Original article
Emerg Med J: first published as 10.1136/emermed-2012-201712 on 15 October 2013. Downloaded from http://emj.bmj.com/ on 3 May 2019 by guest. Protected by copyright.
Department of Emergency
Medicine, Izmir Tepecik
Research and Educational
Hospital, Izmir, Turkey
Correspondence to
Dr Ersin Aksay, Department of
Emergency Medicine, Izmir
Tepecik Research and
Educational Hospital, Izmir,
Turkey; ersin.aksay@gmail.com
Received 5 July 2012
Revised 11 March 2013
Accepted 21 September 2013
Published Online First
16 October 2013
To cite: Zorbalar N,
Yesilaras M, Aksay E. Emerg
Med J 2014;31:e66–e70.
e66
Carbon monoxide poisoning in patients presenting
to the emergency department with a headache
in winter months
Nilay Zorbalar, Murat Yesilaras, Ersin Aksay
ABSTRACT
Background Carbon monoxide (CO) poisoning is an
important reason for emergency department (ED) visits
during winter months, but because there are no specific
symptoms it can be difficult to diagnose. We aimed to
determine the frequency of CO poisoning in patients
presenting to the ED with headaches during winter
months and evaluate the ability of non-invasive
carboxyhaemoglobin measurement (SpCO) to screen for
CO poisoning in these patients.
Methods SpCO measurement values of adult patients
were measured non-invasively with a Rad-57 Pulse CO-
Oximeter. Patients whose initial SpCO reading was over
10% underwent a venous blood draw for laboratory
determination of invasive carboxyhaemoglobin (COHb)
measurement. Patients with a invasive COHb level of over
10% were diagnosed with CO poisoning. Percentage of
screened patients with suspected and occult CO
poisoning, the distribution of patients with CO poisoning
by time of day of the ED visit and the positive predictive
value of SpCO to detect CO poisoning were calculated.
Results 483 patients presenting with headaches were
screened with SpCO measurement. Thirty-eight had a
mean SpCO value of over 10%, 31 (6.4% of the study
population) of which had elevated COHb confirmed by
laboratory determination. SpCO measurement, therefore,
had a positive predictive value of 82% for CO poisoning.
Twenty-four (77%) of the CO poisoning cases were
suspected and seven (23%) were occult. CO poisoning
was detected more frequently in patients visiting the ED
after midnight and during morning hours.
Conclusions CO poisoning should be kept in mind in
patients presenting to the ED with a headache. SpCO is
an effective screening tool to detect CO poisoning in
these patients.
INTRODUCTION
As signs and symptoms of carbon monoxide (CO)
poisoning are often vague, it is difficult to diagnose in
patients presenting to the emergency department
(ED). Headaches are the most common symptom of
CO poisoning (37–85%) and so may be the best
single indicator to prompt screening for suspected
and occult CO poisoning. Clinical research and case
studies reveal that the incidence of occult CO poison-
ing in patients visiting ED is more frequent than
expected.1–3 There are only a small number of studies
on ED frequency of occult CO poisoning in the litera-
ture.4 5 We aimed to determine the frequency of CO
poisoning in patients presenting to the ED with head-
aches during winter months, and to test the utility of
non-invasive carboxyhaemoglobin (SpCO) as a
Zorbalar N, et al. Emerg Med J 2014;31:e66–e70. doi:10.1136/emermed-2012-201712
method of screening for CO poisoning in these
patients.
METHODS
This is a prospective cross-sectional study con-
ducted on patients who presented to the ED due to
headache symptoms between 1 February and 31
March 2011. Our hospital (Izmir Tepecik Traning
and Research Hospital) is a tertiary facility with
over 200 000 patients being treated in 2011.
Approval from the local ethics committee was
obtained before the start of the study and all
patients signed an informed consent form before
being included in the study. Inclusion criteria
included consecutive patients 14 years and older
with one of their chief symptoms being headache.
Patients were excluded from the analysis if three
consecutive SpCO’s could not be performed for
any reason or if an invasive SpCO could not be
performed for any reason.
