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.

Insulin resistance is fundamentally a physiological condition where target cells in muscles, fat, and liver become less responsive to insulin, a hormone essential in regulating blood glucose levels. This impaired responsiveness necessitates higher insulin secretion to maintain homeostasis, eventually leading to hyperinsulinemia and metabolic dysregulation. The complexity of insulin resistance lies not only in its multifactorial etiology but also in its asymptomatic nature during the early stages, which delays diagnosis and therapeutic measures.

Traditional methods for predicting insulin resistance predominantly rely on biochemical markers, including fasting insulin and glucose levels, homeostatic model assessment (HOMA-IR), and oral glucose tolerance tests. While these diagnostic tools provide valuable insights, they fall short in capturing the nuanced heterogeneity of insulin resistance across diverse populations, especially among pediatric patients. Shah and Tryggestad emphasize that reliance on static, single-time-point measurements often overlooks dynamic metabolic changes, reducing the predictive reliability of these methods.

Moreover, the authors delve into the limitations inherent in current clinical practices, such as the gold-standard hyperinsulinemic-euglycemic clamp technique. Although highly accurate, this technique is labor-intensive, expensive, and impractical for widespread clinical use, particularly in large-scale screening programs. Its complexity restricts its application to specialized research settings, thus widening the gap between research findings and clinical reality.

Recognizing these challenges, Shah and Tryggestad advocate for the integration of emerging technologies and multi-dimensional data sources to enhance model accuracy. Advances in genomics, metabolomics, and proteomics have opened new avenues for identifying molecular signatures associated with insulin resistance. These omics-based approaches, combined with robust computational models, promise a more holistic understanding of individual risk profiles, facilitating personalized medicine.

One of the pivotal points highlighted in the article is the potential role of machine learning and artificial intelligence in revolutionizing predictive modeling for insulin resistance. These computational tools excel at detecting complex, non-linear patterns within high-dimensional datasets that traditional statistical methods may miss. The authors suggest that employing AI-driven algorithms can improve the sensitivity and specificity of risk prediction, enabling earlier interventions and better clinical outcomes.

The paper also underscores the critical need to validate new models across diverse demographic cohorts. Insulin resistance manifests differently depending on age, ethnicity, genetic background, and lifestyle factors. Therefore, inclusive and representative datasets are paramount to developing universally applicable prediction tools. The authors stress that pediatric populations, often underrepresented in research, warrant special attention since early-life metabolic disturbances can set the trajectory for chronic diseases later in life.

Environmental and behavioral factors, including diet, physical activity, and endocrine disruptors, are also integral to understanding insulin resistance but remain challenging to quantify accurately. Shah and Tryggestad encourage the incorporation of real-world longitudinal data, such as continuous glucose monitoring and wearable device metrics, to capture lifestyle influences dynamically. This approach could provide a richer context for interpreting biological measurements.

In reviewing the landscape of current predictive methods, the article notes the discrepancies in cut-off values and diagnostic criteria across different clinical guidelines, which further complicate diagnosis and treatment planning. Standardizing definitions and harmonizing methodologies are crucial steps to ensure consistency in research and patient care. The authors call for collaborative efforts among clinicians, researchers, and policymakers to establish universally accepted standards.

The potential impact of improved prediction models extends beyond individual patient care; it has profound implications for public health initiatives. Better forecasting of insulin resistance prevalence could inform targeted prevention programs, reduce healthcare costs, and ultimately curb the global burden of diabetes and related metabolic disorders. Shah and Tryggestad argue that investment in this area is not merely academic but a societal imperative.

Furthermore, the article highlights the challenges faced in translating research insights into clinical practice. Regulatory hurdles, data privacy concerns, and the need for clinician education on emerging tools must be addressed to ensure successful implementation. The authors advocate for an interdisciplinary framework that bridges bioinformatics, clinical expertise, and patient engagement to harness the full potential of novel predictive models.

In their concluding remarks, Shah and Tryggestad envision a future where insulin resistance prediction is embedded within a precision medicine paradigm. This vision includes routine screening augmented by sophisticated algorithms that integrate biological, environmental, and behavioral data streams. Such a paradigm shift would enable proactive, personalized management strategies to halt or reverse metabolic dysfunction before it progresses.

In summary, the review presented by Shah and Tryggestad serves as a clarion call for innovation in the prediction of insulin resistance. While acknowledging the strides made with existing methods, the authors illuminate the pressing need for more accurate, accessible, and dynamic models that reflect the complexity of this metabolic condition. Their work lays a robust foundation for future research aimed at transforming insulin resistance from a silent affliction to a manageable clinical target, ultimately improving health outcomes globally.

Subject of Research: Insulin resistance prediction methods and the necessity for improved predictive models.

Article Title: Insulin resistance: current methods of predicting and need for improved models.

Article References:
Shah, R., Tryggestad, J.B. Insulin resistance: current methods of predicting and need for improved models. Pediatr Res (2026). https://doi.org/10.1038/s41390-026-04976-8

Image Credits: AI Generated

DOI: https://doi.org/10.1038/s41390-026-04976-8