M. Thevis et al., Eur. J. Mass Spectrom. 14, 117–125 (2008)
Received: 4 June 2008 n Revised: 28 July 2008 n Accepted: 28 July 2008 n Publication: 1 August 2008
117
European
Journal
of
Mass
Spectrometry
Special Issue: Sports Drug Testing by Mass Spectrometry
Quantitative analysis of urinary glycerol
levels for doping control purposes using gas
chromatography-mass spectrometry
Mario Thevis,* Sven Guddat, Ulrich Flenker and Wilhelm Schänzer
Center for Preventive Doping Research, Institute of Biochemistry, German Sport University Cologne, Carl-Diem Weg 6, 50933 Cologne,
Germany. E-mail: [email protected]
The administration of glycerol to endurance athletes results in an increased fluid retention and improved performance, particularly
under hot and humid conditions. Consequently, glycerol is considered relevant for sports drug testing and methods for its detection
in urine specimens are required. A major issue in this regard is the natural occurrence of trace amounts of glycerol in human urine,
which necessitates a quantitative analysis and the determination of normal urinary glycerol levels under various sporting conditions. A
quantitative method was established using a gas chromatography/isotope-dilution mass spectrometry-based approach that was validated with regard to lower limit of detection (0.3 µg mL–1), lower limit of quantification (0.9 µg mL–1), specificity, linearity (1.0–98.0 µg mL–1),
intraday and interday precision (<20% at 2.4, 24.1 and 48.2 µg mL–1) as well as accuracy (92–110%). Sample aliquots of 20 µL were enriched
with five-fold deuterated glycerol, dried and derivatised using N-methyl-trimethylsilyltrifluoroacetamide (MSTFA) before analysis. The
established method was applied to a total of 1039 doping control samples covering various sport disciplines (349 endurance samples,
286 strength sport samples, 325 game sport samples and 79 other samples) in- and out-of-competition, which provided quantitative
information about the glycerol content commonly observed in elite athletes’ urine samples. About 85% of all specimens yielded glycerol
concentrations < 20.0 µg mL–1 and few reached values up to 132.6 µg mL–1. One further sample, however, was found to contain 2690 µg mL–1,
which might indicate the misuse of glycerol, but no threshold for urinary glycerol concentrations has been established yet due to the
lack of substantial data. Based on the results obtained from the studied reference population, a threshold for glycerol levels in urine set
at 200 µg mL–1 is suggested, which provides a tool to doping control laboratories to test for the misuse of this agent in elite and amateur
sport.
Keywords: sport, doping, mass spectrometry, plasma volume
Introduction
Glycerol (1,2,3-propanetriol) is the structural backbone of
numerous lipids, and its biochemically most important form is
glycerol-3-phosphate, which is essential for the biosynthesis
of triacylglycerols. Normal serum concentrations of glycerol in humans range between 4.6 µg mL–1 and 27.6 µg mL–1,1
and the majority of the circulating substrate results from
de-­esterification of triacylglycerols in adipose tissue and blood,2
which is utilised as a gluconeogenic precursor or oxidised via
mito­chondrial respiration. Its urinary excretion was ­considered
­ egligible at healthy physiological serum glycerol levels,
n
whereas an increased blood glycerol concentration caused by
oral or intravenous administration correlated significantly with
an elevated urinary elimination.3 In addition to the central role
of glycerol in lipid biosynthesis, glycerol had a widespread clinical utility in the treatment of cerebral oedema, glaucoma and
intracranial hypertension4–6 and since 1987, the co-administration of glycerol and fluid has been employed for hyperhydration purposes.7 The increased amount of total body water was
ISSN: 1469-0667 © IM Publications LLP 2008
doi: 10.1255/ejms.919 All rights reserved
118
reported in numerous studies; however, controversial results
concerning ergogenic effects and influences on thermoregulation or cardiovascular strains were reported.8–16 Nevertheless,
glycerol was considered relevant for doping controls and prohibited in sports, in particular in light of studies demonstrating
the performance-enhancing properties of glycerol loading for
instance in Olympic distance triathlon10 competition as well as
cycling,9 but enhanced sporting performance was only reported
for endurance exercise as comparable studies with tennis
players did not yield performance benefits.17
Numerous studies describing the gas chromatography-mass
spectrometry (GC-MS)-based analysis of glycerol in human
plasma were published for the purpose of monitoring lipolysis
and fat mobilisation in vivo. Various derivatives and ionisation
conditions were tested, including electron ionisation, positive
and negative chemical ionisation of triacetate, trimethylsilyl
(TMS), tert-butyldimethylsilyl (tBDMS) and heptafluorobutyr
(HFB) derivatives.18–22 Moreover, an assay employing liquid chromatography mass spectrometry (LC-MS) was shown to allow for
the determination of glycerol in human plasma,23 but only few
articles dealt with the qualitative or quantitative measurement
of glycerol in human urine.24,25 In order to test for normal physiological glycerol levels of a particular population composed by
elite athletes of different sport disciplines, 1039 doping control
urine samples were quantitatively analysed for glycerol using
a rapid and sensitive isotope-dilution GC-MS approach, which
was validated according to commonly accepted ICH guidelines.26 Based on these data, a threshold for urinary glycerol
concentrations might be established, enabling the identification
of glycerol administration to athletes.