COHb and methaemoglobin levels of the
patients were measured non-invasively using an
adult or paediatric reusable finger probe (rainbow
DCI or rainbow DCIP, rev G) connected to a
Rad-57 Pulse CO-Oximeter (software V.1713;
Masimo Corp, Irvine, California, USA). The
Rad-57 and probes were obtained from Masimo
Corporation and returned after the data collection.
Emergency residents who performed the mea-
surements received standardised training in the
proper use of the device before the study.
Step-by-step directions for non-invasive measure-
ments on patients were developed and proper use
of the equipment and use of the correct technique
were enforced. When feasible, the third or fourth
finger of the patient’s non-dominant hand was used
for the measurement site. The thumb and little
fingers, finger nails with nail polish or fingers with
henna were not used. Demographic information,
measurement time, accompanying symptoms, vital
signs and values measured by Rad-57 (carboxy-
haemoglobin and methaemoglobin) were recorded
on standardised study forms.
Two additional SpCO measurements were per-
formed sequentially on patients whose initial SpCO
values were greater than 10% within 5 min after
arrival. The mean of three consecutive measure-
ments was recorded as the COHb value in those
patients. Furthermore, patients whose mean COHb
values were more than 10% underwent an invasive
measurement in which a sample of venous blood
was taken by a nurse from the brachial vein of the
same arm which was used for the SpCO. COHb and
lactate levels in venous blood were measured using a

blood gas analyser (GEM Premier 3000-GEM OPL
CO-Oximeter; Instrumentation Laboratory), within 5 min of the
SpCO measurement by the physician. Calibration of the
CO-Oximeter was performed every 15 days by the distributor
for the manufacturer, according to the manufacturer’s
recommendations.
In cases in which there was inconsistency between results of
the SpCO and the invasive COHb measurements, the values
obtained from venous blood samples were taken as gold stand-
ard to diagnose CO poisoning. Patients with a COHb level
greater than 10% were diagnosed with CO poisoning irrespect-
ive of their smoking status. All patients enrolled in the study
were treated based on their individual clinical conditions.
All data were recorded on a computer using Microsoft Office
Excel 2007. Statistical analyses were performed using the statis-
tics for biomedical research software (Med Calc Software, V.12,
Belgium). Quantitative data were summarised with frequencies
and percentages, whereas qualitative data were presented with
mean±SD. Fisher’s exact test was used for the analysis of the dif-
ference between syncope frequencies in patients with and
without CO poisoning. Student’s t test was used for analysis of
the difference between invasive COHb levels of the patients with
and without syncope, the difference between lactate levels in
false-positive patients and true-positive patients and the differ-
ence between venous blood carboxyhaemoglobin levels of
smoking patients and non-smoking patients.
All analyses were conducted within 95% CI and a p value of
less than 0.05 was considered significant. The positive predictive
value (PPV) of SpCO to detect CO poisoning, defined as the
percentage of cases of SpCO level greater than 10% when
COHb level was greater than 10% was determined.
RESULTS
A total of 29 439 patients presented to our ED during the study
period and of those, 485 (1.6%) reported a headache as one of
their chief symptoms. Four hundred and eighty-three consecu-
tive patients consented to be in the study but in one of these
patients a venous blood sample could not be obtained within
5 min. Of the 482 patients who made up the study population,
186 (38.6%) were men and the mean age was 38.0±15.6 years.
Thirty-eight patients (7.9%) had a mean SpCO level above
10%, 31 (6.4%) of whom had a invasive COHb level greater
than 10% and were diagnosed with CO poisoning. SpCO meas-
urement, therefore, had a positive predictive value of 82% for
the detection of elevated carboxyhaemoglobin. Symptoms,
demographics, vital signs and laboratory results of patients with
mean SpCO values greater than 10% are shown in table 1.
Twenty-three (%74.2) patients with CO poisoning were
women and the mean age was 33.6±14.9 years. Vital signs of
these patients were as follows (mean±SD): systolic blood pres-
sure 129.2±19.2 mm Hg, diastolic blood pressure 76.6
±14.3 mm Hg, pulse rate 95.6±14.6/min, respiratory rate 18.9
±2.3/in, temperature 36.3±0.5/°C and oxygen saturation 97.8
±1.3% on room air. The mean SpCO value for patients with
CO poisoning was 20.6±4.8% while the mean of invasive
COHb level was 20.9±6.1%. Associated symptoms were
nausea/vomiting (25.8%), syncope (19.4%), weakness (9.7%),
dizziness (9.7%), chest pain (6.5%) and abdominal pain (3.2%).
