Practice variation in anti-epileptic drug use for neonatal hypoxic-ischemic encephalopathy among regional NICUs

Although seizures occur in 26–65% of neonates with hypoxic-ischemic encephalopathy (HIE), it is well known that anti-epileptic drug (AED) management is variable among centers [1‐4]. There are several possible reasons for this variability. Neonatal seizures are often subclinical, difficult to detect and cannot be predicted adequately by clinical variables alone [5, 6]. Furthermore, limitations in available resources to detect seizures, as well as a lack of consensus for seizure management among treating neonatologists and child neurologists lead to inconsistent recognition and treatment of neonatal seizures [7, 8]. Continuous electroencephalographic (cEEG) monitoring is therefore recommended in the management of neonates with encephalopathy [9]. However, cEEG is resource intensive and may not be available in all cooling centers. Even when available, factors such as time to application and interpretation may not be uniform across centers. Amplitude-integrated EEG (aEEG) is an alternative form of easily interpretable neuromonitoring that is routinely used in many but not all centers. Finally, the use of selective head cooling for treatment of HIE may temporarily preclude continuous EEG monitoring during therapeutic hypothermia (TH). Detection of subclinical seizures is important because treatment of subclinical seizures reduces seizure burden, and longer duration of seizures is associated with more severe brain injury on MRI and lower performance scores in all domains of the Bayley Scares of Infant Development-III [10, 11].

Variation also exists in the choice of AEDs. Phenobarbital is the first-line AED for treatment of neonatal seizures despite limited evidence to support its use over other agents, [12‐14] either for treatment or for seizure prophylaxis [15‐17]. Common second-line AEDs for persistent seizures include phenytoin (with similar effectiveness as phenobarbital) [14] and benzodiazepines. More recently, levetiracetam and topiramate are increasingly being used in NICUs as second-line AEDs [8, 18] and are under investigation for potential neuroprotective qualities [19]. Lidocaine has also been described as an AED [14, 20]. Unfortunately, the field has few randomized trials in neonates proving safety or efficacy of one AED over another. A clinical trial of bumetanide as a second-line AED for electrographic seizures not responsive to phenobarbital did not show efficacy but did show the serious side-effect of hearing impairment [21]. The recently completed clinical trial of levetiracetam as first-line therapy for neonatal seizures (NEOLEV2 NCT01720667) reportedly did not show greater efficacy of levetiracetam over phenobarbital (Child Neurology Society Annual Meeting, Chicago, IL, October 16, 2018). Consistent and rational use of these drugs is important as pre-clinical and clinical studies have raised concern regarding AED-associated neurotoxicity in the developing brain, with detrimental effects on neurogenesis, cell proliferation and migration, apoptosis, synaptogenesis and white matter integrity [22‐26].

Reducing intercenter variation (ICV) through standardization of care has been demonstrated to improve outcomes across NICU populations [27]. Importantly, several centers have shown that protocol-driven management of neonates at risk for seizures results in improvements in care including diagnosis of seizures [28], decreased phenobarbital levels, progression to status epilepticus, length of hospital stay [29] and discharge on AED [30]. (Improvement in outcomes due to protocolized approaches has been shown in management of other neonatal diseases as well, including congenital diaphragmatic hernia [31] and short bowel syndrome [32]). An important step to improving consistency of care is to understand determinants of variability in AED prescribing practices. Recent studies have reported exposure trends over time and ICV in AED use for neonatal seizures [7, 8, 33, 34]. A consistent message from these reports is the widespread ICV in AED practices, which is not surprising given that prior investigations have evaluated populations of mixed diagnoses and data from various NICUs with different levels of care. Even though neuromonitoring and neuroimaging technology and child neurology specialists are readily available, CHND NICUs do not share standardized treatment protocols. Therefore, we hypothesized that seizure treatment for HIE would vary among the quaternary care Children’s Hospitals in our large consortium. Our objective was to identify sources of ICV in AED utilization with the plan to identify opportunities for quality improvement (QI).

Using linked data from the Children’s Hospital Neonatal Database (CHND) and Pediatric Health Information Systems (PHIS) we quantified ICV in the use of AEDs (initiation, selection and duration) and AED cost as another proxy measurement of AED use for neonates with HIE.

The purpose of this project was to evaluate the variability that exists across regional NICUs caring for a large burden of neonates with HIE in the US, in order to inform a future QI intervention [41]. In a contemporary cohort of neonates with HIE treated with TH at 20 US regional NICUs, we observed significant ICV in AED utilization. We looked at utilization from a number of different perspectives including selection, any AED exposure, duration of exposure, discharge on AEDs and AED cost as another proxy for utilization. This study of our very large consortium highlighted unwarranted variation [42] in the management of AEDs in HIE, particularly in neonates without electrographic evidence of seizures. This work therefore supports a future QI collaborative across the CHND consortium targeting neonates with HIE who do not have electrographic evidence of seizures. It is important to note that CHND NICUs do not share standardized treatment protocols although all centers have similar levels of care and availability of specialty services. Although best practices have been designated by the state of California (https://​www.​dhcs.​ca.​gov/​services/​ccs/​Documents/​ccsnl061116.​pdf), only 2 California sites were involved in this study and more globally accepted guidelines are not available. Nonetheless, all CHND centers involved in this study met recommendations by the American Academy of Pediatrics Committee on Fetus and Newborn for centers that provide TH, including level III or higher NICU care, neurologic consultation, neuromonitoring with aEEG or cEEG, neuroimaging by MRI, systems for monitoring longitudinal neurodevelopmental outcome, training programs and infrastructure including written protocols and monitoring of outcomes as well as outreach to community hospitals [43].

