Evaluating the Efficacy of Hyperbaric Oxygen Therapy for Acute Carbon Monoxide Poisoning: A Systematic Review and Meta‐Analysis

ABSTRACT

Aim

To reassess the efficacy of hyperbaric oxygen (HBO) therapy for acute carbon monoxide (CO) poisoning through a systematic review and meta‐analysis, including a subgroup analysis of treatment pressures exceeding 2.5 atm absolute (ATA).

Methods

We conducted a systematic review and meta‐analysis of randomized controlled trials (RCTs) evaluating HBO therapy for acute CO poisoning. A literature search was performed in MEDLINE, CENTRAL, and Ichushi‐Web databases, focusing on RCTs published up to June 2024. Only adult patients (≥ 18 years) were included, and studies were screened following PRISMA guidelines. Data extraction and quality assessment were conducted by two independent reviewers using the Cochrane risk of bias tool and the GRADE approach. Statistical analysis used a random effects model, with outcomes expressed as odds ratios (OR) and 95% confidence intervals (CI).

Results

Six studies were included, and no significant benefit of HBO therapy was observed in terms of reducing mortality or improving neurological outcomes. The subgroup analysis of HBO at ≥ 2.5 ATA also showed no significant advantage over control treatments. Moderate to significant heterogeneity was found across included studies, and the quality of evidence was rated as low to very low.

Conclusions

The efficacy of HBO therapy, even at ≥ 2.5 ATA, for improving outcomes in acute CO poisoning remains unproven. Despite these findings, HBO therapy may still hold potential benefits that require further exploration. High‐quality, multicenter RCTs are necessary to better define its role in the treatment of CO poisoning.

Trial Registration

UMIN Clinical Trials Registry: UMIN000054641

Meta‐analysis indicates that hyperbaric oxygen therapy (HBO) does not significantly improve mortality, neurological outcomes, or sequelae in acute conditions compared to standard care. These findings underscore the need for further high‐quality research to establish HBO's clinical utility.

1. Introduction

Since a 2002 report by Weaver et al. [1], hyperbaric oxygen (HBO) therapy has been considered effective in preventing delayed neuropsychiatric sequelae (DNS) following acute carbon monoxide (CO) poisoning. However, subsequent randomized controlled trials (RCTs) have provided inconsistent results regarding its efficacy, leaving the therapeutic value of HBO therapy for acute CO poisoning uncertain [2, 3]. Reflecting this uncertainty, the indications and treatment protocols for HBO therapy in acute CO poisoning vary considerably across Western countries, lacking standardization [4, 5]. Significant variability in the use of HBO therapy for CO poisoning across Europe has been reported, highlighting the inconsistency in treatment approaches [4]. Similarly, variability in HBO treatment protocols for acute CO poisoning in the United States has been documented, further emphasizing the lack of consensus [5].

Similarly, in Japan, the application and protocols of HBO therapy for acute CO poisoning are not well‐defined, leading to confusion in clinical practice. A multicenter retrospective survey revealed significant variability in the treatment of CO poisoning across Japan [6]. Moreover, a prospective observational study suggested the use of HBO therapy for preventing DNS, but the protocols varied significantly, contributing to clinical uncertainty [7]. Despite the absence of clear evidence, some experimental studies suggest that a treatment pressure of 2.5 atm absolute (ATA) or higher may be effective [8]. It has been suggested that the efficacy of HBO therapy, particularly at pressures of 2.5 ATA or higher, is likely linked to its ability to significantly reduce oxidative stress and lipid peroxidation, which are critical in mitigating neurological damage following CO poisoning [8]. However, its clinical effectiveness remains unclear and requires further validation.

In this study, we performed a systematic review and meta‐analysis to reassess the efficacy of HBO therapy for acute CO poisoning. Furthermore, we conducted a subgroup analysis to evaluate the effectiveness of treatment pressures exceeding 2.5 ATA.

2. Methods

2.1. Ethics and Approval

This systematic review and meta‐analysis protocol has been registered in the UMIN Clinical Trials Registry. The systematic review and meta‐analysis were reported in accordance with Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidelines [9], and do not require ethical approval.

