Blood Ethanol Estimation: A Comparison of Three Methods

Full text

Letters to the Editor
85
Blood Ethanol Estimation:
A
Comparison
of
Three
Methods
To
the
Editor:-Quick and accurate
determination of ethanol intoxication is
frequently needed
in
the ED. Clinical
estimation of intoxication is reportedly
However, the “criterion
standard,” laboratory blood alcohol
concentration (BAC) analysis, often lacks
the timeliness necessary for clinical use
with
emergently
ill
patients.
Various methods for point-of-care
estimation of BAC have been applied
in
the ED. Previous studies have ad-
dressed the correlation of clinical esti-
mation (CE),1-4 breath analysis
(BRE),s-’O and saliva alcohol (SAL)
analysis”,” with BAC analysis. The ac-
curacy of each of these methods is af-
fected by several patient and technician-
dependent factors.
No
previous study has compared all
of these modalities of alcohol level es-
timation (ALE)
to
one another
in
the
same clinical setting. We performed a
blinded prospective analysis of these
modalities
in
a cohort of ED patients.
Methods
Study
Design.
This was a prospec-
tive, blinded study comparing three dif-
ferent methods of point-of-care ALE
(CE, BRE, and SAL analysis)
with
lab-
oratory BAC determination.
Setting.
The study was performed
from July
5
to December
14,
1993,
at a
university-based hospital
with
an ED
annual census of approximately
55,000
visits. The ED is staffed by residents
from emergency medicine (EM) and
other training programs who are super-
vised by full-time
EM
attending physi-
cians.
PopulationlHuman Subject
Com-
mittee Review.
Any ED patient for
whom a BAC determination was
or-
dered was considered eligible for the
study. Patients were enrolled
as
a con-
venience sample as time and personnel
permitted. Verbal consent was obtained
from patients, and they were given an
information sheet regarding the study at
the time of their release from the ED.
The study was approved by
our
insti-
tution’s Committee on Research Involv-
ing Human Subjects.
Experimental Protocol.
It was the
authors’ intent to assess these measures
in
our
normal routine ED practice.
Therefore, no additional training was
provided to participating individuals re-
garding CE
or
BRE of alcohol intoxi-
cation. In
our
ED, nurses and nursing
assistants routinely perform BRE
on
our
patients as indicated. These staff mem-
bers receive one in-service training ses-
sion. Staff members were cognizant that
the study was ongoing and were asked
to make a special effort to obtain reli-
able BRE measurements. Blood sam-
ples were drawn, sent to
our
hospital
laboratory, and analyzed per preexisting
institutional protocols. The
only
depar-
ture from routine was the addition
of
SAL estimation as performed by trained
research assistants and physicians after
receiving one in-service training session.
After skin preparation with povi-
done-iodine (Betadine), a BAC sample
was sent to the hospital laboratory in
a sodium-fluoride-containing Vacu-
tainer (Bectin-Dickenson, Franklin
Lakes, NJ). Within
15
minutes
of
a blood
draw, three ALEs were performed by
separate individuals blinded to each
other’s results.
The
three procedures
were performed sequentially
in
the fol-
lowing order: an ED registered nurse
(RN) recorded a clinical estimate. A dif-
ferent RN
or
a nursing assistant per-
formed a BAC determination with an
Alcosensor I11 (Intoximeters Incorpo-
rated, St. Louis,
MO),
which was cali-
brated on a monthly basis and used per
the manufacturer’s specifications. The
SAL was determined using a QED
A350
Saliva Alcohol Test (Enzymatics Incor-
porated, Horsham, PA). This device has
a “QIA spot,” which darkens only on
contact with an adequate volume
of
sa-
liva sample, thus allowing the user to
determine the adequacy of the test. The
SAL determinations were performed by
previously trained research assistants and
physicians. Laboratory analysis of BAC
was performed by trained laboratory
technicians using the TDx Immunoassay
(Abbott Laboratories, Chicago, IL) per
routine at
our
institution.
Data Measurements
and
Anafy-
sis.
Data from ALE for each tech-
nique and the criterion standard, BAC
determination, were recorded by the re-
search assistants. Additional data in-
cluded patient demographics, clinical in-
dication for blood ethanol, and
turnaround time for BAC results. The
turnaround time was defined as the in-
terval from the time the blood was drawn
to the time the BAC was reported.
