The capacity of neurological pupil index to predict absence of somatosensory evoked potentials after cardiac arrest—A study protocol

Anoxic‐ischemic brain injury is the most common cause of death after cardiac arrest (CA). Robust methods to detect severe injury with a low false positive rate (FPR) for poor neurological outcome include the pupillary light reflex (PLR) and somatosensory evoked potentials (SSEP). The PLR can be assessed manually or with automated pupillometry which provides the neurological pupil index (NPi). We aim to describe the interrelation between NPi values and the absence of SSEP cortical response and to evaluate the capacity of NPi to predict the absence of cortical SSEP response in comatose patients after CA.


| BACKG ROUND
The major cause of death among comatose survivors of cardiac arrest (CA) is global hypoxic-ischemic brain injury. 1,2 The initial phase of anoxemia/ischemia damages the regions of the brain most sensitive to hypoxia: the hippocampus, thalamus, basal ganglia, and cerebral cortex. 3 In the cortex, the synaptic activity ceases within seconds after CA. 4 With prolonged CA, the injury also affects areas less sensitive to hypoxia, such as the brainstem. 5 After successful resuscitation, the reperfusion may worsen the injury in all parts of the brain during the next several days. [6][7][8] To assess the neurological prognosis in comatose patients after primary resuscitation and stabilisation, international guidelines advocate a multimodal approach. 9 Most tests are applied to identify brain injury consistent with poor neurological outcome.
The recommended methods are electroencephalography (EEG), computed tomography, magnetic resonance imaging (MRI), pupiland corneal reflex, somatosensory evoked potentials (SSEP), and biomarkers of brain injury, such as neuron specific enolase (NSE). 9 Each method tests different neurological functions, reflexes, or the extent of injury. A combination of methods enables neurological prognostication with high specificity. 10 A stringent prognostication approach is vital as withdrawal of life-sustaining treatment (WLST) based on inadequate assessment of prognosis can result in death as a "self-fulfilling prophecy". 11,12 However, it is both time and resource consuming to include all the recommended methods in everyday practice prognostication. To establish the interrelation of NPi and SSEP have the potential to save resources in the multimodal approach to prognostic assessment by using only one of two clearly interrelated methods.

| Pupillary light reflex
The pupillary light reflex (PLR) arc is dependent of an intact brainstem function. The afferent limb involves the optic nerve, which synapses on neurons in the pretectal nucleus and then the Edinger-Westphal nucleus located in the midbrain. The efferent limb involves the oculomotor nerve axons, and the ciliary ganglion which enables pupillary muscle contraction. A bilateral PLR absence indicates a severe injury in the midbrain and signifies a poor neurological outcome in comatose CA patients. 13 The manually assessed PLR, also called standard PLR (sPLR), is associated with imprecision due to human error, environmental and patient factors. 14 Pupillometry is a semiautomated method that provides a quantitative measure of the PLR. With a standardized light stimulation, the handheld pupilometer measures the pupil size, latency, constriction-, and dilation velocity. Based on the values obtained, it calculates an index of the PLR known as the neurological pupil index (NPi). 15 An NPi of 0 means that the PLR is absent, and an NPi of 1-5 ranges from an abnormally slow to normal PLR. 15 A transiently absent PLR is common during the first hours following the CA. If the circulatory arrest was brief, the brainstem may recover fully. Patient factors that could make the PLR unreliable are ophthalmological pathologies such as asymmetric glaucoma, Argyll Robertson pupil, preexisting optic neuropathies. 16

| Somatosensory evoked potential
Somatosensory evoked potentials measures the cortical response to median nerve stimuli. Stimulating electrodes are placed on the forearm and receiver electrodes are placed on the neck and scalp.
The electrically induced signal travels from the median nerve along to the spinal cord, via the medial lemniscus in the midbrain to the contra-lateral somatosensory cortex. A signal peak can be detected by the electrodes at specific anatomical locations at specific time points. The N13 corresponds to the cervical spine and the N20 corresponds to the primary somatosensory cortex. A bilaterally absent N20 signal concomitant with a confirmed N13 signal is a strong predictor of poor neurological outcome in comatose patients after CA. 17

| Rationale and hypothesis
The PLR and SSEP are robust methods with low false positive rates (FPR) for poor prognosis in the multimodal prognostication after CA. 18,19 As explained above, both PLR and SSEP depend on intact midbrain function. Anatomically, the medial lemniscus, the Trial registration: ClinicalTrials.gov, NCT04720482, Registered 21 January 2021, retrospectively registered.

