Study population
The Survivorship Work and Sexual Health (SWEET)-study was a nation-wide questionnaire study of Norwegian long-term BCSs with the primary study objective to explore work life experiences >five years after diagnosis. Women aged 20-65 years who had been diagnosed with early-stage (I-III) BC in 2011 or 2012 were identified by the Cancer Registry of Norway (CRN) and invited to participate. They had to be free of prior or subsequent malignancies with the exception of ductal carcinoma in situ and non-melanoma skin cancer. In total, 2803 women were invited, and 1361 (49%) survivors responded after receiving one written reminder.
In this sub-study, survivors who were older than the legal retirement age in Norway at time of survey (≥ 67 years, n=349) were excluded, as were survivors with incomplete consent (n=3), self-reported BC recurrence (n=3), missing data on work status (n=31) or missing data on the work ability score (n= 49), leaving a final sample of 926 BCSs.
Primary outcomes
WA was assessed using the work ability index (WAI) which is based on self-report [13]. Three of the original seven items of the WAI were considered relevant in this cancer specific setting [18]. Item one, known as the work ability score (WAS), asked survivors to rate their current WA on a scale from 0 to 10 points, where 10 points reflect highest possible WA and zero implies extinguished WA. As recommended, WAS were categorized into four; poor (WAS category 1, score 0-5), moderate (WAS category 2, score 6-7), good (WAS category 3, score 8-9) and excellent (WAS category 4, score 10). The WAS has been identified as a valid indicator of WA [19].
In item two of the WAI survivors were asked to describe their current WA with respect to the physical and the mental demands of their work. Response categories range from very bad (1) to very good (5). In item three, survivors were asked to rate to what extent their physical and mental WA has been affected by cancer. Response categories range from a lot (1) to not at all (5). The WAI has been reported to have good validity and has been used extensively in clinical occupational health research [20].
Explanatory variables
Cancer-related variables including age at diagnosis and information on surgical treatment (mastectomy or breast conserving therapy) were supplied by the CRN. All other variables were based on self-report.
Sociodemographic information included age at survey, living arrangements (with or without a partner) and educational attainment (long; >12 years and short; ≤12 years).
Work status was categorized as paid work, disability pension, retired or ‘other statuses’ (including job seeker, temporarily laid off, work allowance, education or military service and homemaker). Paid work included full- and part time work, self-employment and sick leave. In order to assess change in WA from diagnosis to survey, we asked survivors to rate their WA at time of diagnosis on the same scale (0-10) as used for the WAS.
Social support during cancer therapy was described using the 12-items version of the Structural and Functional Support Scale (SFSS) [21] in which survivors rate to what extent they received support from their supervisor (items 1-4), colleagues (items 5-8) and GP (items 9-12). Each item had five response categories, ranging from 1 (not at all) to 5 (a lot). For each of the three support providers a sum score was calculated, ranging from 4-20 [14]. We also dichotomized the score into high (score 11-20) and low (score 4-11) support. The SFSS is reported to have good psychometric properties [21]. Cronbach’s alpha for the SFSS was 0.9.
Current social support at the work place was described using the support sub-scale from the Demand Control Support Questionnaire (DCSQ) [22], which includes six positive statements concerning the general atmosphere at the work place and the support provided from supervisors and colleagues. Survivors rated their level of agreement with these statements on a four-point scale from complete agreement (1) to no agreement at all (4), yielding a sum score from 6-24, where lower values indicate higher social support. Cronbach’s alpha was 0.9 for the DCSQ.
Use of primary health care services was explored by asking survivors to report the number of visits to the GP during the last 12 months.
Life style factors included overweight (defined as body mass index (BMI) ≥ 25kg/m2) and physical activity. The latter was explored using a modified version of the Godin Leisure Time Exercise Questionnaire [23], classifying the survivors as either physically active (≥150 minutes of moderately intense- or ≥ 75 minutes of vigorously intense weekly activity) or inactive.
