A directed acyclic graph (DAG)53 we constructed shows the anticipated relationships between menopausal status, low sexual desire/hypoactive sexual desire disorder (HSDD), and the other variables being evaluated in this analysis. The direction of the arrows between any 2 variables indicates the known (as indicated by the literature) or assumed directionality of the relationship. We considered body mass index (BMI) as a confounder because there is evidence suggesting an association between menopausal status and BMI. Heavier women tend to go through the menopausal transition at a later age,66 perhaps due to a relationship between adiposity and hormonal production. In addition, women with a higher BMI may have lower self-image, which may result in low sexual desire or HSDD. Although the mechanism is not always apparent, smoking status was included in this DAG because of its relationship with menopausal status. The age at natural menopause is influenced by whether and how much a woman currently smokes67 and may be associated with low sexual desire and HSDD either causally or because of its associations with other health outcomes or risky behaviors. Depression was included in the DAG based on a review by West et al2 suggesting that women with psychiatric comorbidities, especially depression and anxiety, are at higher risk of sexual dysfunction. There is also literature suggesting menopause may lead to depression, perhaps because of hormonal fluctuations.68 Similarly, antidepressants were included in the DAG because the most popular type, the selective serotonin reuptake inhibitors, are known to have sexual dysfunction as a critical adverse effect, and they are used for the treatment of depression and menopausal hot flashes.69 Note that the directionality of the arrows indicates that depression, antidepressants, and exogenous hormones are potential intermediates along the causal pathway between menopause and low sexual desire or HSDD and cannot be confounders of this relationship. The directionality of the smoking status and depression association is unclear because some studies indicate that depression leads to smoking and other studies indicate that smoking leads to depression, perhaps through some unmeasured confounders.70 Also, the age at which smoking begins may affect the directionality of its relationship with depression.
West SL, D’Aloisio AA, Agans RP, Kalsbeek WD, Borisov NN, Thorp JM. Prevalence of Low Sexual Desire and Hypoactive Sexual Desire Disorder in a Nationally Representative Sample of US Women. Arch Intern Med. 2008;168(13):1441-1449. doi:10.1001/archinte.168.13.1441
Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008
We sought to estimate the prevalence of low sexual desire and hypoactive sexual desire disorder (HSDD) in US women, focusing on their menopausal status.
We performed a cross-sectional study. From a probability sample of households, 2207 US women aged 30 to 70 years and in stable relationships (≥3 months) were interviewed by telephone. The analysis focused on 755 premenopausal women and 552 naturally and 637 surgically menopausal women. Low sexual desire was defined using the Profile of Female Sexual Function desire domain, and HSDD was defined using the Profile of Female Sexual Function and the Personal Distress Scale.
Prevalence of low sexual desire ranged from 26.7% among premenopausal women to 52.4% among naturally menopausal women. The prevalence of HSDD was highest among surgically menopausal women (12.5%). Compared with premenopausal women and adjusting for age, race/ethnicity, educational level, and smoking status, the prevalence ratios for HSDD were 2.3 (95% confidence interval, 1.2-4.5) for surgically menopausal women and 1.2 (0.5-2.8) for naturally menopausal women; the prevalence ratios for low sexual desire were 1.3 (0.9-1.9) and 1.5 (1.0-2.2) for surgically and naturally menopausal women, respectively.
Prevalence of low sexual desire is elevated among surgically and naturally menopausal women vs premenopausal women. Distress about low desire (HSDD) appears to be more than twice as prevalent among surgically menopausal women vs premenopausal women, although the estimate is fairly imprecise.
