The Effect of Sodium Chloride Concentration, Water Content,
and Protein on the Gas Chromatographic Headspace Analysis
of Ethanol in Plasma
MARK T. WATTS, PH.D. AND OLIVIA L. MCDONALD, CLT(HEW)
The authors used gas chromatographic headspace analysis to
study the sodium chloride concentration dependence of the partitioning of acetonitrile, ethanol, n-propanol, and t-butanol from
water and plasma to headspace vapor. Increasing the sodium
chloride concentration caused logarithmic increases in the partitioning. At 25 °C the slopes (logi0|peak height|/mol sodium/
L) obtained with the use of water or plasma were as follows:
acetonitrile, 0.064 (0.059); ethanol, 0.126 (0.125); n-propanol,
0.152 (0.149); and t-butanol, 0.200 (0.183). Differences in water
content between the two liquids may contribute to the small differences in the regression data. More importantly, saturation
with sodium chloride at 25 °C produced solutions with different
sodium molarities: 5.2 mol/L for water and 4.8 mol/L for plasma.
This difference in salt concentration at saturation and the volatile
dependent slopes can account for a large part of the error in
plasma ethanol concentrations when measured with the use of
aqueous external standardization and internal standardization
with any of the other volatiles. Deproteinization of the plasma
abolished the liquid phase-dependent differences in saturated
salt concentration and partitioning. (Key words: Gas chromatography; Headspace; Ethanol; Plasma; Sodium chloride) Am J
Clin Pathol 1990;93:357-362
GAS CHROMATOGRAPHIC HEADSPACE analysis is
a common technique for the quantitation of ethanol and
other volatile compounds in biologic fluids.2,4,7 The type
of specimen (e.g., whole blood, serum, urine, vitreous humor) plays a role in the analysis because the nature of the
liquid phase containing a volatile compound affects the
degree of partitioning into the headspace vapor.4 Toxicologists frequently saturate aqueous biologic specimens
with sodium chloride in an attempt to equalize the partitioning from different liquid phases.4,7 This significantly
enhances the partitioning into the vapor phase (saltingout) and tends to negate solute concentration differences
between specimen types.4
However, we previously reported that notwithstanding
the use of saturated sodium chloride, the type of biologic
specimen influenced the partitioning between liquid and
headspace vapor of several volatiles.8 We showed that
substantial error can be made in the analysis of ethanol
Department of Pathology, Texas Tech University Regional
Academic Health Center and Thomason Hospital,
El Paso, Texas
and that the magnitude of the error depended on the liquid
phase used for preparation of standards and the choice of
internal standard. We undertook the present investigation
for the purpose of determining the cause(s) of this "matrix" effect.
This report documents our findings on the salt concentration dependence of the partitioning of ethanol and
three other volatiles that are frequently used as internal
standards in the analysis of ethanol:3 acetonitrile, n-propanol, and t-butanol. The results indicate that for a given
sodium molarity, the partitioning of the volatiles is very
similar for both water and plasma. However, the simple
but striking finding was that the sodium molarities in saturated solutions of water and plasma were different.
Plasma exhibited lower saturated salt concentrations and
correspondingly lower partitioning of volatiles into headspace vapor. The difference in salt concentrations correlates with the degree of unequal partitioning previously
observed.8 In addition, the partitioning of the various volatiles was affected to differing extents by this difference
in salt concentrations. This is consistent with the observed
inability of internal standardization to negate this matrix
effect.8
We also investigated the difference in water content
between plasma and water for a given salt concentration
and found that this may contribute to the small matrix
effects observed at equal salt concentrations. Finally, the
extent to which protein affects saturated salt concentrations, water content, and volatile partitioning was assessed
by removing protein from plasma. This treatment abolished the aforementioned matrix-dependent effects.
Materials and Methods
Received March 29, 1989; received revised manuscript and accepted
for publication August 21, 1989.
Address reprint requests to Dr. Watts: Department of Pathology, Texas
Tech University Regional Academic Health Center, El Paso, Texas 79905.
