Chao, Chieh‐Ju

Migraine with aura(MwA) is associated with an increased risk of stroke and adverse vascular outcomes compared to those with migraine without aura (MwoA). AI-ECG prediction models developed at our institution can evaluate the probability of atrial fibrillation (AF) and estimate a patient’s age based on a normal sinus rhythm (NSR) ECG. Delta age, AI-ECG estimated age minus chronological age, indicates endothelial dysfunction and is associated with cardiovascular mortality. We aim to assess how AI-ECG prediction model outputs, specifically the AF probability and delta age, are associated with adverse vascular outcomes in patients with migraine. Adult patients diagnosed with MwA and MwoA from 2000-2020 with at least one digital, standard 12-lead ECG were identified. The first ECG that showed NSR is used as the index ECG. Patients with any adverse vascular outcomes before the index ECG were excluded. Adverse vascular outcomes used as endpoints include acute ischemic stroke, acute myocardial infarction, deep vein thrombosis/pulmonary embolism, AF, and carotid artery dissection. A total of 31301 patients (13783 MwA, 17518 MwoA) were included in our analysis. The mean age at the time of the index ECG was 44.3 (14.5), and the average follow-up time was 71.6 (69.8) months. Based on the index ECG, the mean AF probability was significantly higher in MwA compared to MwoA (2.39 vs 2.06, p< 0.001), while the delta age was lower in MwA compared to MwoA (3.7 vs 4.2, p <.001). The prevalences of all vascular outcomes are summarized in Table 1. Multivariate Cox regression models showed that a higher AF prediction model output (every 10% increase), as well as a higher delta age (every 10-year increase), are associated with an increased risk of composite vascular outcomes (HR1.20, 95%CI 1.17-1.23, p<.001, and HR1.19, 95%CI 1.14-1.23, p < .001) [Table 2]. The Kaplan-Meier curve of adverse vascular outcome-free survival by AF prediction model output, using 1% as the cutoff (HR 1.49, 95% CI 1.41, 1.57, p<.001), and by delta age, using 2-year as the cutoff (HR 1.14, 95% CI 1.08, 1.21, p <. 001) are presented in Figure 1. Having aura, older age, male, coronary artery disease, congestive heart failure, diabetes, hypertension, and tobacco use were also associated with a higher risk of adverse vascular outcomes. Our results demonstrated the potential of using AI-ECG AF prediction model output and delta age to identify migraine patients at risk for adverse vascular outcomes.

Abstract Aims The increasing need for diagnostic echocardiography tests presents challenges in preserving the quality and promptness of reports. While Large Language Models (LLMs) have proven effective in summarizing clinical texts, their application in echo remains underexplored. Methods and results Adult echocardiography studies, conducted at the Mayo Clinic from 1 January 2017 to 31 December 2017, were categorized into two groups: development (all Mayo locations except Arizona) and Arizona validation sets. We adapted open-source LLMs (Llama-2, MedAlpaca, Zephyr, and Flan-T5) using In-Context Learning and Quantized Low-Rank Adaptation fine-tuning (FT) for echo report summarization from ‘Findings’ to ‘Impressions.’ Against cardiologist-generated Impressions, the models’ performance was assessed both quantitatively with automatic metrics and qualitatively by cardiologists. The development dataset included 97 506 reports from 71 717 unique patients, predominantly male (55.4%), with an average age of 64.3 ± 15.8 years. EchoGPT, a fine-tuned Llama-2 model, outperformed other models with win rates ranging from 87% to 99% in various automatic metrics, and produced reports comparable to cardiologists in qualitative review (significantly preferred in conciseness (P &lt; 0.001), with no significant preference in completeness, correctness, and clinical utility). Correlations between automatic and human metrics were fair to modest, with the best being RadGraph F1 scores vs. clinical utility (r = 0.42) and automatic metrics showed insensitivity (0–5% drop) to changes in measurement numbers. Conclusion EchoGPT can generate draft reports for human review and approval, helping to streamline the workflow. However, scalable evaluation approaches dedicated to echo reports remains necessary.

