a b s t r a c t

Background: Acute brain lesions on diffusion-weighted-magnetic resonance imaging (MRI) after acute Carbon monoxide poisoning were associated with delayed neurological sequelae. This study was conducted to identify the risk factors associated with acute brain lesions on MRI after acute CO poisoning and to help select pa- tients who need acute-phase brain MRI after Acute CO poisoning in the emergency department (ED).

Methods: This retrospective observational study included 103 adult patients who were hospitalized at a tertiary- care hospital between November 2016 and September 2019 and underwent brain MRI because of acute CO poi- soning. Multivariable logistic regression analysis was applied to identify predictive factors for acute brain lesions on MRI after acute CO poisoning.

Results: Multivariable logistic regression analysis showed that Glasgow Coma Scale score of b9 at ED pre- sentation (odds ratio [OR] 17.749, 95% confidence interval [CI] 3.098-101.690, P = 0.001) and the initial troponin-I level at presentation in the ED (OR 13.657, 95% CI 1.415-131.834, P = 0.024) were predictive factors for acute brain lesions on MRI in acute CO poisoning. The receiver operating characteristics curve for initial troponin-I showed an area under the curve of 0.761 (95% CI 0.638-0.883, P b 0.001) and the optimal cutoff value was 0.105 ng/mL.

Conclusions: Acute-phase brain MRI in acute CO poisoning can be considered for patients who present at the ED with a GCS score b9 or troponin-I level N0.105 ng/mL.

(C) 2020

Introduction

carbon monoxide poisoning is one of the main suicide methods in Far East Asia and has an increasing incidence [1-4]. CO poisoning not only results in acute poisoning because of hypoxia, but also confers chronic neurological sequelae in 30% of patients after recovery from acute poisoning [5-7]. Hyperbaric oxygen therapy (HBOT) is frequently used to treat acute CO poisoning, although its effectiveness remains controversial [8,9]. Moreover, the precise indications as well as the ap- propriate time and frequency for the application of HBOT have not yet been established; therefore, the indications for HBOT are based on ex- pert recommendations, including loss of consciousness, ischemic car- diac changes, neurological abnormalities, severe metabolic acidosis, or

Abbreviations: DW-MRI, diffusion-weighted-magnetic resonance imaging; ED, emergency department; CO, carbon monoxide; GCS, Glasgow Coma Scale; OR, odds ratio; CI, confidence interval; ROC, receiver operating characteristics; AUC, area under the curve; HBOT, hyperbaric oxygen therapy.

* Corresponding author at: Department of Emergency Medicine, Inha University Hospital, 27, Inhang-ro, Jung-gu, Incheon 22332, South Korea.

E-mail address: [email protected] (J.H. Kim).

N25% of carboxyhemoglobin (COHb) [9]. There are two types of chronic neurological sequelae: manifestations that occur from the initial stage of CO poisoning and persist, and Delayed neuropsychiatric sequelae (DNS). The optimal therapy to prevent DNS has not been standardized, al- though neuroprotective bundle therapy, comprising antioxidants, ste- roids, erythropoietin, and Targeted temperature management, is being proposed in addition HBOT [9,10]. Predictive factors of the development of chronic neurological sequelae have been studied, although no consis- tent predictors have been identified [3,11]. In two recent studies, acute brain lesions observed upon diffusion-weighted-magnetic resonance imaging (DW-MRI) undertaken during acute-phase CO poisoning were associated with DNS [12,13]. Given the absence of established Treatment protocols with proven efficacy to prevent DNS, the selection of a treatment from the various available strategies, including HBOT, could potentially be determined based on acute brain lesions on MRI in patients with acute CO poisoning. Studies on predictive discernment of acute brain damage on MRI in patients with acute CO poisoning are rare. Indications for brain MRI during acute-phase CO poisoning have not yet been established. Therefore, this study was conducted to deter- mine the necessity of an acute brain MRI in patients with acute CO

https://doi.org/10.1016/j.ajem.2020.05.076

0735-6757/(C) 2020

poisoning in the emergency department (ED) by identifying risk factors associated with acute brain lesions.