Many of the materials and methods are the same as
those described previously.8 We used an Aerograph Series
2100 Gas Chromatograph® (Varian, Walnut Creek, CA),
with flame ionization detection. The oven was operated
357
358
WATTS AND MCDONALD
isothermally at 165 °C, and the detector and injector ports
were set at 200 °C. The column packing was 80/100 mesh
Poropak S (Supelco, Bellefonte, PA), and prepurified nitrogen was used as the carrier gas (45 mL/minute). One
milliliter of headspace vapor was injected with the use of
a Pressure-Lok® series "A" 2-mL gas syringe (Dynatech
Precision Sampling Corporation, Baton Rouge, LA). In
these experiments all headspace equilibrations were performed at 25 °C.
The biologic specimen chosen for study was plasma
because we could obtain a large supply and because we
previously found a large matrix effect for serum. Several
bags of outdated fresh-frozen plasma (from whole blood
preserved with citrate-phosphate-dextrose-adenine-one
[CPDA-1 ]) were pooled, aliquoted, and stored frozen so
that all the experiments could be performed on the same
plasma matrix. Headspace vapor above the plasma pool
alone contained no interfering chromatographic peaks.
Individual aqueous stock standards of each of the volatiles were prepared by mixing either 13.4 mL acetonitrile,
20 mL ethanol, 22 mL n-propanol, or 7.5 mL t-butanol
with sufficient deionized water to total 100 mL. These
concentrations were chosen to provide approximately
equal headspace peak heights for all the compounds in
deionized water.
The experimental procedure was designed to provide
maximum accuracy and precision in the preparation of
water and plasma specimens that were closely matched
in volatile concentration and sodium concentration. We
first labeled and weighed on an analytic balance (Mettler
Instrument Corporation, Hightstown, NJ) 10-mL draw
Vacutainer® evacuated blood collection tubes (BectonDickinson, Rutherford, NJ). We then weighed amounts
of sodium chloride into two sets of these tubes that would
produce solution with sodium concentrations of approximately 150, 500, 1,000, 2,000, 3,000, and 4,000 mmol/L.
The plasma matrix already contained sodium at a concentration near 150 mmol/L, so no salt was added to
produce the lowest concentration, and for the higher concentrations less sodium chloride was added to allow for
the endogenous sodium. Six milliliters of the liquid phase
(either deionized water or plasma) was then added with
the use of class A volumetric pipettes. The tubes were
weighed again. Because of the addition of the salt, the
tubes now contained different volumes. We refer to these
tubes as "salt tubes." Another set of empty 10-mL tubes
was labeled and weighed; into each tube we pipetted 5
mL (class A volumetric pipettes) of the contents of the
salt tubes. These tubes, which we refer to as "spike tubes,"
were then reweighed. We then added 30 yiL (Pipetman®,
Rainen Instrument Company, Woburn, MA) of the
aqueous volatile standard and reweighed the spike tubes.
At this point we had six tubes each for plasma and water
that contained the same volatile concentration but different salt concentrations. The sodium concentrations for
A.J.C.P. • March 1990
the two liquid phases were closely matched and could be
calculated from the gravimetric data. Saturated solutions
of sodium chloride in plasma and water were prepared at
25 °C by addition of an excess of sodium chloride to
approximately 5 mL of the liquid phase in a 10-mL tube.
The tubes were capped and rotated (Multi-Purpose Rotator Model 150V, Scientific Industries, Inc., Bohemia,
NY) for approximately 30 minutes. The tubes were centrifuged and 5 mL withdrawn and placed in preweighed
10-mL tubes. These spike tubes were then weighed, and
30 fiL of the volatile standard was added and a final
weighing performed. We also prepared a spike tube containing only 5 mL of deionized water and 30 nL of the
volatile standard. The final nominal millimolar/liter (mg/
dL) concentrations of the volatiles were as follows: acetonitrile, 15.6(64); ethanol, 20.6(95); n-propanol, 17.5
(105); and t-butanol, 4.9 (36).
