The Cervantes Scale for Measuring Health-Related Quality of Life in Postmenopausal Breast Cancer Survivors During Adjuvant Endocrine Therapy

Background: Potentially, instruments to manage adverse effects may improve adherence and persistence of treatment, health-related quality of life (HRQL) and breast cancer (BC) outcomes. In this sense, we consider the Cervantes Scale an appropriate option to assess HRQL in BC survivors undergoing adjuvant endocrine therapy with Aromatase Inhibitors (AI), as it is a specic HRQL questionnaire that considers particularities of the perimenopausal and postmenopausal women. The purpose of this study was to analyze the psychometric properties of the 31-item Cervantes Scale (CS-31), 16-item Cervantes Short-Form Scale (CS-16) and 10-item Cervantes Scale (CS-10) and to promote the additional validation for breast cancer (BC) survivors during adjuvant endocrine therapy. Methods: This prospective study included 89 postmenopausal BC survivors in endocrine therapy. The women completed the CS, Functional Assessment of Chronic Illness Therapy - fatigue (FACIT-F) and Hospital Anxiety and Depression Scale (HADS). The internal consistency, convergent validity, responsiveness analyses and known-group validity of CS were evaluated. Also, simple linear regression assessed the predictive capacity of the CS in relation to the FACIT-F. Results: The internal consistency was good (Cronbach's alpha: CS-10=0.76; CS-16=0.80; CS-31=0.89). For convergent validity, in general, we identied negative and strong correlations (r>0.6) with exception of moderate correlation between Social/Family Well-Being of FACIT-F and Sexual (r=-0.453) and Couple (r=-0.436) of CS-31, and weak correlation between Social/Family Well-Being of FACIT-F and Sexual (r=-0.289) and Couple (r=-0.287) of CS-16. A prospective improvement in health-related quality of life (HRQL) were identied, except for domains Sexual and Couple of CS-16/CS-31, as well as FACT-G, Social/Family and Functional Well-Being of FACIT-F. The CSs were able to predict endocrine with a worse HRQL. The CS Global score, the Menopause and Health score (CS-16, p < 0.001; CS-31, p = 0.004) and Psychic score (CS-31, p = 0.002) were higher in T0 compared to T1 and T2, and did not differ between T1 and T2. The Psychic score (CS-16, p = 0.011), Emotional Well-Being (p = 0.029), Physical Well-Being (p = 0.016) and FACIT-F TOI (p = 0.017) differed statistically between T0 and T2, with higher scores at baseline. The Fatigue Subscale had higher at T0 when compared to T1 (p = 0.012). The domains Sexual and Couple (CS-16 and CS-31), as well as FACT-G, Functional and Social/Family Well-Being (FACIT-F) did not differed statistically.

Thus, the aim of this study was to analyze the psychometric properties of the 31-item Cervantes Scale (CS-31), 16-item Cervantes Short-Form Scale (CS-16) and 10-item Cervantes Scale (CS-10) and to promote the additional validation of these scales for BC survivors during adjuvant endocrine therapy. To the best of our knowledge, this study is the rst one that used the CS to assess HRQL in BC women, especially during AI use. The study was carried out from January 2016 to August 2018 with postmenopausal BC survivors undergoing adjuvant endocrine therapy with AI at the Clinical Hospital of the Federal University of Uberlandia, Minas Gerais, Brazil. The follow-up time was 24 months, and the face-to-face assessments were performed at three time points: T0, initial follow-up period; T1, intermediate follow-up period, 12 months after T0; and T2, nal follow-up period, 24 months after T0, with interviews carried out by properly trained nutritionists.
The recruitment, exclusion criteria and the Cervantes' sample are described in Fig. 1. Clinical and sociodemographic data were obtained through the analysis of medical records or interviews.

