This cross-sectional study included women with and without breast cancer recruited from outpatient clinics at a tertiary public hospital in metropolitan Melbourne, Australia. Data were collected between March - September 2019. For the group with breast cancer, we included women who had undergone primary treatment and/or adjuvant therapy for a histologically confirmed breast tumour and were ≥ 18 years of age. The inclusion criterion for the control group was women with no history of breast cancer. Women who were pregnant, unable to communicate in English, with neurological disorders, or with severe physical or psychiatric impairments were excluded from both groups.

Socio-demographic and medical variables collected included age, body mass index (BMI), parity, relationship status, social situation, educational level achieved, employment, smoking, menopausal status, Charlson Comorbidity Index,12 and medications. Menopausal status for the control group was categorised according to the description outlined by the Australasian Menopause Society.13 As breast cancer treatments may interfere with ovarian function, menopausal status fluctuates in women with breast cancer.14 Therefore, menopausal status in women with breast cancer was categorised according to the description within breast cancer literature15 in consultation with an endocrinologist (AV). Additional clinical data including breast cancer stage, treatment status, and current medications for cancer treatment were extracted from the medical records of consenting participants.

The occurrence of and distress related to PF dysfunction were assessed using the Pelvic Floor Distress Inventory (PFDI-20).10 This instrument consists of 20 questions in three domains: urinary (6 questions), pelvic organ prolapse (6 questions), and colorectal-anal (8 questions).10 The PFDI-20 has been shown to have good face, content, and construct validity16 and test-retest reliability.17 The impact of PF dysfunction was assessed using the Pelvic Floor Impact Questionnaire (PFIQ-7).10 This questionnaire consists of seven questions in three domains (urinary, pelvic organ prolapse, and colorectal-anal)10 and has established good face, content, and construct validity16 and test-retest reliability.17

The primary outcome of this study was the presence of UI which was established from a positive response to either question 16 or 17 on the PFDI-20.10 An affirmative response to questions 9 or 10 of the PFDI-2010 was used to quantify the presence of faecal incontinence while question 3 was used to quantify the presence of pelvic organ prolapse.10 The PFDI-2010 was also used to quantify the severity of bother (distress) of PF dysfunction as per the scoring algorithm.10 Summary scores ranged from 0 (least distress) to 300 (greatest distress).10 The impact of PF dysfunction on the daily lives of women was quantified using summary scores of the PFIQ-7.10 The PFIQ-7 summary score was calculated according to the scoring algorithm and ranged from 0 (least impact) to 300 (greatest impact).10

The sample size for this study was based on our primary outcome which was the occurrence of UI. The occurrence of UI within the last four weeks has been found to be 45–55% in women receiving endocrine therapy for breast cancer18,19 compared to 25–36% in women without breast cancer20,21; a 20% difference in occurrence rates. To achieve a between-group difference in occurrence of UI of 20%, with 80% power and 95% confidence intervals, we calculated a sample size of 240 participants would be required i.e. 120 in each group.

Descriptive statistics were used to summarise participant demographics and the magnitude of PF dysfunction in women with and without breast cancer. Independent t–tests for continuous data and χ2 tests for categorical variables were used to compare the differences in participant demographics and prevalence, distress, and impact of PF dysfunction between groups.

Logistic regression models were used to examine the association between breast cancer and the presence of PF dysfunction. A two-step modelling approach was applied. Firstly, univariable regression models for potential predictors and the presence of PF dysfunction were computed. Secondly, factors with a moderate association (p ≤ 0.1)22 on univariable analysis were entered into a multivariable model. Logistic regression models were adjusted for known covariates of the occurrence of PF dysfunction: age, BMI, and parity.23 The final model included the variables associated with the occurrence of PF dysfunction, variables of interest to the study (eg. presence of breast cancer), and the variables arising from sample differences (eg. age, BMI, and parity). Model findings were reported with odds ratios (OR) together with the corresponding 95% confidence intervals (CI). A p ≤ 0.05 was considered statistically significant.

