Sexual Desire and the Female Sexual Function Index (FSFI): A Sexual Desire Cutpoint for Clinical Interpretation of the FSFI in Women with and without Hypoactive Sexual Desire Disorder

Abstract

Introduction. A validated cutpoint for the total Female Sexual Function Index scale score exists to classify women with and without sexual dysfunction. However, there is no sexual desire (SD) domain-specific cutpoint for assessing the presence of diminished desire in women with or without a sexual desire problem.Aims. This article defines and validates a specific cutpoint on the SD domain for differentiating women with and without hypoactive sexual desire disorder (HSDD).Methods. Eight datasets (618 women) were included in the development dataset. Four independent datasets (892 women) were used in the validation portion of the study.Main Outcome Measures. Diagnosis of HSDD was clinician-derived. Receiver-operator characteristic (ROC) curves were used to develop the cutpoint, which was confirmed in the validation dataset.Results. The use of a diagnostic cutpoint for classifying women with SD scores of 5 or less on the SD domain as having HSDD and those with SD scores of 6 or more as not having HSDD maximized diagnostic sensitivity and specificity. In the development sample, the sensitivity and specificity for predicting HSDD (with or without other conditions) were 75% and 84%, respectively, and the corresponding sensitivity and specificity in the validation sample were 92% and 89%, respectively.Conclusions. These analyses support the diagnostic accuracy of the SD domain for use in future observational studies and clinical trials of HSDD. Gerstenberger EP, Rosen RC, Brewer JV, Meston CM, Brotto LA, Wiegel M, and Sand M. Sexual desire and the female sexual function index (FSFI): A sexual desire cutpoint for clinical interpretation of the FSFI in women with and without hypoactive sexual desire disorder. J Sex Med 2010;7:3096–3103.

Full text

Page 1

Sexual Desire and the Female Sexual Function Index (FSFI):
A Sexual Desire Cutpoint for Clinical Interpretation of the FSFI
in Women with and without Hypoactive Sexual Desire Disorderjsm_18713096..3103
Eric P. Gerstenberger, MS,* Raymond C. Rosen, PhD,* Jessica V. Brewer, MPH,*
Cindy M. Meston, PhD,†Lori A. Brotto, PhD,‡Markus Wiegel, PhD,§and Michael Sand, PhD¶
*New England Research Institutes, Inc., Watertown, MA, USA;†Department of Psychology, University of Texas at Austin,
Austin, TX, USA,‡University of British Columbia—Obstetrics/Gynaecology, Vancouver, BC, Canada;§Behavioral
Medicine Institute of Atlanta, Atlanta, GA, USA;¶Boehringer Ingelheim Pharmaceuticals, Inc—General Medicine,
Ridgefield, CT, USA
DOI: 10.1111/j.1743-6109.2010.01871.x
A B S T R A C T
Introduction. A validated cutpoint for the total Female Sexual Function Index scale score exists to classify women
with and without sexual dysfunction. However, there is no sexual desire (SD) domain-specific cutpoint for assessing
the presence of diminished desire in women with or without a sexual desire problem.
Aims. This article defines and validates a specific cutpoint on the SD domain for differentiating women with and
without hypoactive sexual desire disorder (HSDD).
Methods. Eight datasets (618 women) were included in the development dataset. Four independent datasets (892
women) were used in the validation portion of the study.
Main Outcome Measures. Diagnosis of HSDD was clinician-derived. Receiver-operator characteristic (ROC)
curves were used to develop the cutpoint, which was confirmed in the validation dataset.
Results. The use of a diagnostic cutpoint for classifying women with SD scores of 5 or less on the SD domain as
having HSDD and those with SD scores of 6 or more as not having HSDD maximized diagnostic sensitivity and
specificity. In the development sample, the sensitivity and specificity for predicting HSDD (with or without other
conditions) were 75% and 84%, respectively, and the corresponding sensitivity and specificity in the validation
sample were 92% and 89%, respectively.
Conclusions. These analyses support the diagnostic accuracy of the SD domain for use in future observational
studies and clinical trials of HSDD. Gerstenberger EP, Rosen RC, Brewer JV, Meston CM, Brotto LA, Wiegel
M, and Sand M. Sexual desire and the female sexual function index (FSFI): A sexual desire cutpoint for
clinical interpretation of the FSFI in women with and without hypoactive sexual desire disorder. J Sex Med
2010;7:3096–3103.
Key Words. Sexual Desire; Cutpoint; Hypoactive Sexual Desire Disorder; Female Sexual Function Index
Introduction
T
sexual function developed for use in clinical trials
and epidemiological studies of sexual dysfunction
in women. The FSFI was designed as a multidi-
mensional questionnaire measure, with subscales
to assess the major components of sexual function
in women, including sexual desire, arousal,
he Female Sexual Function Index (FSFI)
[1–3] is a 19-item, self-report measure of
orgasm, pain, and satisfaction. Development of the
questionnaire included both qualitative and quan-
titative studies, with individual items based on
qualitative interviews in women with and without
sexual dysfunction [1]. Separate validation studies
have been reported by a number of authors using
independent samples of women [2,3]. The FSFI
has been translated into multiple languages and
has been widely used to assess female sexual func-
tion in a variety of clinical and nonclinical settings
3096
J Sex Med 2010;7:3096–3103© 2010 International Society for Sexual Medicine