The earliest attempt to quantify the prevalence of female sexual disorders was in 1929 when Davis1 reported on the sex lives of 2200 women. Many researchers have queried women about their sexuality since the study by Davis was published,2 but they have used different terms, including the clinical definition of hypoactive sexual desire disorder (HSDD) from the Diagnostic and Statistical Manual for Mental Disorders, Fourth Edition.3- 35 This definition has 3 criteria: a deficiency or absence of sexual fantasies and desire for sexual activity; the deficiency causes marked distress or interpersonal difficulty; and the dysfunction cannot be attributed to another Axis I disorder or result from the physiological effects of a substance or a general medical condition. The Hite Report,36 the first study, to our knowledge, to survey a convenience sample of women about sexual interest, found that only 1% had no interest in sex. The study by Laumann et al18 was the first to use a nationally representative survey to address female sexual problems. Their definition of sexual desire disorders as lacking interest in sex for a period of several months in the past 12 months18 does not conform with current definitions and does not address distress. The more recent studies by Dennerstein et al3 and Leiblum et al6 of HSDD prevalence among US and European women derived the cutoff scores on 2 validated instruments, the Profile of Female Sexual Function (PFSF)37,38 and the Personal Distress Scale (PDS),39 and used their cutoff scores to classify women with low sexual desire (hereinafter called “low desire”) and HSDD. Both studies surveyed women from a market research database, in contrast with our study, which used a national probability sample.
Because little is known about the prevalence of sexual desire disorders, including HSDD, we conducted a telephone interview of women aged 30 to 70 years who were living in the continental United States. Although we also collected data on relationship factors and comorbid conditions, this report focuses on the demographic factors that influenced the prevalence of low desire and HSDD among premenopausal and menopausal (natural or surgical) women.
We conducted a cross-sectional study with a nationally representative sample of women aged 30 to 70 years in steady relationships for 3 months or longer. We required women to be in relationships for 3 months or longer for 2 reasons: so that they would have the opportunity for sexual intercourse and to control for partner factors in the sexual relationship. We stratified based on oophorectomy status and age (groups: 30-39, 40-49, 50-59, and 60-70 years), oversampling women who underwent bilateral oophorectomy. Our sample size estimates were based on attrition data from our pilot study40 and the frequency of oophorectomies in the general population.41 After approval by the University of North Carolina Public Health Institutional Review Board, the Survey Research Unit (SRU) at The University of North Carolina at Chapel Hill conducted the survey between September 9, 2004, and March 30, 2005. A midcourse revision occurred on January 11, 2005, to cease interviewing women 60 years and older and oversampling women aged 30 to 39 years who underwent oophorectomies.
We chose a probability sample of households with telephone line access in the continental United States. We selected the sample using a dual-frame approach in which 10% of the sample was chosen from a list-assisted random-digit dialing frame of households with telephones, whereas the remainder was chosen from a complementary (ie, nonoverlapping) targeted-list frame of directory-listed residential telephone numbers for which recent household demographic information was available. The targeted frame was stratified by 4 age groups (30-39, 40-49, 50-59, and 60-70 years) and proportionally allocated to corresponding oophorectomy rates.
Upon contact with a household member, the SRU administered a standardized introduction to determine study eligibility (based on age, oophorectomy status, and relationship status). The selected woman was read a description of the study and the required elements for informed consent; all women who were interviewed gave verbal consent.
Data were collected through computer-assisted telephone interviews administered by formally trained female interviewers selected for their comfort in ascertaining sensitive data; 10% of all interviews were monitored by supervisory staff. We chose computer-assisted telephone interview technology rather than telephone audio computer-assisted self interviews based on a pilot study comparing these 2 modalities for data collection.40 The pilot study showed better correlations between overall sexual satisfaction responses and the PFSF desire domain estimates when data were collected by female interviewers.
Of 45 562 telephone numbers put into calling, 28 320 were ineligible because the number was not for a household or no woman in the household met the study's inclusion requirements. We did not screen 13 927 telephone numbers due to nonresponse or screening refusals. Among the eligible households screened, 1108 selected women refused to participate; 2207 women completed the interview. The overall response rate was 56.7% and accounted for ineligibility, unknown status, and the different types of eligibility status codes (partial, refusal, no further contact, and incomplete) as suggested by the American Association for Public Opinion Research guidelines.42
Base sample weights were predicated on original probabilities of choosing households from a national frame of telephone numbers. We adjusted for the number of phone lines reaching the household, the retention probabilities for achieving sample size targets by age and oophorectomy status, nonresponse based on age stratum, and differential phone coverage to reduce coverage bias.43 The sample design resulted in variable weights that required trimming to redistribute extreme weights so that summing the new weights remained roughly the same.44 This trimming process reduced the variable effect of the weights with limited effect on the weighted estimates. We aligned our sample to the key demographic variables of the population by poststratification adjustments using the March 2004 Current Population Survey.41 We used SRU weights that were normalized to the total sample size for most analyses. For the multivariable analyses, we standardized the SRU weights to sum to the sample size for each age-menopausal status stratum, with the standard errors within each stratum corresponding to the applicable sample size for that stratum.
