Nonalcoholic fatty liver disease (NAFLD) was originally conceived as a binary entity and was specifically defined as >5% fat in hepatocytes among individuals who drink little or no alcohol (defined as <20 g/day for female and <30 g/day for male).¹ However, given the alarming rise in NAFLD prevalence as well as global alcohol consumption (AC), such a binary definition is now considered flawed, and NAFLD and AC are understood to be more intertwined.^2-4 Indeed, it is now recognized that there are overlapping biological processes that may contribute to both NAFLD and alcohol-related liver disease. Metabolic dysfunction-associated fatty liver disease was proposed based solely on metabolic criteria without the prerequisite to exclude patients with alcohol intake or other chronic liver diseases.⁵ However, this all-encompassing definition was met with skepticism, which led to reconsideration of the nomenclature based on a multi-society consensus, and the term “steatotic liver disease (SLD)” was introduced.⁶ SLD is a hierarchical term that further subdivides into metabolic dysfunction-associated steatotic liver disease (MASLD) and alcohol-associated liver disease (AALD) at two extremes, and another entity that overlaps with MASLD and AALD termed as MASLD with increased alcohol intake (MetALD), which includes people with MASLD and AC between 20 and 60 g/day.⁶ The importance of differentiating AALD from MASLD lies in its multiple implications. First, the entities have contrasting disease drivers, which alters the natural history of fibrosis progression, cirrhosis development, incident decompensation, and risk of hepatocellular carcinoma.^1,4 Furthermore, although lifestyle modifications, including improvement in metabolic profile in MASLD and alcohol cessation in AALD, are cornerstones of disease management, it is first necessary to establish primary drivers to determine the best pharmacotherapy, especially considering the evolving therapeutic options for MASLD.^1,6

Validated objective tools to determine the relative contribution of alcohol or metabolic risk factors in patients with features of overlap are lacking, and clinicians must rely heavily on self-reported alcohol intake.⁶ The Alcoholic Liver Disease/Nonalcoholic Fatty Liver Disease Index (ANI) was proposed to differentiate alcoholic liver disease (ALD) from NAFLD, wherein scores greater than zero incrementally favor ALD and scores less than zero incrementally favor NAFLD.⁷ The ANI score was shown to have excellent diagnostic accuracy (area under the curve [AUC]=0.97) in the derivation cohort and subsequently replicated cohorts.⁸ However, in its original description and subsequent validation, it was used primarily as a binary discriminant. Given this context, we aimed to identify subtypes of SLD and assess the performance of the ANI as a non-invasive test point-of-care score in differentiating SLD subtypes.

Adult patients (>18 years) presenting to the outpatient liver clinic for evaluation at a tertiary care center were screened for hepatic steatosis using magnetic resonance imaging-proton density fat fraction (MRI-PDFF; 3 Tesla Phillips mDixon Quant software). MRI-PDFF values >6.4% were indicative of significant steatosis.⁹ All patients underwent a standardized clinical evaluation in the form of history and anthropometric examination, which included age, sex, height, weight, body mass index (BMI). Asia-Pacific cut-offs for BMI were used: normal 18–22.9 kg/m², overweight 23–24.9 kg/m², and obese ≥25 kg/m².¹⁰

Patients were evaluated for features of metabolic syndrome, including diabetes mellitus, hypertension, high-density lipoprotein (HDL), and triglyceride levels. A detailed work-up was performed to rule out alternate causes of liver disease including chronic viral hepatitis, autoimmune hepatitis, and Wilson’s disease. MASLD, MetALD, and AALD were defined according to a multi-society consensus document.⁶ The cardiometabolic criterion for defining MASLD was at least one of the five following traits: (1) BMI≥23 kg/m² (as per Asian cut-offs); (2) fasting blood glucose ≥100 mg/dL, glycosylated hemoglobin ≥5.7%, or treatment for diabetes mellitus; (3) blood pressure >135/80 mmHg or treatment for hypertension; (4) plasma triglycerides >150 mg/dL or treatment with lipid-lowering drugs; and (5) plasma HDL <40 mg/dL for males or <50 mg/dL for females or treatment with lipid-lowering drugs.

Self-reported AC history was collected by two physicians and quantified in grams/day or per week as applicable. Significant AC was defined as consuming more than 50 g/day in females and more than 60 g/day in males, while moderate AC was defined as consuming one drink (20 to 50 g) or less in a day for females and two drinks (30 to 60 g) or less in a day for males. AC thresholds for the intermediate group with MetALD were 140 to 350 g/week (or 20 to 50 g/day) for females and 210 to 420 g/week (or 30 to 60 g/day) for males.^6,11 Those consuming alcohol amounts less than the threshold were categorized as MASLD and those greater than as AALD. Moderate AC was defined as alcohol intake up to 350 or 50 g/day for females and up to 420 or 60 g/day for males.

