Content uploaded by Donald J Mcmahon
Author content
All content in this area was uploaded by Donald J Mcmahon
Content may be subject to copyright.
Differences in Free Estradiol and Sex Hormone-
Binding Globulin in Women with and without
Premenstrual Dysphoric Disorder
Susan Thys-Jacobs, Don McMahon, and John P. Bilezikian
St. Lukes-Roosevelt Hospital (S.T.-J.) and Division of Endocrinology (S.T.-J., D.M., J.P.B.), Department of Medicine, College of Physicians
and Surgeons, Columbia University, New York, New York 10019
Context: Over the years, different hypotheses involving the ovarian steroid hormones have been
proposed to explain the luteal phase occurrence of severe premenstrual syndrome symptoms.
Although it had been strongly suspected that differences in the concentrations of the ovarian
steroids may underlie the mood and psychological imbalance of this disorder, the evidence for this
hypothesis has been inconsistent and remains controversial.
Objective: Our objective was to measure the ovarian steroid hormones across the menstrual cycle
in women with and without luteal phase symptoms consistent with premenstrual dysphoric dis-
order (PMDD).
Design: We measured estradiol (E2), progesterone, and SHBG in women with and without PMDD
using a cross-sectional and prospective experimental design. Participating women underwent
2-month self-assessment symptom screening and 1-month hormonal evaluation.
Results: Overall means for LH, progesterone, E2, peak E2, and free E2 were not different between
groups. Across the menstrual cycle, overall percent free E2 was significantly lower and SHBG
significantly greater in the PMDD group compared with controls (1.39 ⫾ 0.26 vs. 1.50 ⫾ 0.28, P ⫽
0.03; 61.4 ⫾ 25.1 vs. 52.4 ⫾ 21.3 nmol/liter, P ⫽ 0.046, respectively). During the luteal phase, free
E2 was significantly lower in the PMDD group compared with controls (PMDD 7.6 ⫾ 7.0 vs. controls
8.9 ⫾ 8.4 pmol/liter; P ⫽ 0.032). For both follicular and luteal phases, SHBG was significantly higher
in the PMDD group (follicular phase 60.5 ⫾ 31.7 vs. 51.4 ⫾ 38.2 nmol/liter, P ⫽ 0.047; luteal phase
65.1 ⫾ 32.3 vs. 55.1 ⫾ 38.9 nmol/liter, P ⫽0.03). In both groups, SHBG significantly increased from
the follicular to luteal phase.
Conclusion: Luteal phase concentrations of free E2, percent free E2, and SHBG differ significantly
between women with and without PMDD. (J Clin Endocrinol Metab 93: 96–102, 2008)
P
remenstrual dysphoric disorder (PMDD), also described as
severe premenstrual syndrome (PMS), is a poorly under-
stood disorder characterized by the cyclical occurrence of be-
havioral, psychiatric, and physical symptoms during the luteal
phase of the menstrual cycle. Women with PMDD experience
mood changes, irritability, and depression so severe that per-
sonal and professional role performance is frequently disrupted.
Severe PMS is unique to premenopausal women because it is
related to the ovarian cycle and does not occur before puberty,
during pregnancy, or after menopause. Indeed, there appears to
be compelling evidence that PMDD is related to the ovarian
hormonal rhythmicities of the menstrualcycle (1–3), assymptom
relief ensues with suppression of ovulation and ablation of these
cyclical ovarian hormonal changes. Bilateral ovariectomy, estra-
diol (E2) implants, oral contraceptives, and GnRH agonists have
significantly diminished PMS symptoms (4– 8). However, the
precise role of the ovarian steroid hormones remains a contro-
versial one because most investigations have failed to discern any
0021-972X/08/$15.00/0
Printed in U.S.A.
Copyright © 2008 by The Endocrine Society
doi: 10.1210/jc.2007-1726 Received August 2, 2007. Accepted October 11, 2007.
First Published Online October 23, 2007
Abbreviations: E2, Estradiol; 5-HT2A, 5-hydroxytryptamine; PMDD, premenstrual dyspho-
ric disorder; PMS, premenstrual syndrome.
ORIGINAL ARTICLE
Endocrine Care
96 jcem.endojournals.org J Clin Endocrinol Metab. January 2008, 93(1):96 –102
overall difference in ovarian hormone concentrations across the
menstrual cycle. However, several studies have reported higher
luteal phase E2 concentrations, whereas others have reported, in
contrast, lower concentrations between women with and with-
out PMDD (9–13).
In viewof the fact that estrogen has a profound effect on mood
(14), we expected that women with PMDD would have lower
luteal phase free E2 concentrations compared with healthy
asymptomatic controls. Our aims were to: 1) measure fluctua-
tions in total, free, and bound ovarian steroid hormone levels
across the menstrual cycle in women with and without PMDD;
and 2) determine whether differences in the ovarian steroid hor-
mones existed between these two groups of women.
Subjects and Methods
Study subjects
Recruitment and study design have previously been described (15).
This investigation, conducted between 2000 and 2005 was approved by
the institutional review board of St. Luke’s-Roosevelt Hospital Center.