We measured female sexual function using the 37-item PFSF and the 7-item PDS, which were developed by Procter & Gamble Pharmaceuticals and are not in the public domain. The PFSF was validated as a self-report survey for use in menopausal women with low desire: Cronbach α was .94 for surgical menopause and .89 for natural menopause.37,38 The PFSF incorporates 7 domains (sexual desire, sexual pleasure, sexual arousal, orgasm, sexual responsiveness, sexual concerns, and sexual self-image) that are relevant to women's sexual function45- 49; however, we used only the sexual desire domain for this research. Women were asked about sexuality issues during the past 30 days and responded using a scale from 1 to 6, corresponding to always, very often, often, sometimes, seldom, and never. Low scores on the PFSF desire domain indicate low desire. The PDS has been validated as a self-report survey for use in menopausal women and uses the same scoring as the PFSF, with low scores indicating greater distress owing to lack of sexual desire.39
The study questionnaire addressed demographic factors such as age, current marital status, race/ethnicity, educational level, and income. We also collected data on reproductive history, relationship factors, sexual frequency, and numerous medical conditions. To screen for current depression, we used a score of 8 or more on the 8-item Center for Epidemiologic Studies instrument (CES-D-8).50 We ascertained quality-of-life characteristics using the validated Short Form 12 (SF-12)51 and the EuroQoL.52
Our analysis focused on the relationship of menopausal status with low desire and HSDD prevalence, and evaluated specific sociodemographic and biological factors as confounders, including age, educational level, race/ethnicity, smoking status, body mass index, current depression, antidepressant use among those with a previous depression diagnosis, and menopausal estrogen use (only among menopausal women). We evaluated confounders by using a directed acyclic graph (DAG)53 (Figure) that considered the covariates potentially influencing the relationship for menopausal status with low desire and HSDD prevalence, according to the published literature.2
We assessed menopausal status by accounting for a woman's last menstrual period, recent or current pregnancy, reproductive surgeries (bilateral oophorectomy and/or hysterectomy), age, reasons for no menses in the past 12 months, and current birth control or hormone therapy (estrogens and progesterones). We defined low desire as a PFSF desire domain transformed score of less than 40, whereas HSDD was defined as low desire with a PDS transformed score of less than 60. The derivation of these cutoff scores is described in the studies by Dennerstein et al3 and Leiblum et al.6 We used published guidelines for scoring the SF-12 in which we evaluated each of the 8 items and used the algorithm of Ware et al54 to develop norm-based scores based on 1998 US population means and standard deviations.
We conducted univariable analyses using PROC SURVEYMEANS in SAS statistical software55 with SRU weights to compute population-based prevalence estimates for low desire and HSDD. For our multivariable analyses, we used a weighted log-binomial model that applied a linear regression model to logarithms of the prevalence of low desire and HSDD to estimate prevalence ratios with 95% confidence intervals for menopausal status associations with low desire and HSDD. We used generalized estimating equations within PROC GENMOD from SAS statistical software to fit the model for “events/trials” syntax with “trials” as the standardized SRU weights (w) and with events as the products of w × outcomes (y), with y equal to 1 if HSDD (or low desire) present and 0 if otherwise. Our weighted log-binomial model estimated regression coefficients and corresponding robust standard errors using SAS55 that are considered comparable to what would be obtained with SUDAAN statistical software56 if its scope included this type of model.57 We considered each covariate as a modifier of the association for menopausal status with low desire and HSDD by including multiplicative interaction terms (menopausal status by each covariate) in addition to individual covariates within the final adjusted log-binomial models. For each interaction term, we evaluated the Wald test statistic for type 3 generalized estimating equation analysis with appropriate degrees of freedom (α = .10).
