Psychometric validation of the female sexual function index (FSFI) in cancer survivors
The Female Sexual Function Index (FSFI) is the most commonly used self-report instrument to measure sexual functioning among women cancer survivors. Despite this, the validity and reliability of the FSFI for use in cancer populations has not been established.
Data were combined from 3 separate institutional review board-approved studies of the psychosexual adjustment of women cancer survivors conducted at Memorial Sloan-Kettering Cancer Center. Psychometric analysis was applied to the FSFI responses from 181 women comprising 4 cohorts, including survivors of: gynecologic cancer (all types; 2 cohorts), malignancies requiring bone-marrow/stem cell transplantation, and early stage cervical cancer.
A factor analysis supported the subscale structure of the FSFI, yielding results nearly identical to those from the original FSFI validation study. Internal consistency reliability was 0.94 for the FSFI total score and ranged from 0.85 to 0.94 for the domain scores. Corrected item-total correlations ranged from 0.44 to 0.79 for the total score and from 0.62 to 0.88 for the domain scores. FSFI scores were correlated negatively with measures of depression, distress, and menopausal symptoms and were correlated positively with quality of life. FSFI scores exhibited a preliminary ability to discriminate between women based on whether or not they received chemotherapy and/or radiation.
The FSFI demonstrated strong psychometric properties in this study, supporting its continued use for monitoring sexual function and cancer-related dysfunction among sexually active women who are cancer survivors. Cancer 2012. © 2012 American Cancer Society.
Female sexual dysfunction (FSD) is quite common among women who have survived cancer,1-5 and the most commonly used measure to assess sexual functioning in this population is the Female Sexual Function Index (FSFI).6 However, the FSFI originally was developed and validated in healthy women, and the validity and reliability of the FSFI in cancer populations has not been established.
FSD resulting from cancer may differ qualitatively and quantitatively from FSD experienced in otherwise healthy women. Cancer treatments, including surgery, chemotherapy, radiotherapy, and hormone therapy, can directly affect sexual organs and hormone levels,1 resulting in vaginal shortening and stenosis, decreased vaginal lubrication and elasticity, as well as disruption of the sexual response cycle.2 A survey of long-term vaginal and cervical cancer survivors reported rates of FSD well over twice those of matched healthy controls; in particular, survivors had >14 times the odds of dyspareunia compared with controls.4 FSD also has been identified as a primary source of psychological distress after cervical cancer treatment7 and as an independent predictor of depressive symptoms.8 FSD can contribute to poor body image, reinforce stigma felt by those with a cancer diagnosis, and disrupt relationship closeness and intimacy.5 Indeed, some studies have demonstrated that up to one-third of relationships were ended within 1 year after completing cancer treatment.9, 10 Collectively, these studies highlight the necessity of integrating sexual function assessment into routine oncologic clinical care and the importance of identifying FSD for early intervention.
Unfortunately, few psychometrically sound measures of sexual function have been used consistently in female cancer populations.5 The exception is the FSFI, a 19-item instrument assessing 6 domains of sexual function: Desire, Arousal, Lubrication, Orgasm, Satisfaction, and Pain. The original validation study6 established the FSFI's reliability and construct validity in women diagnosed with sexual arousal disorder and in women without sexual difficulties. Later studies demonstrated its validity in women diagnosed with the other major categories of FSD.11, 12 Recently, validation of the FSFI was extended further to women with vulvar intraepithelial neoplasia, a preinvasive skin lesion of the vulva that can become cancerous if left untreated.13 In all these studies, the FSFI total and domain scores displayed excellent internal consistency reliability coefficients (ie, Cronbach α>.80) as well as the ability to differentiate between clinical and nonclinical samples.
The FSFI items and domains were developed to reflect the key features of the major categories of FSD6 as defined by the current diagnostic classification systems in the United States14 and internationally.15 This correspondence with well established theories and diagnostic systems of sexual dysfunction is likely why the FSFI is 1 of the most widely administered measures of female sexual function. It is noteworthy that the FSFI has been cited in studies with gynecologic,16-22 breast,23, 24 rectal,25, 26 and urologic27 cancer patients/survivors, although other specialized questionnaires exist addressing sexual and vaginal problems specific to certain types of cancer and/or treatment.28-30 Despite its widespread use among cancer survivors, it is concerning that no studies to date have empirically evaluated and documented the reliability and validity of the FSFI in female cancer populations. The etiology of sexual dysfunction and symptoms in cancer survivors differs from that experienced by women in the general population; therefore, it is imperative to carefully examine the psychometric properties of the FSFI for use among cancer survivors. The objective of this study was to provide the first systematic evaluation of the factor structure, reliability, and construct validity of the FSFI for measuring the sexual functioning of female survivors.
