Primary breast cancer and health related quality of life in Spanish women: The EpiGEICAM case-control study

This study evaluates the impact of breast cancer (BC) in health related quality of life (HRQL) and in psychological distress (PD) during the initial phases of the disease and looks for contributing factors. A multicentric case-control study, EpiGEICAM, was carried out. Incident BC cases and age- and residence- matched controls were included. Clinical, epidemiological, HRQL (SF-36) and PD information (GHQ-28) was collected. We used multivariable logistic regression models to estimate OR of low HRQL and of PD in cases compared to controls, and to identify factors associated with low HRQL and with PD. Among 896 BC cases and 890 control women, cases had poorer scores than both, the reference population and the control group, in all SF-36 scales. BC women with lower education, younger, active workers, never smokers, those with comorbidities, in stage IV and with surgical treatment had lower physical HRQL; factors associated with low mental HRQL were dissatisfaction with social support, being current smoker and having children. Cases had a fivefold increased odds of PD compared to controls. Managing comorbidities and trying to promote social support, especially in younger and less educated women, could improve well-being of BC patients.

The main characteristics of BC cases and controls included in the study are presented in Table 1. BC cases and controls had a similar age (mean: 50 years), more than 90% were born in Spain, more than 70% were married or with a partner, 20% were nulliparous, around 55% of them had no comorbidities, 41-43% were non-smokers and 45% were caring for someone. On the other hand, BC cases reported more social support, had lower education, were more often retired or housewives, had lower adherence to WCRF/AICR recommendations and a higher frequency of BC family history (25% vs 19%).
With respect to clinical characteristics of the cases, 67% had hormone receptor positive and HER2 negative tumours, and 80% were in stages 0-II. At the time of answering the questionnaire, 79% had received surgery, and 40% were receiving radiation therapy or chemotherapy (Table 1).
Health-related quality of life. In women with BC, mean scores in each of the eight SF-36 scales and in physical and mental component summaries were under 50, showing that these women perceived their health as somewhat poorer than did women in the general population. The scale with the lowest mean was role-physical (36. 8 Fig. 1). In the analysis that differentiates cases in terms of TNM stage and surgical status, differences with respect to the control group in the bodily pain scale were seen only in operated cases; in physical functioning, role-physical and the PCS differences were Be as lean as possible within the normal range of body weight, 2) be physically active as part of everyday life, 3) limit consumption of energy-dense foods and avoid sugary drinks, 4) eat mostly foods of plant origin, 5) limit intake of red meat and avoid processed meat, 6) limit alcoholic drinks, 7) limit consumption of salt and avoid moldy cereals or pulses, 8) aim to meet nutritional needs through diet alone, and 9) mothers to breastfeed. The tenth recommendation, targeted to cancer survivors was not included in the construction of the score of adherence, because it was not applicable to the recently diagnosed population included in our study. Thus, the score ranges from 0 to 9. (2020) 10:7741 | https://doi.org/10.1038/s41598-020-63637-w www.nature.com/scientificreports www.nature.com/scientificreports/ higher in cases operated than in not operated. On the other hand, for role-emotional and the MCS, differences with respect to controls were higher in patients in the earliest tumour stages, independent of surgical status. For general health, vitality, social functioning and mental health, all cases showed higher odds of low scores compared to controls, being the magnitude of the ORs similar in the four subgroups (Fig. 3).
Factors related to HRQL. In women with BC, those older, retired, more educated, and smokers or former smokers had lower odds of low PCS. On the other hand, women that reported any comorbidity, those in stage IV and patients who had received surgery for BC treatment showed increased odds of low PCS (Table 2). These associations were similar among controls except for comorbidities, which showed a stronger association with low physical health in controls (p-interaction = 0.016). With respect to the mental component summary, BC women less satisfied with their social support and current smokers had higher odds of low MCS, while nulliparity was associated with reduced odds of low MCS. Among controls, younger women and those reporting first-degree BC relatives showed higher odds of low mental health (p-interaction: 0.005 and 0.019, respectively). The association with social support was stronger in controls than in BC cases, and smoking status and parity were not associated with mental health in control women. Results for specific SF-36 scales are reported in Supplementary Table S2. The sensitivity analyses including only non-metastatic BC patients produced similar results (data not shown).
Psychological distress. For  Factors related to psychological distress. In women with BC, lower satisfaction with social support, lower education, working with nightshifts and having children were related to higher odds of PD. More advanced tumour stage and chemotherapy were also associated with increased odds of PD (Table 2). In controls, the association with perceived social support was stronger than in cases, while no associations were observed with education, working status or parity. Reporting more than two comorbidities and low adherence to the WCRF/AICR recommendations were related to higher odds of PD in control women.