Abstract Aims : Direct oral anticoagulants (DOACs) are widely used for stroke prevention in people with non-valvular atrial fibrillation/flutter (NVAF). Anticoagulants have different bleeding profiles. Apixaban is associated with the lowest bleeding risk. This study evaluates the interaction between DOAC used and right ventricular systolic pressure (RVSP) on echocardiogram and bleeding rates. It was hypothesized that pulmonary hypertension may be associated with excess bleeding based on anticoagulant-specific pharmacologic profiles. Methods & Results : A retrospective analysis of a database was conducted. Multivariable regression models for bleeding rates were developed using a causative analysis approach among participants who had undergone an echocardiogram, and interaction terms for DOAC choice were incorporated. Stratum specific estimates were calculated using linear computation. Patients taking Apixaban had lower bleeding rates than those taking Rivaroxaban and Dabigatran. Among those without pulmonary hypertension, there was no difference in bleeding rates. Where RVSP was ≥45mmHg, there was a 90% increase in bleeding rates with use of Dabigatran (HR 1.99, p=0.024) and Rivaroxaban (HR 1.86, p=0.042) compared to Apixaban (Figure 1). Conclusions : Elevated RVSP is associated with excess bleeding in patients taking Rivaroxaban and Dabigatran, but not Apixaban, after controlling for confounding variables.

Abstract Objective To develop machine learning models using patient and migraine features that can predict treatment responses to commonly used migraine preventive medications. Background Currently, there is no accurate way to predict response to migraine preventive medications, and the standard trial‐and‐error approach is inefficient. Methods In this cohort study, we analyzed data from the Mayo Clinic Headache database prospectively collected from 2001 to December 2023. Adult patients with migraine completed questionnaires during their initial headache consultation to record detailed clinical features and then at each follow‐up to track preventive medication changes and monthly headache days. We included patients treated with at least one of the following migraine preventive medications: topiramate, beta‐blockers (propranolol, metoprolol, atenolol, nadolol, timolol), tricyclic antidepressants (amitriptyline, nortriptyline), verapamil, gabapentin, onabotulinumtoxinA, and calcitonin gene‐related peptide (CGRP) monoclonal antibodies (mAbs) (erenumab, fremanezumab, galcanezumab, eptinezumab). We pre‐trained a deep neural network, “TabNet,” using 145 variables, then employed TabNet‐embedded data to construct prediction models for each medication to predict binary outcomes (responder vs. non‐responder). A treatment responder was defined as having at least a 30% reduction in monthly headache days from baseline. All model performances were evaluated, and metrics were reported in the held‐out test set (train 85%, test 15%). SHapley Additive exPlanations (SHAP) were conducted to determine variable importance. Results Our final analysis included 4260 patients. The responder rate for each medication ranged from 28.7% to 34.9%, and the mean time to treatment outcome for each medication ranged from 151.3 to 209.5 days. The CGRP mAb prediction model achieved a high area under the receiver operating characteristics curve (AUC) of 0.825 (95% confidence interval [CI] 0.726, 0.920) and an accuracy of 0.80 (95% CI 0.70, 0.88). The AUCs of prediction models for beta‐blockers, tricyclic antidepressants, topiramate, verapamil, gabapentin, and onabotulinumtoxinA were: 0.664 (95% CI 0.579, 0.745), 0.611 (95% CI 0.562, 0.682), 0.605 (95% CI 0.520, 0.688), 0.673 (95% CI 0.569, 0.724), 0.628 (0.533, 0.661), and 0.581 (95% CI 0.550, 0.632), respectively. Baseline monthly headache days, age, body mass index (BMI), duration of migraine attacks, responses to previous medication trials, cranial autonomic symptoms, family history of headache, and migraine attack triggers were among the most important variables across all models. A variable could have different contributions; for example, lower BMI predicts responsiveness to CGRP mAbs and beta‐blockers, while higher BMI predicts responsiveness to onabotulinumtoxinA, topiramate, and gabapentin. Conclusion We developed an accurate prediction model for CGRP mAbs treatment response, leveraging detailed migraine features gathered from a headache questionnaire before starting treatment. Employing the same methods, the model performances for other medications were less impressive, though similar to the machine learning models reported in the literature for other diseases. This may be due to CGRP mAbs being migraine‐specific. Incorporating medical comorbidities, genomic, and imaging factors might enhance the model performance. We demonstrated that migraine characteristics are important in predicting treatment responses and identified the most crucial predictors for each of the seven types of preventive medications. Our results suggest that precision migraine treatment is feasible.