All participants provided written informed consent. Briefly, healthy, pre-
menopausal women between the ages of 18 and 45 yr, with either severe
PMS or asymptomatic menstrual cycles were recruited from the New
York City metropolitan and tristate area. Inclusion criteria for all par-
ticipants were general good health and regular menstrual cycles 21–35 d
in length (⫹5 d). Exclusion criteria for women recruited to either group
were: current major medical condition or serious medical illness in the
past; major psychiatric disorder, including active depression; menstrual
cycle irregularity; oral contraceptive medication; bone and calcium dis-
orders; and any overt metabolic disorder, such as diabetes mellitus. Pre-
screening by telephone excluded subjects not meeting inclusion and ex-
clusion criteria. To be included as an asymptomatic control required a
woman to have: 1) no prior medical history of PMDD or PMS, 2) a
prospective and consecutive assessment with a 2-month daily diary dem-
onstrating minimal symptoms during both the luteal and intermenstrual
phases with no evidence of functional impairment, and 3) no more than
two symptoms rated in either the moderate or severe range on 3 or more
days during any one screening cycle. For the PMDD group, a woman
required a diagnosis of PMDD meeting the Diagnostic and Statistical
Manual of Mental Disorders: DSM-IV (16) criteria based upon a pro-
spective and consecutive self-assessment with a 2-month daily symptom
diary. Inclusion diagnosis for the PMDD group required five or more
symptoms rated at least moderate in severity for the last week of the
menstrual cycle with evidence of functional impairment for at least 2 d
and a minimal symptom intermenstrual phase occurring after menses.
Protocol design
Eligibility evaluation included a medical, menstrual, and gynecolog-
ical history, as well as a physical examination and baseline laboratory
screening with complete blood count, chemistry panel, thyroid function
panel, pregnancy test, and 24-h urine calcium measurement. Two
months of prospective, consecutive daily symptom screening with
1-month hormonal evaluation were conducted. If two consecutive men-
strual cycles did not fulfill the inclusion criteria, the asymptomatic con-
trol candidate was disqualified. PMDD participants were allowed three
menstrual cycles to qualify, if either of the first 2 screening months did
not meet inclusion criterion. Daily symptom assessment continued
throughout the menstrual cycle when hormonal evaluation occurred. A
modified structured psychiatric interview (Structured Clinical Interview
for DSM-III-R) was conducted to exclude current or recent mental dis-
orders within the previous 6 months (17).
Symptom assessment
The PMS diary, a self-assessment symptom rating scale, was used for
the symptom screening, monitoring, and diagnosis of both the PMDD
and control participants (18). The PMS diary is a concise, validated
questionnaire comprised of 18 physical and emotional symptoms: 11
items with item 11 having eight physical symptoms. Each symptom is
rated on a four-point rating scale: 0, absent; 1, mild; 2, moderate; and 3,
severe. The average of the components of symptom 11 counted as a single
item when determining qualification for inclusion criteria, whereas the
severity rating of all 18 symptom items were summed for a total symptom
severity score. The luteal mean was defined as the 7-d symptom mean
before the onset of menstruation. The menstrual mean was defined as the
mean scores during days of menstruation and the intermenstrual mean
as the 7-d symptom mean immediately after menses.
Measurements
Timed hormone and biochemical samples across the menstrual cycle
were drawn based on the length of the qualifying subject’s two previous
menstrual cycles. Serum and urine samples were collected at eight points
in time (i.e. d 2, 7, 12–15, 22, and 27 based on a 28- to 30-d cycle). Two
samples were obtained within the early follicular phase, four within the
periovulatory period, and two samples within the mid-to-late luteal
phase. The timing and frequency of the blood samples were adjusted
according to the average duration of the individual subject’s two qual-
ifying menstrual cycles. Serum samples were assayed for E2 with the
extraction chromatography RIA (normal range depending on phase, 20 –
750 pg/ml), progesterone with chemiluminescence assay (normal range
depending on phase, 50–2500 ng/dl), and LH with immunochemilumi-
nometric assay (normal range 0.5–76.3 IU/liter). SHBG was measured
with the immunochemiluminescent assay (normal range 17–120 nmol/
liter; intraassay and interassay variations 4.3 and 4.7%, respectively).
Laboratory tests were performed at Nichols Institute, Quest Diagnostics
(San Juan Capistrano, CA). Blood samples were collected between 0700
and 1000 h after an overnight fast. Blood samples were allowed to clot
for1hatroom temperature, and serum was obtained by centrifugation.
All timed hormone samples were frozen at ⫺70 C and assayed simulta-
neously for an individual subject. Ovulation was determined by the LH
surge and the highest E2 level, on examination and review of E2 and LH
levels from d 10–14 along with a minimum luteal phase serum proges-
terone concentration of 500 ng/dl.
Statistical analysis
Descriptive statistics were calculated with means, SD values, and 95%
confidence intervals for all continuous measures and counts and per-
centages for categorical measures. Dependent measures were grouped
into content domains (demographics, reproductive histories, habitus,
serum chemistries, and urine chemistries). LH and progesterone were log
transformed before analysis and reported in untransformed units. To
maintain a family-wise fixed type I error rate, a multivariate ANOVA
model for the measures within each content domain was used to test the
hypothesis that PMDD and control groups differed on at least one mea-
sure within a content domain, and a P value less than 0.05 for the fixed
effect of group was used to warrant univariate analysis of group differ-
ences. Univariate tests of baseline differences between PMDD and con-
trol groups used Student’s t tests for continuous measures or Fisher’s
exact test for categorical variables. Differences between PMDD and con-
trols in the temporal pattern of repeated measures were estimated with
linear mixed models (SAS Proc MIXED; SAS Institute Inc., Cary, NC).
