Authors: Ömer Yüceer
Categories: Original Research, carbon monoxide poisoning, hyperbaric oxygen therapy, carboxyhemoglobin, lactate, troponin, Glasgow coma scale
Source: International Journal of General Medicine
Doi: 10.2147/IJGM.S604040
Authors: Ömer Yüceer
Carbon monoxide (CO) poisoning remains a major global cause of morbidity and mortality. Although normobaric oxygen therapy (NBOT) is the standard initial treatment, hyperbaric oxygen therapy (HBOT) is often considered in patients with severe neurological or cardiac involvement. This study aimed to evaluate clinical, biochemical, electrocardiographic, and imaging factors associated with the selection of HBOT in patients presenting to the emergency department with CO poisoning.
This retrospective cohort study included 272 adult patients diagnosed with CO poisoning between January 2020 and January 2025. Of these, 169 patients (62.1%) received NBOT and 103 patients (37.9%) received HBOT. Demographic characteristics, laboratory parameters (COHb, lactate, troponin, creatine kinase-MB [CK-MB]), Glasgow Coma Scale (GCS) scores, electrocardiographic (ECG) findings, and magnetic resonance imaging (MRI) results were analyzed. Multivariable logistic regression and receiver operating characteristic (ROC) analyses were performed.
Patients receiving HBOT had significantly lower GCS scores, higher cardiac biomarkers, and more frequent ischemic ECG findings. In multivariable analysis, lower GCS (OR 0.66, 95% CI 0.57–0.76, p<0.001), ischemic ECG findings (OR 7.31, 95% CI 1.56–34.30, p=0.012), and lactate levels (OR 0.81, 95% CI 0.65–0.99, p=0.043) were independently associated with HBOT selection. Lactate and COHb demonstrated limited discriminative ability (AUC 0.554 and 0.599, respectively).
HBOT selection in CO poisoning is primarily associated with neurological status and cardiac involvement rather than biochemical parameters alone. These findings support a multidimensional clinical approach in emergency settings.
The most prevalent sources of carbon monoxide (CO), an odorless, colorless, and tasteless gas, are heating-related sources such coal stoves, fires, and water heaters.1,2 Carbon monoxide poisoning accounts for approximately 8–34% of all poisoning cases worldwide.3 Diagnosis is based on carboxyhemoglobin (COHb) levels in combination with clinical findings, which may range from mild headache to coma and death.4 Clinical manifestations and fatal outcomes are primarily associated with inhalation of CO through the respiratory tract from smoke or faulty heating systems.5–7
About 10–15% of breathed CO binds to myoglobin and cytochrome oxidase enzymes, despite the lungs eliminating the majority of it. Cellular hypoxia results from carbon monoxide’s approximately 250-fold stronger affinity for haemoglobin than oxygen, which forms COHb and significantly reduces the blood’s capacity to carry oxygen.7 Particularly in organs with high oxygen demands including the heart, brain, and skeletal muscle, impaired oxygen supply may result in lactate buildup and increases in creatine kinase and troponin levels.7,8 Fatal outcomes and serious systemic problems could result from these pathophysiological effects.9,10 Clinical manifestations of severe CO poisoning can vary from relatively minor symptoms like headache, nausea, and dizziness to dyspnea, chest discomfort, unconsciousness, and death.5 Patients with severe CO poisoning should receive hyperbaric oxygen treatment (HBOT).11 Under room air circumstances, the half-life of COHb is roughly 3–4 hours; with normobaric oxygen therapy (NBOT) and HBOT provided at 2.5 atmospheres absolute, this time is shortened to 30–90 minutes and 15–25 minutes, respectively.11,12 HBOT may be started without preceding NBOT in patients who arrive with loss of consciousness, cardiac involvement, or severe respiratory compromise.13 Despite a large body of research on CO poisoning, there is still a dearth of information defining precise indications for HBOT, especially in severe instances.12,13 This study aimed to identify factors associated with the selection of HBOT in patients with carbon monoxide poisoning.
Patients 18 years of age and older who arrived at the emergency room between January 2020 and January 2025 with symptoms of carbon monoxide (CO) poisoning and needed hospitalization were included in this retrospective analysis. The study was carried out in compliance with the 2013 edition of the Declaration of Helsinki.
