Rethinking noninvasive steatosis indices: Structural limitations and misclassification across the body mass index spectrum

Abstract

The rebranding of nonalcoholic fatty liver disease (NAFLD) to metabolic dysfunction-associated steatotic liver disease (MASLD) has shifted the clinical focus toward underlying metabolic drivers. This necessitates a re-evaluation of noninvasive diagnostic tools to better align with metabolic dysfunction as the primary driver of hepatic steatosis. Commonly used indices, including the fatty liver index, hepatic steatosis index, and NAFLD liver fat score, were originally developed to estimate liver fat in general populations but are now applied to diagnose or screen for NAFLD in diverse clinical settings. However, their diagnostic performance is not consistent across different patient populations. In individuals with obesity, these diagnostic indices tend to over-identify hepatic steatosis due to excessively high sensitivity and low specificity. Conversely, in lean populations, these same indices often have low sensitivity, leading to under-diagnosis. These findings indicate that steatosis indices are strongly influenced by body composition and metabolic context. This review proposes a unified conceptual framework in which these indices are interpreted as body composition-dependent tools rather than direct measures of hepatic fat. Their structural reliance on anthropometric variables may result in systematic misclassification across the body mass index spectrum, limiting their utility as universal diagnostic classifiers. Accordingly, a shift toward context-dependent interpretation and integrated diagnostic strategies is required. Effective MASLD management in clinical practice requires combining noninvasive markers with imaging, fibrosis staging, and metabolic assessment to refine risk stratification, particularly in atypical cases such as severe obesity or lean MASLD.

Key Words: Metabolic dysfunction-associated steatotic liver disease; Fatty liver index; Hepatic steatosis index; Nonalcoholic fatty liver disease liver fat score; Body mass index; Misclassification; Noninvasive indices; Hepatic steatosis

Core Tip: While noninvasive steatosis markers are widely used, their diagnostic accuracy varies across the body mass index spectrum. Their reliance on anthropometric parameters leads to systematic errors, resulting in overestimation of steatosis and underestimation in lean individuals. Consequently, these tools should be utilized in a context-dependent manner, interpreted alongside imaging and metabolic assessments.

INTRODUCTION

Metabolic dysfunction-associated steatotic liver disease (MASLD), formerly nonalcoholic fatty liver disease (NAFLD), has been redefined to better reflect the central role of metabolic dysfunction in hepatic steatosis. According to recent consensus statements, most patients previously classified as having NAFLD meet the criteria for MASLD, with the updated definition incorporating cardiometabolic risk factors into the diagnostic framework[1-4]. MASLD represents a major global health burden and is increasingly recognized as a systemic metabolic disease closely linked to obesity, insulin resistance[5], and type 2 diabetes mellitus[6-10].

In clinical practice, noninvasive indices such as the fatty liver index (FLI), hepatic steatosis index (HSI), and NAFLD liver fat score (NLFS) remain widely used because they are simple, inexpensive, and based on routinely available clinical parameters. These indices were originally developed and validated in general populations, demonstrating acceptable diagnostic performance[11-13]. However, their applicability across diverse patient populations remains uncertain, particularly across different ethnic and metabolic backgrounds[14]. Even among individuals with a relatively low body mass index (BMI), hepatic steatosis and insulin resistance may coexist, highlighting the limitations of anthropometric-based indices[15]. Furthermore, genetic factors, including patatin-like phospholipase domain-containing protein 3 variants and loci identified through genome-wide association studies, drive hepatic fat accumulation independently of obesity, contributing to heterogeneity in metabolic phenotypes[16,17]. In addition, sex differences influence disease presentation and progression, adding another layer of complexity not captured by conventional indices[18].