Experimental
Chemicals and reagents
Glycerol (86%, p.a.) was purchased from Carl Roth
GmbH (Karlsruhe, Germany), 1,1,2,3,3-2H5-glycerol (98%)
from ISOTEC (Miamisburg, OH, USA), and N-methyl-Ntrimethylsilyltrifluoroacetamide (MSTFA, technical grade) from
Karl Bucher GmbH (Waldstetten, Germany). All solutions were
prepared in MilliQ grade water (Millipore, Eschborn, Germany).
Stock and working solutions
A stock solution of the internal standard (IS) was prepared
from 10 µL of 2H5-glycerol and 10 mL of water, which yielded
a concentration of 1.33 mg mL–1. The working solution was
obtained by a 1 : 20 dilution resulting in a concentration of
66.6 µg mL–1. Accordingly, a stock solution for glycerol was
made from 112.0 mg dissolved in 100 mL of water. Considering
the content of 86%, a concentration of 0.96 mg mL –1 was
obtained, which was further diluted to yield working solutions
of 1.9 µg mL–1 (solution 1) and 19.2 µg mL–1 (solution 2).
Urine samples
A total of 1039 urine samples was analysed for the presence
and quantity of glycerol. All specimens were regular doping
GC-MS Quantitation of Glycerol
control samples tested negative in all mandatory screening
procedures and derived from in-competition (IC) as well as
out-of-competition (OOC) testing of different sport disciplines:
349 endurance sport samples were obtained from biathlon,
speed skating, canoeing, athletics, pentathlon, cycling, swimming, rowing and triathlon with 207 IC and 142 OOC samples;
286 strength sport specimens were collected from boxing,
weight lifting, judo and wrestling with 190 IC and 96 OOC
samples; 325 game sport specimens included American
football, badminton, basketball, ice hockey, soccer, handball, hockey, rugby, tennis and volleyball with 206 IC and 119
OOC samples; 79 other urine specimens were obtained from
aerobic, bob sleigh, bowling, curling, fencing, gymnastics,
shooting, motor cycling, sailing, taekwondo and rock climbing
with 50 IC and 29 OOC samples.
Sample preparation
The sample preparation consisted of three steps only. A total of
20 µL of urine was placed in a glass reagent tube, fortified with
20 µL of the IS working solution (1.3 µg of 2H5-glycerol) and
dried in a dessiccator over phosphorus pentoxide in vacuo over
night. The dry residue was subsequently dissolved in 100 µL of
MSTFA, heated for 20 min at 60°C and transferred to GC-MS
vials after cooling to ambient temperature.
GC-MS analysis
GC-MS measurements of all samples were conducted on an
Agilent 6890 gas chromatograph interfaced to a 5973 mass
selective detector. The GC was equipped with a HP-5MS column
(inner diameter 0.2 mm, film thickness 0.2 µm, length 17 m),
and a temperature program starting at 95°C increasing by
20°C min–1 to 175°C and then by 40°C to 320°C was employed.
Helium was used as carrier gas (0.8 mL min–1, constant pressure) and the injector temperature was set to 300°C, the interface temperature to 320°C and the ion source temperature to
230°C. Two µL of the derivatised sample were injected in split
mode (1 : 10) and glycerol and the IS were detected after electron ionisation (EI) at 70 eV using selected ion monitoring (SIM)
at dwell times of 20 ms each. Three diagnostic ions per analyte
were considered for qualitative analyses according to current
WADA guidelines: glycerol m/z 293, 218 and 205; 2H5-glycerol:
m/z 298, 222 and 208. The quantifier ions were m/z 218 and 222
for glycerol and the IS, respectively.