Methods

Study design and population

This retrospective observational study was conducted at a single tertiary-care hospital which has a monoplace Hyperbaric chamber. The study design was approved by the institutional review board of the study center, and the need for written informed consent was waived as no patient-identifying data were included in the manuscript.

The medical records of consecutive patients who were admitted to our hospital with CO poisoning via the ED between November 2016 and September 2019 were reviewed. Inclusion criteria were age 19 years or higher and presentation at the ED within 24 h of the last CO ex- posure, hospitalization for longer than 2 days, and acute-phase brain MRI data (conducted within 240 h of the last CO exposure). Patients who went into cardiac arrest before brain MRI or with insufficient ob- servation periods were excluded.

Management

All patients who were included in this study received HBOT in accor- dance with the institutional protocol. The institutional HBOT protocol for patients with CO poisoning is as follows. HBOT is administered in a

monoplace chamber with a target atmospheres absolute (ATA) value set to 2.8 for 90 min, although the total execution time per session is 130 min because the pressure is gradually increased and decreased at 20 min before and after the target ATA is reached.

The indications for HBOT at our hospital are based on expert recom- mendations [9], but the number of HBOT applications varies based on stratification into high- and low-risk groups by patient indications. Pa- tients with loss of consciousness, CO-exposure duration of N4 h or un- known duration, endotracheal intubation at presentation in the ED, a Glasgow Coma Scale score b13, elevation of cardiac enzymes, or abnormal brain imaging tests were assigned to the high-risk group and received three sessions of HBOT within the first 24 h of hospitaliza- tion and one session of HBOT per day during their hospital stay. Patients in the low-risk group received two sessions of HBOT during the first 24 h followed by one session of HBOT every day until the third day of hospi- talization. However, during actual application, the timing and frequency of HBOT were adjusted according to the patient’s condition and compli- ance. Therefore, these data were collected and analyzed after confirming the actual timing and number of HBOT sessions in each patient.

Brain MRIs were recommended in all patients demonstrating any one of the indications for HBOT at our hospital as abnormal findings on brain imaging is one of the criteria for inclusion in the high-risk group based on our protocol for HBOT. Both 3-T (Signa Architect, Dis- covery MR750; GE Healthcare) and 1.5-T (Signa HDxt, Signa Excite; GE Healthcare) MRI scanners were used in this study.

Fig. 1. Flow diagram of study population.

Data collection

The patient medical records were retrospectively reviewed by two emergency physicians, who collected clinical data, including gender, age, height, weight, comorbidity, intention of exposure, co-ingestion of alcohol or other drugs, duration of CO exposure, interval from last CO exposure to ED presentation, GCS score on initial ED presentation, interval from last CO exposure to first HBOT, interval from last CO expo- sure to brain MRI, number of HBOTs before brain MRI, brain MRI find- ings, and occurrence of DNS. In addition, laboratory data, including white blood cell count, levels of COHb, hemoglobin (Hb), C- reactive protein (CRP), albumin, lactic acid, troponin-I, and creatine ki- nase (CK) at presentation in the ED, were collected. When the patient was transferred to our hospital from another hospital for HBOT, the time of arrival and examinations at the ED were based on patient arrival at the ED of the referring hospital. When medical records were insuffi- cient or laboratory results were not available, only existing data were included for statistical analysis. The assessment of DNS occurrence was conducted from a neuropsychiatric battery or medical records after at least 6 weeks following CO exposure, and a telephonic survey

was conducted for patients with no records.

Table 1

Baseline characteristics of the study population.

