Citations

Abstract Patients with rare genetic disorders often wait years for a correct diagnosis, highlighting the urgent need for early and efficient triage for those at risk. Electronic health records (EHRs) are rich in useful information, yet the extent to which this data is sufficient to generate prediagnostic signals is uncertain. We trained machine learning models on billing codes and phenotypes extracted from clinical notes and applied these across eight conditions in a longitudinal cohort of roughly 3 million patient records from the Mayo Clinic, seeking early identification of conditions. The number of patients identified early varied by condition, ranging from 10% to 89% at 99% specificity. Median lead times exceeded one year prior to the first diagnostic billing code for most conditions, including Fabry disease (1,408 days), hereditary angioedema (2,755 days), neurofibromatosis type 1 (1,067 days), and hereditary hemorrhagic telangiectasia (2,320 days). We show that ML models trained on EHR data can prioritize patients for clinician review and confirmatory genetic testing and that the integration of diverse phenotypic data sources provides superior predictive value.

Background. Rare diseases collectively affect an estimated 300 million people worldwide, yet patients wait an average of 4-8 years for a correct diagnosis after visiting multiple physicians, accumulating unexplained findings, and suffering preventable disease progression before a unifying diagnosis is reached. Electronic health records (EHRs) capture this pre-diagnostic trajectory in rich detail, but conventional methods typically collapse years of longitudinal phenotypic signals into a single static snapshot; our approach instead models this temporal structure directly. Methods. We conducted a retrospective cohort study using longitudinal EHR data from approximately 3 million patients at Mayo Clinic Platform_Accelerate. Cases were patients with confirmed diagnoses of 10 rare diseases; controls (~5,200 per disease) had no record of any rare disease. HPO phenotype terms were extracted from clinical notes and laboratory results using a medspaCy NLP pipeline; ICD codes were grouped at three-character level. We developed and trained two temporal models, GRU-Attn and Conformer, that encode patient histories as quarterly time-binned sequences spanning up to 30 years, and compared these against three static baselines (CatBoost, XGBoost, logistic regression). Models were evaluated systematically at prediction horizons h = 1-10 years before diagnosis. Results. In the full held-out cohort benchmark, temporal models outperformed static baselines across the large majority of diseases and prediction horizons, with the advantage widening substantially as the horizon lengthened. At 1-2 years prior to diagnosis, mean AUROC was 0.928 (GRU-Attn) versus 0.871 (CatBoost); at 6-10 years prior to diagnosis, 0.726 versus 0.693, with static models degrading to near-chance for some diseases. Combining HPO and ICD features yielded the strongest performance, with a mean gain of +0.063 AUROC over the best single source at 1-2 years prior to diagnosis. Separately, in a manually curated cohort of 143 confirmed-undiagnosed cases with extended EHR histories, GRU-Attn identified 138 cases before first disease mention and all 143 cases before formal diagnosis at the prespecified 99th-percentile specificity threshold, with median lead times of 1.1-8.3 years and 1.9-21.7 years, respectively. Conclusions. By learning from the temporal trajectory of phenotypic signals rather than their static aggregate, temporal EHR models improved retrospective early detection performance across several rare diseases, most powerfully for conditions where pre-diagnostic findings accrue gradually and heterogeneously over years. These results show that longitudinal EHR trajectories carry rare-disease signal beyond static phenotypic burden and support prospective evaluation of temporal EHR modeling for rare-disease risk stratification.

