See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/21375454

The Effect of Breath Alcohol Simulator Solution Volume on Measurement

Results

Article  in  Journal of Analytical Toxicology · November 1991

DOI: 10.1093/jat/15.6.332 · Source: PubMed

CITATIONS READS

8 429

3 authors, including:

Rod Gullberg
Northwest University
61 PUBLICATIONS   630 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Teaching View project

All content following this page was uploaded by Rod Gullberg on 28 August 2016.

The user has requested enhancement of the downloaded file.

 Journal of Analytical Toxicology, Vol. 15. November/December 1991

Technical Note

The Effect of Breath Alcohol Simulator Solution Volume

on Measurement Results

P a u l R. S p e c k , A n t h o n y J. M c E I r o y , a n d R o d G. G u l l b e r g
Washington State Patrol, Breath Test Section, 6431 Corson Avenue South, Seattle, Washington 98108-3462

the impact of ethanol depletion over 10-20 tests is not signifi-

Abstract j
cant. The question arises, however, as to the influence of solution
Breath alcohol simulator devices (simulators) containing volume on measured results in a breath-testing instrument. The
alcohol and water solutions are used to calibrate and test purpose of this paper is to determine the effect of simulator so-
breath-testing instruments. The manufacturers of simulators lution volume upon measured results.
design them to contain 500 mL of solution. This study
evaluated the variability observed among three different
groups of data: one group (experimental) in which the solution
volume varied from 400 mL to 600 mL, one group (control) in
which the simulator volume remained constant at 500 mL for Experimental
each aliquot, and finally one group consisting of field
simulator measurements collected over time. The infrared
The breath-testing instrument used was a B A C Verifier Data-
breath test instrument employed was a BAC Verifier
Datamaster. A one-way analysis of variance and the Cochran's Muster (National Patent Analytical Systems, Inc.). The instru-
C for equal variances were applied to the data. The results ment quantitatively measures ethanol concentration in the vapor
indicated that when the solution volume remained at 500 mL, phase by infrared absorption and reports the measurenlent as
there was nearly as much or greater within-run variability as g/210 l.. An internal pump provides positive pressure to an at-
when the volumes were varied from 400 mL to 600 mL. Both tached simulator, thereby forcing the headspace vapor sample
the experimental and control groups showed statistical into the instrument's sample chamber. The simulator employed
significance for the one-way ANOVA and were considered was a Guth Model 34C (Guth Laboratories. Inc.) containing a
within-run measurements. The high within-run instrumental double baffle construction. The same siinulator and instrument
precision (CV approx. 1%) probably accounts for these results.
were used throughout for both the field and laboratory data. Op-
The field data, considered between-day measurements, had
erators are relied upon to confirm the simulator temperatt, re of 34
larger within-group variability and resulted in a nonsignificant
ANOVA. Small variations from the 500-mL volume in a _+ 0.2~ for field evidentiary tests. The temperatures were also
simulator do not result in a statistically significant difference confirmed for each group of laboratory tests.
where between-day measurements are evaluated. Three different groups of measurements were collected. Group
I, the experimental group, contained five alkluots of different so-
lution w~lumes; Group 2, the control group, contained five
aliquots each with the same solution volume: and Group 3. the
field group, consisted of the first 10 measurements (except one
Introduction
group where p~ = 8) from five different aliquots of solution pre-
Breath alcohol simulators are devices designed to simulate end pared lk)r field evidentiary use. The ethanol concentration for so-
expiratory human breath in that they provide a known concen- lutions used in Groups 1 and 2 was determined by gas chro-
tration of ethanol per 210 L of effluent to a breath-testing in- matography to be 0.1234 g/100 m L providing 0.102 g/210 L
strument for purposes of calibration and testing. The ethanol vapor concentration headspace. The ethanol concentration for the
concentration in the effluent vapor can be precisely controlled be- solution used in Group 3 was determined to be 0.1243 g/100 mL,
cause the temperature is regulated (34 _+0.2~ and the partition providing 0.103 g/210 L vapor concentration headspace.
coefficient for ethanol in water (k,/,,) is known (1,2). Group 1 comprised seven aliquots of solution consisting of
Simulators are the standard devices used for calibrating and 350, 400, 450, 500. 550, 600, and 650 mL. Ten measurements
testing all types of breath-testing instruments (3). Some of their were made of the headspace vapor sample tor each aliquot of so-
deficiences in precision have been discussed ( I ). However, many lution (with the exception of 350 mL and 650 mL) with the
of these have been corrected in modern simulator devices (4). BAC Vertifier DataMaster. Each group of 10 measurements look
Simulator manufacturers design the device to hold 500 mL of from 9 to 11 rain to complete and all measurements were com-
solution (5). This appears to to pro,,ide an adequate volume so pleted in one day.

