APMIS 110: 819–24, 2002 Copyright C APMIS 2002

Printed in Denmark . All rights reserved

ISSN 0903-4641

Trimethylamine content in vaginal secretion and its relation

to bacterial vaginosis
HELEN WOLRATH,1 HANS BORÉN,2 ANDERS HALLÉN3 and URBAN FORSUM1
1
Division of Clinical Microbiology, Dept. of Molecular and Clinical Medicine, 2Division of Chemistry, Dept.
of Physics and Measurement Technology, Linköpings Universitet, 3Dept. of Dermatology and Venereology,
University Hospital, Uppsala, Sweden

Wolrath H, Borén H, Hallén A, Forsum U, Trimethylamine content in vaginal secretion and its re-
lation to bacterial vaginosis. APMIS 2002;110:819–24.
The presence of a fishy odor emanating from women who present with a malodorous vaginal discharge
is well known. The odor is due to bacterial reduction of trimethylamine oxide to trimethylamine
(TMA) in vaginal secretion. The release of TMA from specimens of vaginal fluid following the ad-
dition of alkali is often used in making a clinical diagnosis of bacterial vaginosis (BV). We now report
a sensitive gas chromatographic method for analysis and quantification of TMA in vaginal fluid in
which weighed samples were used. In addition, a proper diagnosis of BV was obtained using Gram-
stained smears of the vaginal fluid according to the method of Nugent et al. (R. P. Nugent et al., J
Clin Microbiol 1991;29:297–301). We also diagnosed BV according to Hallén et al. (A. Hallén et al.
Genitourin Med 1987;63:386–9). TMA was present in all women with a Nugent score between 7 and
10 and in almost all women diagnosed with BV according to the method of Hallén et al. TMA was
not found or was only found in very low concentrations in vaginal fluid from women with Nugent
scores of 0 to 3. TMA was also found in four women with a negative sniff test. It seems that high
levels of TMA in samples of vaginal fluid are typical for BV regardless of the scoring method used
for diagnosis. However, low levels of TMA, ⬍5 mg/g vaginal fluid, do not always correlate with BV.
Key words: Bacterial vaginosis; trimethylamine; Nugent scores.
Urban Forsum, Dept. of Molecular and Clinical Medicine, Division of Clinical Microbiology, Linköp-
ing University, 581 85 Linköping, Sweden. e-mail: urban.forsum/imk.liu.se

Clinicians have long noticed a fishy odor ema-
nating from women with vaginal discharge. Col-
lectively with other symptoms, the condition
has often been designated as nonspecific vagin-
itis. Nonspecific vaginitis was first introduced as
a syndrome including foul smelling discharge,
but a widely accepted concept including the syn-
drome is now recognized as bacterial vaginosis
(BV). In 1978 Pheifer et al. (1) reported for the
first time on the presence of the same fishy,
amine-like odor in samples of vaginal fluid from
women with nonspecific vaginitis when 10%
Received May 30, 2002.
Accepted September 19, 2002.
KOH was added. However, the amine respon-
sible for the odor was not elucidated. It is well
known that a ‘‘fishy odor’’ appears when fish
become rancid, and this smell is due to trime-
thylamine (TMA) (2). In 1986, Brand et al. (3)
verified that the amine responsible for causing
the fishy odor in vaginal samples from women
with BV is indeed TMA. BV-positive samples
with a fishy odor were analyzed and compared
with non-smelling samples from healthy
women. Further, Cruden et al. (4) later showed
in a study of some isolates of Gardnerella va-
ginalis, Bacteroides and Mobiluncus species, i.e.
bacteria that occur in BV, that Mobiluncus spp.
in culture can reduce trimethylamine oxide
819
 WOLRATH et al.
(TMAO) to TMA. In addition to TMA, at least
three other amines are apparent in vaginal
samples, i.e. putrescine, cadaverine, and tyr-
amine (3, 5–7). Whether these three amines con-
tribute to the amine smell when 10% KOH is
added to samples of vaginal fluid has never been
established.
In humans, intestinal bacteria form TMA
due to the metabolic breakdown of choline. The
primary sources of choline are egg, liver and
soybeans. Intestinal bacteria also form TMA
from TMAO, which is present in saltwater fish.