Background and Purpose Stroke is a frequent manifestation of non-bacterial thrombotic endocarditis (NBTE), yet current knowledge largely stems from case reports and small case series. This systematic review summarizes stroke presentation, imaging features, management, and outcomes in patients with NBTE-related stroke. Methods We conducted a patient-level systematic review of published case reports and case series. PubMed, MEDLINE, Embase, Web of Science, and PsycINFO were searched from inception through August 2025 for studies reporting individual patients with NBTE-related stroke. Extracted data included demographics, vascular risk factors, stroke presentation, imaging features, NBTE etiology, treatment (unfractionated heparin/low-molecular-weight heparin [UFH/LMWH], vitamin K antagonists [VKA], direct oral anticoagulants [DOACs], and/or valvular surgery), and outcomes. Multivariable regression was used to explore factors associated with recurrent stroke and mortality. Results We included 236 patients with NBTE and stroke from 218 studies (median age 56 years [IQR: 43–63], 63% female). Left hemispheric stroke syndromes predominated among clinical presentations (27%), and multi-vessel territory ischemic infarcts (>2 territories) were observed in 65%. Conventional vascular risk factors were infrequent (0%−21%), whereas deep vein thrombosis (43%) and systemic embolism (41%) frequently preceded or coincided with stroke onset. Malignancy (70%) and autoimmune disease (25%) were the main NBTE etiologies. Anticoagulation was initiated in 70%, and 21% underwent valvular surgery. Over a median follow-up of 2.0 months, stroke recurred in 43% and death occurred in 49% of patients. Malignant etiology (aOR 10.3, P < 0.001) and metastases (aOR 3.9, P < 0.001) were associated with increased mortality, while VKA use (aOR 0.08, P < 0.001) and autoimmune etiology (aOR 0.1, P < 0.001) were associated with lower mortality. Malignancy (aOR 2.8, P < 0.001) and DOAC use (aOR 5.2, P = 0.012) were associated with increased risk of recurrent stroke. Conclusion NBTE should be considered in stroke patients lacking conventional vascular risk factors, particularly those with malignancy or autoimmune disease. Malignancy and DOAC use were associated with higher recurrent stroke risk, while VKA use was associated with lower mortality; however, these anticoagulant-class findings should be considered hypothesis-generating and interpreted cautiously given the case-report evidence base, small treatment subgroups, potential confounding, and limited follow-up. What this study adds By synthesizing patient-level data from published case reports and case series, this review provides a stroke-focused characterization of NBTE, including clinical presentation, infarct topography, underlying etiologies, management strategies, recurrent cerebrovascular events, and mortality. The findings highlight multifocal infarction, malignancy-associated NBTE, and recurrent stroke despite anticoagulation as key clinical concerns, while emphasizing that observed associations with anticoagulant class should be viewed as hypothesis-generating rather than definitive treatment guidance

Background: Patients with spontaneous coronary artery dissection (SCAD) are at increased risk of cardiovascular events, including dissection in other arterial territories such as cervical arteries. Those with combined coronary and cervical arterial dissection (CACAD) may have more widespread arterial abnormalities and worse outcomes, underscoring the need for better risk stratification and personalized care. Objective: This study examined the prevalence of cervical artery dissection (CvAD) in patients diagnosed with SCAD, evaluated potential risk factors for developing CvAD, and assessed their impact on clinical outcomes, including all-cause mortality, SCAD recurrence, and stroke. The goal was to enhance risk stratification by identifying SCAD patients at increased risk for CvAD. Design: In this retrospective cohort study, patients diagnosed with SCAD between 2018 and 2024 across Mayo Clinic sites were identified using ICD-10 codes. Manual chart review confirmed CvAD diagnoses. Patients were stratified based on the presence or absence of CvAD. Chi-square and independent t-tests were used to compare risk factors and comorbidities. Univariate screening and multivariable logistic regression were used to identify predictors of CvAD. Results: Among 1380 patients with SCAD, 131 (9.5%) had CvAD. The median follow-up was 2.02 years (IQR: 1.1–5.2 years). All CvAD diagnoses were identified after the index SCAD event through systematic post-SCAD vascular imaging. The mean age was similar between groups (51.5 ± 11.3 years in CACAD vs. 50.9 ± 10.7 years in SCAD-only, p = 0.540), and the majority were female (93.9% vs. 91.9%, p = 0.529). Multivariable regression identified extremity arterial dissection (OR: 8.13, 95% CI: 3.41–19.36, p < 0.001), fibromuscular dysplasia (FMD) (OR: 7.52, 95% CI: 4.67–12.10, p < 0.001), connective tissue disorders (CTDs) (OR: 6.85, 95% CI: 2.35–19.98, p < 0.001), and anxiety (OR: 1.69, 95% CI: 1.01–2.81, p = 0.044) as significant predictors of CvAD. CvAD was a strong independent predictor of stroke (OR: 6.37, 95% CI: 2.93–13.87, p < 0.001). Mortality and SCAD recurrence did not differ significantly between groups. Conclusions: Patients with CACAD demonstrate a systemic vascular phenotype with significantly higher rates of FMD, connective tissue disease, extremity arterial dissection, and stroke. These findings highlight the importance of comprehensive vascular screening and individualized monitoring in at-risk SCAD patients.

Variants in the Glucokinase Regulator ( GCKR ) gene are increasingly being detected with growing applications of genomic sequencing technologies. Despite the crucial roles of the GCKR gene in energy homeostasis, the clinical applications of GCKR variants are currently limited. We observed that a large proportion of GCKR variants that are available in gnomAD database ( https://gnomad.broadinstitute.org/ ) lacked adequate evidence of pathogenicity or benign impact. We highlight the clinical need for an improved understanding of GCKR variants by presenting 2 compelling cases of unrelated families (3 individuals in total) who harbor rare GCKR variants in the context of cornea arcus and hyperlipidemia. Our observations underscore the burgeoning need for functional as well as family segregation studies to facilitate the clinical applications of GCKR variants.