Characteristics Total patients

(N

= 103)

Age (mean +- SD), years 43.0

(29.0-54.0)

Male, N (%) 64.0 (62.1%)

Height (mean +- SD), cm 168.14 +- 7.43

Weight (median, IQR), kg 64.8

(56.6-73.9)

BMI (mean +- SD), kg/m² 23.08 +- 3.71

Transfer from another hospital 59 (57.3%)

Intentional exposure 74 (71.8%)

Co-ingestion

Alcohol 31 (30.1%)

Sedatives 30 (29.1%)

Comorbidities 21 (20.4%)

GCS score b9 at presentation in the ED 18 (17.5%)

GCS score at presentation in the ED; 9 to 14 31 (29.5%)

COHb at presentation in the ED (mean +- SD), % 26.69 +- 23.86

duration of exposure to CO (median, IQR), hours 2.81

(1.00-7.50) n = 70

Formal readings by neuro-radiologic specialists were used for brain MRI findings. High signal intensities on T2/flair-weighted imaging with diffusion restriction or with typical findings of CO poisoning with- out previous hypoxic brain injury or CO exposure were considered to be acute brain lesions caused by acute CO poisoning.

The authors who reviewed the medical records were blinded to the

Interval from last CO exposure to presentation in the ED (median, IQR), minutes

Interval from last CO exposure to first HBOT (median, IQR), hours

Interval from last CO exposure to brain MRI (median, IQR), hours

49.0

(30.0-104.0)

3.13

(1.93-4.81)

n = 100

53.21

(39.50-67.45)

results of MRI and the DNS outcomes at the time of the initial review of the medical records.

Statistical analyses

Patients were assigned to two groups based on the presence or ab- sence of acute brain lesions on MRI (acute brain lesions group and no acute brain lesion group). We compared each variable between the two groups. The t-test was used for continuous variables with normal distribution, and the values were expressed as means and standard de- viations. The Mann-Whitney U test was used for continuous variables with non-normal distribution and the values were expressed as me- dians and interquartile ranges (IQR). For categorical variables, groups were compared using the Fisher’s exact test. Variables with a P-value b0.05 were included in a multivariable logistic regression model based on a backward elimination method. A receiver operating characteristic (ROC) curve was used to determine the predictive accuracy of the vari- able for acute brain lesions on MRI. P-values b0.05 were considered sta- tistically significant. All statistical analyses were undertaken using SPSS

19.0 (IBM Corp. Armonk, NY, USA).

Number of HBOT before brain MRI (median, IQR) 3 (3-4)

GCS score of 15 at discharge 99 (96.1%)

DNS occurrence 9 (18.4%)

n = 49

SD: standard deviation, IQR: interquartile range, BMI: body mass index, GCS: Glasgow Coma Scale, ED: emergency department, COHb: carboxyhemoglobin, HBOT: hyperbaric oxygen therapy, MRI: magnetic resonance imaging, CK: creatine kinase.

GCS score of 99 patients at discharge was 15, with the exclusion of one patient who died during hospitalization and three patients who had continuous altered mental status (Table 1).

Acute-phase brain lesions on MRI after acute CO intoxication

Brain MRI after acute CO intoxication revealed acute lesions in 28 pa- tients, of which 25 (89.3%) had lesions at the globus pallidus (GP). Eight

Table 2

Acute brain lesions on magnetic resonance imaging after acute carbon monoxide intoxication.

Region Total number (n = 28) Patient

Results

Baseline patient characteristics

During the study period, 460 patients (age >= 19 years) visited our ED with acute CO poisoning, and 229 were hospitalized for at least 2 days. Of the 229 hospitalized patients, MRIs were recommended in a total of 200 patients on the basis of the indications for HBOT; 95 patients re- fused to undergo the investigation, mostly because of economic reasons. Therefore, only 105 patients agreed to undergo a brain MRI. However, two patients were excluded because of cardiac arrest and cardiopulmo- nary resuscitation prior to brain MRI and for MRI conducted N240 h after acute CO poisoning, respectively. Therefore, 103 patients were finally included in this study (Fig. 1). Of the 103 patients, 64 (62.1%) were male and 21 (20.4%) had predisposing medical problems (Table 1). Ad- ditionally, 74 (71.8%) patients had intentionally been exposed to CO in suicide attempts, and all of them used briquettes. Three out of 103 pa- tients, for various reasons, did not receive HBOT before brain MRI. The

Globus pallidus (n = 25)

PVWM: periventricular white matter.