Headspace tubes were prepared by pipetting 1 mL
(Medical Laboratory Automation, Inc., Mount Vernon,
NY) of the contents of the spike tubes into each of two
preweighed 10 mL tubes. These tubes were then weighed,
vortexed, and placed in the water bath to equilibrate. Two
injections of headspace were performed for each headspace tube, resulting in four peak height measurements
for every salt concentration and liquid phase. These data
were averaged and normalized for each volatile by assigning 100 to the value of the peak height from the headspace vapor above deionized water. This allowed easy
comparison between volatiles.
Coefficients of variation (CVs) for the various steps in
the preparation and analysis of the headspace tubes were
as follows: addition of 6 mL to the salt tube, < 0.1%;
addition of 30 nL of volatile standard, < 1.0%; addition
of 1 mL to the headspace tube, < 1.0%; and peak height
averages, < 2.0%. Similarly, the CV for the percentage by
weight of added sodium chloride for the experiments with
the different volatiles was < 0.5%.
Sodium molarities of the solutions in the spike tubes
were measured with a CIBA-CORNING 614® ion-selective electrode analyzer (Medfield, MA). For all concentrations above 150 mmol/L, dilutions with deionized water were required to bring the sodium concentration into
the measurement range of the analyzer (80-200 mmol/
L). The analyzer uses direct potentiometry, but the large
dilutions required for the higher concentrations greatly
reduced the possibility of differences between water and
plasma because of differences in excluded volume.5 Measured sodium concentrations agreed between experiments
with the different volatiles with a CV of < 2%.
We placed a portion of our plasma pool into a boiling
water bath to coagulate the protein. Expression of the
fluid from the congealed solution produced a solution
having a low protein concentration (approximately 5 g/
L). We refer to this solution as deproteinized plasma.
Measurements of several analytes in the plasma before
GAS CHROMATOGRAPHIC ANALYSIS OF ETHANOL
Vol. 93 • No. 3
and after this treatment were performed with the use of
a PARAMAX® Analytical System (Baxter Healthcare
Corporation, Irvine, CA). Osmolality was determined with
an Advanced Laboratory Wide-Range Osmometer®,
Model 3W11 (Advanced Instruments, Inc., Needham
Heights, MA).
Water content of the solutions in the spike tube was
assessed by pipetting 200 nh (Pipetman®) into preweighed
13 X 100 mm glass tubes (American Scientific Products,
McGraw Park, IL). The tubes were reweighed and placed
in a 110 °C TempCon® oven (American Scientific Products) overnight, then cooled and reweighed. The water
content, Cw (g/L), was calculated from the difference in
weights before and after heating, and the volume of the
specimen.
We also performed an experiment with randomly selected patient serum specimens. Twenty-one specimens
were saturated with NaCl at 25 °C. Spike tubes were then
prepared as described above (scaled down from 5 to 2
mL) with the use of a volatile standard containing both
ethanol and n-propanol. Headspace peak heights, salt
concentrations, and water content were determined for
each as described above. A water specimen was also saturated with NaCl and carried through the same series of
steps.
Results
Figure 1 displays the normalized peak heights versus
the measured sodium concentration for the four volatiles.
Each volatile displayed a logarithmic increase in partitioning with linear increases in the sodium concentration.
Linear regression analysis yielded the data shown in Table
1. All correlation coefficients were greater than or equal
to 0.998. For each of the volatiles there was close agreement in the slopes and intercepts between water and
WATER
1000-
o
PLASMA
800-
o
a
•
600-
0
»
ETHANOL
A
A
AcETONITRIlE
•
r-BuTAN0L
H-PROPANOL
•
0
#
400-
4
•
A
0
9
1(1(1-
0
A
A
200-
|l
A
0
-
1
1
.
—
—
• •••
2
•
• • • — • ,
•
3
SODIUM
•
4
5
6
MOU/L
FIG. 1. Headspace peak height (log scale) versus sodium concentration.
Peak heights have been normalized by assigning 100 to the peak height
observed using deionized water as the liquid phase. All headspace equilibrations were at 25 °C. The highest sodium concentrations are those
obtained with saturation by sodium chloride at 25 °C.
