g/L for women and 120 g/L for men, serum creatinine > 1.5 times the upper limit of normal, concurrent hepatitis A or B infection, evidence of decompensated liver disease, severe psychiatric disease, clinically significant coexisting medical conditions that would preclude HCV therapy, previous treatment with interferon or ribavirin [as evidenced by past medical history on Plexus (enotes)].

Then, the adherence to the BSG guidelines, confirming MASLD prior to calculation of FIB-4, was also assessed via Plexus and results reviewed to see if calculation of a FIB-4 was indicated. The largest effect was assessed by applying each of the criteria separately to the data to determine the effect on the population. Then, the most significant factor was selected and then the rest of the criteria were applied to the reduced cohort.

RESULTS

In total 462 patients had a FIB-4 requested and calculated by the laboratory in December 2025. The number of patients that matched the validation laboratory values are shown in Table 2. Only 89 (19%) met age criteria (33-47 years) and of these the number meeting each biochemical criteria are shown in the third column. Applying all 5 criteria, only 24 (5%) remained (Table 2).

Table 2 Breakdown of patients meeting initial validation criteria, n (%).

Criterion	Number (n = 462)	Number (n = 89)
Age 33-47 (years)	89 (19)
ALT 14-156 (U/L)	288 (62)	82 (92)
AST 13-97 (U/L)	439 (95)	86 (97)
Platelets 130-258	224 (48)	39 (44)
Albumin 38-46 (U/L)	349 (76)	64 (72)
Met all 5 criteria	24 (5)	24 (27)

ALT: Alanine aminotransferase; AST: Aspartate aminotransferase.

By applying the above filters and removing patients that would have been excluded in the initial study as per the extended list of exclusion criteria only eight patients remained (9% of the correct age group, and 2% of the whole cohort). Applying BMI criteria, only four people were left.

Of the final four patients, only one had LFT derangement at the time of calculation of the FIB-4 score. Some of these patients had a FIB-4 score calculated on normal LFTs because the initial derangement of LFTs did not include all the factors in the FIB-4 calculation and thus had them repeated. Others never had any derangement in LFTs at all. The following table shows the above 4 patients with their respective FIB-4 scores and radiological evidence of steatosis (Table 3).

Table 3 Features of remaining patients.

ID	Note on patient	FIB-4 calculated	Radiological evidence of steatosis
I	Had no present nor historical LFT derangement	0.55	No US abdomen performed
II	FIB-4 incalculable on previously deranged LFTs; FIB-4 calculated on normal LFTs	0.72	No US abdomen performed
III	Had no present nor historical LFT derangement	0.87	No US abdomen performed
IV	FIB-4 incalculable on previously deranged LFTs; FIB-4 calculated on deranged LFTs	Fibro scan suggested	Mild heterogenous liver echotexture with no obvious focal pathology identified

FIB-4: Fibrosis-4; LFT: Liver functions tests; US: Ultrasound.

DISCUSSION

Of the few patients resembling the validation cohort, only one had a FIB-4 calculation indicated. Patients without evidence of steatosis and no other risk factors for liver fibrosis represent a patient population that the FIB-4 score has not been validated upon and the value of the calculation uncertain. Two of these patients also had no record of any LFT derangement; according to BSG guidelines, there would therefore have been no indication to follow the MASLD pathway, and it is unclear why a NILS would have been requested. In this limited dataset, however, this did not lead to any unnecessary referrals, only unnecessary investigations. This data is suggestive that the FIB-4 score is applied haphazardly and that inclusion of the FIB-4 in the NILS may be a source of diagnostic clutter. Recognising that the FIB-4 score is being applied inappropriately is important to adjust predetermined screening panels to eliminate calculation of scores that are not clinically applicable to particular patient groups. The guidelines on the management of abnormal LFTs[7] recommend investigating FT derangement without repeating them, yet all remaining patients had a FIB-4 calculated on a repeat set of LFTs. This may be inevitable when balancing slimmed down test profiles to prevent overdiagnosis but can be targeted by a more reflexive approach to screening such as used by the iLFT algorithm[11]. However, one benefit to repeating LFTs could be the revealing of a dynamic shift more suggestive of an acute derangement, which could inform the clinician’s diagnostic question. However, since the FIB-4 was requested to be calculated on the repeat LFTs, the clinician is then faced with diagnostic noise as they may recognise it may not have been indicated to calculate. This concern is amplified when applying FIB-4 to patients who do not belong to a validated population (namely, without evidence of steatosis), generating artefactual findings by producing a hypothetical cirrhosis risk calculated in the context of an improper patient sample on inappropriate lab results. From our limited data, one could conclude that repeating investigations in patients with an incomplete panel of deranged LFTs will likely not alter the FIB-4 calculation. However, a repeat panel is only truly useful, without further validation of the FIB-4, if it is used to assess applicability of FIB-4. Routinely including a full panel for standard investigations outside the context of MASLD is likely not cost-effective; further research is required to determine the feasibility of repeat full panels vs expansion of standardised panels. One possible workaround to avoid repetition of LFTs for their further analysis is by the use of an automation technique like iLFT[12] which could potentially reduce the need for repeating LFTs by reflexively triggering expansion of panels if incompletely ordered, initially. Implementation of the iLFT[12] algorithm has the potential to reduce inappropriately ordered FIB-4 calculations by, for example, excluding those with cholestatic liver markers. Artificial intelligence (AI) models could be an alternative to recognise patient groups and their validation status before applying scoring systems to non-validated patient groups.