Sample Size
The survivors were recruited at any time of the AI treatment, through the non-probabilistic sampling for convenience. The volunteers were accrued consecutively to avoid selection bias.
The sample size of a group of individuals and three measurements was calculated with the G*Power software, version 3.1 (Düsseldorf, Germany) [12]. An F test was conducted using ANOVA repeated measures, based on an effect size f of 0.25, an alpha level of 0.05 and at 80% power, being required a total of 28 women at each time.

Cervantes Scale
All participants replied by interview to the CS. Figure 2 provides the items contained in the three versions of the CS.
The CS-31 was developed in 2004 [9] and validated in Brazil in 2012 [13], with Cronbach's alpha for the global score of 0.91 and 0.83, respectively. The CS-31 consists of 31 items rated on a Likert scale from 0 to 5 and divided into four domains, namely Menopause and Health (subdivided into Vasomotor Symptoms, Health and Aging), Sexual, Couple and Psychic, with scores ranging from 0 to 155 points. As recommended, the questionnaires were considered invalid if three or more questions were left unanswered. But, if one or two unanswered questions, the score was obtained by multiplying by a correction factor [9].
In 2013, the CS-31 was reduced to a short 10 item version (CS-10), with Cronbach's alpha of 0.78. The sum of scores of the 10 items provides a global score which can range from 0 to 50 points [14].
In 2015, a 16-item short version (CS-16) was developed as of CS-31, with Cronbach's alpha for the global score of 0.82. This scale maintained the original structure with four domains and the psychometric properties, but was established a rescaling of correction method to convert the original scores into a 0-100 scale [15,16]. The CS-16 can be administered faster, 2.5 minutes compared to the average application time of 7 minutes of CS-31. Three or more items without answer invalided the questionnaires.

Page 4/22
Only CS-31 was validated in Brazil, however, both CS-10 and CS-16 were developed from CS-31, including only issues already present in it. In all the CS, a higher score represents a worse HRQL.

Psychometric evaluation
All participants replied by interview to the Functional Assessment of Chronic Illness Therapy -fatigue (FACIT-F) and Hospital Anxiety and Depression Scale (HADS), which are validated instruments and already well established for BC patients.
Validation instruments.
FACIT-F (version 4). This scale of 40-item includes the 27-item Functional Assessment of Cancer Therapy-General (FACT-G) that assess the HRQL and 13 items that assess self-reported fatigue [17]. This instrument measures four well-being subscales (physical, social/family, emotional and functional), one fatigue subscale, and derives to calculate the FACIT-F Trial Outcome Index (TOI) (score range 0-108), the FACT-G total score (score range 0-108) and the FACIT-F total score (score range 0-160).
Items are rated on a Likert scale from 0 (not at all) to 4 (very much). A higher score represents a better HRQL. The FACIT-F has been previously validated in Brazil [18]. In the present study, the Cronbach's alpha were FACIT-F α = 0.93 (95% con dence HADS. This self-reported questionnaire comprising seven items target anxiety (subscale HADS-A) and seven items target depression (subscale HADS-D) [19]. Items are rated using a 4-point Likert-type scale with scores of 0 (minimally present) to 3 (maximally present). The scores range for HADS-A and HADS-D from 0 to 21, with higher scores indicating greater distress. We adopted the following cut-off for both scales: <8 for non-cases, ≥ 8 for doubtful cases and ≥ 11 for the identi cation of cases [19]. The HADS has been previously validated in Brazil [20]. In this study, the Cronbach's alpha were HADS-A α = 0.76 (95% [CI] = 0.67-0.83) and HADS-D α = 0.80 (95% [CI] = 0.72-0.85).