MATERIALS AND METHODS
For the current analysis, data were combined from 3 separate institutional review board-approved research protocols conducted at Memorial Sloan-Kettering Cancer Center. These studies obtained informed consent from each participant and investigated psychosexual adjustment and reproductive health of women who were cancer survivors (ages 18-50 years). The different purposes and eligibility criteria of the 3 studies resulted in 4 distinct cohorts of survivors when combined for the current analysis. FSFI responses were extracted from 217 women in these 4 distinct cohorts of survivors of gynecologic cancer (all types) (the GYN1 cohort [n = 35] and the GYN2 cohort [n = 51]), malignancies requiring bone marrow/stem-cell transplantation (the BMT cohort [n = 70]), and early stage (stage IB1 or lower) cervical cancer (the CERV cohort [n = 61]). Eligibility criteria for the GYN1 cohort included a history of gynecologic cancer, loss of fertility secondary to cancer treatment, no evidence of disease for at least 1 year, and no biologic or adopted children. The GYN2 and BMT cohorts came from 1 study. The GYN2 survivors had eligibility criteria similar to those for the GYN1 study, except they were not excluded if they had biologic or adopted children, and they had the additional requirement that they had to be eligible for third-party reproductive assistance or had partial fertility options (eg, removal of uterus [hysterectomy] but functioning ovaries or ovaries removed [or ovarian failure] with an intake uterus). The BMT cohort eligibility criteria differed from those of the GYN2 cohort only in the type of cancer required for enrollment. Finally, the CERV cohort came from a longitudinal study that was restricted to survivors of early stage cervical cancer who underwent radical surgery (radical hysterectomy or trachelectomy) alone and received no adjuvant radiotherapy or chemotherapy. In addition, CERV study participation did not depend on fertility status, and patients were enrolled presurgery and were assessed multiple times up to 2 years postsurgery. Many of the clinical differences among the cohorts are apparent in Table 1. Greater detail is available in the published reports of these individual studies.17-19
Table 1. Medical and Demographic Descriptives Among Women With Valid Female Sexual Function Index Assessments, Overall and by Study Group (N = 181)
|Median age [IQR], y|
| At FSFI completion||42.0 [37.0-47.0]||39.0 [36.0-42.0]||33.0 [28.0-38.0]||33.8 [31.0-38.0]||36.2 [31.0-41.0]|
| At cancer diagnosis||40.7 [34.4-45.3]||35.4 [31.8-39.4]||24.9 [14.5-32.8]||32.7 [30.2-37.3]||32.8 [27.8-38.3]|
|Median time since cancer diagnosis [IQR]|
| Years||1.6 [0.9-2.8]||2.3 [1.3-5.4]||6.5 [4.4-13.0]||0.5 [0.3-0.8]||1.7 [0.7-5.2]|
| Months||19.5 [10.6-34.0]||27.6 [15.3-64.3]||78.5 [53.0-155.5]||5.7 [3.2-10.0]||20.7 [7.8-62.1]|
| Cervical||6 (27.3)||40 (85.1)|| ||61 (100)||107 (59.1)|
| GTN||1 (4.5)||3 (6.4)|| || ||4 (2.2)|
| Ovarian||5 (22.7)||1 (2.1)|| || ||6 (3.3)|
| Uterine/endometrial||10 (45.5)||3 (6.4)|| || ||13 (7.2)|
| Hodgkin lymphoma|| || ||10 (19.6)|| ||10 (5.5)|
| Leukemia|| || ||25 (49)|| ||25 (13.8)|
| Other lymphoma|| || ||12 (25.5)|| ||12 (6.6)|
| Pediatric sarcomas|| || ||4 (7.8)|| ||4 (2.2)|
| Hysterectomy||22 (100)||40 (85.1)|| ||21 (34.4)||83 (45.9)|
| Trachelectomy|| ||3 (6.4)|| ||40 (65.6)||43 (23.8)|
| Unknown|| ||4 (8.5)|| || ||4 (2.2)|
| NA|| || ||51 (100)|| ||51 (28.2)|
|Chemotherapy and/or RT|
| Neither||10 (45.5)||26 (55.3)||0 (0)||61 (100)||97 (53.6)|
| Chemotherapy only||8 (36.4)||5 (10.6)||9 (17.6)||0 (0)||22 (12.2)|
| RT only||2 (9.1)||4 (8.5)||1 (2)||0 (0)||7 (3.9)|
| Both||2 (9.1)||12 (25.5)||41 (80.4)||0 (0)||55 (30.4)|
| Non-Hispanic white||20 (90.9)||36 (76.6)||40 (78.4)||47 (77.1)||143 (79)|
| Married/cohabitating||11 (50)||36 (76.6)||26 (51)||39 (63.9)||112 (61.9)|
| ≤High school graduate||2 (9.1)||7 (14.9)||5 (9.8)||7 (11.5)||21 (11.6)|
| Some college||5 (22.7)||5 (10.6)||11 (21.6)||11 (18)||32 (17.7)|
| College graduate||6 (27.3)||18 (38.3)||15 (29.4)||18 (29.5)||57 (31.5)|
| Graduate school||9 (40.9)||17 (36.2)||20 (39.2)||25 (41)||71 (39.2)|
| Full time||14 (63.6)||31 (66)||28 (54.9)||41 (67.1)||114 (63)|
Several additional instruments were administered with the FSFI. For the purposes of the current investigation, we were interested only in the correlations of these instruments with the FSFI scores.