The three ALEs were compared with
BAC determination using the kappa sta-
tistic and Pearson correlation. In addi-
tion, sensitivity and specificity (and
95%
CIS) for detecting a BAC
of
100
mg/dL
were calculated for each ALE. Since the
upper
limit
of detection for alcohol for
the SAL analysis used was
350
mg/dL,
all the BACs
>
350
mgidL were ad-
justed to
350
mg/dL for the comparison
with the SAL level. Only those patients
for whom attempts were made to collect
data
for
all three ALE methods were
subjected to analysis. Patients who did
not have adequate breath and/or saliva
samples were noted but not included
in
the final analysis.
Results
Data collection was attempted for
78
patients. Six were excluded because one
or
more
of
the ALE methods were not
performed. Twenty-eight patients
(36%)
were not included
in
the final analysis
because adequate specimens could not
be obtained for either of
or
both the
BRE and SAL techniques. Of those ex-
cluded, five
(6%)
patients could not
provide sufficient specimens
for
either
method;
14
(18%)
could not provide an
adequate breath specimen only; while
nine
(12%)
patients could not provide
an adequate saliva specimen only.
Of
the remaining
44
patients in-

86
ACADEMIC EMERGENCY MEDICINE JAN 1996 VOL 3/NO 1
500
-
400
9
UI
5
300
0
-
200
100
0
1
3
5
7
9
11
13
15
17
19
21 23 25 27 29 31
33
35 37 39 41 43
Patient Number
I
FIGURE
1.
Distribution
of
blood
alcohol concentration
(BAC)
for
the study population.
cluded
in
the analysis, the mean age was
43.7
1?:
14.5 years (SD), with a range of
22-92 years. There were 36 (82%) men
and eight (12%) women. The BAC de-
terminations were ordered
in
the follow-
ing settings: evaluation
of
depressed level
of consciousness
(n
=
23), trauma/head
injury
(n
=
11). part
of
a drug screen
(n
=
3), seizure evaluation
(n
=
3),
other
(n
=
4). Clinical estimation was
done by a total
of
27 registered nurses,
with
no
nurse participating more than
four
times. The mean turnaround time
for
blood alcohol values was 90
5
47
(SD)
minutes.
The BACs ranged from
0
to 576 mg/
dL (Fig. 1). The mean BAC was 288
t
129 (SD) mg/dL. Kappa values com-
paring blood ethanol values and the three
ALE methods were 0.84 for SAL anal-
ysis, 0.69 for BRE, and 0.55 for CE.
Pearson correlation coefficients were 0.90
for SAL analysis, 0.77 for BRE, and
0.58 for CE. Lower BACs showed a bias
toward more favorable comparisons.
For
SAL analysis and CE, the sensitivity and
specificity for detecting a BAC
>
100
mg/dLwere both 1.0(95%CI0.97,1.0).
For BRE, the sensitivity was 1.0 (95%
CI 0.97,l.O) and the specificity was 0.98
(95% CI 0.94, 1.0) for detecting a BAC
>
100 mg/dL.
Discussion
In
the present study, three modali-
ties: CE, BRE, and SAL analysis were
compared with the determination
of
BAC
by
use
of
the kappa statistic (concor-
dance) and Pearson correlation.
In
both
comparisons, SAL determination out-
performed the other two modalities.
However, a significant number
of
pa-
tients were unable to perform either BRE
(n
=
14, 18%), SAL testing
(n
=
9,
12%),
or
both
(n
=
5,6.4%). Although
this is problematic in terms
of
statistical
analysis, we consider it an important re-
sult that points out a significant limita-
tion
of
BRE and SAL analysis
in
the
ED setting.
Breath analysis requires a certain
degree of ability on the part of both the
patient and the technician. Patients’
abilities to provide breath specimens are
influenced by their abilities to follow di-
rections and to accurately execute the
procedure while in the intoxicated state.
Previous studies have reported a drop
in
correlation
of
orally obtained and na-
sally obtained breath alcohol samples
between “cooperative” and “uncoop-
erative” patientss and between con-
scious and unconscious patients.”
