Breast sensorisexual function: a novel patient-reported outcome measure of sexual sensory functions of the breast

Abstract

Background

A validated measure assessing sexual sensory functions of the breast is needed to optimize sexual and other health outcomes after breast procedures.

Aim

To describe the development of a patient-reported outcome measure (PROM) to assess breast sensorisexual function (BSF).

Methods

We applied the PROMIS standards (Patient Reported Outcomes Measurement Information System) for measure development and evaluation of validity. An initial conceptual model of BSF was developed with patients and experts. A literature review yielded a pool of 117 candidate items that underwent cognitive testing and iteration. Forty-eight items were administered to an ethnically diverse, national panel–based sample of sexually active women with breast cancer (n = 350) or without (n = 300). Psychometric analyses were performed.

Outcomes

The main outcome was BSF, a measure that assesses affective (satisfaction, pleasure, importance, pain, discomfort) and functional (touch, pressure, thermoreception, nipple erection) sensorisexual domains.

Results

A bifactor model fit to 6 domains—excluding 2 domains with only 2 items each and 2 pain-related domains—revealed a single general factor representing BSF that may be adequately measured by the average of the items. This factor, with higher values denoting better function and with the standard deviation set to 1, was highest among women without breast cancer (mean, 0.24), intermediate among women with breast cancer but not bilateral mastectomy and reconstruction (−0.01), and lowest among those with bilateral mastectomy and reconstruction (−0.56). Between women with and without breast cancer, the BSF general factor accounted for 40%, 49%, and 100% of the difference in arousal, ability to orgasm, and sexual satisfaction, respectively. Items in each of 8 domains demonstrated unidimensionality (ie, they measured 1 underlying BSF trait) and high Cronbach’s alphas for the entire sample (0.77-0.93) and the cancer group (0.71-0.95). Correlations with sexual function, health, and quality of life were positive for the BSF general factor and mostly negative for the pain domains.

Clinical Implications

The BSF PROM can be used to assess the impact of breast surgery or other procedures on the sexual sensory functions of the breast in women with and without breast cancer.

Strengths and Limitations

The BSF PROM was developed by using evidence-based standards, and it applies to sexually active women with and without breast cancer. Generalizability to sexually inactive women and other women warrants further study.

Conclusion

The BSF PROM is a measure of women’s breast sensorisexual function with evidence of validity among women affected and unaffected by breast cancer.

Introduction

Of >3.8 million breast cancer survivors in the United States today,¹ about one-third have undergone mastectomy.^2,3 Of these, 530 000 survivors have had or are considering breast reconstruction procedures.^2,4 Women who undergo mastectomy commonly experience loss of breast sensation: nonsexual (eg, pressure sensation with a hug or ability to sense heat) and sexual (eg, nipple erection).⁵ Despite loss of breast sensation, chronic postmastectomy breast pain is also common, affecting 25% to 60% of patients.^6,7 Loss of breast sensation and pain are known to interfere with sexual function,^8–10 a domain of physical function valued by women across the life course.^11,12 However, the best available treatments to preserve breast sensation following mastectomy, including skin- and nipple-sparing autologous flaps and nerve-grafting techniques, have targeted mainly nonsexual sensory outcomes,^5,13 and only a subset of patients is eligible for these procedures.¹⁴ Furthermore, sensory function outcomes of these interventions have been mixed,^15,16 and few studies have examined sexual function outcomes.^15,17,18

The sexual physiology of the breast includes erectile or tumescent responsiveness to a variety of tactile stimuli.¹⁹ During sexual activity, arousal modulates the experience of breast and nipple sensation, and pleasurable breast and nipple sensation typically promotes clitoral sensitivity, vaginal lubrication, and orgasm.^20-22 In addition, breast and nipple responsiveness to sexual stimulation can promote sexual response and satisfaction of the partner.²³ Lack of attention to preservation of breast sexual sensation is reflected in the ways that breast sensation has been measured, including lack of assessment of the known sexual functions of the female breast.²⁴ In addition, female sexual function is often categorized in dominant measurement regimes as a quality-of-life outcome,²⁵ rather than a physical function or health outcome.

A systematic review of 124 patient-reported outcome measures (PROMs) developed to assess breast sensation outcomes following mastectomy and reconstruction found only 4 validated measures.²⁴ In that review, Smeele and colleagues provided a useful working definition for the concept of breast sensation, composed of 4 sensory components:

(1) ability to feel sensory stimuli (eg, touch, pressure, vibration, temperature, and pain) in the breast, the nipple, or both, (2) ability to experience affective aspects of touch (eg, pleasant, sexual) in the breast, the nipple, or both, (3) ability of the nipple to become erect upon a sensory stimulus (eg, cold, touch) and (4) symptoms indicating end-organ sensory loss (eg, hypoesthesia, numbness, and anesthesia) in the breast, the nipple, or both.^24(p436)

Additionally, the definition provided by Smeele and colleagues distinguishes sensations felt on the breast from those felt on the nipple. These anatomic distinctions are salient to the development and assessment of insterventions to preserve and restore nonsexual and sexual breast sensory functions.

Development and testing of interventions designed to preserve or recover breast function require outcome measures that can comprehensively assess both the nonsexual and sexual functions of the breast. The National Cancer Institute’s Patient-Reported Outcomes Measurement Information System (PROMIS) provides standards for developing PROMs appropriate for the general population and people with illnesses, including cancer.^26,27 PROMs validated for use among healthy control samples are needed to establish meaningful clinical differences in breast reconstruction outcomes.²⁸ In this study, we apply the PROMIS standards²⁷ to develop a PROM to assess the sexual functions of the breast—specifically, breast sensorisexual function (BSF)—for use in studies of women with and without cancer.