The ANI was calculated using an online available calculator (https://www.mayoclinic.org/medical-professionals/transplant-medicine/calculators/the-alcoholic-liver-disease-nonalcoholic-fatty-liver-disease-index-ani/itt-20434726). The score constitutes simple components of age, mean corpuscular volume, alanine transaminase (ALT), aspartate transaminase (AST), BMI, and gender. From these variables, an ANI is generated along with a percent probability of ALD.

Patients with established cirrhosis, alcoholic hepatitis, hepatocellular carcinoma or other malignancies, known hematological disorders, end-stage renal disease, advanced heart failure, pregnancy, and those with contraindications or technical limitations to MRI were excluded. This study was approved by the appropriate Institutional Review Board of Apollo Multispeciality Hospitals (IEC/BR/2023/10), and written informed consent was obtained from each patient prior to study inclusion. The study protocol conformed with the ethical guidelines of the 1975 Declaration of Helsinki.

Statistical analyses

According to the original study by Dunn et al.⁷, ANI validation sensitivity and specificity were 93.5% and 92%, respectively; thus, given alpha=0.05 and a marginal error rate of 0.05, the minimum sample size required was n=190. Descriptive parameters were reported as number (%), mean±standard deviation, or median (interquartile range [IQR]). Categorical variables were compared using Pearson’s χ² test. Continuous data comparison between two groups was performed using the Mann-Whitney U-test, and comparisons of more than two groups used the analysis of variance (ANOVA) or Kruskal-Wallis test as per the normality of the data. Area under the receiver operating characteristic curve (AUROC) with bootstrapped 95% confidence interval (CI) was generated for the ANI of MASLD and ALD. A P-value <0.05 was considered significant. All statistical analyses were performed with SPSS version 26.0 (IBM Co.), MedCalc 20.0 (MedCalc), and GraphPad Prism software (GraphPad Software Inc.).

A total of 202 patients satisfied the inclusion and exclusion criteria and were included. Of these, 77.2% were obese (BMI=27.5 kg/m² [IQR, 25.2 to 30.4]), while 42.1% and 38% had diabetes and hypertension, respectively, and 40.5% of the SLD cohort reported AC. The baseline characteristics of the SLD patients are shown in Table 1.

Of 202 patients, 120 (59%) satisfied the definition for MASLD, while 50 (24.7%) and 32 (15.8%) satisfied the definition for MetALD and ALD, respectively. Table 2 shows differences in individual parameters between MASLD, MetALD, and AALD. Hemoglobin, transaminase levels, and gamma glutamyl transpeptidase (GGT) were significantly different between the groups. However, cardiometabolic risk factors were similar across the groups. The ANI was higher in the AALD group (0.71 [IQR, –1.3 to 4.8]) in comparison with the MetALD (–1.45 [IQR, –2.4 to 0.28]) and MASLD (–3.7 [IQR, –7 to –1.6]) groups. Also, ANI for the MetALD group was higher than that of the MASLD group (Fig. 1). Liver fat content as measured by MRI-PDFF was similar across the groups (Supplementary Fig. 1).

The diagnostic AUROC for the ANI for MASLD was 0.79 (IQR, 0.72 to 0.84) with a sensitivity of 54.2% and specificity of 92.6% (cut-off <–3.5). The ANI for AALD was highest, with an AUROC of 0.80 (IQR, 0.74 to 0.86), sensitivity 78.1%, and specificity 71.1% at a cut-off point of >–1.49 (Fig. 2). As, AST/ALT ratio, and GGT have also been used as predictors of AALD. We compared the AUROCs of ANI, GGT, and AST/ALT, and ANI (0.80 [IQR, 0.74 to 0.86]) showed better performance compared with AST/ALT (0.75 [IQR, 0.69 to 0.81]) and GGT (0.74 [IQR, 0.67 to 0.80]) (Fig. 3). As GGT level was also significantly different between the groups, the ANI with the addition of GGT was also evaluated but was not improved (AUC_diff=0.004 [IQR, –0.004 to 0.01]; P=0.33) (Supplementary Fig. 2). The ANI was originally conceived as a binary score, with those greater than zero incrementally favoring AALD. In our cohort, 15 patients (46.8%) with AALD had ANI <0, which further suggested the impact of metabolic factors even in patients with AALD. Importantly, 13 (86.6%) of these patients were obese, and seven (46.6%) were diabetic. In contrast, among patients with AALD and ANI >0, eight (47%) were obese and 10 (58.8%) were diabetic (P=0.02 for obesity, P=0.49 for diabetes).