Only globus pallidus PVWM	17 5		#1 to #17 #19, #21, #22, #24,
Frontal	2	1	#25 #24, #25, #27
Parietal	3		#19, #21, #24
Temporal	3		#19, #21, #24
Occipital	2		#21, #24
Hippocampus	3		#18, #20, #25
Thalamus	2		#23, #25
Hypothalamus	3		#23, #24, #25
Corpus callosum	1		#25
Caudate nucleus	1		#25
Putamen		1	#28
Insular		1	#28
Pons	1		#24
Cerebellum	2	1	#19, #24, #26

Other lesion without globus

pallidus lesion (n= 3)

number

of the 25 patients had other brain lesions besides the GP lesions (Table 2).

Comparison between the groups with and without acute brain lesions on MRI

The following variables differed significantly between groups: hy- pertension (P = 0.023), duration of exposure to CO (P = 0.010), GCS score b 9 at presentation at ED (P b 0.001) and time interval to first HBOT after last CO exposure (P = 0.001). Additionally, WBC, CRP, initial troponin-I level, and initial CK level at presentation at ED were signifi- cantly higher in the acute brain lesion group than in the group without acute brain lesions (P = 0.005, P = 0.001, P b 0.001 and P b 0.001, re- spectively; Table 3).

Multivariable logistic regression analysis for acute brain lesions on MRI in acute carbon monoxide poisoning

Multivariable logistic regression analysis showed that initial troponin-I level at presentation in the ED (OR 13.657, 95% CI 1.415-131.834, P = 0.024) and GCS score b9 at presentation at ED (OR 17.749, 95% CI 3.098-101.690, P = 0.001) were independent pre- dictive factors for acute brain lesions on MRI in acute CO intoxication (Table 4). The ROC curve for initial troponin-I to determine the predic- tive accuracy for acute brain lesions on MRI demonstrated a fair AUC of 0.761 (95% CI 0.638-0.883, P b 0.001; Fig. 2) and the optimal cutoff

value according to Youden Index was 0.105 ng/mL, with a sensitivity of 70.4% and a specificity of 79.2%.

Prognosis of patients based on presence of acute brain lesions on MRI

Follow-up details were available for 74 of the 103 patients, and the rate of loss of follow-up did not different between the two groups (P

= 0.083; Table 5). Three patients who had sustained neurological ab- normalities from the presentation at the ED, 21 patients who had al- ready been diagnosed with depression before CO exposure, and one dementia patient were excluded because it was difficult to determine whether DNS had occurred. DNS was significantly more common in the acute brain lesion groups (P = 0.002; Table 5). The GP was the commonest site of abnormal lesions in the DNS patients (Table 6).

Discussion

Clinical symptoms and neurological complications of CO poisoning appear to be caused not only by hypoxia, but also by various effects of CO on cellular mitochondrial respiration, energy utilization, inflamma- tion, and free-radical generation [9]. Many studies have investigated the factors that can predict chronic neurological sequelae of CO poison- ing, but there is no neuropsychiatric battery that is customized for the diagnosis of cognitive impairment or Neurological problems caused by CO poisoning [2]. Moreover, all studies conducted to date have used dif- ferent neuropsychiatric assessments to evaluate the chronic

Table 3

Comparison between acute brain lesion group and group without acute brain lesions on MRI.