Most of the TMA formed is then enzymatically
converted to TMAO in the liver, and subse-
quently excreted in the urine (8). The concen-
trations of TMA and TMAO in urine are on
average 1 and 40 mg/day, respectively. These
values are markedly influenced by current diet,
and the ability to convert TMA to TMAO is
also a genetically linked trait (9). Furthermore,
there is a rare metabolic anomaly, trimethylami-
nuria, in which the patient smells like fish due
to excretion of TMA through their breath,
sweat, urine and vaginal secretion (10).
Many of the amine studies were performed
before or in parallel with studies that defined
the syndrome of BV using Amsel’s clinical cri-
teria (11), a well-defined polythetic concept of
major importance to women’s health. The four
criteria for the diagnosis of BV according to
Amsel are: the release of a fishy odor when
KOH is added to the vaginal secretion, a va-
ginal pH above 4.5, the presence of an adherent
white discharge, and clue cells in the vaginal wet
smear. Three of these four criteria must be met
for a diagnosis of BV. Amsel’s criteria have been
accepted as the ‘‘gold standard’’ for the clinical
diagnosis of BV. The concept of BV has been
further refined, from a diagnostic point of view,
with the Nugent scoring system based on micro-
scopically detectable changes in Gram-stained
smears of vaginal fluid (12). The use of Gram-
stained vaginal smears is well documented in
BV diagnostics and has been validated against
Amsel’s criteria (13, 14) in several different con-
texts, e.g. obstetrics and gynecology. Several
other diagnostic schemes for BV have been de-
vised that rely on microscopic findings and/or
the fishy smell, vaginal pH, and the character of
the vaginal fluid. All scoring systems have been
criticized for having variables that are not inde-
820
pendent of each other and for including vari-
ables with low sensitivity and specificity.
The so-called sniff test has proven to be an
important and useable criterion for the diag-
nosis of BV. However, it is possible that the con-
tent of TMA in vaginal secretion can be influ-
enced by the diet, as is the case for urine. Since
there is individual variation in normal TMA N-
oxidation, it is also feasible that the concen-
tration varies in vaginal secretions. Several
issues regarding the usefulness of smelling the
fishy odor in patients or recognizing the odor
of vaginal secretion samples when 10% KOH is
added thus require further clarification. Do all
BV patients have TMA in their vaginal secre-
tion? Is it possible to detect TMA in vaginal
fluid from BV patients even though the sniff test
is negative? Furthermore, in the investigations
analyzing TMA in vaginal fluid, no proper
quantification has been done.
Using a column suitable for TMA, we have
now developed a headspace gas chromatogra-
phy flame ionization detection (headspace GC-
FID) method for more precise analysis. At the
same time we quantified the concentration of
TMA in vaginal fluid by measuring the weight
of all samples. We also wanted to examine
whether TMA alone is responsible for the odor
in vaginal fluid in women with BV. We therefore
assessed the extent to which other amines that
have been identified in vaginal fluid, i.e. pu-
trescine, cadaverine and tyramine, in concen-
trations present in vaginal fluids, contribute to
the characteristic odor.
MATERIALS AND METHODS
Sampling
Women of childbearing age (nΩ61), with various
lower genital tract disorders, who attended a sexually
transmitted disease (STD) clinic in Uppsala, Sweden,
during a 3-month period volunteered to participate
in this study. Both cases with and without BV were
included. A glass vial together with a plastic loop tool
was weighed on a 0.000 g scale. Vaginal samples from
the lateral fornix were taken with the plastic loop,
which was then placed in the vial, the new weight
was noted, and the weight of the vaginal fluid was
calculated. Two ml of 1 M HCl was added to the
sample and the vial was closed and then frozen at
ª20 æC until analysis.
In addition, another sample of vaginal fluid was
taken with a plastic loop from the lateral fornix and
 TRIMETHYLAMINE CONTENT IN VAGINAL SECRETION
smeared onto a glass slide, air-dried, and saved for
batch-wise staining later by heat fixing and Gram
staining according to standard procedures. The
Gram-stained smears were scored according to Nug-
ent et al. (12), in a fully blinded manner, by another
member of the research group after the clinical visit
and TMA determinations. At the STD clinic a diag-
nosis was made according to the method of Hallén et
al. (15), based on pH, the sniff test, and clue cells.
