A Validation Study of Non-invasive Scoring Systems for Assessing Severity of Hepatic Fibrosis in a Cohort of South Indian Patients With Non-alcoholic Fatty Liver Disease

Abstract

Introduction

Liver biopsy is the gold standard for diagnosing and staging non-alcoholic fatty liver disease (NAFLD), but liver biopsy has its limitations. Non-invasive tests (NITs) eliminate many of the drawbacks of liver biopsy. We did a retrospective observational study to validate the NAFLD Fibrosis Score (NFS score) and Fibrosis Score 4 (FIB-4 index) against the gold standard liver biopsy in a cohort of South Indian patients with NAFLD.

Aims

The aim of this study was to validate the diagnostic accuracy of non-invasive fibrosis scoring systems (FIB-4 index and NFS), compared to that of liver histology, to predict AF in a cohort of south Indian patients with NAFLD.

Material and methods

A retrospective observational analytical study of patients who had a liver biopsy with a diagnosis of NAFLD and had all the data for aetiology assessment and NIT calculation within 4 weeks of biopsy were included in the study. On liver biopsy, NAFLD was scored as per NIH's NASH committee grading system. NFS and FIB-4 index were calculated, and scores more than 0.676 and 2.67, respectively, were taken as the cut-off to predict advanced fibrosis (AF). The sensitivity, specificity, positive predictive value, negative predictive value, and area under the receiver operating characteristic curve for NFS score and FIB-4 score to diagnose AF were calculated.

Results

A total of 147 patients were included in the study. Of these, 56 (38.1%) patients had AF (Stage 3, 4). Patients with AF were more likely to be older and have diabetes mellitus (DM). Patients with AF had lower platelet count, higher aspartate aminotransferase (AST), lower albumin, and higher AST/alanine aminotransferase ratio. An NFS of >0.676 had a sensitivity of 68% and specificity of 100%, and an FIB-4 index of >2.67 had a sensitivity of 67% and specificity of 95.6 % in diagnosing AF in our study.

Conclusion

The non-invasive scoring systems NFS and FIB-4 index can be used as a bedside tool for diagnosing liver fibrosis in NAFLD allowing liver biopsy to be used in a more targeted manner for patients diagnosed with AF on NITs.

The prevalence of non-alcoholic fatty liver disease (NAFLD) in Asian population is around 27% and in India ranges from 8.7% to 53.5%.^1,2 The prevalence is lower in rural areas and higher in urban populations. The country has seen a rise in the prevalence of metabolic risk factors such as diabetes, obesity, and dyslipidemia making NAFLD a leading cause of liver-related morbidity and mortality. A recent population-based study from South India showed a prevalence of 49.8%. Urban domicile was found to be a risk factor for NAFLD. NAFLD is a spectrum of diseases ranging from simple steatosis to steatohepatitis, fibrosis, and cirrhosis. Cardiovascular diseases and non-hepatic malignancies followed by liver diseases are the leading cause of mortality in patients with NAFLD.^{1, 2, 3} In patients with NAFLD, the presence of fibrosis is an independent predictor of liver-related and all-cause mortality. It is important to identify NAFLD patients with fibrosis so that early and timely interventions are done.

Liver biopsy is the gold standard for diagnosis and staging of NAFLD.⁴ This method has a number of drawbacks. Liver biopsies only sample a very small area of the liver (1/50,000), which can lead to sampling errors. It is an invasive procedure carrying significant morbidity including haemorrhage or hemobilia in 0.5%.^1,4 It is not readily acceptable to patients as well as physicians. To address this, several non-invasive diagnostic methods based on serum biomarkers and imaging modalities have been developed. Non-invasive tests (NITs) eliminate many of the drawbacks of liver biopsy, are simple to implement in clinical practice, and are less expensive and more acceptable to patients. Simple serum tests (e.g., aspartate aminotransferase [AST] to Platelet Ratio Index [APRI], body mass index [BMI], AST/alanine aminotransferase [ALT] ratio, diabetes, Fibrosis Score 4 [FIB-4 index], NAFLD Fibrosis Score [NFS score]) are made up of readily available biochemical surrogates and clinical risk factors for liver fibrosis^{5, 6, 7, 8} (Table 1). However, they are less accurate than more sophisticated and patented serum tests, which include direct measures of fibrogenesis or fibrolysis such as the FibroTest® (Biopredictive, Paris, France), Fibrometer®, (Echosens, France), enhanced liver fibrosis score™ score (Siemens Healthcare, Germany), and Hepascore (PathWest, University of Western Australia, Australia).^9,10 NITs based on simple serum tests (FIB-4, NFS) and clinical parameters have not been validated in the Indian population. We did a retrospective observational study to validate the NFS score and FIB-4 index against the gold standard liver biopsy in a cohort of South Indian patients with NAFLD.

