Predictive Utility of Changes in Optic Nerve Sheath Diameter after Cardiac Arrest for Neurologic Outcomes

The optic nerve sheath diameter (ONSD) can help predict the neurologic outcomes of patients with post-cardiac arrest (CA) return of spontaneous circulation (ROSC). We aimed to investigate the effect of ONSD changes before and after CA on neurologic outcomes in patients with ROSC after CA using brain computed tomography (CT). The study included patients hospitalized after CA, who had undergone pre- and post-CA brain CT between January 2001 and September 2020. The patients were divided into good and poor neurologic outcome (GNO and PNO, respectively) groups based on their neurologic outcome at hospital discharge. We performed between-group comparisons of the amount and rate of ONSD changes in brain CT and calculated the area under the curve (AUC) to determine their predictive value for neurologic outcomes. Among the 96 enrolled patients, 25 had GNO. Compared with the GNO group, the PNO group showed a significantly higher amount (0.30 vs. 0.63 mm; p = 0.030) and rate (5.26 vs. 12.29%; p = 0.041) of change. The AUC for predicting PNO was 0.64 (95% confidence interval = 0.53–0.73; p = 0.04), and patients with a rate of ONSD change >27.2% had PNO with 100% specificity and positive predictive value. Hence, ONSD changes may predict neurologic outcomes in patients with post-CA ROSC.


Introduction
Ischemia/reperfusion cerebral injury after cardiac arrest (CA) may cause cerebral edema. [1,2] This results in an increase in intracranial pressure (ICP) and contributes to poor neurologic outcomes in patients with post-CA return of spontaneous circulation (ROSC). [3,4] In these survivors, there is a need for early detection of increasing ICP and neurologic outcome prediction to allow appropriate post-resuscitation care. [5] This allows priority to be given to patients with expected good neurologic outcomes given the limited medical resources. There have been studies on various predictive factors for post-CA neurologic outcomes, including neurologic examination of brainstem re exes, electrophysiological tests, and serum biomarkers, such as neuron speci c enolase and S-100B. [6][7][8] However, these factors have been recommended as prognostic factors at 72 post-CA hours. [8][9][10] Moreover, early brain computed tomography (CT) of patients with post-CA ROSC may play a crucial role as a prognostic predictor. Further, the American Heart Association guidelines recommend early post-CA brain CT scans and con rm that a decrease in the grey-to white matter ratio (GWR) can help predict neurologic outcome. [10][11][12][13] Additionally, there have been studies with optic nerve sheath diameter (ONSD) on brain CT for predicting neurologic outcomes in post-CA survivors. [14][15][16] Previous studies have shown that the ONSD on brain CT could be a useful tool as a non-invasive method of ICP measurement. [17,18] Additionally, recent studies have demonstrated that the ONSD on brain CT is useful for early neurologic outcome prediction through evaluation of increased ICP in patients with post-CA ROSC. [15,19] A recent meta-analysis con rmed the utility of ONSD as a prognostic factor for neurologic outcomes in post-CA patients. [20] However, most of the studies included in the meta-analysis indicated that sole use of ONSD had limited predictive utility for prognosis. Furthermore, all these studies only measured the post-CA ONSD without considering individual differences.
This study aimed to assess differences between pre-and post-CA ONSD in patients with ROSC after CA using brain CT imaging. Additionally, we aimed to investigate the impact of the amount and rate of post-CA ONSD changes on the neurologic outcome at discharge.

Study design and population
This retrospective observational cohort study investigated brain CT scans of patients hospitalized after CA who visited the emergency department of a single university-a liated hospital in Korea between January 2001 and September 2020. This study was approved by the Institutional Review Board of Hanyang University Guri Hospital (IRB No. GURI 2020-12-008), which waived the requirement of informed consent.
We included adult patients hospitalized after CA who underwent pre-and post-CA brain CT. The exclusion criteria were as follows: (1) being transferred to another hospital after ROSC, (2) age < 19 years, (3) having traumatic/non-traumatic brain hemorrhage or brain tumor, (4) a history of ophthalmological disorders or surgeries that could affect ONSD, and (5) having the most recent pre-CA brain CT obtained at an age < 19 years. Finally, eligible patients were divided into the GNO (good neurologic outcome) and PNO (poor neurologic outcome) groups based on the neurologic outcome at discharge; subsequently, we measured ONSD changes and performed between-group analysis. The primary outcome was the association between ONSD changes and the neurologic outcomes of patients hospitalized after CA.