However, accumulating evidence suggests that the diagnostic performance of these indices varies across patient populations. Studies indicate that diagnosing hepatic steatosis in obese individuals often leads to overestimation, necessitating higher diagnostic thresholds to maintain accuracy due to reduced specificity[19-22]. Conversely, the diagnostic accuracy of these indices may differ in lean patients with NAFLD/MASLD, raising concerns about under-detection of the disease[23,24]. These findings indicate that the diagnostic performance of steatosis scores is significantly influenced by body composition and metabolic context, limiting their generalizability across diverse populations. Importantly, existing literature reports these limitations sporadically across different clinical contexts, failing to provide a comprehensive, unified framework. Consequently, the core structural issue – the reliance of these indices on anthropometric and metabolic variables – remains poorly understood. This dependence may lead to systematic misclassification of hepatic steatosis across the BMI spectrum, with potential underestimation in lean individuals and overestimation in obese individuals.

This review integrates current findings to establish a unified framework for understanding the limitations of noninvasive steatosis indices. By highlighting population-dependent behavior and structural constraints, this work clarifies the appropriate clinical role of these tools in the evolving MASLD landscape.

NLFS, HSI, AND FLI

Noninvasive steatosis indices such as the NLFS, HSI, and FLI have been widely used in clinical practice; however, their diagnostic performance is not uniform across different populations. Previous studies have highlighted important limitations of steatosis biomarkers, including variability in performance and dependence on metabolic and anthropometric factors[25,26]. In addition, external validation studies have demonstrated that these indices often perform less accurately when applied to populations different from the original development cohort, raising concerns about their generalizability[27]. Growing evidence indicates that these indices are unreliable at extreme BMI levels, as their values tend to reflect general body measurements (e.g., weight or waist circumference) rather than actual liver fat content.

In individuals with obesity, several studies have reported reduced specificity and the need for higher thresholds to maintain diagnostic accuracy, reflecting a tendency toward overestimation of hepatic steatosis. In this context, Farina et al[19] comprehensively evaluated these indices in an obese cohort, demonstrating that conventional cut-offs lacked clinical utility due to excessive sensitivity and reduced specificity. Similar findings have been reported in bariatric surgery cohorts, where conventional indices required recalibration to detect clinically meaningful hepatic steatosis[20,21]. Importantly, substantial weight loss following bariatric surgery has been shown to markedly reduce hepatic steatosis and improve steatohepatitis, highlighting the dynamic relationship between adiposity and liver fat content[28]. In addition, weight reduction achieved through lifestyle or dietary interventions has been consistently associated with improvement in hepatic steatosis and metabolic parameters[29]. These findings suggest that the relationship among BMI, lipid metabolism, and hepatic fat accumulation differs substantially in individuals with severe obesity and is highly responsive to changes in metabolic status, thereby affecting the performance of standard scoring systems.

Conversely, the diagnostic performance of these indices may also be compromised in lean populations. The applicability of conventional cut-offs at the lower end of the BMI spectrum remains uncertain, raising concerns regarding potential under-recognition of hepatic steatosis in lean NAFLD/MASLD[23,24]. Importantly, lean NAFLD represents a clinically distinct phenotype characterized by specific metabolic features and risk factors despite the absence of overt obesity[6]. Furthermore, lean individuals with NAFLD have an increased risk of metabolic disorders, indicating that this phenotype is not benign despite the absence of excess adiposity[30]. Notably, meta-analytic evidence has revealed that lean and obese individuals with NAFLD share a remarkably similar cardiometabolic risk profile, suggesting that relying on BMI alone fails to capture the underlying metabolic dysfunction[31]. Moreover, because non-obese individuals with NAFLD face an increased risk of mortality, the clinical significance of this leaner phenotype may be underestimated[24]. These findings suggest that failing to recognize steatosis in lean individuals is not merely a diagnostic issue but may have serious long-term health consequences. Taken together, these findings indicate that steatosis scores do not behave consistently across the BMI spectrum but are instead strongly influenced by body composition and metabolic context.