Medical records were retrospectively reviewed using the hospital information management system. Discharge summaries, laboratory parameters like COHb, lactate, troponin T, and creatine kinase–MB (CK-MB) levels, ECG findings suggestive of ischemia (ST-segment elevation, ST-segment depression, and T-wave inversion), and ischemic findings detected on MRI were among the data extracted.
Demographic and clinical variables recorded were age, sex, mode of hospital admission (self-presentation or ambulance transport), time of admission (day or night), and seasonal distribution. For analytical purposes, the calendar year was divided into four 21 December–20 March (winter), 21 March–20 June (spring), 21 June–22 September (summer), and 23 September–20 December (autumn). The source of poisoning (stove, water heater, fire, exhaust, etc.) and presenting symptoms were also documented.
Lactate levels were categorized as >1.6 mmol/L or ≤1.6 mmol/L, and COHb levels as >20% or ≤20%. Treatment modalities were classified as NBOT or HBOT. Lactate levels were categorized as >1.6 mmol/L or ≤1.6 mmol/L, and COHb levels as >20% or ≤20%.
The cut-off values for lactate (>1.6 mmol/L) and COHb (>20%) were selected based on previously reported clinical thresholds in the literature.
The standard initial management for CO poisoning in the emergency department consisted of administering 100% normobaric oxygen until symptom resolution. As no HBOT unit was available at the study center, patients requiring HBOT were referred to the nearest facility, Kayseri City Hospital. At that center, patients received a single 2-hour HBOT session at 2.5 atmospheres absolute, after which they were transferred back to Niğde Ömer Halisdemir University Emergency Department.
Indications for HBOT included respiratory distress; cardiovascular and/or neurological symptoms; elevated COHb, troponin T, CK-MB, or lactate levels; ischemic ECG changes; and ischemic findings on MRI. Pregnant patients, individuals with chronic renal or hepatic disease, patients with concurrent trauma, and asymptomatic patients who did not require hospitalization were excluded from the study.
The Shapiro–Wilk test was used to analyze the distributional characteristics of continuous variables. While qualitative variables are summarized as counts and percentages, quantitative variables are provided as mean ± standard deviation (SD), along with their minimum and maximum values. The independent samples t-test was used to compare two independent groups, and the chi-square test was used to examine relationships between categorical variables. Fisher’s exact test was used when more than 25% of the anticipated cell counts were less than five.
Pearson’s correlation coefficient was used to investigate relationships between continuous variables.
IBM SPSS Statistics (version 22; SPSS Inc., Chicago, IL, USA) was used to conduct statistical analyses. Statistical significance was defined as a two-tailed p-value of less than 0.05. Lactate and COHb levels were assessed using receiver operating characteristic (ROC) curve analysis, which yielded the area under the curve (AUC), ideal threshold values, sensitivity, and specificity. MedCalc software (version 22.018; MedCalc Software, Ostend, Belgium) was used to perform ROC analyses. A post hoc power analysis was performed based on the primary outcome, demonstrating that the study had sufficient statistical power (>80%) to detect clinically meaningful differences. Variables with a p-value <0.10 in univariate analyses and those considered clinically relevant were included in the multivariable logistic regression model to identify factors associated with HBOT selection.
Variables that were not included in the multivariable model were excluded due to lack of statistical significance or potential multicollinearity. Cases with missing data were excluded from the respective analyses, and the proportion of missing data was low (<5%) and unlikely to affect the overall results.
The Çukurova University Non-Interventional Clinical Research Ethics Committee granted approval for this study (Approval No: 09.01.2026/162) before the data were extracted and analyzed. The study was carried out in compliance with the 2013 edition of the Declaration of Helsinki. The ethics committee disregarded the need for informed consent because of the retrospective design and use of anonymised data.
A total of 272 patients were included in the study, of whom 103 (37.9%) received HBOT and 169 (62.1%) received NBOT. Comparisons between the HBOT and NBOT groups were performed to identify differences in clinical, laboratory, and electrocardiographic parameters.