Data collection
We retrospectively collected the following data from electronic medical records: age, gender, comorbidities (hypertension, diabetes, myocardial infarction), etiology (cardiac, respiratory), location of CA, whether the CA was witnessed, bystander CPR, rst monitored shockable rhythm, CA duration including the no-ow (time between CA and CPR initiation) and low-ow time (time between active CPR and ROSC), and administered targeted temperature management (TTM). Based on the medical records, we determined the interval between the latest pre-CA brain CT and ROSC (month), which was termed as 'CT to ROSC', and between ROSC and post-CA brain CT (minute), which was termed as 'ROSC to CT'. Additionally, we collected data regarding the neurologic outcomes on discharge using the Glasgow-Pittsburgh Cerebral Performance Categories (CPC). Based on the CPC scale, we de ned good (GNO) and poor neurologic outcomes (PNO) as a CPC 1 or 2 and 3-5, respectively. ONSD measurements using brain CT Brain CT scans were performed based on standard protocols using non-contrast 4-mm contiguous slices parallel to the orbital oor from the skull base to the vertex. The pre-CA and post-CA ONSDs were bilaterally measured at 3 mm behind the globe on brain CT using the picture archiving and communication system (PACS) ruler tool (PiView STAR, INFINITT, Seoul, Korea). Images were magni ed at 450% and changed to the 'mediastinum' window (window width: 440; window level: 45) using a PACS tool. The ONSDs of the right and left eyes were averaged to obtain the mean value. All measurements were performed using emergency physicians blinded to the patient information, including neurologic outcome. Additionally, we calculated the amount and rate of ONSD change. We de ned the amount of change as the difference between the pre-CA and post-CA ONSD. Moreover, the rate of ONSD change was calculated as follows: We used the following three CT equipment: Somatom Sensation 16, SOMATOM De nition DS, and SOMATOM De nition Edge (Siemens Healthcare, Erlangen, Germany). The following parameters were used: 120 kVp, 250-500 mAs, and 4 to 4.5-mm slice thickness. All CT images were stored as the Digital Imaging and Communication in Medicine format in the PACS.

Sample size
We calculated the sample size based on a pilot study on 33 participants using G*Power (3.1.9.6; Heine Heinrich University, Düsseldorf, German). The mean ONSD of patients with GNO and PNO was 4.75 ± 1.45 mm and 5.63 ± 1.85 mm, respectively. The required sample size was calculated as 90 participants (effect size: 0.53, a-error: 0.05, power: 0.8); nally, considering a 10% drop-out rate, 99 participants were required.

Statistical analysis
Continuous and categorical variables were reported as the median with interquartile range (IQR) and number with percentages. Normally distributed variables were analyzed using the Mann-Whitney U-test and Wilcoxon rank-sum test while non-normally distributed variables were analyzed using the Shapiro-Wilk test. Chi-square tests or Fisher's exact test were used to analyze categorical variables. Statistical signi cance was set at P < 0.05. Multivariable analysis with logistic regression was used to determine the risk factors for poor neurologic outcomes with adjustment for confounding variables found signi cant on univariate analysis. Variables with p < 0.2 on univariate analysis with the rate of ONSD change were included in the multivariable analysis. Further, the Hosmer-Lemeshow test was used to con rm the logistic model calibrations. The predictive performance of the main outcome was assessed using the area under the receiver operating characteristic ([ROC] AUC) of the sensitivity over 1 -speci city. Results were obtained using the Youden index and presented as a 95% con dence interval (CI) of AUC with sensitivity, speci city, positive predictive value (PPV), and negative predictive value (NPV). ROC analysis was performed using MedCalc Statistical Software (version 17.2, MedCalc Software, Ostend, Belgium) while the other statistical analyses were performed using SPSS software (version 25.0, IBM, Armonk, NY).

Baseline characteristics
Among 145 post-CA survivors who underwent brain CT before and after CA, 49 patients were excluded as follows: 40 patients who transferred to another hospital, seven patients with intracranial or subarachnoid hemorrhage, one patient with a brain tumor, and one patient aged ≤ 18 years. Finally, we enrolled 96 patients and allocated them to the GNO group (n = 25, 26.0%) and PNO group (n = 71, 74.0%) (Fig. 1). Table 1 summarizes the demographic and clinical characteristics. The median age of the included patients was 70 (IQR: 58-79) years with 56.3% being male. The GNO group was signi cantly younger than the PNO group. Moreover, the GNO group showed a signi cantly higher frequency of cardiac etiology and shockable rhythm, as well as a shorter no-ow and low-ow time, than the PNO group. Contrastingly, out-of-hospital cardiac arrest was more frequent in the PNO group. * The interval between the latest pre-CA brain CT and ROSC. † The interval between ROSC and post-CA brain CT.
Comparison of pre-CA and post-CA ONSDs In both groups, the post-CA ONSD was signi cantly higher than the pre-CA ONSD (Fig. 2). In the GNO group, there was a signi cant difference between the pre-CA ONSD and post-CA ONSD (5.06 vs. 5.50 mm, p < 0.001) in the GNO group. Similarly, in the PNO group, there was a signi cant difference between the pre-CA ONSD and post-CA ONSD (5.07 vs. 5.72 mm, p = 0.001) (Supplemental Table 1).
The association between ONSD changes and neurologic outcomes   * The interval between ROSC and post-CA brain CT.
Diagnostic value of ONSD changes for predicting the neurologic outcome The AUC for predicting PNO was 0.64 (95% CI = 0.53-0.73; p = 0.04) in the ROC curve for the rate of ONSD change (Fig. 3). Patients with a rate of ONSD change > 27.2% had PNO with speci city and PPV of 100%. GNO could be predicted using a cut-off value of ≤ 5.83% in the ROC curve for the rate of ONSD change, with a sensitivity and speci city of 60.0% and 76.06%, respectively. The PPV and NPV were 46.9% and 84.4%, respectively (Table 4).

