Best Evidence Topics
  • Send this BET as an Email
  • Make a Comment on this BET

Can non-invasive measurement of carboxyhaemoglobin be used to rule out significant carbon monoxide toxicity in patients presenting to the Emergency Department?

Three Part Question

In [emergency department patients with possible carbon monoxide toxicity] is [non-invasive measurement of carboxyhaemoglobin saturation] [sufficiently sensitive to rule out significant toxicity and sufficiently accurate to guide treatment]?

Clinical Scenario

A 45 year old woman in brought into the emergency department following a house fire, during which she was trapped in a smoke-filled room for a period before being extricated by the fire service. She has no thermal burns and is fully conscious and orientated but is complaining of a headache and dizziness. You are concerned about possible toxicity secondary to carbon monoxide inhalation. The department has invested in a non-invasive fingertip carbon monoxide saturation meter, and you wonder whether this is sufficiently sensitive to rule out significant carbon monoxide toxicity and is sufficiently accurate to guide treatment.

Search Strategy

Medline 1950-08/2011 and EMBASE 1980-11/2012
(exp CARBOXYHEMOGLOBIN/ OR exp CARBON MONOXIDE/ OR OR OR OR OR OR (CO AND oximetry).af OR OR (carbon AND monoxide AND poisoning).af OR (carbon AND monoxide AND detection).af) AND (exp DIAGNOSIS/ OR AND ( OR OR (non AND invasive).af) Limited to human and English language

‘Google Scholar’ was searched using the terms “non-invasive carbon monoxide measurement”, “non-invasive carboxyhaemoglobin/carboxyhemoglobin measurement” and “fingertip carboxyhaemoglobin/carboxyhemoglobin measurement”.
Review of references and manual searching was undertaken to identify any further relevant papers.

Search Outcome

Medline returned 159 results and embase returned 219 results. Google scholar and reference review identified a further two papers.
Abstracts were reviewed and nine unique papers relating to the three part question were identified.

Relevant Paper(s)

