The study, published in Health Science Reports, set out to quantify how frequently insulin resistance occurs among overweight patients with sonographically confirmed fatty liver disease, and to identify clinical markers that might flag the condition at the bedside. Between February and October 2022, the team recruited 50 adults aged 18 or older from the outpatient department of Hepatology. All participants had a body mass index above the Asia-Pacific threshold of 23 kg/m², ultrasound evidence of hepatic steatosis, and—critically—no established diabetes. Anyone who consumed 20 grams or more of alcohol daily, showed impaired glucose tolerance on a standard oral glucose tolerance test, or carried a prior diagnosis of chronic liver disease such as viral hepatitis, Wilson’s disease, autoimmune hepatitis, cholestatic liver disease, hemochromatosis or portal hypertension was excluded, ensuring the results reflected the earliest, pre-diabetic stages of metabolic dysfunction.

Insulin resistance was measured using the Homeostasis Model Assessment of Insulin Resistance, or HOMA-IR, a widely validated surrogate index calculated from fasting plasma glucose and fasting insulin concentrations. Participants fasted for at least eight hours before venous blood was drawn, and samples were analyzed on a Siemens Atellica Solution immunoassay and clinical chemistry platform. A HOMA-IR value of 2.5 or above was used as the cutoff for insulin resistance, a threshold supported by validation studies against the hyperinsulinemic-euglycemic clamp—the gold-standard technique for measuring whole-body insulin sensitivity—and by reference data from a normoglycemic Bangladeshi cohort in which the 75th percentile HOMA-IR was 2.6.

The results were striking. Sixty percent of participants—30 of the 50—were biochemically insulin resistant, with a 95 percent confidence interval ranging from 45.2 to 73.6 percent. The median HOMA-IR across the cohort was 2.85, and the mean was 3.24 ± 1.9, nearly double the mean of 1.66 reported in a large earlier study of normoglycemic Bangladeshi adults. The distribution of HOMA-IR values was markedly right-skewed, with skewness of 1.739, indicating that a subset of participants carried very high degrees of resistance even while their fasting glucose remained within normal limits. In other words, the pancreas was still compensating, pumping out enough insulin to keep blood sugar in check—but at a metabolic cost that eventually runs out.

Among the clinical signs examined, one stood out dramatically: acanthosis nigricans, the velvety, darkened skin plaques that classically appear on the neck, axillae and other flexural surfaces in states of chronic hyperinsulinemia. Every single participant with acanthosis nigricans—100 percent—was insulin resistant by HOMA-IR, and their mean HOMA-IR of 5.79 was roughly double that of patients without the skin finding, who averaged 2.89. This association reached statistical significance at p = 0.037, leading the authors to suggest that a simple skin examination could serve as a powerful, zero-cost screening tool in resource-constrained outpatient settings where laboratory insulin assays are not routinely available.

By contrast, several surrogate markers that have gained popularity in metabolic research performed poorly in this specific cohort. The triglyceride-to-HDL cholesterol ratio, the METS-IR score—a novel index developed and clamp-validated in Mexican populations as a predictor of visceral adiposity and incident type 2 diabetes—and BMI itself all failed to show statistically significant correlations with HOMA-IR. Spearman correlation between BMI and HOMA-IR yielded a weak positive coefficient of 0.239 (p = 0.095), and log-transformed regression on BMI accounted for only about 5 percent of the variance. The METS-IR index showed only a marginal, non-significant trend (R² = 0.072, p = 0.062). Waist circumference fared somewhat better—72.7 percent of participants with increased waist circumference were insulin resistant compared with 50 percent of those with normal waist measurements—but the difference did not reach significance either.

The sonographic severity of fatty liver also told a surprising story. Using internationally accepted ultrasound criteria—increased hepatic echogenicity relative to the right kidney, posterior acoustic enhancement, and blurring of the portal vein walls—an expert radiologist graded each patient’s fatty infiltration. Yet the degree of steatosis bore no relationship to insulin resistance: median HOMA-IR was 2.998 in grade I, 2.795 in grade II, and actually lowest, at 2.31, in grade III, with no statistical significance by the Kruskal-Wallis test. This finding challenges the intuitive assumption that “more fat on ultrasound means more metabolic derangement,” and aligns with a growing body of evidence suggesting that insulin sensitivity depends less on the sheer quantity of liver fat than on the distribution and metabolic activity of adipose tissue and on hepatic insulin extraction.