SLD was introduced as an overarching term to encompass all entities leading to hepatic steatosis. However, metabolic risk factors or alcohol are the most common causes of steatosis, and the new terminologies of MASLD, MetALD, and AALD help understand the disease as a spectrum and not binaries.¹² While liver biopsy remains the gold standard for differentiating these entities, it has limitations of invasiveness and overlapping histological features between NAFLD and ALD.^13,14 There remains a need for objective non-invasive markers for differentiation, for which the ANI has been useful. We analysed a cohort of patients with SLD across the MASLD–MetALD–AALD spectrum to assess the performance of the ANI in distinguishing the entities.

The ANI score was originally devised to differentiate between patients with ALD and NAFLD based on a biopsy based cohort.⁷ In the original description, ANI had a c-statistic of 0.989 in the derivation cohort.⁷ In the original description, the score was predominantly used as a binary cut-off that incrementally favored a diagnosis of ALD when ANI >0 and of NAFLD when ANI <0.⁷ Subsequently, the index was tested in a Chinese cohort of 139 patients, showing a sensitivity, specificity, and AUROC for ALD of 87.1%, 92.5%, and 0.934 (95% CI, 0.879 to 0.969), respectively, at a cut-off score of –0.22.⁸ The index was also tested in a Serbian cohort of 135 patients using a cut-off of <–0.66 to identify patients with NAFLD and yielded 96.7% specificity and 84.1% sensitivity.¹⁵

We conducted the first study of the ANI in a cohort of patients with SLD as defined by MRI-PDFF, which has been considered one of the most accurate measures of fat assessment, and assessed its performance for the MASLD, MetALD, and AALD spectrum.¹⁶ The ANI differed across the spectrum with significantly different scores between AALD, MetALD, and MASLD. The score showed excellent specificity and sensitivity for MASLD and MetALD, respectively. Additionally, we identified separate cut-offs, <–3.5 and >–1.49, for better identification of MASLD and AALD, respectively, with those falling between to be representative of MetALD. However, the diagnostic performance of the index in our cohort was inferior to previous findings (0.79 and 0.80 for MASLD and AALD, respectively). In the original study by Dunn et al.⁷, the c-statistic of the ANI score was 0.767 in one of the validation cohorts, indicating performance heterogeneity for different cohorts.

Despite performing inferior to original derivation studies, with a diagnostic AUC >0.8, the ANI appears to be a good additional tool based upon simple parameters to identify AC. However, while the ANI identifies significant and no-AC at the two extremes, better tools are needed in the overlap zone. Other biomarkers of AC including ethyl glucuronide, carbohydrate-deficient transferrin, and phosphatidyl ethanol may be of use in this setting, although such tools are not available for commercial use.¹⁷

A previous study showed that addition of GGT to the ANI further increased the AUC from 0.93 to 0.96.⁸ However, we failed to demonstrate any performance improvement with the addition of GGT. While GGT has often been labelled as a marker of ALD, a recent study form Hossain et al.¹⁸ indicated GGT to be a significant determinant of NAFLD after adjusting for the effects of glycaemic control and weight. Hence, while GGT may be elevated solely in AALD cohorts, in an amalgamation of a mixed cohort of MASLD an AALD, GGT alone loses its diagnostic utility.

Our study has several strengths, being the first to use the ANI in an SLD cohort. Additionally, the cohort was specifically defined using MRI-PDFF as an entry point and was stratified by AC with prospectively documented history. We suggest different cut-offs for excluding MASLD and AALD, with an intermediate zone indicating MetALD.

Our study also has certain limitations of which the key is the lack of liver biopsy, although there remain challenges in liver biopsy itself in differentiating between these entities.¹³ The history of alcohol intake was self-reported, and may be errors of underreporting, which remains a universal challenge and a key area for future research.⁶ Although our overall cohort was large, the number of subjects with AALD was small, which may be one of the reasons for a lower diagnostic performance in comparison to previous studies. However, in a real-life setting in patients with SLD, this is often the presentation. Last, an interesting aspect and limitation that emerges is the entire cohort having some degree of “metabolic dysfunction” as per the current MASLD criteria, which may be too liberal considering even one of the five risk factors to be enough for diagnosing metabolic dysfunction. A similar observation has recently been made by Anirvan et al.¹⁹, which calls for close analysis of the defining criteria especially in the face of the related global public health challenge.²⁰

The ANI differs significantly among MASLD, MetALD, and AALD and can perform well in identifying MASLD and AALD. Additional scoring systems and biomarkers are required to further stratify individual contributions of alcohol or metabolic risk factors in the MetALD group.

Supplementary materials can be found online at https://doi.org/10.7570/jomes23063.

The authors declare no conflict of interest.

Study concept and design: AR and MKG; acquisition of data: AR, AT, and NS; analysis and interpretation of data: AR and AVK; drafting of the manuscript: AR; critical revision of the manuscript: AD, AVK, and MKG; statistical analysis: AR and AD; administrative, technical, or material support: SJ and UG; and study supervision: AT, NS, and MKG.