The impact of health symptoms on health-related quality of life in early-stage breast cancer survivors

First, the respondent characteristics (age, year of diagnosis, type of surgery, stage of disease, and type of hospital) were compared to non-respondent characteristics to assess generalizability (χ², level of significance p < 0.05). Respondent characteristics, health symptoms, and global health and functioning were reported. Second, multivariable linear regression analyses were performed to determine the effect of health symptoms on global health and functioning. Variables were selected based on significance; levels of significance of 0.10 and 0.05 were applicable for univariable and multivariable analyses, respectively. Backward selection was applied to reach parsimonious models. We corrected for time since diagnosis [6, 39], age [6, 27, 39], presence of comorbid disease(s) [6, 27], and the level of education [6, 27]; these variables were selected based on the literature and availability in our dataset. We also corrected for breast reconstruction, as we expected that this treatment modality had an independent and positive effect on HRQoL [40, 41]. We did not correct for other treatment modalities.

HRQoL scores for the respondents and norm population were compared through T testing. Cohorts were matched 1:1 based on age (categories: < 50, 50–59, 60–69, 70+). Furthermore, scores were stratified by time since diagnosis (> 2, 2–4, 4+ years) and tested through one-way ANOVA. For both, a level of significance of p < 0.05 was practised.

All analyses were performed in STATA SE14.2 [42].

Figure 1 reports the mean global health and functioning for the respondents and reference population. The mean global health score for respondents (76.3 ± 17.2) did not differ significantly from the general population score (75.6 ± 16.9). Although mean scores for all individual functioning domains were 80 or higher, these scores were significantly lower than those in the general population for role (80.3 ± 23.6 vs. 84.3 ± 23.6; p = 0.016), emotional (82.9 ± 19.9 vs. 85.4 ± 16.9; p = 0.053), cognitive (80.6 ± :21.9 vs. 90.0 ± 15.9; p < 0.001), and social (85.6 ± 21.6 vs. 91.8 ± 17.7; p < 0.001) functioning. Excluding cancer patients from the general population sample did not alter our findings.

Specifically in younger patients (age < 50 years), physical functioning was significantly better (5 points difference), while cognitive (11 points) and social (9 points) were significantly worse than in older patients (results not shown). When stratified by years since diagnosis, the mean HRQoL scores did not statistically differ over time (Supplementary Table 2).

Table 3 presents the associations between prevalent health symptoms and functioning, adjusted for covariates and after backward selection. A β of -x indicated a negative association between the health problem and global health or functioning of x points. All included domains were significantly negatively associated with either one of the categories of health symptoms. Cognitive functioning was affected by a range of health symptoms, including cardiac (β: − 4.4, p = 0.002), gastrointestinal (β: − 4.3, p = 0.042), renal and urinary (β: − 9.7, p = 0.006), central nervous system (β: − 16.2, p < 0.001), and psychological (β: − 5.2, p = 0.017) health symptoms. Furthermore, health symptoms burdened different aspects of quality of life. For instance, musculoskeletal health symptoms affected global health status (β: − 4.2, p = <0.028), and physical (β: − 7.6, p = <0.001), role (β: − 9.5, p < 0.001), emotional (β: − 5.0, p = 0.019), and social (β: − 6.3, p = 0.016) functioning. Last, associations were of different magnitudes. The largest effects were found for the associations between health symptoms of the central nervous system and cognitive functioning (β: − 16.2, p < 0.001), and fatigue and role functioning (β: − 14.1, p = <0.001).

Supplementary Table 5 includes all covariates included in the multivariate analyses. Time since diagnosis was not significantly associated with functioning. Higher age was associated with significantly lower physical functioning (60–69: β: − 6.8, p = 0.002; 70+ : β: − 13.5, p < 0.001) and better cognitive functioning (60–69: β: 7.4, p = 0.014; 70+ : β: − 7.8, p = 0.018). Presence of comorbid disease was negatively associated with global health (β: − 5.5, p = 0.001), physical functioning (β: − 6.1, p < 0.001), and role functioning (β: − 8.5, p < 0.001). Immediate breast reconstruction was associated with a higher global health (β: 7.7, p = 0.005) and emotional functioning (β: 8.4, p = 0.007).