A key explanation for this phenomenon is the structural design of these indices. Because many common scores rely on BMI or BMI-related variables, they are inherently tied to overall body shape rather than actual liver fat content. Figure 1 shows that this dependency can result in systematic errors in classifying hepatic steatosis. This phenomenon is consistent with the pathophysiological role of adiposity in NAFLD, in which obesity drives hepatic fat accumulation through complex metabolic and inflammatory pathways[32,33]. In this context, indices incorporating anthropometric variables may primarily capture the metabolic burden associated with obesity rather than directly quantifying hepatic steatosis[27]. In individuals with lower BMI, these indices demonstrate reduced sensitivity, leading to potential underestimation of liver fat. Conversely, in higher BMI populations, they tend to produce elevated values independent of actual fat accumulation, resulting in a risk of overestimation.

Figure 1 Conceptual illustration of misclassification of hepatic steatosis across the body mass index spectrum. Commonly used non-invasive indices for hepatic steatosis – such as the fatty liver index (FLI), hepatic steatosis index (HSI), and nonalcoholic fatty liver disease liver fat score (NLFS) – include body mass index (BMI) or BMI-related components as key elements in their calculation. In individuals with lower BMI, these indices tend to yield lower values, reducing sensitivity and leading to underestimation of steatosis. Conversely, in individuals with higher BMI, these indices yield elevated values independent of hepatic fat content, resulting in overestimation. The arrows indicate the direction of increasing misclassification risk across the BMI spectrum. It illustrates the structural dependence of these indices on BMI and the resulting bidirectional misclassification.

This limitation reflects a more fundamental issue in the development of these indices. Most hepatic steatosis scores are developed through regression modeling of general populations, where BMI and waist circumference act as the primary predictors. While this approach improves statistical accuracy within derivation cohorts, it effectively positions these indices as proxies for metabolic load rather than direct measurements of hepatic fat content. Consequently, their validity may be compromised when applied to populations with divergent phenotypes, such as individuals with severe obesity or lean MASLD. This conceptual limitation suggests that simple recalibration of cut-offs may not fully resolve the problem.

In addition, controlled attenuation parameter (CAP), obtained via vibration-controlled transient elastography, enables the noninvasive assessment of hepatic steatosis by measuring ultrasound attenuation in the liver[34]. CAP was introduced as a noninvasive tool for assessing hepatic steatosis, having been evaluated against histological and imaging standards[21,22]. Magnetic resonance (MR)-based techniques, including MR spectroscopy, have been established as reliable methods for quantifying hepatic triglyceride content in vivo[35]. MR-based methods, particularly MR imaging (MRI)-proton density fat fraction (PDFF), provide a precise, noninvasive, and quantitative method for measuring liver fat, strongly correlating with histology and demonstrating high accuracy across a wide range of hepatic steatosis levels[24,36,37]. Comparative studies have demonstrated that MRI-PDFF offers superior diagnostic accuracy compared with CAP, particularly across different grades of steatosis[38]. However, despite their superior accuracy, the availability and cost of these imaging modalities limit their widespread use in routine clinical practice, necessitating a balanced approach that integrates multiple diagnostic tools[39,40].

These findings underscore that imaging-based methods offer a more direct, accurate assessment of hepatic fat than biomarker-based indices, although their clinical utility is limited by high costs, restricted availability, and complex technical requirements. Although imaging provides a superior reference standard, noninvasive indices remain essential in clinical practice. They should be utilized as contextual, complementary tools integrated with imaging and metabolic assessments, particularly for patient populations at risk of misclassification.

CLINICAL IMPLICATIONS

From a clinical perspective, the limitations of noninvasive steatosis tests hinder effective MASLD assessment and management. Given the escalating metabolic and clinical burden of MASLD, accurate risk stratification that looks beyond mere fat detection is crucial[5]. While hepatic steatosis represents a key feature of MASLD, fibrosis has been consistently identified as the strongest predictor of long-term outcomes, including liver-related complications and overall mortality[41-44]. Differentiating simple steatosis from advanced stages like nonalcoholic steatohepatitis is clinically important, as they vary in disease progression and treatment response[45]. Furthermore, studies indicate that advanced fibrosis correlates with increased mortality, driven primarily by complications related to the liver and cardiovascular system[42,43]. In this context, composite noninvasive tools such as the FibroScan-aspartate aminotransferase score have been developed to improve the identification of patients with active steatohepatitis and significant fibrosis, illustrating the shift toward integrated diagnostic strategies[46]. Therefore, reliance on indices that primarily reflect hepatic fat content may not adequately capture clinically relevant disease severity.