Author, date and country Patient group Study type (level of evidence) Outcomes Key results Study Weaknesses
Zaouter, C & Zavorsky, G
9 healthy subjects inhaling CO to COHb of 10-14% on 2 days. COHb measured simultaneously with SpCO Diagnostic accuracy studyBias-1Small numbers, healthy volunteers in test conditions unlikely to be representative of ED patients
Limits of agreement-6 to +4% COHb
Sensitivity (for COHb 10-14%)54%
Ruppel et al
149 subjects undergoing pulmonary function testing including DLCO and ABGCohort StudyBias-0.39 Different population from ED patients. 10 patients excluded from analysis due to technical problems (eg. poor oximeter trace)
Limits of agreement-8.20 to 7.41
Roth et al
1578 patients attending the ED with unrelated complaints who underwent non-invasive testing for COHb at triage and then had venous or arterial sample taken at discretion of treating physician.Cohort studyBias (Mean difference non-invasive minus venous)2.32% (95%CI 2.11-2.54%) for all patients, 1.41% (95%CI 1.01-1.81%) for smokers and 2.80% (95%CI 2.52-3.07%)for non smokersVenous and non-invasive measurement not taken simultaneously (within 1 hr, blood taken later than SpCO) which may account for apparent ‘over-reading’ of SpCO meter. Not all patients screened at triage had blood samples taken. Only 17 patients in this cohort had diagnosis of CO poisoning.
Precision4.01% (4.40% smokers, 4.28% non-smokers)
Limits of agreement (mean of difference +/-1.96 SD from mean)-5.7% to 10.37% COHb
Touger, Birnbaum, Wang, Chou, Pearson, & Bijur,
Convenience sample of 120 adult and paediatric ED patients with suspected CO poisoning who had venous sampling simultaneously with non-invasive measurementCohort studyBias (mean difference venous minus non invasive)1.4% (95%CI 0.2-2.6%)Majority of patients (97/120) had COHb <15% reducing precision of sensitivity calculation for COHb >15%. Not all patients had repeated measurements –possibly introducing selection bias to repeatability coefficient Not clear how ‘suspected CO toxicity ‘was defined for purposes of patient selection
Limits of agreement -11.6%-14.4%
Sensitivity (for COHb >15%)48% (95%CI 27-69%)
Specificity99% (95%CI 94-100%)
Repeatability coefficient (1.96 x SD differences between repeated measurements on same pt)7.6% COHb
33.3% of patients had difference between laboratory and non-invasive measurement >+/- 5%
Piatkowski, Ulrich, Grieb, & Pallua,
20 patients admitted to burn centre with CO intoxication & 5 healthy volunteer controls underwent simultaneous hourly non-invasive and venous COHb measurement Diagnostic accuracyBias (Mean difference non invasive minus venous)3.15% Very small numbers limits accuracy of precision estimate
Coulange et al
12 non-smoking adult patients admitted to ED with suspected CO poisoning underwent simultaneous non-invasive and blood COHb measurementCohort studyBias (Mean difference venous minus non invasive) -1.5%Small numbers
Suner, et al.,
10,856 patients attending ED with unrelated complaints who underwent non-invasive testing for COHb. 64 (13%) of those with elevated COHb saturation had paired venous and non-invasive COHb testing at the discretion of the treating physician.Cohort studyCorrelation co-efficientR= 0.72Only 13% of patients with elevation of SpCO on non-invasive testing (and no patients without) went on to have blood COHb tested hence assessment of sensitivity and specificity not valid and number of false negatives undefined. 22/64 false positives. ‘Cut-off’ values arbitrarily defined. Mean time between non invasive measurement and blood sampling 67 mins (95% CI 53.1-81.3 min). Manufacturer funded.
Bias (Mean difference, non invasive minus lab) -4.2% (95%CI -2.8-5.6)
Limits of agreement -16%-7.5% COHb
Sensitivity (using cut off for non smokers of >9%, smokers >13%)94%
Barker, Curry, Redford, & Morgan
Ten healthy volunteers breathing 500ppm CO until COHb reached 15%. Samples taken simultaneously from arterial line as non-invasive SaCO measuredHuman volunteer methods comparison studyCorrelation co-efficientR=0.867Very small study in healthy volunteers. Small numbers limit accuracy of precision estimate. Results only compared up to COHb 15%. Manufacturer sponsored.
Bias (mean difference non invasive minus lab)-1.22%
Precision (SD of difference)+/-2.19%