Comorbidities told a quieter story. HOMA-IR values trended higher among the 11 participants with hypertension (4.17 ± 3.01) and the two with hypothyroidism (4.19 ± 3.36) compared with the 34 who had no comorbidity (2.93 ± 1.36), though these differences did not reach statistical significance. Metabolic syndrome, however, was rampant: 53.2 percent of the cohort met the revised NCEP ATP III criteria, substantially higher than the roughly 30 percent baseline prevalence reported for the general Bangladeshi population. Among women in the study, 74.1 percent had metabolic syndrome versus 25 percent of men—a highly significant difference—and insulin-resistant participants carried a prevalence of 69 percent compared with just 27.8 percent among insulin-sensitive ones (p = 0.006).

The findings arrive against a sobering demographic backdrop. MAFLD affects roughly a quarter of adults worldwide and spans a spectrum from simple steatosis through non-alcoholic steatohepatitis to cryptogenic cirrhosis and liver failure. South Asian populations carry a disproportionate share of this burden, exhibiting distinctive risk phenotypes—often characterized by visceral adiposity and insulin resistance at lower body weights than in European populations. Bangladesh-specific data indicate a MAFLD prevalence of 33.86 percent, and the World Health Organization reports that global obesity has tripled since 1975. The new study places these figures in a clinical context: in a tertiary hepatology clinic, the majority of overweight, non-diabetic patients with fatty liver were already well along the insulin resistance pathway that precedes type 2 diabetes.

The results echo findings from neighboring India, where comparable studies in MAFLD populations reported insulin resistance prevalence ranging from 60 to 97.5 percent depending on the cutoffs used. In Bangladesh itself, prior studies of obese patients without fatty liver screening found IR rates of 60.2 to 78.8 percent, while the only previous Bangladeshi study specifically linking MAFLD and insulin resistance—conducted in patients with impaired glucose tolerance rather than normoglycemic individuals—found 57.1 percent prevalence. The new work is notable precisely because it targets the window before glucose intolerance sets in, when intervention is most effective.

The authors caution that the single-center design and modest sample size of 50 patients limit generalizability, and that ultrasonography, while inexpensive and widely available, is less sensitive than liver biopsy for quantifying steatosis—limitations that may partly explain the weak correlations with BMI and surrogate lipid indices. Nevertheless, the clinical message is unambiguous. Regular screening for hyperglycemia and insulin resistance among overweight patients with fatty liver disease could identify individuals at high risk of progressing to type 2 diabetes, enabling timely lifestyle and pharmacologic interventions before irreversible complications—cardiovascular disease, neuropathy, nephropathy and advanced liver scarring—take hold. In a health system already straining under the burden of chronic disease, catching insulin resistance early may prove one of the most cost-effective investments available.

Fatty liver disease, characterized by excessive fat accumulation in liver cells, poses a significant health risk. The importance of early diagnosis cannot be overstated, as undiagnosed fatty liver disease can progress to severe complications such as cirrhosis and liver cancer. Traditional diagnostic methods, including ultrasonography, computed tomography (CT), and magnetic resonance imaging (MRI), while accurate, often necessitate sophisticated equipment that can be both costly and logistically challenging to implement in many medical settings.

In contrast, chest X-rays represent a more accessible diagnostic tool. These imaging studies are frequently utilized to evaluate the lungs and heart but simultaneously capture portions of the liver. This creates an opportunity for alternative diagnostic approaches for fatty liver disease, an area that has been largely underexplored until now. The synergistic potential of leveraging chest X-ray data to assess liver conditions underscores a critical evolution in medical practice, driven by innovative technology.

The research group utilized a retrospective approach, analyzing a total of 6,599 chest X-ray images derived from 4,414 patients. A noteworthy aspect of their methodology involved the application of controlled attenuation parameter (CAP) scores, which provide quantitative measurements of liver fat content. By training their AI model on this extensive dataset, the researchers were able to enhance its diagnostic accuracy substantially. The area under the receiver operating characteristic curve (AUC) for their model ranged impressively between 0.82 and 0.83, indicating a high level of reliability in distinguishing affected individuals from those without the condition.

Professor Uchida-Kobayashi expressed enthusiasm regarding the implications of their findings. She noted that the introduction of diagnostic methods utilizing readily available chest X-rays suggests the potential for widespread improvements in fatty liver disease detection. The future of such technology could lead to earlier interventions, ultimately altering the disease trajectory for many patients worldwide. Given the current global burden of fatty liver disease, this approach could be transformative in varying healthcare contexts.

The AI-driven detection model represents a convergence of art and science, melding the extensive capabilities of deep learning algorithms with the foundational principles of medical imaging. By enhancing diagnostic pathways, such innovations could streamline patient outcomes limiting the risk of severe disease progression. Such technology resonates with ongoing efforts to integrate artificial intelligence into various healthcare sectors, advocating for smarter, data-driven decision-making processes.