Variables Acute brain lesion group (n = 28) Group without acute lesions on MRI (n = 75) P value

Age (median, IQR), years 52.0 (35.0-59.5) 39.0 (28.5-53.0) 0.055

Gender (male), % 57.1% 64.0% 0.649

Height (mean +- SD), cm 166.9 +- 6.7 168.5 +- 7.7 0.323

Weight (median, IQR), Kg 62.5 (54.25-67.85) 65.0 (57.65-74.58) 0.083

BMI (median, IQR), Kg/m² 21.94 (19.69-23.86) 23.07 (20.07-25.99) 0.063

Comorbidities (%)

Hypertension 21.4% 5.3% 0.023^a

DM 10.7% 6.7% 0.680

Cardiac disease 10.7% 1.3% 0.072

Dyslipidemia 7.1% 0.0% 0.060

Intentional exposure (%) 67.9% 73.7% 0.624

Co-ingestion

Alcohol (%) 25.0% 32.0% 0.631

Drugs (%) 35.7% 26.7% 0.465

Duration of exposure to CO (median, IQR), hours 7.00 (3.00-8.25) (n = 19)

2.00 (0.70-4.50)

(n = 51)

0.010^b

Transferred from another hospital (%) 67.9% 53.3% 0.263

Time interval from last CO exposure to presenting at ED (median, IQR), minutes 60.50 (35.50-121.50) 46.00 (29.00-100.00) 0.319

GCS score at presentation at ED b 9, (%) 42.9% 8.0% b0.001^a

GCS score at presentation at ED; 9 to 14, (%) 32.1% 29.3% 0.812

COHb at presentation at ED (mean +- SD), % 27.1 +- 16.7 27.0 +- 14.5 0.966

Time interval from last CO exposure to first HBOT (median, IQR), hours 4.15 (3.36-5.28)

(n = 26)

2.93 (1.73-4.60)

(n = 74)

^b

Time interval from last CO exposure to taking brain MRI (mean +- SD), hours 63.32 +- 41.47 53.09 +- 28.89 0.160

Number of HBOT before brain MRI (median, IQR) 5.0 (3.0-5.0) (n = 26)

3.0 (3.0-4.0)

(n = 74)

0.362

WBC, x10³/uL, (median, IQR) 13.830 (9.380-20.120) 10.080 (7.515-13.575) 0.005^b

Hb (median, IQR), g/dL 14.2 (12.9-15.5) 14.9 (13.7-16.3) 0.121

Lactic acid (median, IQR), mmol/L 3.26 (1.80-7.42) (n = 25)

CRP (median, IQR), mg/dL 0.31(0.13-1.15)

(n = 26)

Albumin (median, IQR), g/dL 4.35 (4.00-4.60)

(n = 28)

Troponin- I (median, IQR), ng/ml 0.48 (0.10-0.99) (n = 27)

CK (median, IQR), IU/L 367.0 (154.5-1665.0)

(n = 28)

2.71 (1.49-4.39)

(n = 74)

0.08(0.03-0.21)

(n = 64)

4.30 (4.10-4.60)

(n = 74)

0.10 (0.10-0.10)

(n = 72)

114.0 (74.0-186.0)

(n = 74)

0.198

^b

0.845 b0.001^b b0.001^b

SD; standard deviations, IQR; interquartile range, BMI; body mass index, CO; carbon monoxide, ED; emergency department, GCS; Glasgow coma scale, COHb; carboxyhemoglobin, HBOT; hyperbaric oxygen therapy, MRI; magnetic resonance imaging, WBC; white blood cell, Hb; hemoglobin, CRP; C-reactive protein, CK; Creatinine kinase.

^a Fisher’s exact test.

^b Mann-Whitney U test.

Table 4

Independent predictors for acute brain lesions on magnetic resonance imaging in patients

Table 5

Prognosis based on presence of acute lesions on MRI.

GCS score b9 at presentation in the ED 17.749 3.098-101.690 0.001

Initial troponin-I at ED 13.657 1.415-131.834 0.024

OR: odds ratio, CI: confidence interval, GCS: Glasgow Coma Scale, ED: emergency department.