We interviewed 2207 women: 22.2% aged 30 to 39 years, 26.9% aged 40 to 49 years, 45.5% aged 50 to 59 years, and 5.4% aged 60 to 70 years. Most women were non-Hispanic white (85.5%), were married or living with someone (90.9%), had an annual household income of more than $60 000 (54.6%), and had high school or less, vocational school, or some college as their highest educational level (62.4%). Because only 24 women were pregnant and we could not determine whether the perimenopausal women were in early or late transition, our analysis focuses on the 1944 nonpregnant women who were premenopausal or menopausal (natural or surgical) at the time of the survey.
Because smoking status, age, body mass index, race/ethnicity, and educational level were potential confounders according to the DAG (Figure), we adjusted for them in our multivariable analyses of the association for menopausal status with low desire and HSDD. We excluded body mass index from our final log-binomial models because adjusting for it did not affect the estimated associations. From our DAG, current depression status, antidepressant use, and exogenous hormones cannot be confounders because they are intermediates in the causal pathway between menopausal status and low desire/HSDD; however, we did evaluate them as effect modifiers (ie, whether depression modifies the association of menopausal status with HSDD). Depression and antidepressants were strongly associated with low desire and HSDD, but we found little evidence that depression or antidepressants modified the association for menopausal status with low desire or HSDD. Also, current use of hormonal therapy among menopausal women did not modify the association between type of menopause and low desire or HSDD.
Table 1 provides the norm-based SF-12 domain and summary scores for the 1944 women in our study. The means for all the SF-12 domain and summary scores are very close to 50 with narrow 95% confidence intervals, indicating that the women in our analysis were similar to the 1998 US female population used by Ware et al.54
Table 2 shows the study characteristics and the prevalence estimates with 95% confidence intervals for low desire and HSDD. These results incorporate the original SRU weights to achieve representativeness for the female population residing within the continental United States. Black, non-Hispanic women had a lower prevalence of reduced desire and HSDD than either white non-Hispanic or Hispanic women, but less than 15% of our study population was black or Hispanic. In general, the prevalence of low desire was at least 3 times higher than the prevalence of HSDD, indicating that many more women reported low desire than were distressed by this condition.
The prevalence of low desire and HSDD, stratified by markers of endogenous hormones, including age, menopausal status, and timing of oophorectomy, is shown in Table 3. The lowest prevalence of reduced sexual desire was in younger women, specifically premenopausal women and those younger than 45 years at interview and when their ovaries were removed. Low desire increased with age, but surgically menopausal women had higher HSDD prevalence than either premenopausal or naturally menopausal women regardless of their age at interview or when their ovaries were removed. Younger women who had their ovaries removed appeared to have the highest HSDD prevalence.
The estimated prevalence ratios for the association of menopausal status with low desire and HSDD, adjusted for age, race/ethnicity, educational level, and smoking status, are shown in Table 4. Compared with premenopausal women, those who underwent surgical menopause had more than twice the prevalence of HSDD, although the estimate is fairly imprecise. Prevalence of low desire in both surgically and naturally menopausal women was slightly elevated compared with premenopausal women. There was a suggestion that the prevalence ratios for both low desire and HSDD differed by race/ethnicity (results not shown); however, given the relatively small number of nonwhite women interviewed, this may be a spurious finding.
The mean number of sexual encounters in the 30 days before the interview was similar between those with low desire and HSDD (Table 5). However, regardless of menopausal status, mean sexual encounters were reduced for women with low desire or HSDD compared with women without each condition, respectively. Because of study design, we cannot determine whether low desire or HSDD leads to less intercourse or vice versa.