Page 2

[4–7]. The FSFI is a well-characterized instrument
and its psychometric properties have been previ-
ously described [2,3,8].
Based on early validation studies, a cutpoint for
the total FSFI scale score (26.5) was proposed by
Wiegel et al. [3] in order to classify women with
and without sexual dysfunction. Despite the use of
the total score cutpoint in differentiating women
with or without sexual dysfunction, it does not
provide a domain-specific cutpoint for assessing
the presence of diminished desire in women
[8–10]. Sexual desire (SD) is a complex, multifac-
eted component of sexual response in men and
women, which requires independent assessment
by means of a validated self-report scale.
There are two important reasons for consider-
ing the need for a domain-specific cutpoint. First,
diminished desire can be assessed in women who
are not currently sexually active, whereas other
domains and the total sexual function score are
predicated on sexual activity with or without a
partner [12]. Second, since hypoactive sexual
desire disorder (HSDD) is the most frequently
diagnosed sexual problem in women and fre-
quently occurs concurrently with other sexual dis-
orders in women [13,14], a validated diagnostic
cutpoint for the SD domain would be valuable
for use in future clinical and research studies, par-
ticularly in assessing desire-related outcomes of
treatment.
Methods
Development Sample
To establish a diagnostic cutpoint for the SD
domain of the FSFI, datasets were obtained from
independent clinical investigators at geographi-
cally diverse sites in the United States and Canada.
In order to be included in this analysis, a dataset
had to meet several criteria, including: (i) an inde-
pendent clinical assessment plus diagnosis for all
women classified as positive cases of HSDD; (ii)
documentation of the presence or absence of other
sexual dysfunctions; (iii) information on age or
menopausal status of the women; and (iv) complete
FSFI data on all study participants. Eight datasets
(four datasets each from Dr. Lori Brotto and Dr.
Markus Wiegel), totaling 618 women, met these
criteria and were considered for inclusion in the
analysis. Diagnosis of HSDD was made by an
independent clinician assessment using the Diag-
nostic and Statistical Manual of Mental Disorders,
4th edition (DSM-IV) criteria, which defines
HSDD as “persistent or recurrent deficiency (or
absence) of sexual fantasies and desire for sexual
activity, which causes marked distress or interper-
sonal difficulty, and which is not better accounted
for by a medical, substance-related, psychiatric, or
other sexual condition.” [15] The FSFI question-
naire was administered in paper and pencil form
using standard instructions and scoring procedures
[1–3]. The diagnostic interview by clinicians was
not controlled across studies in this retrospective
analysis.
Validation Sample
An independent sample of participants was used to
confirm and validate the cutpoint identified. Four
separate studies (not published) conducted by Boe-
hringer Ingelheim (the sponsor) were used, result-
ing in a total sample size of 892 participants. In
these studies, the diagnosis of HSDD was made
using a structured in-depth clinical interview con-
ducted by a trained clinician according to the text
revision of the DSM IV criteria. The FSFI ques-
tionnaire was administered in the standardized
(i.e., paper and pencil) form [1].
Human Subjects Protection
The studies used in this analysis were approved by
the appropriate local institutional review boards,
andsubjectssigned study-specific
consent forms. Each study was conducted in accor-
dance with the standards of the Helsinki Declara-
tion of 1975, as revised in 1983.
informed
Statistical Methods
Internal consistency for the two items comprising
the SD domain of the FSFI was assessed using
Cronbach’s alpha [16]. Receiver-operator charac-
teristic (ROC) curves [17] were constructed using
logistic regression to assess the sensitivity and
specificity of each cutpoint on the scale as com-
pared with the independent clinician-derived diag-
nosis of HSDD. A series of analyses were
performed in which the presence or absence of an
HSDD diagnosis in women without any concomi-
tant sexual dysfunction was regressed on the SD
domain score, which was dichotomized at each of
the possible values of the scale. The ROC curve
wascreatedbyplotting
1—specificity, where sensitivity is the proportion
of participants correctly identified as having
HSDD and specificity is the proportion of partici-
pants correctly identified as not having HSDD.
The curve was used to identify the cutpoint in the
SD scale that simultaneously optimized sensitivity
and specificity. These analyses were repeated com-
sensitivity against
Sexual Desire Cutpoint for the FSFI
3097
J Sex Med 2010;7:3096–3103