The patients were also screened for STD agents and
Candida.
For the odor test, samples of the amines putrescine,
cadaverine, and tyramine were prepared at pH 6, and
samples of TMA were prepared at pH 5, in three dif-
ferent concentrations. The highest concentrations
were equal to the highest concentrations of pu-
trescine, cadaverine, tyramine, and TMA, respec-
tively, found in vaginal fluid from women with BV.
Two trained chemists with experience in threshold
odor number determinations of drinking water vol-
unteered to act as a test panel to evaluate the odor
of the amine samples. All samples were examined
both at pH 5–6 and at pH ⬎11.
Reagents and standards
Trimethylamine hydrochloride was obtained from
the Aldrich Chemical Co. (Milwaukee, WI, USA);
N,N-dimehylethylamine and triethylamine were ob-
tained from Acros Organics (Geel, Belgium); potas-
sium hydroxide and sodium sulphate were obtained
from Merck-Schuchardt (Hohenbrunn, Germany).
Stock standard solutions of 0.1 mg/ml TMA were
prepared in 0.2 M HCl and stored at 4 æC. A stock
solution of the internal standard, N,N-dimethylethyl-
amine of 0.05 mg/ml, and a solution of triethylamine
of 2.5 mg/ml were both prepared in 0.2 M HCl and
stored at 4 æC.
Samples of 324, 32.4 and 3.24 mg/ml of putrescine,
892, 89.2 and 8.92 mg/ml of cadaverine, and 1495,
149.5 and 14.95 mg/ml of tyramine at pH 6, and
samples of 88, 8.8 and 0.88 mg/ml of TMA at pH 5
were prepared for the olfactory test.
Analytical procedure
The frozen samples were thawed and transferred to
a 6 ml test vial. The volume was adjusted to 3 ml
with distilled water. As internal standard, 5 mg N,N-
dimethylethylamine (0.1 ml of a solution of 0.05 mg/
ml) was added together with 0.1 ml of the solution of
2.5 mg/ml of triethylamine in order to obtain a high
constant amine concentration in all samples. For cali-
bration from 0 to 2 mg TMA in the sample, the cali-
bration solution, the internal standard and triethyla-
mine were mixed in 3 ml of distilled water in the vial.
Approximately 1 g Na2SO4(s) and 0.4 g KOH(s) was
added to the vials before they were sealed. The vials
were heated to 60 æC for 1 h before analysis.
Headspace, 0.5 ml, was injected into the GC.
GC parameters
A Hewlett Packard 6890 GC was equipped with
a Flame Ionization Detector and a fused silica plot
column (poraplot coating from Chrompack, 25
m¿0.32 mm I.D, 0.10 mm phase thickness). Helium
was used as carrier gas at a pressure of 8 psi (split
injection, split ratio 15.0, split flow at 10.2 ml/min,
injector temperature 100 æC, temperature program
150 æC for 21 min, 20 æC/min to 200 æC).
RESULTS
In order to see which BV-associated amines
could be sensed by humans, an olfactory test
was done. The test panel could clearly smell the
odor, relevant for vaginal fluid samples, from
TMA at concentrations of 88 and 8.8 mg/ml
TMA at pH 5, but were unable to sense any
odor from the samples with putrescine, cadaver-
ine and tyramine (324, 892 and 1495 mg/ml) at
pH 6. They could also detect the TMA odor
from the sample with a concentration of 0.88
mg/ml at pH ⬎11, but they could still not detect
any odor from the other three amines at pH
⬎11. It was also observed that not everyone in-
volved in conducting the studies recognized the
odor of TMA, although this did not include
those on the test panel.
The concentration of TMA was measured in
one sample from each of 41 patients and in 2
samples from 20 patients. Results from the 20
patients with two samples are presented as the
average for each patient, since the results were
similar. The weight of the vaginal fluid in each
sample was measured and found to be in the
range 0.001–0.057 g except for two samples that
weighed 0.9 and 2.842 g, respectively. The cali-
bration curve for TMA was linear (R2Ω0.9938)
in the concentration range from 0 to 2 mg for
each sample. The lowest concentration on the
calibration curve was 0.01 mg.