The remarkable implications of this study not only extend to improving diagnostic practices but resonate with larger shifts within the healthcare system. As healthcare professionals navigate the dual challenges of accessibility and accuracy, the AI model provides a beacon of hope for enhancing the standard of care. In particular, hospitals and clinics equipped with basic X-ray technology can bolster their diagnostic capabilities without incurring the substantial costs typically associated with ultrasound or MRI procedures.

Moreover, the researchers’ commitment to verifying their findings through rigorous standards signifies an important step in the journey from innovative research to clinical application. Validation in clinical settings remains paramount for adoption in routine practice, as it helps establish trust and efficacy among healthcare providers and patients alike. As these models continue to evolve, the ethical considerations surrounding AI in healthcare are equally pertinent, calling for responsible integration and transparency.

Furthermore, the initiative encourages a broader dialogue surrounding public health strategies, emphasizing the proactive management of liver health through accessible diagnostic tools. As the global healthcare landscape continues to evolve and respond to increasing rates of liver disease, innovative solutions like the AI model from Osaka Metropolitan University are invaluable. Such endeavors align seamlessly with the vision of empowering healthcare systems worldwide to incorporate cutting-edge technologies that cater to diverse patient needs.

This moment in healthcare reflects not only a technological advancement but also a cultural shift towards embracing innovation for improved health outcomes. By leveraging existing resources—like chest X-rays—this research exemplifies the capability of modern science to reimagine traditional practices in a way that is both efficient and effective. The future landscape of disease diagnosis may very well pivot on the marriage of artificial intelligence with established imaging techniques, heralding a new era in medical diagnostics.

As we witness increased interest in artificial intelligence applications in medicine, this research stands out as a noteworthy example of how AI can serve to bridge gaps in healthcare delivery. Researchers and practitioners alike may draw inspiration from this study, as it highlights the persistent necessity for innovation in health diagnostics, particularly for conditions that carry significant disease burdens like fatty liver disease.

In conclusion, the advent of AI-driven diagnostic tools capable of analyzing chest X-rays for fatty liver disease detection marks a pivotal moment in medical research and clinical practice. This transformative approach aligns seamlessly with the push for more accessible, accurate, and timely diagnoses, ultimately aiming to safeguard patient health and improve quality of life for millions. These advancements exemplify how science continually strives toward solutions that bridge the gaps between technology, medicine, and public health.

Subject of Research: Fatty liver disease
Article Title: Performance of a Chest Radiograph-based Deep Learning Model for Detecting Hepatis Steatosis
News Publication Date: 20-Jun-2025
Web References: Link to the Journal
References: None
Image Credits: Credit: Osaka Metropolitan University

Keywords

AI, Fatty Liver Disease, Chest X-rays, Deep Learning, Imaging Analysis, Medical Research, Healthcare Innovation, Osaka Metropolitan University.

Metabolic-associated fatty liver disease represents a spectrum of liver disorders characterized by excessive fat accumulation in the liver parenchyma, unrelated to significant alcohol consumption but tightly linked to metabolic dysregulation. The study by Chen et al. probes the intricate relationship between serum uric acid, a byproduct of purine metabolism, and creatinine, a muscle metabolism derivative, to compute a ratio that may serve as an accessible biomarker indicative of hepatic fat infiltration and ensuing metabolic derangements.

The relevance of SUA/Cr extends beyond its routine clinical measurement. Uric acid, while a well-known risk factor in adult metabolic syndromes, is increasingly recognized for its pro-inflammatory and pro-oxidative properties that could contribute to pathogenesis in juvenile cohorts. However, the influence of renal function on circulating uric acid levels warrants normalization against creatinine concentrations, thus justifying the use of the SUA/Cr ratio for enhanced specificity and reliability.

Chen and their team meticulously conducted a cross-sectional observational analysis involving a diverse cohort of children, stratifying participants based on the presence or absence of MAFLD confirmed via imaging techniques and biochemical assays. Their approach integrated comprehensive metabolic profiling, encompassing lipid panels, insulin resistance indices, and anthropometric measurements, to elucidate correlations that transcend simplistic associations.

One noteworthy aspect of the study lies in its methodological rigor. By excluding confounding factors such as overt renal impairment, acute infections, or genetic predispositions affecting uric acid metabolism, the researchers ensured that the observed associations reflect genuine pathophysiological mechanisms pertinent to MAFLD. The inclusion criteria, therefore, were stringent enough to bolster the validity and generalizability of their findings.

Their data compellingly demonstrate that children diagnosed with MAFLD exhibit significantly elevated SUA/Cr ratios compared to their non-MAFLD counterparts. This elevation correlates robustly with markers of insulin resistance, dyslipidemia, and adiposity indices, underscoring the SUA/Cr ratio as a promising surrogate marker for metabolic dysfunction underlying fatty liver pathology at a young age.