The non-specific presentation of CO poisoning can make diagnosis difficult and requires high level of clinical suspicion followed by confirmation with carboxyhaemoglobin testing. Laboratory testing for venous carboxyhaemoglobin is the accepted gold standard for diagnosis, however requires venepuncture and introduces a significant delay into the diagnostic process. The availability of a rapid, non-invasive test for presence of significant levels of carboxyhaemoglobin would therefore be clinically valuable. Conventional pulse oximetry uses two wavelengths of light and therefore the calculation of fractional absorption relies on the assumption that only two species of haemoglobin (oxy- and deoxy- haemoglobin) are present. If other light absorbing species are present the calibration and calculated value of the sensor may be invalid. It is well recognised that conventional saturation monitoring is unreliable in the presence of dyshaemoglobins such as carboxyhaemoglobin, due to the similarity in the wavelengths of light absorption between carboxyhaemoglobin and oxyhaemoglobin. Due to the potentially serious consequences of a false negative reading and a consequently ‘missed’ diagnosis of carbon monoxide poisoning, a high sensitivity and degree of accuracy is required for this test to be clinically useful. A number of studies have attempted to evaluate the accuracy of non-invasive measurement of carboxyhaemoglobin in a variety of different situations. These studies demonstrated a systematic bias within limits which may be considered clinically acceptable although the size and direction of bias was variable. In the majority of studies the non-invasive measurement was found to read slightly higher than the venous carboxyhaemoglobin assay; however the delays in venous blood sampling may account for this difference. The calculated ‘limits of agreement’ should be within a range which is considered clinically acceptable and is therefore to a degree a matter of judgement (Bland & Altman, 1986). The studies which calculated limits of agreement all found limits which are sufficiently broad to be clinically unacceptable in the context of possible carbon monoxide toxicity given the effect that differences of this magnitude would exert on clinical management. Touger et al (2010), the only paper to calculate a repeatability coefficient, found a value of 7.6% which may also be considered unacceptably high in the context. Sensitivity for the detection of elevated CO in these studies varies between 48-99%. The two highest quality studies available differ significantly in the values presented for sensitivity and specificity. This may be due to the different populations used in these studies or to the different ‘cut-off’ values used to define carbon monoxide poisoning. The study by Touger et al (2010) used a sample of Emergency Department attendees suspected of suffering from carbon monoxide poisoning and used a cut-off value of 15% COHb, giving a sensitivity of only 48% and a specificity of 99%. The paper by Roth et al (2011) refers to an unselected group of emergency department patients presenting with unrelated complaints and used a ROC curve to define a cut-off of 6.6% COHb, giving a sensitivity of 94% and a specificity of 77% for the study population, however for the subgroup of smokers sensitivity and specificity were both lower at 89% and 71% respectively. Of concern in the paper by Touger et al (2010) a number of patients with potentially toxic levels of COHb had false negative (0%) readings using the non-invasive monitor.

Clinical Bottom Line

The broad range of the limits of agreement and the inadequate sensitivity in the available studies mean that at the present time the evidence does not support the use of non-invasive carboxyhaemoglobin saturation monitors as a rule-out tool in the context of Emergency Department patients suspected of being at risk for carbon monoxide poisoning and confirmatory arterial or venous carboxyhaemoglobin testing should be undertaken.


  1. Zaouter C, Zavorsky GS The measurement of carboxyhaemoglobin and methaemoglobin using a non-invasive pulse co-oximeter. Respiratory Physiology & Neurobiology 2012; 182(2):88-92
  2. Ruppel G, Wilson H, Gall V, Hempkens J Multi-wavelength pulse oximeter is not suitable for adjusting DLCO Measurements. Respiratory Care 2011 56(8):1115-1121
  3. Roth, D., Herkner, H., Schreiber, W., Hubmann, N., Gamper, G., Laggner, A., et al. Accuracy of non-invasive multiwave pulse oximetry compared with carboxyhaemoglobin from blood gas analysis in unselected emergency department patients. Annals of Emergency Medicine 2011; 58 (1), 74-79
  4. Touger, M., Birnbaum, A., Wang, J., Chou, K., Pearson, D., & Bijur, P. Performance of the RAD-57 pulse co-oximeter compared with standard laboratory carboxyhemoglobin measurement. Annals of Emergency Medicine 2010; 56 (4): 382-388.
  5. Piatkowski, A., Ulrich, D., Grieb, G., & Pallua, N. A new tool for the early diagnosis of carbon monoxide intoxication. Inhalation Toxicology 2009; 21 (13): 1144-1147
  6. Coulange, M., Barthelemy, A., Hug, F., Thierry, A., & De Haro, L. Reliability of new pulse CO-oximeter in victims of carbon monoxide poisoning. Undersea and Hyperbaric Medicine 2008; 35 (2): 107-111.
  7. Layne, T., Snyder, C., Brooks, D., & Enjeti, S. Evaluation of a new pulse co-oximeter; Non invasive measurement of carboxyhemoglobin in the outpatient pulmonary lab and emergency departments. Respiratory Care Open Forum Abstracts 2006
  8. Barker, S., Curry, J., Redford, D., & Morgan, S. Measurement of carboxyhemoglobin and methemoglobin by pulse oximetry. A human volunteer study. Anesthesiology 2006; 105 (5): 892-897