Several forms of bias may apply to patient-reported data, including recall bias, report bias, selection bias, and survivorship bias. As a result of recall bias, the reported symptoms may have been underrepresented when patients were not able to recall all symptoms they encountered in the previous year. Although recall bias was not applicable for the assessment of HRQoL at time of survey, response shift may lead to different interpretations of HRQoL over time [48]. Also, symptoms may have been either overestimated if only patients with many health symptoms participated, or underestimated by reluctance towards reporting for instance sexuality problems [38]. Furthermore, we excluded patients who could not read or write Dutch, thereby risking to exclude a vulnerable part of the patient population: non-responding patients in observational patient-reported studies have different sociodemographic and clinical characteristics, and may have systemically lower HRQoL scores [49]. As patients with less favourable sociodemographic characteristics may have been exposed to other risk factors affecting both breast cancer risk and HRQoL, information about healthy behaviour would be beneficial in further interpreting the results. As a result of survivorship bias, patients with relatively favourable disease characteristics may have been included.

We were able to confirm that a higher symptom burden was associated with a lower HRQoL [16, 17]. Therefore, it may be most effective to detect and manage or unburden the treatment-related health symptoms that were significantly associated with a lower HRQoL, and thus provide follow-up care that is valuable to the daily functioning of breast cancer patients. For instance, guidelines provides recommendations for physical therapy for the effects of musculoskeletal health symptoms as arm/shoulder function and lymphoedema [18, 19]. However, such clear recommendations are not available for all reported negative associations. Fatigue was prevalent in 9–100% of cancer survivors [9, 15, 17, 26, 27, 44] and has a large impact on HRQoL and functioning [9, 15, 17, 27, 32]. Exercise, cognitive-behavioural therapy, and education about coping were found to be effective against cancer-related fatigue [10, 50, 51], however, fatigue was found to be a very persistent problem, lasting up to 10 years in one-fourth to one-third of breast cancer patients [15]. Moreover, of the 63% of respondents who reported fatigue, only a third reported this to her physician [26], suggesting underdetection. Furthermore, a large negative association was found between health symptoms of the central nervous system and cognitive functioning, probably caused by the large proportion of patients (42.6%) reporting memory and concentration symptoms. Even though cognitive dysfunction is frequently reported among (breast) cancer patients, not much is known about its aetiology, and proven effective interventions are lacking [10, 15, 52]. Last, we found negative associations between gastrointestinal problems and global health, and physical, role, and cognitive functioning. In literature, gastrointestinal symptoms as diarrhoea, nausea, appetite loss, and constipation were reported less frequently over time, by 15% to 20% of patients [44], which we could confirm [26]. Gastric or abdominal complaints were reported in 13% of patients. However, over 50% of them reported healthcare use for this symptom, which was nearly the highest reported use of care for the range of included symptoms [26]. Literature reported that the majority of gastro-complaints declined over time [11, 44], still, the high use of care could indicate many patients struggle with these symptoms long after diagnosis.

These examples illustrate not only the importance of detection and acknowledgement of health symptoms, but also the importance of research towards successful detection and intervention for these symptoms [53], so that guidelines can provide guidance in dealing with them. Survivorship research is a priority research area for breast cancer, including follow-up tools to assess quality of life in long-term survivors [23].

We measured symptoms through the SaP questionnaire [36], rather than through the EORTC-QLQ-C30 symptom domains. The SaP assesses a broader range of non-specific symptoms, aiming to include symptoms from all organ systems. We hypothesized that a broader selection of symptoms than currently included in the QLQ-C30 is of relevance for longer-term survivors [27]. Our results confirm this hypothesis. For instance, menopausal (40%) and memory and concentration complaints (43%) were commonly reported, but are not or limitedly included in the QLQ-C30 and BR-23 breast cancer module. This hypothesis was confirmed by Van Leeuwen et al. [11]. We endorse Van Leeuwen et al. that development of a survey specifically for cancer survivors would address a gap in assessing issues of a more chronic nature [11].