Underlying depression or dementia

Continuous altered mental status

n = 28

6 (25.0%) n = 24

with acute carbon monoxide intoxication.						With acute	Without acute	P-valuea
Variables	Multivariate analysis					lesion	lesions
	OR 95% CI	P-value			Lost to follow-up	4b (14.3%)	25 (33.3%)	0.083

3 (12.5%) n = 24

n= 75

16 (32.0%) n= 50

0 (0.0%) n= 50

0.597

0.031

DNS 7 (46.7%)

n = 15

2 (5.9%) n= 34

0.002

neurological sequelae of CO poisoning, and the results have varied; therefore, it is unclear which factors predict chronic neurological se- quelae. In two recent studies of a relatively large number of patients with CO poisoning, the presence of acute brain lesions was a predictor of the risk of DNS occurrence [12,13]. Although our study did not evalu- ate acute brain lesions on MRI after CO poisoning as a predictor of DNS, the findings showed that the occurrence of DNS was higher in the acute brain lesion group than in the group without acute brain lesions on MRI, which supports the findings of the two previous studies. In most studies, the commonest site of MRI abnormalities in CO poisoning was the GP, which was consistent with our findings [12-14]. Jeon et al. reported that the occurrence of DNS was only related to whether acute brain le- sions were present on MRI regardless of the lesion location, whereas an- other study showed that abnormalities in deep white matter and GP were related to DNS [12,13]. Most of our patients with DNS had abnor- mal lesions in the GP on acute-phase brain MRI, but the study sample was insufficient to determine the relationship between the abnormal regions and the occurrence of DNS.

In studies of TIA and ischemic stroke, both 1.5-T (Tesla) and 3-T MRI showed no false negatives when undertaken after 24 h of injury and 3 h of symptom appearance, respectively [15]. Therefore, the initial timing of the brain DWI-MRI in patients with CO poisoning should be deter- mined in reference to this time point. Shono K et al. reported that using 3-T MRI scanner, undertaking DWI within short latency (<=2 h) from Ischemic symptoms is an independent risk factor for false negative findings [15]. In this study, there was no significant difference between the two study groups in the time interval from the last CO exposure to obtaining a brain MRI. Moreover, no MRIs were taken within 2 h of ex- posure to CO. With regard to MRIs conducted within 24 h after CO poi- soning, there were 4 patients in the acute brain lesion group and 10 patients in the group without acute brain lesions. However, of these

MRI: magnetic resonance imaging, DNS; delayed neuropsychiatric sequelae.

^a Fisher’s exact test.

^b Including one patient who died during hospitalization.

14 patients, there were 2 and 7 patients who underwent 1.5-T MRIs in the groups with and without acute brain lesions, respectively, without any statistical differences between the two groups; the remaining 5 pa- tients underwent 3-T MRIs (Supplementary Table S1).

The relationship between consciousness at presentation in the ED or duration of CO exposure and chronic neurological sequelae or brain MRI abnormalities varies among studies [3,6,11-13]. In our study, there were significant differences in both duration of CO exposure and state of con- sciousness at ED presentation between the acute brain lesion group and the group without brain lesions on MRI; however, on multivariate anal- ysis, acute-phase brain MRI abnormalities and duration of CO exposure were unrelated, although the state of consciousness at the time of pre- sentation in the ED was related to acute-phase brain MRI abnormalities. In this study, we divided the state of consciousness based on a cutoff of the GCS score (b9) based on the study by Liao et al. where GCS score b 9 was identified as a risk factor for DNS [16]. For assessing the state of con- sciousness at the time of ED visit in patients with acute CO poisoning, the GCS score of 9 appears to be useful not only for predicting the occur- rence of DNS but also for predicting acute brain lesions on MRI.

There was no significant difference in the number of HBOT sessions before MRI between the two groups, although the time to the first HBOT after the last CO exposure was significantly longer in the acute brain le- sion group. This may be interpreted to infer that it is better to administer the first HBOT as soon as possible; however, this finding may have been the result of greater need for intervention before HBOT in patients who were in worse clinical condition. On multivariate analysis, acute-phase

Fig. 2. The receiver operating characteristic (ROC) curve for initial troponin-I level at presentation at emergency department.

Table 6 Symptoms of delayed neuropsychiatric sequelae and distribution of acute brain lesions on magnetic resonance imaging.

