Longitudinal changes of health-related quality of life over 10 years in breast cancer patients treated with radiotherapy following breast-conserving surgery

HRQoL was prospectively assessed using the EORTC QLQ-C30 version 3.0 [35]. It is composed of 30 items: global health status/quality of life (GHS/QoL), five functional domains (physical, role, cognitive, emotional and social functioning), three combined symptom domains (pain, fatigue and nausea/vomiting) and six single symptom items (dyspnea, insomnia/sleep problems, appetite loss, constipation, diarrhea and financial difficulties). In accordance with the EORTC QLQ-C30 scoring manual, raw scores were calculated as arithmetic means if at least half of the items for the scale were answered, then linearly transformed into 0 to 100 scale [36]. Higher scores indicate a better functioning or a better GHS/QoL, whereas higher scores on the symptom scale represent a heavier burden of symptoms. Based on the original study design with focus on radiation toxicity, the HRQoL data was collected at baseline (pre-RT, after breast-conserving surgery), during RT, end of RT, 6 weeks after RT and at the 10-year follow-up.

Linear-mixed models (LMM) were used to evaluate differences between selected time points while accounting for the correlation of multiple measurements within a subject (i.e., random intercept for subject). For this part of the analysis, all patients with HRQoL data at t10 were included as part of the regression models. The time points were included as a categorical variable to test mean HRQoL score differences between the time points t1, t5 and t10. All models were adjusted for age at baseline. In addition to LMM, linear regression models were used to evaluate the specific effect of baseline age on mean score differences between 6 weeks after RT (t5) and 10 years after (t10) for those HRQoL domains with significant differences at t10 vs. t5. Wilcoxon rank-sum test was used to assess differences between patients and controls stratified by age groups (< 65, 65– < 75, ≥ 75 years) at the 10-year follow-up. For the last objective, multivariable linear regression models were obtained to identify determinants for HRQoL domains at the 10-year follow-up in BC survivors. Potentially relevant factors including age (< 65, 65– < 75, > 75), BMI (≥ 25 vs. < 25 kg/m²), living with others, smoking status (never, former, current smokers), endocrine therapy (tamoxifen, aromatase inhibitors), recurrence status (yes/no) and selected comorbidities (depression, hypertension, stroke and diabetes) and symptoms (pain/lymphedema and fibrosis) assessed at follow-up were analyzed in these models.

Figure 1 shows the changes of GHS/QoL and functional performances (Fig. 1a) as well as the symptom domains (Fig. 1b) from baseline (pre-RT, t1) through 6 weeks after RT (t5) until the 10-year follow-up (t10) among BC survivors. When evaluating changes between selected time points (Table 2), age-adjusted scores at baseline (t1), 6 weeks after RT (t5) and the 10-year follow-up (t10) were compared. For most functional domains, there was an overall decrease in functional scores during the acute treatment period (t2 to t4) after baseline (t1), with a recovery at 6 weeks after RT (t5), which was followed by deteriorations (e.g., physical functioning) or stable scores at 10 years post-diagnosis (t10) for most functional domains except for role functioning. A significant improvement was observed between t1 and t5 for GHS/QoL, emotional and social functioning. Role functioning scores significantly improved by 7.4 points during the 10-year follow-up compared to post-RT scores (77.0 vs. 69.6). On the other hand, for GHS/QoL, physical and cognitive functioning, mean scores were significantly lower at t10 compared to t5, with mean score differences of 3.2, 9.1, and 3.6 points, respectively. When comparing baseline (t1) and the 10-year follow-up (t10), a significant improvement was observed for role, emotional and social functioning, whereas physical and cognitive functioning mean scores were significantly lower at t10 compared to t1.

Regarding the symptom domains, dyspnea scores increased by 10.6 points from 11.7 at t1 to 22.3 at t10. Pain scores were relatively stable between 29.5 at t1 and 28.7 at t10, respectively. Fatigue had a peak during RT (41.0 at t2) but returned to baseline level (32.4 at t10). Insomnia scores increased and remained relatively stable after baseline with mean scores of 44.2 at t10. Appetite loss, nausea/vomiting, constipation and diarrhea scores remained relatively stable at low levels of up to 15.0. Financial burden significantly improved from t5 to t10, at low levels though.

The specific effect of age at baseline and its relationship with previously observed significant differences between t10 and t5 was also evaluated. For GHS/QoL, physical and role functioning, baseline age was associated with worsening in functional scores between t5 and t10, with statistically significant differences in mean scores between time points after the age adjustment. For diarrhea and financial difficulty, baseline age was associated with an increase in mean scores between t5 and t10. Similar to the functional domains, mean differences were also still significant between time points after adjustment for baseline age. For cognitive functioning and dyspnea, age was not significantly associated with mean differences between t5 and t10.

Patients with recurrences/metastasis had generally lower GHS /QoL, lower functional and higher symptoms scores than patients without disease progression (Fig. S2).

Sleep problems, fatigue, pain and dyspnea were the most prevalent symptoms in BC survivors at the 10-year follow-up (Fig. 2). About a third to almost half of patients reported moderate to severe fatigue and sleep problems, respectively.

