Inductive reasoning with large language models: a simulated randomized controlled trial for epilepsy DOI Creative Commons
Daniel M. Goldenholz,

Shira R. Goldenholz,

Sara Habib

et al.

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: March 19, 2024

Abstract Importance The analysis of electronic medical records at scale to learn from clinical experience is currently very challenging. integration artificial intelligence (AI), specifically foundational large language models (LLMs), into an pipeline may overcome some the current limitations modest input sizes, inaccuracies, biases, and incomplete knowledge bases. Objective To explore effectiveness using LLM for generating realistic data other LLMs summarizing synthesizing information in a model system, simulating randomized trial (RCT) epilepsy demonstrate potential inductive reasoning via chart review. Design An LLM-generated simulated RCT based on treatment with anti-seizure medication, cenobamate, including placebo arm full-strength drug arm, evaluated by LLM-based versus human reader. Setting Simulation seizure diaries, effects, reported symptoms notes generated multiple different neurologist writing styles. Participants Simulated cohort 240 patients, divided 1:1 arms. Intervention Utilization generation synthesis these notes, aiming evaluate efficacy safety cenobamate control either evaluator or AI-pipeline. Measures AI focused identifying number seizures, symptom reports, efficacy, statistical comparing 50%-responder rate median percentage change between arms, as well side effect rates each arm. Results closely mirrored analysis, demonstrating drug’s marginal differences (<3%) both symptoms. Conclusions Relevance This study showcases accurately simulate analyze trials. Significantly, it highlights ability reconstruct essential elements, identify recognize symptoms, within framework. findings underscore relevance future research, offering scalable, efficient alternative traditional mining methods without need specialized training. Key Points Question Can (LLMs) effectively trial, relevant symptoms? Findings In generate treatment, AI-driven analyses were found match expert evaluations. process demonstrated capture effects minimal outcomes analyses. Meaning use analyzing trials offers promising approach developing systems records. could revolutionize way are conducted analyzed, enabling rapid, accurate assessments therapeutic

Language: Английский

The present and future of seizure detection, prediction, and forecasting with machine learning, including the future impact on clinical trials DOI Creative Commons
Wesley T. Kerr, Katherine N. McFarlane,

Gabriela Figueiredo Pucci

et al.

Frontiers in Neurology, Journal Year: 2024, Volume and Issue: 15

Published: July 11, 2024

Seizures have a profound impact on quality of life and mortality, in part because they can be challenging both to detect forecast. Seizure detection relies upon accurately differentiating transient neurological symptoms caused by abnormal epileptiform activity from similar with different causes. forecasting aims identify when person has high or low likelihood seizure, which is related seizure prediction. Machine learning artificial intelligence are data-driven techniques integrated neurodiagnostic monitoring technologies that attempt accomplish those tasks. In this narrative review, we describe the existing software hardware approaches for forecasting, as well concepts how evaluate performance new future application clinical practice. These include long-term without electroencephalography (EEG) report very sensitivity reduced false positive detections. addition, implications evaluation novel treatments seizures within trials. Based these data, machine could fundamentally change care people seizures, but there multiple validation steps necessary rigorously demonstrate their benefits costs, relative current standard.

Language: Английский

Citations

8
Inductive reasoning with large language models: a simulated randomized controlled trial for epilepsy DOI
Daniel M. Goldenholz,

Shira R. Goldenholz,

Sara Habib

et al.

Epilepsy Research, Journal Year: 2025, Volume and Issue: 211, P. 107532 - 107532

Published: Feb. 24, 2025

Language: Английский

Citations

0
Challenges and directions in epilepsy diagnostics and therapeutics: Proceedings of the 17th Epilepsy Therapies and Diagnostics Development conference DOI Creative Commons
Samuel W. Terman, Laura Kirkpatrick, Wesley T. Kerr

et al.

Epilepsia, Journal Year: 2023, Volume and Issue: 65(4), P. 846 - 860

Published: Dec. 23, 2023

Abstract Substantial efforts are underway toward optimizing the diagnosis, monitoring, and treatment of seizures epilepsy. We describe preclinical programs in place for screening investigational therapeutic candidates animal models, with particular attention to identifying eliminating drugs that might paradoxically aggravate seizure burden. After development, we discuss challenges solutions design regulatory logistics clinical trial execution, develop disease biomarkers interventions may be not only seizure‐suppressing, but also disease‐modifying. As disease‐modifying treatments designed, there is clear recognition that, although represent one critical target, targeting nonseizure outcomes like cognitive development or functional requires changes traditional designs. This reflects our increasing understanding epilepsy a profound impact on quality life patient caregivers due both themselves other factors. review examines selected key future directions diagnostics therapeutics, from drug discovery translational application.

Language: Английский

Citations

8
Placebo response in patients with Dravet syndrome: Post-hoc analysis of two clinical trials DOI Creative Commons
Orrin Devinsky,

Kerry Hyland,

Rachael Loftus

et al.