332 Reproduction (photocopying) of editorial content of this journal is prohibited without publisher's permission
Journal of Analytical Toxicology, Vol. 15, November/December 1991

Group 2 comprised five aliquots of the solution placed into cause the simulator had an inadequate solution volume to allow
the simulator, consisting of 500 mL each. Ten measurements test completion, and the heating element in the simulator began
were again made with each of these solutions. Each group of to overheat. Sirnilarly, data from the 650-mL aliquot in Group 1
10 measurements took 10 rain to complete. All measurements in could not be collected because solution was emitted from the
Group 2 were completed in the one day following Group I mea- outlet port during pump operation.
surements. The mean values for the field solutions (Group 3) are shifted
Group 3, the field group, consisted of the first l0 simulator from those for the experimental and control groups, because a
measurements (one group consisted ofn = 8) on each of five dif- different solution batch having a different ethanol concentra-
ferent aliquots of simulator solutions under field testing condi- tion was employed in the field group.
tions. The volumes of these solutions were approximated by the Table II shows the results of the one-way ANOVA for each of
personnel preparing them from the State Toxicology Labora- the three groups. An important assumption in analysis of vari-
tory. Each group of 10 measurements was conducted over a pe- ance is the equality of variance among the sets within each
riod of time ranging from approximately one to four days. This group. This was tested with Cochran's C test and the results
field group represented the typical long-term variability expected were not significant, as shown in Table It. The volumes of the 10
under field conditions and would be composed of both within- solution aliquots selected at random were 480 mL + 10 mL
run and between-day variability. (Mean + SD) and ranged from 428 to 502 mL.
The mean, standard deviation, and coefficient of variation
were calculated for each sample of 10 measurements in all
groups. A one-way analysis of variance (ANOVA) was per-
formed within each of the three groups resulting in an F-statistic.
A Cochran's C statistic was then calculated to evaluate the ho- Discussion
mogeneity of variances within each of the three groups.
The volumes from 10 solution aliquots prepared for field use The purpose of the present study was to evaluate the effect of
were measured by type A volumetric cylinder with scale reso- varying simulaltor solution volume upon measurement results.
lution of_+10 mL. The solution aliquots were selected at random Changes in both the measurement means and variability were
from 10 different batches. The jars were visibly sealed with ev- factors considered. The one-way ANOVA evaluates a significant
idence tape. change in measurement means while the Cochran's C test eval-
uates significant differences in variability.
Table I and Figure 1 reveal that the range for group means in
both the experimental group and the control group were similar
at 0.0020 and 0.0017 g/210L, respectively. Varying the simulator
Results volume did not appear to affect how the measurement means
varied. The standard deviations lor each set of measurements, on
Table I shows the descriptive data analysis resulting from the other hand, appeared to vary more between the groups. The
measurements within each of the three groups. Figure 1 shows standard deviations for the experimental group ranged from
the mean along with the range in a vertical bar chart format for 0.0006-0.0010 g/210 L while the control group varied from
each solution aliquot in the three groups. We were unable to 0.0006-0.0016 g/210 L. This greater variability among mea-
collect data from the aliquot containing 350 mL in Group 1 be- surement standard deviations in the control group resulted in the

Table I. Descriptive Statistics for Measurements from Simulator Solutions of Varying and Constant Volumes
Collected under Various Conditions

Solution
Group volume (mL) Mean (g/210 L) Range SD CV (%)

