Daily Endocrinology Research Analysis
Three endocrinology-related studies stand out today: a systematic review confirms pyridoxal-5'-phosphate as the most effective therapy for PNPO deficiency while highlighting liver safety monitoring; a large multi-cohort analysis shows that adding self-reported sleep traits does not improve type 2 diabetes risk prediction beyond existing models; and a real-world machine learning approach identifies individuals likely to have familial hypercholesterolemia who would be missed by LDL-C thresholds al
Summary
Three endocrinology-related studies stand out today: a systematic review confirms pyridoxal-5'-phosphate as the most effective therapy for PNPO deficiency while highlighting liver safety monitoring; a large multi-cohort analysis shows that adding self-reported sleep traits does not improve type 2 diabetes risk prediction beyond existing models; and a real-world machine learning approach identifies individuals likely to have familial hypercholesterolemia who would be missed by LDL-C thresholds alone.
Research Themes
- Therapeutic optimization in rare metabolic epilepsies (vitamin B6 pathway)
- Limits of incremental risk predictors in type 2 diabetes
- Machine learning case-finding for familial hypercholesterolemia in real-world EHRs
Selected Articles
1. Effectiveness of Pyridoxal-5'-Phosphate in PNPO Deficiency: A Systematic Review.
Across 30 studies reporting 49 PNPO-deficient patients treated with PLP, 77.6% achieved clinical seizure responsiveness and survival improved versus untreated siblings. Most PLP-responsive patients did not respond to pyridoxine, and liver toxicity occurred in approximately 20%—suggesting the need for careful dosing and liver monitoring while ensuring access to pharmaceutical-grade PLP.
Impact: This synthesis consolidates therapeutic evidence in an ultra-rare neurometabolic disorder and quantifies both benefit and safety signals, directly informing clinical management and regulatory needs for PLP access.
Clinical Implications: Use PLP as first-line therapy for PNPO deficiency with baseline and periodic liver function tests; avoid assuming pyridoxine responsiveness; advocate for reliable, regulated PLP formulations to ensure dosing accuracy.
Key Findings
- PLP achieved clinical seizure responsiveness in 38/49 (77.6%) PNPO-deficient patients.
- PLP treatment was associated with significantly improved survival versus untreated siblings (p < 0.001).
- Most PLP-responsive patients did not respond to pyridoxine (30/33, 90.9%), and liver toxicity occurred in 10/49 (20.4%), likely dose-related.
Methodological Strengths
- Registered protocol (PROSPERO) and multi-database search strategy.
- Quantitative synthesis of clinical response and survival with explicit adverse event reporting.
Limitations
- Evidence base relies on observational reports and case series with inherent bias.
- Dose regimens and product quality of PLP varied; limited mechanistic safety data on hepatotoxicity.
Future Directions: Prospective registries with standardized PLP dosing, pharmacovigilance for liver safety, and head-to-head comparisons with pyridoxine in genotype-defined PNPO deficiency.
Pyridox(am)ine 5'-phosphate oxidase (PNPO) deficiency is an ultrarare inherited neurometabolic disease, characterized by primarily neonatal-onset B6-responsive epileptic encephalopathies. Treatment often requires sustainable access to high-quality pyridoxal-5'-phosphate (PLP, i.e., active vitamin B6), although some patients (also) respond to pyridoxine (PN). While PN is authorized as a medicinal product, PLP is not, and this forces reliance on lesser-regulated food supplements, which risks dosing inaccuracies. This systematic review evaluates the effecti
2. Clinical utility of self-reported sleep duration and insomnia symptoms in type 2 diabetes prediction.
Extremes of sleep duration and insomnia symptoms were associated with higher incident T2D risk. However, adding these sleep traits to the QDiabetes risk score did not improve discrimination (C statistic ~0.893) or reclassification in UK Biobank and replication cohorts. A T2D polygenic score marginally improved prediction, with little incremental value from sleep traits.
