Due to advances in treatment and care, the five-year survival rate of breast cancer has reached 85% or above (1). Of note, endocrine therapy – including tamoxifen and aromatase inhibitors – is essential for reducing recurrence and mortality rates (2). Approximately 84% of patients with breast cancer express hormone receptors. Endocrine therapy (with a standard five-year duration after breast surgery, chemotherapy, or radiation therapy) has been proven to be helpful in treating endocrine-sensitive breast cancer (3, 4). However, reports have shown that breast cancer patients who receive endocrine therapy often experience a high level of symptom burden (5, 6). These symptoms inversely can influence drug compliance, anatomical status, and quality of life (7, 8).

Endocrine therapy-related symptoms (ESs) are defined as treatment-related side effects experienced by women receiving endocrine therapy that negatively affect health-related quality of life and adherence to therapy (9). Previous studies (3, 5, 9–11) have reported that ESs mainly include gynecologic symptoms (e.g., vaginal discharge, sexual dysfunction, and vaginal dryness), menopausal symptoms (e.g., hot flashes, night sweats, cold sweats, mood swings, irritability), and musculoskeletal symptoms (e.g., joint pain, bone pain). Among these symptoms, hot flashes were the most common side effects regardless of the treatment received (9). Other frequently reported side effects, such as vaginal discharge, vaginal dryness, dyspareunia, and arthralgia, vary in prevalence between different types of medicines (9).

According to previous studies (11, 12), patients undergoing endocrine therapy usually suffer from more than 10 cooccurring symptoms. A considerable amount of work has been performed to investigate cooccurring symptoms (5, 13, 14), symptom clusters (15–17) or some prevalent individual symptoms (18–21) for ESs. Understanding how symptoms interact with each other will be helpful for preventing the occurrence of related symptoms. Recent advances in network analysis provide a new way to gain insight into the complex nature of comorbid symptoms and clusters of symptoms and to identify core symptoms. The paradigm in symptom science called “symptom networks” has been used to explore the intricate connections between symptoms linked to chronic illnesses and psychopathology (22–25).

Network analysis could be used to construct a partial correlation model of the relationship between observed variables and then visualize the importance of each variable in the network and its complex association from the overall perspective in the form of a graph (26, 27). The partial correlation network model is based on the weighted correlation network, which accounts for the possibility that the correlation between two nodes is affected by another node and provides a way to further accurately explore the relationship between nodes (28). In recent years, the partial correlation network model has been used to study symptoms among cancer patients. For instance, de Rooij et al. (29) demonstrated that fatigue was the most common core symptom among cancer survivors, with moderate links to other symptoms such as emotional or cognitive symptoms, appetite loss, dyspnea, and pain.

Investigating core symptoms related to endocrine therapy could activate other symptoms in the network, which could be helpful for identifying targets for symptom intervention and delivering efficient symptom management (28, 30). Therefore, the main aim of the current study was to 1) explore the core symptoms of breast cancer patients with endocrine therapy and 2) assess differences in symptom networks among different demographic and clinical covariates. To our knowledge, this is the first study to explore the core symptoms among breast cancer patients undergoing endocrine therapy.

To the best of our knowledge, this is the first report to analyze the ESs network among breast cancer patients. The main finding of this study is that irritability and mood swings were the most prevalent and central symptoms. Developing interventions targeted at emotional symptoms may be essential for reducing the overall symptom burden among breast cancer patients undergoing endocrine therapy.

In the overall symptom networks, we identified five clusters: emotional symptoms (i.e., irritability and mood swings), sexual symptoms (i.e., lost interest in sex, vaginal dryness, and pain or discomfort with intercourse), vaginal symptoms (i.e., vaginal discharge and vaginal itching/irritation), vasomotor symptoms (i.e., hot flashes, night sweats, and cold sweats), and neurological symptoms (i.e., dizziness and headaches). There were a few inconsistencies with previous results. For example, Li et al. (15) identified seven clusters of musculoskeletal, vasomotor, urinary, sexual, psychological, neurocognitive, and weight symptoms across the 18-month follow-up period of aromatase inhibitor therapy using exploratory factor analyses. Wen et al. (17) identified three groups of breast cancer patients receiving aromatase inhibitors by principal component analysis, namely, the disease symptom cluster, treatment-related psychological symptom cluster, and gastrointestinal symptom cluster. The inconsistencies may be caused by the instruments and data analysis method. Li et al. (15) evaluated 47 symptoms by the Breast Cancer Prevention Trial Symptom Checklist, Patient’s Assessment of Own Functioning Inventory, Beck Depression Inventory-II, and Profile of Mood, States Tension/Anxiety and Fatigue/Inertia subscales, and Wen et al. (17) assessed 13 symptoms by the MD Anderson Symptom Inventory; however, we used the Functional Assessment of Cancer Therapy-Endocrine Subscale (ES). The content of symptoms that were included to conduct clusters was different. Moreover, previous studies identified symptom clusters usually assuming a common underlying factor, while network analysis provided a more dynamic approach with the assumption that symptoms cluster because they mutually interact (29).