Here, group (PMDD vs. control), time (day of sample or menstrual cycle
phase, see below), and group by time interactions were entered as fixed
effects, subject and error were entered as random effects, and a com-
pound symmetry covariance structure was used to model the within-
subject autocorrelation among repeated measures. P values for the fixed
effects are reported from the mixed model estimates; P values for group
differences at specifictimes (and within-group differences between times)
are reported from the mixed model estimates of the differences, and
means and SEs are shown. Calculation of the free and bound fractions of
J Clin Endocrinol Metab, January 2008, 93(1):96–102 jcem.endojournals.org 97
E2 was derived from the mathematical model of Sodergard et al. (19)
using total E2, SHBG, and albumin concentrations.
Each subject’s menstrual cycle length was determined by the onset of
menses at the beginning of the cycle sampled and the onset of menses at
the end of the cycle. To facilitate analysis of grouped data from women
with different menstrual cycle lengths, each woman’s menstrual cycle
was divided into five phases, and phase mean values were used for anal-
ysis: 1) phase 1, the first 5 d from onset of menses (menstrual/early
follicular phase); 2) phase 2, menstrual cycle d 6 –2 d before ovulation (by
primary ovulation date)(late follicular phase);3) phase 3,day before, day
of, and day after ovulation (ovulatory phase); 4) phase 4, 2 d after ovu-
lation through midpoint of luteal phase (early luteal phase); and 5) phase
5, midpoint of luteal phase to onset of next menses (late luteal phase).
Combined phases were then used to represent the follicular, midcycle,
and luteal phases of the cycle. Final menses of the sampled hormone cycle
was used to determine cycle length of that particular cycle, and individual
phases were then identified.
Results
A total of 4924 women were screened for this study, with 4034
of these women telephone screened disqualified for failure of
inclusion/exclusion criteria involving ethnicity, oral contracep-
tive usage, irregular menstrual cycles, geographic location, or
taking psychotropic medication. A total of 890 subjects signed
consent and were enrolled in the study for symptom monitoring
and diary evaluation. One hundred twenty-nine women com-
pleted the timed biochemical and hormone evaluation. Unantic-
ipated use of oral contraceptives or antidepressants, pregnancy,
anovulatory or irregular menstrual cycles, and other calcium
derangements such as primary hyperparathyroidism led to the
exclusion of 14 subjects from analysis. The remaining 115 (68
women with PMDD and 47 controls) met all enrollment criteria.
The demographic data and reproductive history for the overall
group of115 premenopausal women are shown in Table 1. There
were no significant differences in age, age at menarche, race,
height, weight, body mass index, menstrual cycle length, or preg-
nancies between groups. The mean age of the PMDD participant
was 29.5 ⫾ 6.1 compared with 27.9 ⫾ 6.0 yr for the control
subjects (P ⫽ 0.17). There was no difference between groups in
smoking history, current smoking history, or pack years of
smoking history. By definition of the inclusion criteria, the
PMDD group had significantly higher luteal symptom scores
than controls, as well as higher menstrual and intermenstrual
scores.
Reproductive steroid hormones–transverse means
The overall hormonal means of the PMDD and control
groups are depicted in Table 2. The transverse means (overall
means across the menstrual cycle) for LH, progesterone, E2,
peak E2, and free E2 were not different between groups. Overall
SHBG was significantly greater in the PMDD group compared
with controls (61.4 ⫾ 25.1 vs. 52.4 ⫾ 21.3 nmol/liter; P ⫽
0.046). Overall percent free calculated E2 was significantly
greater in the asymptomatic control group compared with the
PMDD group (1.50 ⫾ 0.28 vs. 1.39 ⫾ 0.26; P ⫽ 0.03). Calcu-
lated E2 bound to SHBG and to albumin were not significantly
different between groups. However, calculated percent E2
bound to SHBG was higher in the PMDD group (45.1 ⫾ 9.7 vs.
41.5 ⫾ 10.3; P ⫽ 0.05), whereas calculated percent E2 bound to
albumin was higher in the asymptomatic group (57.0 ⫾ 0.1 vs.
53.5 ⫾ 9.4; P ⫽ 0.06).
Mixed linear model analysis with fixed effects for group,
time, and group by time interaction showed that the ovarian
steroid hormones varied significantly across the menstrual cycle
in both groups of subjects with ovulatory cycles. E2, progester-
one, and LH changed significantly over the menstrual cycle phas-
es: for E2, F
(4, 404)
⫽ 129.1, P ⬍ 0.001; for progesterone,
F
(4, 415)
⫽ 235.6, P ⬍ 0.001; and for LH, F
(4, 415)
⫽ 105.2, P ⬍
0.001). SHBG showed a significant main effect of phase (F
(4, 397)
⫽ 12.5; P ⬍ 0.001) and a significant effect of group (F
(1, 113)
⫽
4.323; P ⫽ 0.040). Free E2 showed a significant effect of phase
(F
(4, 409)
⫽ 133.1; P ⬍ 0.001). Percent free E2 showed a signif
-
icant effect of group (F
(1, 115)
⫽ 5.15; P ⫽ 0.025) and phase
(F
(4, 397)
⫽ 16.5; P ⬍ 0.001).