I. Experimental 400 0.1045 0.103-0.106 0.0010 0.96

450 0.1032 0.102-0.104 0.0006 0.58
500 0.1052 0.104-0.106 0.0008 0.76
550 0.1034 0.102-0.104 0.0008 0.77
600 0.1041 0.102-0.105 0.0009 0.86

II. Control 500 0.1024 0.100-0.104 0.0016 1.6

500 0.1031 0.101-0.105 0.0011 1.1
500 0.1039 0.103-0.105 0.0006 0.58
500 0.1034 0.102-0.105 0.0010 0.97
500 0,1041 0.I02-0.106 0.0011 1.1

t11. Field 500" 0.0985 0.096-0.102 0.0018 1.8

500" 0.0996 0.095-0.103 0.0025 2.5
500" 0.1001 0.097-0.105 0,0023 2.3
500* 0.0982 0.095-0.103 0.0027 2.7
500* 0.0981 0.096-0.103 0.0021 2.1

"Approximatevalue.

333
 Journal of Analytical Toxicology, Vol. 15, November/December1991

lower F Statistic seen in Table II. The F Statistic for the one-way group where simulator volumes remained constant and was
ANOVA evaluates the ratio of between-group variability to therefore probably due to other random factors. This demon-
within-group variability and the control group showed the strates that varying the simulator volume from 400 to 600 mL did
greatest within-group variability. not significantly alter the measurement system precision when
Standard deviations ranging from 0.0006--0.0016 g/210 L (CV compared to solution volumes that remained constant.
0.58-1.6%) are indicative of a very precise measurement system. Perhaps the most significant result is seen when considering
This is particularly significant since the Federal National the statistical tests done upon the field group of data. Tables I and
Highway Traffic Safety Administration standards for evidential II reveal larger standard deviations when compared with the
breath test instruments allow standard deviations of 0.0042 laboratory sets of measurements. For every set of field mea-
g/210 L for 10 measurements at 0.100 g/210 L (3). surements the standard deviations were greater and comprised
For both the experimental and control groups the F Statistic both within-run and between-day components of variability.
was significant at the 0.01 level, indicating significant differences Separating these components of variability can be done by
between the individual measurement means within each group. adding standard deviations or coefficients of variation in quadra-
This has resulted, however, from the highly precise measurement ture (9,10). The greater variability resulted in an F Statistic that
system with low standard deviations and needs to be considered was not significant (p > 0.05) and a Cochran's C value that was
in light of statistical versus real differences (6-8). likewise not significant, thus validating the assumption of equal
An important consideration in the one-way ANOVA is the variances.
assumption of equal variances among the measurement groups Measurement variability can result from many random and in-
evaluated. The Cochran's C statistic tests for equality of vari- determinate causes (11). One very important consideration in in-
ances and thus evaluates for proper application of the one-way terpreting precision estimates is the time frame in which the
ANOVA. For both the experimental and control groups, the as- measurements were made. Within-run measurements are typi-
sumption of equal variances hold at p > 0.05. The p values in cally made during short time intervals and at least on the same
Table II, however, show that the control group had greater vari- day. It is known that within-run measurements result in greater
ability among the measurement variances. This again was the precision than between-day (12). This is partly due to the reduced
time frame in which random factors can influence results, be-
Table II. Results of Statistical Analysis Among the cause they tend to remain more stable over shorter time intervals.
Three Data Groups The larger variability due to between-day measurements would
probably obscure any variability that changes in simulator
0ne-way ANOVA C0chran's volume would contribute.
Group F df p C p The present study demonstrates that when keeping the simu-
lator solution volume constant from aliquot to aliquot there is
Experimental 9.7 49 <0001 0.27 0.91 greater variability in the measurements that when varying the
Vol = 400-600 mL volumes from 400 to 6(X) mL. This result, however, is a statistical
Control 4.8 49 0.003 0.41 0.074
artifact and not a real difference. Likewise, the F and Cochran's
Vol = 500 mL
Field 1.5 47 0.23 0,27 1.0
C statistics need to be evaluated in light of highly precise within-
run data. The real world of between-day breath test instrument

Measurement
Results

g/210 L GROUP 1 GROUP 2 GROUP 3

EXPERIMENTAL CONTROL FIELD
RV
0.107 0.i03

0.106 0.102

0.105 [ 0.I01

0.104 0.I00

0.103 0.099

0.102 0.098

0.I01 0.097

0.i00 0.096

400 450 500 550 600 500 500 500 500 500 500 500 500 500 500

RV = REFERENCE VALUE SOLUTION VOLUMES (ml)

Figure 1. Vertical bar chart showing mean _+one standard deviation for each set of 10 measurements collected under the different conditions.