Table 1.

Non-invasive Liver Tests (NILTs).

Scoring system	Clinical and biochemical parameters
1.NAFLD Fibrosis Score	Age, BMI, presence of diabetes, AST/ALT ratio Platelet count and albumin.
2. FIB-4	Age, AST and Platelet count
3. AST to Platelet Ratio Index	AST, Platelet count
4. BARD	BMI, AST/ALT ratio, Diabetes

Abbreviations: ALT = alanine aminotransferase; AST = aspartate aminotransferase; BMI = body mass index; BARD = BMI AST/ALT ratio, Diabetes; FIB-4 = Fibrosis Score 4; NAFLD = non-alcoholic fatty liver disease.

MATERIALS AND METHODS

A retrospective observational analytical study was done in the hepatology department of a tertiary care hospital in South India. The study duration was 2 years from July 2018 to July 2020. All patients aged >18 years who had a liver biopsy with a diagnosis of NAFLD, non-alcoholic steatohepatitis (NASH), or NASH cirrhosis and had all the data for aetiology analysis and NIT calculation done within 30 days of the biopsy were included in the study. An informed consent was also obtained. The demographic parameters, biochemistry, imaging, and liver biopsy findings were retrieved from records. Any patient with incomplete data that prevented the calculation of FIB-4 index and NFS scores or making a diagnosis of NAFLD was excluded. Patients with dual aetiology and hepatocellular carcinoma (HCC) were excluded. Patients on hepatotoxic drugs and complementary and alternate medicines were also excluded.

The liver biopsy was evaluated by the same pathologist. Diagnosis and severity of NAFLD were made as per NIH's NASH committee (NASH CRN) grading system. The severity of steatohepatitis was assessed using the NAFLD activity score (NAS) score (Steatosis, hepatocellular ballooning and lobular inflammation). Fibrosis stage was scored separately according to the NASH CRN system: fibrosis stage 0 = no fibrosis; stage 1 = centrilobular pericellular fibrosis; stage 2 = centrilobular and periportal fibrosis; stage 3 = bridging fibrosis; and stage 4 = cirrhosis. Alternate aetiology was ruled out using a history of significant alcohol or drug intake, HBsAg, anti–hepatitis C virus (HCV), ferritin, transferrin saturation, antinuclear antibody (ANA), and ceruloplasmin in all patients. Height, weight, and BMI were measured in all patients.

On basis of histological finding, the study population was divided into two groups. NASH CRN fibrosis stages 0–2 were considered mild to moderate fibrosis, and stages 3–4 were considered as advanced fibrosis (AF). NFS was calculated using the formula (−1.675 + 0.037 × (age [years]) + (0.094 × BMI [kg/m2]) + (1.13 × impaired fasting glucose or diabetes [yes = 1, no = 0]) + (0.99 × AAR) -(0.013 × platelet [10⁹/L]) -(0.66 × albumin [g/dl]). A score of more than 0.676 was taken as the cut-off to predict AF in accordance with previous studies. FIB-4 index was calculated using the formula (Age [years] × AST [IU/L])/(platelet count [ × 10⁹/L] × √ALT[IU/L]), and a score of more than 2.67 was taken as the cut-off to predict AF in accordance with the previous studies. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the receiver operating characteristic (ROC) curve for NFS score and FIB-4 index to diagnose AF were calculated.

The Youden Index was used to determine the optimal upper cut-off. Data were entered using Microsoft Excel, and statistical analyses were performed using IBM Statistical Package for the Social Sciences (SPSS) Statistics version 26. A Shapiro–Wilk test was performed to analyse the distribution of the continuous data. Normally distributed data were expressed as mean ± standard deviation and were compared using the unpaired t-test (independent sample t test). Other continuous data were expressed as median and range and were compared using non-parametric tests (Mann–Whitney U Test). Qualitative data were presented as frequency and percentage. The chi-square test and Fischer exact test were used to determine associations between the variables categorised. For all tests, a P-value of <0.05 was considered as statistically significant. Overall diagnostic accuracy was evaluated by determining the area under the ROC curve (AUROC). Diagnostic performances were determined by sensitivity, specificity, PPV, and NPV. ROC curves were compared according to Hanley and McNeil's method and the performances of the two scores were compared using the McNemar test.