TABLE 1. Demographic and clinical characteristics of PMDD and control subject groups completing hormone sampling and
between-group comparison
Measure PMDD Control P value
n6847
Age (yr) 29.5 ⫾ 6.1 (28.0 –31.0) 27.9 ⫾ 6.0 (26.1–29.7) 0.17
Age at menarche 12.9 ⫾ 1.2 (12.6 –13.2) 12.6 ⫾ 1.2 (12.2–12.9) 0.14
No. of Caucasians (%) 49/68 (72) 35/47 (77) 0.59
Height (m) 1.66 ⫾ 0.06 (1.65–1.68) 1.65 ⫾ 0.07 (1.63–1.70) 0.32
Weight (kg) 63.8 ⫾ 13.0 (60.7– 66.9) 62.0 ⫾ 9.3 (59.3– 64.8) 0.40
BMI (kg/m
2
)
23.1 ⫾ 4.8 (21.9 –24.2) 22.5 ⫾ 3.7 (21.4 –23.6) 0.50
Years PMS symptoms 10.8 ⫾ 6.2 (9.3–12.3) N/A
Menstrual cycle length (d) 28.8 ⫾ 2.8 (28.1–29.5) 28.4 ⫾ 2.5 (27.6 –29.1) 0.45
PMDD symptom median
Luteal phase 21.2 ⫾ 10.3 (18.70 –23.65) 1.0 ⫾ 1.5 (0.60 –1.49)
Menstrual phase 14.9 ⫾ 10.5 (12.37–17.46) 1.1 ⫾ 1.1 (0.79 –1.46)
Intermenstrual phase 1.89 ⫾ 3.8 (0.97–2.81) 0.3 ⫾ 0.5 (0.13– 0.45)
Values are expressed as mean ⫾ SD (95% confidence interval) or number (%) within group. The Student’s t test comparison of differences between-group means for
continuous measures and Fisher’s exact test comparison of differences between-group proportions for categorical measures were performed. BMI, Body mass index;
N/A, not applicable.
98 Thys-Jacobs et al. Free E2 and SHBG in Premenstrual Dysphoric Disorder J Clin Endocrinol Metab, January 2008, 93(1):96 –102
Reproductive hormones–phase means
Between-group comparisons of the different phases of the
menstrual cycle (Fig. 1) demonstrated that the LH concentra-
tions were not significantly different during the individual fol-
licular or midcycle phases of the menstrual cycle. However, lu-
teal phase LH concentrations were greater in the asymptomatic
control group (LH 7.5 ⫾ 17.9 vs. 5.7 ⫾ 14.7; P ⫽ 0.13) com-
pared with the PMDD group.
SHBG wassignificantly higher in the PMDD group (Fig. 2)for
both follicular and luteal phases, with a trend toward signifi-
cance at midcycle (follicular phase 60.5 ⫾ 31.7 vs. 51.4 ⫾ 38.2
nmol/liter, P ⫽ 0.047; luteal phase 65.1 ⫾ 32.3 vs. 55.1 ⫾ 38.9
nmol/liter, P ⫽ 0.03; and midcycle phase 60.4 ⫾ 33.9 vs. 51.7 ⫾
40.8, P ⫽ 0.06). In both groups, SHBG significantly increased
from the follicular to luteal phase (PMDD 60.5 ⫾ 31.7 vs. 65.1 ⫾
32.3 nmol/liter, P ⬍ 0.001; controls 51.4 ⫾ 38.2 vs. 55.1 ⫾ 38.9
nmol/liter, P ⫽ 0.002).
Total E2 and progesterone concentrations were not different be-
tween groups during the individual phases of the menstrual cycle (Fig.
1). Free E2 was significantly lower in the PMDD group compared with
controls duringthe luteal phase (PMDD 7.6 ⫾ 7.0 vs. control 8.9⫾ 8.4
pmol/liter; P ⫽ 0.032; Fig. 2). Throughout all phases of the menstrual
cycle, percent free E2 was significantly lower in the PMDD group com-
pared with controls (follicular phase 1.384 ⫾ 0.348 vs. 1.492 ⫾ 0.422,
P ⫽ 0.033; midcycle phase 1.408 ⫾ 0.372 vs. 1.518 ⫾ 0.451, P ⫽
0.034; and luteal phase 1.348 ⫾ 0.351 vs. 1.470 ⫾ 0.427, P ⫽ 0.018).
Although calculated E2 bound to SHBG was not different between
groups during the luteal phase, E2 bound to albumin was significantly
greater in the control group during this same period (337 ⫾ 323 vs.
293 ⫾ 264 pmol/liter; P ⫽ 0.045). Percent E2 bound to SHBG and
bound to albumin were significantly different between groups. Percent
E2 bound to SHBG was greater in all phases of the menstrual cycle for
the PMDD group compared with controls, with asignificant difference
during the luteal phase (follicular phase 45.2 ⫾ 13.2 vs. 41.6 ⫾ 15.8,
P ⫽ 0.06; midcycle 44.3 ⫾ 14.0 vs. 40.6 ⫾ 16.8, P ⫽ 0.06; and luteal
phase 46.7 ⫾ 13.4 vs. 42.5 ⫾ 16.1, P ⫽ 0.03; Fig. 3). In contrast,
percent E2 bound to albumin was significantly greater for controls
during the luteal phase compared with the PMDD group (follicular
phase 56.9 ⫾ 15.4 vs. 53.4 ⫾ 12.8, P ⫽ 0.06; midcycle 57.9 ⫾ 16.4 vs.
