A multicentre, randomised, controlled, clinical Investigation of a standalone decision support Algorithm for Neonatal Seizure Recognition

Seizures are the most common neurological emergency encountered in the neonatal
intensive care unit (NICU) and are the hallmark of neonatal encephalopathy.
Despite this they remain difficult to diagnose clinically.
Electroencephalography (EEG) monitoring is the gold standard for seizure
detection but real-time interpretation is challenging for most in the acute
setting.
Neonates can have peculiar movement patterns that are not associated with any
EEG *signature* of seizure. Alternatively, continuous EEG monitoring can show
status epilepticus in neonates who have little, if any, clinical
manifestations. The occurrence of *electro- clinical dissociation* - what you
see is not what you get- is a particular problem in neonates. In this group,
80-90% of electrographic seizures are *EEG-only* seizures. We have previously
shown that only 9% of the time-summed total seizure burden in a group of 51
neonates was accompanied by clinical manifestations.
The accurate diagnosis of seizures matters a great deal. Research suggests that
*seizures beget seizures* and those seizures contribute to neuronal damage in
neonates with hypoxic ischaemic encephalopathy. In clinical practice, many
neonates are treated on the basis of clinical diagnosis alone and this may
result in both under treatment and over treatment (because small normal
movements are misinterpreted as seizures). Over treatment carries the risk of
extending the duration of ventilation and intensive care stay, and increasing
the neurotoxic effects of antiepileptic drugs.
All experts agree that there is an urgent need for an accurate and robust
method of seizure detection, which can provide information in a readily
digestible form, in real time at the cotside. Translating automated seizure
detection from the bench to the cotside is challenging, and has not so far been
achieved.

The primary objective of this investigation is to assess the performance of the
ANSeR Software System by quantifying and comparing the diagnostic accuracy of
investigation personnel using the ANSeR Software System with the diagnostic
accuracy of investigation personnel not using the ANSeR Software System for the
diagnosis of neonatal seizures. The Expert Committee will be used as the
diagnostic reference standard. The secondary objectives are to quantify and
compare seizure burden and AED use in the algorithm group and non-algorithm
group.

This will be an open, two arm, parallel group, randomised, controlled
investigation of the ANSeR Software System as a stand-alone neonatal seizure
recognition decision support tool. Term neonates requiring EEG monitoring will
be stratified by recruiting site and then randomised to receive either EEG
monitoring with the ANSeR Software System or EEG monitoring without the ANSeR
Software System.

EEG monitoring with the ANSeR Software System as a seizure detection algorithm.

Neonates randomised to the control arm of this investigation will receive
standard clinical care and continuous EEG monitoring as per protocols at the
local sites. There is no additional risk envisaged in this group.*Neonates
randomised to the active arm will receive standard clinical care and continuous
EEG monitoring with the support of the ANSeR Software System. The ANSeR
software analyses EEG data that is collected as normal standard of care and
emits a continuous output of the ANSeR Seizure Detection Algorithm at the cot
side in real time. The ANSeR Software System does not provide any diagnostic
conclusion.
Treatment of any seizures identified by the investigation personnel in both
arms of the investigation will be as per local procedures and guidelines.*The
hypothesis is that the diagnostic accuracy of neonatal seizures will improve
when the ANSeR Software System is available to alert healthcare professionals
to potential seizures and inform the diagnostic decision. Real-time continuous
assessment of the aEEG with a maximum of two channels is standard clinical care
in our center. Detailed assessment of a multiple-channel EEG is performed
retrospectively at this moment. Therefore, we would expect an improvement in
the accuracy of seizure detection in neonates randomised to the active arm.
This may lead to improved appropriateness and promptness of the AED treatment
regime.


A retrospective study of 300 paediatric and adult patients indicated that
continuous monitoring of EEG could lead to a change in AED administration in
52% of patients included in the study. While this figure is startling it should
be remembered that in neonates, the clinical indications of seizures are often
not present or are more subtle in comparison to those seen in the adult or
paediatric population making identification of seizures even more difficult in
neonates.
The potential benefits of being randomised to the ANSeR Software System
outweigh the risks. It is anticipated that the ANSeR Software System will be
safe for use and will provide accurate information to assist in seizure
diagnosis, which is to the benefit of the neonate and the healthcare system.