RESULTS

During the study period, 147 patients with NAFLD fulfilled the inclusion criteria. The demographic and biochemical parameters are shown in (Table 2). The mean age was 47 years. The study population was predominantly male (84; 57.1%). The number of patients with NASH CRN fibrosis stage 0 was 55, stage 1 was 15, stage 2 was 21, stage 3 was 10, and stage 4 was 46. AF was seen in 56 patients (38.1%). In the study age, BMI, AST, low platelets, and low albumin were found to be associated with AF (Table 2). Older age was a risk factor for AF (42.06 ± 11.86 vs 55.05 ± 8.74; P < 0.001). Sex was not a risk factor for AF. The mean BMI in mild to moderate fibrosis was 25.55 ± 3.61 and in AF was 27.04 ± 4.27 and was significant (P < 0.03). For one unit increase in BMI, the odds of developing AF increased by 1.108 (95% confidence interval [CI]: 1.01, 1.21; P: 0.026). Mean AST levels (23 vs 50.5; P < 0.001), AST/ALT (0.92 ± 0.39 vs 1.4 ± 0.54; P < 0.001), low Platelet count (137,750 vs 264,253), and low albumin level (3.29 ± 0.7 vs 4.5 ± 0.46) were significantly associated with AF.

Table 2.

Clinical Characteristics of patients with biopsy proven NAFLD.

Variables (n = 147)	Mild to moderate fibrosis (n = 91)	Advanced fibrosis (n = 56)	P value	Test
Age	42.06 ± 11.86	55.08 ± 8.74	<0.001	t test
Sex
Male (n = 84) Female (n = 63)	48 (57%) 43 (68%)	36 (43%) 20 (32%)	0.16	Chi-square test
BMI	25.55 ± 3.61	27.04 ± 4.27	0.03	t test
DM
Yes (n = 48) No (n = 99)	12 (25%) 79(79.8)	36 (75%) 20 (20.2%)	<0.001	Chi-square test
Height (cm)	161.4 ± 9.58	164.15 ± 8.31	0.06	t test
Platelet	264,253 ± 60,166	137,750 ± 78,886	<0.001	t test
AST	23 (12, 82)	50.5 (13, 182)	<0.001	Mann–Whitney U Test
ALT	25.2(10,88)	30.5 (13,145)	0.19	Mann–Whitney U Test
AST/ALT	0.92 ± 0.39	1.40 ± 0.54	<0.001	t test
Albumin	4.5 ± 0.46	3.29 ± 0.7	<0.001	t test
INR	1.07 ± 0.231	1.08 ± 0.236	0.62	t test
NFS	−3.13 (−5.7,0.66)	1.4 (−3.33, 4.44)	<0.001	Mann–Whitney U Test
FIB-4	0.67 (0.17, 5.21)	5.03 (0.56, 17.5)	<0.001	Mann–Whitney U Test

Abbreviations: ALT = alanine aminotransferase; AST = aspartate aminotransferase; BMI = body mass index; DM; FIB-4 = Fibrosis Score 4; INR = Indian rupee; NFS = non-alcoholic fatty liver disease (NAFLD) Fibrosis Score.

The mean value for NFS in mild to moderate fibrosis was – 3.13(−5.70 to 0.66), whereas for AF, it was 1.4 (−3.33 to 4.44) (Table 2). Using 0.676 as the cut-off, NFS had a sensitivity of 67.86%, a specificity of 100%, an NPV of 83.49 %, and a PPV of 100% to diagnose AF with an AUROC of 0.951 (Figure 1). The FIB-4 index had a mean value of 0.67 (range: 0.17–5.21) in mild fibrosis group and 5.03 (0.56–17.50) in the AF group (Table 2). Using the cut-off as 2.67 to predict AF, the FIB-4 index had a sensitivity of 67.8%, specificity of 95.6%, NPV of 82.86%, and PPV 90.48% to diagnose AF with AUROC of 0.891 (Figure 1).

Figure 1.

AUROC of NFS and FIB-4. Abbreviations: AUROC = area under the receiver operating characteristic curve; FIB-4 = Fibrosis Score 4; NFS = non-alcoholic fatty liver disease (NAFLD) Fibrosis Score.