In
1981,
McDermott and Evans reported a cor-
relation
of
0.89 and wrote, “It is nec-
essary to stress that in order to obtain
these results a considerable effort is
needed to train staff members in the use
of the machine, and to maintain and
standardize the unit.”x In the current
study, a significant number of patients
(n
=
19) could not provide adequate
specimens
for
BRE.
Saliva specimen acquisition also was
impaired by lack
of
an adequate amount
of saliva in patients with dry mucous
membranes
(n
=
14) in the current study
population. Jones et al. reported the same
difficulty using the QED Saliva Alcohol
Test, even among healthy volunteers.
l2
This difficulty becomes particularly
problematic in view
of
the pharmaco-
logic tendency of alcohol to contribute
to clinical dehydration. Therefore, al-
though SAL determination and BRE may
provide an accurate means
of
ALE under
carefully controlled conditions, we have
found the performance
of
these screen-
ing methods to be significantly variable,
and thus they are of limited clinical util-
ity
in
the ED.
The present study is limited by the
fact that data were collected on
a
con-
venience rather than a consecutive ba-
sis. It is possible that certain patients
were not included because they were too
uncooperative
or
considered too intox-
icated to provide breath and/or saliva
specimens. If this were the case, an even
larger percentage
of
patients would have
been excluded because
of
the lack
of
adequate specimens. This would further
support the limitations of BRE and/or
SAL testing
in
a clinical setting.
In
the current study, several indi-
viduals performed BRE and SAL anal-
ysis, mimicking typical day-to-day prac-
tice. All had limited training, but
differences in experience and technical
proficiency were not assessed. The study
design called for the individuals per-
forming the three different ALES
to
be
blinded to one another’s results. Any
violation in this portion
of
the protocol
would have potentially invalidated the
results reported. Since one
of
the tech-
niques tested (SAL analysis) had an up-
per limit
of
detection, all BACs
>
350
mg/dL
(n
=
13) were adjusted to
350

Letters to the Editor
87
mg/dL for the comparison with SAL
level. This artificially improved our kappa
statistic and Pearson correlation. Fi-
nally. the study was limited by the rel-
atively low number of patients sampled.
Further studies might address the ef-
fect of additional years of experience on
CE of BAC. Our own future efforts in-
clude assessment
of
the use of serum
instead of saliva in the
QED
device. This
method of sampling promises to be less
dependent on patient compliance level
and hydration status.
Conclusion
Our findings suggest that SAL es-
timation using a QED
350
is
superior to
BRE using an Alcosensor
111
and to
CE
for rapid estimation of BAC in the
ED
patient. However,
all
tests seem to dis-
criminate between patients with
BAC
levels
>
100
mg/dL equally well.
A
sig-
nificant number of patients are unable
to provide a sufficient quantity
of
saliva
to test. This is a significant factor lim-
iting the current utility of SAL estima-
tion for
ED
patients. A similar failure
to cooperate with BRE was noted.
MARK
E.
KEIM,
MD
JOEL
M.
BARTFIELD.
MD
NANCY
RACCIO-ROBAK, RN, MPH
Albany Medical College, Albany,
NY
Department of Emergency Medicine
Prior presentation: ACEP Research Forum,
San Francisco, CA, February 1995.
The authors thank John Lekas, BS, EMT-P,
and Annette Liu, BS, for their help with study
implementation. Special appreciation to the
nursing staff of the Emergency Department
of
Albany
Medical Center for their help with
data collection. The authors also
thank
Terry
L.
Peters, MS,
for
her help in data analysis.
Key words: ethanol; alcohol; blood levels;
intoxication; measurement.
REFERENCES
1.
AMA Council on Scientific Affairs. Al-
cohol
and
the driver. JAMA. 1986;
2. Rutherford WH. Diagnosis
of
alcohol
ingestion
in
mild head injuries. Lancet.
3.
Bogen EJ. Drunkeness. JAMA. 1927;
4. Maio RF,
Wu
A, Blow FC, Zink BJ.
255 522-7.
1977; 1:1021-3.
89:1508-11.
Prehospital care providers
do
not accu-
rately identify motor-vehicle crash
pa-
tients
with
positive serum alcohol con-
centrations: a brief report. Prehosp
Disaster Med. 1994; 9(suppI):S74.
5.
Gibb KA, Yee AS, Johnston
CC,
Martin
SD, Nowak
RM.