Impact: Clarifies that commonly collected sleep measures, despite risk associations, do not justify inclusion in clinical T2D risk calculators beyond established variables—preventing model overfitting and complexity creep.
Clinical Implications: Maintain use of established risk calculators without adding self-reported sleep traits; consider polygenic scores as an adjunct in select contexts while acknowledging marginal gains.
Key Findings
- Sleep duration extremes and insomnia symptoms were associated with higher incident T2D risk.
- Adding sleep traits to QDiabetes did not improve discrimination (C statistic 0.8933 vs 0.8931–0.8939) or NRI (near zero with wide CIs).
- Including a T2D polygenic score modestly improved prediction (C 0.8945; NRI 0.20), with negligible further gains from adding sleep traits.
Methodological Strengths
- Large primary cohort with independent replication in multiple US cohorts.
- Robust performance metrics (Harrell’s C, NRI) and comparison with polygenic risk scores.
Limitations
- Sleep traits were self-reported, potentially introducing measurement error.
- Abstract does not report exact sample size or follow-up duration; residual confounding possible.
Future Directions: Evaluate objective sleep metrics (actigraphy, polysomnography) and circadian disruption markers; test calibration and clinical utility in diverse populations and care settings.
AIMS/HYPOTHESIS: Suboptimal sleep health is linked to higher risks for incident type 2 diabetes. We aimed to assess the clinical utility of adding self-reported sleep traits to a type 2 diabetes prediction model. METHODS: In this cohort study, we used UK Biobank data and Cox proportional hazards models to examine how self-reported sleep duration and insomnia symptoms were associated with incident type 2 diabetes risk. Harrell's C statistic and net reclassification improvement (NRI) were used to assess whether sleep traits improved the incident type 2 d
3. Performance of the FIND-FH machine learning algorithm for the identification of individuals with suspected familial hypercholesterolemia.
Among 93,418 individuals, FIND-FH flagged 340 as high probability for FH; 20–32% met modified Simon-Broome or DLCN criteria based on EHR. Over half (56%) warranted FH outreach, and notably 53% of those had peak LDL-C <190 mg/dL—patients likely missed by LDL-based screening alone.
Impact: Demonstrates that ML-based case-finding can prioritize FH evaluation beyond LDL thresholds, addressing underdiagnosis and enabling targeted preventive therapy.
Clinical Implications: Health systems can deploy FIND-FH to trigger outreach and confirmatory evaluation (clinical/genetic), particularly for patients with sub-190 mg/dL LDL-C yet high phenotypic risk.
Key Findings
- FIND-FH identified 340/93,418 individuals as high probability for FH in EHR data.
- Manual review found 20–32% met modified Simon-Broome or DLCN criteria for at least possible FH.
- Among outreach-eligible patients, 53% had LDL-C <190 mg/dL, indicating LDL-based strategies would miss many candidates.
Methodological Strengths
- Very large EHR-derived cohort with manual chart adjudication.
- Evaluation against established diagnostic criteria (modified Simon-Broome, DLCN).
Limitations
- Single-center setting with potential selection and documentation biases.
- Lack of routine genetic confirmation; many patients did not meet full diagnostic criteria based on available data.
Future Directions: Prospective, multi-center validation with genetic testing endpoints; integration of algorithm-triggered care pathways to measure impact on diagnosis rates and statin/PCSK9 uptake.
BACKGROUND: Familial hypercholesterolemia (FH) is an inherited cholesterol disorder that is markedly underdiagnosed. OBJECTIVE: This study evaluated the real-world performance of the Find, Identify, Network, Deliver-FH (FIND-FH) score, a novel machine learning algorithm, in identifying individuals with high likelihood of FH. METHODS: The FIND-FH model was applied to electronic health record (EHR) data from UT Southwestern Medical Center. Manual chart review was performed on those deemed high probability of FH (score >0.35) to assess accuracy of FH diagnosi