DISCUSSION

NAFLD is emerging as a common liver disease in India with prevalence in the general population ranging from 25% to 49%.^11,12 Given the severity of the condition, it is crucial to identify patients with high rates of morbidity and mortality early. Among the factors determining the severity of NAFLD, the degree of hepatic fibrosis is the most essential factor allowing clinicians to estimate the long-term prognosis, such as the development of hepatocellular carcinoma, liver-related death or cardiovascular mortality.^13,14 Therefore, it is vital to promptly identify AF (≥stage 3 fibrosis) in patients with NAFLD. Liver biopsy is the gold standard for staging and identifying fibrosis in NAFLD patients.^13,14 However, it is not suitable for a routine screening use due to its invasive nature, complications, possibility of sampling error, and high cost.¹⁵ Techniques such as transient elastography and magnetic resonance elastography are not easily available in community setting and are expensive for routine screening.^14,16,17 Therefore, a simple, inexpensive, and NITs to identify and quantify liver fibrosis is necessary. Non-invasive fibrosis scoring systems–based easily available and cheap serologic tests such as the FIB-4 index and NFS have been developed as screening tools to assess the degree of fibrosis in liver disease. However, a validation study on such scoring systems in the Indian population is essential as NFS and FIB-4 score were developed and validated in the Western population. The aim of this study was to validate the diagnostic accuracy of non-invasive fibrosis scoring systems (FIB-4 index and NFS), compared to that of liver histology, to predict AF in a cohort of South Indian patients with NAFLD.

The study was conducted in the department of hepatology at a tertiary care hospital. The clinical, laboratory, and histology data of all patients who had undergone liver biopsy in our institution between July 2018 and July 2020 were retrieved. Those who had a biopsy diagnosis of NAFLD, NASH, NAFLD with fibrosis, and NASH cirrhosis and had all laboratory data to calculate NITs and etiologic workup for cirrhosis done within 30 days of liver biopsy formed the study cohort. The diagnosis of NAFLD was independently verified in all patients. During the study period, we had 147 patients who were confirmed to have NAFLD on laboratory parameters and liver biopsy. Diagnosis and severity of NAFLD were made as per NASH CRN grading system. On basis of histological finding, the study population was divided into two groups. NASH CRN fibrosis stages 0–2 were considered mild to moderate fibrosis, and stages 3–4 were considered as AF. We calculated the NFS and FIB-4 index for all patients using an online calculator. We used cut-off points of 0.676 for NFS and 2.67 for FIB-4 index to diagnose AF. These cut-off points were previously validated in the Western and other Asian population. We then proceeded to validate these cut-offs in our cohort of South Indian NAFLD patients with liver biopsy as the gold standard.

Angulo et al. in 2007 first published the utility of NFS in identifying patients with NAFLD and AF.¹⁸ In a cohort of 733 patients divided into estimation group and validation group, using a higher cut-off of 0.676, they could diagnose AF with a PPV of 90% in the estimation group and of 82% in the validation group. By applying this model, they could have avoided biopsy in 75% with correct prediction in 90%.¹⁸ The FIB-4 index was first validated in a cohort of HIV/HCV coinfected patients by Sterling et al. They showed a PPV of 65% and a specificity of 97% when an upper cut-off of 3.25 was used to diagnose AF.¹⁹ Similar findings were shown by Valet-Pichard A et al. in a cohort of chronic HCV patients.²⁰ Shah AG et al. validated the FIB-4 index in NAFLD and found that a FIB-4 index score of >2.67 had a PPV of 80%, and an AUROC of 0.802 (95% CI: 0.768–0.847) for diagnosis of AF in NAFLD. FIB-4 was shown to be superior to NFS, AST/ALT ratio, and APRI in the diagnosis of AF in NAFLD.²⁰ Since then, the FIB-4 index with an upper cut-off of >2.67 has been validated in NAFLD across various ethnic groups.²¹