Twenty-four patients with CO poisoning stated that they sus-
pected they had been exposed to CO, whereas the remaining
seven (1.4% of all patients) stated that they never suspected
such a possibility (occult cases). The mean invasive COHb level
in the occult CO poisoning patients was 19.9±5.5%. The
Zorbalar N, et al. Emerg Med J 2014;31:e66–e70. doi:10.1136/emermed-2012-201712
Original article
Emerg Med J: first published as 10.1136/emermed-2012-201712 on 15 October 2013. Downloaded from http://emj.bmj.com/ on 3 May 2019 by guest. Protected by copyright.
source of CO for all patients was using stoves inappropriately
for heating.
One patient in whom an elevated troponin level was detected
underwent hyperbaric oxygen treatment. The remaining 30
patients received high flow supplemental oxygen in the ED.
The mean length of ED stay for CO poisoning patients was
402±249 min. None of the patients enrolled in the study died.
Ten patients experienced syncope before to arrival in the ED
and six of these were subsequently diagnosed with CO poisoning.
Syncope was found to occur in 19.4% of the patients with CO
poisoning and 0.9% of patients who complained of a headache
and did not have CO poisoning ( p<0.001). There was no signifi-
cant difference between the invasive COHb levels of the patients
with and without syncope (20.5±6.5%, n=6 vs 22.9±4.8%,
n=25, p>0.05). Eight patients with CO poisoning had been
using tobacco. There was no significant difference in venous carb-
oxyhaemoglobin levels between smokers and non-smokers (22.4
±4.8% vs 20.4±6.6%, p=0.45).
The distribution of patients diagnosed with CO poisoning by
the time of day of the ED visit is shown in figure 1. CO poison-
ing was detected more frequently in patients visiting the ED
after midnight (00:00–04:00) and during morning hours
(07:00–10:00).
Of the 38 patients whose SpCO levels were above 10%, seven
had invasive COHb levels less than 10%. All of these ‘false-
positive’ patients had methaemoglobin measurement (SpMet)
levels greater than 1.57% but the mean±SD SpMet for these
patients was lower than those of the ‘true-positive’ patients (1.8
±0.17% vs 2.1±0.46%). Sixteen out of 19 patients whose
SpMet levels were found to be high (>%2) were diagnosed
with CO poisoning.
While mean lactate levels in venous blood was found to be 1.02±
0.26 mmol/L in false-positive patients, it was 1.71±0.83 mmol/L in
‘true-positive’ patients (p=0.056).
DISCUSSION
The main aim of the study is to determine the frequency of CO
poisoning in patients presenting to the ED with headaches
during winter months using non-invasive SpCO as a screening
tool. 6.4% of our study population was diagnosed with CO poi-
soning. The PPV of SpCO was found to be 82%.
CO poisoning is one of the leading causes of death resulting
from accidental poisoning worldwide.6 The use of fossil fuels as
a heating source is common in Izmir (which is the third biggest
city in Turkey, its population is over 3 900 000), frequent ED
visits due to CO poisoning in winter months are observed.
Patients who do not suspect exposure to CO can be difficult to
diagnose because symptoms are variable and sometimes vague.
Lack of a timely diagnosis can delay treatment in those patients
or worse, patients discharged with an incorrect diagnosis can be
re-exposed to CO.
Among the most common symptoms of CO poisoning are
headache (37–85%), ataxia (18–69%), weakness (9–69%) and
vomiting (18–52%).1 7 A headache typically presents when the
COHb level is between 10% and 20%.3 The mean venous
blood COHb level in our patients was 20.9%. The majority of
our patients experienced a mild to moderate level of CO
poisoning.
Suner et al5 performed SpCO using aRAD-57 in 10 856
patients presenting to the ED irrespective of their symptoms.