Discussion
EpiGEICAM BC cases had lower mean scores in all the HRQL domains evaluated by the SF-36, compared with both population reference values and the sample of controls. Role-physical, physical functioning and social functioning were the most affected domains. Psychological distress was very prevalent in BC cases, with more than 50% classified as such by the GHQ-28. www.nature.com/scientificreports www.nature.com/scientificreports/ Factors associated with poorer scores in physical health were younger age, lower education, never smoking, active working status, comorbidities, advanced tumour stage and surgery. The association with comorbidities in BC cases was weaker than in controls. This could reflect a response shift among BC women after cancer diagnosis, leading them to give less importance to limitations derived from other chronic diseases. Factors associated with poorer physical health specifically in BC cases were active working and never smoking. Active workers may have a more physically demanding activity than retired women, putting in evidence the limitations caused by the disease or treatments. With respect to smoking, we hypothesise that current and former smokers might attribute some of their physical limitations to consequences of smoking instead of the oncological disease and consequently give higher scores than never smokers in these domains. For mental health, low satisfaction with social support, smoking, and parity were associated with lower scores in BC women, while in controls, a younger age and family history of BC were the factors associated with lower scores. Controls in our study were age-matched to cases, and  Continued therefore, younger controls could have had a stronger psychological impact derived from their paired case having received the diagnosis at an earlier age. In contrast, for BC cases, having children seems to have a higher impact on mental health than age, which could reflect that women with children have additional worries related to the impact that her disease can have in their children.
The higher impairment in the physical areas compared to the psychological aspects observed in our results has been reported in other studies [12][13][14][15] , and could be related to the time of the HRQL measurement, which took place in the first months after surgery or during radiation therapy or chemotherapy in most participants.
With respect to the high prevalence of PD observed in BC cases in our study, even if it can be overestimated, as has been described with GHQ-28 in people with somatic diseases 16,17 , its magnitude and the differences observed with the control group support the importance of including appropriate screening and attention to psychological problems as part of the integral care for BC patients. Other studies that used the GHQ questionnaire to assess PD in BC women shortly after surgery have also reported elevated prevalences 11,18 . Dissatisfaction with social support, lower education, comorbidities, parity, advanced tumour stage and chemotherapy were the main determinants of PD in our patients. Interestingly, the associations with education and parity were observed only in BC cases, possibly reflecting differences in coping strategies in less educated women and greater concern in patients with children.
Perceived social support was associated in our study with both, PD and the mental component of HRQL. It is worth noting that the percentage of women very satisfied with their social support was much higher among BC cases than among controls, which may reflect a good response from the social networks of these women facing a hard life event. However, there were still some 12% of patients very unsatisfied with their social support, for which interventions aimed at improving it could have a positive impact.
These results should be interpreted in the context of study limitations. First of all, the cross-sectional nature of the design implies that the temporal sequence cannot be stablished for some variables, and we cannot know whether a factor, such as social support is a cause and/or a consequence of the poor HRQL or the PD. However, perceived social support has been found to predict HRQL during follow-up in other studies, giving support to an interpretation of our results in terms of perceived social support preceding HRQL 19,20 . Secondly, even though women participation was high (82%), the sample was not randomly selected and therefore it cannot be considered representative of all the BC patients. In addition, women that did not complete the SF-36 or the GHQ questionnaires were not included in the analyses, leaving women older, less educated and with poorer scores in certain HRQL scales under-represented in our results. To mitigate these limitations we have provided a detailed description of the characteristics of women included in our study in order to facilitate the assessment of external validity. An additional limitation to the external validity is that, by design, women older than 70 years were not invited to participate in our study. Thirdly, although we were able to adjust by several variables, residual confounding could still affect our estimates due to unmeasured variables (e.g. concomitant stressful life events) or to insufficient discriminative capacity of the variables included (e.g. comorbidities were self-reported through an open question). Also, the generic nature of the SF-36 may limit the possibility of evidencing certain disease-or treatment-specific adverse events or consequences. Nevertheless, this instrument has allowed us detecting differences with respect to the reference population and to the group of controls. On the other hand, given that we showed a comprehensive description of SF-36's results, including the eight individual scales and the two component summaries, and that we did not perform multiple testing adjustments, probability of finding some spurious statistically significant associations could not be ruled out. In this regard, consistency with other studies may help to assess the validity of our findings. The differences observed in ceiling and floor effects between cases and controls may have influenced our results, underestimating differences between these groups. Lastly, it should be noticed that controls were mostly selected among in-law relatives, friends and colleagues of BC cases, and therefore, they could be affected by the situation of their paired case. This would contribute to underestimate the differences in HRQL between cases and controls. However, given the small differences observed in HRQL between EpiGEICAM controls and www.nature.com/scientificreports www.nature.com/scientificreports/ the reference population, with the exception of physical functioning (Fig. 1), we consider that the magnitude of this underestimation would not be very high.