Assay validation
The qualitative and quantitative determination of glycerol in
human urine was validated regarding specificity, lower limit of
detection (LLOD), lower limit of quantification (LLOQ), intraday
and interday precision, accuracy and linearity according to
ICH guidelines.26 Respective items were defined and tested
as follows:
Specificity
Ten different blank urine specimens of known origin (six male
and four female urine samples) were prepared as described
in order to probe for interfering peaks in the selected ion
M. Thevis et al., Eur. J. Mass Spectrom. 14, 117–125 (2008)
chromatograms at the expected retention times for glycerol
and the IS.
Lower limit of detection (LLOD) and lower limit of
quantification (LLOQ)
The LLOD was defined as the “lowest content that can be
measured with reasonable statistical certainty” at a signalto-noise ratio ≥ 3, while the LLOQ represents the lowest
quantifiable amount of an analyte, which is characterised
by a signal-to-noise ratio ≥ 9. Aliquots of ten different blank
urine samples with no detectable amount of glycerol were
spiked with the IS and an additional ten aliquots were fortified with 2.4 µg mL–1 of glycerol plus the IS. The samples
were prepared and analysed according to the established
protocol providing the data necessary to estimate the LLOD
and LLOQ.
Intraday precision
Within one day, ten urine samples spiked to low (2.4 µg mL–1),
medium (24.0 µg mL–1) and high (48.1 µg mL–1) concentrations
of glycerol were prepared and analysed and the intraday precision was calculated for each concentration level.
Interday precision
On three consecutive days, a total of 90 urine samples of low,
medium and high concentrations (2.4 1 µg mL–1, 24.0 1 µg mL–1
and 48.1 µg mL–1, respectively) were prepared and analysed
randomly and the assay precision was calculated for each
concentration level.
Linearity
A calibration curve prepared from a urine sample with no
detectable amount of glycerol was measured from 1.0 µg mL–1
to 98.0 µg mL–1 using ten calibration points.
Accuracy
119
The software used was R in the latest version (04-22-2008,
http://cran.r-project.org/src/base/R-2/R-2.7.0.tar.gz.)
Results and discussion
The utility of GC-MS for the determination of glycerol after
adequate derivatisation was demonstrated several times in the
past.18–22,24,25 In the present study, trimethylsilylation of glycerol
by means of MSTFA was chosen, which allowed a rapid and
straight-forward sample preparation consisting of the addition of the IS, evaporation and reconstitution in the derivatising
agent only. After incubation, an immediate analysis is possible
avoiding any extraction or additional purification step. The EI
mass spectra of glycerol and its five-fold deuterated analog
after trimethylsilylation are illustrated in Figure 1, which contain
typical fragment ions for instance at m/z 293, m/z 218, m/z 205
and m/z 103 [(a) glycerol-tris-TMS] or m/z 298, m/z 222, m/z 208
and m/z 105 [(b) 2H5-glycerol], the proposed origins and structures of which are depicted in Scheme 1. Common losses of 15 u
(–CH3) and 90 u (–TMSOH) from the molecular ion (not visible
in EI spectra) yield the fragment ions at m/z 293 and m/z 218
and cleavages of glycerol C–C bonds give rise to m/z 103 and
m/z 205 as complementary fragments. In addition, the ion at
m/z 117 is suggested to originate from the molecular ion by the
elimination of a methyl radical from a TMS residue (presumably
yielding an intermediately formed m/z 293) accompanied by the
release of TMS–CH2–O–TMS (–176 u). All proposed structures
are in accordance with the observed presence or absence of
deuterium atoms in corresponding fragment ions obtained
from 2H5-glycerol [Figure 1(b)]; however, definite structure identification would require additional experiments, for instance
using high resolution/high accuracy tandem mass spectrometry, which was not within the scope of the present study. In
addition to ions characterising glycerol, abundant fragments
originating from trimethylsilylation such as m/z 147, m/z 133
and m/z 73 were detected.
The accuracy of the method was determined from urine aliq470 Scheme 1
uots (100 mL each) spiked to 2.5 1 µg mL–1, 24.5 1 µg mL–1 and
49.0 µg mL –1, which were prepared and analysed471
ten-fold,
and the obtained results were correlated to the calibration
472
curve.
Data analysis
473
H2C O
TMS
m/z 205
HC O
TMS
m/z 103
H2C O
TMS
474
The effects of the type of control (IC vs OOC) and the sports
discipline (endurance sport, game sport, strength
475 sport,
other) were investigated by means of a two-factor analysis
476
of variance (ANOVA). Subsequently, a corresponding general
linear model was fitted. These calculations were based
477 on logtransformed concentrations, where samples below the LLOQ
arbitrarily were assigned a value of 0.9 µg mL–1 (see478Results
and discussion). The upper 99% reference limit for479
the total
population and the corresponding 99% confidence interval (CI)
480 bootwere estimated by bootstrap statistics. A total of 1999
strap replicates were performed and the 99% quantile
481served
as estimate for the reference limit. The CI was calculated
based on Gaussian approximation of the bootstrap 482
statistics.