Methods

We developed the BSF PROM in 3 phases. University of Chicago’s Institutional Review Board approved this study, and all participants provided written documentation of informed consent.

Phase 1: defining BSF and development of conceptual model

Using an iterative process, we combined expert clinician observations with data obtained from a registry of female patients with cancer and sexual function concerns to develop an initial conceptual model of BSF.²⁹ We then engaged a patient advisory board and conducted a focus group³⁰ with 3 clinicians with combined expertise in breast oncology and sexual function to refine the model. Then, we conducted a literature review to identify existing descriptions and models of breast function, as well as measures relevant to assessing BSF in women with and without cancer.^{9,13,21,31–42} Several nonsexual and nonsensory breast functions were identified (eg, signaling femininity, supporting clothing, lactation), but we focused on sexual sensory functions of the breast. BSF was defined to include the following conceptual areas: the ability to feel touch, temperature, and pressure; the ability of the nipple to become erect; and sensation associated with movement or mobility of the breast (Figure 1).

Figure 1.

Conceptual model of breast sensory function. Gray box with dashed outline represents conceptual areas of focus for breast sensorisexual function (BSF) measure development.

Phase 2: candidate item development and cognitive testing

We categorized items from existing measures by BSF conceptual area and removed redundant items (“binning” and “winnowing”; Supplementary Data A, Figure A1)^26,43; then, we drafted new items to fill conceptual gaps in the extant literature for assessment. For each conceptual area, we included items to assess affective aspects of breast sensation, such as feelings of importance, satisfaction, pleasure, and pain related to breast or nipple sensory function (individually), as well as intensity of sensations (eg, how much) and frequency (eg, how often). An overall 117 candidate items were advanced for further testing. Cognitive testing⁴⁴ was conducted with women affected and unaffected by breast cancer by using the cognitive model of survey response (Supplementary Data B). Data were analyzed via cognitive coding of text summaries.⁴⁵ Because the objective of cognitive testing was reparative, data saturation was achieved and cognitive interviews were concluded when we determined that any serious problems with the item bank had been identified and resolved.⁴⁵ Items from all conceptual areas, except mobility due to its elimination following cognitive testing, were advanced for psychometric evaluation.

Phase 3: psychometric evaluation

Forty-eight BSF items (Supplementary Data C, Table C1) were administered via a web-based survey to a sample of respondents affected by breast cancer (n = 350) or unaffected (n = 300).

Study sample

Respondents were recruited from Opinions 4 Good,⁴⁶ a national internet-based survey research panel used for prior women’s health and other studies.^47–51 Quota sampling by age, race/ethnicity, educational attainment, and breast surgery history was used to ensure a uniform distribution of respondents across key subgroups. Additionally, a targeted sample of respondents affected by breast cancer were recruited through the Program in Integrative Sexual Medicine (PRISM)²⁹ research registry, from patient advisors, and via social media and other social network channels. Individuals were eligible if they were ≥18 years old, English speaking, assigned female sex at birth, sexually active within the last 30 days, not currently pregnant, and had not breast-fed a child within the last 30 days.

Measures

BSF items are listed in Supplementary Data C, Table C1. Each item used an ordinal response scale ranging from “none/not at all/never” to “a lot/very/always.” Items assessed the breast (or chest wall for respondents whose breast was removed and not reconstructed) and nipple separately; only respondents with 1 or both nipples (93.5% overall) were eligible to answer nipple-related questions. Because BSF items assessed breast or nipple sensory function in the context of sexual activity, a response choice, “My breasts (or nipples) were not a part of my sexual activity,” was provided (these responses were excluded from the analyses). In addition to the BSF items, the survey instrument assessed sociodemographic,^52,53 health,^52,54–56 and psychosocial characteristics.⁵⁷ Self-reported health was assessed with a single item from the Medical Outcomes Study (1 = poor, 5 = excellent).⁵⁵ Sexual function was assessed with the PROMIS SexFS v2.0 Brief Profile (Female) (mean = 50, SD = 10).^54,58 Quality of life was assessed with the Neuro-QoL Item Bank v1.0 Positive Affect and Well-being–Short Form,⁵⁷ and higher scores indicated better quality of life.

Our patient advisory board recommended that we prompt women to report on each breast separately if their breasts look, feel, or function differently from each other (n = 106); this ensured inclusivity and applicability of the measure to most women. Additionally, respondents who had only 1 breast (n = 42) reported separately for the breast and contralateral chest wall. In both cases, just those responses pertaining to the first or only breast described were analyzed. While assessing both breasts separately is important for real-world application of the BSF PROM, the limited size of the subgroup reporting separately about each breast precluded incorporating both breasts appropriately in the analyses presented here.

Survey data collection

Data collection took place in 2 waves: wave 1 recruited 550 participants (February-March 2020; 250 affected and 300 unaffected) and wave 2 recruited 100 participants with bilateral mastectomy and reconstruction (June 2020). Of the latter group, 89 participants were recruited via Opinions 4 Good and 11 from the PRISM registry or other internal recruitment channels. Individuals were invited by email to complete an online survey (median time, 15 minutes; IQR, 12–21 minutes). A follow-up survey to assess test-retest reliability was offered to all 650 respondents. The follow-up survey was administered 2 weeks after the end of data collection for each wave (range, 10 days–2 months) and completed by 44% of participants (n = 286). Older participants and those with poorer quality of life were more likely to complete the follow-up survey. A small incentive was paid following completion of each survey.