Page 3

paring women without an HSDD diagnosis with
women with an HSDD diagnosis and possible
concomitant sexual dysfunctions, overall and
withinpremenopausal
women. The area under the curve is reported as a
measure of diagnostic accuracy.
The sensitivity and specificity of the proposed
cutpoint, along with 95% exact binomial confi-
dence intervals, were subsequently confirmed in
the validation sample.
All analyses were carried out using SAS version
9.1.3 (SAS Institute, Cary, NC, USA), except for
the exact 95% binomial confidence intervals,
which were calculated using Stata version 7 (Stata-
Corp LP, College Station, TX, USA).
and postmenopausal
Results
Development of a Diagnostic Cutpoint
Since each dataset contained a relatively small
number of participants and a limited number of
women with HSDD (either alone or in combina-
tion with another sexual disorder), the eight
datasets were combined for this analysis. Charac-
teristics of the eight datasets, separately and com-
bined, are given in Table 1.
The age range of women in the pooled sample
was 18–74, with an average age of 37.8 years (stan-
dard deviation of 13.5 years). Approximately
72.5% were premenopausal, 14.4% perimeno-
pausal, and the remaining 13.1% were postmeno-
pausal. Menopausal status was collected for 248
(40.1%) women, and the remaining women were
categorized as follows: women aged 45 or younger
were considered premenopausal, women between
the ages of 46 and 59 were considered perimeno-
pausal, and women aged 60 or older were consid-
ered postmenopausal. Of the 618 participants, 159
participants had an HSDD diagnosis and 254 par-
ticipants were sexually healthy controls; the
remaining participants were diagnosed with sexual
disorders other than HSDD.
The SD domain of the FSFI consists of two
Likert-type items, one item on frequency and
another item on intensity of sexual desire. The
internal consistency of the SD domain was
assessed and found to have a Cronbach’s alpha of
0.92. This high value implies that the SD scale is
one-dimensional, and the single cutpoint arising
from a Classification and Regression Trees
(CART) analysis with a unidimensional scale
would have yielded results similar to those
obtained by using a conventional ROC analysis.
Figure 1A shows an ROC curve predicting
HSDD as a function of SD domain scores com-
paring women without sexual dysfunction (con-
trols, n = 254) with women with HSDD only
(n = 49). Classifying women with SD scores of 5 or
less as positive cases (i.e., having the condition)
and women with SD scores of 6 or above as nega-
tive cases (i.e., not having the condition) simulta-
neously maximizes sensitivity (75.5%, 37 out of 49
women with HSDD were correctly classified) and
specificity (83.5%, 212 of the 254 women without
sexualdysfunction were
Figure 1B shows an ROC curve predicting HSDD
as a function of SD domain scores comparing
women without sexual dysfunction (controls,
n = 254) with HSDD women with or without
other sexual dysfunctions (n = 159). A similar clas-
sification scheme maximized sensitivity (74.8%,
119 out of 159 women with HSDD were correctly
classified) and specificity (83.5%, 212 of 254
women without sexual dysfunction were correctly
classified as not having HSDD).
Separate ROC curves (Figure 1C, D) that com-
pared women with and without sexual dysfunction
and by menopausal status were constructed. Clas-
sifyingwomenwithSDscoresof5orlessaspositive
cases and women with SD scores of 6 or above as
negative cases, simultaneously maximized sensitiv-
ity (premenopausal: 70.0%, 70 out of 100 women
with HSDD were correctly classified; postmeno-
pausal: 82.4%, 14 out of 17 women with HSDD
were correctly classified) and specificity (premeno-
pausal: 87.6%, 177 of the 202 women without
HSDD were correctly classified; postmenopausal:
62.1%, 18 of the 29 women without HSDD were
correctly classified) in both subsets of women.
All four ROC curves provided similar results.
Specifically, the use of a diagnostic cutpoint to
classify women with SD scores of 5 or less on the
SD domain as having HSDD and those with SD
scores of 6 or more as not having HSDD maxi-
mized diagnostic sensitivity and specificity.
correctly classified).
Validation of the Diagnostic Cutpoint
The cutpoint was validated in four separate studies
and in a fifth analysis that combined all four
studies. The age range was 18–68, with an average
age of 41.3 years (standard deviation of 11.9 years).
Approximately 59.1%
15.0% perimenopausal, and the remaining 25.8%
were postmenopausal. Menopause status was col-
lected for 770 (86.3%) women, and the remaining
women were categorized in the manner described
above for the development sample. Of the 892
were premenopausal,
3098
Gerstenberger et al.
J Sex Med 2010;7:3096–3103