The Female Sexual Function Index
The FSFI is a multidimensional measure of female sexual functioning with 19 items that have ordinal, Likert-type response formats and are scored from 0 (or 1) to 5. The scoring algorithm sums items on each domain/subscale and then scales the sums so that each subscale has a maximum score of 6. The FSFI total score is the sum of the 6 domain/subscale scores and has a maximum score of 36. Higher scores indicate better functioning. A total score ≤26.0 has been validated as a cutoff score for diagnosing FSD.12 The initial FSFI validation study6 reported excellent reliability for the FSFI total score (Cronbach α = .97) and subscales (Cronbach α range, from .89 to 0.96).
The Functional Assessment of Cancer Therapy-Cervix
The Functional Assessment of Cancer Therapy-Cervix (FACT-Cx) questionnaire31 is the Functional Assessment of Cancer Therapy-General (FACT-G) questionnaire32 plus the Cx subscale. The Fact-G version 4 is a 27-item scale that measures quality-of-life (QOL) in patients with 4 subscales that measure physical, functional, social/family, and emotional well being. The FACT-Cx subscale contains 15 additional items specific to symptoms related to cervical cancer, including questions related to body image and sexual functioning.
The Medical Outcomes Study Short-Form Health Survey
The Medical Outcomes Study short-form health survey (SF-12)33 is a self-report survey with 12 questions that measure 8 functional health concepts: physical functioning, role limitations because of physical problems, bodily pain, general health perceptions, vitality, social functioning, role limitations because of emotional problems, and mental health. The SF-12 yields 2 summary scales: the Physical Component Summary (PCS), and the Mental Component Summary (MCS).34
The Center for Epidemiologic Studies Depression Scale
The Center for Epidemiologic Studies Depression Scale (CES-D)35 assesses the frequency of 20 depressive symptoms, each rated on a 4-point scale (0, rarely or none of the time; 1, some of the time; 2, occasionally; and 3, most of the time). Scores ≥16 are suggestive of depression.
Impact of Events Scale
The Impact of Events Scale (IES)36 is a 15-item scale that measures distress levels in response to a specific, potentially traumatic event. Items assess the frequency of intrusive and avoidant thoughts and behaviors about the event. The GYN1, GYN2, and BMT cohorts were asked to endorse items with respect to their cancer-related infertility, whereas the CERV cohort answered the items with respect to their cancer and treatment experience.
Menopausal Symptom Checklist
The Menopausal Symptom Checklist (MSCL)37 has 36 items that assess symptoms associated with menopause. Women rate how bothered they were by symptoms over the last 4 weeks on a scale from 0 to 4 on which anchors are given at 0 (not at all bothered), 2 (somewhat bothered), and 4 (very bothered).
The Abbreviated Dyadic Adjustment Scale
The Dyadic Adjustment Scale (DAS)38 is an instrument that was designed to assess the quality of and satisfaction with relationships as perceived by married or cohabiting couples. The abbreviated DAS (ADAS)39 is a 7-item short-form of the DAS.
The Reproductive Concerns Scale
The Reproductive Concerns Scale (RCS)40 has 14 items that assess the degree of concern and emotional distress over impaired reproductive ability in female cancer survivors. Women rate the relevance of statements regarding possible thoughts and feelings about pregnancy, fertility, and reproduction during the past month on a scale from 0 (not at all relevant) to 4 (very much relevant).
All participants completed the FSFI, CES-D, and IES as part of their participation in larger studies. Completion of the other instruments differed across the cohorts according to the assessment protocols of the larger studies from which the cohorts were extracted. Table 2 displays which of the instruments each of the cohorts completed. These specific instruments were chosen for administration in their respective studies based on their established psychometric validity and reliability, their prior use in studies of cancer survivors and/or other medical populations, their relevance to the goals of their respective studies, and investigator preference. Higher scores on the CES-D, IES, MSCL, RCS, and FACT-Cx subscales indicate relatively poor functioning, whereas high scores on the FSFI, SF-12 MCS and PCS, ADAS, and FACT-G total and subscale scores indicate relatively high functioning.
Table 2. Patient-Reported Outcome Questionnaires Completed by Each Cohort
|ADASa|| ||X||X|| |
|RCS|| ||X||X|| |
|SF-12 PCS|| ||X||X|| |
|SF-12 MCS|| ||X||X|| |
|FACT-Cx|| || || ||X|
FSFI responses from women who reported little or no sexual activity were considered invalid and were excluded from the analysis. Of the 19 FSFI items, 15 items have a response option of “no sexual activity” or “did not attempt intercourse,” which is assigned a score of zero if selected. Although the FSFI scoring algorithm assumes that the zero category indicates the lowest level of functioning on each item's ordinal response scale, there is strong empirical evidence to the contrary.41, 42 Because these 15 items measure sexual activity in addition to their intended sexual functioning domain, they violate the psychometric assumption of item unidimensionality when administered to women with no recent sexual activity. Women who indicated no sexual activity/intercourse or who had missing responses to ≥8 items were excluded from the current analysis, because these women most likely were not sufficiently sexually active for the FSFI to be a valid assessment of their sexual functioning.