Sample size justification

Our originally proposed sample size of 500 met or exceeded common recommendations for factor analyses of ≤50 items.⁵⁹ In fact, as shown by MacCallum et al,⁵⁹ with a sample size of 500 and the residual degrees of freedom available here (at least 100 for all models), we had nearly 100% power to detect a lack of fit even against closely fitting alternatives. For this reason, we report 90% CIs for the root mean square error of approximation (RMSEA) to indicate the precision with which we are able to evaluate the lack of fit. In addition, we note that our original sample size was at the maximum of the range (150–500) suggested by Bader et al⁶⁰ to achieve acceptable parameter and SE estimates when fitting a bifactor model, assuming general factor loadings of 0.5 and explained common variance of 0.5 (both of which are less than those for our final model). Sample size was later increased to 650 to provide adequate power and precision for subgroup analyses. For establishing construct validity, we had 80% power (2-sided alpha = 0.05) to detect a correlation of 0.26 between BSF and sexual function among women with breast cancer who underwent bilateral mastectomy and reconstruction (n = 114) and 0.16 among unaffected women.

Statistical analysis

Responses for each item were coded with integers from 1 to 5, with higher values indicating a greater amount of the construct being measured. Pairwise polychoric correlations were computed among the items, and exploratory factor analyses were performed with the principal factor method. Orthogonal varimax and oblique oblimin rotations were examined. Consistent with our hypothesis of a superordinate construct corresponding to overall BSF with several specific domains, we also performed a bifactor rotation in which each item is assumed to depend on a single general factor, as well as 1 other domain-specific factor, and all factors are assumed uncorrelated.^61–64 Informed by these exploratory analyses, we then fit and compared 4 confirmatory factor models: (1) a single-factor model to establish a baseline; (2) a model in which each item loads on 1 of 6 factors, with all factors assumed correlated; (3) a second-order model in which each of the individual first-order factors loads on a single second-level factor with no residual correlation among the first order factors; and (4) the bifactor model (Supplementary Data C, Figure C1). Three fit indices are reported: the RMSEA, the standardized root mean squared residual, and the comparative fit index (CFI). In addition, we performed likelihood ratio tests comparing nested models. Estimates of the general factor for each respondent were obtained from the bifactor model by computing the posterior means (ie, empirical Bayes).

Ten nonoverlapping subsets of items, each corresponding to a conceptually distinct domain, were identified by our conceptual model and the exploratory analyses just described. Each was evaluated separately to provide information for those who may wish to use these independently—including monotonicity (ie, for each item, whether a greater response is associated with a higher mean score for the other items in that domain) and unidimensionality (ie, whether a single underlying factor accounts for the observed interitem correlation)—as determined by fitting a single-factor model to the items in each domain according to maximum likelihood.²⁷ Additionally, Cronbach’s alpha was used to estimate the reliability of each subscale. Test-retest reliability over a 2-week period was assessed for each domain with intraclass correlation coefficients (ICCs), based on a 2-way mixed effects model measuring consistency of agreement. A coefficient of 0.40 to 0.59 indicates fair reliability, 0.60 to 0.74 good reliability, and ≥0.75 excellent reliability.⁶⁵

Construct validity of the general factor from the bifactor model and each subscale was assessed by comparing 3 subgroups based on cancer status and level of treatment (ie, known-group validity): respondents without breast cancer, respondents with breast cancer but without bilateral mastectomy and reconstruction (BMR–), and respondents with breast cancer who underwent bilateral mastectomy and reconstruction (BMR+). We hypothesized that affected respondents would have decreased BSF when compared with unaffected respondents and that affected respondents who had undergone more invasive treatment would have worse BSF than those who had undergone less invasive treatment. These comparisons were performed through ordinary least squares regression, adjusting for age and self-reported health. A joint test comparing all 3 groups (2 df) was performed; only if this test was statistically significant were pairwise comparisons performed. Additionally, Pearson correlations were calculated between the general factor and subscale scores, on one hand, and scores from the PROMIS SexFS (sexual satisfaction, orgasm ability, and lubrication as an indicator of arousal), health, and quality of life, on the other. Because of the known associations between sexual function and both health and quality of life,^12,66 we hypothesized that BSF would also be associated with these constructs, albeit to a lesser degree. Finally, we regressed each of the 3 sexual function measures (satisfaction, orgasm ability, and lubrication) on cancer status (affected vs unaffected)—by itself and then by adjusting for the BSF general factor—to determine approximately how much of the difference in sexual function between groups might be accounted for by differences in BSF.

All statistical analyses were performed with Stata version 16.1 (StataCorp LLC). Polychoric correlations were calculated with the polychoric package,⁶⁷ and confirmatory factor models were fit with the sem command.

Results

Sample characteristics

Table 1 summarizes the sociodemographic and health characteristics of survey respondents, stratified by cancer status. Respondents with breast cancer reported significantly lower sexual satisfaction, reduced ability to orgasm, and less lubrication as well as significantly poorer health and quality of life. More than a third of respondents (35%) had undergone bilateral mastectomy with reconstruction of 1 or both breasts. Nearly 1 in 5 (18%) of all respondents with 2 breasts reported that their 2 breasts looked, felt, or functioned differently from each other. Forty-two respondents (6%) had only 1 breast.