The baseline demographic characteristics and temporal distribution of emergency department admissions are summarized in Table 1.Table 1Baseline Characteristics (Overall)SectionCategoryn%GenderFemale17464.0GenderMale9836.0SeasonWinter16761.4SeasonAutumn4717.3SeasonSpring4616.9SeasonSummer124.4Time00:01–06:0015055.1Time06:01–12:003211.8Time12:01–18:00207.4Time18:01–00:007025.7
Female patients constituted the majority of the study population (64.0%), and most admissions occurred during the winter season (61.4%). Regarding the time of presentation, more than half of the patients were admitted between 01 and 00 (55.1%). The distribution of carbon monoxide exposure sources is presented in Table 2.Table 2Source of Exposure (Overall)Exposure Sourcen%Stove23787.1Combi boiler2810.3Fire72.6
The most common source of exposure was stove-related poisoning (87.1%), followed by combi boiler exposure (10.3%) and fire-related poisoning (2.6%). Comparisons of categorical clinical variables between HBOT and NBOT groups are summarized in Table 3.Table 3Categorical Variables by Treatment GroupVariableHBOT n (%)NBOT n (%)Total n (%)pIn-hospital mortality22 (21.4)12 (7.1)34 (12.5)<0.001Intubation34 (33.0)17 (10.1)51 (18.8)<0.001Any comorbidity40 (38.8)10 (5.9)50 (18.4)<0.001COPD11 (10.7)4 (2.4)15 (5.5)0.004Ischemia on ECG35 (34.0)3 (1.8)38 (14.0)<0.001Ischemia on MRI7 (6.8)12 (7.1)19 (7.0)0.924COHb >20%63 (61.2)67 (39.6)130 (47.8)<0.001
In-hospital mortality was significantly higher among patients receiving HBOT compared with those receiving NBOT (21.4% vs 7.1%, p<0.001). Similarly, intubation rates were significantly greater in the HBOT group (33.0% vs 10.1%, p<0.001). The prevalence of comorbidities (38.8% vs 5.9%, p<0.001) and chronic obstructive pulmonary disease (10.7% vs 2.4%, p=0.004) was also higher in patients treated with HBOT. In addition, ischemic electrocardiographic findings were significantly more frequent in the HBOT group (34.0% vs 1.8%, p<0.001). The proportion of patients with COHb levels greater than 20% was significantly higher in the HBOT group (61.2% vs 39.6%, p<0.001). However, ischemic findings detected on MRI did not differ significantly between the two treatment groups (p=0.924). Continuous demographic, clinical, and laboratory variables are presented in Table 4.Table 4Continuous Variables by Treatment GroupVariableHBOT Mean±SDNBOT Mean±SDTotal Mean±SDpAge (years)54.83±16.9449.52±15.8151.53±16.420.011GCS11.37±3.4014.26±1.4413.17±2.76<0.001Lactate2.64±1.952.40±1.922.49±1.930.319COHb25.26±8.2722.83±8.3123.75±8.360.020Troponin29.38±37.479.92±9.4217.29±25.94<0.001CK-MB4.57±4.472.01±1.362.98±3.20<0.001pH7.28±0.157.34±0.117.32±0.13<0.001Length of stay (days)1.31±1.700.29±0.970.68±1.38<0.001
Patients receiving HBOT were significantly older than those receiving NBOT (54.83±16.94 vs 49.52±15.81 years, p=0.011). The HBOT group also had significantly lower Glasgow Coma Scale (GCS) scores (11.37±3.40 vs 14.26±1.44, p<0.001). Laboratory analyses demonstrated significantly higher COHb levels (25.26±8.27 vs 22.83±8.31, p=0.020) and markedly elevated cardiac biomarkers, including troponin (29.38±37.47 vs 9.92±9.42, p<0.001) and CK-MB (4.57±4.47 vs 2.01±1.36, p<0.001), among patients treated with HBOT. In addition, arterial pH values were significantly lower in the HBOT group (7.28±0.15 vs 7.34±0.11, p<0.001). The length of hospital stay was also significantly longer in patients receiving HBOT (1.31±1.70 vs 0.29±0.97 days, p<0.001). Although lactate levels were numerically higher in the HBOT group, this difference did not reach statistical significance (p=0.319). Multivariable logistic regression analysis was performed to identify independent factors associated with HBOT selection. The results are presented in Table 5.Table 5Multivariable Logistic Regression for HBOT SelectionPredictorOR (95% CI)pGCS0.66 (0.57–0.76)<0.001COHb1.04 (0.99–1.09)0.100Lactate0.81 (0.65–0.99)0.043Troponin1.02 (0.99–1.04)0.154Ischemic ECG7.31 (1.56–34.30)0.012