334
Journal of AnalyticalToxicology,Vol. 15, November/December1991

performance, which is actually the better evaluation of system there will not be measurable differences due to variations in
precision (12), is largely variable and will tend to obscure any simulator volumes. Solution volumes of approximately 500 mL
statistical significance produced by highly precise within-run are adequate for forensic purposes.
measurements. It is important to distinguish between statistical
and real differences when evaluating data subjected to inferen-
tial analysis.
The evaluation of simulator solution volumes prepared for field
evidentiary use indicates an adequate approximation to 500 mL. References
The variability is certainly not enough to measurably alter in-
strument results. 1. K.M. Dubowski. Breath-alcohol simulators: Scientific basis and
actual performance. J. Anal ToxicoL 3:177-82 (1979).
2. A.W. Jones. Determination of liquid air partition coefficients for di-
lute solutions of ethanol in water, whole blood, and plasma. J.
Anal ToxicoL 7:193-97 (1983).
Conclusions 3. National Highway Traffic Safety Administration, DOT. Highway
safety programs; model specifications for evidential breath testing
The present study evaluated the effect of varying simulator so- devices: Publication of a conforming products list. Fed. Reg.
lution volumes upon measurement results. These results were 49:48855-864 (1984).
then compared to a group of field data collected under typical 4. K.M. Dubowski and N.A. Essary. Evaluation of commercial
breath alcohol measurement conditions. The data reveals that the breath-alcohol simulators: Further studies. J. Anal ToxicoL 15:
group maintaining constant simulator volumes (500 mL) showed 272-75 (1991).
more within-group (n = 10) variability than the group that varied 5. Guth Laboratories, Inc., Model 34C Simulator'. Owners Manual
volumes from 400-600 mL. As a result, the F Statistic was R.U. Guth, Harrisburg, PA, 1985.
larger for the control group due to less within-group variability. 6. R.M. Royall. The effect of sample size on the meaning of signif-
icance tests. The American Statistician 40(4): 313-14 (1986).
Both the experimental and control groups showed exceptionally
7. D.A. Grant. Testing the null hypothesis and the strategy and tac-
high precision (CV < 1.64%) due to the short time interval of
tics of investigating theoretical models. Psychological Rev.
measurement acquisition. Caution needs to be exercised in in- 69(1): 54-61 (1962).
terpreting statistical significance under these conditions since 8. D.R. Cox. Some problems connected with statistical inference,
they may not be indicative of a real difference due to solution Ann. Mathematical Statistics 29:357-72 (1958).
volume. Further insight is gained by viewing the data collected 9. J.R. Taylor. An Introduction to Error Analysis: The Study of Un-
under field conditions over a longer time interval and using vol- certainties in Physical Measurements, University Science
umes (approximately 500 mL) prepared under normal operating Books, Mull Valley, CA, 1982, pp. 40-98.
conditions. The within-group variability was much larger (al- 10. C.G. Fraser. Acceptable analytical error: A different opinion. Clin.
though still very good with CV < 2.7%) and resulted in a non- Chem. 35 (7): 1553-54 (1989).
11. N.C. Barford. Experimental Measurements: Precision, Error
significant F Statistic for the one-way ANOVA. The larger vari-
and Truth, John Wiley and Sons, New York, 1985, p. 86.
ability associated with between-day field testing conditions will
12. J.K. Taylor. Quality Assurance of Chemical Measurements,
exceed the variability associated with variations in the simu- Lewis Publishers, Chelsea, Michigan, 1987, p. 83.
lator solution volume. Under field testing conditions, where sim- Manuscript received August 2, 1991 ;
ulator tests are conducted with each breath alcohol analysis, revision received February 12, 1991.

335
View publication stats