Descriptive statistics were tabulated for demographic and medical characteristics of survivors. Medians and interquartile ranges summarized continuous variables. Frequencies and percentages summarized categorical variables.
The internal consistency reliability of the FSFI total and subscale scores was evaluated by using Cronbach α coefficients. Differences in the reliabilities of the scales across the 4 cohorts were tested using a test43 for differences between k independent coefficient alphas. Corrected item-total correlations between each item score and the scale score with the item removed from the total were calculated as additional measures of scale and item reliability. These were computed both for the total score and for subscale scores. Proportions of item responses that scored ≥4 were calculated as a measure of item severity.
The validity of the FSFI was examined in 3 ways. First, factorial validity was assessed by conducting an exploratory factor analysis to determine whether the FSFI factor structure identified in the original validation and, by extension, the structure of the 6 domain subscales were valid among female cancer survivors. Similar to the original study, factors were extracted using principal components analysis with a varimax rotation applied to the extracted factors. Factors with eigenvalues >1.0 were retained in the final solution, and an item-factor loading of at least 0.30 was considered the threshold for considering a loading salient.
Second, construct validity was assessed by examining the patterns of Pearson correlations between the FSFI scales and the other completed self-report measures. We hypothesized that scores on these instruments would vary in the degree and direction of their association with the content domains thought to be measured by the FSFI in a pattern consistent with theoretical expectations. In general, the Desire, Arousal, and Satisfaction domains were expected to correlate strongly with measures of psychological and emotional constructs (eg, CES-D, SF-12 MCS); whereas the Lubrication, Orgasm, and Pain domains were expected to correlate strongly with measures of physical health-related constructs (eg, FACT Physical Well Being, SF-12 PCS).
Third, discriminant validity was evaluated by comparing FSFI scores between groups of survivors that theoretically should differ in their levels of sexual functioning; specifically, between groups defined by whether or not they received chemotherapy and/or radiation. Scores were compared across treatment groups using independent-sample t tests and analyses of variance. Proportions of patients meeting the FSFI cutoff for FSD were compared using Fisher exact tests. Because chemotherapy/radiation treatment is confounded with cohort, the analysis of discriminant validity should be considered exploratory.
The threshold for statistical significance throughout this study was P < .05. All analyses were conducted in SAS (version 9.2; SAS Institute, Inc., Cary, SC) and PASW (version 19.0; SPSS, Inc., Chicago Ill) software.
Of the 217 women originally examined, 36 (16.6%) were excluded from the analysis because they had ≥8 zero/missing FSFI responses. Table 1 summarizes the medical and demographic characteristics of the 181 survivors who had valid FSFI assessments. The median age at FSFI completion and cancer diagnosis, respectively, was 36.2 years and 32.8 years. There was a median of 1.7 years between the diagnosis of cancer and completion of the FSFI. The BMT cohort was diagnosed at a younger age (median, 24.9 years) and had more time since diagnosis (median, 6.5 years) than the other cohorts. Patients in the CERV cohort were diagnosed most recently, with a median of only 5.7 months since diagnosis. Cervical cancer was the most common diagnosis (59.1%) followed by leukemia (13.8%) and uterine/endometrial cancer (7.2%). Most patients underwent hysterectomy (45.9%) or trachelectomy (23.8%). None of the 51 patients in the BMT cohort underwent surgery; however, >80% received both radiation and chemotherapy. The entire CERV cohort underwent surgery alone (protocol requirement). Approximately half of the patients in the GYN1 and GYN2 cohorts received either radiation, chemotherapy, or both in addition to undergoing surgery. The sample was predominantly non-Hispanic white (79%), well educated (70.7% had graduated college), and employed full time (63%). Almost two-thirds were married/cohabitating at time of FSFI completion.