Table 1.

Sociodemographic and health characteristics of survey respondents, stratified by breast cancer status.^a

	Patients, mean ± SD or No. (%)
Survey respondent characteristics	Overall (N = 650)	Unaffected (n = 300)	Affected (n = 350)
Age, y	46.9 ± 13.7	47.9 ± 14.1	46.1 ± 13.3
Race and ethnicity
Non-Hispanic White	204 (31.5)	101 (33.9)	103 (29.4)
Non-Hispanic Black	118 (18.2)	73 (24.5)	45 (12.9)
Hispanic	239 (36.9)	78 (26.2)	161 (46.0)
Asian	46 (7.1)	26 (8.7)	20 (5.7)
Other	41 (6.3)	20 (6.7)	21 (6.0)
Education
<High school	184 (28.3)	78 (26.1)	106 (30.3)
High school/some college	194 (29.9)	96 (32.1)	98 (28.0)
Bachelor degree or higher	271 (41.8)	125 (41.8)	146 (41.7)
Partnered	527 (83.9)	225 (78.1)	302 (88.8)
Sexual function scores
Satisfaction	46.6 ± 7.7	47.5 ± 8.3	45.8 ± 7.1
Lubrication (arousal)	44.5 ± 8.3	47.0 ± 8.3	42.3 ± 7.6
Ability to orgasm	40.7 ± 10.8	43.6 ± 11.0	38.3 ± 10.1
Self-reported health
Excellent/very good	222 (34.2)	129 (43.1)	93 (26.6)
Good	290 (44.7)	127 (42.5)	163 (46.6)
Fair/poor	137 (21.1)	43 (14.4)	94 (26.9)
Quality of life	50.7 ± 7.9	52.4 ± 7.9	49.2 ± 7.5
Breast cancer treatment history
Chemotherapy			177 (55.3)
Radiation			146 (45.8)
Use of tamoxifen			90 (30.0)
Use of aromatase inhibitor			89 (29.1)
Use of raloxifene			38 (12.7)
Use of leuprolide acetate			30 (10.5)
Lumpectomy	79 (12.9)	4 (1.4)	75 (23.3)
Lymph node removal	64 (10.5)	2 (0.7)	62 (19.3)
Mastectomy	139 (22.8)	1 (0.4)	138 (42.9)
Bilateral mastectomy with reconstruction	114 (18.7)		114 (35.4)

^aData presented here exclude don’t know, refused, and missing responses from analyses. Blank cells indicate not applicable.

Psychometric evaluation

Factor analysis

BSF item missingness was low (<1% overall). Applying the principal factor method to 46 items (2 items with different response options were excluded) yielded 2 dominant eigenvalues. Following a varimax rotation, the first factor explained 38% of the variance with the loadings >0.3 for all nonpain items (most loadings >0.5); the second explained 21% of the variance with loadings >0.3 for the pain items only (all but 2 loadings were >0.5); and the third factor explained only an additional 7% of the variance. Given these findings, we dropped 13 pain-related items and repeated the analysis, yielding a single clearly dominant eigenvalue. A bifactor rotation yielded specific factors corresponding closely to the predetermined conceptual areas. We then removed an additional 5 items loading weakly on the general factor, leaving 2 domains with 2 items each (general importance of the breast during sexual activity and nipple erection), which we also removed. A bifactor model with 6 specific factors fit to the remaining 24 items yielded a reasonable fit (RMSEA = 0.079 [90% CI, 0.074–0.083], CFI = 0.92).⁶⁸ This represented a substantial improvement in fit as compared with a single-factor model fit to the same items (RMSEA = 0.134 [90% CI, 0.130–0.139], CFI = 0.74). While the fit of the second-order model was similar to that of the bifactor model (Supplementary Data C, Table C2), a likelihood ratio test comparing the 2 was significant (135.7 on 18 df, P < .001) in support of the latter. A likelihood ratio test comparing the second-order model with the first-order correlated factors model was significant (135.2 on 9 df, P < .001), though the fit of both was also similar. The 6 specific factors corresponded to (1) general satisfaction/pleasure derived from the breasts/nipples during sexual activity, (2) ability to feel touch on the breasts/nipples, (3) importance of one’s ability to feel touch on the breasts/nipples, (4) touch-related pleasure on the breasts/nipples, (5) ability to feel temperatures on the breasts/nipples, and (6) general pressure (sexual and nonsexual) felt on the breasts—with the last of these loading least strongly on the general factor (Supplementary Data C, Table C3).

Scale properties computed separately for each of the 10 domains are provided in Table 2. Eight of these domains assessing sensation during sexual activity demonstrated reasonable fit for the single-factor model when fit to the entire sample: (1) general satisfaction/pleasure derived from the breasts/nipples during sexual activity, (2) general pain/discomfort in the breasts, (3) ability to feel touch on the breasts/nipples, (4) importance of one’s ability to feel touch on the breasts/nipples, (5) touch-related pleasure on the breasts/nipples, (6) ability to feel temperatures on the breasts/nipples, (7) pain in the breasts/nipples in response to stimuli, and (8) general pressure (sexual and nonsexual) felt on the breasts. The CFI among unaffected respondents occasionally fell below the PROMIS-suggested cutoff of 0.95 (though the standardized root mean squared residual met the suggested cutoff in all cases except 1). Cronbach’s alpha for the entire sample exceeded 0.90 in 5 domains and 0.85 in another 2: pain in response to stimuli on the breast/nipple during sexual activity and the ability to feel temperature on the breast/nipple during sexual activity. Reliability was lowest for the general pressure domain (0.77). Reliability among affected respondents met or exceeded 0.85 in all but 2 domains: pain in response to stimuli and general pressure during sexual activity or with contact like a hug.

