Daily Endocrinology Research Analysis
Three papers stand out today: an AI model (DiaCardia) that detects prediabetes from single-lead ECGs with robust internal and external validation; a large multicenter MASLD analysis showing that liver stiffness measurement better reflects true severity when FIB-4 and LSM are discordant; and a mechanistic study revealing a perilysosomal Ca2+-mTORC1-TRPML1 pathway by which palmitate impairs β-cell autophagy, suggesting new therapeutic angles.
Summary
Three papers stand out today: an AI model (DiaCardia) that detects prediabetes from single-lead ECGs with robust internal and external validation; a large multicenter MASLD analysis showing that liver stiffness measurement better reflects true severity when FIB-4 and LSM are discordant; and a mechanistic study revealing a perilysosomal Ca2+-mTORC1-TRPML1 pathway by which palmitate impairs β-cell autophagy, suggesting new therapeutic angles.
Research Themes
- AI-enabled metabolic screening
- Non-invasive fibrosis risk stratification in MASLD
- β-cell lipotoxicity mechanisms and autophagy
Selected Articles
1. Artificial intelligence identifies individuals with prediabetes using single-lead electrocardiograms.
DiaCardia, a LightGBM-based model using ECG features, identified prediabetes with AUROC 0.851 internally and 0.785 externally; a single-lead (lead I) version achieved AUROC 0.844. Predictive features included higher R-wave amplitudes in leads aVL/I and lower peak interval dispersion, and performance remained strong after propensity score matching.
Impact: This work enables scalable, non-invasive, and low-cost prediabetes screening using ubiquitous ECGs and wearables, potentially shifting detection upstream.
Clinical Implications: ECG-based AI could pre-screen high-risk individuals in primary care or via wearables, guiding confirmatory lab testing and early preventive interventions.
Key Findings
- LightGBM-based DiaCardia achieved AUROC 0.851 internally and 0.785 in external validation.
- Single-lead (lead I) ECG model retained high accuracy (AUROC 0.844; sensitivity 82.3%; specificity 70.2%).
- Predictive ECG features included higher R-wave amplitude in leads aVL and I, and smaller peak interval dispersion.
- Performance was robust after propensity score matching for six confounders.
Methodological Strengths
- External validation across cohorts and propensity score matching.
- Single-lead ECG model demonstrates practical scalability for wearables.
Limitations
- Sample size and cohort details are not specified in the abstract; true generalizability across diverse ethnicities and devices requires further study.
- No prospective outcome data to show that ECG screening improves clinical endpoints.
Future Directions: Prospective pragmatic trials integrating ECG-AI triage into primary care and wearable platforms to assess impact on detection rates, time-to-diagnosis, and preventive outcomes.
BACKGROUND: Early detection of prediabetes is crucial for diabetes prevention, yet it remains challenging due to its asymptomatic nature and low screening rates. This study aimed to develop and rigorously validate artificial intelligence (AI) models to identify individuals with prediabetes solely using electrocardiograms (ECGs). METHODS: We defined prediabetes/diabetes based on fasting plasma glucose ≥ 110 mg/dL, hemoglobin A RESULTS: The best-performing model, a LightGBM-based algorithm we termed DiaCardia, achieved an area under the receiver operating characteristic curve (AUROC) of 0.851 in the internal test dataset (sensitivity: 85.7%, specificity: 70.0%). The model demonstrated robust generalizability, achieving an AUROC of 0.785 in the external validation cohort. Furthermore, DiaCardia maintained substantial predictive ability (AUROC: 0.789) after adjustment for six major confounders using propensity score matching. Higher R-wave amplitude in leads aVL and I, and smaller peak interval dispersion were prominent predictors. Notably, a version of DiaCardia using only single-lead (lead I) ECG data achieved a comparable AUROC of 0.844 (sensitivity: 82.3%; specificity: 70.2%). CONCLUSIONS: This study establishes that an AI model, DiaCardia, can accurately identify individuals with prediabetes from an ECG alone, with performance that is robust across different patient cohorts and independent of major clinical confounders. Our highly generalizable, single-lead DiaCardia model offers a promising solution for scalable prediabetes screening via wearable devices, potentially enabling early, home-based detection and transforming diabetes prevention strategies.
2. Histological severity and hepatic outcomes in patients with MASLD and discrepant FIB-4 and liver stiffness measurement.
In 12,950 MASLD patients from 16 centers, about 30% showed discordant FIB-4 and LSM. Advanced fibrosis and liver-related events aligned more strongly with high LSM, even when FIB-4 was low, while high FIB-4 with low LSM was not associated with increased events.
Impact: Clarifies how to interpret discordant non-invasive tests, supporting LSM-driven adjudication when FIB-4 and LSM disagree, with implications for surveillance and referral.
Clinical Implications: When FIB-4 and LSM are discordant, prioritize high LSM for risk stratification and closer follow-up; high FIB-4 with low LSM may not warrant escalation.
Key Findings
- Among 12,950 patients, 30% in tertiary settings had discordant FIB-4 and LSM results.
- Advanced fibrosis prevalence was highest in high-FIB-4/high-LSM (62.4%) and substantial in low-FIB-4/high-LSM (30.6%).
- Liver-related event rates were 0.67, 1.19, 2.58, and 21.30 per 1,000 person-years across low/low, high/low, low/high, high/high groups.
- Low-FIB-4/high-LSM (aSHR 4.2) and high-FIB-4/high-LSM (aSHR 21.3) had higher LRE risk versus low/low, whereas high-FIB-4/low-LSM did not.
Methodological Strengths
- Large multicenter cohort with biopsy subset and long median follow-up (47.4 months).
- Clear predefined cutoffs and competing risk modeling for LREs.