2.2. Search Strategies

Database searches were conducted in MEDLINE (via PubMed), CENTRAL, and Ichushi‐Web (Japan Medical Abstracts Society) to retrieve relevant articles for the literature review. We searched for full‐text randomized controlled trials (RCTs) published up to June 2024. A combination of key terms was used to establish a full search strategy (Table S1).

2.3. Study Selection and Inclusion Criteria

Only RCTs were included. Conference proceedings, animal studies, and studies without sufficient data for outcome measurement were excluded. The study population of interest was adults (> 18 years old) with acute CO poisoning. There were no restrictions on the country or severity of poisoning. Only studies published in English or Japanese were included.

2.4. Data Extraction and Management

The following data were extracted: author(s), title, journal name, year of publication, and relevant outcomes. After removal of duplicates, two independent reviewers screened abstracts and titles and subsequently reviewed full‐text articles for inclusion using an electronic screening form. Disagreements were resolved through discussion or by a third reviewer. The flow diagram of the study selection process followed the PRISMA guidelines.

2.5. Assessment of Risk of Bias

The risk of bias will be independently assessed by two reviewers using the Revised Cochrane Risk of Bias Tool for Randomized Controlled Trials (RoB 2) [10], which evaluates five key domains: the randomization process, deviations from intended interventions, missing outcome data, outcome measurement, and selection of reported outcomes. For each domain, reviewers will answer signaling questions that guide classification into categories of “low risk,” “some concerns,” or “high risk.” The study's overall risk of bias will then be determined based on these domain‐level assessments. Discrepancies among reviewers will be resolved through discussion with a third reviewer.

2.6. Rating the Quality of Evidence Using the GRADE Approach

We used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) tool to rate the quality of evidence on the effect of hyperbaric oxygen therapy on important outcomes for patients with carbon monoxide poisoning [11]. The quality of evidence was assessed for each outcome and categorized as high, moderate, low, or very low using the GRADEpro Guideline Development Tool.

2.7. Statistical Analysis

We performed a meta‐analysis using a random effects model to pool estimates of treatment effects. Odds ratios (OR) and 95% confidence intervals (CI) were used for dichotomous outcomes, while mean differences and 95% CIs were used for continuous outcomes. Heterogeneity was evaluated using the I ² statistic, with significant heterogeneity defined as I ² greater than 50%. A funnel plot was used to assess the potential for publication bias. All analyses were performed using Review Manager software (RevMan 5.3).

3. Results

We identified 762 records from electronic databases, including PubMed (167), CENTRAL (578), and Ichushi‐Web (17). After removing 56 duplicates, 706 records were screened, of which 695 were excluded based on study design or lack of relevant data. Eleven full‐length reports were assessed for eligibility, and six studies were ultimately included in the final analysis (Figure 1) [1, 2, 12, 13, 14, 15].

FIGURE 1.

Flow diagram of study selection.

3.1. Study Characteristics

The six studies included in this meta‐analysis were randomized controlled trials evaluating the efficacy of HBO therapy for acute CO poisoning. The total number of participants across these studies was 1050, with patients assigned to either the HBO group or a control group. Study details, including patient characteristics, interventions, and evaluation timings, are summarized in Table 1.

TABLE 1.

Baseline characteristics of eligible studies.