Linear regression models were used to examine associations between the distress and impact of PF dysfunction in women with and without breast cancer using a similar two-step modelling approach. All variables were standardised before computing linear regression models. Initial testing showed that assumptions for linear regression were met for both the PFDI-20 and PFIQ-7. Highly correlated predictors were identified when the variance inflation factor (VIF) was higher than 2.5 in all regression models, in which case only the variable with the higher R2 was retained in the multivariable model. All analyses were performed using Stata v16.0/IC (StataCorp, LLC).

Figure 1 illustrates the recruitment flowchart. 697 women with breast cancer were identified by medical staff to be eligible for this study and 120 consented to participate (17% consent rate). All women who consented to participate subsequently completed the questionnaires (100% completion rate). In the control group, the consent rate was calculated by dividing the number of women who consented to completing the questionnaire by the number of women who viewed the questionnaire online. Seventy-six percent consented to participate (332/428) and 51% completed the questionnaire (170/332). There were 162 responses that were either completely blank (n = 149) or partially completed (n = 13) from the control group. Only complete responses were included in this analysis.

Figure 1.

Recruitment flowchart.

(0.33MB).

The demographic and clinical characteristics of all participants, and occurrence of PF dysfunction are presented in Table 1. Women with breast cancer were significantly older (53 ± 12 years old) and had a higher BMI (31 ± 7 kg/m2). A larger proportion of women with breast cancer were post-menopausal (86/120, 72%) than women without breast cancer who were mostly pre-menopausal (109/170, 64%).

Seventy-four percent of women with breast cancer (88/120) experienced any type of UI, compared to 57% in the control group (97/170) which was a statistically significant difference between groups (percentage difference = 17%; 95% CI: 7, 29) (Table 1). Further analysis of the occurrence of UI according to menopausal status in women with breast cancer demonstrated that pre-menopausal women (7/7, 100%) experienced higher rates of UI compared to peri or post-menopausal women (77/109, 70%). Conversely, women in the control group who were either peri or post-menopausal experienced UI at higher rates (47/57, 82%) than women who were pre-menopausal (46/109, 42%). The factors associated with the presence of PF dysfunction in both groups are presented in Table 2. After controlling for known risk factors of PF dysfunction and factors with a moderate association on univariable analysis, there were no statistically significant associations between breast cancer and the presence of incontinence or pelvic organ prolapse.

The distress and impact of PF dysfunction are presented in Table 1. The PFDI-20 urinary domain score showed the highest distress value in both groups, suggesting that urinary symptoms were the most distressing in both groups of women. Women with breast cancer experienced a significantly higher distress of urinary symptoms than women without breast cancer (mean difference = 5.5; 95% CI: 0.3, 10.6).

There was a significant difference in impact summary scores between groups; women with breast cancer experienced higher impact of PF dysfunction than women without breast cancer (mean difference = 18.2; 95% CI: 8.9, 27.7). The urinary domain score of the PFIQ-7 was the one with the highest impact value in both groups. Women with breast cancer experienced a significantly higher impact of urinary symptoms than women without breast cancer (mean difference = 12; 95% CI: 8.6, 41.2).

In Table 3, univariable analyses showed that higher age, higher BMI, not having a university degree and not currently working were moderately associated with distress of PF dysfunction, based on the overall PFDI-20 score. Of the variables that were entered into the multivariable model, age (β 0.21; 95%CI: 0.06, 0.36) and not having a university degree (β −0.45; 95%CI: −0.77, −0.13) were the strongest predictors of distress of PF dysfunction. The presence of breast cancer was not associated with the distress of PF dysfunction in the multivariable model. This model explained 6.3% of variance in overall distress scores.