Ten of the sixty-one participating women had
a diagnosis of Chlamydia or Candida. In the
sniff test, done at the STD clinic at the time of
the patient’s visit, 13 of the women were re-
garded as positive. Chemical analysis of the va-
ginal fluid from the 61 women showed more
than 0.1 mg TMA/g vaginal fluid in 26 of the
women. Two patients with a negative sniff test
had high concentrations of TMA in their va-
ginal fluid (Fig. 1).
Samples of vaginal fluid were used to prepare
821
 WOLRATH et al.

Fig. 1. Concentration of TMA in vaginal fluid in all

61 patients studied: 13 with a positive sniff test, where
none was below the detected limit and 5 were under
5 mg/g vaginal fluid, and 48 with a negative sniff test,
where 34 were below the detection limit and 2 were
over 5 mg/g vaginal fluid.
wet smears for scoring according to Hallén et
al. and to prepare air-dried smears for staining
and scoring according to Nugent et al. The
Nugent scoring resulted in 12 samples diag-
nosed as BV (Nugent scores of 7–10), 11 as in-
termediate (Nugent scores of 4–6), and 38 as
normal (Nugent scores of 0–3). The scoring ac-
cording to Hallén et al. resulted in 23 women
positive for BV and 38 negative for BV (Fig.
2). For the 12 patients diagnosed as having BV
according to Nugent et al., only one was diag-
nosed as non-BV by the Hallén et al. score and
two had a negative sniff test. Three of the thirty-
eight patients diagnosed as normal with the
Nugent score system were diagnosed as having
BV according to Hallén et al. None of the
women had a positive sniff test. In the group of
11 patients diagnosed as intermediate with the
Nugent score system, 9 were diagnosed as BV
and 2 were diagnosed as non-BV according to
Hallén et al. Of the 11 patients with an inter-
mediate score, one also had Chlamydia, and
among the 38 women diagnosed as normal with
the Nugent score system, 6 had a positive Chla-
mydia test and 3 had Candida.
Fig. 3 depicts an example of GC chromato-
grams derived from a sample from a patient
with a Nugent score of 8 (BV). The concen-
tration of TMA was 88 mg TMA/g vaginal fluid.
The concentrations of TMA in the vaginal fluid
are summarized in Figs. 1 & 2. TMA was found
in the vaginal fluid from all women with a Nug-
ent score indicative of BV, even though five had
concentrations below 5 mg/g vaginal fluid. In the
822
Fig. 2. Concentration of TMA in vaginal fluid in all
61 patients studied: 12 BV, with a Nugent score of 7
to 10 (BV), where none was below the detection limit
and 7 were over 5 mg/g vaginal fluid; 11 Intermediate,
with a Nugent score of 4 to 6 (Inter), where 2 were
under the detection limit and 3 were over 5 mg/g va-
ginal fluid; and 38 normal, with a Nugent score of 0
to 3 (Normal), where 32 were under the detection
limit. 23 BV diagnosed at the STD clinic according
to Hallén et al., where 3 were under the detection
limit and 10 were over 5 mg/g vaginal fluid, and 38
normal, where 34 were below the detection limit.
Fig. 3. GC-FID chromatogram from the analysis of
TMA in vaginal fluid from a patient with BV, Nugent
score 8.
23 samples diagnosed as BV according to Hall-
én et al., TMA was found in all but 3 patients.
Ten patients had low concentrations of TMA,
below 5 mg/g vaginal fluid. TMA was found in 9
out of 11 women with intermediate BV (Nugent
scores of 4–6). These were the same nine women
who were diagnosed as having BV using the
method of Hallén et al., but only five out of
these nine had a positive sniff test. The two
women in whom no TMA was found were diag-
nosed as normal by the Hallén et al. scoring
 TRIMETHYLAMINE CONTENT IN VAGINAL SECRETION
system and both had a negative sniff test. TMA
was found in low concentrations, under 5 mg/g
vaginal fluid, in 6 of the 38 women with Nugent
scores of 0–3, indicating the absence of BV. In
the three women diagnosed with BV using the
method of Hallén et al., but diagnosed as nor-
mal by the Nugent scoring system (score 0–3),
no TMA was found.