Table 2.

Domains of breast sensorisexual function and fit indices, with PROMIS-recommended fit benchmarks.

				PROMIS-recommended fit benchmarks
Domain and conceptual definition	Items included a	No. of observations	No. of items	CFI	SRMR	Cronbach alpha
				>0.95	<0.08	>0.90 (individual) >0.70 (group)
General satisfaction/pleasure assesses how satisfied a person is with their breast or nipple sensation and how pleasurable physical contact with the breast or nipple has felt to them during sexual activity.	SB101r, SN101r, PIB101, PIN101	535	4	0.96	0.021	0.93
Unaffected only		259	4	0.83	0.027	0.95
Affected only		276	4	0.99	0.016	0.89
General pain/discomfort is the degree and frequency of physical discomfort, including pain, of the breast experienced with sexual activity.	DB201, DB202, PaB101, PaB102	592	4	0.99	0.011	0.92
Unaffected only		265	4	0.98	0.014	0.92
Affected only		327	4	>0.99	0.011	0.89
Ability to feel touch assesses the degree to which a person is able to feel various intensities of touch on the breast or nipple during sexual activity.	TN103r, TN104r, TB101r, TB102r	552	4	0.98	0.021	0.91
Unaffected only		264	4	0.93	0.029	0.92
Affected only		288	4	>0.99	0.014	0.87
Importance of touch is the importance of a person’s ability to feel light or firm touch on the breast or nipple during sexual activity.	TB205, TB206, TN205, TN206	554	4	0.95	0.027	0.91
Unaffected only		267	4	0.87	0.036	0.92
Affected only		287	4	0.99	0.019	0.88
Touch-related pleasure assesses how pleasurable various intensities of touch on the breast or nipple have felt during sexual activity.	TN203, TN204, TB203, TB204	556	4	0.97	0.021	0.91
Unaffected only		267	4	0.92	0.028	0.93
Affected only		289	4	0.99	0.016	0.87
Ability to feel temperatures assesses the degree to which a person is able to feel various temperatures on the breast or nipple during sexual activity.	ThN101r, ThN102r, ThB101r, ThB102r	550	4	0.90	0.043	0.89
Unaffected only		264	4	0.85	0.045	0.91
Affected only		286	4	0.94	0.041	0.85
Pain in response to stimuli is the degree of physical discomfort, including pain, in the breast or nipple experienced with physical contact during sexual activity.	TB207, TB208, TN207, TN208, PB202	556	5	0.96	0.031	0.86
Unaffected only		266	5	0.81	0.081	0.85
Affected only		290	5	>0.99	0.015	0.84
General pressure is the degree to which a person is able to feel pressure in the breast with physical contact (sexual and nonsexual), how satisfied they are with their ability to feel pressure, and how pleasurable pressure in the breast has felt to them.	PB101ar, PB103ar, PB201, PB204,	598	4	0.96	0.042	0.77
Unaffected only		267	4	0.96	0.044	0.81
Affected only		331	4	0.95	0.043	0.71
General importanceb is the importance of the role that the breast or nipple plays during sexual activity.	IB101, IN101		2
Nipple erectionb is the ease and frequency with which a person’s nipples become erect during sexual activity and how pleasurable their nipple erection has felt.	NE101r, NE301		2

Abbreviations: CFI, comparative fit index; PROMIS, Patient-Reported Outcomes Measurement Information System; SRMR, standardized root mean squared residual.

^aSee Supplementary Data C (Table C1) for all items tested for inclusion in the breast sensorisexual function patient-reported outcome measure, including response distributions and missingness.

^bBecause there were only 2 items in these domains, these domains and their items were not subjected to further psychometric analyses following factor analyses.

Scoring the BSF PROM

The correlation between the BSF general factor, as obtained from the bifactor model, and the score generated by taking the average of the 24 items included in the bifactor model was 0.97, indicating that the simple average may be used to provide a BSF score instead of fitting the bifactor model. Cronbach’s alpha among these 24 items was 0.96. Each domain of BSF can be administered and scored as the average of all items in that domain. The BSF PROM, with instructions for administration and scoring, is available upon request.

Test-retest reliability

In analyses that included all 10 domains, test-retest reliability differed markedly among those without breast cancer, those with cancer BMR–, and those with cancer BMR+ (Table 3). ICCs for the BSF general factor were 0.48, 0.34, and 0.91 for the 3 groups, respectively. A similar pattern but with somewhat higher ICCs was evident for sexual satisfaction: 0.61 (unaffected), 0.53 (BMR–), and 0.97 (BMR+).

Table 3.