483
484
485
486
H2C O
TMS
C O
TMS
H2C
m/z 218
.+
H2C O
TMS
HC O
TMS
CH3
+
H2C O Si
CH3
.+
m/z 293
HC O
CH3
+
H2C O Si
CH3
m/z 117
Scheme 1. Suggested origins and structures of characteristic
ions derived from glycerol-tris-TMS after electron ionisation.
120
GC-MS Quantitation of Glycerol
434
Figure 1
435
(a)
436
(b)
147
73
100
205
40
20
4559
20
439
440
117
103
133
60
60
218
177 191
89
100
Intensity (%)
Intensity (%)
438
208
140
180
60
105 120
40
133
20
45 60
293
220
260
300
40
340
443
445
446
1300
298
200
240
280
320
(d)
m/z 298
m/z 293
2.06
7000
2.04
75000
1000
5000
50000
2.04
60000
m/z 293
40000
3000
m/z 298
20000
500
300000
m/z 218
4000
200000
2000
100000
60000
900000
2.06
50000
2.04
m/z 222
2.04
240000
m/z 218
200000
m/z 222
30000
100000
40000
447
449
160
m/z (Da)
(c)
m/z 205
10000
2.04
130000
m/z 208
100000
2.06
2.04
800000
m/z 208
m/z 205
500000
600000
50000
20000
448
120
m/z (Da)
6000
444
222
179192
92
80
441
442
147
80
80
437
73
100
200000
300000
1.96
2.00
2.04
2.08
2.12
1.96
2.00
2.04
time (min)
2.08
2.12
1.96
2.00
2.04
2.08
time (min)
time (min)
2.12
1.96
2.00
2.04
2.08
2.12
time (min)
(e)
2.06
960000
2.06
5200
m/z 218
m/z 222
4000
600000
450
451
452
2000
300000
1.96
2.00
2.04
time (min)
2.08
2.12
1.96
2.00
2.04
2.08
2.12
time (min)
Figure 1. EI-MS spectra of (a) glycerol-tris-TMS (mol wtmonoisotopic = 308) and (b) 2H5-glycerol-tris-TMS (mol wt = 313); extracted ion chromatogram of (c) a blank urine sample, (d) a blank urine spiked to 2.4 µg mL–1 of glycerol and (e) a urine specimen containing approx.
100 µg mL–1 of glycerol but no IS to demonstrate the negligible influence of high glycerol levels on the IS quantifier ion at m/z 222.
453
454
455
456
By means of the observed characteristic fragment ions, the
detection, identification and quantification of glycerol in human
urine was accomplished. Typical extracted ion chromatograms
of a blank urine sample containing the internal standard only,
and a blank urine specimens spiked to 2.4 µg mL–1 with glycerol
are shown in Figure 1(c) and (d), respectively. Here, the adequate
chromatographic separation and identification of glycerol from
other urinary components using three diagnostic fragments
(m/z 293, m/z 218 and m/z 205) is demonstrated. Moreover, a
urine enriched with 50 µg mL–1 of glycerol was analysed without
IS to probe for the influence of high urinary glycerol levels on
the quantifier ion of the IS at m/z 222. As depicted in Figure 1(e),
a signal at m/z 222 is found resulting from unlabelled glycerol,
which accounts for approximately 5–7% of the normal intensity of the internal standard. However, due to the fact that the
enhancement of the quantifier ion of the IS is present in regular
urine specimens as well as in samples of the calibration curve,
the quantitative result is not distorted.
Assay validation
The method was validated according to ICH guidelines and the
obtained results are summarised in Table 1.
20
M. Thevis et al., Eur. J. Mass Spectrom. 14, 117–125 (2008)
121
Table 1. Summary of assay validation.
Intraday precision
(n = 10 + 10 + 10)
Interday precision
(n = 30+30+30)
Accuracy
(n = 10 + 10 + 10)
LLOD
(µg mL–1)
LLOQ
(µg mL–1)
Concentration
(µg mL–1)
CV (%)
Concentration
(µg mL–1)
CV (%)
Concentration
(µg mL–1)
%
0.3
0.9
  2.4
16.5
  2.4
16.8
  2.5
109.6
24.0
48.1
11.6
  8.7
24.0
48.1
15.9
  9.7
24.5
49.0
110.3
  92.3
Specificity
Although the renal elimination of glycerol was reported negligible under normal physiological conditions,3 trace amounts
were detected in most urine specimens analysed in the present
study. Hence, “blank” urine samples without any signal for glycerol were rare, but as the identity of the target analyte was characterised by gas chromatographic retention time and diagnostic
fragment ions it was shown that no other compound co-eluted
or interfered with the target analyte or the IS.