Statistical Analysis
Internal consistency was studied considering the global score, including all items, and for each domain. We use Cronbach's alpha coe cient, considering adequate values between 0.70 and 0.95 [21]. Convergent analyses were assessed by determining the correlation between the CS (global score and domains) and a speci c related measure from FACIT-F. We hypothesized that global score between CS and FACIT-F, as they are general measures, as well as constructs indirectly related (Sexual and Couple of CS with Social/Family Well-Being of FACIT-F) should correlate with r > 0.4. Regarding that similar constructs, as Psychic of CS and Emotional Well-Being of FACIT-F, our hypothesis was correlations with r > 0.6. Correlation coe cients < 0.4 were considered weak correlations, between 0.4 and 0.6, moderate correlations, and > 0.6, strong correlations [22].
For the responsiveness analyses, we assessed the change in global and domains scores of FACIT-F and CS between the three time points (T0, T1 and T2). Differences were evaluated using One-way ANOVA test with repeated measures and Sidak posthoc, or the non-parametric Friedman with multiple comparison tests.
Simple linear regression was performed to verify the predictive capacity of the CS Global in relation to the FACIT-F Total score, FACIT-F TOI and FACT-G.
The known-group validation analysis was performed to assess if the instrument would be able to discriminate between subgroups of women, using one-way ANOVA with Sidak post-hoc. The CS global scores were compared between non-cases, doubtful cases and cases of anxiety and depression by HADS. We hypothesized that those women with higher scores for both anxiety and depression, would have a worse HQRL. In BC patients, the HRQL is closely related with these psychological disorders [23].
All statistical analyzes were performed using IBM SPSS Statistics (Chicago, IL, USA), software package (SPSS Statistics for Windows, version 21.0), considering statistically signi cant p-values of less than 0.05. Table 1 shows the demographic and clinical characteristics of the 89 BC survivors.

Results
The median (p25-p75) of age was 65 (58.5-69.5) years, the median of time using AI was 29.5 (18.1-41.8) months, of time diagnosis was 4 (2-5) years, and of climacteric period was 16 (8-20) years. Most women were not married (56.2%, n = 50), but the most have partner (75.3%, n = 67). Regarding adjuvant endocrine therapy, 44.9% (n = 40) of women used tamoxifen prior to starting AI.  [21], indicating low internal reliability (Table 3).     (Table 5).     Simple linear regression analyses. All analyses were statistically signi cant (p < 0.001). The results indicated that the CS Global scores, especially the CS-31, were able to predict changes in the FACIT-F Total score, FACIT-F TOI and FACT-G Total score (Table 7). Table 7  Known-group validity. As previously hypothesized, those women with higher scores for both anxiety and depression by HADS presented worse HRQL by CS when compared to subgroup non-cases. Regarding the anxiety, the analysis of CS-16 revealed that the subgroup cases had worse HRQL even when compared to the subgroup doubtful cases (Table 8).

