Health-related quality of life among breast cancer patients compared to cancer survivors and age-matched women in the general population in Vietnam

We conducted an online survey and a hospital-based survey to reach BC patients and survivors in different geographical locations in Vietnam. Respondents, respectively, either accessed and completed the self-administrative web-based questionnaire or were approached and interviewed face-to-face. The hospital-based survey was conducted at Hanoi Oncology Hospital and Oncology Center of Hue Central Hospital (tertiary cancer treatment hospitals for the Northern and Central regions of the country, respectively). The EQ-5D-5L instrument developed by the EuroQol Group [14] was used to measure HRQoL of BC patients/survivors. The EQ-5D-5L was chosen because of its wide use internationally [15] which enhanced the opportunities to compare results with other studies.

We then used raw data¹ of an EQ-5D-5L standardised valuation study [16] conducted in 2017 with a nationally representative sample of Vietnamese general population to match each BC patient/survivor (as case) with a peer (woman in general population) by age (as control) (1-to-1 match). Firstly, data were randomly sorted, each case was matched with a ‘nearest’ control with same age or difference within a calliper of 0.25*standard deviation of age (recommended calliper by literature [17]). The matched pair of case and control was removed from the pool before next matching was performed. In short, the nearest neighbour matching within a calliper and nonreplacement was the matching algorithm [17]. After matching, HRQoL was compared amongst groups of BC patients, survivors, and age-matched peers.

Both the online and hospital-based surveys targeted: (1) BC patients (who were receiving hospital treatment) and (2) BC survivors (who finished treatment and discharged from hospital). Patients who were undergoing investigation for suspected BC but had not received a diagnosis were excluded. BC patients/survivors for the online survey were recruited through official websites and/or social media of national BC organizations as well as (breast) cancer patient/survivor clubs. In addition, all BC patients/survivors who presented at the named tertiary treatment centres whilst the survey was underway were approached and asked to participate. Bearing in mind available resources and logistics of conducting face-to-face interviews, data collection were restricted to three months for the online survey and three days for the hospital surveys. The sample size comprised the number of people who responded by the end of the time restriction.

The leaders (who are cancer survivors) of the two biggest BC and cancer (in general) patient/survivor ‘clubs’ in the country were invited to the advisory board of the study. They were not involved in the questionnaire development as the EQ-5D-5L is a standard instrument and its license forbids any changes though they were asked to assess the sensitivity of the questions and whether the time required to answer the questionnaire may cause any burden to the patients during their treatment. The leaders actively contributed to the recruitment process by distributing information related to the study and encouraging their peers to participate. We intend to disseminate the main results to the participants of the study and the general population of BC patients and survivors in Vietnam. The patient leaders will be consulted to choose an appropriate and effective method of dissemination.

Sociodemographic characteristics of BC patients and survivors were compared using t-tests and χ² tests. (1) The EQ-5D health profile; (2) The overall self-rated health status EQ-VAS; and (3) The utility scores (EQ-5D values) were used to compare HRQoL among BC patients, survivors, and age-matched women in the general population.

Regarding EQ-5D health profile, the frequencies and proportions of BC patients and survivors reporting each level of problem on each dimension was presented and χ² tests were used to assess between-group differences. The level of problems on each dimension were dichotomised into ‘no problems’ and ‘any problems’ to undertake comparisons with age-matched women from the general population.

Corresponding to EQ-VAS and utility scores, descriptive statistics used mean value and the standard deviation (SD). Kruskal Wallis tests assessed differences in BC patients’ median EQ-VAS/utility scores by stage of treatment. The most common regression techniques for analysing EQ-5D data, ordinary least squares (OLS), Tobit, and generalized linear model (GLM) were considered [21] in relation to conducting an assessment of the determinants of EQ-VAS and utility scores. Results from these models were similar. We decided to report the results from Tobit model as it takes account of the censored nature of EQ-5D data (bounds at full health and worst health state) [21]. Results from OLS and GLM models are provided in the Supplementary file 2. Two Tobit-derived models were presented. Model 1 contains all covariates based on a literature review and binary analyses including treatment status (patients vs survivors), age, education level, residence area (urban vs rural), marital status, and household monthly income. Model 2 consisted only of those independent variables that were statistically significantly associated with EQ-VAS/utility scores (identified via a backward elimination approach). Model goodness-of-fit was compared using Akaike and Bayesian information criteria (AIC and BIC).