DISCUSSION
There have been several studies on the occur-
rence of amines in vaginal fluid associated
with BV (3, 5–6, 16–18), but few have focused
specifically on the content of TMA. However,
both Chen et al. (5) and Jones et al. (18) re-
ported having found methylamine in vaginal
fluids. Furthermore, Brand and colleagues (3)
concluded that TMA was the amine respon-
sible for the fishy odor in BV. But no detailed
study of the concentration of TMA in vaginal
fluid in women with and without BV has yet
been done.
In this study we used the GC technology with
a plot column, developed for analysis of volatile
amines, to improve the determination of TMA.
As the headspace analysis is not completely pre-
cise, we used a chemically similar amine as an
internal standard. This headspace analysis
methodology is essentially similar to that used
by Brand & Galask (3), and requires saturated
solutions with a very high pH. However, instead
of using K2CO3 for saturation, we used
Na2SO4. The reason for using Na2SO4 was to
avoid evaporation of TMA in the CO2 formed
upon acidification of the solution. We also
added a third amine, triethylamine (TEA), to
the sample to obtain a constantly high amine
concentration. This addition was independent
of the TMA concentration, but was a prerequi-
site in order to hinder the sticky amines from
adhering to the surrounding surfaces. This ad-
dition of triethylamine drastically increased the
sensitivity of the analysis method. We were
thereby able to analyze levels as low as 0.01 mg
TMA/sample. As we used weighed samples, it
was possible to determine the precise TMA con-
centration in the vaginal fluids.
Since TMA is normally excreted in the urine
(8), and in other body fluids in people suffering
from trimethylaminuria, one may speculate that
even people who are not suffering from trime-
thylaminuria also excrete small amounts of
TMA and TMAO in body fluids other than
urine, e.g. vaginal fluid. No investigations have
been done on the TMAO and TMA in vaginal
fluid associated with BV. Sardas and colleagues
(10) are thus the only research group that has
studied TMAO metabolism in association with
BV, but this research was done with urine and
not vaginal fluid. They studied the TMA to
TMAO metabolism in women with a fishy-
smelling vaginal discharge, clinically diagnosed
BV based on Amsel’s criteria, and compared it
with women with a fishy-smelling vaginal dis-
charge but without BV, and healthy women with
no smelly discharge. They determined the
amount of TMA and TMAO in urine and con-
cluded that the fishy-smelling women, with or
without a BV diagnosis, excreted significantly
more free TMA in urine and that they had a
similar, significantly reduced capacity to N-ox-
idize TMA as compared to healthy control
women. Sardas et al. (10) determined that the
percentage of total TMA excreted as TMAO in
urine is highly individual and ranges from 35 to
100%. It can be assumed that the same relation-
ship between TMA and TMAO is also found in
vaginal fluid.
In our study, we found women with amounts
of TMA up to 5 mg TMA/g vaginal fluid but
with no BV diagnosis according to the Nugent
or Hallén et al. scoring systems. This TMA con-
tent could be due to their diet or their individual
capacity to metabolize TMA to TMAO. We also
had women with a negative sniff test who had
TMA in their vaginal fluid. Two had very high
amounts, 46.5 and 51.8 mg TMA/g vaginal fluid,
respectively, and both were diagnosed as having
BV according to the Nugent or Hallén et al. sys-
tems. There were additional women with a nega-
tive sniff test who had low concentrations of
TMA in their vaginal fluid, 0.31 to 4.36 mg
TMA /g vaginal fluid, and who were diagnosed
either as BV or intermediate according to Nug-
ent. An explanation for these negative sniff tests
could be differences in the ability of the clinical
staff to actually smell TMA. Since the ability to
detect the smell of TMA is individual, a more
precise diagnostic methodology would be pref-
erable. TMA is a possible marker, but a more
thorough investigation of the cut-off concen-
tration of TMA is required. This is especially so
823
 WOLRATH et al.
because TMA varies as a result of individual
differences in N-oxidation and diet.