BSF test-retest reliability analysis: ICCs.^a

	ICCs
	Overall	Unaffected	Affected	BMR+ only	BMR– only
BSF domains
Importance of touch	0.62	0.61	0.55	0.91	0.40
General satisfaction/pleasure	0.65	0.63	0.56	0.85	0.44
General pressure	0.40	0.39	0.42	0.82	0.22
Ability to feel touch	0.67	0.55	0.67	0.92	0.53
Ability to feel temperatures	0.62	0.58	0.51	0.78	0.44
General pain/discomfort	0.60	0.43	0.55	0.93	0.37
Pain in response to stimuli	0.52	0.26	0.69	0.92	0.53
Touch-related pleasure	0.56	0.50	0.53	0.87	0.34
Nipple erection	0.63	0.59	0.53	0.86	0.45
General importance	0.55	0.53	0.55	0.89	0.40
BSF general factor scoreb	0.56	0.48	0.54	0.91	0.34

Abbreviations: BMR–, women with breast cancer but without bilateral mastectomy and reconstruction; BMR+, women with breast cancer who underwent bilateral mastectomy and reconstruction; BSF, breast sensorisexual function; ICC, intraclass correlation coefficient.

^aAll ICCs were based on a 2-way mixed effects model assessing consistency of agreement.

^bAverage of 24 items included in bifactor model.

Construct validity

Scores for each of the 10 domains differed among the 3 clinical groups (Table 4). As expected, women in the BMR+ group had the lowest sensory function, and the affected women in the BMR– group had somewhat lower sensory function for several domains than unaffected respondents. Pain was higher among those with breast cancer, as expected, but was similar between the cancer treatment groups. Also as expected, the predicted mean of the BSF general factor was higher among unaffected respondents than the affected groups; among the affected groups, the BSF general factor was lowest among those in the BMR+ group. The BSF general factor was less variable for the BMR+ group than the other groups, and nearly all respondents in this group had a score <0 (representing the average over the entire sample; Figure 2).

Table 4.

Predicted means of BSF domains, comparing women affected and unaffected by breast cancer, adjusted for age and self-reported health.

Domain: group	Predicted mean	Regression coefficient (95% CI)	Overall P value (2-df test)
General satisfaction/pleasure			<.001
Unaffected	3.28	Reference
Affected, BMR–	2.98**	−0.30 (−0.49, −0.10)
Affected, BMR+	2.37***^^^	−0.92 (−1.15, −0.68)
General pain/discomfort			<.001
Unaffected	1.50	Reference
Affected, BMR–	2.09***	0.59 (0.44, 0.73)
Affected, BMR+	2.27***	0.77 (0.59, 0.94)
Ability to feel touch			<.001
Unaffected	3.55	Reference
Affected, BMR–	3.27**	−0.28 (−0.46, −0.10)
Affected, BMR+	2.49***^^^	−1.06 (−1.29, −0.84)
Importance of touch			<.001
Unaffected	3.22	Reference
Affected, BMR–	3.04	−0.18 (−0.38, 0.02)
Affected, BMR+	2.56***^^^	−0.67 (−0.91, −0.43)
Touch-related pleasure			<.001
Unaffected	3.27	Reference
Affected, BMR–	3.02*	−0.24 (−0.44, −0.05)
Affected, BMR+	2.44***^^^	−0.82 (−1.06, −0.59)
Ability to feel temperatures			<.001
Unaffected	3.30	Reference
Affected, BMR–	3.04**	−0.26 (−0.44, −0.08)
Affected, BMR+	2.41***^^^	−0.89 (−1.11, −0.66)
Pain in response to stimuli			<.001
Unaffected	1.78	Reference
Affected, BMR–	2.15***	0.37 (0.21, 0.53)
Affected, BMR+	2.16***	0.39 (0.19, 0.58)
General pressure			.01
Unaffected	2.98	Reference
Affected, BMR–	2.89	−0.09 (−0.24, 0.07)
Affected, BMR+	2.70**	−0.28 (−0.47, −0.09)
General importance			<.001
Unaffected	3.03	Reference
Affected, BMR–	3.01	−0.02 (−0.22, 0.19)
Affected, BMR+	2.54***^^	−0.48 (−0.74, −0.23)
Nipple erection			<.001
Unaffected	3.27	Reference
Affected, BMR–	2.84***	−0.43 (−0.64, −0.22)
Affected, BMR+	2.16***^^^	−1.10 (−1.38, −0.82)
Overall BSF factor scorea			<.001
Unaffected	0.24	Reference
Affected, BMR–	−0.01**	−0.24 (−0.40, −0.08)
Affected, BMR+	−0.56***^^^	−0.80 (−1.00 to −0.61)

Abbreviations: BMR–, women with breast cancer but without bilateral mastectomy and reconstruction; BMR+, women with breast cancer who underwent bilateral mastectomy and reconstruction; BSF, breast sensorisexual function; df: degrees of freedom.

^aThe overall BSF factor score (standardized) is derived from the bifactor model that excluded the following domains: general importance, nipple erection, general pain/discomfort, and pain in response to stimuli (range, −2 to 2; see Figure 2).

^* P < .05.

^** P < .01.

^*** P < .001 vs unaffected.

^{^^} P < .01.

^{^^^} P < .001 vs affected BMR–.

Figure 2.

Distribution of the breast sensorisexual function (BSF) general factor score by breast cancer status. Green line, respondents with breast cancer and more invasive treatment (ie, bilateral mastectomy with reconstruction); red line, respondents with breast cancer and less invasive treatment (ie, other than bilateral mastectomy with reconstruction); blue line, respondents without breast cancer.