In addition, the misclassification of hepatic steatosis associated with these indices may directly affect clinical decision-making. In individuals with obesity, overestimation of steatosis may lead to unnecessary further testing or overdiagnosis, whereas in lean individuals, underestimation may result in missed or delayed recognition of MASLD. Such bidirectional misclassification may contribute to inappropriate risk stratification and suboptimal management in clinical practice.

Moreover, MASLD is increasingly recognized as a systemic, multisystem disorder rather than just a liver condition, with strong links to cardiovascular disease and metabolic complications[47]. Cardiovascular disease represents a leading cause of mortality in this population[48-50], and meta-analytic evidence has demonstrated that NAFLD/MASLD is associated with an increased risk of incident cardiovascular events[51]. Furthermore, NAFLD/MASLD contributes to atherosclerosis and cardiometabolic dysfunction[52-54]. Because steatosis scores are influenced by metabolic and anthropometric factors, they may partly reflect cardiometabolic burden; however, their ability to predict cardiovascular outcomes is not well established when used in isolation.

These considerations highlight the need for a more integrated diagnostic approach. Recent advances in noninvasive diagnostic methods have been proposed to improve the accuracy and clinical applicability of MASLD assessments[55]. Noninvasive steatosis scores should not be used in isolation; rather, they must be interpreted alongside patient characteristics, including BMI, metabolic profiles, and specific clinical context. When available, advanced imaging techniques such as MRI-PDFF or CAP should be used to complement these diagnostic indices, especially in cases when the diagnosis remains unclear. Importantly, histological classification remains the reference standard for characterizing disease activity and distinguishing steatosis from steatohepatitis[56]. Furthermore, longitudinal studies have demonstrated that NAFLD/MASLD may progress over time, highlighting the need for dynamic and risk-oriented assessment rather than reliance on a single cross-sectional measure[57].

Ultimately, noninvasive steatosis indices should not be regarded as definitive diagnostic tools but as context-dependent screening instruments. Their appropriate use requires understanding their structural limitations and population-specific behavior, underscoring the importance of individualized interpretation in the clinical management of MASLD. These considerations also highlight the need for more comprehensive diagnostic strategies that integrate multiple domains of disease assessment and risk stratification[58-60].

CONCLUSION

While noninvasive indices are vital for evaluating hepatic steatosis, their limitations are intrinsic rather than incidental. Because these indices rely on body composition and metabolic factors, their diagnostic accuracy is heavily influenced by how they are built, leading to consistent misclassification across the entire BMI range. This fundamental constraint challenges their validity as universal diagnostic tools.

Importantly, the poor performance of BMI at its extreme ranges stems from fundamental conceptual flaws in the design of these indices rather than just inaccurate thresholds. Therefore, simply adjusting the cut-off points, while beneficial in limited scenarios, will not fully resolve the issue.

Noninvasive steatosis scores should be re-evaluated rather than expanded in clinical decision-making. Instead of serving as definitive diagnostic classifiers, these scores should be repositioned as context-dependent tools that offer indirect insights into metabolic burden. Their use should be complemented by imaging modalities, fibrosis assessment, and comprehensive metabolic evaluation, particularly in populations with atypical phenotypes such as severe obesity or lean MASLD.

As the MASLD framework evolves, we must move beyond simple score-based classifications toward integrated, phenotype-aware assessment strategies. Recognizing the structural limitations of current indices is essential to avoid misinterpretation in clinical practice and to guide the development of more reliable and clinically relevant diagnostic tools.