Discussion
The psychometric properties revealed that CS-31, CS-16 and CS-10 are valid instruments for assessing HRQL in BC survivors during adjuvant endocrine therapy. All the CS presented adequate internal consistency; satisfactory convergent validity for CS global and for almost all domains, except for Sexual and Couple of CS-16; and known-group validity, with statistical signi cance between anxiety and depression and worse HRQL. Furthermore, we identi ed a prospective improvement in HRQL of the baseline for the other time points. The CS were able to predict changes in the FACIT-F, especially the CS-31. To the best of our knowledge, our study is the rst one that used the CS to assess HRQL in BC women, especially during adjuvant endocrine therapy.
CS-10 contains a small number of items (32% of the original length), is simple to apply and does not include sexual and couple relationship issues, which can often be embarrassing for respondents. However, the items on this scale are not divided into domains, allowing more restricted analyses. The authors of CS-10 constructed this instrument to a quick menopausal symptom assessment, not speci cally to assessment of HRQL [14], although in the present study, it proved to be valid for this purpose. As for the CS-16 and CS-31, as they are multidimensional, they allow more in-depth analysis of adverse effects and general HRQL. Even though assessments of general HRQL may serve a purpose, knowing details that permeate the HRQL of these women, certainly, can be useful in clinical management. We identi ed superior psychometric properties of CS-31 over CS-16, which may be related both to the characteristics of sample and to the greater number of items presented in the rst. Furthermore, the CS-16 does not include any item related to symptoms in the musculoskeletal system, and aching in muscles and joints is one of the main adverse effects of AI use [24].
We identi ed higher scores for anxiety and depression in women with worse HQRL and improvement in HRQL over the study.
Recently, Martino and collaborators [25] identi ed that after 6 months of treatment with AI, BC patients presented a signi cantly higher perceived HRQL for both physical and mental components, added to a signi cant reduction of anxious and depressive symptoms, possibly due to the decline of the physical and psychological effects of recent diagnosis and previous treatments [25].
Regarding Sexual and Couple domains of CS, as well as Social/Family Well-Being of FACIT-F, the latter which also presents items related to sexual life and couple relation, have not changed over time.  [4], and some urogenital effects, such as vaginal dryness, persist lifelong if untreated [29]. Possibly, the treatment has a more lasting impact on sexuality and a longer follow-up would be necessary to investigate changes in these domains. The main recommendation for the management of sexual health in BC survivors is that a multidisciplinary team needs to include the sexuality as an integral part of treatment, contributing to an improvement of HRQL [26].
We hypothesized that constructs indirectly related as Sexual and Couple of CS with Social/Family Well-Being of FACIT-F should present moderate correlation. However, this expected correlation has not been achieved by CS-16. Probably, due to the greater number of items investigated, the CS-31 is more sensitive to capture nuances that involve the sexual function and couple relation. It should be noted that sexuality is considered a biopsychosocial concept, and therefore it is believed to be associated with biological and psychosocial factors [30].
We must recognize that our study has some limitations as a small sample size. Several correlations were weak and could have become signi cant with larger sample sizes. Even though the CS was not designed for this population, these women are considered BC survivors, i.e. currently free of the disease, and have predominantly adverse effects like to those of other postmenopausal women, although intensi ed by the AI use. We need to consider that the CS' target population is women aged 45 to 64 years [9] and our sample includes women aged 47 to 79 years. For this purpose, we divided women into two age groups (47 to 64 and 65 to 79 years) and observed that age had no effect on the CS scores by performing the Generalized Linear Model (GLzM) analysis (data not shown). In addition, the CS is a self-reported questionnaire, however in this study, all participants replied by interview, which may have inhibited responses to items in the Sexual and Couple domains. Even so, the standardization for this type of application was a methodological care considering that in our sample there were illiterate women.
As pointed out by others [13], we identi ed that most of invalid questionnaires were lled by women that were not married or without a partner, referring to a sexually inactive life. In fact, it was the reason for the sample difference between CS-31 and CS-16, considering that the rst present more items in these domains. This seems to be a limitation of the multidimensional CS, and adaptations to these instruments are necessary to contemplate all climacteric women, irrespective of their marital status and sexual activity.
The strengths of the current research include that our study has evaluated the psychometric properties of the three CS at the same time and used them to assess HRQL at three time points, with a 2-year follow-up.
Adjuvant endocrine therapy adherence is suboptimal in BC patients. It is negatively associated with treatment adverse events [8] and associated with increased early tumor recurrence and mortality rates [31]. Potentially, clinical interventions to manage these adverse effects may improve HRQL and BC outcomes [32]. Future studies are necessary to con rm whether the implementation of CS in routine medical oncology is able to contribute to improvements in HRQL and in prognosis of these women. Even so, we suggest that CS-31 is used in outpatient service to investigate HRQL and the CS-10 to early screening of adverse effects or to population studies that seek to investigate HRQL in BC patients in AI use, as it is the fastest to apply. We believe that the use of CS can optimize the attendance time and the health outcomes, since physicians could focus on individual adverse effects and monitor, through graphic summaries, the evolution of these effects after speci c interventions.

Conclusions
We identi ed that the CS, highlighting the CS-31, proved to be an appropriate option for use in routine medical oncology with BC survivors during adjuvant endocrine therapy, although larger studies are needed to con rm these results.