Page 4

Table 1
Selected characteristics of women included in the sexual desire cutpoint development datasets
Wiegel datasets
Brotto datasets
Total
Study 1
Study 2
Study 3
Study 4
Study 1
Study 2
Study 3
Study 4
Sample size in analysis dataset (n)
185
34
41
259
26
24
27
22
618
Diagnosis
Sexually healthy controls (n)
86
17
16
131
2
0
2
0
254
HSDD only (n)
8
0
6
0
9
13
4
9
49
HSDD + other dysfunction (n)
49
4
7
0
12
10
15
13
110
Other (e.g., sexual pain disorder) (n)
42
13
12
128
3
1
6
0
205
Key variables
Age
Mean ? std. deviation
28.5 ? 8.7
30.8 ? 9.1
52.2 ? 10.2
40.1 ? 13.0
37.0 ? 9.8
55.0 ? 7.8
42.3 ? 12.2
49.4 ? 11.0
37.8 ? 13.5
Range
18–53
18–53
26–74
21–69
24–55
31–64
25–65
23–68
18–74
Menopausal status*
Premenopausal (n)
170
30
25
174
19
4
18
8
448
Perimenopausal (n)
15
4
0
33
7
12
8
10
89
Postmenopausal (n)
0
0
16
52
0
8
1
4
81
FSFI sexual desire domain score (scaled, range 1.2–6)
Overall
(mean ? std. deviation)
4.08 ? 1.16
4.13 ? 1.19
2.85 ? 1.20
3.50 ? 1.37
2.15 ? 0.93
1.65 ? 0.85
2.00 ? 1.03
2.40 ? 0.89
3.43 ? 1.41
Sexually healthy controls (mean ? std. deviation)
4.63 ? 0.90
4.69 ? 0.74
3.15 ? 1.49
4.14 ? 1.13
2.10 ? 0.42
—
1.80 ? 0.85
—
4.24 ? 1.15
HSDD Only (mean ? std. deviation)
3.38 ? 0.71
—
2.60 ? 0.98
—
2.07 ? 1.28
1.80 ? 1.07
1.95 ? 0.57
2.67 ? 0.91
2.38 ? 1.10
HSDD + other dysfunction (mean ? std. deviation)
3.16 ? 0.98
4.35 ? 1.14
1.97 ? 0.75
—
2.10 ? 0.65
1.50 ? 0.51
2.00 ? 1.06
2.22 ? 0.86
2.59 ? 1.12
Other (mean ? std. deviation)
4.14 ? 1.21
3.32 ? 1.29
3.10 ? 0.84
2.83 ? 1.27
2.60 ? 1.25
1.20
2.10 ? 1.41
—
3.12 ? 1.35
*For datasets without information on menopausal status, women 45 or younger were assumed to be premenopausal, women aged 46–59 were assumed to be perimenopausal, and women 60 or older were assumed to be
postmenopausal.
HSDD = hypoactive sexual desire disorder; FSFI = Female Sexual Function Index.
Sexual Desire Cutpoint for the FSFI
3099
J Sex Med 2010;7:3096–3103