Sleep problems were defined as having more than three episodes per week of difficulty falling asleep and/or waking up too early and not being able to fall asleep again during the last three months (yes/no) [24].
Neuropathy was assessed using two items from the scale of chemotherapy induced long-term neurotoxicity (SCIN) [25], which asks survivors to describe the presence or absence of neuropathy in hands and feet with response alternatives ranging from 0 (not at all) to 3 (very much), yielding a sum score from 0-6. The sum score is then, as recommended by the authors [25], dichotomized into high (≥4) and low (≤3) degree of neuropathy.
Using subscales from the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC QLQ-C30) version 3 [26], cognitive function, pain and fatigue were reported, while arm and breast symptoms were assessed using the breast cancer specific module (EORTC BR 23)[27]. Items were rated from 1 (not at all) to 4 (very much) and transformed to a 0-100 scale as recommended in the EORTC scoring manual. More severe pain, fatigue, arm and breast symptoms result in higher symptom scores, while higher cognitive function yields a higher functional score. Cronbach’s alphas were 0.7 for cognitive functioning, 0.9 for pain and fatigue, 0.8 for arm- and breast symptoms.
Depressive symptoms was explored using the Patient Health Questionnaire, the PHQ-9 [28], which rates the frequency of depressive symptoms over the last two weeks. Responses range from not at all (0) to nearly every day (3), providing a sum score from 0-27. We imputed values for the PHQ-9 by substituting missing values with mean values if no more than two items were missing. Cronbach’s alpha for the PHQ-9 was 0.8.
Symptoms of anxiety were evaluated using the General Anxiety Disorder 7-item tool (GAD-7)[29], which rates the presence of anxiety during the last two weeks, scored from 0 (not at all) to 3 (nearly every day) resulting in a sum score from 0-21. We imputed values for the GAD-7 by substituting missing values with mean values when more than 50% of items had been answered. Cronbach’s alpha for GAD-7 was 0.9.
Fear of cancer recurrence was assessed using four items from the Concern About Recurrence Questionnaire (CARQ) [30], which includes an assessment of the frequency and degree of distress (first three items) and a self-perceived risk of recurrence assessment from 0-100% (fourth item). As recommended, the fourth item was converted to a 0-10 scale to be in accordance with the other items. A sum score from 0-40 was calculated. Cronbach’s alpha for CARQ was 0.7.
Statistical analyses
Descriptive analyses were performed for the total sample, describing continuous variables using means and standard deviations (SD) and categorical variables as numbers and percentages (%).
To describe the relationship between social support during cancer therapy and WA, we performed cross tabulations between the different items of SFSS for each provider of support and WAS. P-values from Pearson’s chi-squared test and Cramér’s V were reported.
To explore factors associated with excellent WA we applied logistic regression models with WAS as the dependent variable (score 10 vs. score 0-9). BMI was considered a continuous variable. Explanatory variables associated with excellent WA at significance level of ≤0.2 in univariable analyses were included in the multivariable model.
Collinearity statistics were performed for sets of variables which theoretically may overlap. This included age at diagnosis, cognitive function, fatigue, depression, anxiety and fear of cancer recurrence, with highest observed variation inflation factor (VIF) of 3.3 which is considered acceptable. As fear of cancer recurrence and anxiety has overlapping symptomatology, we chose to exclude anxiety from the final model. Furthermore, as the majority of BCSs who were disabled at time of survey rated poor current WAS, and work status and WA is closely related, we excluded work status from the regression analyses. Results were presented as odds ratios (ORs) with 95% confidence intervals (95%) and accompanying p-values.
In order to assess the potential risk of non-response bias, the CRN provided data concerning cancer-related variables for the non-responders. Comparisons of responders vs. non-responders were performed by comparing mean values for these selected variables.
All analyses were performed using IBM SPSS version 26.0 (SPSS, Chicago, IL).