Factorial Validity in Cancer Survivors
The exploratory factor analysis yielded a factor solution nearly identical to that reported in the original validation studies.6, 12 Five factors were retained in the final factor solution (Table 3) and accounted for 81.4% of the total variance in responses. The 5 factors corresponded well to the 6 FSFI domains, with the Desire and Arousal domains combining to make the first factor (the FSFI authors separated this factor into 2 domains for clinical reasons6). Each item loaded most strongly on the single factor associated with its domain. Notably, the pattern of factor loadings corresponded more closely with the pattern observed previously in women with sexual dysfunction compared with the solution observed in women with no dysfunction.12
Table 3. Rotated Factor Analysis Solution of the 19 Female Sexual Function Index Questions
|1. Desire: Frequency||0.85a||0.21||0.06||0.06||0.11|
|2. Desire: Level||0.86a||0.19||0.11||0.10||−0.01|
|3. Arousal: Frequency||0.73a||0.21||0.27||−0.01||0.25|
|4. Arousal: Level||0.79a||0.27||0.30a||−0.01||0.16|
|5. Arousal: Confidence||0.69a||0.28||0.34a||0.18||0.15|
|6. Arousal: Satisfaction||0.60a||0.28||0.53a||0.11||0.17|
|7. Lubrication: Frequency||0.36a||0.83a||0.14||0.15||0.09|
|8. Lubrication: Difficulty||0.30a||0.84a||0.14||0.22||0.05|
|9. Lubrication: Frequency of maintaining||0.21||0.83a||0.23||0.15||0.17|
|10. Lubrication: Difficulty in maintaining||0.19||0.85a||0.18||0.15||0.16|
|11. Orgasm: Frequency||0.19||0.20||0.87a||0.01||0.11|
|12. Orgasm: Difficulty||0.26||0.18||0.83a||0.16||0.14|
|13. Orgasm: Satisfaction||0.22||0.14||0.81a||0.16||0.22|
|14. Satisfaction with amount of closeness with partner||0.05||0.16||0.23||0.03||0.87a|
|15. Satisfaction with sexual relationship||0.35a||0.16||0.17||0.30a||0.78a|
|16. Satisfaction with overall sex life||0.52a||0.15||0.18||0.37a||0.57a|
|17. Pain: Frequency during vaginal penetration||0.17||0.16||0.07||0.88a||0.08|
|18. Pain: Frequency after vaginal penetration||−0.07||0.17||0.04||0.86a||0.15|
|19. Pain: Level during or after vaginal penetration||0.11||0.15||0.18||0.86a||0.08|
|% Total variance accounted for by each factor||22.95||17.95||15.53||14.26||10.74|
|% Total variance accounted for by 5-factor solution||81.43|| || || || |
Female Sexual Function Index Item Statistics
The percentages of responses scored 4 or 5 (ie, item severity or difficulty), indicating very high sexual functioning, are listed in Table 4 by item. High functioning was most difficult for respondents in the Desire domain, in which only 32% scored 4 or 5. Item 14, which assesses satisfaction with the amount of emotional closeness with their partner during sexual activity, had the largest proportion of women reporting high functioning (71.8%). The percentage of responses scored ≥4 to the remaining items ranged from 44% to 68%.
Table 4. Item Proportions and Corrected Item-Total Correlations for Female Sexual Function Index Total Scores and Domain Scores
|1. Desire: Frequencyb||32||0.61||0.82|
|2. Desire: Levelb||31.5||0.59||0.82|
|3. Arousal: Frequencyb||55.2||0.70||0.80|
|4. Arousal: Levelb||47||0.76||0.88|
|5. Arousal: Confidenceb||44.2||0.76||0.79|
|6. Arousal: Satisfactionb||55.2||0.79||0.81|
|7. Lubrication: Frequency||55.2||0.74||0.87|
|8. Lubrication: Difficulty||62.4||0.71||0.87|
|9. Lubrication: Frequency of maintaining||56.9||0.72||0.85|
|10. Lubrication: Difficulty in maintaining||66.9||0.69||0.84|
|11. Orgasm: Frequency||56.4||0.56||0.81|
|12. Orgasm: Difficulty||68||0.65||0.84|
|13. Orgasm: Satisfaction||61.9||0.61||0.76|
|14. Satisfaction with amount of closeness with partner||71.8||0.48||0.62|
|15. Satisfaction with sexual relationship||59.1||0.67||0.85|
|16. Satisfaction with overall sex life||49.7||0.74||0.70|
|17. Pain: Frequency during vaginal penetration||60.2||0.53||0.86|
|18. Pain: Frequency after vaginal penetration||67.4||0.44||0.85|
|19. Pain: Level during or after vaginal penetration||62.4||0.53||0.87|
Corrected item-total correlations with the FSFI total score all were relatively large, ranging from 0.44 to 0.79 (Table 4). The Pain items, as a group, had the weakest associations with the FSFI total score (mean correlation coefficient [r] = 0.50), whereas the Arousal items had the strongest (mean r = 0.75). Subscale-specific item-total correlations ranged from 0.62 to 0.88. Satisfaction items had the weakest association with their total (mean r = 0.73), whereas Lubrication and Pain items had the strongest (both sets of items had a mean subscale item-total correlation of 0.86).
Female Sexual Function Index Total and Domain Score Reliability and Descriptive Statistics
The Cronbach α reliability coefficients for each FSFI score were calculated first separately within the 4 study cohorts and then were analyzed for significant differences.43 No significant differences (P < .05) were identified across the cohorts. Table 5 presents the reliabilities of the scales for the cohorts combined. All FSFI scores had very high reliability. The Cronbach α value for the FSFI total score was .94 (95% confidence interval [CI], .92-.95) and, among the subscales, ranged from a low of .85 for Satisfaction (95% CI, .81-.89) to a high of .94 for Lubrication (95% CI, .92-.95).