TABLE 2. Reproductive steroid and hormone averages for PMDD (n ⫽ 68) and control (n ⫽ 47) subjects completing menstrual
cycle serum sampling and between-group comparison
Measure PMDD Control P value
Serum LH (IU/liter) 10.9 ⫾ 4.0 (9.9 –11.9) 11.5 ⫾ 5.3 (10.0 –13.1) 0.52
Serum LH peak (IU/liter) 33.5 ⫾ 23.5 (27.8 –39.2) 34.2 ⫾ 23.6 (27.1– 41.3) 0.29
Luteal phase LH (IU/liter) 5.7 ⫾ 14.7 ( 4.2–7.2) 7.5 ⫾ 18.0 (5.7–9.3) 0.064
Serum SHBG (nmol/liter) 61.4 ⫾ 25.1 (55.4 – 67.5) 52.4 ⫾ 21.3 (46.2–58.7) 0.046
Serum albumin (g/dl) 4.4 ⫾ 0.2 (4.3– 4.4) 4.3 ⫾ 0.21 (4.26 – 4.39) 0.24
Serum progesterone (ng/ml) 366.2 ⫾ 164 (326 – 406) 346.4 ⫾ 181.2 (293– 400) 0.34
Luteal phase progesterone (ng/ml) 1110 ⫾ 769 (1034 –1187) 1143 ⫾ 940 (1050 –1235) 0.59
Serum E2 (pg/ml) 153.2 ⫾ 46.4 (142–164) 150 ⫾ 57.8 (133–167) 0.76
Serum E2 peak (pg/ml) 276.4 ⫾ 125 (246 –306.7) 261.0 ⫾ 143 (218.1–303.9) 0.55
Free E2 (calculated pmol/liter) 7.7 ⫾ 2.4 (7.1– 8.3) 8.0 ⫾ 2.3 (7.3– 8.7) 0.52
Percent free E2 (calculated) 1.39 ⫾ 0.26 (1.32–1.45) 1.50 ⫾ 0.28 (1.41–1.58) 0.03
E2 bound to SHBG (calculated pmol/liter) 257 ⫾ 108 (231–283) 239 ⫾ 139 (199 –280) 0.45
Percent E2 bound to SHBG (calculated) 45.1 ⫾ 9.7 (45.8- 47.5) 41.5 ⫾ 10.3 (38.5– 44.5) 0.05
E2 bound to albumin (calculated pmol/liter) 298 ⫾ 91.7 (249 –288) 304 ⫾ 86 (279 –329) 0.71
Percent E2 bound to albumin (calculated) 53.5 ⫾ 9.4 (51.2–55.7) 57.0 ⫾ 10.1 (54.1– 60.0) 0.06
Luteal phase E2 (pg/ml) 156 ⫾ 134 (141–171) 170 ⫾ 164 (152–189) 0.238
Luteal phase free E2 (calculated pmol/liter) 7.6 ⫾ 7.0 (6.8–8.3) 8.9 ⫾ 8.4 (8.0 –9.8) 0.032
Luteal phase percent free E2 (calculated) 1.35 ⫾ 0.35 (1.28–1.41) 1.47 ⫾ 0.43 (1.39 –1.55) 0.018
Values are expressed as mean ⫾ SD (95% confidence interval). The Student’s t test comparison of differences between-group means was performed. To convert E2
pg/ml to pmol/liter, multiply by 3.67.
FIG. 1. Serum LH and progesterone means for PMDD (solid line) and control subjects
(dotted line) at the follicular phase, midcycle, and luteal phases of the menstrual
cycle (means and
SE). Significant differences are shown for between-group
differences and within-group phase differences. The within-group comparison in LH
for the midcycle compared with follicular and luteal phases for both the PMDD and
control groups with a P value less than 0.05 (
␣
). A within-group comparison in
midcycle progesterone compared with follicular phase for the PMDD and control
groups with a P value less than 0.05 (

).Within-group comparisons in luteal
progesterone for both PMDD and control groups compared with follicular and
midcycle phases with a P value less than 0.05 (c). LH is in IU/liter (normal range 0.5–
76.3), and progesterone is in ng/ml (normal range 50–2500).
J Clin Endocrinol Metab, January 2008, 93(1):96–102 jcem.endojournals.org 99
54.3 ⫾ 13.6, P ⫽ 0.06; and luteal phase 56.0 ⫾ 15.7 vs. 52.0 ⫾ 13.1,
P ⫽ 0.03). Within-group comparisons between phases demonstrated
the well-known fluctuations among the various menstrual cycle phases
in both groups for LH, E2, free E2, percent free E2, and progesterone
(Figs. 1–3).
Discussion
PMDD, also described as severe PMS, is a luteal phase disorder
characterized by the cyclical recurrence of mood and physical
symptoms during the latter half of the menstrual cycle. To date,
the underlying pathophysiology of this phase-related disorder
has remained unexplained (20, 21). This investigation that sys-
tematically screened a large number of potential candidates for
the distinct PMDD and control populations has found that luteal
phase concentrationsof free E2, percent free E2, and SHBG differ
significantly between women with and without PMDD. Both
free E2 and percent free E2 were significantly lower, and SHBG
concentrations significantly higher in PMDD subjects compared
with asymptomatic controls. These differences in percent free E2
and SHBG between groups were found throughout the men-
strual cycle but became clinically relevant in the PMDD group
with the statistically significant lower concentrations of free E2
FIG. 2. Serum E2, free E2, and percent free E2 means for PMDD (solid line) and
control subjects (dotted line) at the follicular phase, midcycle, and luteal phases
of the menstrual cycle (means and
SE). Significant differences are shown for
between-group differences and within-group phase differences. An asterisk (*)
above phase means indicates a between-group comparison difference with a P
value less than 0.05 at that phase. Within-group comparisons in midcycle E2 and
free E2 for PMDD and control groups compared with follicular and luteal phases
with a P value less than 0.05 (
␣
). Within-group comparisons in luteal E2 and free
E2 for PMDD and control groups compared with follicular phase with a P value
less than 0.05 (

). Within-group comparisons in midcycle percent free E2 for the
PMDD group compared with follicular and luteal phases with a P value less than
0.05 (c). A within-group comparison in midcycle percent free E2 for the control
group compared with luteal phase with a P value less than 0.05 (d). E2 is in pg/
ml (normal range 20 –750).
