Abstract
Delayed onset of neuropsychiatric symptoms after apparent recovery from acute carbon monoxide (CO) poisoning has been described as delayed neuropsychiatric sequelae (DNS). To date, there have been no studies on the utility of serum neuron-specific enolase (NSE), a marker of neuronal cell damage, as a predictive marker of DNS in acute CO poisoning. This retrospective observational study was performed on adult patients with acute CO poisoning consecutively treated over a 9-month period. Serum NSE was measured after emergency department arrival, and patients were divided into two groups. The DNS group comprised patients with delayed sequelae, while the non-DNS group included patients with none of these sequelae. A total of 98 patients with acute CO poisoning were enrolled in this study. DNS developed in eight patients. The median NSE value was significantly higher in the DNS group than in the non-DNS group. There was a statistical difference between the non-DNS group and the DNS group in terms of CO exposure time, Glasgow Coma Scale (GCS), loss of consciousness, creatinine kinase, and troponin I. GCS and NSE were the early predictors of development of DNS. The area under the curve according to the receiver operating characteristic curves of GCS, serum NSE, and GCS combined with serum NSE were 0.922, 0.836, and 0.969, respectively. In conclusion, initial GCS and NSE served as early predictors of development of DNS. Also, NSE might be a useful additional parameter that could improve the prediction accuracy of initial GCS.
Introduction
Carbon monoxide (CO) competes with oxygen for hemoglobin binding; the affinity of hemoglobin for CO is 250 times higher than that for oxygen, leading to a reduction in oxygen delivery to tissues and cellular hypoxia. 1 Therefore, CO poisoning can cause damage to organs that especially rely on oxygen such as the brain, heart, muscle, and kidney.
CO-induced brain damage can occur to varying degrees of severity. Among them, delayed onset of neuropsychiatric symptoms after apparent recovery from acute CO poisoning has been described as delayed neuropsychiatric sequelae (DNS). There are various symptoms and signs of DNS, including mental deterioration, cognitive dysfunction, amnesia, gait disturbance, mutism, urinary or fecal incontinence, psychosis, depression, and Parkinsonism. 2 –4 Some investigators have estimated that DNS occurs in 10–30% of victims, but the reported incidence varies widely. Choi reported that the lucid interval before appearance of neurologic sequelae is generally from 2 to 40 days, although various results have been reported in other studies. 2,5,6 Therefore, a major challenge for physicians is to identify patients who are likely to develop DNS. In particular, it would be helpful for clinicians if a laboratory test could predict DNS.
Among different isoenzymes of enolase, neuron-specific enolase (NSE) is a glycolytic enzyme, whose isoform is found in the neuronal cytoplasm of the central nervous system. It is assumed to be released from neuronal and glial tissue to the blood when cell membrane integrity is lost. Because NSE is not physiologically secreted, but rather is formed only when axons are damaged, serum NSE level can be used as a marker of neuronal cell damage in patients with a variety of conditions, including traumatic and hypoxic brain damage, status epilepticus, and cardiac arrest. 7 –11 Although several studies performed on adults have reported that serum NSE might act as an indicator of CO-induced hypoxic brain damage, 7,12 –14 studies on the usefulness of serum NSE as a marker to predict DNS in acute CO poisoning patients are lacking.
The aim of this study was to evaluate the ability of serum NSE measured at the emergency department (ED) to serve as a biochemical marker for early predicting DNS in acute CO poisoning.
Methods
Patients
This retrospective observational study was performed on adults >18 years of age with acute CO poisoning. Patients were consecutively treated over a 9-month period (from April 2015 to May 2016), and serum NSE was measured as part of the standard treatment protocol for CO poisoning during this period. The ED was located in a single, urban, tertiary-care hospital (Wonju, Republic of Korea), has more than 43,000 annual visits, and is staffed 24 h a day by board-certified emergency physicians. A diagnosis of CO poisoning was based on patient medical history and carboxyhemoglobin (CO-Hb) >5% (>10% in smokers). All patients presenting with acute CO poisoning upon arrival to the ED were treated with 100% high-flow oxygen therapy through a face mask with reservoir bag and hyperbaric oxygen therapy (HBOT), if indicated. 15,16 HBOT could not be performed in patients who were hemodynamic unstable, had uncontrolled irritability, or could not undergo equalization of the middle ear. Patients were educated on DNS at the time of hospital discharge and were encouraged to return to the hospital if they experienced DNS symptoms. Discharged patients were followed for at least 2 months based on Choi’s observation that the lucid interval is generally from 2 to 40 days. 2
Exclusion criteria for patients in this study were as follows: (1) presence of other conditions that could alter serum NSE level, including neural chest-derived tumors like small-cell lung cancer and neuroendocrine tumors 17,18 ; (2) failure for recover from decreased mental status (i.e., permanent neurologic injury); (3) previous neurocognitive dysfunction including dementia, psychiatric disease, or Parkinson disease; (4) failure to follow-up after discharge; (5) lack of data on serum NSE level; and (6) presentation more than 24 h after acute CO poisoning because the half-life of serum NSE is 24 h. 12
Study variables and definitions
Data were collected by retrospective review of electronic medical records of patients by two emergency physicians who were blinded to the study objectives and hypothesis. Evaluators were also blinded to the categorization of the patient groups and were trained prior to data collection in order to reduce possible bias from the data collection procedure. We used explicit case report forms in this study. Chart evaluators and study coordinators met periodically to resolve any disputes and to review coding rules. Study coordinators monitored the performance of the evaluators. This study was performed retrospectively and observationally, and patient records and information were anonymized prior to analysis.