Table 5. Internal Consistency Reliability and Score Summaries of Female Sexual Function Index Total Score and Subscales
|FSFI total score||181||.94 (.92-.95)||24.75 ± 6.75||25.90 [19.50-30.30]|
|FSFI domain score|
| Desire||181||.90 (.87-.93)||3.49 ± 1.28||3.60 [2.40-4.80]|
| Arousal||181||.92 (.90-.94)||4.06 ± 1.43||4.20 [3.00-5.10]|
| Lubrication||180||.94 (.92-.95)||4.26 ± 1.62||4.65 [3.30-5.70]|
| Orgasm||180||.90 (.87-.92)||4.31 ± 1.57||4.80 [3.60-5.60]|
| Satisfaction||179||.85 (.81-.89)||4.26 ± 1.47||4.80 [3.20-5.60]|
| Pain||181||.93 (.91-.95)||4.39 ± 1.79||4.80 [3.60-6.00]|
|Met the FSFI clinical dysfunction cutoff score of ≤26.0a||181|| || || |
| No.||94|| || || |
| %||51.93|| || || |
| 95% CI, %||44.59-59.28|| || || |
There were no significant differences across the cohorts in the distributions of the FSFI scores; therefore, scale descriptive statistics are presented in Table 5 for the cohorts combined. The mean (±standard deviation [SD]) FSFI total score was 24.75 ± 6.75—within the established diagnostic range of ≤26.0 that suggests clinically significant FSD. Similarly, over 51.9% of women (n = 94 met this FSD threshold. The median (25th to 75th percentile) FSFI score was 25.9 (19.5-30.3), and the scores ranged from a low of 5.5 to a high of 36. Scores were lowest on the Desire subscale (mean = SD, 3.49 ± 1.28) and highest on the Pain subscale (mean = SD, 4.39 ± 1.79).
Convergent Construct Validity
The construct validity of the FSFI scales generally was supported by the pattern of small-to-moderate correlations with the other instruments. The FSFI scores tended to be correlated negatively with depression (CES-D), distress (IES), menopausal symptoms (MSCL), and reproductive concerns (RCS), and generally were correlated positively with QOL (FACT-G, FACT-Cx), functional health status (the SF-12 PCS and MCS), and relationship satisfaction (ADAS) (Table 6).
Table 6. Convergent Construct Validity of the Female Sexual Function Index Full Score and Domain Scores With Other Patient-Reported Outcome Measures
| No. of score pairs in analysis||180||180||180||179||179||178||180|
| No. of score pairs in analysis||180||180||180||179||179||178||180|
| No. of score pairs in analysis||119||119||119||119||119||118||119|
| No. of score pairs in analysis||78||78||78||78||77||78||78|
| No. of score pairs in analysis||95||95||95||95||94||94||95|
| No. of score pairs in analysis||89||89||89||89||88||88||89|
| No. of score pairs in analysis||89||89||89||89||88||88||89|
| No. of score pairs in analysis||60||60||60||59||60||59||60|
| No. of score pairs in analysis||61||61||61||60||61||60||61|
|FACT Physical Well Being|
| No. of score pairs in analysis||60||60||60||59||60||59||60|
|FACT Social Well Being|
| No. of score pairs in analysis||60||60||60||59||60||59||60|
|FACT Emotional Well Being|
| No. of score pairs in analysis||61||61||61||60||61||60||61|
|FACT Functional Well Being|
| No. of score pairs in analysis||61||61||61||60||61||60||61|
The Satisfaction subscale correlated significantly with the largest number of scales and had the only significant correlation with the ADAS, consistent with it being considered a measure of global sexual and relationship satisfaction, and the QOL domain on the FSFI.6 The SF-12 PCS was correlated significantly with Orgasm and Pain, whereas the MCS was correlated significantly with Arousal and Satisfaction.
Exploratory Investigation of the Discriminant Validity of the Female Sexual Function Index
The FSFI classified 59.5% of survivors who received chemotherapy and/or radiation as having clinically significant FSD compared with 45.4% of survivors who received neither treatment (P = .073) (Table 7). Survivors who received neither chemotherapy nor radiation had significantly better Lubrication (P = .004) and Pain (P = .046) scores and had marginally better FSFI total scores (P = .058) than survivors who received chemotherapy, radiation, or both.