Among women unaffected by cancer, the BSF general factor was positively correlated with sexual satisfaction (0.61), ability to orgasm (0.54), and lubrication (0.44; Figure 3). Among the BMR– group, the BSF general factor was associated with all domains of sexual function (sexual satisfaction, 0.61; ability to orgasm, 0.59; lubrication, 0.53). Similarly, among the BMR+ group, the BSF general factor was correlated with sexual satisfaction (0.60) and lubrication (0.49) but not with ability to orgasm (−0.01). As expected, sexual function was lower among affected than unaffected respondents. Adjusting for the BSF general factor reduced these differences: BSF explained 40%, 49%, and 100% of the difference in arousal, ability to orgasm, and sexual satisfaction, respectively, between women with and without breast cancer (Table 5).

Figure 3.

Correlations between domains of breast sensorisexual function (BSF) and sexual function, health, and quality of life, stratified by cancer status and cancer treatment. General importance: the importance of the role that the breast or nipple plays during sexual activity.

Table 5.

Impact of the BSF general factor on the relationship between sexual function and breast cancer status.

	Mean ± SD		Difference a  (95% CI), R2
	Unaffected	Affected	Unadjusted	Adjusted for BSF b
Sexual satisfaction	47.5 ± 8.3	45.8 ± 7.1	−1.7, (−2.9 to −0.5), 0.01	0.7, (−0.4 to 1.7), 0.29
Lubrication (arousal)	47.0 ± 8.3	42.3 ± 7.6	−4.7, (−5.9 to −3.5), 0.08	−2.8, (−4.0 to −1.6), 0.23
Ability to orgasm	43.6 ± 11.0	38.3 ± 10.1	−5.3, (−7.0 to −3.6), 0.06	−2.7, (−4.4 to −1.1), 0.20

Abbreviation: BSF, breast sensorisexual function.

^aDifference = affected – unaffected.

^bAdjusted for BSF general factor (average of 24 items included in bifactor model).

Among unaffected women, the BSF general factor was positively correlated with quality of life (0.47) and self-reported health (0.31); this pattern was similar for women in the BMR– group (quality of life, 0.54; health, 0.35; Figure 3). For women in the BMR+ group, the BSF general factor was more weakly correlated with quality of life (0.25) and health (0.20); the 2 pain-related domains (breast pain and discomfort and pain with stimulation), however, were negatively correlated with these constructs.

Discussion

The BSF measure is a new PROM, developed by using the National Cancer Institute’s PROMIS standards,²⁷ with evidence of validity to comprehensively assess the sexual sensory functions of the breast. It is designed for use among women with and without a history of breast cancer to enable assessment of clinically meaningful differences in breast sensation²⁸ between affected and unaffected groups. The BSF general factor explains between 40% and 100% of the difference in sexual function observed between women with and without breast cancer. Six domains of the BSF instrument yield a general factor score that may be calculated simply as the average of the 24 items in those domains. The 10 individual BSF domains may be administered and scored on their own, as needed. The BSF PROM is timely in light of innovations to preserve and recover breast sensation after mastectomy and reconstruction.^69,70

The Breast-Q Sensation Module,⁷¹ published after Smeele’s systematic review was completed,²⁴ is a new, high-quality PROM that focuses primarily on assessing the nonsexual sensory functions of the breast. The module includes items particularly useful to assessing 2 components of Smeele’s definition of breast sensation: ability to feel sensory stimuli and symptoms of sensory loss. While the Breast-Q Sensation Module includes assessment of the degree of feeling in the breast area with sexual touch, this instrument does not assess affective valence, such as the degree to which sexual touch is pleasurable. Furthermore, the Breast-Q items do not distinguish sensations felt on the nipple/nipple areolar complex or other breast areas. The BSF PROM complements the Breast-Q Sensation Module with more specific assessment of the sexual sensory functions of the breast and nipple. Clinicians and researchers in the field can choose to administer 1 or both measures, including the whole BSF instrument or select domains of interest, depending on their clinical focus and outcomes of interest.

Test-retest reliability was high in the BMR+ group: the observed ICC of 0.91 was equivalent to that reported by Tsangaris et al among women who had undergone breast reconstruction.⁷¹ Tsangaris and colleagues assessed nonsexual and sexual aspects of breast sensation among >600 participants twice, 1 week apart. Test-retest reliability among BMR– and unaffected individuals was lower in our study than in that by Tsangaris et al. This difference may reflect, as evident in our cognitive testing, that women without breast cancer and those treated for breast cancer but without bilateral mastectomy and reconstruction may have less experience thinking explicitly about their breast sensation. It is also possible that this difference is due to (1) real temporal fluctuations in BSF during the 2 weeks between our assessments, (2) a failure to account for differences between breasts among the BMR– group, or (3) characteristics of the individuals who completed both surveys in these groups (eg, possible cognitive issues among those older and in poorer health, less attention to the items among those unaffected). The latter possibility is consistent with lower observed test-retest reliability among these groups in sexual satisfaction, an unexpected finding given the high test-retest reliability of the sexual satisfaction measure.¹¹ Further work is necessary to elucidate this issue.