Background: Loeys–Dietz syndrome (LDS) is an autosomal dominant aortopathy characterized by aggressive aneurysm formation and arterial dissections. Pregnancy-related outcomes and timing of LDS diagnosis remain poorly characterized. Methods: Demographics, genetic, obstetric, and vascular data was collected from genetically or clinically confirmed individuals with LDS seen at the three Mayo Clinic sites from 2018 to 2025. Aneurysm progression, new aneurysm formation, and arterial dissections were recorded across all vascular beds. Vascular events were assessed during pregnancy, within 12 months postpartum, and during breastfeeding. Comparative analyses were performed between women with and without a history of pregnancy, and a single-arm descriptive analysis was conducted among patients who experienced vascular complications during the peripartum period. Continuous variables were compared using the Mann–Whitney U test, while categorical variables were analyzed using chi-square or Fisher exact tests. Results: Of 47 women with LDS, 24 had a history of pregnancy, accounting for 54 pregnancies. In the comparative analysis, age at LDS diagnosis differed significantly between women with and without a prior pregnancy: women without prior pregnancy were diagnosed at a younger age (median 23.5 years [IQR 10.8–41.0], n = 23) than those who had been pregnant (median 53.5 years [IQR 43.0–59.3], n = 24). Among pregnant women, the median age at first pregnancy was 28 years (IQR 23–34); only 4 (16.7%) knew their diagnosis before pregnancy. Of 54 pregnancies, 40 (74.1%) resulted in live birth, with 23 (57.5%) vaginal and 17 (42.5%) cesarean deliveries; preterm delivery occurred in 1 (2.5%) pregnancy, and postpartum hemorrhage in 2 (5.0%). No maternal deaths, aortic dissections, or uterine ruptures occurred during gestation or the first postpartum year. In addition, 14 women (58.3%) developed aneurysms, 6 (25.0%) experienced at least one arterial dissection, and 7 (29.2%) required surgical repair, predominantly involving the ascending and abdominal aorta. The prevalence of vascular complications did not differ significantly between groups. Conclusions: In this LDS cohort, pregnancy and the early postpartum period were not accompanied by acute aortic catastrophes, despite frequent diagnostic delay. Although women without prior pregnancy were diagnosed at a younger age, the overall burden of vascular and morphologic complications did not differ significantly by pregnancy history. These findings highlight the importance of long-term cardiovascular follow-up in women with LDS.

Abstract Background Vascular Ehlers-Danlos syndrome (VEDS) and Marfan’s syndrome are associated with cardiovascular complications, including arrhythmias. However, atrial fibrillation (AF) risk in these populations has not been frequently reported. This study assessed the AF risk in VEDS and Marfan's syndrome patients using propensity score-matched analysis from an international database. Methods The TriNetX Research Network, an international database with de-identified patient data from over 100 centers, was used to pull patients with a diagnosis of VEDS and Marfan's Syndrome. Baseline characteristics (e.g., demographics, cardiovascular comorbidities, etc.) were collected, and propensity score matching was performed to balance the cohorts. Rates of AF were collected across both cohorts, and risk analysis and Kaplan-Meier survival were used to assess outcomes. Results A total of 1,114 and 23,172 patients had VEDS and Marfan’s syndrome, respectively. After propensity score matching, each cohort included 1,091 balanced patients regarding demographics and clinical characteristics. When compared in the matched cohorts, AF incidence was lower in VEDS (4.6%) compared to Marfan’s syndrome (8.6%), with a risk difference of -0.040 (95% CI: -0.061, -0.020; z = -3.794; p &lt; 0.001). The risk ratio was 0.532 (95% CI: 0.381, 0.742), and the odds ratio was 0.509 (95% CI: 0.358, 0.726), indicating a significantly lower risk of AF in VEDS. Kaplan-Meier analysis showed an AF survival probability of 84.23% in VEDS and 78.20% in Marfan’s syndrome (log-rank p = 0.071). The hazard ratio was 0.723 (95% CI: 0.508, 1.029), which didn’t reach statistical significance (p = 0.579). Conclusion Patients with VEDS have a significantly lower risk of developing AF compared to those with Marfan’s syndrome. However, no significant difference in AF survival probability was observed. These findings suggest significant differences in the arrhythmic burden between these connective tissue disorders, which can direct future risk stratification and management strategies.Figure 1