FIG. 3. Serum SHBG, SHBG-bound E2, percent SHBG-bound E2 means for PMDD
(solid line) and control subjects (dotted line) at the follicular phase, midcycle, and
luteal phases of the menstrual cycle (means and
SE). An asterisk (*) above phase
means indicates a between-group comparison difference with a P value less than
0.05 at that phase. A within-group comparison in luteal SHBG for PMDD and
control groups compared with midcycle and follicular phases with a P value less
than 0.05 (
␣
).Within-group comparisons in luteal SHBG-bound E2 for PMDD and
control groups compared with follicular and luteal phases with a P value less than
0.05 (

). Within-group comparisons in luteal SHBG-bound E2 for PMDD and
control groups compared with follicular phase with a P value less than 0.05 (c).
Within-group comparison in luteal percent SHBG-bound E2 for the PMDD and
control groups compared with midcycle with a P value less than 0.05 (d). SHBG is
in nmol/liter (normal range 17–120 nmol/liter).
100 Thys-Jacobs et al. Free E2 and SHBG in Premenstrual Dysphoric Disorder J Clin Endocrinol Metab, January 2008, 93(1):96 –102
during the luteal phase of the cycle. No luteal phase differences
were noted in progesterone concentrations.
Estrogen, although known to be important in both ovulation
and reproduction, has been recognized to enhance well-being
with major effects on the central nervous system in areas other
than the hypothalamus (22, 23). In this regard, estrogen appears
to exert actions in areas of the brain that are important for learn-
ing and memory, emotions and affective state, and pain sensi-
tivity. It has widespread regulatory effects throughout the brain,
including the brainstem and midbrain catecholaminergic neu-
rons, midbrain serotonergic pathways, midbrain dopaminergic
activity, and the basal forebrain cholinergic system (24). Evi-
dence suggests that estrogen modulates the metabolism of mono-
amine and neuropeptide transmitter pathways, and may affect
those mechanisms involved in mood and affective disorders (25,
26). Estrogen stimulates a significant increase in dopamine (D2)
receptors in the striatum of the brain, and an increase in the
density of 5-hydroxytryptamine (5-HT2A) binding sites in
the anterior frontal, cingulate, and primary olfactory cortex of
the brain (14), areas concerned with mood and behavior. In one
investigation, a single pulse of E2 in female rats induced a sig-
nificant increase in the density of 5-HT2A receptors in the fore-
brain (27). In another investigation, serotonin 5-HT2A receptor
binding in the prefrontal cortex of the brain as assessed by
positron emission tomography was significantly increased in 10
postmenopausal women after estrogen therapy. Acute changes
in ovarian function or even relatively lower estrogen concentra-
tions during the luteal phase of the menstrual cycle may influence
serotonergic metabolism and neurotransmitter reuptake, both
thought to mediate mood and emotion (28).
Over the years, different hypotheses involving the ovarian
steroid hormones have been proposed to explain the luteal phase
occurrence of severe PMS symptoms. Most investigations to date
have not substantiated consistent differences in the concentra-
tions of either E2 or progesterone between symptomatic or
asymptomatic women. Explanations for normal serum hormone
levels have even included the possibility of variations in the es-
trogen receptor with genotypical differences likely to mediate
differential behavioral responses. A recent investigation by Huo
et al. (29) has now demonstrated preliminary evidence for this
hypothesis with four single nucleotide polymorphisms in intron
of estrogen receptor 1 found to be significantly different in
PMDD and control subjects.
Our investigation with careful subject selection and well-
matched controls is the largest study of women with and without
PMDD in which multiple-timed measures of hormones were ob-
tained across the menstrual cycle. Although we did not find sig-
nificant differences in overall serum E2 or luteal phase E2 con-
centrations, free E2, percent free E2, and SHBG were found to
differ significantly between women. Differences in free E2 or
bioavailable E2 rather than total E2 may help explain the myriad
of PMDD symptoms, such as depression, moodiness, and irri-
tability. Our findings provide support to the hypothesis that a
relative estrogen deficiency may have a role in the pathogenesis
of PMS. These findings are similar in part to those demonstrating
lower serum E2 levels and diminished urinary E2 excretion dur-
ing the luteal phase in those with premenstrual symptoms, and
may also help explain the beneficial effect of E2 therapy admin-
istered in PMS (12, 30, 31). In contrast, our data differ from the
findings of Wang et al. (10), who reported higher E2 and lower
progesterone concentrations during the luteal phase in 12
women with PMS compared with controls, and from the data of
Seippel and Backstrom (11), who related higher luteal phase-E2
concentrations with greater PMS symptom severity. Both of
these studies were relatively small, and neither free E2 nor SHBG
was measured. It is unclear why so many studies have reported
different results. Methodological limitations may exist as to the
definition of what is a true asymptomatic control, actual sample
size, and to issues related to menstrual cycle length. The extreme
variability of menstrual cycle length and failure to properly align
the individual cycle phases may be the single most important
confounder in conflicting PMDD data.