They reported that the sensitivity and specificity of SpCO using
invasive COHb as the gold standard test were 94% and 54%,
respectively. In that study, occult CO toxicity was defined as a
SpCO value of 9% or greater for non-smokers and 13% or
e67
e68

Table 1 Symptoms, demographics, vital signs and laboratory results of patients with mean SpCO greater than 10%

Original article
Systolic blood Diastolic blood Pulse Body Oxygen Mean Venous Mean
Patient Age Associated Smoking pressure pressure rate Respiratory temperature saturation SpCO carboxyhaemoglobin SpMet Occult
no. Gender (years) symptoms status (mm Hg) (mm Hg) (/dk) rate (/dk) (°C) (%) (%) (%) (%) poisoning

1 F 36 Vomiting chest − 124 68 88 16 36 99 17.33 12.5 1.93 +

pain
2 F 33 Vomiting chest + 124 56 104 19 39.7 94 16 0 2.1 NP
pain weakness
syncope
139 F 38 – + 163 102 112 19 36 97 24.33 22.8 2.23 −
165 F 38 Syncope − 110 70 65 16 36.8 100 15.33 22.4 1.2 +
177 M 31 Vomiting − 132 74 94 19 36.4 97 25.33 23.9 2.63 −
245 F 68 – − 151 82 96 19 36 98 28.67 24.4 2.93 −
246 F 39 Vomiting weakness − 171 85 127 20 36 96 19.67 18 2.53 −
247 F 68 – − 177 83 77 18 36.8 98 15.67 0 1.57 NP
261 M 32 – − 135 78 99 19 36.1 100 22.67 16.8 2.13 −
293 F 22 Vomiting dizziness − 107 67 86 14 36 97 16.67 30 1.8 +
294 F 14 Syncope − 109 69 142 16 37.3 97 14 5 1.73 NP
295 M 27 – − 125 67 81 18 36.1 99 14 2.2 1.93 NP
308 M 28 – − 134 93 120 19 37.5 97 16.33 12 1.67 −
309 F 25 Vomiting − 144 86 90 20 36 97 14.67 13.1 1.63 −
310 M 27 Weakness − 128 69 83 14 37.5 97 13.67 0 1.83 NP
319 M 68 – − 126 71 90 23 36.4 98 25.67 21.2 2.63 −
Zorbalar N, et al. Emerg Med J 2014;31:e66–e70. doi:10.1136/emermed-2012-201712

320 F 40 Dizziness − 150 90 81 20 36.4 98 23.33 13.2 2.33 −

361 F 14 Vomiting − 161 71 124 18 36 97 21.33 17.7 2.27 +
362 F 16 Syncope − 110 78 85 22 36.7 99 17 20.1 1.83 +
363 F 28 Syncope + 121 68 85 18 36 99 19.67 23.8 1.73 −
375 F 40 Abdominal pain + 135 82 79 19 36.2 99 20.33 16.6 2.17 +
376 F 45 – − 147 97 101 20 36.8 96 16.67 32 1.6 −
377 F 22 – + 106 47 107 18 36.7 97 11.67 31 1.37 −
378 F 14 Syncope − 100 56 81 18 36.2 99 19.33 19 1.97 −
379 M 15 – − 160 80 99 20 36.1 97 26.33 31.2 2.63 −
380 F 31 Syncope − 125 81 78 18 36 100 21 20 2.20 +
399 M 44 Weakness + 106 64 93 19 36 98 29.33 24.2 2.80 −
400 F 16 – − 122 103 105 19 36 95 21.67 25.7 2.07 −
405 F 43 – + 115 105 112 18 36 97 23.67 24.7 2.37 −
406 F 44 – + 132 72 98 19 36.1 97 10.67 0 1.67 NP
415 F 31 – − 121 70 111 17 37 98 19 19 1.93 −
450 M 36 Nausea + 130 80 90 20 36 99 18 16.5 1.97 −
451 F 69 Syncope vomiting − 105 65 96 27 36 98 26 32 2.77 −
452 F 17 Weakness − 118 50 88 19 37.6 99 17 10.9 1.73 −
453 F 38 – − 139 63 106 18 36 96 12 12.4 1.27 −
474 F 45 – + 143 74 110 18 36 98 11.67 0.7 1.8 NP
482 M 28 Dizziness − 119 74 77 15 36 98 23.67 22.3 2.33 −
485 F 29 – + 75 100 19 36 97 25.33 19.4 2.57 −
F, female; M, male; NP, patients without poisoning; SpCO, carboxyhaemoglobin measurement; SpMet, methaemoglobin measurement.