The most central symptoms of breast cancer patients during endocrine therapy were emotional symptoms regardless of treatment regimens (selective estrogen receptor modulators versus aromatase inhibitors). Previous studies also revealed that breast cancer patients undergoing endocrine therapy experienced a range of emotional distress (11, 38) The widely used endocrine treatment reduces the postmenopausal estrogen levels by nearly complete inhibition of the enzyme aromatase or anti-estrogenic effects on breast cancer cells that contain estrogen receptors by blocking this receptor (3). Possibly beneficial effects of estrogens on brain function may be, among others, the result of estrogenic activity through estrogen receptors in brain regions that are important for cognitive functioning, effects on neurotransmitters, protection against ischemic damage, and increased survival of brain cells (21, 39). According to network analysis theory (40), core symptoms have the greatest impact on the other symptoms over the entire network. Regarding those patients undergoing endocrine therapy, emotional symptoms might have strong interactions with sexual symptoms, vaginal symptoms, vasomotor symptoms, and neurological symptoms. The findings were also confirmed in the network analysis of gastric cancer patients, which showed that treating psychological distress and enhancing emotional well-being can be high-impact intervention targets throughout the cancer trajectory (41). Therefore, further interventions targeting emotional symptoms, such as psychosocial support, may be the optimal strategy to reduce the overall symptom burden.

Although previous studies provided evidence that receiving aromatase inhibitors could induce musculoskeletal symptoms frequently and vaginal discharge was more frequent in selective estrogen receptor modulator therapy (2, 9, 13, 42), we did not detect any significant differences in the network structure and global strength between treatment subgroups. One reason was network approaches in which constructs were modeled in terms of interactions between their constituent factors rather than the incidence or severity (36), and the other reason might be that we could not control all the other clinical covariates in the subgroup network analyses (29). However, it is worth noting that the edge difference test suggested that compared with the selective estrogen receptor modulator group, the network of patients treated with aromatase inhibitors had increased connections among vaginal and sexual symptoms, which implied that selective estrogen receptor modulators and aromatase inhibitors potentially have distinctive effects on vaginal and sexual domains. To overcome vaginal and sexual issues, education consulting, using vaginal lubricant, and combining pelvic floor muscle relaxation exercises may be potential approaches (9, 43).

First, according to Epskamp et al. (32), in a 20-node and 15-node network, 210 parameters and 120 parameters needed to be estimated, respectively. To reliably estimate these parameters, the number of observations needed typically exceeds the number available in characteristic psychological data. Therefore, our sample size was appropriate. Nevertheless, network analysis is a data-driven approach, and generally, the larger the sample size is, the more stable the network. Therefore, it is necessary to verify the results using different algorithms in other independent data with larger samples. Second, we excluded four symptoms based on low prevalence and clinical experience, and we did not include clinical covariates in our subgroup networks because they were not consistent across treatment regimens (36), which could introduce information bias and variable selection bias. Third, we applied an online survey to collect data using convenience sampling, which attracted many younger and higher education level participants, thus potentially introducing participant selection bias. Last but not least, it was a cross-sectional study that limited causality determination among symptoms and the generalizability of the findings. Thus, it is necessary to conduct longitudinal research to develop dynamic networks in the future.

In conclusion, we found that emotional symptoms (i.e., mood swings and irritability) were frequently reported by breast cancer patients during endocrine therapy and were consistently central in the symptom networks across not adjusting and adjusting clinical covariates and treatment subgroups. It was suggested that emotional symptoms could be an important target for reducing the overall symptom burden in breast cancer patients during endocrine therapy. Although causal conclusions cannot be drawn, if the interrelatedness of these symptoms is assumed, then developing interventions targeting emotional symptoms may reduce multiple other symptoms through a negative feedback loop of other interrelated symptoms. The findings of this study need to be verified by larger samples and using different algorithms.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. Requests to access these datasets should be directed to Feng Jing, 21111170001@m.fudan.edu.cn.

The studies involving human participants were reviewed and approved by Institutional Review Boards of Fudan University School of Nursing (No.IRB#2017-05-01). The patients/participants provided their written informed consent to participate in this study.

The manuscript was written by FJ. ZZ provided statistical method guidance. JQ and LT contributed to the data collection. LX provided critical feedback. YH and WX directed the writing and revision of the paper. All authors contributed to the article and approved the submitted version.