Editorial comment
The interrelation between loss of the pupillary light reflex (PLR) and the loss of cortical response to a somatosensory evoked potential (SSEP) in comatose cardiac arrest patients is not known. This exploratory prospective study is designed to evaluate whether a specific degree of attenuated PLR, as measured by semiautomated pupillometry, can predict the bilateral loss of cortical SSEP response in severe anoxic/ ischemic brain injury. Such an interrelation between the two methods would enable the use of pupillometry rather than the more resource demanding SSEP for neurologic prognostication in post cardiac arrest patients.
Edinger-Westphal nucleus, and the oculomotor nerve nucleus are neighboring structures in the midbrain. A severe hypoxic lesion could affect these structures causing an absence of both PLR and SSEP response. Because the brainstem is more resistant to hypoxia than the cortex, the cortical response to SSEP is likely to cease prior to the PLR. Even so, if the ischemic injury is severe enough, it will affect the PLR. In theory, the reaction is first slowed down, before it finally fades completely-corresponding to a progressively decreasing NPi value.
Our reasoning is that there should be an interrelation between PLR and SSEP. A patient with ischemic brain injury who has an abnormal or absent PLR will most likely have a discontinued cortical SSEP response.
We hypothesize that a certain NPi value is predictive of a bilaterally absent N20 SSEP response. Ideally, there should be a numeric NPi cut-off, with a high specificity for absent cortical SSEP response. If the hypothesis can be confirmed, it would allow SSEP testing to be excluded in certain patients. Instead, the technically less demanding pupillometry could be used as its proxy in the multimodal prognostication after CA.

| Ethical approval
The study is approved by the Swedish Ethical Review Authority (DNR 2019-00823, 2020-00506). A consent is obtained from patients regaining consciousness or the next of kin when appropriate after receiving verbal and written information.

| Study design
An explorative, prospective, noninterventional, observational, cohort study with consecutive inclusion. The patients are submitted to clinical routine prognostication procedures, including pupillometry and SSEP. The study data will be collected from medical records and patient monitoring systems.

| Research aim
To describe the association between NPi and bilateral absence of the cortical SSEP N20 signal in patients remaining comatose after CA, and to define a NPi cut-off value that renders a <5% FPR for a bilaterally absent SSEP cortical response.

| Setting
The study will be conducted in the Sahlgrenska University Hospital, the referral center for specialized cardiac care for the 1.7 million inhabitants of western Sweden. The yearly admittance of resuscitated survivors of CA in the general ICU is 50-60 patients, and around 50% will remain comatose on day 3.

| Postcardiac arrest intensive care
The local standard operations procedure (SOP) for comatose survi-

| Population and study size
We will include a minimum of 50 adult (>18 years) survivors of CA remaining comatose > 48 hours during post resuscitation care in the intensive care unit.
Exclusion criteria will be regained consciousness <48 hours after CA, or before examination by pupillometry or SSEP is performed; pregnancy; intracranial bleeding; traumatic brain injury, palliative care and lack of next of kin.

| Outcome measures
Primary outcome: sensitivity, specificity, and odds ratio for NPi to predict bilateral absence of the N20 SSEP signal, including the NPi value corresponding to a 5% FPR for absent cortical SSEP response.
Secondary outcome: the PLR and SSEP sensitivity, specificity and odds ratio for poor neurological outcome at hospital discharge and death at 30 days.

| Procedures
Pupillometry is routinely performed together with SSEP (±1 hour) >48 hours after CA using a handheld device for automated infra-

| Neurological and survival outcome
Neurological outcome is assessed at hospital discharge from available medical records data according to the Modified Rankin Scale (mRS). 20,21 A poor neurological outcome is defined as mRS 4-6 (unable to walk and attend bodily need without assistance, bedridden requiring constant nursing care, dead). Survival status at 30 days.

| Additional data collected
Demographic and clinical data along with test results will be retrieved from medical records. Physiological data during intensive care will be retrieved from patient monitoring systems.

| Statistical considerations and data analysis
Existing data applicable to a calculation of sample size were identified in the online supplement of a study by Oddo et al, reporting the accuracy of NPi as compared to manual assessment of PLR. 14 Their data included the association between PLR, SSEP, and outcome in patients included by criteria similar to ours. Provided NPi values were median and interquartile range (IQR) on day 3 after CA, categorised according to neurological outcome at 3 months.
NPi was 4.5 (4.2-4.7) in the favorable outcome group and 3.7 (3.3-4.2) in the unfavorable outcome group respectively. From the study by Moseby-Knappe, based on the "TTM trial" cohort we estimated the prevalence of absence of SSEP response to be 37% correspondingly. 10,24 In order to find a significant difference in NPi of 0.7 with a power of 95% with two-sided Fisher's nonparametric permutation test, 45 patients are needed, assuming a 2:1 allocation and unequal SD in the groups (0.37 and 0.67 calculated from the IQR above), and significance level 0.01. To account for uncertainty within these estimates, we aim to include 50 patients with a complete protocol.
A receiver operating characteristics curve will be used to find the NPi cut-off values resulting in a <5% FPR for absent SSEP to predict poor neurological outcome. NPi values below the cut-off, that is, values consistent with poor outcome, will be used to calculate the predictive value for SSEP at its given prevalence. Fisher's exact test will be used to assess correlation between NPi and SSEP.