Study	Year	Country	Design	Total patients	Entry criteria	Intervention	Control	Outcome evaluation timing
Raphael et al. [12]	1989	France	RCT	629	CO‐Hb ≥ 10% (smokers) or ≥ 5% (nonsmokers); admitted within 12 h postexposure	HBO (2.0 ATA for 2 h, 1–2 sessions)	NBO (100% O2 for 6 h)	1 month posttreatment
Ducassé et al. [13]	1995	France	RCT	26	Acute CO poisoning with no coma; admitted < 12 h postexposure	HBO (2.5 ATA, 2 h)	NBO (100% O2 for 6 h)	1 month posttreatment
Thom et al. [14]	1995	USA	RCT	65	Acute CO poisoning with confirmed symptoms (e.g., neurological dysfunction) within 6 h	HBO (2.8 ATA for 30 min, then 2.0 ATA)	NBO until symptoms resolved	4 weeks after exposure
Scheinkestel et al. [15]	1999	Australia	RCT	191	All patients with CO poisoning regardless of severity	HBO (2.8 ATA, 60 min/day, up to 6 days)	NBO (100% O2, 100 min)	At end of treatment
Weaver et al. [1]	2002	USA	RCT	152	Acute CO poisoning with symptoms such as loss of consciousness, confusion, metabolic acidosis, etc.	HBO (3 ATA for the first session, followed by 2 ATA for two additional sessions)	NBO (1 ATA, compressed air)	6 weeks, 6 months, 12 months
Annane et al. [2]	2010	France	RCT	179 (Trial A only)	Acute CO poisoning with loss of consciousness (coma excluded); CO‐Hb ≥ 10% (smokers) or ≥ 5% (nonsmokers); admitted within 12 h postexposure	HBO (2.0 ATA, 60 min)	NBO (6 h)	1 month after treatment

3.2. Overall Outcomes

The results of the overall analysis, including the quality of evidence for each outcome, are detailed in Table 2.

TABLE 2.

Summary of certainty of evidence and effect estimates for HBO therapy in acute carbon monoxide poisoning.

Certainty assessment							№ of patients		Effect		Certainty	Importance
No. of studies	Study design	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	HBO	NBO	Relative	Absolute
									(95% CI)	(95% CI)
Mortality
3	RCTs	Very serious a	Not serious	Not serious	Serious b	None	3/342 (0.9%)	2/309 (0.6%)	RR 1.25	2 more per 1000	⨁◯◯◯	Critical
									(0.21 to 7.34)	(from 5 fewer to 41 more)	Very low a , b
Favorable neurological outcome
3	RCTs	Serious c	Serious d	Not serious	Not serious	None	173/342 (50.6%)	168/321 (52.3%)	RR 0.99	5 fewer per 1000	⨁⨁◯◯	Critical
									(0.79 to 1.23)	(from 110 fewer to 120 more)	Low c , d
Delayed neurological sequalae
3	RCTs	Very serious e	Serious d	Not serious	Serious b	None	33/160 (20.6%)	33/158 (20.9%)	RR 0.94	13 fewer per 1000	⨁◯◯◯	Critical
									(0.12 to 7.53)	(from 184 fewer to 1000 more)	Very low b , d , e
Neuropsychiatric sequalae
5	RCTs	Serious g	Serious d	Not serious	Serious f	None	180/451 (39.9%)	181/429 (42.2%)	RR 0.90	42 fewer per 1000	⨁◯◯◯	Critical
									(0.65 to 1.25)	(from 148 fewer to 105 more)	Very low d , f , g

Abbreviations: CI, confidence interval; HBO, hyperbaric oxygen; NBO, normobaric oxygen; RCTs, randomized trials; RR, risk ratio.

^a A single high‐risk study contributed a substantial weight to the analysis.

^b The optimal information size criterion was not met.

^c One high‐risk study was included, but its weight was small; therefore, the rating was downgraded by one level.

^d Moderate heterogeneity was observed based on the I ² statistic.

^e Two high‐risk studies contributed a substantial weight to the analysis.

^f The confidence interval included a significant benefit.

^g Two high‐risk studies were included, but their combined weight was small; therefore, the rating was downgraded by one level.

3.3. Mortality

Three trials assessed mortality outcomes. The pooled analysis showed no statistically significant difference between the HBO and control groups (RR: 1.25; 95% CI: 0.21 to 7.34) (Figure 2A). This translates to an absolute effect of “2 more deaths per 1000 patients” (95% CI: 5 fewer to 41 more deaths). Heterogeneity was low (I ² = 0%), and the quality of evidence was rated as very low due to serious concerns about risk of bias and imprecision.

FIGURE 2.