Table 7. Ability of Female Sexual Function Index Scales to Discriminate Between Chemotherapy and Radiation Treatment Groups
|Chemotherapy and/or RT|
| Neither, n = 97b||25.64 ± 6.63||3.59 ± 1.23||4.15 ± 1.45||4.58 ± 1.60||4.43 ± 1.54||4.28 ± 1.43||4.64 ± 1.58||45.4|| |
| Either/both, n = 84b||23.73 ± 6.78||3.37 ± 1.33||3.95 ± 1.39||3.89 ± 1.57||4.18 ± 1.60||4.25 ± 1.53||4.10 ± 1.97||59.5||.073|
| t(df)||t(179) = 1.91||t(179) = 1.13||t(179) = 0.96||t(178) = 2.92||t(178) = 1.10||t(177) = 0.14||t(179) = 2.01|| || |
| P||.058||.45||.34||.004||.27||.89||.046|| || |
|Chemotherapy, RT, both, or neither|
| Neither, n = 97b||25.64 ± 6.63||3.59 ± 1.23||4.15 ± 1.45||4.58 ± 1.60c||4.43 ± 1.54||4.28 ± 1.43||4.64 ± 1.58||45.4|| |
| CT only, n = 22||22.47 ± 6.86||3.46 ± 1.25||3.91 ± 1.47||3.45 ± 1.55c||3.82 ± 1.58||4.21 ± 1.60||3.62 ± 1.94||72.7|| |
| RT only, n = 7||22.11 ± 7.15||3.43 ± 1.79||3.73 ± 1.52||3.47 ± 1.37||3.77 ± 1.20||4.00 ± 1.58||3.71 ± 1.69||71.4||.087|
| Both, n = 55d||24.44 ± 6.73||3.33 ± 1.33||3.99 ± 1.37||4.12 ± 1.58||4.37 ± 1.63||4.30 ± 1.52||4.35 ± 2.00||52.7|| |
| F(df numerator,df denominator)||F(3,177) = 1.81||F(3,177) = 0.49||F(3,177) = 0.38||F(3,176) = 3.98||F(3,176) = 1.23||F(3,175) = 0.10||F(3,177) = 2.38|| || |
| P||.15||.69||.77||.009||.30||.96||.07|| || |
The FSFI classified 73% and 71% of women who received chemotherapy only and radiation therapy only, respectively, as having FSD compared with only 45% of those who received neither and 53% of those who received both (P = .087). There also were significant differences in Lubrication scores (P = .009) and marginally significant differences in Pain scores (P = .07) across these 4 mutually exclusive treatment groups. Specifically, patients who received neither therapy had the best scores, particularly compared with those who received either chemotherapy only or radiation only. Although scores were expected to be lowest among patients who received both chemotherapy and radiation, 75% of survivors who received both treatments were from the BMT cohort, the only group that did not have malignancies of the reproductive and/or urologic organs and the group for which the most time had passed since their diagnoses.
The objective of the current study was to perform the first empirical validation of the FSFI in female cancer survivors. Our results indicate that the FSFI is a reliable and valid instrument for measuring levels of sexual functioning and cancer-related FSD in this population. The psychometric properties obtained in this study are very similar to those reported in the initial FSFI validation. Specifically, the original FSFI factor structure, on which the domain subscales are based, was replicated in this group of cancer survivors, as were the reliability coefficients of the total and domain scores. The FSFI demonstrated convergent validity by correlating strongly with validated measures of menopausal symptoms, reproductive concerns, QOL, and depression. Finally, the FSFI exhibited preliminary evidence of discriminative validity, in that marginally significantly (P < .10) higher proportions of survivors who had received chemotherapy and/or radiation were classified as having FSD by the FSFI diagnostic cutoff compared with survivors who did not receive those treatments. In addition, the chemotherapy/radiation-treated patients tended to have lower mean FSFI scores than the untreated patients, particularly in the Lubrication and Pain domains.
The sound psychometric properties of the FSFI among cancer survivors were clearly evident in this study. More subtly, however, the results suggest that emotional intimacy with a partner may play an important role in the sexual functioning of cancer survivors that is conceptually distinct from overall satisfaction with sexual life. Although only 32% of survivors endorsed strong sexual desire (Table 4), and only 50% were very satisfied overall with their sexual life, 72% reported being very satisfied with their emotional closeness with partners during sexual activity. In addition, this “emotional closeness” item had the lowest and second lowest corrected item-total correlations in the domain-specific analysis and the FSFI total score, respectively. Although these properties are within acceptable ranges for a psychometric scale, the findings seem to suggest that this item is measuring a facet of satisfaction that may be particularly salient among cancer survivors and is somewhat distinct from overall satisfaction. This is consistent with a recent qualitative study in which, although survivors reported mainly negative effects of cancer on sexual functioning, many indicated improvements in intimacy after their cancer diagnosis, highlighting the complex relations among overall satisfaction with sex life, intimacy, and sexual functioning among cancer survivors.44
Although these results support the continued use of the FSFI among cancer survivors, clinicians and researchers should be acutely aware of the threat to the validity of the instrument's scores when administered to women with little or no recent sexual activity. In general, the FSFI scores are not valid measures of sexual functioning for sexually inactive women, because of 15 of the 19 items that have response options of “no sexual activity” or “did not attempt intercourse” are assigned a score of zero by the FSFI scoring algorithm. From a psychometric viewpoint, treating these responses as the lowest possible level of functioning on the ordinal response scale is justifiable only when it is explicitly known that a given respondent avoided sexual activity specifically because of extreme dysfunctions in the symptoms assessed by these items. Even in these instances, however, it is possible that women who attribute their avoidance of sexual activity to sexual dysfunction represent a qualitatively distinct clinical subgroup in need of more in-depth sexual rehabilitation and should not have their sexual functioning measured on the same metric as survivors who are not avoidant of sexual activity. More clearly, the FSFI is not a valid measure of sexual functioning for female survivors who are sexually inactive because of lack of a partner, poor relationship quality, or other reasons unrelated to changes secondary to cancer treatment (eg, estrogen deprivation).
