Neonatal Seizures—Perspective in Low-and Middle-Income Countries

Neonatal seizures are the most common neurological emergency in the neonatal period and often pose diagnostic and management challenges for clinicians across the world [1]. Recently, a number of published studies have been aimed at improving the diagnosis, treatment, and outcome of neonatal seizures. However, the majority of these studies originate from high-income countries (HIC). There is evidence that the incidence and etiologies of neonatal seizures differ in HIC and low- and middle-income countries (LMIC),which may have far-reaching implications for their management [2]. Furthermore, most studies from LMIC solely rely on clinical diagnosis for seizure identification due to the limited availability of electroencephalography (EEG) and amplitude-integrated EEG (aEEG) [2‐5]. In this review, we address the diverse manifestations of neonatal seizures, their incidence, etiology, the diagnostic role of aEEG/EEG, evaluation, and management within low resource settings. As high-quality evidence from LMIC is lacking, much of the authors' recommendations are based on the evidence from HIC.

The International League Against Epilepsy (ILAE) defines seizures as a transient occurrence of signs and/or symptoms due to abnormally excessive or synchronous neuronal activity in the brain. This definition excludes the electrographic-only seizures, which constitute a 40%–60% of all seizures occurring in critically ill neonates [6‐8]. The American Clinical Neurophysiology Society classifies seizures into clinical-only, electroclinical, or electrographic-only seizures. Electrographic seizures are defined as a paroxysmal abnormal, sustained change in the EEG characterized by a repetitive and an evolving pattern with a minimum 2 µV voltage (peak to peak) and a duration of at least 10 s [9]. "Evolving" means here an unequivocal, gradual change in frequency, amplitude, morphology, and location. This definition is now generally accepted by the experts and has major implications for the diagnosis of seizures, as it requires the availability of aEEG or EEG [2, 10].

The incidence of neonatal seizures is estimated to be 1–3 per 1000 live births in HIC and widely varies from 36–90 per 1000 live births in LMIC (Supplementary Table S1). Reports from LMIC are not only scarce but methodology is variable and seizure detection are mostly reliant on clinical diagnosis [2]. In an extensive literature review for the case definite of neonatal seizures by the Brighton Collaboration [2], only four studies on incidence of neonatal seizures from LMIC were described [3, 4, 11, 12], all of which relied solely on clinical diagnosis (Supplementary Table S1).

Most of the neonatal seizures are acute provoked seizures indicating that they are secondary to an acute brain injury or systemic insult. Only 10%–15% of seizures in the neonatal period are the first manifestation of an epilepsy syndrome (unprovoked seizures), typically due to an underlying structural or genetic etiology [13]. Hypoxic–ischemic encephalopathy (HIE) remains the most common cause of neonatal seizures. Other causes include perinatal stroke, intracranial hemorrhage, metabolic, and electrolyte disturbances, systemic and central nervous system infections, and inborn errors of metabolism or genetic epilepsy syndromes [14]. A history and physical examination complemented with video EEG/aEEG can aid early diagnosis of the underlying etiology, which is important for management and prognostication.

The etiological spectrum of neonatal seizures is different in HIC and LMIC (Supplementary Table S2). While HIE is the major cause in both HIC and LMIC (although more frequent in LMIC), infections have been reported as the second most common cause of seizures in LMIC [5, 15]. Although intracranial hemorrhage and perinatal stroke have been reported to be less common than infections in the LMIC, it is likely that there is under-diagnosis due to a lack of neuroimaging. Overall, the infection rate is similar in babies with encephalopathy across HIC and LMIC (Supplementary Table S2). Seizures resulting from metabolic abnormalities such as hypoglycemia and kernicterus used to be more common in HIC but are now more common in the LMIC compared to HIC, reflecting the better overall supportive care in the HICs [2]. It is possible that there is relevant comorbidity in LMIC, for example, a neonate with HIE may also have an infection or hypoglycemia, which may impact outcome [16]. However, no study has specifically evaluated the effect of comorbidity on outcome.

Neonatal seizures are different to seizures in older children in their presentation, with a substantial proportion being electrographic-only, lacking a clinical correlate. Mizrahi and Kellaway studied seizure patterns in 71 neonates using video EEG, 11 of which never showed a clinical correlate [6]. Focal clonic seizures, certain myoclonic seizures, and focal tonic seizures were most reliably associated with an electrographic correlate. Most so-called “subtle seizures”[1] were poorly correlated with EEG seizure activity [6] and are now considered not to be seizures [6, 17]. In another study evaluating neonates with encephalopathy undergoing therapeutic hypothermia with video EEG monitoring, 43% failed to show any clinical event among the 14 infants who developed electrographic seizures [18]. A recent prospective study involving 426 neonates with suspected clinical seizures found that 62% of the subjects had at least 1 electrographic-only seizure and 16% had electrographic seizures without any clinical manifestations [8]. Types of clinical seizures are summarized in Supplementary Table S3 [17].

Recently, the ILAE published a new classification of neonatal seizures (Fig. 1) which was adapted from the new classification of seizures for adults and older children [19]. It emphasizes the role of EEG or aEEG for the confirmation of suspected clinical events and is based on an electroclinical relationship of seizures (electroclinical, electrical-only). It then classifies the seizure according to semiology (motor: automatisms, focal clonic, focal tonic, myoclonic, epileptic spasms, and nonmotor: autonomic and behavioral arrest) each with additional modifiers [17]. Those events which fail to demonstrate a seizure activity in EEG are not considered to be seizures [17, 20, 21].