We examined parameters that are meaningful in acute CO poisoning. The clinical parameters assessed were age; sex; source of CO; intentionality of poisoning; CO exposure time; time elapsed from rescue to arrival at the ED; history of smoking and co-ingestion of alcohol; use of HBOT; past medical history; initial Glasgow Coma Scale (GCS), vital signs, and symptoms and signs; and complications during admission such as rhabdomyolysis, 19 acute kidney injury (AKI), 20 and pneumonia. Laboratory parameters were levels of serum NSE, creatinine kinase (CK), and high-sensitivity troponin I (TnI) measured after ED arrival. Level of initial CO-Hb was checked from prehospital sources or obtained at our hospital. Serum NSE level (reference range: <16.3 ng/mL) was obtained using cobas® (Roche Diagnosis, Indianapolis, Indiana, USA) and was measured within 2 h after ED arrival.
Patients were divided into two groups. The DNS group comprised patients with delayed sequelae of mental deterioration, cognitive dysfunction, gait disturbance, mutism, urinary or fecal incontinence, psychosis, depression, and Parkinsonism, which were confirmed in follow-up. The non-DNS group included patients with none of these sequelae. This study was approved by the Institutional Review Board of Wonju College of Medicine, Yonsei University.
Study endpoints
The primary goals of this study were to compare serum NSE level measured at the ED associated with the presence of DNS and to investigate early predictors of DNS in acute CO poisoning.
Statistical analysis
Categorical variables are presented as frequencies and percentages, while continuous variables are presented as means and standard deviations or as medians and interquartile ranges. χ 2 or Fisher’s exact tests were used to compare categorical variables, while two-sample t- or Mann–Whitney U-tests were used to compare continuous variables. Normality was first assessed using the Shapiro–Wilk test. Logistic regression analysis was used to identify the factors to predict the DNS, and the results are expressed as odds ratio (OR) with 95% confidence interval (CI). Correlation analysis was performed with Spearman test. The area under the receiver operating characteristic (ROC) curve was used to evaluate the ability of serum NSE to determine the presence of DNS. In addition, comparison of the area under the curve (AUC) was used to compare the predictive ability of each method for the presence of DNS in acute CO poisoning. p-Values less than 0.05 were considered statistically significant, and all statistical analyses were conducted using SPSS Statistics for Windows version 23.0 (IBM, Armonk, New York, USA) and SAS version 9.2 (SAS Institute Inc., Cary, North Carolina, USA).
Results
Patient characteristics
A total of 139 consecutive adult patients with acute CO poisoning were identified during the study period. Forty-one patients were excluded based on the following criteria: failed to recover from neurologic complications including decreased mental function (6 patients); failed to follow-up (17 patients); presented more than 24 h after acute CO poisoning (4 patients); and did not have serum NSE measured (14 patients). After exclusion, a total of 98 patients with acute CO poisoning were enrolled in this study (Figure 1).
Flow diagram of patient selection. CO: carbon monoxide; NSE: neuron-specific enolase.
Baseline characteristics and laboratory findings are shown in Table 1. Of the 98 analyzed patients, 60 (61.2%) were men. The median age was 56 years (range, 18–83 years). Charcoal was the most common source of CO (62 patients, 63.3%), and intentionality was present in 34 patients (34.7%). The median CO exposure time was 3.0 h, and HBOT was performed in 87 patients (88.8%). Hypertension (25 patients, 25.5%) was the most common underlying disease. The most common symptoms and signs included headache (51 patients, 52.0%) and loss of consciousness (46 patients, 46.9%). Neurologic abnormalities including dysarthria, aphasia, motor weakness, and sensory change developed in 28 patients (28.6%). Of the complications during hospitalization, rhabdomyolysis (17 patients, 17.3%) was the most common. The median NSE and CO-Hb of the total patients were 21.5 ng/mL (reference range: <16.3 ng/mL) and 19.0%, respectively (Table 1).