Various meta-studies have looked at the diagnostic utility of NFS and FIB-4 index in predicting AF. In a recent meta-analysis examining the diagnostic accuracy of the two methods for AF, which included 13,764 patients from 32 trials for the FIB-4 index and 13,337 patients from 33 studies for NFS, the FIB-4 index had a combined pooled sensitivity of 0.42 (95%CI: 0.33 to 0.51) and a pooled specificity of 0.93 (95% CI: 0.91 to 0.95).²¹ The area under the receiver operating characteristic curve (AUROC) of summary receiver operating characteristic (SROC) for FIB-4 index was 0.76 (95% CI: 0.74 to 0.81). For NFS to predict AF, the pooled sensitivity was 0.38 (95% CI: 0.28 to 0.50), and the pooled specificity was 0.94 (95% CI: 0.90 to 0.96). The AUC of SROC for NFS was 0.74 (95% CI: 0.71 to 0.79).²² In another meta-analysis that included 37 studies with 5393 cases with FIB-4 index and 3248 with NFS, the AUROCs for FIB-4 index was 0.76 and that for NFS was 0.73 for identifying AF.²³ In an earlier meta-analysis of four studies with 1038 adult patients, at a 0.676 cut-off, NFS had a pooled sensitivity of 0.27 (95% CI: 0.19–0.35) and a specificity of 0.98 (95% CI: 0.96–0.98), and the AUROC of 0.647 ± 0.2208 was used to predict AF.²⁴ In our study, there were 91 patients (62%) with mild to moderate fibrosis (stages 0–2) and 56 patients (38%) with AF(stages 3–4). Using 0.676 as the upper cut-off value to diagnose AF, NFS had a sensitivity of 67.86%, a specificity of 100%, an NPV of 83.49 %, and a PPV of 100%, with an AUROC of 0.951 (Figure 1). Using 2.67 as the upper cut-off to diagnose AF, FIB-4 index had a sensitivity of 67.8%, specificity of 95.6%, NPV of 82.86%, and PPV 90.48%, with an AUROC of 0.891(Figure 1). Our study findings show a higher sensitivity and similar specificity with a higher AUROC for both NFS and FIB-4 index to predict AF than the two previously published meta-analyses.

Both NFS and FIB-4 index are cheap and easy to use in out-patient settings and help to identify patients who need further evaluation and reference to a specialist. Previous studies have shown that sequential testing with FIB-4 index followed by shear wave elastography was cost effective in the community setting.²⁵ In a resource-poor country as India, NFS and FIB-4 index can be used to better utilise the resources and save cost. NFS and FIB-4 index are calculated using easily available serologic tests that are routinely done as part of evaluation of patients with metabolic syndrome and diabetes. The cost for doing NFS or FIB-4 index in an out-patient setting would be around Indian rupee (INR) 1000 compared to INR 5000 for a fibroscan, INR 12,000 for Magnetic resonance elastography (MRE), and INR 15,000 for a liver biopsy at our institute. We have not done a cost effectiveness analysis in our present study, and further studies specifically designed to look at cost effectiveness of NITs are needed. The present study also showed that advanced age, high BMI, and the presence of diabetes were risk factors for AF. A high AST, higher AST/ALT ratio, low platelet, and low albumin also indicated more AF in NAFLD (Table 2). In the current study, AF was seen in 33 % of patients with a BMI of <23 indicating lean NASH in the cohort of South Indian patients.

The main drawback of our study is its retrospective nature and that the study population was selected from a cohort of patients who underwent a liver biopsy for various reasons ranging from evaluation of deranged liver function, tests, aetiology of cirrhosis, and liver-donor evaluation. But still, it's the first study attempting to validate NFS and FIB-4 index in the South Indian population for diagnosis of AF (stages 3–4). It showed that NFS and FIB-4 index can be used to identify high-risk patients with NAFLD who require specialist care. Further studies are needed to see if these findings can be reproduced in the primary care setting.

NITs are an easy method to diagnose AF in NAFLD patients. In the current study, both NFS and FIB-4 index showed good diagnostic accuracy in predicting AF in NAFLD patients from South India. Both NFS and FIB-4 index are calculated using easily available parameters and are cheap and easy to administer in a busy clinical practice. In a resource-poor country as India, it's important to differentiate between patients who require specialist care and those who can be followed-up in the primary care pathway. The current study showed that both NFS and FIB-4 index helps to identify high-risk patients with AF who need specialist care. It can, therefore, help in resource utilisation and make healthcare cost effective. Further studies are needed to see if these diagnostic accuracies can be reproduced in primary care and to assess the cost effectiveness of these studies in primary care.

Credit authorship contribution statement

Conceptualisation: Charles Panackel.

Data Curation: Joe Francis Mathew.

Formal Analysis: Joe Francis Mathew.

Methodology: Charles Panackel.

Project Administration: Charles Panackel.

Resources: Nita John, Mathew Jacob.

Supervision: G N Ramesh.

Validation: Joe Francis Mathew.

Visualisation: Joe Francis Mathew.

Writing – Original Draft: Joe Francis Mathew.

Writing – Review and Editing: Charles Panackel.

Conflicts of interest

The authors have none to declare.

Acknowledgements

None.

Funding

The authors have nothing to declare.