AbstractBackgroundDespite optimal medical therapy and cardiac resynchronization therapy (CRT), significant functional mitral regurgitation (MR) persisted in 30% of the patients and labeled as CRT nonresponders.AimsWe sought to study the impact of transcatheter edge‐to‐edge repair (TEER) in patients with symptomatic grade III and IV functional MR despite CRT.MethodsA retrospective analysis was conducted of all patients who had prior CRT for at least 6 months and underwent TEER for significant residual functional MR (grade ≥3) and symptomatic heart failure (HF) at our institution. The primary outcomes were the change in New York Heart Association classification (NYHA), MR grade, echo parameters, and NT‐ProBNP from baseline to 1‐year post‐procedure.ResultsA total of 28 patients were identified, mean age of 73 ± 6.7 years and 89% males. Procedure success was achieved in all patients. At 1‐year follow‐up, patients had lower MR grade (median 2, IQR 1 [1,2] vs. 4, IQR 1 [3,4]; p < 0.001), NYHA class (median 2, IQR 1 [2,3] vs. 3, IQR 1 [3,4]; p < 0.001), and NT‐ProBNP (7658 ± 11322 vs. 3760 ± 4431; p = 0.035) compared to before the TEER procedure. The left ventricular end‐diastolic volume (255 ± 59 vs. 244 ± 66 mm; p = 0.016) and the right ventricular systolic pressure (52 ± 14 mmHg vs. 37 ± 13 mmHg, <0.001) decreased.ConclusionPatients who remain symptomatic after CRT with severe functional MR had improved functional status and MR grade at 1‐year following TEER. There was a signal toward reverse remodeling.

Background: Conduction abnormality post-transcatheter aortic valve implantation (TAVI) remains clinically significant and usually requires chronic pacing. The effect of right ventricular (RV) pacing post-TAVI on clinical outcomes warrants further studies. Methods: We identified 147 consecutive patients who required chronic RV pacing after a successful TAVI procedure and propensity-matched these patients according to the Society of Thoracic Surgeons (STS) risk score to a control group of patients that did not require RV pacing post-TAVI. We evaluated routine echocardiographic measurements and performed offline speckle-tracking strain analysis for the purpose of this study on transthoracic echocardiographic (TTE) images performed at 9 to 18 months post-TAVI. Results: The final study population comprised 294 patients (pacing group n = 147 and non-pacing group n = 147), with a mean age of 81 ± 7 years, 59% male; median follow-up was 354 days. There were more baseline conduction abnormalities in the pacing group compared to the non-pacing group (56.5% vs. 41.5%. p = 0.01). Eighty-eight patients (61.6%) in the pacing group required RV pacing due to atrioventricular (AV) conduction block post-TAVI. The mean RV pacing burden was 44% in the pacing group. Left ventricular ejection fraction (LVEF) was similar at follow-up in the pacing vs. non-pacing groups (57 ± 13.0%, 59 ± 11% p = 0.31); however, LV global longitudinal strain (−12.7 ± 3.5% vs. −18.8 ± 2.7%, p < 0.0001), LV apical strain (−12.9 ± 5.5% vs. 23.2 ± 9.2%, p < 0.0001), and mid-LV strain (−12.7 ± 4.6% vs. −18.7 ± 3.4%, p < 0.0001) were significantly worse in the pacing vs. non-pacing groups. Conclusions: Chronic RV pacing after the TAVI procedure is associated with subclinical LV systolic dysfunction within 1.5 years of follow-up.

Abstract Aims Lipoprotein(a) [Lp(a)] has demonstrated its association with atherosclerosis and myocardial infarction. However, its role in the development of in-stent restenosis (ISR) after percutaneous coronary intervention (PCI) is not clearly established. The aim of this study is to investigate the association between Lp(a) and ISR. Methods and results A retrospective study of adult patients who underwent successful PCI between January 2006 and December 2017 at the three Mayo Clinic sites and had a preprocedural Lp(a) measurement was conducted. Patients were divided into two groups according to the serum Lp(a) concentration [high Lp(a) ≥ 50 mg/dL and low Lp(a) &lt; 50 mg/dL]. Univariable and multivariable analyses were performed to compare risk of ISR between patients with high Lp(a) vs. those with low Lp(a). A total of 1209 patients were included, with mean age 65.9 ± 11.7 years and 71.8% were male. Median follow-up after baseline PCI was 8.8 [interquartile range (IQR) 7.4] years. Restenosis was observed in 162 (13.4%) patients. Median serum levels of Lp(a) were significantly higher in patients affected by ISR vs. non-affected cases: 27 (IQR 73.8) vs. 20 (IQR 57.5) mg/dL, P = 0.008. The rate of ISR was significantly higher among patients with high Lp(a) vs. patients with low Lp(a) values (17.0% vs. 11.6%, P = 0.010). High Lp(a) values were independently associated with ISR events (hazard ratio 1.67, 95% confidence interval 1.18–2.37, P = 0.004), and this association was more prominent after the first year following the PCI. Conclusion Lipoprotein(a) is an independent predictor for long-term ISR and should be considered in the evaluation of patients undergoing PCI.