Lower limit of detection (LLOD) and lower limit of
quantification (LLOQ)
The detection and quantification limits were defined by signalto-noise ratios of ≥3 and ≥9, respectively, which were determined at 0.3 µg mL–1 (LLOD) and 0.9 µg mL–1 (LLOQ).
Intraday and interday precision
At three concentration levels (2.4 µg mL–1, 24.0 µg mL–1 and
48.1 µg mL –1) the intraday and interday precisions of the
method were determined, which were all <20% as outlined
in Table 1.
Linearity
Linearity of the quantitative analyses was demonstrated by
calibration curves prepared from spiked urine samples, the
natural glycerol content of which was below the LLOD. Within
a range from 1.0 µg mL–1 to 98.0 µg mL–1, the calibration curve
was found linear (y = 0.0163x + 0.056, r2 = 0.9941) according to the
Mandel-test.27
Accuracy
The accuracy was determined at three concentration levels
(2.5 µg mL–1, 24.5 µg mL–1 and 49.0 µg mL–1) by means of measured calibration curves and was found between 92.3% and
110.3% (Table 1).
Authentic doping control urine samples
A total of 1039 doping control urine samples tested negative in all routine screening procedures were measured
using the established assay for the quantification of glycerol, in order to determine normal physiological levels of
the target analyte in elite athletes’ urine specimens. An
entity of 653 IC and 386 OOC samples obtained from various
sport disciplines was studied to obtain a comprehensive set
of data that shall represent the population of interest and
their particular condition. In 237 samples (22.8%), glycerol
levels were below the LLOQ of 0.9 µg mL–1, 263 specimens
(25.3%) contained less than 4.5 µg mL–1, 236 (22.7%) between
4.5 µg mL–1 and 10 µg mL–1, 145 (14.0%) between 10.0 µg mL–1
and 20.0 µg mL –1 , 77 (7.4%) between 20.0 µg mL –1 and
30.0 µg mL–1, 67 (6.4%) between 30.0 µg mL–1 and 60.0 µg mL–1
and 13 (1.4%) between 60.0 µg mL–1 and 140.0 µg mL–1 (Table
2). A differentiation into IC and OOC samples further demonstrated that glycerol concentrations higher than 4.5 µg mL–1
were found significantly more often in IC samples (Figure
2), which is in accordance with literature data that demonstrated increased glycerol eliminations under demanding
physical exercise conditions that result in increased lipolysis
and fat ­mobilisation.1,11
The evaluation of specimens with regard to sport disciplines
and IC vs OOC samples is depicted in Figure 3. The endurance
sport specimens [Figure 3(a)] were composed by 207 IC and 142
OOC samples, and 190 (91.8%) of the IC as well as 141 (99.3%)
of the OOC specimens contained ≤20 µg mL–1. Strength sport
samples (160 IC, 93 OOC) yielded 160 (84.2%) IC and 93 (96.9%)
specimens with ≤20 µg mL–1 [Figure 3(b)], game sport samples
(206 IC, 119 OOC) gave rise to 127 (61.7%) IC and 103 (86.6%)
specimens with glycerol concentrations below 20 µg mL –1
[Figure 3(c)] and other sport discipline samples composed of
50 IC and 29 OOC tests showed a total of 40 (80.0%) IC and 28
(96.6%) OOC specimens with less than 20 µg mL–1 of glycerol in
urine [Figure 3(d)]. In summary, more than 94% of all OOC and
Table 2. Urinary glycerol concentrations found in a total of 1039
doping control urine samples.
Samples
Urinary concentration
(µg mL–1)
IC
OOC
Total
< 0.9
< 4.5
4.5–10.0
10.0–20.0
20.0–30.0
30.0–60.0
60.0–140.0
Sum
144
120
134
119
  65
  59
  12
653
  93
143
102
  26
  12
   8
   2
386
  237
  263
  236
  145
   77
   67
   14
1039
122
456
457
GC-MS Quantitation of Glycerol
Figure 2
79% of all IC specimens demonstrated glycerol concentrations
≤20 µg mL–1, and the highest amount determined in the 1039
urine samples serving as reference population was measured
at 132.6 µg mL–1 in a game sport specimen.