We conducted the first study to provide a nationally representative estimate of sexual function among US women aged 30 to 70 years using validated self-reported outcome measures, the PFSF and the PDS. The overall prevalence of low desire was 36.2%, but for HSDD, characterized by distress about low desire, the prevalence was 8.3%. There was a reduction in the mean number of sexual encounters in the past month between women with low desire or HSDD vs those without each condition; however, there was very little difference between those with low desire and those with HSDD. Both low desire and HSDD prevalence varied by age, race/ethnicity, educational level, body mass index, current smoking status, current depression assessed by the CES-D-8, current use of antidepressants in those with a previous depression diagnosis, hormonal therapy (in menopausal women), and menopausal status.
In a recent US study of 952 premenopausal and menopausal (surgical and natural) women in relationships for 3 months or longer who were surveyed using the PFSF and PDS, 24% to 36% reported low desire, whereas 9% to 26% reported HSDD.6 Similar to the present study, Leiblum et al6 reported the HSDD prevalence as highest among surgically menopausal women.
The most frequently cited research on female sexual function was published in 1999 by Laumann et al.18 Of 1749 women aged 18 to 59 years (median age group, 30-39 years) who had been sexually active in the previous 12 months, 43% reported sexual problems. The authors found that emotional problems, marital and socioeconomic status, and educational level were predictors of dysfunction, but they did not address menopausal status or distress about sexual problems. Subsequently, Bancroft et al7 identified women in heterosexual relationships for 6 months or longer using random-digit dialing to measure distress about sexuality, with 44.3% reporting impaired sexual response and 24.4% reporting that they were distressed by their impairment.
We estimated differences between the prevalence of low desire and HSDD in certain subgroups. Women with a graduate degree and those aged 60 to 70 years had very low HSDD prevalence estimates, but their prevalence estimates of low desire were 10 times higher. Although this suggests that low desire may be acceptable, ie, not distressing to these women, the number of women in these strata was relatively small. The women who were younger than 45 years when interviewed and when their ovaries were removed had the highest HSDD prevalence of all women studied, with a slightly lower prevalence in women 45 years or older when interviewed and when their ovaries were removed. These results are contrary to those in studies by Teplin et al58 and Aziz et al,59 which both reported no effect of oophorectomy on sexual function. However, these studies used different sexual function measures than the PFSF and PDS. The high HSDD rates among surgically menopausal women who had their ovaries removed prior to but in close proximity to their interview may be affected by the sudden hormonal changes with an oophorectomy, in contrast to the more gradual hormonal changes with natural menopause.60,61 Naturally menopausal women had a high prevalence of low desire but the lowest prevalence of HSDD. Besides having more time to physically adjust to the hormonal changes, women undergoing natural menopause may be expecting changes in their sexual response, which would reduce their distress about having low desire.
The prevalence of low desire among menopausal women who currently used exogenous hormones was substantially lower than in those who did not use hormones; this relationship was reversed for women with HSDD, although differences were minimal. However, the association for menopausal status with low desire or HSDD did not differ in women who did and did not use exogenous hormones. An important consideration is whether the women who use exogenous estrogens differ from those who do not with respect to factors not evaluated, such as propensity for risk taking and underlying risk of breast cancer or cardiovascular disease. The difference between exogenous hormone users and nonusers may be a function of risk-benefit considerations incorporating the importance of sexuality in their lives—ie, that the hormone therapy users actively seek ways to improve their sexual function despite the potential health risks with these medications.62
Our study methods and results correspond with previously published surveys but have several improvements.2 Like the studies published by Dennerstein et al,3 Hayes et al,5 and Leiblum et al,6 we used the validated PFSF and the PDS to estimate the prevalence of low desire and HSDD. However, we used a probability sample to identify our study population, which allowed us to calculate nationally representative prevalence estimates. Our results were comparable to those from Leiblum et al for HSDD in surgically menopausal women despite the difference in the data collection method.