Limitations
- Tertiary-center population may limit generalizability; referral bias possible.
- Overall LRE incidence was low; thresholds (FIB-4 1.3; LSM 8 kPa) may not apply universally.
Future Directions: Prospective algorithms integrating LSM, serum markers, and imaging to resolve discordance; cost-effectiveness analyses for surveillance strategies.
BACKGROUND/AIMS: Current guidelines recommend a 2-step approach for identifying advanced fibrosis in metabolic dysfunction-associated steatotic liver disease (MASLD), using FIB-4 followed by liver stiffness measurement (LSM) via vibration-controlled transient elastography. However, some patients may exhibit discordant results. This study evaluates the histological severity and outcomes in patients with discordant FIB-4 and LSM results. METHODS: This secondary analysis of the VCTE-Prognosis study included 12,950 patients evaluated for MASLD at 16 tertiary centers, of whom 2,915 underwent liver biopsy. Patients were categorized into four groups based on established FIB-4 (1.3) and LSM (8 kPa) cutoffs. RESULTS: F3-F4 fibrosis was observed in 6.4%, 13.7%, 30.6%, and 62.4% in low-FIB-4-low-LSM (n=6,403), high-FIB-4-low-LSM (n=3,017), low-FIB-4-high-LSM (n=1,363), and high-FIB-4-high-LSM (n=2,167) groups, respectively. During a median follow-up of 47.4 months, 248 patients experienced hepatic decompensation, hepatocellular carcinoma, liver transplantation, or liver-related death. The incidence rates of liver-related events (LREs) were 0.67, 1.19, 2.58, and 21.30 per 1,000 person-years, respectively. Compared to low-FIB-4-low-LSM patients, those with low-FIB-4-high-LSM (adjusted subdistribution hazard ratio [aSHR] 4.2) and high-FIB-4-high-LSM (aSHR 21.3) had a significantly higher risk of LREs, while high-FIB-4-low-LSM patients did not. Similar findings were observed when hepatic decompensation and hepatocellular carcinoma were analyzed separately. CONCLUSIONS: Approximately 30% of patients in tertiary centers exhibit discordant FIB-4 and LSM results, with LSM more likely reflecting true severity. While some patients with discordant results may have advanced fibrosis, the overall incidence of LREs remains low.
3. Palmitate impairs autophagic degradation via oxidative stress-perilysosomal Ca2+ overload-mTORC1 activation in pancreatic β-cells.
Palmitate triggers perilysosomal Ca2+ overload, sustained lysosomal mTORC1 activation, TRPML1 suppression, and autophagosome accumulation in β-cells. Pharmacologic mTORC1 inhibition, mitochondrial ROS scavenging, Ca2+ entry reduction, KATP opening, or ER Ca2+ ATPase activation restored TRPML1 activity, autophagic flux, and β-cell survival.
Impact: Identifies a mechanistic oxidative stress–Ca2+–mTORC1 axis linking lipotoxicity to autophagy failure via TRPML1, revealing actionable nodes to protect β-cells.
Clinical Implications: Suggests repurposable strategies (e.g., L-type Ca2+ channel blockers, KATP openers) and mTORC1/TRPML1 modulation to preserve β-cell autophagy under lipotoxic stress; requires translational validation.
Key Findings
- Palmitate increased perilysosomal Ca2+, sustained lysosomal mTORC1 activation, suppressed TRPML1, and caused autophagosome accumulation in β-cells.
- mTORC1 inhibition or mitochondrial superoxide scavenging prevented Ca2+ abnormalities and autophagy defects.
- Reducing Ca2+ entry, opening KATP channels, or activating ER Ca2+ ATPase restored TRPML1 activity, autophagic flux, and β-cell survival.
Methodological Strengths
- Multi-level mechanistic interrogation linking organellar Ca2+, mTORC1, TRPML1, and autophagy.
- Pharmacologic rescue experiments across multiple intervention points.
Limitations
- Predominantly in vitro mechanistic study; human translational validation is needed.
- Lipotoxic conditions and doses may not fully recapitulate in vivo human exposures.
Future Directions: Validate the Ca2+-mTORC1-TRPML1 axis in human islets and metabolic disease models; assess β-cell protective effects of repurposed Ca2+ modulators and mTORC1/TRPML1-targeted therapies in vivo.
Saturated fatty acids impose lipotoxic stress on pancreatic β-cells, leading to β-cell failure and diabetes. In this study, we investigate the critical role of organellar Ca2+ disturbance on defective autophagy and β-cell lipotoxicity. Palmitate, a saturated fatty acid, induced perilysosomal Ca2+ elevation, sustained mTORC1 activation on the lysosomal membrane, suppression of the lysosomal transient receptor potential mucolipin 1 (TRPML1) channel, and accumulation of undigested autophagosomes in β-cells. These Ca2+ aberrations with autophagy defects by palmitate were prevented by an mTORC1 inhibitor or a mitochondrial superoxide scavenger. To alleviate perilysosomal Ca2+ overload, strategies such as lowering extracellular Ca2+, employing voltage-gated Ca2+ channel blocker or ATP-sensitive K+ channel opener effectively abrogated mTORC1 activation and preserved autophagy. Furthermore, redirecting perilysosomal Ca2+ into the endoplasmic reticulum (ER) with an ER Ca2+ ATPase activator, restores TRPML1 activity, promotes autophagic flux, and improves survival of β-cells exposed to palmitate-induced lipotoxicity. Our findings suggest oxidative stress-Ca2+ overload-mTORC1 pathway involvement in TRPML1 suppression and defective autophagy during β-cell lipotoxicity. Restoring perilysosomal Ca2+ homeostasis emerges as a promising therapeutic strategy for metabolic diseases.