Emerg Med J: first published as 10.1136/emermed-2012-201712 on 15 October 2013. Downloaded from http://emj.bmj.com/ on 3 May 2019 by guest. Protected by copyright.

Figure 1 The distribution of cases
according to the hours. CO, carbon
monoxide.
greater for smokers. Twenty-eight cases of CO poisoning were
detected (0.26% of all patients screened), 11 of which (39%)
were unsuspected. Only three out of those 11 patients had a
headache among their list of chief symptoms.
A number of studies in the literature evaluate the frequency
of CO poisoning in patients presenting to the ED complaining
of a headache.7 Eberhardt et al7 measured CO levels in the
exhaled breath of patients who presented to the ED in winter
months with headache symptoms in 2006. They found that 12
patients out of 170 (7.1%) had CO poisoning (cut-off value for
smokers 5%, for non-smokers 2%). In their study, the mean
carboxyhaemoglobin level in patients’ exhaled breath was 5%,
whereas invasive COHb levels of the same patients averaged
7.2%. Patients who were suspected of being exposed to CO
were excluded from the study, which left all patients they
detected with CO poisoning as occult poisoning. In our study,
however, we accepted carboxyhaemoglobin levels greater than
10% in venous blood as the gold standard in diagnosing CO
poisoning. As the diagnostic value threshold was higher in our
study, the occult CO poisoning frequency was lower. While
Eberhardt et al7 presented the need for breath holding for meas-
urement as a limitation, the non-invasive method used in our
study does not have such a limitation.
All of the CO poisonings observed in our study were due to
the use of fossil fuels for heating. Our hospital is located in an
area where the use of stoves for heating is common. Previous
studies indicate that many patients with elevated carboxyhaemo-
globin levels are not diagnosed in the ED because they present
with non-specific symptoms, so the actual incidence of CO poi-
soning is greater than reported.1–3 While 6.4% of the patients in
our study group were diagnosed with CO poisoning, seven
(1.5%) of them had occult CO poisoning. We suggest that SpCO
be used as a screening tool during the triage of patients present-
ing to the ED with a headache in winter months. Sixteen out of
19 patients whose SpMet levels were found to be high (>%2)
were diagnosed with CO poisoning. Elevated SpMet
levels should therefore alert clinicians to the presence of CO
poisoning.
As not all of the patients who presented to the ED during the
study period had their carboxyhaemoglobin levels measured, we
were not able to evaluate the specificity of SpCO (number of
false-negative results). On the other hand, seven of the 38 patients
with high SpCO had normal venous carboxyhaemoglobin levels
(false positives). All these patients had SpMet levels greater than
1.57%. Our hospital is located in a city in the Mediterranean
region where thalassaemia is common and the incidence of
Zorbalar N, et al. Emerg Med J 2014;31:e66–e70. doi:10.1136/emermed-2012-201712
Original article
Emerg Med J: first published as 10.1136/emermed-2012-201712 on 15 October 2013. Downloaded from http://emj.bmj.com/ on 3 May 2019 by guest. Protected by copyright.
haemoglobinopathy is 4.8%.8 Both increased methaemoglobin
and haemoglobinopathies in patients may cause errors in SpCO as
stated in the RAD-57 operator manual.9 We cannot rule out, there-
fore, that the false-positive evaluations in our study were due to
the presence of haemoglobinopathies or the higher than normal
methaemoglobin levels. We therefore suggest that a confirmatory
blood gas measurement be made in patients who have SpCO levels
greater than 10% or SpMet levels greater than 1.6%.
Among the patients who presented to the ED with a head-
ache, CO poisoning was detected more frequently in those pre-
senting after midnight or in the morning hours and in those
who had accompanying syncope. In their retrospective study,
Keles et al4 reported on 323 patients diagnosed with CO poi-
soning between 2002 and 2003. The most commonly pre-
sented symptoms by the patients were headache (55%),
vomiting (44%), syncope (28%) and epileptic seizures (4%).