Calculated values also amplify other errors such as biological and analytical variation, thus the composite calculations are inherently less precise than the individual component values[13]. ALT and AST are measured by at least two major methods (enzymatic assay with or without pyridoxal-5-phosphate) and combining these variable measures, without altering decision thresholds based on analytical method used, in inappropriate patient populations further compounds error[13]. The initial study validating the FIB-4 score does not specify the method by which the ALT and AST were measured and thus the value ranges from the initial study could potentially represent a measuring technique that differs from our population as well as any other population on which the FIB-4 score is calculated.

Extending scoring systems to populations dissimilar from their validation cohorts can be useful, but only when evidence supports transferability. The greatest limiting criterion in our cohort’s resemblance to the validation dataset was age, which reduced the comparable population by over 80%. In 2025, Kimura et al[11] investigated age-specific thresholds and found that FIB-4 overestimated fibrosis risk in patients older than those in the initial validation set; their cohort included substantial non-MASLD causes of steatosis (including viral hepatitis and alcoholic liver disease). They recommended using FIB-3[14-16], a simplified model excluding age, which demonstrated superior diagnostic accuracy in patients > 60. By recognising that patients belonging to an age bracket that is better assessed using FIB-3 rather than FIB-4, there is the opportunity to improve diagnostic accuracy and estimation of risk of liver fibrosis without collecting any further blood tests or information-merely by switching to the most appropriate diagnostic scoring system.

Another relevant subgroup is patients living with obesity, who are at higher risk of MASLD[17,18] and thus potentially also to progression of cirrhosis[16]. In 2024, Green et al[15] compared intraoperative liver biopsy during bariatric surgery with calculated FIB-4 scores and found standard FIB-4 cutoffs led to significant false negatives, meaning that cases of advanced fibrosis were missed. Using the thresholds initially set by Sterling et al[7] led to many severely obese patients falling into indeterminate zones or being misclassified.

Considering the two patient populations above, one may suggest the automatic streamlining of using secondary thresholds (for patients with obesity) or alternate scoring systems (i.e. FIB-3 in patients > 60) in patients of specific groups with scores evidenced to be superior[18].

CONCLUSION

The vast majority of patients who have a FIB-4 calculated are largely dissimilar to the original population that the FIB-4 score was validated on. Since the FIB-4 score is part of the NILS, it is often applied to an undifferentiated population, which may lead to scores of indeterminate significance. Significance of the any score is diluted by the compounded effect of inappropriate application and biological and analytical variation. There are more studies required to better illustrate the applicability of the FIB-4 score to patient groups dissimilar to the validation group (thrombocythaemia, neutropaenia, extremes of age, etc.) and the general population. However, there have been studies that have recommended adaptations to the FIB-4 score to more diverse populations that demonstrate diagnostic superiority, but these are generally not applied. By recognising that there are data delineating which adaptations to the FIB-4 score can be most relevant to which populations, there is the opportunity to improve diagnostic accuracy and provide patient-centred care. AI or reflexive algorithms such as iLFTs, offers the possibility to reflexively trigger further blood analysis without repeating LFTs and appropriate application of scores.