2025 Volume 50 Issue 2 Pages 69-73
Acute carbon monoxide poisoning (ACOP) is a cause of accidental or deliberate deaths worldwide. Subsequent complications, particularly delayed neurological sequelae (DNS), are preventable and treatable based on their pathophysiology. Hyperbaric oxygenation therapy (HBO) is a potential procedure for preventing and treating DNS; however, the effects of HBO on DNS are unclear and debated. In the present study, we investigated which factors are associated with the development of DNS and the effects of HBO in patients with ACOP. We performed retrospective subanalyses of the COP-J registry, focusing on adults who underwent HBO, regardless of whether they developed DNS. The multivariable analysis showed that the Glasgow coma scale (GCS) on admission was significantly associated with DNS (odds ratio 0.736; 95% confidence interval 0.608–0.892; P = 0.002). The receiver operating characteristic curve analysis of GCS for DNS revealed a cutoff value of 12.5 according to Youden’s index (sensitivity 80.8%, specificity 76.9%). This retrospective analysis of a nationwide Japanese registry of ACOP showed that low GCS scores on admission could be a predictive factor for DNS, with a possible cutoff value of ≤12, in patients who undergo HBO.
Acute carbon monoxide poisoning (ACOP) is a cause of accidental or deliberate deaths worldwide. In patients with ACOP, acute and subsequent neurological complications are important issues that need to be addressed to improve their long-term outcomes. These complications include altered mental status, confusion, syncope, and convulsion. Interestingly, the severity of complications is not correlated with the extent of hypoxia (Savioli et al., 2024). Subsequent complications, especially delayed neurological sequelae (DNS), occur in 1%‒47% of patients with ACOP. The clinical symptoms of DNS include mobility disorders, gait disturbance, autonomic dysfunction, seizure, and blindness. Based on the pathophysiology of these symptoms, they are potentially preventable and treatable (Savioli et al., 2024). Hyperbaric oxygenation therapy (HBO) is a potential procedure for the prevention and treatment of DNS. However, the effects of HBO on preventing DNS are unclear and controversial.
The effects of HBO on DNS were initially reported around 2000 (Scheinkestel et al., 1999; Weaver et al., 2002). However, subsequent systematic reviews and metanalyses did not find an advantage in using HBO to prevent DNS (Buckley et al., 2011; Ho et al., 2022). A Japanese nationwide registry, COP-J, was created to investigate the characteristics of patients with DNS, regardless of the use of HBO, and to identify factors related to the severity of ACOP (Fujita et al., 2021). In the present study, we performed subanalyses of the COP-J registry to investigate the factors associated with DNS after HBO in patients with ACOP.
We used data from the COP-J registry, which registered patients with ACOP who were admitted to 54 Japanese hospitals. The study was approved by the ethics committees at Yamaguchi University, Ube, Japan, the participating institutions, and each participating hospital. We retrieved data for patients registered for retrospective analyses, which was approved by the ethics committee at Fujita Health University, Toyoake, Japan.
PatientsData from a total of 311 patients with ACOP in the COP-J registry were analyzed. We retrieved data for patients aged >15 years old. Patients who did not receive HBO were excluded from the current analyses.
Study outcomes and statistical analysisIn this study, DNS was defined as neurological symptoms that began within 2 months after ACOP. The patients’ age, sex, duration of CO exposure, loss of consciousness at the scene, initial oxygen treatment by the emergency medical services, initial carboxyhemoglobin (COHb) at admission, and the Glasgow coma scale (GCS) at admission. We performed multivariable logistic regression analysis using DNS as the outcome and the following independent variables: age, sex, duration of CO exposure, loss of consciousness at the scene, initial oxygen treatment by the ambulance team, initial COHb at admission, and GCS at admission. Receiver operating characteristic curve (ROC) analysis for DNS was also performed using significant factors. Univariate analysis was performed by Mann-Whitney U test. In all analyses, P values of <0.05 were considered statistically significant. All statistical analyses were performed using SPSS version 25.0 (IBM, Armonk, NY, USA).
The registry consisted of 311 patients, of which 192 met the inclusion criteria (i.e., age >15 years and treatment with HBO; Fig. 1). Table 1 shows the characteristics of the 192 patients with ACOP who underwent HBO, and the results of the logistic regression analysis for DNS. DNS occurred in 13 patients (8%), the only significant factor for DNS was GCS on admission (odds ratio 0.736; 95% confidence interval 0.608–0.892; P = 0.002). Figure 2 and Table 2 show the ROC analysis of GCS on admission for DNS. Using Youden’s index, the cutoff value of GCS was 12.5, with a sensitivity of 80.8% and a specificity of 76.9%. Figure 3 show histogram of each GCS and number of DNS cases. Table 3 shows comparison of each GCS scores between DNS and non DNS cases, all scores (eye, verbal, and motor) are different significantly (P<0.001).