Delving deeper into mechanistic insights, the authors postulate the potential causative role of hyperuricemia-induced oxidative stress in promoting hepatic steatosis. Elevated uric acid levels impel mitochondrial dysfunction and activate inflammatory cascades within hepatocytes, exacerbating lipid accumulation and fibrosis progression. When adjusted for creatinine, the SUA/Cr ratio may magnify sensitivity to these pathological changes, especially in pediatric patients with variable muscle mass and kidney function.

This study also delineates age- and sex-specific reference intervals for the SUA/Cr ratio in children, a feature often overlooked in prior research. Establishing normative data is crucial for clinical translation, enabling practitioners to interpret ratios contextually rather than relying on adult or generalized pediatric cutoffs, which may result in misclassification or underdiagnosis.

Furthermore, the potential utility of SUA/Cr as a non-invasive, cost-effective screening tool for identifying children at elevated risk for MAFLD holds profound implications for public health strategies. Early identification could prompt timely lifestyle interventions aimed at curbing obesity, improving diet quality, and augmenting physical activity, thereby attenuating disease progression before irreversible liver damage ensues.

Beyond screening, the SUA/Cr ratio might serve as a valuable biomarker for monitoring therapeutic response in clinical trials assessing novel pharmacologic or behavioral interventions targeting metabolic dysfunction and hepatic steatosis. Serial measurements could offer dynamic insights into disease trajectory and treatment efficacy, facilitating personalized medicine approaches in pediatric hepatology.

While the study provides compelling evidence, the authors acknowledge inherent limitations, such as its cross-sectional design, which precludes causal inference, and the reliance on liver ultrasonography rather than histological confirmation for diagnosing MAFLD. Nonetheless, the findings lay a robust foundation for longitudinal cohort studies and randomized interventions to explore causality and therapeutic ramifications more conclusively.

This burgeoning field prompts reflection on the broader implications of uric acid metabolism in pediatric health, intertwining nephrology, endocrinology, and hepatology disciplines. It challenges clinicians and researchers alike to rethink conventional biomarkers, embracing integrated indices like the SUA/Cr ratio that may capture complex metabolic interplays more effectively.

Intriguingly, the research also raises questions about genetic variants influencing uric acid handling in children, which could modulate susceptibility to MAFLD. Future studies incorporating genomic analyses alongside biochemical assessments might unravel personalized risk profiles, advancing precision medicine paradigms.

From a global health perspective, the study underscores the urgency of addressing metabolic diseases in children, whose early onset portends a lifetime trajectory of morbidity and mortality. Integrating simple, reliable biomarkers such as SUA/Cr into routine pediatric screenings could revolutionize preventive medicine and resource allocation, prioritizing those most vulnerable.

Moreover, the accessibility of measuring serum uric acid and creatinine in various healthcare settings enhances the feasibility of widespread implementation, especially in resource-limited contexts where advanced imaging and invasive diagnostics remain impractical.

In conclusion, the pioneering work by Chen et al. elevates the serum uric acid to creatinine ratio as a potent, clinically relevant biomarker associated with metabolic-associated fatty liver disease in children. Their findings advocate for a paradigm shift towards incorporating metabolic ratios into early diagnostic frameworks, catalyzing proactive interventions that may mitigate the burgeoning burden of pediatric liver disease globally.

As the epidemic of MAFLD among youth intensifies, integrating biochemical markers such as SUA/Cr promises to refine early detection and personalized management, heralding a new era in pediatric hepatology. The implications ripple beyond hepatology, spotlighting the interconnectedness of metabolic health, renal function, and systemic inflammation in shaping pediatric disease landscapes.

This research not only illuminates a novel biomarker but also galvanizes multidisciplinary efforts to unravel and combat the complex metabolic derangements afflicting children in the modern era. Ongoing and future investigations inspired by these findings will undoubtedly deepen our understanding and inform evidence-based strategies to safeguard the health of generations to come.

Subject of Research: Correlation between serum uric acid/creatinine ratio and metabolic-associated fatty liver disease (MAFLD) in children

Article Title: Serum uric acid/creatinine ratio and metabolic-associated fatty liver disease risks in children

Article References:
Chen, B., Qu, X., Li, T. et al. Serum uric acid/creatinine ratio and metabolic-associated fatty liver disease risks in children. Pediatr Res (2025). https://doi.org/10.1038/s41390-025-04219-2

Image Credits: AI Generated

DOI: https://doi.org/10.1038/s41390-025-04219-2