Forest plot of the comparison: Control (NBO) vs. HBO for mortality (A), favorable neurological outcome (B), delayed neurological sequalae (C), and neuropsychiatric sequalae (D). CI, confidence intervals. The “Risk of Bias” is represented across six domains: (i) Randomization process, (ii) Deviations from intended interventions, (iii) Missing outcome data, (iv) Measurement of the outcome, (v) Selection of the reported result, (vi) Overall risk of bias. In the figure, these domain headers are displayed as A–F on the right side of the forest plot and are indicated by color codes: Green (+) for low risk, yellow (?) for unclear risk, and red (−) for high risk.

3.4. Favorable Neurological Outcomes

Three studies evaluated favorable neurological outcomes. The pooled RR was 0.99 (95% CI: 0.79 to 1.23), suggesting no significant benefit of HBO therapy (Figure 2B). This corresponds to “5 fewer patients per 1000 achieving favorable outcomes” (95% CI: 110 fewer to 120 more). Moderate heterogeneity was observed (I ² = 46%), and the quality of evidence was rated as low due to inconsistency and risk of bias.

3.5. Delayed Neurological Sequelae (DNS)

Delayed neurological sequelae were assessed in three trials. The pooled RR was 0.94 (95% CI: 0.12 to 7.53), indicating no significant difference between groups (Figure 2C). This translates to an absolute effect of “13 fewer DNS cases per 1000 patients” (95% CI: 184 fewer to 1000 more cases). Heterogeneity was moderate (I ² = 68%), and the evidence quality was very low, primarily due to serious risk of bias and imprecision.

3.6. Neuropsychiatric Sequelae

Five trials assessed neuropsychiatric sequelae, with a pooled RR of 0.90 (95% CI: 0.65 to 1.25), showing no statistically significant difference (Figure 2D). This corresponds to “42 fewer cases per 1000 patients” (95% CI: 148 fewer to 105 more). Heterogeneity was high (I ² = 74%), and the quality of evidence was rated as very low, largely due to high risk of bias and inconsistency.

3.7. Subgroup Analysis for HBO Therapy at ≥ 2.5 ATA

3.7.1. Mortality

One trial analyzed mortality outcomes for HBO therapy at ≥ 2.5 ATA, with no statistically significant difference compared to control (RR: 1.25; 95% CI: 0.21 to 7.34) (Figure 3A). This translates to an absolute effect of “6 more deaths per 1000 patients” (95% CI: 18 fewer to 146 more deaths). Heterogeneity was not applicable. The quality of evidence was very low due to imprecision and risk of bias.

FIGURE 3.

Forest plot of the subgroup analysis for Hyperbaric Oxygen Therapy at ≥ 2.5 ATA, showing comparisons for mortality (A), favorable neurological outcome (B), delayed neurological sequelae (C), and neuropsychiatric sequelae (D). CI, confidence intervals. The “Risk of Bias” is represented across six domains: (i) Randomization process, (ii) Deviations from intended interventions, (iii) Missing outcome data, (iv) Measurement of the outcome, (v) Selection of the reported result, (vi) Overall risk of bias. In the figure, these domain headers are displayed as A–F on the right side of the forest plot and are indicated by color codes: Green (+) for low risk, yellow (?) for unclear risk, and red (−) for high risk.

3.7.2. Favorable Neurological Outcomes

For favorable neurological outcomes, one trial reported an RR of 0.64 (95% CI: 0.38 to 1.07), suggesting no significant benefit of HBO therapy (Figure 3B). “104 fewer patients per 1000 achieving favorable outcomes” (95% CI: 178 fewer to 20 more). Heterogeneity was moderate (I ² = 45%), and the quality of evidence was rated as low.

3.7.3. Delayed Neurological Sequelae (DNS)

Two trials assessed DNS, showing no significant reduction in risk (RR: 0.77; 95% CI: 0.01 to 99.14) (Figure 3C). This translates to an absolute effect of “22 fewer DNS cases per 1000 patients” (95% CI: 96 fewer to 1000 more; the extreme width of the confidence interval makes the upper bound difficult to interpret meaningfully). Heterogeneity was high (I ² = 83%). The quality of evidence was very low due to high risk of bias and inconsistency.