Abstract Aims Doppler mean gradient (MG) can underestimate aortic stenosis (AS) severity in patients with atrial fibrillation (AF) compared with patients with sinus rhythm (SR), potentially delaying intervention in AF. This study compared outcomes in patients with AF and SR following transcatheter aortic valve replacement (TAVR) and investigated delay in TAVR based on computed tomography aortic valve calcium score (AVCS). Methods and results Patients who underwent TAVR from 2013 to 2017 for native valve severe AS were identified from an institutional database. Baseline characteristics and overall survival were compared between those with SR and AF. There were 820 patients (mean age 81 years; 41.6% females) included in this study. AF was present in 356 patients. Patients with AF were older (82.2 vs. 80.5, P = 0.003) and had a lower MG compared with patients with SR (42.0 vs. 44.9, P = 0.002) with similar indexed aortic valve area (0.4 vs. 0.4, P = 0.17). Median AVCS was higher in AF (males: AF 2850.0 vs. SR 2561.0, P = 0.044; females: AF 1942.0 vs. SR 1610.5, P = 0.025). Projected AVCS, assuming the same age of diagnosis, was similar between AF and SR. Median survival post-TAVR was worse in AF compared with SR (3.2 vs. 5.4 years, log rank P &lt; 0.001). AF, lower MG, higher right ventricular systolic pressure, dialysis, diabetes, and significant tricuspid regurgitation were associated with higher mortality (P &lt; 0.05 for all). Conclusion Older age and higher AVCS in patients with AF compared with those with SR suggest that AS was both underestimated and more advanced at TAVR referral.

Abstract Objective To develop a natural language processing (NLP) algorithm that can accurately extract headache frequency from free‐text clinical notes. Background Headache frequency, defined as the number of days with any headache in a month (or 4 weeks), remains a key parameter in the evaluation of treatment response to migraine preventive medications. However, due to the variations and inconsistencies in documentation by clinicians, significant challenges exist to accurately extract headache frequency from the electronic health record (EHR) by traditional NLP algorithms. Methods This was a retrospective cross‐sectional study with patients identified from two tertiary headache referral centers, Mayo Clinic Arizona and Mayo Clinic Rochester. All neurology consultation notes written by 15 specialized clinicians (11 headache specialists and 4 nurse practitioners) between 2012 and 2022 were extracted and 1915 notes were used for model fine‐tuning (90%) and testing (10%). We employed four different NLP frameworks: (1) ClinicalBERT (Bidirectional Encoder Representations from Transformers) regression model, (2) Generative Pre‐Trained Transformer‐2 (GPT‐2) Question Answering (QA) model zero‐shot, (3) GPT‐2 QA model few‐shot training fine‐tuned on clinical notes, and (4) GPT‐2 generative model few‐shot training fine‐tuned on clinical notes to generate the answer by considering the context of included text. Results The mean (standard deviation) headache frequency of our training and testing datasets were 13.4 (10.9) and 14.4 (11.2), respectively. The GPT‐2 generative model was the best‐performing model with an accuracy of 0.92 (0.91, 0.93, 95% confidence interval [CI]) and R 2 score of 0.89 (0.87, 0.90, 95% CI), and all GPT‐2–based models outperformed the ClinicalBERT model in terms of exact matching accuracy. Although the ClinicalBERT regression model had the lowest accuracy of 0.27 (0.26, 0.28), it demonstrated a high R 2 score of 0.88 (0.85, 0.89), suggesting the ClinicalBERT model can reasonably predict the headache frequency within a range of ≤ ± 3 days, and the R 2 score was higher than the GPT‐2 QA zero‐shot model or GPT‐2 QA model few‐shot training fine‐tuned model. Conclusion We developed a robust information extraction model based on a state‐of‐the‐art large language model, a GPT‐2 generative model that can extract headache frequency from EHR free‐text clinical notes with high accuracy and R 2 score. It overcame several challenges related to different ways clinicians document headache frequency that were not easily achieved by traditional NLP models. We also showed that GPT‐2–based frameworks outperformed ClinicalBERT in terms of accuracy in extracting headache frequency from clinical notes. To facilitate research in the field, we released the GPT‐2 generative model and inference code with open‐source license of community use in GitHub. Additional fine‐tuning of the algorithm might be required when applied to different health‐care systems for various clinical use cases.