Lower GCS scores were independently associated with HBOT selection (OR 0.66, 95% CI 0.57–0.76, p<0.001). In addition, the presence of ischemic ECG findings significantly increased the likelihood of HBOT administration (OR 7.31, 95% CI 1.56–34.30, p=0.012). Lactate level was also identified as an independent predictor (OR 0.81, 95% CI 0.65–0.99, p=0.043). However, COHb and troponin levels were not independently associated with HBOT selection after multivariable adjustment. Receiver operating characteristic (ROC) curve analysis was conducted to assess the discriminative performance of lactate and COHb levels for predicting HBOT selection, and the results are summarized in Table 6.Table 6Receiver Operating Characteristic (ROC) Analysis for Predicting HBOT SelectionParameterAUC95% CICut-offSensitivitySpecificitypLactate0.5540.49–0.62>2.1 mmol/L58.251.30.12COHb0.5990.53–0.66>23%61.455.00.03
Lactate demonstrated poor discriminative ability with an area under the curve (AUC) of 0.554 (95% CI 0.49–0.62). Similarly, COHb showed limited predictive performance with an AUC of 0.599 (95% CI 0.53–0.66). The optimal cut-off value for lactate was >2.1 mmol/L, yielding a sensitivity of 58.2% and a specificity of 51.3%, whereas the optimal cut-off for COHb was >23%, with a sensitivity of 61.4% and a specificity of 55.0%. ROC curves for lactate and COHb are illustrated in Figures 1 and 2, respectively. Figure 1Lactate ROC curve for predicting HBOT selection.A line graph titled Lactate. The x-axis label is 100-Specificity, ranging from 0 to 100. The y-axis label is Sensitivity, ranging from 0 to 100. A diagonal reference line runs from (0, 0) to (100, 100). A lactate receiver operating characteristic curve starts near (0, 0), rises to about (5, 5), about (10, 10), about (15, 18), about (20, 30), about (30, 35), about (50, 60), about (58, 62), about (62, 70), about (80, 85) and ends at (100, 100). A boxed annotation reads AUC = 0,554 and P = 0,0132.A line graph showing a lactate receiver operating characteristic curve for predicting HBOT selection. Figure 2COHb ROC curve for predicting HBOT selection.A line graph titled COHb. The x-axis label is 100-Specificity and the unit is percent, ranging from 0 to 100. The y-axis label is Sensitivity and the unit is percent, ranging from 0 to 100. One receiver operating characteristic curve is plotted along with a diagonal reference line from (0, 0) to (100, 100). The receiver operating characteristic curve starts at (0, 0), rises to about (10, 8), (15, 18), (20, 28), (25, 40), (30, 50), (35, 55), (40, 60), (50, 70), (60, 74), (70, 80), (80, 86), (90, 92) and ends at (100, 100). A text box states AUC = 0,599 and P = 0,0052.A line graph showing a COHb receiver operating characteristic curve.
Overall, these findings suggest that although certain laboratory parameters are associated with clinical severity, their standalone predictive performance for HBOT selection remains limited.
Carbon monoxide (CO) poisoning remains a significant clinical problem worldwide and continues to present diagnostic and therapeutic challenges in emergency medicine.2,5 One of the major difficulties in management is the lack of universally standardized criteria for HBOT, particularly in patients with heterogeneous clinical presentations.11–13 In this cohort of hospitalized patients, more than one-third required HBOT, reflecting a substantial proportion of clinically severe cases. HBOT selection is primarily associated with neurological impairment and cardiac involvement rather than isolated biochemical abnormalities. Patients receiving HBOT demonstrated markedly lower GCS scores, and neurological status emerged as an independent factor associated with HBOT selection. These results are consistent with previous literature emphasizing the central role of cerebral hypoxia in CO toxicity and the importance of neurological assessment in treatment decisions.5,9,14,15 In clinical practice, deterioration in consciousness remains one of the most reliable indicators for considering advanced therapy.
Cardiac involvement also played a critical role. Patients treated with HBOT exhibited higher cardiac biomarker levels and a substantially greater frequency of ischemic electrocardiographic findings. Ischemic ECG changes were independently associated with HBOT selection. Cardiovascular manifestations of CO toxicity have been well described and are thought to result from systemic hypoxia and impaired oxygen delivery.16–18 Our findings support the view that evidence of cardiac involvement should strongly influence therapeutic decision-making.