The online survey was promoted via the communication channels of eight different stakeholders and remained open from September 3 to December 3, 2019. During this period, 412 individuals clicked the survey link, 333 (81%) consented, and 230 (69%) completed the questionnaire; 21 were removed because they were non-breast cancer patients/survivors, leaving 209 respondents. In the hospital-based survey, 106 BC patients/survivors were approached, 101 (95%) consented, and 100 (99%) completed interviews. Thus, the combined dataset of breast cancer patients/survivors consisted of 309 observations (Fig. 1).

Table 1 presents the sociodemographic characteristics of participants. There were no significant differences in terms of age, marital status, and occupation between BC patients and survivors. A significantly greater proportion of survivors attained a higher education level/monthly income and lived in urban areas. Compared with national data [22‐24], the combined study sample contained a higher proportion of individuals who lived in urban areas and spent more years in formal education.

Regarding mobility, self-care, usual activities, and anxiety/depression, a significantly greater proportion of current patients reported problems compared to survivors (Table 2). Pain/discomfort was the only dimension for which there was no statistically significant difference between patients and survivors. Overall, pain/discomfort and anxiety/depression were the two dimensions in which the highest proportion of respondents reported having any problems (75 and 57%, respectively). Self-care was the least affected dimension—only 17.6% respondents reported problems in this dimension.

306/309 BC patients/survivors were age-matched with their peers in the general population using the raw dataset of the EQ-5D standardised valuation study [16] conducted in 2017 in Vietnam (which consisted of a national representative sample of 613 women aged 18 +). Figure 2 shows the results regarding EQ-5D health profile after the matching. A significantly greater proportion of BC patients/survivors reported problems in all five EQ-5D dimensions (Chi-square tests, p < 0.001, test results are not shown in the figure) compared to women from the general population. A clear and consistent group hierarchy is evident in terms of experience of problems across domains.

The mean (SD) EQ-VAS and utility scores of BC patients were 64.9 (20.1) and 0.74 (0.22) which were significantly lower than BC survivors, 76.2 (15.7) and 0.84 (0.15), respectively (Mann Whitney U test, Z = − 4.2, p < 0.001) (Fig. 3, test results are not shown). These scores were lower than age-matched women from the general population, 77.9 (15.0) and 0.91 (0.1), respectively.

Table 3 shows BC patients’ mean EQ-VAS and utility scores by stage of treatment. Although EQ-VAS scores and utility scores showed clear differences among those at different stages of treatment, the difference was not statistically significant (Kruskal–Wallis tests). Lowest scores were reported when the patients have just finished mastectomy (mean EQ-VAS score = 54, mean utility score = 0.68). The next stages with low scores included chemotherapy, radiotherapy, and targeted therapy with scores ranged from 61.8 to 70.1 for EQ-VAS and 0.71 to 0.80 for utility. In most cases, scores of those in active treatment were lower than those of the survivors except in the cases of lumpectomy and breast reconstruction surgery where patients had higher EQ-VAS and utility scores than survivors (EQ-VAS: 90.5 and 80.0 vs 76.2; Utility score: 0.88 and 0.92 vs 0.84, respectively).

In the Tobit models, compared with BC patients, the survivors reported a 9.3 point higher score EQ-VAS (β = 9.5, p < 0.001; Model 1: treatment status was adjusted for age group, education level, household monthly income, residence area, and marital status) and 0.07 higher utility score (β = 0.07, p < 0.001; Model 2: treatment status was adjusted for age group, education level, and household monthly income) (Table 4).

Pain/discomfort and anxiety/depression were the dimensions of HRQoL where BC patients/survivors were more likely to report problems (75.4 and 57.3%, respectively)—these problems were 2–2.5 times more prevalent than any problems on other dimensions. In fact, these were also the two dimensions where severe or extreme problems were most likely to be reported by BC patients and survivors. The dimension where problems were least likely to be reported by those with BC was self-care (17.5%). These results are similar to findings from other research such as studies in China (n = 2626 BC patients), Korea (n = 827 BC patients), and Malaysia (n = 150 BC survivors) [19, 20, 25].

Pain/discomfort presents even years after the diagnosis and treatment as BC patients and survivors reported no difference in the level of problems in this dimension (79.2 and 73.5%, respectively). It also indicates that self-reported pain may not decrease for BC survivors after finishing treatment or may recur at given time points. In all other dimensions (mobility, self-care, usual activities, and anxiety/depression), the survivors had significantly better health status though the extent of the difference varied. The dimension with least difference between patients and survivors was usual activities (8%). The biggest difference between patients and survivors with respect to having problems lay on the dimensions of self-care and anxiety/depression (18%).