Patient disposition. A total of 311 patients with carbon monoxide poisoning were considered eligible after applying the eligibility criteria of age ≥15 years and hyperbaric oxygenation therapy.
| Patients (n = 192) | P value | OR (95% CI) | |
|---|---|---|---|
| Age (years) | 48 (31–63) | 0.347 | 1.021 (0.977–1.067) |
| Male (%) | 123 (64%) | 0.194 | 3.275 (0.546–19.656) |
| Duration of CO exposure (min) | 120 (60-240) | 0.605 | 1.000 (1.000–1.001) |
| Loss of consciousness at scene (%) | 80 (50%) | 0.379 | 2.437 (0.335–17.721) |
| Oxygen therapy by EMS (%) | 133 (90%) | 0.684 | 0.519 (0.022–12.236) |
| COHb on admission (%) | 18 (9-28) | 0.708 | 0.989 (0.932–1.049) |
| GCS | 15 (12-15) | 0.002 | 0.736 (0.608–0.892) |
| DNS (%) | 13 (8%) |
Values are median (interquartile range) or n (%) of cases.
OR odds ratio, CI confidence interval, EMS emergency medical services, COHb carboxyhemoglobin, GCS Glasgow coma scale, DNS delayed neurological sequelae
ORs were adjusted by logistic regression analysis as multivariate analysis.
Receiver operating characteristic curve analysis of the Glasgow coma scale on admission for delayed neurological sequelae. The area under the curve and the cutoff value are shown in Table 2.
| Cutoff value | AUC (95% CI) | Sensitivity | Specificity | Youden’s index | P value |
|---|---|---|---|---|---|
| GCS | 0.758 (0.618–0.898) | <0.001 | |||
| 12 | 80.8% | 76.9% | 0.577 | ||
| 12.5 | 80.8% | 76.9% | 0.577 | ||
| 13 | 76.2% | 76.9% | 0.531 |
Sensitivity and specificity for DNS were decided by ≤ cutoff value.
The cutoff value was determined using the largest Youden’s index.
AUC area under the curve, CI confidence interval, GCS Glasgow coma scale, DNS delayed neurological sequelae
Histogram of each Glasgow coma scale (GCS) and number of delayed neurological sequelae (DNS) cases.
| Patients (n = 192) | DNS (n = 13) | Non DNS (n = 179) | P value |
|---|---|---|---|
| GCS (eye) | 3 (2–4) | 4 (4–4) | <0.001 |
| GCS (verbal) | 1 (1–4) | 5 (4–5) | <0.001 |
| GCS (motor) | 4 (4–6) | 6 (6–6) | <0.001 |
Values are median (interquartile range).
GCS Glasgow coma scale, DNS delayed neurological sequelae
DNS generally occur about 40 days after CO exposure, and previously reported risk factors include age of >40 years, cardiovascular disease, prolonged CO exposure, electroencephalographic abnormalities, and coma (Savioli et al., 2024). In this study, we considered the factors of age, sex, duration of CO exposure, loss of consciousness at the scene, initial oxygen treatment by the ambulance team, initial COHb at admission, and GCS at admission as potential risk factors for DNS. Of these, GCS on admission was a significant risk factor for DNS, and the ROC analysis revealed a clear cutoff value of GCS based on the curve shape. These results suggested that HBO was potentially ineffective in patients with a low GCS score on admission.
Previous reports have described the relationship between GCS and DNS. In a retrospective observational study (n = 240; 19.6% received HBO), initial GCS was identified as a potential clinical predictor of DNS (Caballero-Bermejo et al., 2024). Additionally, prospective studies by Choi et al. (2024) (n = 167; 100% received HBO) and Yoo et al. (2023) (n = 1034; 93.9% received HBO) showed that GCS of <9 was a significant risk factor for DNS. Likewise, a meta-analysis of 10 studies (18% of patients received HBO) revealed that GCS of <11 was significant risk factors for DNS (Namgung et al., 2022). In our study, in which 100% of patients underwent HBO, GCS of ≤12 was a risk factor for DNS, and suggests the limited effectiveness of HBO for preventing DNS.
Some studies have also suggested that the interval from CO exposure to initial HBO could be a risk factor for DNS (Choi et al., 2024; Liu et al., 2024), which suggests that delayed HBO therapy could be a risk factor for DNS. The COP-J registry showed the interval from CO exposure to initial HBO was 330 min (median; interquartile range 208–444 min), however, this study investigated the factors on admission, further studies would be needed to address this question.
Finally, the present analyses revealed that GCS on admission is a risk factor for DNS, and that GCS of ≤12 may be predictive of DNS. Further studies are needed to recognize this high-risk subgroup correctly, and to determine the appropriate treatment strategy, such as how to perform HBO.
In conclusion, these retrospective analyses of a nationwide Japanese registry of patients with ACOP showed the possibility that, in patients who undergo HBO, GCS on admission could be predictive factor for DNS, with a cutoff value for GCS of ≤12.
We wish to thank all the members involved in the multicenter, observational COP-J registry. A list of institutions participating in the COP-J registry is available in the article by Fujita et al. (doi: org/10.1371/journal.pone.0253602).
Conflict of interestThe authors declare that there is no conflict of interest.