359
Table 1. Linear Regression Analysis of the
Data in Figure 1 *
Volatile
Liquid Phase
Slope (1 SD)
Intercept (1 SD)
Acetonitrile
Water
Plasma
Water
Plasma
Water
Plasma
Water
Plasma
0.064(0.0012)
0.059 (0.0005)t
0.126(0.0030)
0.125(0.0037)
0.152(0.0043)
0.149(0.0042)
0.200 (0.0029)
0.183 (0.0033)t
2.004(0.0031)
2.036 (O.OOU)t
2.021 (0.0078)
2.050 (0.0098)t
2.034(0.0111)
2.063 (0.011 l)f
2.016(0.0075)
2.060 (0.0087)t
Ethanol
n-Propanol
t-Butanol
• Slopes are expressed in units of Logic (Normalized Peak Height/mole sodium/L); intercepts
in units of Logio {Normalized Peak Height).
t Statistically significant difference from water, P < 0.01 by Mest.
plasma. Therefore, for a given sodium concentration the
partitioning of the volatiles is very similar for the two
liquid phases. However, it was repeatedly observed in the
experiments with the different volatiles that the measured
sodium concentration in the saturated solutions differed
between water and plasma: 4,824 + / - 111 mmol/L in
plasma; 5,236 +/— 91 mmol/L in water (+/— 1 standard
deviation, n = 4, P < 0.01 by Mest). This results in discrepancies between the two liquid phases in the partitioning of the volatiles when the liquids are saturated with
sodium chloride. The magnitude of the discrepancy depends on the slope that a given volatile exhibits. t-Butanol
exhibits the largest matrix effect because it has the highest
slope: for a given change in salt concentration there is a
large change in the partitioning.
Although the agreement of the slopes and intercepts is
close, it is not identical. Close inspection reveals that the
use of plasma as the liquid phase results in slightly lower
slopes and slightly higher intercepts. The data for all four
volatiles yielded statistically significant differences between
liquid phases in the intercepts (see Table 1). Figure 2A
displays the data for acetonitrile on an expanded scale to
clearly show the difference.
One possibility for the difference lies in the fact that
for a given sodium concentration, plasma has a lower
water content than water because of the presence of other
solutes. If one assumes that the volatile molecules are not
bound to or sequestered by any solute molecules,1 then
one can apply a mathematical correction to the data to
allow for the lower water content of the plasma specimens.
It follows from the assumption that the volume occupied
by the other solutes is unavailable to the volatile compound. This "excluded volume" hypothesis predicts that
for equal concentrations of volatile in water or plasma,
expressed in amount per unit volume of total solution,
there is a higher concentration of the volatile in the
aqueous portion of the plasma. This should result in
greater peak heights for plasma specimens than for water
specimens at a given salt concentration. A simple correction can be made by multiplying the peak heights by the
ratio of the water content of the salt solution divided by
A.J.CP. • March 1990
WATTS AND MCDONALD
360
A
Table 2. Sodium Concentrations and Water Content
of the Experimental Solutions
A
200-
Measured Sodium
Concentration
(mmol/L)
A
£
A
A
•
I
O
UJ
I
A
ACETONITRILE IN
&
<
A
PLASMA
A
WATER
a
*
A A
100-
&
1
2
3
SODIUM
4
MOLE/L
3
SODIUM
4
MOLE/L
i
Water
Plasma
Water
Plasma
Difference
148
482
932
1,869
2,768
3,817
5,236
152
497
950
1,861
2,825
3,741
4,824
992
987
977
956
939
913
883
938
931
923
903
884
863
837
54
56
54
53
55
50
46
1
B
200
i
<
&
or.
a.
O
O
1002
Water Content (g/L)
5
FIG. 2. A The data from Figure 1 for acetonitrile plotted on an expanded peak height log scale. B. The same data corrected for water content
(see text for details of correction).
the water content of deionized water. Table 2 lists the
measured sodium concentrations and water contents of
the solutions. At sodium concentrations less than 3,000
mmol/L, the water content of plasma solution is approximately 55 g/L less than the water content of an aqueous
solution with a comparable salt concentration. Figure 2B
displays the data for acetonitrile adjusted for water content. The linear regression data for the corrected peak
heights for all the volatiles are shown in Table 3. The
slopes for plasma are still slightly lower than those for
water, but the differences in the intercepts are greatly reduced (the number of statistically significant differences
is reduced by this correction). This results in very close
agreement between liquid phases in partitioning at any
given sodium concentration.