Page 5

participants, 454 participants had an HSDD diag-
nosis and 300 participants were sexually healthy
controls; the remaining participants were diag-
nosed with sexual disorders other than HSDD.
Characteristics of the four validation datasets,
separately and combined, are given in Table 2.
The range of specificity across the five analyses
was 90.3–97.1% (Table 3). The sensitivity range
A
B
CD
1 -Specificity
Sensitivity
Area under the curve = 0.87Area under the curve = 0.85
Area under the curve = 0.85Area under the curve = 0.77
Figure 1 Receiver-operator characteristic (ROC) curves predicting hypoactive sexual desire disorder (HSDD) as a function
of sexual desire (SD) domain scores (range: 2–10) comparing (A) women without sexual dysfunction (controls, n = 254) to
women with HSDD only (n = 49), (B) women without sexual dysfunction (controls, n = 254) to HSDD women with or without
other sexual dysfunctions (n = 159), (C) premenopausal women without sexual dysfunction (controls, n = 202) to premeno-
pausal women with HSDD or HSDD and concomitant sexual dysfunction (n = 100), and (D) postmenopausal women without
sexual dysfunction (controls, n = 29) to postmenopausal women with HSDD or HSDD and concomitant sexual dysfunction
(n = 17). Each point in the figure is labeled by its corresponding cutpoint (for example, “5–6” means that women with a SD
domain score of 5 or less was predicted to have HSDD and women with a SD domain score of 6 or greater was predicted
to not have HSDD) and shows the sensitivity and 1-specificity of the cutpoint used to predict HSDD. A cutpoint between 5
and 6 (women with SD scores of 5 or less are considered to be positive cases, while women with SD scores of 6 or above
are considered to be negative cases) simultaneously maximizes sensitivity and specificity in all four cases. The area under
each curve has been calculated as a measure of accuracy.
Table 2
Selected characteristics of women included in the sexual desire cutpoint validation datasets
Validation datasets
Study AStudy BStudy CStudy DOverall
Total sample size (n)
Diagnosis
Sexually healthy controls (n)
HSDD only (n)
HSDD + other dysfunction (n)
Other (e.g., sexual pain disorder) (n)
Missing (n)
Key variables
Age
Mean ? std. deviation
Range
Menopausal status*
Premenopausal (n)
Perimenopausal (n)
Postmenopausal (n)
90248258 296892
35
31
0
24
0
62
112
124
118
12
79
145
27
23
22
300
406
48
95
43
9
47
18
1
3
33.3 ? 12.0
18–61
46.2 ? 11.1
18–64
45.3 ? 11.7
20–68
36.3 ? 9.2
18–53
41.3 ? 11.9
18–68
75
12
3
98
60
90
129
19
110
226
43
27
528
134
230
*For datasets without information on menopausal status, women 45 or younger were assumed to be premenopausal, women aged 46–59 were assumed to be
perimenopausal, and women 60 or older were assumed to be postmenopausal.
HSDD = hypoactive sexual desire disorder.
3100
Gerstenberger et al.
J Sex Med 2010;7:3096–3103