Our study had several methodological advantages over past work. We conceptualized our analytic strategy by developing a DAG before conducting our multivariable analyses to determine a priori which variables are potential confounders for the association of menopausal status with low desire and HSDD. In addition, as low desire and HSDD are prevalent conditions, we used log-binomial models to estimate adjusted associations as these models provide prevalence ratios rather than odds ratios, which are prone to overestimating an association when the outcome is common.63 Last, despite not having hormone levels to assess the menopausal transition,64 we were able to evaluate whether most women were surgically or naturally menopausal based on their questionnaire responses.
Our focus on women with a current sexual partner of at least 3 months' duration limits study generalizability to this subset of US women. However, by limiting our study to women who had the potential to have sex, we improved the validity of our findings by minimizing partner issues as a confounder for low desire and HSDD. Limited study participation from women in certain age and oophorectomy status strata resulted in small cell sizes, which affected the precision of our estimates. In particular, women aged 60 to 70 years were reluctant to be interviewed, and the low prevalence of oophorectomy among women aged 30 to 39 years made it difficult to identify women within this stratum to interview. Given the difficulties in reaching participants for telephone interviews, such as caller ID and answering machines, having only 33.4% (1108/3315) of eligible women decline to participate in such a sensitive interview is reasonable by today's survey standards.
In 2000, there were more than 42 million US women older than 50 years.65 Given our estimates of low desire and HSDD prevalence, at least 16 million women aged 50 years or older currently experience low desire, and approximately 4 million are distressed by their low desire. African American women had the lowest prevalence of low desire and HSDD in our study, although estimates were very imprecise because of lack of power for evaluating racial/ethnic differences. Future studies should oversample African Americans to develop more robust estimates for this group.
The present analysis evaluates sociodemographic, menopausal, and related hormonal factors influencing sexual desire and HSDD. Future analyses will consider partner relationship factors in addition to depression and antidepressants, which our study suggested may influence low desire and HSDD among US women.
Correspondence: Suzanne L. West, PhD, MPH, RTI International, 3040 Cornwallis Rd, PO Box 12194, Research Triangle Park, NC 27709-2194 (firstname.lastname@example.org).
Accepted for Publication: January 13, 2008.
Author Contributions: Drs West and D’Aloisio had full access to all the data in the study and take responsibility for its integrity and the accuracy of the data analysis. Study concept and design: West, Agans, Kalsbeek, Borisov, and Thorp. Acquisition of data: West, Agans, and Kalsbeek. Analysis and interpretation of data: West, D’Aloisio, Agans, Kalsbeek, and Thorp. Drafting of the manuscript: West, D’Aloisio, Agans, Kalsbeek, and Thorp. Critical revision of the manuscript for important intellectual content: West, D’Aloisio, Agans, Kalsbeek, Borisov, and Thorp. Statistical analysis: West, D’Aloisio, Agans, and Kalsbeek. Obtained funding: West, Agans, and Kalsbeek. Study supervision: West and Agans.
Financial Disclosure: Drs West and D’Aloisio received research support from Procter & Gamble Pharmaceuticals. Dr Borisov is employed by Procter & Gamble Pharmaceuticals and owns the company's stock and has received stock options. Dr Thorp received research support from the following pharmaceutical companies: Wyeth, Boehringer-Ingelheim, Procter & Gamble, and GlaxoSmithKline.
Funding/Support: This study was funded by Procter & Gamble Pharmaceuticals.
Role of the Sponsor: The funding agency had a limited role in the design of this study and in the review and approval of the manuscript.
Additional Contributions: Nikki McKoy provided administrative assistance in handling the incentives and preparing the manuscript for submission, and Lisa Vinikoor, PhD, helped pilot the questionnaire for the study. We thank Gary G. Koch, PhD, for his statistical consulting and supervising, and Peter Nyangweso, MS, for validating Dr Koch's analytic approach for using log-linear models on survey data. Patricia Koochaki, PhD, Cindy Rodenberg, PhD, and Michael Steinbuch, PhD, from Procter & Gamble Pharmaceuticals provided useful comments regarding the design of the study. Finally, this study could not have been completed without the excellent supervision of Thu-Mai Christian, MA, and the outstanding female interviewers at the SRU.