3.7.4. Neuropsychiatric Sequelae

Three trials examined neuropsychiatric sequelae, with a pooled RR of 0.64 (95% CI: 0.28 to 1.50) (Figure 3D). Heterogeneity was substantial (I ² = 86%). This corresponds to “188 fewer cases per 1000 patients” (95% CI: 377 fewer to 262 more). The evidence quality was rated as very low, primarily due to imprecision and inconsistency.

4. Discussion

Our meta‐analysis did not demonstrate the efficacy of HBO therapy for acute CO poisoning, even at treatment pressures of ≥ 2.5 ATA. Several factors may contribute to these findings. First, the number of RCTs included in the analysis was limited, reducing the statistical power of our conclusions. Second, the quality of the available RCTs was generally low, partly because many of these trials were conducted years ago and lacked the rigorous standards of modern clinical research. Third, variability in study designs, particularly in terms of inclusion criteria, treatment protocols, and outcome definitions, introduced heterogeneity into the analysis, further complicating the interpretation of findings. Moreover, the methods used to assess outcomes varied significantly across studies, underscoring the need for more standardized approaches in future research.

A significant limitation of previous meta‐analyses assessing HBO therapy for CO poisoning, such as that by Buckley et al. [3], lies in their exclusive reliance on neuropsychiatric sequelae as the sole outcome measure. Although these analyses have provided valuable insights, their narrow focus restricts the comprehensiveness of their conclusions. Specifically, clinically meaningful outcomes such as mortality and full recovery without sequelae have not been adequately explored. This limitation may stem from the challenges of measuring long‐term outcomes in heterogeneous patient populations and the availability of data. Additionally, inconsistencies in the definitions and assessment methods for neuropsychiatric sequelae further diminish the generalizability of these findings. In contrast, Lin et al. conducted a meta‐analysis of the same six RCTs and reported that patients who received HBO therapy had lower frequencies of headache, memory impairment, difficulty concentrating, disturbed sleep, and delayed neurological sequelae compared with those who received NBO therapy [16]. However, while their conclusions suggested a benefit of HBO, their pooled analyses did not reach statistical significance, indicating only a nonsignificant trend. Moreover, the outcomes assessed in their study differed from ours, as they focused on individual neuropsychiatric symptoms rather than clinically meaningful composite outcomes such as mortality, DNS, or favorable neurological recovery. Taken together, these differences in outcome selection, along with the lack of statistical significance in Lin's results, suggest that prior meta‐analyses have not provided sufficient evidence to establish the superiority of HBO therapy over NBO in the management of acute CO poisoning.

To address these limitations, our meta‐analysis expanded the scope of outcome measures to include mortality, delayed neurological sequelae (DNS), and favorable neurological outcomes. This broader approach provides a more comprehensive evaluation of the potential benefits and limitations of HBO therapy, offering insights directly relevant to patient care. However, it remains a significant limitation that none of the included RCTs defined mortality or full recovery without sequelae as primary outcomes [1, 2, 12, 13, 14, 15, 17]. Most of these studies prioritized surrogate or intermediate outcomes, likely reflecting the logistical challenges of conducting long‐term follow‐up studies and the complexities inherent in CO poisoning. Furthermore, the exclusion of extreme cases—those at high risk of death or with mild symptoms not requiring treatment—may have introduced a bias toward moderate cases, limiting the generalizability of the findings.

Another critical issue identified in this meta‐analysis is the variability in treatment protocols, including differences in treatment pressures and the number of HBO sessions across studies [1, 2, 12, 13, 14, 15, 17]. This variability reflects the lack of consensus on standardized HBO protocols for CO poisoning and complicates the interpretation of findings. Establishing an optimal protocol to prevent long‐term neuropsychiatric sequelae in CO poisoning is a critical area for future research. Although previous RCTs conducted at pressures below 2.5 ATA have not demonstrated the superiority of HBO therapy [2, 12], our subgroup analysis focusing on pressures ≥ 2.5 ATA suggested a potential benefit in preventing neuropsychiatric sequelae. Although the results were not statistically significant, this analysis underscores the need for further investigation into whether higher pressures might provide greater benefits for specific patient subgroups.

