Authors :
Presenting Author: Rup Sainju, MBBS – University of Iowa
Deidre Dragon, BSS – University of Iowa
Harold Winnike, RRT – University of Iowa
Justin Kuhn, BS RRT, RPFT, RRT-ACCS/NPS, AE-C – University of Iowa
Jennie Michelson, Research specialist – University of Iowa
Eduardo Bravo, PhD – University of Iowa
Linder Wendt, MS – University of Iowa
Patrick Ten Eyck, PhD – University of Iowa
Brian Gehlbach, MD – University of Iowa
George Richeron, MD, PhD – University of Iowa
Rationale:
Central respiratory CO2 chemoreception (CCR), mediated by medullary serotonergic neurons, contributes to CO2 homeostasis by stimulating breathing in response to elevated arterial CO2. CCR can be quantified by measuring the hypercapnic ventilatory response (HCVR). Impaired postictal breathing is frequently obsevered preceeding documented SUDEP cases. Some seizures can inhibit serotonergic neurons in the brainstem, potentially attenuating HCVR and delaying recovery from postictal respiratory depression—conditions that may increase SUDEP risk. Therefore, we hypothesize that postictal attenuation of HCVR is a biomarker for SUDEP. The objective of this study was to examine the determinats of postictal HCVR in people with epilepsy (PWE) admitted to an epilepsy monitoring unit.
Methods:
PWE 18 years or older were monitored with continuous video-EEG, EKG, respiratory impedance plethysmography, nasal airflow, pulse oximetry, and transcutaneous CO2. HCVR was measured using a hyperoxic rebreathing protocol (6% CO2 and 50% O2), during the interictal state and postictally following focal seizures, once subjects were able to cooperate. The HCVR slope— defined as the slope of regression line between changes in minute ventilation and end-tidal CO2— was calculated for each test. The primary outcome was the HCVR slope after focal seizures. Demographics, clinical and seizure related variables were considered as independent predictors. Periictal respiratory changes were excluded from the analysis due to substantial missingness.
Generalized linear models with appropriate link functions were used to construct univariate and multivariate models. Multivariate models were bulit using stepwise regression based on the Akaike Information Criterion, and the number of variables was limited to <