In Table 3, the results of the two-factor ANOVA performed
on the log-transformed glycerol concentrations are summarised, where the type of control and the sports discipline
represent the independent variables. Both factors, as well as
the interaction term were highly significant. Consequently, the
urinary glycerol concentration must be assumed to be largely
dependent on these factors. In particular, differences in value
obtained from OOC samples must be assumed with regard to
the selected sport categories endurance sport, strength sport,
game sports, and other disciplines. Moreover, the magnitudes of the changes of the urinary glycerol excretion during
competition will depend on the respective discipline. Table
4 details these effects. Endurance athletes exhibit relatively
high OOC glycerol values, but any change during competition
was negligible. In strengths sports, the glycerol level in OOC
samples was even higher and a slight, but statistically significant, increase was found in-competition. Representatives of
other disciplines and game sports yielded comparably low
glycerol concentrations in OOC specimens; however, in the
Figure 2. Distributions of the log-transformed urinary glycerol
concentrations grouped by the combinations of independent
variables (IC = in-competition, OOC = out-of-competition).
458
Figure 3
464
Figure 3 (cont.)
459
(a)
465
(c)
Game Sport
Endurance Sport
60
90
80
50
70
samples (n)
50
21
40
samples (n)
40
60
30
20
30
10
20
10
0
<0.9
0
<0.9
<4.5
4.5-10.0
10.0-20.0
20.0-30.0
30.0-60.0
461
4.5-10.0
10.0-20.0
20.0-30.0
30.0-60.0
60.0-140.0
µg/mL
µg/mL
in-competition samples
460
<4.5
60.0-140.0
in-competition samples
466
out-of-competition samples
out-of-competition samples
467
(b)
468
(d)
Strength Sport
Other Sport
60
16
50
14
12
samples (n)
samples (n)
40
30
20
10
8
6
4
10
2
0
0
<0.9
<4.5
4.5-10.0
10.0-20.0
20.0-30.0
30.0-60.0
60.0-140.0
<0.9
<4.5
4.5-10.0
10.0-20.0
µg/mL
462
463
in-competition samples
20.0-30.0
30.0-60.0
60.0-140.0
µg/mL
out-of-competition samples
469
in-competition samples
out-of-competition samples
470by sport sections and IC/OOC samples: (a) endurance sport, (b)
Figure 3. Differentiation of measured urinary glycerol concentrations
strength sport, (c) game sport and (d) other disciplines
22
23
M. Thevis et al., Eur. J. Mass Spectrom. 14, 117–125 (2008)
123
Table 3. Dependence of the log-transformed glycerol concentrations of 1039 doping control urine samples on the factors “type of control
(type)” (i.e. in-competition and out-of-competition) and “sports discipline” (two-factor ANOVA). Df: degrees of freedom, Sum Sq: sum of
squares, Mean Sq: mean sum of squares, F value: quantile of the F distribution, p: probability.
Type
Discipline
Type x discipline
Residuals
Df
Sum sq
Mean sq
F value
p
    1
    3
    3
1026
   71
   80
   84
1299
71
27
28
 1
  55.9
21
  22.2
< 0.001
< 0.001
< 0.001
latter the strongest increase in IC samples of all four studied
categories was observed. The distributions of the log-transformed concentrations grouped by the combinations of the
independent variables are illustrated as box-plots in Figure 2.
The distribution of the total of the urinary glycerol concentrations was far from Gaussian using untransformed as well as
log-scale values. Hence, the calculation of upper reference
limits by parametric statistics was impossible, but bootstrap
statistics represented a suitable alternative. The bootstrap
estimate of the 99%-quantile for the entire set of analyses
was 68 µg mL–1 and the corresponding upper 99%-confidence
interval amounted to 87 µg mL–1. In contrast to the reference
limit, the calculation of the latter parameter could be performed
by Gaussian approximation. Due to the non-parametric nature
of the bootstrap methodology, the fact that numerous samples
fell below the LLOQ was of no relevance. It can be assumed that
at most 1% of the routine samples exceed a concentration of
87 µg mL–1 and virtually no sample will exhibit concentrations
of more than 200 µg mL–1. Consequently, a practical—though
somewhat arbitrary—threshold value could be established
at urinary glycerol concentrations of 200 µg mL–1 for doping
control purposes as the measured data suggest considerably
lower “normal” levels in samples of the considered sport disciplines, if a common sport nutrition and medication is given.