Despite ACNS guidelines for EEGs in neonates [9], significant variability exists in the application of cEEG for seizure detection/monitoring. We did observe a dramatic increase in use of cEEG overall following publication of the guidelines mid-study in 2011. A decrease in use of aEEG use toward the end of the study period may have been related to discontinued sales of the selective head-cooling device in the US. Although 98% of all neonates in our study received some form of neuro-monitoring (cEEG or aEEG or both), it is possible that our seizure rates are underestimated in those who did not receive monitoring, those who received delayed monitoring or those that received aEEG alone, given the superior sensitivity of cEEG for seizure detection (particularly for seizures that are brief, infrequent or of low amplitude, or not central or parietal [44]). The incidence of seizures detected by EEG in our cohort was 28%, lower than for the CoolCap (61% detected by aEEG) [45], TOBY (54% detected by aEEG) [46] and NICHD hypothermia trials (46% clinical seizures) [47]. Details regarding exact timing of seizure detection and EEG acquisition in relation to AED administration were not available, although it is known that the majority of seizures in HIE occur in the first 24–48 h of life [48, 49]. Status epilepticus rates were lower than expected [2] and may be related to the application of TH to mild HIE cases in real practice. That some clinical seizures occurred in the absence of electrographic seizures might be explained by the following scenarios: clinical movements might not be due to epileptiform activity; seizures noted prior to initiation of cEEG might have spontaneously resolved or resolved following AED given; the threshold to treat clinical seizures during TH might be higher if patients are not on cEEG or aEEG for the entire period of TH and rewarming; even if they were, cEEG reading might not be immediately available. We observed relatively low rates of clinical seizures but a rate of EEG seizures of nearly 10% in cases of mild encephalopathy who were cooled. For these cases, we speculate that clinical or EEG seizures might have been noted after initial assignment of severity category without reassignment to the moderate category after seizures were noted. Our data reinforce that cEEG or aEEG should be obtained in all mild cases of encephalopathy as EEG seizures would indicate that the eligibility for TH had been met.

Consistent with AED selection in other studies [8, 33, 34], we observed a similar predominance of phenobarbital use and a higher frequency of levetiracetam compared to phenytoin/fosphenytoin use. We examined levetiracetam use by year and found an increase in levetiracetam over fosphenytoin/phenytoin in the final year of the study. The apparent inverse relationship of levetiracetam and fosphenytoin/phenytoin use suggests that preferential use of these second-line medications varies by center practice; alternative explanations include fosphenytoin shortages as well as the development of an intravenous formulation of levetiracetam. AED costs per patient were highest for levetiracetam, 2.9-fold greater than fosphenytoin/phenytoin, and cost considerations may drive AED choice for some providers. On the other hand, levetiracetam may be preferred by some providers because of its association with decreased respiratory depression.

Although previous studies have shown ICV in AED utilization, given that the NICUs in our consortium are all Level IV, we were nonetheless somewhat surprised to find the magnitude of ICV that we observed. One study that included some of the same referral centers, observed similar ICV in continuation of AEDs at discharge for neonatal seizures of all etiologies. After univariable analysis adjusting for electrographically confirmed seizures, status epilepticus, seizures refractory to the initial loading dose of AED and abnormal neurological exam at discharge, only study site and seizure etiology remained significantly associated with discharge on AEDs. With regards to seizures specifically associated with HIE, this study’s overall rate of discharge on AEDs was 57%, similar to the 56% that we observed in cases of HIE with electrographically confirmed seizures. Treatment duration differences were implied in this study but not directly reported [8].

Frequency of AED at discharge was center-dependent in our study as well, suggesting that physician/center practice drives the decision to continue AEDs. In our study, over half of neonates with electrographic seizures and 7% of neonates without electrographic seizures were discharged on AEDs. Stated otherwise, if a neonate ever received an AED, that neonate had a 1 in 3 chance of being discharged on an AED. This variation is important because, although neonates with HIE, and particularly those with seizures, are at increased risk for later epilepsy [50, 51], emerging evidence suggests that discharge on an AED might not be indicated in all neonates with acute seizures after HIE [52]. It is well recognized that prolonged use of most AEDs is associated with neuronal apoptosis and neurodevelopmental delays [26, 53]. This added risk is even less acceptable for neonates who have never demonstrated seizures by EEG. Unlike previous studies, we showed ICV in other measures of AED utilization, including any exposure and duration of exposure and cost.