Funding

This study was supported by a research grant from Inha University

Patient number

Brain lesions Symptom of

DNS

(grant no. 60705-01).

#1 Only GP Cognitive

dysfunction

#2 Only GP Cognitive

dysfunction

#5 Only GP Cognitive

dysfunction

#17 Only GP Cognitive

dysfunction

#22 GP + PVWM Cognitive

dysfunction

CRediT authorship contribution statement

Jang Hyo Kim:Writing - original draft, Formal analysis, Investiga- tion.Areum Durey:Investigation, Data curation.Seung Baik Han:Inves- tigation, Data curation.Ji Hye Kim:Conceptualization, Methodology, Project administration, Supervision, Writing - review & editing.

#24 GP + PVWM, frontal, parietal, temporal, occipital, hypothalamus, pons, cerebellum

Cognitive dysfunction

Declaration of competing interest

#26 Cerebellum Anxiety and

insomnia

DNS: delayed neuropsychiatric sequelae, MRI: magnetic resonance imaging, GP: globus pallidus, PVWM: periventricular white matter.

brain MRI abnormalities and time to the first HBOT after the last CO ex- posure were unrelated.

Among laboratory findings, initial WBC, CRP, initial CK, and initial troponin-I levels at presentation in the ED were significantly higher in the acute brain lesion group. However, only initial troponin-I level remained an independent factor on multivariate analysis. A previous one study showed that N-terminal pro-brain natriuretic peptide, but not troponin-I, at presentation could predict long-term neurological outcome in acute charcoal-burning CO intoxication; however, we could not include the N-terminal pro-brain natriuretic peptide in our study due data unavailability [17]. Nonetheless, we believe that cardiac muscle damage from CO poisoning is associated with acute brain damage.

Our study has several limitations. First, due to insufficient records, several laboratory results and clinical information were analyzed after exclusion for missing data. Second, this study could not definitively es- tablish an independent relationship between acute brain lesions on MRI and DNS because of the relatively high rate of loss to follow-up and lack of uniform methods in DNS assessment of each patient. Third, there is a possibility of selective bias because not all HBOT-indicated patients underwent brain MRIs. The excluded patients rejected the MRI mostly for economic reasons. Nevertheless, when we compared the variables [gender, age, height, weight, comorbidity, intention of exposure, co- ingestion of alcohol or other drugs, duration of CO exposure, interval from last CO exposure to ED presentation, GCS score on initial ED pre- sentation (GCS score b9 and GCS score from 9 to 14), white blood cell (WBC) count, levels of COHb, hemoglobin (Hb), C-reactive protein (CRP), albumin, lactic acid, troponin-I, and creatine kinase (CK) at pre- sentation in the ED] between 103 patients included in the study and 95 patients who were excluded as they refused to take brain MRIs, gen- der, GCS score b 9, and initial troponin-I level were significantly differ- ent (P = 0.027, P = 0.038, P = 0.007, respectively; Supplementary Table S2). Among three variables, only initial troponin-I level was inde- pendently related to taking brain MRIs on multivariable logistic regres- sion analysis (OR 9.676, 95% CI 1.434-65.296, P = 0.020).

Conclusions

Acute-phase brain MRI in acute CO poisoning can be considered for patients who present to the ED with a GCS score b9 or troponin-I level N0.105 ng/mL.

Supplementary data to this article can be found online at https://doi. org/10.1016/j.ajem.2020.05.076.

None.

Acknowledgements

None.