Along the same line, but counter to the positive treatment one
would have anticipated on E2 therapy, another investigation
demonstrated that administration of higher doses of E2 during
the luteal phase in women with PMS worsened PMS symptom-
atology, and resulted in more negative mood symptoms com-
pared with placebo (32). The reported contradictory effects of E2
on luteal phase symptoms are still not fully explained. E2 may be
influencing another hormonal axis such as the vitamin D path-
way, resulting in disparate symptomatic responses. A previous
investigation by our group (15) demonstrated lower 1,25 dihy-
droxyvitamin D concentrations during the luteal phase in
women with PMDD compared with controls. Lower E2 concen-
trations during the luteal phase, as noted in this study, may ad-
versely affect vitamin D synthesis, resulting lower in 1,25 dihy-
droxyvitamin D concentrations with diminished calcium access
and reserves.
SHBG is a glycoprotein synthesized in the liver, which trans-
ports and regulates the access of sex steroids to their target tissues
(33). Levels of SHBG may be influenced by nutrition, sex ste-
roids, thyroid hormones, peptide hormones as insulin, and ge-
netics (34). In vitro data suggest that both androgens and estro-
gens stimulate SHBG production; however, in vivo androgens
appear to suppress SHBG, whereas estrogens are associated with
elevated levels. One study investigated the effect of various
growth factors on SHBG production in the human hepatoblas-
toma cell line Hep G2. Hep G2 cell cultures were incubated with
IGF-I, epidermal growth factor, and TGF (35). All these growth
factors resulted in a significant decrease in SHBG production in
cell culture medium. Our study found SHBG to be significantly
different between the two study cohorts across the menstrual
cycle, with higher concentrations in the PMDD group compared
with controls, particularly during the luteal phase. Interestingly,
we have also recently found in this same study cohort that IGF-I
concentrations are significantly different as well, with lower con-
centrations in PMDD subjects compared with controls (36). Our
clinical observations of an inverse relationship of IGF-I to SHBG
concentrations appear to support the work of Plymate et al. (35),
in which growth factors such as IGF-I may regulate SHBG.
Higher SHBG concentrations as noted in the PMDD group may
limit the bioavailability of E2 to the brain, liver, and other tissues,
and ultimately affect mood.
In this investigation, subjects were limited to Caucasian and
J Clin Endocrinol Metab, January 2008, 93(1):96–102 jcem.endojournals.org 101
Hispanic women. In addition, funding and feasibility constraints
limited hormonal sampling to8dofthemenstrual cycle, and
hormonal events may have been missed. Free E2 was calculated
and not measured in this study. Although methods used to cal-
culate free E2 have been used in many epidemiological and clin-
ical investigations, calculated E2 is not the same as measured E2.
In conclusion, differences in the concentrations of free E2,
percent free E2, and SHBG were demonstrated across the men-
strual cyclein women withPMDD compared with asymptomatic
controls. The differences in the concentrations of SHBG and free
E2 between women with and without luteal phase symptoms
may partially elucidate the existing controversy on the connec-
tion between ovarian steroid hormones and luteal phase
symptoms.
Acknowledgments
Address all correspondence and requests for reprints to: Susan Thys-
Jacobs, M.D., 343 West 58th Street, Suite 11, New York, New York
10019. E-mail: sthysja@chpnet.org.
This work was supported by the National Institute of Health-Na-
tional Institute of Diabetes and Digestive and Kidney Diseases Grant
DK57869-03 in conjunction with National Institute of Mental Health
and Institute of Women’s Health.
Disclosure Statement: The authors have nothing to disclose.
References
1. Rapkin A 1992 The role of serotonin in premenstrual syndrome. Clin Obstet
Gynecol 35:629 –636
2. Reid RL, Yen SS 1981 Premenstrual syndrome. Am J Obstet Gynecol 139:
85–104
3. Steiner M 1992 Female specific mood disorders. Clin Obstet Gynecol 35:599–
611
4. Rubinow DR 1992 The premenstrual syndrome: new views. JAMA 268:1908–
1912
5. Casson P, Haln PM, Van Vugt DA, Reid RL 1990 Lasting response to ovari-
ectomy in severe intractable premenstrual syndrome. Am J Obstet Gynecol
162:99–105
6. Graham CA, Sherwin BB 1992 A prospective treatment study of premenstrual
symptoms using a triphasic oral contraceptive. J Psychosom Res 36:357–366
7. Brown C, Ling F, Andersen R, Farmer R, Arheart K 1994 Efficacy of depot
leuprolide in premenstrual syndrome: effect of symptoms severity and type in
a controlled trial. Obstet Gynecol 84:779–786
8. Schmidt PJ, Nieman LK, Danaceau MA, Adams LF, Rubinow DR 1998 Dif-
ferential behavioral effects of gonadal steroids in women with and in those
without premenstrual syndrome. N Engl J Med 338:209–216
9. Rubinow D, Hoban C, Grover GN, Galloway DS, Roy-Byrne P, Andersen R,
Merriam G 1988 Changes in plasma hormones across the menstrual cycle in
patients with menstrually related mood disorder and in control subjects. Am J
Obstet Gynecol 158:5–11
10. Wang M, Seippel L, Purdy R, Backstrom T 1996 Relationship between symp-
tom severity and steroid variation in women with premenstrual syndrome.