Baseline characteristics and laboratory findings of patients with acute carbon monoxide poisoning.
DNS: delayed neuropsychiatric sequelae; LPG: liquefied petroleum gas; CO: carbon monoxide; ED: emergency department; HBOT: hyperbaric oxygen therapy; DM: diabetes mellitus; HTN: hypertension; GCS: Glasgow Coma Scale; SBP: systolic blood pressure; DBP: diastolic blood pressure; PR: pulse rate; RR: respiratory rate; BT: body temperature; MMSE: minimal mental status examination; NSE: neuron-specific enolase; CO-Hb: carboxyhemoglobin.
aMedian (interquartile range).
bCognitive dysfunction was assessed using the mini-mental status examination (MMSE) within 24 h after ED arrival in 84 (85.7%) of 98 patients.
Comparisons of general characteristics and laboratory findings based on the presence of DNS
DNS developed in eight patients (8.2%). Symptoms of DNS were mental deterioration, cognitive dysfunction, gait disturbance, and mutism. Comparisons of baseline characteristics and laboratory findings are shown in Table 1. CO exposure time and initial GCS significantly differed between patients in the non-DNS and DNS groups (3.0 vs. 10.5 h, p = 0.001 and 15 vs. 9 h, p < 0.001, respectively). Of the symptoms and complications evaluated in this study, loss of consciousness, rhabdomyolysis, AKI, and pneumonia were significantly more common in the DNS group. The median serum NSE value was significantly higher in the DNS group than in the non-DNS group (45.6 vs. 21.5 ng/mL, p = 0.002). The median serum CK and TnI level were also significantly higher in the DNS group than in the non-DNS group (740 vs. 124 U/L, p = 0.005 and 0.948 vs. 0.015 ng/mL, p < 0.001, respectively). Initial CO-Hb level did not differ significantly between groups. Although we analyzed with age, CO exposure time, GCS, CO-Hb, CK, and TnI, there was no variables showing significant correlation with level of serum NSE (Table 2).
The result of bivariate correlation between level of NSE and continuous variables in the study population.
NSE: neuron-specific enolase; CO: carbon monoxide; GCS: Glasgow Coma Scale; CO-Hb: carboxyhemoglobin.
Early predictors of DNS in acute CO poisoning
CO exposure time, initial GCS, loss of consciousness, serum NSE, CK, and TnI were analyzed by multiple logistic regression to identify predictors related to the development of DNS; initial GCS at the ED (OR, 3.336; 95% CI, 0.130–0.867, p = 0.024) and serum NSE (OR, 1.105; 95% CI, 1.019–1.199, p = 0.016) were found to be significant early predictors. Although the areas under the ROC curve for initial GCS and NSE for differentiating the DNS group from the non-DNS group were 0.922 (95% CI: 0.850–0.967) and 0.836 (95% CI: 0.748–0.903), respectively, there was no significant difference between GCS and NSE (p = 0.337). The AUC for early prediction of DNS using a combination of initial GCS and NSE (0.969 (95% CI: 0.913–0.994)) was higher than the AUC for GCS or NSE alone (Tables 3 and 4).
Predictors of delayed neuropsychiatric sequelae in patients with acute carbon monoxide poisoning as determined by multivariate logistic regression analysis.
OR: odds ratio; CI: confidence interval; CO: carbon monoxide; GCS: Glasgow Coma Scale; NSE: neuron-specific enolase.
Values of receiver operating characteristic curves for GCS and NSE for predicting delayed neuropsychiatric sequelae in patients with acute carbon monoxide poisoning.
GCS: Glasgow Coma Scale; NSE: neuron-specific enolase; AUC: area under the curve; CI: confidence interval.