The current work has also some strengths, such as the high sample size, the exhaustiveness of data collection and the high participation rate. Also, we used the SF-36, which has shown good performance in BC patients, both in the literature 21,22 and in our own sample. The availability of population based reference values for this instrument allows a richer interpretation of the scores obtained, including an approximation to the impact of BC on HRQL. In addition, the recruitment of a control group also contributed to discern which limitations in HRQL could be more specifically derived from BC.
Results from this study contribute to characterise the impact of a recent breast cancer diagnosis in women' quality of life and the factors associated with impaired HRQL and with PD. This information could aid to advise women newly diagnosed with BC. Clinical and sociodemographic characteristics found to be related to HRQL and PD, such as age, education and parity could be assessed in clinical practice in order to identify women at increased risk of impaired HRQL and mental status, who may need specific interventions to improve them. Controlling comorbidities and reinforcing perceived social support could contribute, in the context of integral health care, to reduce the impact of BC diagnosis on women' quality of life. This is particularly important in less educated women, who have greater psychological distress.

Methods
EpiGEICAM is a case-control study carried out in nine Spanish regions. The study methods have been previously described 23,24 . Briefly, staff of the Oncology Departments of 23 public hospitals, members of the Spanish Breast Cancer Group (GEICAM, http://www.geicam.org/), recruited incident BC cases and controls between 2006 and 2011. To be included, women had to be 18-70 years old, able to answer an epidemiological questionnaire, and give written informed consent. For cases, BC had to be histologically confirmed and diagnosed in the last three months before recruitment. Women with previous history of BC were excluded. Controls were mainly selected by each BC case among their in-law relatives, friends, coworkers, and neighbours, and were matched by age (+/−5 years) and place of residence in a 1:1 ratio. All the participants answered an epidemiological questionnaire. Health related quality of life was measured by the SF-36 (version 1), psychological distress by the GHQ-28 25 and perceived social support by the Duke-UNC scale [26][27][28] . A more detailed description of the SF-36 and GHQ-28 instruments and the scoring methods applied is provided in the Supplementary Information (FORES), and by the Regional Institutional Review Boards of Burgos and Soria, Aragón-CEICA, Galicia, Cantabria and Jaén). The study was conducted following the principles of the Declaration of Helsinki and subsequent updates, and all the participants gave written informed consent to participate. Statistical methods. Psychometric characteristics of the SF-36 and the GHQ-28 were assessed by exploring the pattern of missing items, floor and ceiling effects, and the internal consistency by calculating the Cronbach's alpha.
As the analyses were restricted to women with information in all the SF-36 scales and in the GHQ-28, a comparison was made between these women and those not included in the analyses due to incomplete SF-36 or GHQ-28, using Chi-squared and Mann-Whitney U tests for categorical and quantitative variables, respectively. We performed a similar descriptive analysis of the included cases and controls.
For each SF-36 scale and component summary score, we defined a dichotomous variable classifying women as having low HRQL when their Norm-Based Score (NBS) was <= 45 points, that is, five or more points lower than the reference population's mean, fixed at 50 points. To compare the odds of having low HRQL and of PD between BC cases and controls we used multivariable logistic regression models adjusted by age, country of birth, region of residence, marital status, perceived social support, education, working status, caregiving, number of comorbidities, smoking status, nulliparity, BC family history, and adherence to the World Cancer Research Fund/ American Institute for Cancer Research (WCRF/AICR) lifestyle recommendations 29 . This last variable was calculated following previously described methodology 30 . We additionally explored the presence of differences in the odds of low HRQL or PD between cases and controls depending on cases' tumour stage and surgery status. For this purpose we reclassified cases in four groups according to the combination of TNM tumour stage (0-II vs. III-IV) and surgical status (operated vs. no operated), and fitted models like those mentioned above.
To identify factors associated with low HRQL and with PD, we performed separated analyses for cases and controls. We used multivariable logistic regression analyses to estimate odds ratios (OR) and 95% confidence intervals (95% CI), including the same covariates as mentioned above and, in the analysis of BC cases, also tumour type, TNM stage, and type of treatment received. Lastly, we explored if the observed associations differed between cases and controls in a statistically significant manner. For this purpose, in the multivariable logistic regression models we included interaction terms between each studied factor and case/control status, and tested their statistical significance. To allow tumour factors to be included in these models, controls were arbitrarily assigned to the categories of no surgery, no radiation therapy, no chemotherapy and stage 0/I in the corresponding variables. As a sensitivity analysis, we applied these same models to the subgroup of non-metastatic BC women.