The presence of breast cancer, higher age, higher BMI, not having a university degree, not currently working and being a current smoker were moderately associated with the impact of PF dysfunction on univariable analysis (Table 3). Multivariable analysis showed that these variables accounted for 13.3% of the variance in overall impact scores with not currently working (β −0.50; 95%CI: −0.77, −0.23), being a current smoker (β 0.33; 95%CI: 0.10, 0.59) and the presence of breast cancer (β 0.09; 95%CI: 0.02, 0.18) being statistically significant predictors of impact scores. Having breast cancer was positively associated with impact scores demonstrating that women with breast cancer reported a higher impact of PF dysfunction than women in the control group.

As we observed higher distress and impact scores in the urinary domain of the PFDI-20 and PFIQ-7 in women with breast cancer than women without breast cancer, the association between breast cancer and the distress and impact of urinary symptoms were also computed (see Supplemental material online 1). The presence of breast cancer was not associated with the distress of urinary symptoms. Only age (β 0.21; 95%CI: 0.04, 0.29) and not having a tertiary education (β −0.47; 95%CI: −0.80, −0.15) were statistically significant predictors in the multivariable model that explained 7.2% of the variance in urinary distress scores. In contrast, having breast cancer (β 0.33; 95%CI: 0.08, 0.51) and currently working (β 0.21; 95%CI: 0.05, 0.37) were significantly associated with the impact of urinary symptoms on multivariable analysis. This model explained 5.8% of variance in urinary impact scores.

Despite our efforts to minimise selection bias by recruiting a larger number of women without breast cancer, we observed statistically significant differences in the known risk factors of PF dysfunction (age, BMI, and parity) between groups. Post hoc matched analyses were conducted to minimise the potential confounding effects of these variables. A total of 77 pairs were identified when women with and without breast cancer were sequentially matched according to five-year age groups, BMI (categorised into underweight, normal, overweight, and obese) and parity (nulliparous or parous) using SPSS v25 (IBM Corp, Armonk, NY). Differences in demographic characteristics, prevalence, distress, and impact of PF dysfunction between the two paired groups were assessed using paired t-tests for continuous data and McNemar's test for categorical variables.

Accuracy of matching between the two groups was confirmed as there were no significant differences found in age, BMI, and parity. A similar proportion of women were post-menopausal in each group: women with breast cancer (n = 49/77, 63%); women without breast cancer (n = 43/77, 56%) (Table 4).

Sixty-eight percent (n = 52/77) of women with breast cancer experienced UI compared to 39% (n = 30/77) in women without breast cancer which was significantly different between groups (percentage difference = 29%; 95%CI: 9, 37%). There were no differences in the occurrence of faecal incontinence and pelvic organ prolapse between groups. Women with breast cancer reported a significantly higher distress of PF dysfunction (in all domains) than women without breast cancer (mean difference = 23.5; 95%CI: 11.8, 35.2). They also reported a significantly greater impact of PF dysfunction than women without breast cancer (mean difference = 21.5; 95%CI: 10.4, 32.6).

Some study limitations need to be noted. Firstly, the cross-sectional design limits our ability to draw causal inferences regarding the association between breast cancer and PF symptoms. Our method for selecting variables (based on differences between groups) into the final multivariable regression model may have resulted in a lower goodness of fit compared to other methods such as step-wise or hierarchical regression. However, all significant variables were likely captured as we included moderately associated variables (p ≤ 0.1) into the final model. Previous history of PF symptoms prior to breast cancer diagnosis was not collected as part of this study. Therefore, there is a possibility that women with breast cancer experienced PF dysfunction prior to cancer treatment. Further studies with a longitudinal design are warranted to determine the incidence of PF symptoms throughout the breast cancer treatment trajectory. A large proportion of participants in the control group also did not respond to any question in the questionnaires. As these responses were missing at random and not included in our analyses, we were unable to determine whether this may have produced different outcomes in this group, leading to invalid conclusions. However, this study had high participation rates and a large sample size compared to other similar studies.18 It was also powered to assess the difference in occurrence of UI in women with and without breast cancer.