We were surprised to find that a high proportion of neonates without seizures confirmed by EEG received AEDs, many through discharge. This may partly reflect AED use for clinical seizures not confirmed electrographically, and may occur more frequently when EEG is not immediately obtainable as not all centers have 24/7 EEG technician and neurophysiologist capabilities. High rates of AED use in neonates without electrographic seizures, as high as 60% at one center, might also reflect attempts at neuroprotection or seizure prophylaxis by some sites. A recent Cochrane Database meta-analysis did not support the use of prophylactic barbiturates for perinatal asphyxia because, although this practice seemed to reduce seizures, it did not reduce mortality or neurodevelopmental impairment [17]. Our data suggests a need to identify sites that use AEDs for neuroprotection or seizure prophylaxis and to stop this practice.

That a small proportion of neonates with electrographic seizures did not receive AEDs during their hospitalization is also surprising. As our data reflects only medications received at CHND hospitals, it is possible that these neonates received AEDs at the referral hospital that were not continued upon admission to the CHND NICU. It is also possible that limited real-time availability of neurophysiologists across centers may be associated with delayed EEG interpretation and reporting, so that some seizures clinically resolved by the time of recognition on EEG would not lead to AED initiation. Finally, although benzodiazepines are often used to treat intractable seizures or status epilepticus, we did not report the use of benzodiazepines that may have been used to treat seizures; given the nature of the registry, we were unable to confirm whether benzodiazepines were given for seizures or for sedation. The use of AEDs without EEG evidence of seizures offers an opportunity for intervention and change in practice(s).

The major strength of our study was the linkage of clinical data with PHIS data which enabled us to evaluate utilization and cost of AEDs over the course of hospitalization in neonates with HIE. Although a previous study used PHIS data to evaluate AED use, its subjects had neonatal seizures due to various etiologies and were hospitalized during an epoch when TH was not yet standard of care and costs were not evaluated [4]. As TH has led to centralization of care of neonates with HIE to regional NICUs, describing practice variation in this setting is important. Indeed, not all centers that provide TH provide related services such as cEEG or aEEG [54]. We capitalized on detailed clinical information from CHND not available from PHIS alone that allowed us to observe that AED use was significantly affected by gestational age, HIE severity, EEG seizures and mortality, in contrast to the previous study [4]. After controlling for these clinical covariates, ICV in AED use for neonates with HIE persisted.

Another major strength of our investigation was that we only studied neonates with HIE, the most common etiology of neonatal seizures in the current era, who were cared for at regional NICUs. By contrast, prior studies have compared dissimilar groups such as preterm infants or infants with central nervous system disease [4, 33]. Similarly, prior survey and registry-based studies have evaluated data from various NICUs where availability of neurodiagnostic studies (MRI, EEG, etc.) and child neurology specialists may contribute to variations in care [7, 8]. Our study included only regional NICUs meeting criteria for participation in the CNHD [27], and thus highlights the true conundrum of unexplained practice variation with regards to AED use in HIE.

Our study has some limitations. Referral biases exist because some neonates may have died prior to referral to the CHND NICU. Coding differences in AED use may exist between centers despite electronic acquisition of data but processes are in place to insure quality [27]. Unfortunately, we were also unable to link EEG findings temporally to AED initiation and discontinuation. Likewise, details regarding timing of seizure detection and EEG performance in relation to discharge were not available, although it is known that the majority of seizures in HIE occur within the first 24–48 h of life [48, 49]. Developmental outcomes and detailed seizure information is not presently available in CHND. Additionally, given that this study only involved care in regional referral sites, our findings may not be generalizable to community hospitals.

Interestingly, we observed a significantly higher unadjusted rate of seizures in neonates who were selectively head-cooled in contrast with those who received whole body cooling (Table 1). We speculate that delay in obtaining cEEG may result in delay in treatment and a higher rate of seizures at first cEEG. This observation warrants further study given the relatively small number of infants who received selective head cooling, multiple comparisons and unadjusted rates.

Significant variation exists in AED utilization in neonates with HIE treated with TH within our Children’s Hospital regional NICUs. This data indicates a multicenter QI project within the CHND is in order, with the goal of increasing timely neuromonitoring and eliminating exposure to AEDs without proof of EEG seizures. We believe the rate of exposure to AEDs without EEG seizures should approach 0%. Specific practices to be targeted in our initial QI project will include: 1) observation or use of lorazepam for clinical seizure without EEG confirmation, 2) cEEG or aEEG on admission for all neonates transported for TH (metrics will also include time from admission to placement of cEEG or aEEG), 3) cEEG or aEEG confirmation of seizures prior to phenobarbital, and 4) time from cEEG or aEEG confirmation of seizures to infusion of phenobarbital. Such an effort will improve adherence to evidence-based practices within member hospitals of the CHND.