The addition of sodium chloride to plasma decreases
the concentration of all other analytes because the volume
increases. Measurements of glucose, cholesterol, triglyceride, total protein, and albumin showed steady decreases
in concentration with increasing sodium concentration.
Our gravimetric data allowed us to calculate the densities
of the solutions in the spike tubes and thereby correct the
measured concentrations of these analytes for volume expansion. This calculation demonstrated that there were
no large changes in the amounts of these analytes as salt
was added. Table 4 lists the corrected concentrations. The
data indicate that a small amount of protein may be lost
to precipitation (slight cloudiness was observed with saturated solutions). The decrease in the difference in the
water contents at high salt concentrations (Table 2 data)
also suggests this. However, there is the possibility that
high salt concentrations affect the PARAMAX methods
(e.g., note the "apparent" increase in the corrected glucose
concentration at saturation).
Heat treatment of the plasma produced a solution that
differed from plasma in protein concentration. Table 5
lists some of the characteristics of the two fluids. Cholesterol and triglyceride demonstrated 75% and 90% reductions, respectively, on heat treatment, presumably because
of the precipitation of lipoproteins. Other small molecular
weight compounds and electrolytes (glucose, potassium,
urea nitrogen, total carbon dioxide, creatinine, calcium,
phosphorus, magnesium, and iron) showed no change,
thereby maintaining an osmolality similar to that of the
untreated plasma. Protein was not completely removed
Table 3. Linear Regression Analysis of the Data in
Figure 1 after Correction for Differences
in Water Content*
Volatile
Liquid Phase
Slope (1 SD)
Intercept (1 SD)
Acetonitrile
Water
Plasma
Water
Plasma
Water
Plasma
Water
Plasma
0.051 (0.0014)
0.048 (0.0010)t
0.116(0.0031)
0.115(0.0036)
0.142(0.0040)
0.138(0.0045)
0.187(0.0031)
0.172 (0.0036)t
2.005 (0.0037)
2.012 (O.0027)f
2.020 (0.0079)
2.027 (0.0097)
2.032(0.0104)
2.039(0.0120)
2.018(0.0082)
2.036 (0.0095)f
Ethanol
n-Propanol
t-Butanol
* Slopes are expressed in units of Log,0 (Normalized Peak Height/mole sodium/L); intercepts
in units of Logio (Normalized Peak Height),
t Statistically significant difference from water, /* < 0.01 by /-test.
vol.93-No.3
GAS CHROMATOGRAPHIC ANALYSIS OF ETHANOL
361
Table 4. Effect of Sodium Chloride Addition on the Concentrations of Plasma Solutes
Corrected Concentrations*
Sodium
(mmol/L)
Density
(g/mL)
Glucose
(mmol/L)
Cholesterol
(mmol/L)
Triglyceride
(mmol/L)
Total Protein
(g/L)
Albumin
(g/L)
152
497
950
1,861
2,825
3,741
4,824
1.022
1.034
1.054
1.090
1.127
1.163
1.198
26.5
26.4
26.5
26.5
26.7
26.3
28.5
4.14
4.12
4.07
4.07
4.09
4.07
4.22
1.21
1.19
1.21
1.18
1.13
1.13
1.03
59
58
58
58
56
57
57
36
34
33
32
31
29
30
• The values (other than sodium and density) have been corrected for the volume expansion caused by the addition of sodium chloride: tabulated value = measured value X (1.022/density).
but was greatly diminished in concentration. This caused
the water content to closely approach that of a 150 mmol/
L solution of sodium chloride in water. Partitioning experiments with t-butanol and acetonitrile in the heattreated plasma yielded the data shown in Figure 3. Close
agreement between water and heat-treated plasma is observed in the partitioning of these volatiles at a given salt
concentration. Similar data were obtained for ethanol and
n-propanol in saturated solutions. No slope or intercept
data gave statistically significant differences at the 99%
confidence level. Also, the concentration of sodium at
saturation was indistinguishable for the two liquid phases.