Although COHb and lactate levels were elevated in patients receiving HBOT, their discriminatory performance was modest. This observation suggests that laboratory thresholds alone may not adequately reflect clinical severity. Similar findings have been reported in prior studies, where lactate and COHb were associated with severity but demonstrated limited standalone predictive value.14,18,19 Variability in exposure duration, delayed presentation, and individual physiological differences may explain why COHb levels do not consistently correlate with tissue hypoxia.2,5 Therefore, laboratory parameters should be interpreted within the broader clinical context.
The higher mortality and intubation rates observed in the HBOT group likely reflect baseline disease severity rather than adverse effects of treatment. HBOT was preferentially administered to patients with neurological compromise or cardiac involvement, indicating that these patients represented a more critically ill subgroup. Similar patterns have been reported in previous cohort studies evaluating HBOT utilization.18,20,21 Accordingly, the observed mortality differences should be interpreted as markers of initial clinical status rather than treatment-related risk. However, the timing of intubation was not available in this study; therefore, it remains unclear whether intubation was a contributing factor in HBOT selection or a consequence of clinical deterioration.
MRI findings did not significantly differ between treatment groups. This may be attributable to the timing of imaging during the acute phase, as hypoxic–ischemic changes may not be immediately apparent.22–24 The absence of significant differences does not diminish the diagnostic value of MRI but rather highlights the dynamic evolution of neurological injury in CO poisoning.
Although comorbid conditions have been suggested to worsen outcomes in CO poisoning,25–28 no independent association between comorbidity and HBOT selection was identified in our study. Differences in study design and patient populations may account for this discrepancy.
Several limitations should be acknowledged. The retrospective single-center design introduces potential selection bias, as HBOT decisions were based on clinical judgment rather than standardized criteria. Our findings suggest that HBOT selection in CO poisoning is primarily associated with multidimensional clinical assessment. Data on exposure duration, smoking status, and long-term neurological outcomes were unavailable. Furthermore, the modest discriminatory performance of certain laboratory parameters reinforces the need for cautious interpretation. Nevertheless, the relatively large sample size and the integrated evaluation of neurological, cardiac, laboratory, and imaging findings represent important strengths. Overall, our findings suggest that HBOT selection in carbon monoxide poisoning is primarily associated with multidimensional clinical assessment. Neurological deterioration and cardiac involvement appear to be the most meaningful factors associated with HBOT selection, whereas isolated laboratory abnormalities are insufficient to guide management independently. Integrating clinical examination, electrocardiographic findings, and laboratory markers may improve risk stratification and support clinical decision-making in the emergency department. These findings should not be interpreted as evidence of treatment necessity or effectiveness, but rather as factors associated with HBOT selection.
In the absence of universally standardized HBOT criteria, treatment decisions in carbon monoxide poisoning remain largely clinician-dependent. Our findings highlight that neurological impairment and cardiac involvement represent the most clinically meaningful determinants for HBOT referral. Laboratory markers alone demonstrated limited discriminative ability, underscoring the importance of multidimensional clinical assessment.
These results may support emergency physicians in prioritizing patients for hyperbaric consultation and may contribute to more rational utilization of HBOT resources, particularly in centers where hyperbaric facilities are not readily available.
Baseline characteristics and outcome variables were clearly distinguished and presented separately to improve interpretability and consistency with reporting standards.
This study has several limitations that should be considered when interpreting the findings. First, its retrospective single-center design introduces the possibility of selection bias, as the decision to initiate HBOT was based on clinical judgment rather than predefined standardized criteria. Although this reflects real-world clinical practice, it may limit the generalizability of the results.
Second, detailed information regarding carbon monoxide exposure duration, environmental conditions, smoking status, and time to hospital presentation was not consistently available. These factors may influence both clinical severity and laboratory parameters and could potentially affect HBOT decision-making.
Third, long-term neurological and functional outcomes were not evaluated. Therefore, the present study focuses primarily on in-hospital parameters and does not address delayed neurocognitive sequelae, which are clinically relevant in carbon monoxide poisoning.
Additionally, although laboratory parameters such as lactate and carboxyhemoglobin were analyzed, their modest discriminatory performance suggests that unmeasured clinical variables may also contribute to treatment decisions. Finally, external validation in prospective multicenter cohorts is necessary to confirm the reproducibility and clinical applicability of these findings.
In this retrospective cohort study, factors associated with HBOT selection were primarily related to neurological status and cardiac involvement. These findings suggest that treatment decisions in carbon monoxide poisoning are largely influenced by clinical severity rather than biochemical parameters alone.