Currently, a tool to determine the misuse of glycerol is not available, but the administration was suspected in a doping control
sample measured in 2007. Here, a urinary glycerol content of
2690 µg mL–1 was observed, but no medication explaining the
unusually high level was provided. As a valid threshold has not
been established yet, the report and sanction of an abnormal
analytical finding such as the unusually high amount of glycerol
of 2690 µg mL–1 has not been possible but should be considered
in future sports drug testing.
Conclusion
Urinary glycerol levels of 1039 elite athletes were measured to
provide reference data for “normal” concentrations found in inas well as out-of-competition samples of various sport disciplines. The consideration of inter-individual variations based on
different physiological disposition, amount and type of exercise
and physical performance, as well as nutrition and medication,
was accomplished by the randomly selected samples obtained
from endurance sport, strength sport, game sport and other
sections and none of these specimens showed glycerol concentrations higher than 140 µg mL–1. Consequently, a threshold
for urinary glycerol levels might be established by regulatory
bodies of sports drug testing to enable doping control laboratories to report abnormal analytical findings such as a urine
sample found to contain 2690 µg mL–1 of glycerol in 2007. In
addition, administration studies should be conducted and
samples obtained from patients receiving glycerol in therapeutic
amounts should be measured to obtain more data helping to
modify or fix the suggested cut-off level of 200 µg mL–1.
Acknowledgements
The authors thank David Schäfer for technical assistance and
the Manfred-Donike-Institute for Doping Analysis, Cologne,
for supporting the presented work.
Table 4. Coefficients of the linear model fitted to the log-transformed glycerol concentrations of 1039 drug testing specimens. OOC: out-ofcompetition samples, IC: in-competition samples.
Endurance, OOC
Strength, OOC
Game, OOC
Other, OOC
Endurance × IC
Strength × IC
Game × IC
Other × IC
Estimate
Std error
t value
p
  1.28
  1.46
  1.08
  1.03
–0.08
  0.31
  1.34
  0.68
0.1
  0.11
0.1
  0.21
  0.12
  0.14
  0.13
  0.26
13.34
12.75
10.44
  4.94
–0.67
  2.21
10.35
  2.59
  < 0.001
  < 0.001
  < 0.001
  < 0.001
  0.51
< 0.05
  < 0.001
< 0.01
124
References
1.R.A. Robergs and S.E. Griffin, “Glycerol—biochemistry,
pharmacokinetics and clinical and practical applications”, Sports Med. 26, 145 (1998). doi: 10.2165/00007256199826030-00002
2. W.M. Bortz, P. Paul, A.C. Haff and W.L. Holmes,
“Glycerol turnover and oxidation in man”, J. Clin. Invest.
51, 1537 (1972). doi: 10.1172/JCI106950
3.S. Sommer, R. Nau, E. Wieland and H.W. Prange,
“Pharmacokinetics of glycerol administered orally in
healthy volunteers”, Arzneimittelforschung 43, 744 (1993).
4. M.S. Frank, M.C. Nahata and M.D. Hilty, “Glycerol: a
review of its pharmacology, pharmacokinetics, adverse
reactions, and clinical use”, Pharmacotherapy 1, 147 (1981).
5.S.L. Wald and R.L. McLaurin, “Oral glycerol for the
treatment of traumatic intracranial hypertension”, J.
Neurosurg. 56, 323 (1982).
6. J.S. Meyer, J.Z. Charney, V.M. Rivera and N.T. Mathew,
“Treatment with glycerol of cerebral oedema due to acute
cerebral infarction”, Lancet 298, 993 (1971). doi: 10.1016/
S0140-6736(71)90321-7
7. B. Freund, S. Montain, A. Young, M. Sawka, J. DeLuca, K.
Pandolf and C. Valeri, “Glycerol hyperhydration: hormonal, renal, and vascular fluid responses”, J. Appl. Physiol.
79, 2069 (1995).
8.T. Scheett, M. Webster and K. Wagoner, “Effectiveness of
glycerol as a rehydrating agent”, Int. J. Sport Nutr. Exerc.
Metab. 11, 63 (2001).
9.S. Hitchins, D. Martin, L. Burke, K. Yates, K. Fallon, A.
Hahn and G. Dobson, “Glycerol hyperhydration improves
cycle time trial performance in hot humid conditions”,
Eur. J. Appl. Physiol. Occup. Physiol. 80, 494 (1999). doi:
10.1007/s004210050623
10.A. Coutts, P. Reaburn, K. Mummery and M. Holmes,
“The effect of glycerol hyperhydration on olympic distance triathlon performance in high ambient temperatures”, Int. J. Sport Nutr. Exerc. Metab. 12, 105 (2002).
