Epilepsy as a Dynamic Disease: Statistical Modeling of Transitions Between Seizure Risk States
Abstract number :
2.168
Submission category :
4. Clinical Epilepsy / 4B. Clinical Diagnosis
Year :
2017
Submission ID :
349001
Source :
www.aesnet.org
Presentation date :
12/3/2017 3:07:12 PM
Published date :
Nov 20, 2017, 11:02 AM
Authors :
Sharon Chiang, Baylor College of Medicine; Marina Vannucci, Rice University; Daniel M. Goldenholz, National Institutes of Health; Robert Moss, SeizureTracker.com; and John M. Stern, UCLA
Rationale: A fundamental challenge in treating epilepsy is that changes in seizure count do not necessarily reflect changes in the underlying risk of another seizure. Rather, seizures often occur as probabilistic variation around an underlying seizure risk state due to normal fluctuations from natural history, leading to unnecessary medication adjustments in managing patients with epilepsy. However, no rigorous statistical approach exists to systematically estimate the underlying seizure risk state from observed seizure events and characteristics. Methods: Using data from SeizureTracker.com, a patient-reported seizure diary tool containing over 1.2 million recorded seizures across eight years, a novel Bayesian mixed effects hidden Markov model for zero-inflated count data (BME-HMM-ZIP) was developed to estimate underlying seizure risk states on the individual patient level. Accuracy of underlying seizure risk assessment and projected impact on medication management was evaluated through a simulation study. Clinical utility was demonstrated with Tuberous Sclerosis Complex (TSC) seizure diary data from SeizureTracker.com. Results: The proposed statistical model, BME-HMM-ZIP, achieved superior identification of underlying seizure risk compared to quantitative simulations of current clinical practice. Use of BME-HMM-ZIP to estimate seizure risk led to a projected impact of 5.7-15.5 times lower rates of unnecessary medication adjustments. Applied to seizure diary data, four underlying seizure risk states were identified among patients with TSC. The expected duration of each seizure risk state was less than 12 months for all identified states in TSC patients, providing data-driven support for the current ILAE definition of seizure freedom. Conclusions: This study proposes a rigorous statistical approach for systematic identification of underlying seizure risk based on observed seizure counts. The proposed approach resulted in substantial improvements in ability to differentiate true improvement/worsening in seizure risk from those caused by natural variability. Incorporation of systematic statistical approaches into clinical practice may have the potential to improve the timing of unnecessary interventions in the treatment of patients with epilepsy. Funding: Funding/support for this research was provided by (1) the National Library of Medicine Training Fellowship in Biomedical Informatics, Gulf Coast Consortia for Quantitative Biomedical Sciences (Grant #2T15-LM007093-21) (SC); (2) the National Institute of Health (Grant #5T32-CA096520-07) (SC); the National Institutes of Neurological Disorders and Stroke, Intramural Research Division (DMG); (3) NIH-NINDS K23 Grant NS044936 (JMS); (4) The Leff Family Foundation (JMS). Use of the data was facilitated by the International Seizure Diary Consortium (https://sites.google.com/site/isdchome/).
Clinical Epilepsy