We would like to thank the staff of the STD depart-
ment at the University Hospital in Uppsala for all
their help with the patient samples. We also thank
Bodil Carlsson for Gram staining all smears and for
scoring of the smears. This research was supported
by the Foundation for Strategic Research (graduate
student position for Helen Wolrath via Forum Sci-
entum).
REFERENCES
1. Pheifer TA, Forsyth PS, Durfee MA, Pollock
HM, Holmes KK. Nonspecific vaginitis – role of
Haemophilus vaginalis and treatment with me-
tronidazole. N Engl J Med 1978;298:1429–43.
2. Barrett EL, Kwan HS. Bacterial reduction of tri-
methylamine oxide. Annu Rev Microbiol 1985;
39:131–49.
3. Brand JM, Galask RP. Trimethylamine: the sub-
stance mainly responsible for the fishy odor often
associated with bacterial vaginosis. Obstet Gyne-
col 1986;68:682–5.
4. Cruden DL, Galask RP. Reduction of trimethyla-
mine oxide to trimethylamine by mobiluncus
strains isolated from patients with bacterial vag-
inosis. Microbial Ecology in Health and Disease
1988;1:95–100.
5. Chen KCS, Forsyth PS, Buchanan TM, Holmes
KK. Amine content of vaginal fluid from un-
treated and treated patients with nonspecific
vaginitis. J Clin Invest 1979;68:828–35.
6. Kubota T, Sakae U, Takeuchi H, Usui M. Detec-
tion and identification of amines in bacterial vag-
inosis. J Obstet Gynaecol 1995;21:51–5.
7. Wolrath H, Forsum U, Larsson PG, Borén H.
Analysis of bacterial vaginosis-related amines in
vaginal fluid by gas chromatography and mass
spectrometry. J Clin Microbiol 2001;39:4026–31.
824
8. Spellacy E, Watts RW, Goolamali SK. Trimethyl-
aminuria. J Inherit Metab Dis 1980;2:85–8.
9. Al-Waiz M, Ayesh R, Mitchall SC, Idle JR,
Smith RL. A genetic polymorphism of the N-
oxygenation of trimethylamine in humans. Clin
Pharmacol Ther 1987;42:588–94.
10. Sardas S, Akyol D, Green, RL, Mellon T,
Gökmen O, Cholerton S: Trimethylamine N-oxi-
dation in Turkish women with bacterial vag-
inosis. Pharmacogenetics 1996;6:459–63.
11. Amsel R, Totten PA, Spiegel CA, Chen KC, Es-
chenbach D, Holmes KK. Nonspecific vaginitis:
diagnostic criteria and microbial and epidemio-
logic associations. Am J Med 1983;74:14–22.
12. Nugent RP, Krohn MA, Hillier SL. Reliability of
diagnosing bacterial vaginosis is improved by a
standardized method of gram stain interpreta-
tion. J Clin Microbiol 1991;29:297–301.
13. Schwebke JR, Hillier SL, Sobel JD, McGregor
JA, Sweet RL. Validity of the vaginal gram stain
for the diagnosis of bacterial vaginosis. Obstet
Gynecol 1996;88:573–6.
14. Forsum U, JakobssonT, Larsson PG, SchmidtH,
Beverly A, Bjørnerem A, et al. An international
study of the inter-observer variation between the
interpretations of vaginal smear criteria of bac-
terial vaginosis. APMIS (in press). 2002.
15. Hallén A, Påhlson C, Forsum, U. Bacterial vag-
inosis in women attending STD clinic: diagnostic
criteria and prevalence of Mobiluncus spp. Geni-
tourin Med 1987;63:386–9.
16. Chen KCS, Amsel R, Escenbach DA, Holmes
KK. Biochemical diagnosis of vaginitis: detec-
tion of diamines in vaginal fluid. J Infect Dis
1982;145:337–45.
17. Sanderson BE, White E, Balsdon MJ. Amine
content of vaginal fluid from patients with
trichomoniasis and gardnerella associated non-
specific vaginitis. Br J Vener Dis 1983;59:302–5.
18. Jones BM, Al-Fattani M, Gooch H. The deter-
mination of amines in the vaginal secretions of
women in health and disease. Int J STD AIDS
1994;5:52–5.