11.J.L. Nelson and R.A. Robergs, “Exploring the potential
ergogenic effects of glycerol hyperhydration”, Sports
Med. 37, 981 (2007). doi: 10.2165/00007256-20073711000005
12.E.D. Goulet, M. Aubertin-Leheudre, G.E. Plante and
I.J. Dionne, “A meta-analysis of the effects of glycerolinduced hyperhydration on fluid retention and endurance
performance”, Int. J. Sport Nutr. Exerc. Metab. 17, 391
(2007).
13.F.E. Marino, D. Kay and J. Cannon, “Glycerol hyper­
hydration fails to improve endurance performance and
thermoregulation in humans in a warm humid environment”, Pflugers Arch. 446, 455 (2003). doi: 10.1007/
s00424-003-1058-3
14.P. Montner, D.M. Stark, M.L. Riedesel, G. Murata, R.
Robergs, M. Timms and T.W. Chick, “Pre-exercise glycerol hydration improves cycling endurance time”, Int. J.
Sports Med. 17, 27 (1996). doi: 10.1055/s-2007-972804
GC-MS Quantitation of Glycerol
15.M.J. Anderson, J.D. Cotter, A.P. Garnham, D.J. Casley
and M.A. Febbraio, “Effect of glycerol-induced hyperhydration on thermoregulation and metabolism during
exercise in heat”, Int. J. Sport Nutr. Exerc. Metab. 11, 315
(2001). doi: 10.1097/00005768-200105001-01769
16.E.D. Goulet, R.A. Robergs, S. Labrecque, D. Royer and
I. Dionne, “Effect of glycerol-induced hyperhydration
on thermoregulatory and cardiovascular functions and
endurance performance during prolonged cycling in a
25 degrees C environment”, Appl. Physiol. Nutr. Metab. 31,
101 (2006). doi: 10.1139/h05-006
17.M. Magal, M. Webster, L. Sistrunk, M. Whitehead,
R. Evans and J. Boyd, “Comparison of glycerol and
water hydration regimens on tennis-related performance”, Med. Sci. Sports Exerc. 35, 150 (2003). doi:
10.1097/00005768-200301000-00023
18.C.D. Gilker, G.R. Pesola and D. Matthews, “A mass
spectrometric method for measuring glycerol levels
and enrichments in plasma using 13C and 2H stable
isotopic tracers”, Anal. Biochem. 205, 172 (1992). doi:
10.1016/0003-2697(92)90595-X
19.F. Magni, L. Arnoldi, L. Monti, P. Piatti, G. Pozza and
M.G. Kienle, “Determination of plasma glycerol isotopic
enrichment by gas chromatography-mass spectrometry:
an alternative glycerol derivative”, Anal. Biochem. 211,
327 (1993). doi: 10.1006/abio.1993.1279
20.M.T. Ackermans, A.F.C. Ruiter and E. Endert,
“Determination of glycerol concentrations and glycerol
isotopic enrichments in human plasma by gas chromatography/mass spectrometry”, Anal. Biochem. 258, 80
(1998). doi: 10.1006/abio.1997.2535
21.M. Beylot, C. Martin, B. Beaufrere, J.P. Riou and R.
Mornex, “Determination of steady state and nonsteadystate glycerol kinetics in humans using deuteriumlabeled tracer”, J. Lipid Res. 28, 414 (1987).
22.P.J. Flakoll, M. Zheng, S. Vaughan and M.J. Borel,
“Determination of stable isotopic enrichment
and concentration of glycerol in plasma via gas­chromatography-mass spectrometry for the estimation
of lipolysis in vivo”, J. Chromatogr. B 744, 47 (2000). doi:
10.1016/S0378-4347(00)00209-7
23.T. McIntosh, H. Davis and D. Matthews, “A liquid
­chromatography-mass spectrometry method to measure stable isotopic tracer enrichments of glycerol and
glucose in human serum”, Anal. Biochem. 300, 163
(2002). doi: 10.1006/abio.2001.5455
24.C.E. Dalgliesh, E.C. Horning, M.G. Horning, K.L.
Knox and K. Yarger, “A gas-liquid-chromatographic
­procedure for separating a wide range of metabolites
occurring in urine or tissue extracts”, Biochem. J. 101,
792 (1966).
25.J. Purevsuren, Y. Hasegawa, H. Kobayashi, M. Endo and
S. Yamaguchi, “Urinary organic metabolite screening of
children with influenza-associated encephalopathy for
inborn errors of metabolism using GC/MS.” Brain Dev 30,
in press (2008). doi: 10.1016/j.braindev.2008.01.001