Discussion
In this study, serum NSE (measured within 24 h of acute CO poisoning) was significantly higher in patients in the DNS group, and serum NSE was a significant early predictor of DNS. NSE release due to neuronal cell injury occurs in CO poisoning by two main mechanisms: (1) because of the high affinity of CO for hemoglobin, cells undergo hypoxia 1 and (2) CO exposure can cause inflammation through independent pathways, such as post-ischemic reperfusion injury, CO effects vascular endothelium, oxygen radical-mediated lipid peroxidation, and nitric oxide liberated from platelets at the time of CO exposure, all of which can culminate in neurologic injury. 3,21 –24 Because patients with DNS have more neuronal cell injury than those without DNS, patients in the DNS group would have higher NSE than those in the non-DNS group. Of all previous studies on NSE in CO-poisoned patients, none have evaluated serum NSE in terms of DNS. Previous studies have only compared patients with CO poisoning to healthy controls. Yardan et al. found that although NSE levels in patients with CO poisoning were similar to those of healthy controls, NSE levels were higher in unconscious CO poisoning patients than in conscious CO poisoning patients. 13 In contrast to studies conducted with adults, Akelma et al. reported that serum NSE levels of children with CO poisoning were significantly higher than those of healthy control children. Additionally, serum NSE level increased with CO-associated hypoxic brain damage in accordance with clinical findings in children. 14 In this study, the median serum NSE (a marker of neuronal and astroglial cell death) of all patients, regardless of the presence of DNS, was higher than the reported reference range. 7,8,10,12 –14 This suggests that CO itself causes neuronal cell injury.
In this study, we observed that serum NSE can also be early predictor DNS in the setting of acute CO poisoning. Park et al. previously reported that serum S100B protein is an independent predictor of DNS after acute CO poisoning and has been the only biochemical marker studied in DNS. 25 We thought that, because NSE has a higher molecular weight than S100B and a longer half-life (24 h vs. 30 min–2 h), an increased level of NSE might be more sensitive and reflect the ongoing destruction of neurons and the clinical status in acute CO poisoning with late admission to the ED. 10,12,26,27
In this study, initial GCS at the ED was significantly lower and loss of consciousness was significantly more frequent in the DNS group. Also, GCS was an early predictor of DNS in acute CO poisoning. Various mechanisms including hypoxic and inflammatory responses by CO can cause neuronal injury. 3 This may mean that CO poisoning, which affects consciousness in the acute phase, can more easily develop into DNS in the subacute or chronic phase. In a report by Choi, most cases of DNS were associated with loss of consciousness in the acute phase of CO poisoning. 2
Initial GCS combined with serum NSE had the highest diagnostic power in our study (Table 4). Even though initial serum NSE should not by itself replace careful clinical examination including GCS, we suggest that it could be a useful addition to improve the predictive accuracy of initial GCS for early prediction of DNS.
In this study, CO exposure time was significantly longer in the DNS group than in the non-DNS group. This may be because of increased neuronal damage with longer CO exposure time. It is known that CO exposure lasting more than 24 h is a risk factor of DNS in acute CO poisoning patients. 28,29 We also observed that CK and TnI were significantly elevated in the DNS group compared to the non-DNS group. Because the DNS group had a longer CO exposure time, more muscle and myocardial damage may have occurred, resulting in elevation of CK and TnI by CO-induced hypoxia and post-ischemic reperfusion injury. This is in line with the observation of greater incidence of rhabdomyolysis and AKI in the DNS group than in the non-DNS group. In addition, patients with DNS experienced more pneumonia complications than their non-DNS counterparts. We thought that aspiration pneumonia caused by decreased mental status might be more developed in the DNS group than in the non-DNS group because initial GCS at the ED was significantly lower in the DNS group.
This study had several limitations. First, it was limited by its retrospective design. Missing data during collection was also a limitation. Second, as this study was conducted at the emergency center of a single hospital, the sample size was small. However, we investigated all patients with acute CO poisoning and measured serum NSE in all applicable patients starting in April 2015, in order to reduce this possible bias. Third, because there were 17 patients lost to follow-up, the prevalence of DNS may have been affected. However, we think that the absolute patient number of the DNS group was probably not affected, as discharged patients would be likely to return to the hospital if they experienced DNS, as instructed at the time of hospital discharge. However, “prebiased” could be caused due to this instruction because patients may return to hospital if any subjective, perceived neurological symptoms occur. Fourth, the follow-up period was not the same in all patients. However, all patients were followed for at least 2 months in this study, and Choi has reported that the lucid interval before appearance of neurologic sequelae is generally from 2 to 40 days. 2 Fifth, we did not measure serial serum NSE values. Therefore, we could not investigate serial changes in the DNS group. A well-designed prospective study is necessary to clarify these limitations.
In conclusion, initial serum NSE, which was measured in the ED, was found to be significantly higher in the DNS group than in the non-DNS group, and initial GCS and NSE served as early predictors of development of DNS. Also, serum NSE might be a useful additional parameter that could improve the prediction accuracy of initial GCS.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.