Compared to age-matched peers in the general population, BC patients and survivors were significantly more likely to have problems in every dimension, especially in relation to anxiety/depression. The difference between patients and general population in this dimension was approximately 46 percentage points whilst the difference between survivors and general population was 27 percentage points. Although the differences are larger than a similar study in Malaysia (compared survivors and general population), the trend is the same [25]. These results indicate the potentially profound psychological impact of BC on those who experienced the disease and imply the importance of not only care for physical health but also mental health of BC patients/survivors.

The mean (SD) EQ-VAS and utility scores of BC patients were 64.9 (20.1) and 0.74 (0.22) which were significantly lower than that of BC survivors at 76.2 (15.7) and 0.84 (0.15), respectively. EQ-VAS and utility scores of both BC patients and survivors were also lower than that of age-matched women from the general population at 77.9 (15.0) and 0.91 (0.1), respectively, which speaks to the face validity of the findings. The pattern is clear and consistent as patients have the lowest HRQoL and even when they survived and recovered from cancer, their HRQoL is still lower than age-matched women drawn from the general population. Both cross-sectional studies with matched women from general population [25, 26] and longitudinal study (using other measurement) [27] reported the same pattern of HRQoL. Compared with similar studies that used EQ-5D-5L, mean utility scores of BC patients in our study is quite similar with China and Malaysia but much lower than Korea (0.75 vs 0.78, 0.71, and 0.92, respectively) [19, 20, 25].

Apart from status of treatment (patients vs survivors), those who are younger, have higher education and household monthly income are more likely to report higher HRQoL. These results are similar to findings from previous studies, including studies that used different instruments for measurement [27‐31] and are again suggestive of face validity. Other sociodemographic characteristics such as marital status and living residence area (urban/rural) are not significantly related to either EQ-VAS or utility scores.

Among BC patients, the observed trend in HRQoL amongst different stage of treatment was clear and consistent for both EQ-VAS and utility scores. Lowest HRQoL were reported by patients who have just gone through mastectomy (EQ-VAS = 54, utility score = 0.68). Patients who were being treated with chemotherapy, radiotherapy, or targeted therapy reported the next three lowest HRQoL (respectively, EQ-VAS = 61.8, 70.1, and 69.6; utility score = 0.71, 0.77, and 0.80). Patients who had just completed lumpectomy or breast reconstruction surgery reported better HRQoL than the survivors and on par or even higher than general population (in order of lumpectomy, breast reconstruction surgery, survivors, general population: 90.5, 80.0, 76.2, 77.9 for EQ-VAS; 0.88, 0.92, 0.84, 0.91 for utility score). The results were similar with a recent qualitative study in which Vietnamese BC patients claimed that mastectomy and chemotherapy were the two biggest challenges during their treatment (manuscript under minor revision). Participants in this study whilst expressing the hope to regain their body image with breast reconstruction surgery also shed light on the impact of mastectomy. One said “I only know how bad it is after having a mastectomy… I feel so terrible. To be honest, whenever my husband holds me, I feel the emptiness… I hope that I can have reconstructive surgery because I cannot bear the defective body like that. I want my body to be normal again… Even if I die, I still want to die with a normal body… It is worse than losing my arm. Breast is a special symbol of women's beauty and every woman wants to be beautiful”. Although the results were not statistically significant due to small sample size, it is consistent with previous findings of original studies and systematic reviews with meta-analysis [29, 30, 32‐35].

This is the first study to report EQ-5D-5L data for BC patients and survivors in Vietnam. The use of the standardised EQ-5D-5L instrument also enabled the first comparative study of HRQoL of BC patients, survivors, and age-matched women from the country’s general population. It is important to note that the study was not informed by a formal power calculation though the sample size is comparable to similar studies in Vietnam and elsewhere [10]. Despite the small sample size, the study is able to provide novel and valuable insights into the HRQoL of BC patients and survivors in Vietnam, including its determinants as well as comparison with the general population. The results should be treated with some caution however as they may not generalize to the whole country as the study could not conduct a hospital-based survey in the South region (due to limited resources) which led to the under-representation of this region in the sample. In addition, it is important to note that our sample was dominated by respondents who lived in urban areas, were more educated, and had higher household monthly income. Those mostly came from the online survey (See Supplementary file 3 for sociodemographic characteristics of samples from online and hospital-based surveys). Thus, we acknowledge that there may exist differences between those who completed the online survey compared to those who completed the hospital-based survey related to education, rurality, and age that could impact the reported HRQoL.