Accuracy and useful-
ness of
a
breath alcohol analyzer. Ann
Emerg Med. 1984; 13516-20.
6. Lester D. Breath tests for alcohol.
N
Engl
J
Med. 1971; 284:1269-70.
7.
Alobaidi
TAA, Hill DW, Payne JP. Sig-
nificance of variations
in
blood-breath
partition coefficient of alcohol. Br Med
8.
Evans RP, McDermott
FT.
Use of al-
cometer
in
a casualty department. Med
9. Wenzel
J,
McDermott
FT.
Accuracy
of
blood alcohol estimations obtained with
a
breath alcohol analyzer
in
a casualty
department. Med
J
Aust.
1985;
142:627-
8.
10.
Gerberich
SG,
Gerberich BK, Fife D,
Ciceno J, Lilja
P,
Van Bertcom L. Anal-
ysis of the relationship between blood
alcohol and nasal breath alcohol concen-
trations: implications
for
assessment of
trauma cabes. J Trauma. 1989; 29:338-
43.
11.
Christopher TA. Zeccardi JA. Evalua-
tion of the Q.E.D. saliva alcohol test-
a new rapid accurate device for
mea-
suring
ethanol in saliva. Ann Emerg Med.
12.
Jones AW. Evaluation of Q.E.D. saliva
alcohol test- final report. 1992, De-
partment
of
Alcohol Toxicology,
Uni-
versity Hospital, Linkoping, Sweden;
Enzymatin Incorporated, Horsham,
PA.
J. 1976; 2: 1479-81.
J
Aus~. 1981; 1:185-6.
1992; 21:120-2.
Transport
of
Assaulted
Patients Using Nonmedical
Personnel
To
the
Editor:-At first glance, the
article “Urban Trauma Transport of
As-
saulted Patients Using Nonmedical Per-
sonnel” by Branas and colleagues may
not seem applicable to many emergency
physicians or emergency medical ser-
vices (EMS) medical directors. After
all, few EMS systems interface with po-
lice officers who are willing to get ac-
tively involved with the out-of-hospital
care and transport of injured patients.
In even fewer cases
is
it true that “police
are explicitly permitted
to
transport
penetrating injuries and
often
do
[em-
phasis added]
.
. .
,”
as in Philadelphia.
In fact, I know
of
no other
EMS
system
that has a “recognized and defined role
for police transport
of
assaulted trauma
victims.” However, whether or not the
local police department transports pa-
tients, the article is relevant to
all
phy-
sicians involved in the out-of-hospital care
of the trauma victim.
The authors concluded that as-
saulted patients have generally equiva-
lent outcomes regardless of whether they
were treated and transported by fire
medics or simply transported by
police
officers. This finding goes against the
current dogma of fluid resuscitation and
other measures designed to augment
central intravascular volume and raise
the blood pressure (BP) of hypotensive
trauma victims. However, it is well in
keeping with the clinical experience of
many EMS, emergency medicine, and
trauma specialists. It also indirectly sup-
ports recent literature on out-of-hospital
care of the trauma ~ictim.~-~
The authors state that “the effec-
tiveness of out-of-hospital endotracheal
intubation, fluid resuscitation, and con-
trol
of
hemorrhage in improving the sur-
vival
of
trauma patients has been re-
ported.” However, the effectiveness of
these modalities also has been ques-
tioned. An evolving theory
of
out-of-
hospital trauma management holds that
attempts to normalize BP in the patient
with uncontrolled hemorrhage (using
preoperative fluid resuscitation and/or
with adjuncts such
as
the pneumatic an-
tishock garment) may be detrimental.
These measures
may
lead to accelera-
tion
of
hemorrhage, hydraulic clot dis-
lodgment, and/or dilution of clotting
factors. The study of Branas et
al.
pro-
vides indirect support for that evolving
paradigm. Therein lies the importance
of
this paper, at least in my mind.
However, the authors focus on the
issue of
“a
trade-off between using per-
sonnel who are less-equipped and less-
trained for emergency medical response
against the increased costs of equipment
and training
for
transport by specialized
EMS personnel.” Isn’t the issue not the
increased cost of training and equipping
these personnel, but their benefit?
Branas et al. found that fire medics
generally transported sicker patients than
did nonmedical police officers. Not sur-