Finally, the water content was found to differ between
liquid phases at a given salt concentration by less than
6 g/L.
The saturated solution matrix effects described above
for the pooled plasma sample were also observed with the
use of individual patient serum specimens. Table 6 displays data showing that peak heights for n-propanol were
lower in all serum specimens compared with water.
Ethanol peak heights were equivocal for the two specimen
types, similar to our previous findings at 25 °C. 8 Sodium
chloride concentrations and water content were also lower
for the serum specimens. Because the degree of effect was
different for ethanol and n-propanol, the ratio of their
peak heights differed between serum and water. This
means that the concentration of ethanol in these patient
specimens, calculated with the use of the water specimen
as a calibrator and n-propanol as the internal standard,
is in error by approximately 10% (see values in Table 6).
Discussion
The results obtained on the plasma pool and individual
patient specimens corroborate our earlier findings with
saturated sodium chloride solutions: volatile partitioning
is dependent on the sample matrix despite the saturation
by NaCl.8 Each volatile displays a different degree of this
"matrix" effect, thereby rendering internal standardization
ineffective at negating the effect. For instance, our data
demonstrate that serum or plasma ethanol concentrations
derived from 25 °C headspace analysis with the use of
aqueous standardization, sodium chloride saturation, and
n-propanol internal standardization will be falsely elevated
by at least 10% (patient samples spiked with ethanol to
20.6 mmol/L [95 mg/dL] yielded an average of 23.1
mmol/L [106 mg/dL]—Table 6). Our previous studies
also demonstrated this magnitude of error over a wide
range of ethanol concentrations in postmortem blood.8
At 25 °C, ethanol in saturated sodium chloride solution
showed only a small matrix-dependent difference in partitioning (Fig. 1 and Watts and McDonald8). However,
the alcohols frequently used as internal standards in
ethanol assays, n-propanol and t-butanol, demonstrated
WATEB
IOOO800-
DEPROTEINIZED
PLASMA
o
•
&
A
O
/-BuTANOL
ACETONITRILE
•o
600-
<>
400-
o
Table 5. Clinical Chemical Parameters for the Plasma
and Heat-Treated Plasma
Parameter
Plasma
Heat-Treated Plasma
Sodium (mmol/L)
Water content (g/L)
Osmolality (mOsmol/kg)
Total protein (g/L)
Albumin (g/L)
152
938
313
60
37
146
990
317
5
None detected
*>
200-
8
IOO-
a
1
^
»a
1
2
3
SODIUM
4
MOLE/L
FIG. 3. Headspace peak height (log scale) versus sodium concentration
obtained for t-butanol and acetonitrile with water and heat-treated
plasma.
WATTS AND MCDONALD
362
Table 6. Matrix Effects Observed with Saturated NaCl
Solutions of Water and Individual Serum Specimens*
Parameter
Peak heights (arbitrary units)
Ethanol
n-Propanol
Ratio of peak heights
(ethanol/n-propanol)
Sodium chloride concentration
(mmol/L)
Water content (g/L)
Target ethanol concentrationf
Calculated ethanol concentrationf
Water
208 ± 1
225 ± 3
Serum
213 ± 10
202 ± 9
0.939 ± 0.027
1.052 ±0.022
5,168 ±29
879 ± 1
20.6 (95)
—
4,880 ± 56
825 ± 6
20.6 (95)
23.1(106)
* Water and 21 scrum specimens, all saturated with NaCl. Means and standard deviations,
t Units of mmol/L (mg/dL). The calculated concentration uses the data for the water specimen
as calibrator and n-propanol as internal standard.
large partitioning differences between water and serum or
plasma saturated with sodium chloride. Therefore, although internal standardization of an ethanol assay is designed to improve the accuracy, it in fact can cause substantial error if this matrix effect is ignored.