The characteristics of the patient cohort as well as the control population at the 10-year follow-up were summarized in Table 1. No significant differences in GHS/QoL in BC survivors and the control population were observed at the 10-year follow-up for the different age groups. However, deficits in physical and emotional functioning and in certain symptoms such as dyspnea, fatigue, sleep problems and constipation were more prevalent in BC survivors than controls (Fig. 3; Table S1). Detriments in functional performance and symptom burden occurred predominantly in younger survivors. Compared to controls of comparable age, survivors under 65 years of age reported significantly worse physical (81.6 vs. 87.1), role (78.3 vs 86.2) and emotional functioning (63.1 vs. 74.5), as well as higher burden of dyspnea (25.6 vs. 18.0), fatigue (35.5 vs. 25.0) and sleep problems (50.7 vs. 32.0). The prevalence of psychosocial dysfunction and symptoms among younger survivors showed that younger patients were at greater risk of lower HRQoL years after RT. Patients aged ≥ 75 years reported similar functional performance and symptom burden compared to the control population except for a deficit in emotional functioning (difference of 5.5 points). Middle aged patients reported worse physical (2.5 points) and emotional functioning (5.6 points), and more severe fatigue and sleep problems compared to controls (difference of 2.6 and 3.6 points, respectively). Besides, gastrointestinal problems mainly affected the middle-age group.

Results from multivariable linear regression (Table S2) showed that a poorer GHS/QoL and physical functioning was associated with being overweight. Living with others was associated with better GHS/QoL and physical functioning, lower levels of dyspnea and pain at the long-term follow-up, while patients above 65 years had decreased levels of dyspnea. Depression was found to be associated with lower GHS/QoL, worse functioning, and a higher level of pain, fatigue, insomnia, gastrointestinal problems, and financial difficulty. Besides, hypertension was associated with more severe dyspnea and fatigue as well as worse physical and role functioning. Other side effects such as pain/lymphedema and fibrosis were associated with worse physical and social functioning, respectively.

Few studies have examined the change in HRQoL scores in BC survivors [4, 10‐14, 16, 17, 19] as in this study with repeated measurements not only during and 6 weeks after post-operative RT but also with measurements up to 10 years after treatment. Previous studies analyzing HRQoL of BC patients during and after RT [37, 38] did not find significantly worsened GHS/QoL in BC patients although fatigue and most symptoms increased during RT, which might be related to side effects of treatment. On the whole, GHS/QoL scores at 5–10 years of follow-up have been reported to be comparable to baseline level [4, 12, 16, 17, 19, 39]. Therefore, our findings based on longitudinal data even after > 10-year follow-up are in line with previously reported results.

At 10-year follow-up, the BC survivors in our study had better role, emotional and social functional performances compared to baseline, but a worse physical and cognitive functioning. When compared to 6 weeks after RT, they showed only improvement in role functioning and restrictions in physical and cognitive functioning. Although increasing age was found to be significantly correlated with worsening of both functional and symptom domains between t5 and t10, mean differences were still significant after age adjustment, suggesting that, although aging plays an important role in long-term HRQoL differences, it is not sufficient to explain improvement or worsening of HRQoL scores over time. Previous studies also found that BC survivors often gradually recovered from most functional restrictions in the long term [4, 12, 19, 40]. Nonetheless, on the symptom domains, our results confirmed that fatigue and sleep problems may persist for years after treatment [41‐44]. The level of pain after RT did not show significant differences compared to 10 years and remained stable during the treatment period, whereas a significant improvement in pain over time was reported by another study [45]. Dyspnea as a potential effect of RT was reported more often by younger BC survivors at the 10-year follow-up in our study and was also the HRQoL symptom domain with the largest deterioration from baseline to the 10-year follow-up. Several comorbidities such as hypertension rather than aging itself were associated with higher levels of dyspnea in our determinants analysis. A previous study on long-term HRQoL (10 years) also found that young BC survivors experience the greatest deterioration in multiple HRQoL domains during follow-up compared to other age groups, and among symptom domains, the greatest differences with a reference population were found for dyspnea and fatigue [17]. Similar findings have also been previously reported for dyspnea when comparing BC survivors and the German reference population [33], although other studies have reported similar late dyspnea levels between BC survivors and a reference group [44, 46]. Most of the significant differences observed when comparing functional and symptoms mean scores over time were associated with “little change” according to the minimal clinically important difference definition proposed by Osoba et al. [47] (i.e., differences of about 5 to 10 points).

To account for the impact of aging, we compared the HRQoL of the long-term BC survivors more than 10 years after diagnosis to a control population that was of comparable age but without a breast cancer diagnosis at baseline and were also followed up for about 10 years. The GHS/QoL in the long-term BC survivors was comparable to that of the controls, which is in line with several previous studies [4, 12, 13, 19]. Yet both our findings and that of others found significant restrictions in certain domains, most commonly for fatigue [18, 20, 44], sleep problems [16, 20], and emotional functioning [18, 19, 44], providing clear evidence that HRQoL may be persistently impaired in cancer patients in the long run. In particular, long-term survivors younger than 65 years at 10-year follow-up reported poorer psychosocial related HRQoL performances in our study, including significantly higher levels of fatigue and sleep problems than controls of comparable age. For emotional functioning, fatigue and sleep problems in survivors under 65 compared to controls, the observed differences can be characterized as “moderate clinically meaningful differences” (i.e., 10–20 points) according to Osaba et al. [47] and Cocks et al. [48].