J Clin Endocrinol Metab 81:1076 –1082
11. Seippel L, Backstrom T 1998 Luteal phase estradiol relates to symptom se-
verity in patients with premenstrual syndrome. J Clin Endocrinol Metab 83:
1988–1992
12. Blum I, Lerman M, Misrachi I, Nordenberg Y, Grosskopf I, Wewizman A,
Levy-Schiff R, Sulkes J, Vered Y 2004 Lack of plasma norepinephrine cyclicity,
increased estradiol during the follicular phase, and of progesterone and go-
nadotropins at ovulation in women with premenstrual syndrome. Neuropsy-
chobiology 50:10 –15
13. Hsiao CC, Liu CY, Hsiao MC 2004 No correlation of depression and anxiety
to plasma estrogen and progesterone levels in patients with premenstrual dys-
phoric disorder. Psychiatry Clin Neurosci 58:593–599
14. Fink G, Sumner BE, Rosie R, Grace O, Quinn JP 1996 Estrogen control of
central neurotransmission: effect on mood, mental state and memory. Cell Mol
Neurobiol 16:325–344
15. Thys-Jacobs S, McMahon D, Bilezikian JP 2007 Cyclical changes in calcium
metabolism across the menstrual cycle in women with premenstrual dysphoric
disorder. J Clin Endocrinol Metab 92:2952–2959
16. Task Force on DSM-IV 1994 Premenstrual dysphoric disorder. In: Diagnostic
and statistical manual of mental disorders: DSM-IV. 4th ed. Washington, DC:
American Psychiatric Association
17. Spitizer RL, Williams JBW, Gibbon M, First MB 1992 The Structured Clinical
Interview for DSM-III-R (SCID). I: History, rationale, and description. Arch
Gen Psychiatry 49:624–629
18. Alvir JMJ, Thys-Jacobs S 1991 Premenstrual and menstrual symptom clusters
and response to calcium treatment. Psychopharmacol Bull 27:145–148
19. Sodergard R, Backstrom T, Shanbhag V, Carstensen H 1982 Calculation of
free and bound fractions of testosterone and estradiol-17B to human plasma
proteins at body temperature. J Steroid Biochem 16:801–810
20. Cronje WH, Vashisht A, Studd JWW 2004 Hysterectomy and bilateral oo-
phorectomy for severe premenstrual syndrome. Hum Reprod 19:2152–2155
21. Yonkers K, Brown C, Pearlstein T, Foegh M, Sampson-Landers C, Rapkin A
2005 Efficacy of a new low dose oral contraceptive with drospirenone in
premenstrual dysphoric disorder. Obstet Gynecol 106:492–501
22. Wiklund I, Karlberg J, Mattson LA 1993 Quality of life of postmenopausal
women on a regimen of transdermal estradiol therapy: a double-blind placebo-
controlled study. Am J Obstet Gynecol 168(3 Pt 1):824 –830
23. Joffe H, Cohen LS 1998 Estrogen, serotonin and mood disturbance: where is
the therapeutic bridge? Biol Psychiatry 44:798– 811
24. McEwen B 1999 Molecular and neuroanatomical basis for estrogen effects in
the central nervous system. J Clin Endocrinol Metab 84:1790 –1797
25. Wagner EJ, Manzanares J, Moore KE, Lookingland KJ 1994 Neurochemical
evidence that estrogen induced suppression of kappa-opioid receptor mediated
regulation of tuberoinfundibular dopaminergic neurons is prolactin indepen-
dent. Neuroendocrinology 59:197–201
26. Kelly MJ, Qui J, Wagner EJ, Ronnekleiv O 2003 Rapid effects of estrogen on
G protein coupled receptor activation of potassium channels in the central
nervous system. J Steroid Biochem Mol Biol 83:187–193
27. Summer BE, Fink G 1995 Estrogen increases the density of 5-hydroxytrypta-
mine (2A) receptors in cerebral cortex and nucleus accumbens in the female rat.
J Steroid Biochem Mol Biol 54:15–20
28. Moses EL, Dreverts WC, Smith G, Mathis CA, Kairo BN, Butters MA, Leon-
dires MP, Greer PJ, Lopresti B, Loucks TL, Berga SL 2000 Effects of estradiol
and progesterone administration on human serotonin 2A receptor binding: a
PET study. Biol Psychiatry 48:854 – 860
29. Huo L, Straub RE, Roca C, Schmidt PJ, Shi K, Vakkalanka R, Weinberger DR,
Rubinow DR 2007 Risk for premenstrual dysphoric disorder is associated with
genetic variation in ESR1, the estrogen receptor alpha gene. Biol Psychiatry
62:925–933
30. Woods NF, Lentz MJ, Mitchell ES, Shaver J, Heitkemper M 1998 Luteal phase
ovarian steroids, stress arousal, premenses, perceived stress and premenstrual
symptoms. Res Nurs Health 21:129 –142
31. Steiner M, Steinberg S, Stewart D, Carter D, Berger C, Reid R, Grover D,
Streiner D 1995 Fluoxetine in the treatment of premenstrual dysphoria. N Engl
J Med 332:1529–1534
32. Dhar V, Murphy BE 1990 Double blind randomized crossover trial of luteal
phase estrogens (Premarin) in the premenstrual syndrome. Psychoneuroendo-
crinology. 15:489 –493
33. Hammond GL 1990 Molecular properties of corticosteroid binding globulin
and the sex steroid binding proteins. Endocr Rev 11:65–79
34. Siiteri PK, Murai JT, Hammond GL, Nisker JA, Raymoure WJ, Kuhn RW
1982 The serum transport of steroid hormones. Recent Prog Horm Res 38:
457–510
35. Plymate SR, Hoop RC, Jones RE, Matej LA 1990 Regulation of sex hormone-
binding globulin production by growth factors. Metabolism 39:967–970
36. Thys-Jacobs S, McMahon D, Bilezikian JP, Lower IGF-1 concentrations in
women with premenstrual dysphoric disorder. Am J Obstet Gyn, in press
102 Thys-Jacobs et al. Free E2 and SHBG in Premenstrual Dysphoric Disorder J Clin Endocrinol Metab, January 2008, 93(1):96 –102
