Data source
We recruited both breast cancer outpatients and inpatients from Sichuan Oncology Hospital from November 2017 to May 2018. Ethical permission was granted by the Ethics Committee, West China School of Medicine/West China Hospital, Sichuan University (approval number 2017-255). We obtained permission to use the FACT-B instrument (Simplified Chinese version) and the EQ-5D-5L questionnaire (Simplified Chinese version).
Study participants
The inclusion criteria were as follows. First, participants were clinically and/or pathologically diagnosed with breast cancer. Second, patients were aged 18 and above. Third, patients did not have any mental problems and have the ability to express. In addition, patients agreed to participate in this study. Informed consent was obtained from all participants. We excluded patients who had comorbidities such as cardiovascular disease and mental health problems.
We calculated the sample size as follows:
See formula 1 in the supplementary files.
As suggested by prior research[21], the standard deviation of the health utility value of breast cancer patients was 0.16, the 95% confidence interval of the mean was 0.82 to 0.85, and δ was 1/2 of the width of the confidence interval. Hence, the final target sample size was 440. We recruited 451 breast cancer patients, and 5 respondents did not complete the survey. Therefore, 446 participants were included in the data analysis.
Measures
We measured participants’ quality of life by using the FACT-B instrument, which assesses quality of life across five dimensions: physiological well-being (PWB), social and family support (SWB), emotional well-being (EWB), functional well-being (FWB), and additional breast cancer symptoms (BCS). The FACT-B instrument consists of 37 questions. Since the scores of these five dimensions differed, we standardized them into a scale ranging from 0 to 100. The validity and reliability of the Chinese version of the FACT-B instrument were examined by prior investigators [22].
Additionally, patients’ health utility was measured by the EQ-5D-5L questionnaire (Simplified Chinese version), which uses five assesses five health dimensions (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression) with five levels of severity (no problems, slight problems, moderate problems, severe problems or extreme problems/unable). The severity of each dimension was coded from 0 to 4 with as the reference group. For example, a score of 0 for mobility represents that individuals have no problems with walking, and a score of 4 represents individuals who cannot walk. In addition, participants were required to report their self-rated health status on a scale ranging from 0 to 100, with 0 representing the worst health status and 100 representing the best health status one can imagine (EQ-VAS). The validity and reliability of the EQ-5D-5L questionnaire (Simplified Chinese version) were examined by previous researchers [23]. We calculated health utility by employing a value set based on Chinese data [24].
The main independent variable of interest is disease states, i.e., P, R, S, and M. In addition, we introduced covariates including TNM stage (0, I, II, III, and IV), surgical approaches (breast conserving surgery, modified radical surgery vs. no surgery), menopause state (yes vs. no), radiotherapy (yes vs. no), chemotherapy (yes vs. no), targeted therapy (yes vs. no), endocrine therapy (yes vs. no), and inpatients (vs. outpatients) to control for clinical confounders that may affect patients’ health via adverse effects and are unrelated to disease states[12,19].
Furthermore, we assessed patients’ demographic attributes (age and marital status) and socioeconomic characteristics, including education attainment, household income, residence (urban vs. rural), occupation, and medical insurance type; these covariates were measured to control for the effect of social deprivation on health[15,17].
Data analysis
Data analysis methods for H1
To assess the differences in variables between the four disease states, descriptive analyses (Chi-square test, Fisher’s exact test, and ANOVA) were performed depending on the characteristics of the variables. We calculated health utilities by a value set developed based on previous research in China[24]. To assess the degree of overlap between instruments, Spearman's rank correlation coefficient was calculated not only between each instrument but also between the domains of the FACT-B instrument.
ANOVA and the Wilcoxon rank-sum test were used to compare quality of life scores and health utility scores between different disease states.
Univariate analysis was conducted to determine potential predictors of participants’ health, which was reflected as overall scores on the FACT-B instrument, scores of each dimension of the FACT-B instrument, self-rated health (EQ-VAS), health utility (total score on the EQ-5D-5L questionnaire), and the scores of all five dimensions of the EQ-5D-5L questionnaire. The univariate analysis included 15 independent variables, such as age, marital status, education attainment, residence, medical insurance, occupation and household income. Independent variables that were statically significant (p< 0.05) in the univariate analysis were then introduced in the multivariate analysis. Variance inflation factors were calculated to examine multicollinearity among independent variables in multivariate models.
We performed multiple regression models according to the distribution of the dependent variables. A linear regression model was performed for the overall FACT-B scores since the data were normally distributed. Ordinal logistic regression models were performed for BCS, FWB, EWB, SWB, PWB, and self-rated health, as the distributions of these variables were highly skewed; we also performed ordinal logistic regression for the degree of mobility, self-care, usual activities, pain, and depression, as these variables were ordinal. For BCS, FWB, EWB, SWB, and PWB, we divided each variable into four balanced groups coded as 0, 1, 2, or 3, with 0 representing the lowest group and the reference group in the model; each group consisted of a similar number of participants. A similar process was used for self-rated health status, with five balanced groups coded as 0, 1, 2, 3, or 4, with 0 representing the group with the worst health. The Tobit model was performed for health utilities, as these data were right-censored.
Data analysis methods for H2
We analysed the correlation between quality of life and health utilities from the EQ-5D-5L questionnaire by employing a rank correlation test. Furthermore, we estimated health utilities from the quality of life (assessed by the FACT-B instrument) by employing a mapping function derived from the Singaporean population[20] and conducted a rank correlation test between estimated health utilities and those directly measured from participants using the EQ-5D-5L questionnaire.
The mapping function based on Singaporean patients was as follows[20]:
Estimated health utility =0.2846+0.0121×PWB+0.0044×FWB+0.0034×BCS
Data analyses were performed with SPSS 23.0. and SAS University Edition. A P value of less than 0.05 was considered statistically significant.