Physical functioning has been reported by patients below 80 years to be significantly poorer than their age-matched reference groups, but they were comparable to that of healthy controls in the oldest age group above 80 years [16]. We observed a similar pattern in physical functioning, which could be explained by natural deterioration with aging. Deficits in role functioning were apparent in the youngest age group, suggesting that younger BC survivors might have more problems to cope with this difficult life situation and may continue to suffer more even 10 years after diagnosis. Yet detriments in emotional functioning persisted at the 10-year follow-up among our BC survivors compared to controls for all age groups. In BC survivors of the MARIE study, however, the youngest age group suffered greater detriments in emotional functioning compared to controls [19]. Our results suggest that emotional well-being might be a problem for BC survivors across a wide age range. In contrast to previous findings showing differences in financial difficulties [12, 21, 44], pain [20], and cognitive functioning [18, 19, 44], there was no significant difference between patients and controls in these domains in any of the age groups in our study. Yet dyspnea was found to be significantly more severe among the youngest patient age group compared to controls.

We confirmed a number of determinants for an impaired long-term quality of life of breast cancer survivors such as younger age [16], obesity [25], or certain comorbidities like depression [26]. In addition to RT and cancer surgery, other parameters like endocrine therapy or BC treatment-related symptoms such as lymphedema or pain were significantly associated with higher burden of certain quality of life domains. In addition, other factors like anxiety, limited mobility of shoulder joint, or factors unrelated to the BC diagnosis may also play a role for impaired long-term HRQoL.

The differences in findings among studies might be explained by different age structures of the study populations and different inclusion criteria. Our patient cohort tended to have a younger age structure and a wider age range compared to other studies, and younger survivors have considered cancer as a bigger threat and have been more affected in social aspects [23]. Furthermore, all BC patients included in this study only received adjuvant RT, while other studies included patients who had also received chemotherapy; thus, the therapy-related symptom burden could be different.

We assessed HRQoL in a unique cohort of breast cancer patients who received only radiotherapy after breast-conserving surgery but not chemotherapy. The study was focused on side effects after radiotherapy and attempted to minimize “contamination” due to chemotherapy. Although RT techniques have improved in the last two decades (e.g., intensity-modulated/image-guided RT), resulting in fewer treatment-related adverse effects due to less damage to surrounding tissues, long-term QoL data in the post-treatment period are scarce. One study from the Netherlands also limited their study population to breast cancer patients treated with RT and without chemotherapy, yet the follow-up period was only 3 years [45]. Second, our study is one of the few studies with prospective multidimensional HRQoL assessments that comprise a baseline (pre-RT) and multiple follow-up evaluations during and post-RT treatment over a period of more than 10 years. Third, we compared the long-term HRQoL of the breast cancer survivors to a group of unaffected women with comparable length of follow-up to account for an aging effect. As comparison group, we used a large control population from the same geographic region [33, 34] rather than German general population norm data [49] that has been used in other studies [4, 11, 17, 20, 33, 43, 45, 50] in order to account for regional differences and increase statistical power. Finally, unlike other studies, e.g., [19, 44], there was no age restriction in our study to capture a wider spectrum of patients and further enhance the representativeness.

The analysis has some limitations: Because of the eligibility criteria of the parent study design, the results might not be generalizable to patients who have received chemotherapy as part of their cancer treatment, acknowledging that chemotherapy is an important risk factor for impaired HRQoL. However, findings were similar between our study and studies including chemotherapy patients. The response pattern of our study might have caused an underestimation of the true difference between survivors and controls as those who had passed away or did not participate due to other reasons tended to have worse health conditions (severe symptoms or worse functional performances) which might lead to a lower HRQoL (survivorship bias). With this loss of patients, the sample size was greatly reduced, entailing a loss of statistical power when analyzing subgroups. Overestimation of the observed HRQoL could also have occurred since patients with poor health might be less likely to participate in the study at baseline and also at the follow-ups. Due to the nature of the collected data (both the BC patient cohort and the control population), the data on the baseline HRQoL measures for controls and the HRQoL data for BC patients prior to the conserving surgery or diagnosis are not available. There were 68 controls who had a previous (non-breast) tumor at recruitment and 31 who developed cancer during follow-up. We did not exclude these individuals from the analysis to better represent the general population. With this approach, we could have underestimated the differences between patients and controls. Due to the age restriction at recruitment in the MARIE study (50–74 years), the age distribution of patient cohort and control population was not balanced, and the controls were primarily postmenopausal, while there was no age restriction at recruitment in ISE study, which could reduce the generalizability in the lowest age group. Besides, for the cross-sectional determinants analysis, additional factors, in particular other comorbidities than depression, hypertension, diabetes, and stroke which we accounted for, may have a relevant impact on HRQoL.