The differences in BSF according to cancer status and intensity of treatment provide evidence of construct validity. As compared with women with cancer, unaffected women had higher overall BSF, lower breast and nipple pain domain scores, and higher scores in the other BSF domains. In contrast, affected women who had undergone bilateral mastectomy with reconstruction had the lowest overall BSF, higher breast pain and discomfort scores, and lower scores in the other BSF domains than other women. These findings are consistent with prior studies focused on sexual and nonsexual breast sensation among women with breast cancer.^31,71–73

The BSF general factor was positively correlated with sexual satisfaction, ability to orgasm, and lubrication; these correlations were similar across cancer status and cancer treatment groups. Approximately half the observed differences in sexual function between affected and unaffected women was accounted for by adjusting for differences in BSF. These results corroborate findings from prior studies demonstrating a positive association between nonsexual breast sensation and sexual function among women with breast cancer^15,17,71 and provide further evidence of construct validity. More important, these findings underscore the impact of deficits in BSF on overall sexual function and suggest that BSF may help to predict sexual outcomes following treatment for breast cancer. In addition to corroborating prior findings, our study, which involved a racially and ethnically diverse sample, adds to findings from prior studies that enrolled mostly White women.^15,17,71

Studies of nonsexual breast sensation commonly describe loss of sensation as a quality-of-life issue and have demonstrated a relationship between postmastectomy sensation and quality-of-life outcomes.^32,42,73,74 In our study, the BSF general factor was moderately correlated with quality of life among women affected and unaffected by cancer. The correlation was weaker among women with bilateral mastectomy and reconstruction; however, the domain assessing the importance of light and firm touch on the breast and nipple during sexual activity was strongly correlated with quality of life among women in this group. The 2 pain-related domains (breast pain and discomfort and pain in the breast/nipple in response to stimuli, both during sexual activity) were weakly to moderately correlated with quality of life among women with bilateral mastectomy and reconstruction, and even weaker correlations were seen among unaffected women and women without bilateral mastectomy and reconstruction. This finding suggests that these domains may be distinct from the other domains of BSF, perhaps given their focus on frequency and intensity of experiences of pain and not women’s affective feelings about those sensations.

Patients with breast cancer relate the importance of loss of sexual breast sensation to physical and mental health concerns (S.T. Lindau, E.A. Pinkerton, E.M. Abramsohn, unpublished data), but we find a gap in knowledge about how breast sensory function measures correlate with overall health. Again, like sexual function and quality of life, the BSF general factor was positively correlated with self-rated health in all groups. Similar to the pattern seen for quality of life, the correlation between the BSF general factor and health was moderate for women with and without cancer but weak for women who had undergone bilateral mastectomy and reconstruction. Again, breast pain and discomfort during sexual activity were moderately correlated with health among this group, similar to the quality-of-life findings. The relationship of BSF with health and quality of life for women with bilateral mastectomy with reconstruction for breast cancer may reflect relatively lower variability in BSF for this group and limitations of the subsample size. Nonetheless, these findings do suggest that efforts to preserve and restore BSF may also benefit women’s overall health.

Certain limitations should be considered. First, we did not assess BSF among affected women who had undergone radiation alone. While radiation has been shown to affect nonsexual breast sensation,¹³ Tsangaris et al⁷¹ found no difference in breast sensation when comparing patients who had undergone radiation or chemotherapy alone. Second, 18% of women reported sensory or other differences between their breasts; we used their response to the first breast-related item for each subsequent question. The BSF measure may not fully capture the breast sensory experiences of these women, and efforts to preserve and restore breast function should consider the prevalence of sensory and other differences between a woman’s breasts. Third, while the BSF was developed with input from racially and ethnically diverse patients, clinical experts, and a diverse national sample to optimize generalizability to the US population, participation was limited to sexually active women. Assessment of BSF among and generalizability to sexual and gender minority groups and non–English-speaking people are important areas for future research. For example, the BSF measure may be particularly useful for optimizing breast reconstruction for male-to-female transgender people.

The Bionic Breast Project, developed in part with support from the US National Cancer Institute, is developing an alternative, neuroprosthetic approach to the problem of sexual and nonsexual sensory function loss after mastectomy.⁶⁹ This interdisciplinary effort aims to preserve and restore breast sensory function and mitigate breast pain after mastectomy. The Bionic Breast Device is an implantable neuroprosthesis that restores touch sensation and reduces pain following oncologic or nononcologic mastectomy through a neural stimulation approach that has been successfully deployed to restore touch via bionic hands and feet in people with upper or lower limb amputation.⁷⁵ Development of the Bionic Breast Device and other innovations to mitigate the burden of postmastectomy numbness and pain require PROMs, validated among a diversity of people with and without breast cancer, that can assess nonsexual and sexual aspects of breast sensation.

Conclusions

The BSF PROM is the first comprehensive measure of sexual breast sensation to be developed by using evidence-based PROMIS standards and fills an important gap in this domain. It adds another tool to the small toolkit of PROMs designed to measure women’s breast sensory function and is also freely available for public use.

Funding

The work described in this manuscript was supported by the National Cancer Institute Grant No. 1R21CA226726 (ST Lindau, PI), pilot funding from the Janet D. Rowley Discovery Fund at the University of Chicago Comprehensive Cancer Center (ST Lindau, PI), the University of Chicago Women’s Board (ST Lindau, PI) and the Bucksbaum Institute (ST Lindau, PI).

Conflicts of interest: S.T.L. is an inventor on 2 pending patent applications related to the Bionic Breast. S.T.L. also discloses that she is a contributor to UpToDate, Inc, which generates royalties to her laboratory at the University of Chicago. S.T.L. was the founder and owner of a social impact company, is president of a nonprofit community asset–mapping organization that employs youth, and holds debt in another corporate entity unrelated to the content of this article. S.T.L. and her spouse hold stocks and mutual funds managed by third parties. M.A. is a consultant for Medtronic, Inc and Boston Scientific, Inc. All other coauthors report no conflicts of interest.