Authors :
Presenting Author: Zachary Grinspan, MD, MS – Cornell Weill Medicine
Nilika Singhal, MD – UCSF
Erika Axeen, MD – University of Virginia
Sonam Bhalla, MD – Emory University/ Children's Healthcare of Atlanta
Jason Coryell, MD – Oregon Health Sciences University
Krista Eschbach, MD – Children Hospital Colorado
Shaun A. Hussain, MD, MS – Division of Pediatric Neurology, Department of Pediatrics, UCLA Mattel Children's Hospital, David Geffen School of Medicine
Amanda Sandoval Karamian, MD – University of Utah
Shavonne Massey, MD – Childrens Hospital of Philadelphia
Juma Mbwana, MD – Childrens National Medical Center
Lindsey Morgan, MD – Seattle Childrens Hospital
Anup Patel, MD – Nationwide Children's Hospital
M. Scott Perry, MD – Cook Children’s Physician Network
Erin Romanowski, DO – University of Michigan
Tristan Sands, MD PhD – Columbia Univeristy
Janet Soul, MDCM – Boston Childrens Hospital
Regina Triplett, MD MS – Washington Univeristy St. Louis
Joyce Wu, MD – Ann and Robert H Lurie Children's Hospital of Chicago
Elissa Yozawitz, MD – Montefiore Medicine
Natasha Basma, MPH – Weill Cornell Medicine
Hannah Glass, MD MS – UCSF
Renée Shellhaas, MD, MS – Washington University in St. Louis
Rationale:
There is equipoise about second line therapy for neonatal seizures. We aimed to compare the effectiveness of phenytoin/fosphenytoin (PHT) vs levetiracetam (LEV) in neonates with acute provoked seizures that continued despite initial treatment with phenobarbital, using causal inference techniques for observational data.Methods:
We reviewed charts of neonates with seizures at 16 institutions using the Pediatric Epilepsy Learning Healthcare System. Inclusion criteria were: term-born, seizures by age 7 days, initial treatment with phenobarbital (PB), post-PB breakthrough seizures within 72 hours, and treated with IV PHT or LEV. Exclusions included discharge before age 3 days and known or suspected monogenic epilepsy. The primary outcome was seizure cessation for 24 hours, confirmed by EEG. Additional outcomes included seizure cessation for 72 hours and treatment with a 3rd medication for seizures. We estimated comparative effectiveness using inverse probability of treatment weighting (IPTW) to calculate the average treatment effect. We used saturated propensity scores, incorporating multiple confounders and covariates (Table 1). We adjusted confidence intervals based on center-to-center variations using generalized estimating equations (GEE). We calculated the E-value to estimate the amount of unmeasured confounding required to explain away observed effects.Results:
There were 254 neonates (104 female) born from 2010 to 2023 who met inclusion criteria (Table 1). Of these, 123 received PHT and 131 LEV. Three clinical factors were associated with treatment selection. Neonates who received PHT were younger at the time of first seizure (PHT median 24 hours [IQR 12, 42] vs LEV 34 [14,66], p = 0.03), less likely to have “electrographic only” among their seizure types (PHT 52% vs LEV 70%, p = 0.004), and more likely to have received the second medication for frequent recurrent seizures or status epilepticus (PHT 79% vs LEV 67%, p = 0.02). The unadjusted response rates were 45.5% for PHT and 29% for LEV. The IPTW/GEE analysis indicated PHT was superior to LEV with an odds ratio of 2.15 (95% CI 1.18 to 3.9). The E-value was 2.6 (lower bound 1.6) (i.e., an unmeasured confounder would need to be associated with both the selection of medication and the outcome, each with a relative risk of 2.6, to explain away the superiority of phenytoin/fosphenytoin). The IPTW/GEE analysis also indicated superiority of PHT over LEV for 72 hours of seizure cessation (OR 1.8 [1.1 to 3.1]) but not for use of a third medication for seizures (OR 1.06 [0.68 to 1.6]).
Conclusions:
Phenytoin/fosphenytoin is superior to levetiracetam for second line therapy for neonatal seizures, for both 24 and 72 hours of seizure cessation, based on causal inference with observational data at multiple centers. Confidence in the finding is strengthened by a large effect size (OR > 2), measured confounding favored LEV (sicker neonates received PHT), and robustness to unmeasured confounding (E-value > 2).Funding: This study was funded by NINDS (R01 NS130113).