A fundamental challenge in evaluating the efficacy of HBO therapy lies in the inconsistent definitions and assessment methods for neuropsychiatric sequelae and DNS. Standardizing these definitions and evaluation protocols is crucial to improving the reliability and comparability of findings. For instance, discrepancies in reported rates of neuropsychiatric sequelae, such as those observed between a 2002 RCT and a subsequent RCT conducted by the same group [1, 17], highlight the urgent need for clear and consistent outcome measures. Addressing these inconsistencies will enhance the accuracy of future research findings.

Furthermore, the feasibility of implementing HBO therapy, including its economic impact, is another critical consideration that remains unexplored. The cost of HBO equipment, the need for specialized facilities, and the availability of trained personnel could all influence the practicality of widespread adoption. Future studies should evaluate not only the clinical efficacy of HBO therapy but also its cost‐effectiveness and feasibility in diverse healthcare settings. Incorporating these factors into future research will provide a more comprehensive understanding of its potential role in clinical practice.

Overall, while the current body of evidence suggests that HBO therapy may help prevent neuropsychiatric sequelae, including DNS, the lack of high‐quality RCTs precludes definitive conclusions. Future research should prioritize well‐designed, multicenter RCTs that incorporate standardized protocols and clinically meaningful endpoints, such as survival rates and full recovery without sequelae. Long‐term follow‐up studies will also be essential to fully capture the potential benefits and limitations of HBO therapy. By addressing these gaps, future studies can establish more reliable evidence to clarify the role of HBO therapy in the management of acute CO poisoning.

The strengths of this meta‐analysis include the use of the GRADE approach, which provides an objective framework for evaluating the quality of evidence in systematic reviews and clinical practice guidelines [11]. Another strength is the inclusion of a comprehensive subgroup analysis focusing on higher treatment pressures (≥ 2.5 ATA), which adds depth to the understanding of HBO therapy's potential efficacy. However, our study has several limitations. The relatively small number of included RCTs, combined with variability in treatment protocols, limits the ability to draw definitive conclusions. Additionally, while HBO therapy at ≥ 2.5 ATA is thought to be more effective, the evidence supporting this claim remains of very low quality due to the limited number of studies and inconsistencies in reported outcomes. Future research must address these gaps by conducting well‐designed, multicenter RCTs with standardized protocols to better determine the efficacy of HBO therapy for acute CO poisoning. In addition, evaluating the feasibility of HBO therapy, including cost‐effectiveness and resource availability, will be crucial to ensuring its practical implementation in clinical settings.

5. Conclusions

This meta‐analysis did not demonstrate a significant benefit of HBO therapy, even at treatment pressures of ≥ 2.5 ATA, for improving outcomes in acute CO poisoning. However, given the critical need for effective treatments, HBO therapy remains a potential option that warrants further investigation. Future high‐quality multicenter RCTs are needed to better clarify its role.

Funding

This study was funded by the Japan Resuscitation Council (JRC) and the Japanese Congress on Neurological Emergencies (JNE), the Japanese Society of Intensive Care Medicine (JSICM), and the Japan Society of Neurosurgical Emergency as Member Societies of the JRC.

Ethics Statement

The authors have nothing to report.

Conflicts of Interest

Dr. Shoji Yokobori is an Editorial Board member of AMS Journal and a co‐author of this article. To minimize bias, he was excluded from all editorial decision‐making related to the acceptance of this article for publication. The remaining authors declare no conflicts of interest for this article.

Supporting information

Table S1: (A) Search Strategy for PubMed (June 13, 2024). (B) Search Strategy for CENTRAL (June 14, 2024). (C) Search Strategy for Ichushi‐Web (June 4, 2024).

Fujita M., Todani M., Ajimi Y., et al., “Evaluating the Efficacy of Hyperbaric Oxygen Therapy for Acute Carbon Monoxide Poisoning: A Systematic Review and Meta‐Analysis,” Acute Medicine & Surgery 13, no. 1 (2026): e70114, 10.1002/ams2.70114.

Data Availability Statement

The datasets used and analyzed during the current study are available from the corresponding author upon reasonable request.