It is likely that many studies reporting FSFI scores, particularly those focused on cancer populations, have inflated prevalence estimates of FSD because of this eccentricity in the scoring of the instrument. For example, if we had not excluded women with ≥8 zero-scored responses, then our FSFI-based estimate of FSD would have been 60%, instead of 52%. Clinical experience with cancer survivors, particularly those still in treatment, suggests that women often indicate very low sexual activity levels in the months after cancer diagnosis and treatment not because of dysfunctions in the domains assessed by the FSFI but because they are mentally and physically focused on the impact of the disease in their lives and on their recovery, and they often are preoccupied with concerns about their cancer recurring. Studies that do not omit sexually inactive women from analysis of the FSFI also will tend to have downward biased mean and median FSFI scores as well as inflated score standard deviations. In an illuminating critique of the FSFI, Meyer-Bahlburg & Dolezal41 note the effect these biases can have on group comparisons. For example, a study comparing women treated for vulvar intraepithelial neoplasia (VIN) with matched controls revealed that the women with VIN had significantly lower FSFI total scores and had lower scores on 4 of the 6 domains.13 However, women in the control group were required to be sexually active, but the women with VIN were not. Indeed, greater than 25% of the women with VIN reported no sexual activity in the 4 weeks before completing the FSFI.
Although the VIN study appears to have been biased toward identifying significant differences in FSFI scores between groups, the greater score variability and inflated SDs that result from including sexually inactive women in the analysis of group comparisons also can obscure meaningful differences that otherwise would have been statistically significant.41, 42 For example, a study comparing surgical technique for resection of rectal cancer revealed no differences in female sexual function as measured by the FSFI despite significant differences in patients' ratings of how the surgery affected their sexual functioning overall.25 This seemingly contradictory failure to identify significant FSFI differences is not surprising considering that >50% of the women reported no sexual activity.
It seems reasonable to expect that women who answer “no sexual activity” to any 1 of the 15 FSFI items with this option would indicate the same response to the other 14. In practice, however, this does not appear to be the case. In the current study, we considered women with nonzero responses to at least 8 of these 15 items (ie, more than half) to be sufficiently sexually active for the FSFI scores to be valid indicators of their sexual functioning. This threshold for sexual activity was chosen by rational means only, and future studies should attempt to evaluate this important validity issue in a more empirical manner. In particular, we strongly recommend that future studies using the FSFI in cancer survivors pay specific attention to sexual activity levels and respondents' reasons for sexual inactivity. The lack of such an assessment in the studies on which the current analysis is based was an unfortunate limitation of the design of those studies.
The current study has 2 important strengths related to the characteristics of the samples. First, 75% of the sample was ≤41 years old, an age range in which sexual functioning and reproductive issues are important components of QOL. Second, the women in this study were diverse in terms of cancer diagnosis, treatment, age at diagnosis, and time since diagnosis, all of which can vary with respect to the severity and type of sexual symptoms caused. In addition to conferring greater generalizability to the results, this kind of diversity is advantageous to psychometric studies, because it ensures that a wide range of the construct being measured is represented in the responses.
Despite this diversity, a limitation of the current study is the lack of data from breast cancer survivors. Although, ultimately, we believe that these results are generalizable to breast cancer survivors, future research is needed to establish this empirically. We also want to note the limits to the generalizability of the discriminant validity analysis. It is important to emphasize that the 4 cohorts of survivors combined for this report came from studies with different eligibility criteria. Because of this, many important characteristics, such as cancer type, treatment, and time since cancer diagnosis, are not random across this combined sample but, instead, are inextricably confounded with the study cohorts. Therefore, the analysis of discriminant validity should be considered exploratory, providing only provisional evidence that the FSFI can accurately discriminate between groups of cancer survivors with different treatment histories. Although a previous study indicated that cervical cancer survivors who received radiotherapy alone had significantly worse FSFI scores compared with those who underwent surgery alone and healthy controls,20 it is not clear how the inclusion of sexually inactive women may have influenced those results. More research is needed with sexually active female cancer survivors to determine the sensitivity of the FSFI to treatment differences.
A final aspect of validity not addressed in the current study is the sensitivity of the FSFI scores to changes over time. The longitudinal study that provided the CERV cohort data, however, indicated a significant increase in the mean FSFI total score from 16.8 at the presurgical baseline to 23.8 at 12-months postsurgery.19 Another study indicated significant FSFI improvements in patients with cervical cancer who received radiotherapy from a mean score of 17.0 at baseline to 29.4 after 3 months of using a clitoral therapy device.21 Although these studies suggest that the FSFI is sensitive to changes over time, this is another area that is ripe for future investigations.
In summary, responses given by women in the current study support the continued use of the FSFI for monitoring sexual function and cancer-related dysfunction of female cancer survivors. Researchers and clinicians should be aware that the FSFI may not provide valid assessments of sexual function among women who have not had recent sexual activity. Researchers must take this into account when designing and reporting studies using the FSFI to avoid reporting artificially low FSFI scores and estimates of FSD prevalence as well as potentially biased group comparisons. With the validity of the FSFI established among cancer survivors, in a future study, we will develop a short form of the instrument to make routine screening for sexual dysfunction more realistic and feasible in busy oncology settings.
This study was funded by a grant from the Lance Armstrong Foundation and by Philanthropic funds, Gynecology Service, Memorial Sloan-Kettering Cancer Center.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.