EEG is considered the gold standard for the diagnosis of seizures in newborns. However, even in HIC, EEG is not available to many neonatal units and rarely 24/7 due to its time-intensity, expensive equipment and the requirement for expert interpretation [22].

The availability of EEG or even aEEG is limited in most LMIC and consequently there is also little expertise in the interpretation of neonatal recordings. Clinical diagnosis is unreliable due to the risk of both under and overdiagnosis of seizures. Underdiagnosis is due missing discreet seizure manifestations and because over 50% of seizures are electrographic-only while overdiagnosis is due to misdiagnosing abnormal nonepileptic movements as seizures [20]. In another study where 20 video clippings of suspected clinical seizures were evaluated by 91 doctors and 46 other professionals, the average number of events correctly identified was found to be just 10 out 20. The study also found that the evaluators were more likely to identify a clonic seizure correctly as opposed to a subtle seizure [21], thus illustrating how EEG is a necessary tool for seizure detection and clinical diagnosis alone cannot be relied on.

When and if EEG is not available, one- or two-channel amplitude integrated EEG (aEEG) can be helpful for long term monitoring in NICU. A survey from Brazil found half of all neonatal units providing performing therapeutic hypothermia for HIE use to some kind of neuromonitoring, but very few (< 5%) have access to continuous video EEG suggesting that aEEG is the most common form of neuromonitoring [23]. The same group presented evidence how useful aEEG can be in a LMIC setting [24]. aEEG provides an assessment of background activity (which provides information on the degree of brain injury for example in HIE) and can identify seizures. However, the sensitivity of aEEG is lower than full EEG as short seizures (< 30 s) or low amplitude seizures are often missed. There is also a risk of erroneous interpretation of artefacts as seizures and hence the recording must be appropriately annotated. Many new aEEG machines have the option of adding 1–2 raw channels for interpretation which increase the seizure detection rate and identification of artefact. A recent study reported a median sensitivity for aEEG with raw channels of 78% and a median specificity of 78% to detect individual seizures, but without raw traces, the median sensitivity dropped to 54% [25]. Hence, it is recommended that a full-montage video EEG be done whenever possible, or if EEG is not available to use aEEG with raw channels. The American Clinical Neurophysiology Society recommends EEG monitoring for 24 h in all neonates who are at high risk for seizures, such as neonates with acute brain injury, clinical encephalopathy, or abnormal paroxysmal events [26].

In cases where EEG is not available, it is recommended to use the levels of diagnostic certainties as given by the GAIA [2] and subsequently adapted by the ILAE [17]. The five levels of diagnostic certainties are (Supplementary Fig. S1):

As with any other neurologic emergency, management of airway, breathing and circulation is of utmost importance. Diagnosis and treatment of underlying etiology is crucial for effective control of neonatal seizures, especially secondary to metabolic derangements. In all suspected seizures, a bedside blood glucose and electrolyte measurement should be done first. Any hypoglycemia must be promptly corrected. Brief seizures secondary to transient metabolic derangements (hypocalcemia, hypoglycemia, hypomagnesemia, or hyponatremia) may not warrant anticonvulsant medication, if seizures cease upon correction. Blood, urine and cerebrospinal fluid analysis with respective cultures should be sent followed by empiric antibiotic therapy, or antiviral therapy, if sepsis is suspected. Any suspicion of a CNS infection must prompt institution of empiric antibiotics or antivirals after taking blood cultures. Supplementary Fig. S2 in the online supporting information illustrates an approach to the evaluation of neonatal seizures.

Neonatal mortality associated with seizures has improved drastically over the last three decades, with a reduction in mortality to approximately 7% [40], which can be explained by the improvement of neonatal outcomes in general. In a recent study from a rural district in Kenya involving 142 children with neonatal seizures, 32 died of various causes which amounts to a mortality of 22.5%, indicating that mortality is much higher in LMIC. Regardless of the cause of neonatal seizures, the long-term outcome remains poor. Seizures on their own increase brain injury [41] and a seizure burden of more than 13 min per hour increases the odds of an abnormal neurological outcome by eightfold (OR: 8.00; 95% confidence Interval: 2.06–31.07) [42, 43]. Mortality and morbidity remain high among the preterm neonates with seizure. In a follow-up study from Istanbul, Turkey, of 112 infants with neonatal seizures, 28% of infants later developed cerebral palsy, 36% epilepsy, and almost 50% developmental delay [44]. There is evidence that in neonates with HIE a reduction of electrographic seizure burden guided by aEEG improves cognitive outcome [45, 46]. Early treatment of electrographic seizure burden has been shown to be associated with fewer cases progressing to status epilepticus and a shorter hospital stay [28, 47]. These data indicate that there is a need for improved diagnosis of both electroclinical and electrographic seizure burden to enable earlier treatment.

In a low-resource setting, telemedicine may play an interesting role in providing remote specialized assistance. Centralized systems can reach a large number of centers in real-time, with educational activities, consultation, and monitoring to leverage the quality of care. After proving the value of aEEG monitoring in a LMIC [24], such a system has successfully been implemented by Protecting Brains and Saving Futures (PBSF) in over 30 hospitals across all regions of Brazil [48]. Babies are monitored with EEG or aEEG and assisted by a team in a remote monitoring center, with encrypted data of EEG transmitted to a secure cloud-based server. This approach is suitable for clinical management or research. However, cost-effectiveness as well as legal and regulatory issues remain important challenges to this approach in LMICs.

In summary, neonatal seizures pose a number of diagnostic and therapeutic challenge in LMIC:

It is well recognized that some of these concerns are also applicable to neonatal units in high-resourse settings. All of these need to be addressed urgently to improve diagnosis, treatment, and consequently, the outcome of seizures during the neonatal period.