| D ISCUSS I ON
The aim of this study is to assess the interrelation between PLR and SSEP, two robust tests used in the multimodal neurological prognostication of comatose patients after CA. Understanding the association between the different methods is essential in order to use the most appropriate and rational methods for the individual patient.
We hypothesize that the PLR and SSEP have a systematic interrelation, and that a certain NPi value has the capacity to predict the absence of cortical SSEP response. If confirmed, the prognostication procedure could be rationalised to save resources in the assessment of neurological prognosis. Similar reasoning has been applied to the association between EEG and SSEP. With a normal or "benign" EEG, the cortical SSEP response was invariably present. 25,26 Thus, it has been advocated that patients with such EEG patterns, SSEP is not required for the assessment of neurological prognosis after CA.
The absence of the PLR is a well-established clinical predictor of poor neurological outcome after CA if unaffected by sedation or other factors obtunding the reflex. 27 Confounders to the manual assessment of PLR includes a small pupil size and darker eye colour. 14,[28][29][30] The use of automated infrared pupillometry is more reliable than manual assessment of PLR as it can detect even small changes, otherwise easily missed. Also, the use of pupillometry has been shown to be reliable, and the staff procedure training uncomplicated. 14,31 However, if the eyes are rolling away or undulating, the examination cannot be performed. Certain ophthalmological pathologies are likely to cause unreliable pupillometry and sPLR results. 16 If present, such obstacles will be recorded in the present study database.
The absence of the cortical N20 SSEP signal a strong predictor for poor neurological outcome after CA and recommended in international guidelines. 9,17 Yet, no prognostication method is 100% certain and false positive cases exists. 10,32,33 One important factor influencing the reliability of the interpretation is the noise level in the recording. 34,35 Several false positive SSEP predictions of poor neurological outcome are associated with avoidable confounders such as noise and inconsistent interpretations. 36 A standard recording technique and protocol for interpretation is preferred to raise the reliability of the method. Noise level can be reduced by administering muscle relaxant, and by turning off nonessential electrical ICU equipment during the SSEP registration. 37

| Limitations
There are some limitations to our study. The inherent limitations of an observational design apply. However, the prospective approach can ascertain good quality data and provide a good estimate for the interrelation of the applied methods. Although bilateral absence of the cortical N20 SSEP response has a high specificity for poor neurological outcome, its sensitivity rarely exceeds 50%. 27 Furthermore, the capacity of NPi to predict the absence of SSEP will depend on the prevalence of SSEP. Consequently, the external validity of our results will be limited to patient populations with similar prevalence of absent SSEP. Although, if we can provide accurate sensitivity and specificity estimates for NPi to predict absent SSEP response, the values obtained can be used to calculate the NPi predictive value in other populations, given their specific prevalence of absent SSEP response.

| Clinical use
The interrelation between PLR and SSEP has not been detailed before and the results of the present study may infer alternative conclusions. If our hypothesis is confirmed, the additional use of SSEP may be superfluous if PLR can be reliably assessed using pupillometry and NPi. However, if the association between the methods turns out irregular, SSEP may be the preferred method in neurological prognostication if the PLR is not entirely absent. Regardless of the robustness of our results, the study will improve the understanding of the prognostication methods and can help to use the available resources more efficiently.

ACK N OWLED G EM ENTS
The study will use departmental funding only.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no competing interests.

E TH I C A L A PPROVA L A N D CO N S E NT TO PA RTI CI PATE
This project is approved by the Swedish Ethical Review Authority (DNR 2019-00823, 2020-00506). Inclusion and management of patients will comply with the Helsinki Declaration for research in human subjects and data will be handled and stored according to the European General Data Protection Regulation (GDPR) 38 . Verbal and written consent to use routinely registered data for the study will be obtained at the earliest possible occasion with respect to the severity of the patient's condition and the likely emotional distress of family members. Consent can be withdrawn at any time. Patient ID will be registered and stored on a physical paper, stored in a locked space in the ICU department. Each patient will be assigned a unique serial number, used for data registration and analysis. The datafiles are stored on secure servers protected by firewalls and requiring authorised access.

AUTH O R S ' CO NTR I B UTI O N S
CR conceptualized the study and obtained ethical approval. LL, SJ and MT designed the protocol. LL, SJ and MT wrote the manuscript, which was revised by JN, PR, PL and CR. All the authors read and approved the final version of the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets generated and analyzed during the current study are not publicly available due to the intrusion of the patient's privacy.
But the datasets are available from the corresponding author on request. The results will be published in a peer-reviewed medical journal.