That study also concluded that syncope was seen more often in
patients with high carboxyhaemoglobin levels. A higher possi-
bility of CO poisoning should be carefully considered in
similar patient groups.
Limitations
As we did not perform invasive COHb tests in all 482 study
patients, we were not able to assess the rate of false negatives
from SpCO. However, despite this limitation, we were able to
detect patients who had occult CO poisoning and we therefore
suggest the use of SpCO as a screening tool in the ED.
The carboxyhaemoglobin threshold for CO poisoning was
chosen as 10% in our study. Higher and lower threshold levels
for diagnosis were reported in the literature. Some studies deter-
mine separate threshold levels for smokers and non-
smokers.5 6 8 10 Our carboxyhaemoglobin threshold level is rela-
tively high compared to others in the literature. This might have
caused missing patients with carboxyhaemoglobin levels slightly
under the threshold.
When the half-life of CO is considered, evaluation of
the time period between exposure to CO and presenting to the
ED could have been appropriate. The carboxyhaemoglobin
levels of some patients who had delayed a visit to the ED
following the exposure could have been measured at less
than 10%.
CONCLUSION
Emergency physicians should bear in mind CO poisoning as a
differential diagnosis in patients presenting to ED with
e69
 Original article
headache. SpCO measurements provide an effective screening
tool to detect occult CO poisoning in those patients.
Correction notice Since this article was published online first at emj.bmj.com
it has been changed. Every instance of screening sensitivity has been changed to
positive predictive value.
Acknowledgements Masimo Corporation loaned the Rad-57 Pulse CO-Oximeters
and sensors used in the study.
Contributors EA, MY and NZ contributed to the planning and conducting of the
study. MY and EA prepared the manuscript. All authors contributed substantially to
its revision. EA takes responsibility for the paper as a whole.
Competing interests None.
Ethics approval Ethics approval was received from the local ethics committee of
Izmir Tepecik Research and Educational Hospital.
Patient consent Obtained.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The authors shared the study data with Masimo
Corporation before submission according to the loan agreement (see loan agreement
form in the supplementary file, available online only).
e70
Emerg Med J: first published as 10.1136/emermed-2012-201712 on 15 October 2013. Downloaded from http://emj.bmj.com/ on 3 May 2019 by guest. Protected by copyright.
REFERENCES
1 Lavonas EJ. Carbon monoxide poisoning. In: Shannon MW, et al., eds. Haddad and
Winchester’s clinical management of poisoning and drug overdose. 4th edn.
Chapter 87. Philadelphia, PA: Saunders Elsevier, 2007:1297.
2 Heckerling PS. Occult carbon monoxide poisoning: a cause of winter headache.
Am J Emerg Med 1987;5:201–4.
3 Ares BM, Casais JL, Dapena D, et al. Headache secondary to carbon monoxide
poisoning. Rev Neurol 2001;32:339–41.
4 Keles A, Demircan A, Kurtoglu G. Carbon monoxide poisoning: how many patients
do we miss? Eur J Emerg Med 2008;15:154–7.
5 Suner S, Partridge R, Sucov A, et al. Non-invasive pulse CO-oximetry screening in
the emergency department identifies occult carbon monoxide toxicity. J Emerg Med
2008;34:441–50.
6 Nikkanen H, Skolnik A. Diagnosis and management of carbon monoxide poisoning
in the emergency department. Emerg Med Pract 2011;13:1–14.
7 Eberhardt M, Powell A, Bonfante G, et al. Noninvasive measurement of carbon
monoxide levels in ED patients with headache. J Med Toxicol 2006;2:89–92.
8 http://www.talasemifederasyonu.org.tr/pdf/tani/cansinTedavi-3.pdf (accessed
7 October 2013).
9 http://www.medcontrol.com/omd_lib/rad%2057%20config%20operators%
20manual-%20aug2010.pdf (accessed 19 April 2012).
10 Bland JM, Altman DG. Statistical methods for assessing agreement between two
methods of clinical measurement. Lancet 1986;1:307–10.
Zorbalar N, et al. Emerg Med J 2014;31:e66–e70. doi:10.1136/emermed-2012-201712