Discussion
This study found that the amount and rate of ONSD change were signi cantly associated with neurologic outcomes. Further, there was no signi cant between-group difference in the post-CA ONSD. However, there was no independent association of the rate of ONSD change with neurologic outcomes after adjusting for confounding variables. Together with other established predictors, the rate of ONSD change may be useful for predicting neurologic outcomes. To our knowledge, this is the rst study to investigate individual differences in ONSD changes among post-CA survivors.
Previous studies have reported an association of neurologic outcomes in critically ill patients, including post-CA survivors, with increased ICP. [3,4,21,22] The optic nerve sheath covers the optic nerve and is comprised of a subarachnoid space layer lled with cerebrospinal uid (CSF). [23] ICP is positively correlated with the CSF pressure and ONSD length. [23,24] ONSD is a potential non-invasive ICP estimator and could be useful for assessing intracranial hypertension. [24] In patients with post-CA hypoxic cerebral injury, increased ICP is associated with neurologic outcome. [3,21,22] Several studies have reported that ONSD can predict neurologic outcomes in post-CA survivors. A retrospective cohort study from Korea reported an association of longer ONSD on initial brain CT with poor neurologic outcome. [15] Chelly et al. demonstrated that ONSD could be an early prediction tool for outcomes in post-CA patients treated using TTM. [19] Other studies have used ONSD combined with other predictors, including GWR or albumin levels, to enhance the predictive value. Moreover, a recent metaanalysis reported that ONSD could be useful for predicting neurologic outcome. [14,20,21] Inconsistent with these ndings, a registry-based multicenter study reported no correlation between ONSD on early unenhanced brain CT and neurologic outcome in post-CA survivors managed using TTM. [25] Previous studies have reported that predicting neurologic outcomes in post-CA survivors using post-CA ONSD alone is insu cient with limitations.
ONSD can allow non-invasive ICP measurement and could serve as a surrogate marker for increased ICP. [17,18] However, in healthy adults, there are differences in the baseline ONSD according to individual characteristics, including sex, body mass index (BMI), race, or eyeball size. [26,27] Most studies of healthy volunteers have reported that the mean ONSD ranges from about 3 mm to 5 mm; moreover, the reported mean or median ONSD values have varied across study cohorts depending on the race or measurement tools. [26][27][28][29] Ultrasonographic evaluation of healthy Asians revealed a longer ONSD in males and individuals with high BMI. [26,27] Therefore, these individual differences could confound the interpretation of post-CA ONSD; moreover, it may be useful to consider the baseline ONSD for improving the prognostic value. Therefore, ONSD changes may be useful markers for ICP measurement changes. A prospective observational study on ONSD changes in patients with hydrocephalus reported a signi cant reduction in ONSD after ventriculoperitoneal shunt operation. [30] In our study, ONSD changes were more reliable than the ONSD itself for predicting neurologic outcomes in patients with post-CA.
A recent meta-analysis reported that compared with CT and MR, sonographic measurement allowed more accurate prediction of neurologic outcome in patients with post-CA. [20] However, obtaining and comparing pre-CA and post-CA ONSD using ultrasound could be limited in clinical settings. Moreover, determining the pre-CA ONSD using brain MR has limitations given its speci c modality. Recent studies have indicated that the axial proton density/T2-weighted turbo spin-echo fat-suppressed sequence is required for ONSD measurements using MR. However, in most post-CA patients, the T2-TSE image is not included in the diffusion-weighted MR. [31,32] Additionally, there is a strong association of ONSD with eyeball transverse diameter (ETD) and ONSD/ETD ratio in healthy adults. [33] There is a need for further studies on the association between ONSD/ETD ratio and neurologic outcome in post-CA patients.
This study has several limitations. First, this was a single-center study with a limited sample size that led to an insu cient statistical power; however, we calculated the sample size, which was relatively large compared with that of other studies. Second, this retrospective study included patients who underwent both pre-and post-CA brain CT, which could lead to selection bias affecting the results. Third, although we attempted to extensively collect variables, there could be hidden confounders. Fourth, there could have been minor measurement errors given the very small size of the ONSD in brain CT. However, to minimize these errors, two blinded emergency physicians performed measurements using a standardized method with consensus. Fifth, current guidelines recommend neurologic outcome assessment at 3 months after discharge. However, we measured the neurologic outcome at discharge and did not determine the longterm outcome. Sixth, this was a retrospective study and the clinical utility of the predictive value for prognosis remains unclear. There is a need for large-scale prospective studies to enhance our ndings.