To further understand the cause(s) of this effect, we
studied the dependence of headspace peak heights on salt
concentration. Our observations of the logarithmic increase in volatile partitioning into the headspace with increasing concentration of sodium chloride are in agreement with known effects of salts on activity coefficients
of nonelectrolyte solutes in aqueous salt solutions.6 The
data also show that for any given salt concentration the
partitioning of each of the volatiles was very similar for
water and plasma. However, the most important finding
was the fact that plasma saturated with sodium chloride
has a different sodium molarity than water saturated with
sodium chloride. This simple result has practical implications for biologic ethanol determinations because the
common approach to the removal of matrix effects involves the use of saturated sodium chloride.3 To use
aqueous standardization and internal standardization with
n-propanol or t-butanol, one would have to obtain the
same salt concentration for the unknown patient specimen
and the aqueous standard. Saturation with sodium chloride does not achieve this goal.
If the predominant variable influencing the partitioning
is sodium molarity, then this finding alone can account
for most of the liquid phase-dependent differences observed with saturated solutions.8 Protein apparently plays
a role in determining the sodium concentration at saturation, because its removal abolished the differences in
partitioning and sodium concentrations. Even at equal
sodium molarities, protein may play a role in causing
small differences in partitioning by its effect on water content. Our results empirically show a better agreement between plasma and water when "corrected" for the difference in water content of the two liquids.
A.J.C.P. • March 1990
A paradoxic hypothesis can be made by further extension of this "excluded volume" idea to include allowance
for the presumption that sodium itself may have access
only to the aqueous compartment. 5 This being the case,
for a given volatile and sodium molarity, plasma will contain higher concentrations of both in the aqueous phase
than a matched aqueous sample. Molality is a convenient
unit for comparison of concentrations of solutes per
amount of solvent. It is very interesting to note that the
sodium molalities (calculated from the measured molarities and water contents) of plasma and water saturated
with sodium chloride are much closer than their molarities: 5,743 mmol/kg, plasma; 5,910 mmol/kg, water. In
other words, at saturation the concentration of sodium
in the aqueous compartment of plasma is near the concentration of sodium in the saturated water solution. The
molarity is substantially less because of the volume occupied by protein. If molality was the predominant variable affecting partitioning, then one would expect similar
partitioning for plasma and water at saturation. In fact,
one might even predict higher partitioning from plasma
because the molality of the volatile is higher in plasma;
however, the opposite is observed. Plots of headspace peak
heights versus sodium molality invariably demonstrated
poorer agreement between the liquid phases.
Application of these findings to routine headspace
analysis of ethanol calls for the preparation of standards
and unknowns with equal salt concentrations. The simple
approach of saturation with sodium chloride will lead to
error in biologic specimens containing protein. Alternately, the use of protein-based standards (e.g., whole
blood) may fortuitously produce standards and unknowns
having the same salt concentration at saturation, but this
should be verified by careful recovery studies.
References
1. Baselt RC. Disposition of toxic drugs and chemicals in man. 2nd
ed. Davis: Biomedical Publications, 1982:299-303.
2. Dubowski KM. Alcohol determination in the clinical laboratory.
Am J Clin Pathol 1980;74:747-750.
3. Dubowski KM. Analysis of ethanol: type C procedure. In: Sunshine
I, ed. Methodology for analytical toxicology. Cleveland: CRC
Press, 1975:145-154.
4. Hachenberg H, Schmidt AP. Gas chromatographic headspace analysis. Philadelphia: Heyden & Son, 1979:3-12.
5. Ladenson JH, Apple FS, Aguanno JL, Koch DD. Sodium measurements in multiple myeloma: two techniques compared. Clin
Chem 1982;28:2383-2386.
6. Long FA, McDevitt WF. Activity coefficients of nonelectrolyte solutes
in aqueous salt solutions. Chemical Reviews 1952;51:119-169.
7. Shaw RF. Methods forfluidanalysis—colorimetric and instrumental.
In: Garriott JC, ed. Medicolegal aspects of alcohol determination
in biological specimens. New York: Year Book Medical Publishers, 1988:131-144.
8. Watts MT, McDonald OL. The effect of biologic specimen type on
the gas chromatographic headspace analysis of ethanol and other
volatile compounds. Am J Clin Pathol 1987;87:79-85.