References

1. Choi YR, Cha ES, Chang S Sen, Khang YH, Lee WJ. Suicide from carbon monoxide poi- soning in South Korea: 2006-2012. J Affect Disord. 2014;167:322-5.
2. Chang YT, Chang WN, Huang SH, Lui CC, Lee CC, Chen NC, et al. Neuroimaging stud- ies in carbon monoxide intoxication. Online referencing https://www.intechopen. com/books/neuroimaging-cognitive-and-clinical-neuroscience/neuroimaging- study-in-carbon-monoxide-intoxication; 2012. [Accessed 20 March 2020].
3. Kudo K, Otsuka K, Yagi J, Sanjo K, Koizumi N, Koeda A, et al. Predictors for Delayed encephalopathy following acute carbon monoxide poisoning. BMC Emerg Med. 2014;14:3. https://doi.org/10.1186/1471-227X-14-3.
4. Huang SH, Chang WN, Chen SF, Wang PW, Lui CC, Tu MC, et al. Tc99m-sestamibi thigh SPECT/CT images for noninvasive assessment of skeletal muscle injury in car- bon monoxide intoxication with clinical and pathological correlation. Clin Nucl Med. 2011;36:199-205.
5. Beppu T, Nishimoto H, Ishigaki D, Fujiwara S, Yoshida T, Oikawa H, et al. Assessment of damage to cerebral white matter fiber in the subacute phase after carbon monox- ide poisoning using fractional anisotropy in diffusion tensor imaging. Neuroradiol- ogy. 2010;52:735-43.
6. Parkinson RB, Hopkins RO, Cleavinger HB, Weaver LK, Victoroff J, Foley JF, et al. White matter hyperintensities and neuropsychological outcome following carbon monoxide poisoning. Neurology. 2002;58:1525-32.
7. Chu K, Jung KH, Kim HJ, Jeong SW, Kang DW, Roh JK. diffusion-weighted MRI and 99mTc-HMPAO SPECT in delayed relapsing type of carbon monoxide poisoning: ev- idence of delayed cytotoxic edema. Eur Neurol. 2004;51:98-103.
8. Weaver LK, Hopkins RO, Chan KJ, Churchill SS, Elliott CG, Clemmer TP, et al. Hyper- baric oxygen for acute carbon monoxide poisoning. N Engl J Med. 2002;347: 1057-67.
9. Rose JJ, Wang L, Xu Q, McTiernan CF, Shiva S, Tejero J, et al. Carbon monoxide poi- soning: pathogenesis, management, and future directions of therapy. Am J Respir Crit Care Med. 2017;195:596-606.
10. Oh S, Choi SC. Acute carbon monoxide poisoning and delayed neurological sequelae: a potential neuroprotection bundle therapy. Neural Regen Res. 2015;10:36-8.
11. Weaver LK, Valentine KJ, Hopkins RO. Carbon monoxide poisoning. Am J Respir Crit Care Med. 2007;176:491-7.
12. Jeon S, Sohn CH, Seo D, Oh BJ, Lim KS, Kang DW, et al. Acute brain lesions on mag- netic resonance imaging and delayed neurological sequelae in carbon monoxide poisoning. JAMA Neurol. 2018;75(4):436-43. https://doi.org/10.1001/jamaneurol. 2017.4618.
13. Moon GS, Kim GT. Acute carbon monoxide poisoning: prognostic factors for delayed neuro-psychiatric sequelae. J Korean Soc Emerg Med. 2014;25:268-76.
14. O’Donnell P, Buxton PJ, Pitkin A, Jarvis LJ. The magnetic resonance imaging appear- ances of the brain in acute carbon monoxide poisoning. Clin Radiol. 2000;55: 273-80.
15. Shono K, Satomi J, Tada Y, Kanematsu Y, Yamamoto N, Izumi Y, et al. Optimal timing of diffusion-weighted imaging to avoid false-negative findings in patients with tran- sient ischemic attack. Stroke. 2017;48:1990-2.
16. Liao SC, Mao YC, Yang KJ, Wang KC, Wu LY, Yang CC. Targeting optimal time for hy- perbaric oxygen therapy following carbon monoxide poisoning for prevention of de- layed neuropsychiatric sequelae: a retrospective study. J Neurol Sci. 2019;396: 187-92.
17. Moon JM, Chun BJ, Shin MH, Lee SD. Serum N-terminal proBNP, not troponin I, at presentation predicts long-term neurologic outcome in acute charcoal-burning car- bon monoxide intoxication. Clin Toxicol. 2018;56:412-20.