Epilepsy & Behavior, Journal Year: 2024, Volume and Issue: 156, P. 109805 - 109805

Published: April 26, 2024

Dravet syndrome is a rare, early childhood-onset epileptic and developmental encephalopathy. Responses to placebo in clinical trials for epilepsy therapies range widely, but factors influencing response remain poorly understood. This study explored its effects on safety, efficacy, quality of life outcomes patients with syndrome.

Language: Английский

Citations

2
A Comprehensive Overview of the Current Status and Advancements in Various Treatment Strategies against Epilepsy DOI Creative Commons
Abdul Waris, Muhammad Siraj, Ayyaz Ali Khan

et al.

ACS Pharmacology & Translational Science, Journal Year: 2024, Volume and Issue: 7(12), P. 3729 - 3757

Published: Nov. 1, 2024

Epilepsy affects more than 70 million individuals of all ages worldwide and remains one the most severe chronic noncommunicable neurological diseases globally. Several neurotransmitters, membrane protein channels, receptors, enzymes, and, recently noted, various pathways, such as inflammatory mTORC complexes, play significant roles in initiation propagation seizures. Over past two decades, developments have been made diagnosis treatment epilepsy. Various pharmacological drugs with diverse mechanisms action other options developed to control seizures treat These include surgical treatment, nanomedicine, gene therapy, natural products, nervous stimulation, a ketogenic diet, gut microbiota, etc., which are developmental stages. Despite plethora options, one-third affected resistant current medications, while majority approved side effects, changes can occur, pharmacoresistance, effects on cognition, long-term problems, drug interactions, risks poor adherence, specific for certain psychological complications. Therefore, development new that no or minimal adverse is needed combat this deadly disease. In Review, we comprehensively summarize explain stages epilepsy well their status clinical trials advancements.

Language: Английский

Citations

2
Factors associated with placebo response rate in randomized controlled trials of antiseizure medications for focal epilepsy DOI Creative Commons
Wesley T. Kerr, Maria Suprun,

Neo Kok

et al.

Epilepsia, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 21, 2024

Abstract Objective Randomized controlled trials (RCTs) are necessary to evaluate the efficacy of novel treatments for epilepsy. However, there have been concerning increases in placebo responder rate over time. To understand these trends, we evaluated features associated with increased rate. Methods Using individual‐level data from 20 focal‐onset seizure provided by seven pharmaceutical companies, associations change frequency participants randomized placebo. We used multivariable logistic regression participant and study factors differing rates 50% reduction during blinded treatment, as compared pre‐randomization baseline frequency. In addition, focused on association country recruitment. Results pooled analysis 1674 placebo, a higher (50RR) was shorter duration epilepsy ( p = .006), lower .002), fewer concomitant antiseizure medications .004), absence adverse events < .001), more trial arms geographic region .001). Mixture modeling indicated significantly 50RR Bulgaria, Croatia, India, Canada (42% group vs 22% comprising all 40 other countries, 10 −15 ). six or seizures per 28 days (29% 21%, .00018). Significance These results can assist future RCTs estimating expected rate, which may lead reliable power estimates. Higher markers less‐refractory There were significant differences well an elevated close minimum eligibility criteria.

Language: Английский

Citations

1
Inductive reasoning with large language models: a simulated randomized controlled trial for epilepsy DOI Creative Commons
Daniel M. Goldenholz,

Shira R. Goldenholz,

Sara Habib

et al.

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: March 19, 2024

Abstract Importance The analysis of electronic medical records at scale to learn from clinical experience is currently very challenging. integration artificial intelligence (AI), specifically foundational large language models (LLMs), into an pipeline may overcome some the current limitations modest input sizes, inaccuracies, biases, and incomplete knowledge bases. Objective To explore effectiveness using LLM for generating realistic data other LLMs summarizing synthesizing information in a model system, simulating randomized trial (RCT) epilepsy demonstrate potential inductive reasoning via chart review. Design An LLM-generated simulated RCT based on treatment with anti-seizure medication, cenobamate, including placebo arm full-strength drug arm, evaluated by LLM-based versus human reader. Setting Simulation seizure diaries, effects, reported symptoms notes generated multiple different neurologist writing styles. Participants Simulated cohort 240 patients, divided 1:1 arms. Intervention Utilization generation synthesis these notes, aiming evaluate efficacy safety cenobamate control either evaluator or AI-pipeline. Measures AI focused identifying number seizures, symptom reports, efficacy, statistical comparing 50%-responder rate median percentage change between arms, as well side effect rates each arm. Results closely mirrored analysis, demonstrating drug’s marginal differences (<3%) both symptoms. Conclusions Relevance This study showcases accurately simulate analyze trials. Significantly, it highlights ability reconstruct essential elements, identify recognize symptoms, within framework. findings underscore relevance future research, offering scalable, efficient alternative traditional mining methods without need specialized training. Key Points Question Can (LLMs) effectively trial, relevant symptoms? Findings In generate treatment, AI-driven analyses were found match expert evaluations. process demonstrated capture effects minimal outcomes analyses. Meaning use analyzing trials offers promising approach developing systems records. could revolutionize way are conducted analyzed, enabling rapid, accurate assessments therapeutic

Language: Английский

Citations

1