Identification of relevant literature

Literature search results

A total of 169 records were initially identified from the databases. After removing the duplicates, 102 articles remained. After title and abstract screening, 41 articles were retained for full-text review. Ultimately, 13 studies met the inclusion criteria and were included in the final analysis [12–24]. The study selection process is illustrated in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow chart for literature search and screening.

https://doi.org/10.1371/journal.pone.0336793.g001

Basic characteristics of the included studies

The included studies were published between 2023 and 2025 and conducted in China (n = 12) and South Korea (n = 1). All studies were quantitative in nature, with 12 cross-sectional studies [12–15,17–24] and one longitudinal study [16]. The treatment modalities included chemotherapy, radiotherapy, endocrine therapy, and combined therapies. A total of 5,633 adult female participants (aged ≥18 years) were included in the study. The detailed characteristics are presented in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Basic Characteristics of the Included Studies (n = 13).

https://doi.org/10.1371/journal.pone.0336793.t001

Symptom assessment tools used in the included studies

Thirteen symptom assessment tools were used across the 13 included studies, comprising four single-symptom scales and nine multi-symptom instruments. The most commonly used multi-symptom tools were the M.D. Anderson Symptom Inventory (MDASI, n = 2), Chinese version of the MDASI (MDASI-C, n = 2), Functional Assessment of Cancer Therapy–Breast (FACT-B, n = 2), Chinese version of the Breast Cancer Prevention Trial Symptom Scale (C-BCPT, n = 2), and European Organization for Research and Treatment of Cancer Quality of Life Questionnaire–Core 30 (EORTC QLQ-C30, n = 2). Four studies [12,16,17,20] used two or more instruments in combination; however, each used a unique combination of scales.

Characteristics of symptom networks in included studies

The symptom networks constructed in the included studies encompassed physiological, psychological, and social dimensions. All studies utilized R software for network construction. The key node metrics analyzed included strength centrality, closeness centrality, betweenness centrality, and expected influence. As there is currently no universally accepted standard for determining the minimum sample size in symptom network models, the sample sizes reported in the literature included in this article exhibited considerable variability. Notably, the study conducted by Zha et al [14]. had the smallest sample size, which was determined based on a criterion of “5-6 individuals corresponding to each symptom node,” ultimately including 168 patients. The findings indicate that the symptom network constructed from this sample size demonstrated good stability and accuracy. Among the included studies, 12 constructed cross-sectional symptom networks [12–15,17–24], whereas one developed a longitudinal (dynamic) symptom network [16].

Core and bridge symptoms identified in included studies

A total of 10 core symptoms were identified across the studies included, with fatigue being the most frequently reported (n = 7) [14–17,19,22,23], followed by difficulty in concentrating (n = 2) [13,21]. Ten bridge symptoms were also extracted, primarily fatigue (n = 5) [12,13,15,16,22] and poor appetite/anorexia/decreased appetite (n = 3) [12,15,19]. Four studies [17,20,21,24] conducted subgroup analyses and identified core symptoms within each subgroup. The core symptoms in the subgroups mainly included pain (n = 3) [17,20,21] and fatigue (n = 2) [17,21]. Among the studies included in this review, there were variations in the methods used to identify core symptoms and bridge symptoms. For core symptom identification, different studies employed distinct centrality indices. Four studies [13,14,18,21] solely relied on strength as the criterion for identifying core symptoms, whereas three [16,19,24] adopted a composite approach using strength, closeness, and betweenness. In the identification of bridge symptoms, four studies [14–16,22] used betweenness centrality as the primary indicator, whereas three used indicators such as the Expected Influence, Coefficient of Bridges, Bridge Strength, and Bridge Compactness for identification [12,13,19]. The details are presented in Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Core symptoms, bridging symptoms and research methods covered in the literature (n = 13).

https://doi.org/10.1371/journal.pone.0336793.t002

Symptom networks in breast cancer patients are complex, fatigue may serve as a key target for intervention

Symptoms such as fatigue, sleep disturbance, decreased appetite, nausea, and anguish are highly prevalent and often severe among patients with breast cancer [12–24], suggesting a widespread burden of symptom distress that warrants comprehensive assessment and management. Although the symptoms experienced by breast cancer patients may differ depending on the treatment methods employed, numerous studies have consistently identified that fatigue ranks highest in terms of incidence or severity when utilizing various symptom assessment tools [12–15,22,23]. Fatigue has also been frequently identified as both a core symptom [14–17,19,21–23] and a bridge symptom [12,13,15,16,22] within symptom networks.

Given its centrality and connectivity in the symptom network, fatigue should be prioritized as a key intervention target in the symptom management of patients with breast cancer. Non-pharmacological interventions are currently recommended as first-line strategies for fatigue and include exercise therapy, psychological support, cognitive behavioral therapy, and nutritional interventions [25].

Symptom networks in breast cancer patients are dynamic: The need to strengthen research on longitudinal networks

In a longitudinal study by He et al. [16] it was demonstrated that the symptom network structure in patients with breast cancer undergoing chemotherapy changes dynamically over time, with variations in both core and bridge symptoms at different treatment stages. Prior to chemotherapy, the median number of reported symptoms was four, with sleep difficulty as the core symptom and depressed mood as the bridge symptom. After the first chemotherapy cycle, the median number of symptoms increased to 11, with appetite loss and pain emerging as core symptoms and headache as a bridge symptom. By the third and sixth chemotherapy cycles, the median symptom count further increased to 12 and 13, respectively, with fatigue consistently identified as both a core and bridge symptom. These findings suggest that symptom networks evolve throughout the treatment trajectory and highlight the importance of stage-specific dynamic symptom management strategies.

However, most studies included in this review were cross-sectional in nature [12–15,17–24] and constructed symptom networks based on data collected at a single time point, referred to as cross-sectional or contemporaneous symptom networks [26]. While useful, these networks cannot capture symptom trajectories or infer the causality between symptoms. In contrast, dynamic symptom networks built from repeated measures across multiple time points in the same cohort can reflect temporal shifts in symptom associations and reveal their underlying mechanisms [27].

In summary, future research should prioritize longitudinal designs to construct dynamic symptom networks. This approach enables a better understanding of symptom evolution and interrelationships over time, ultimately providing personalized and temporally adaptive symptom management in breast cancer care.

Lack of specificity in symptom assessment tools among breast cancer patients: A need for standardization

The symptom assessment tools used in symptom network studies of breast cancer patients vary widely, encompassing both single-symptom scales and multi-symptom instruments. In most cases, researchers independently select instruments based on study objectives, resulting in inconsistencies in applicability and standardization. The current tools are primarily designed for general symptom assessment in patients with breast cancer.

For instance, in studies involving patients undergoing endocrine therapy, He et al. [13] used the C-BCPT and identified “difficulty concentrating” as the core symptom, while Jing et al. [18], using the FACT-ES, found “emotional lability” and “irritability” to be core symptoms. Liang et al. [19] applied the MDASI, identifying “fatigue” as the core symptom, whereas Chang et al. [20], using four separate single-symptom tools (NRS, PSQI, CFS, HADS), identified “panic” as the core symptom. These discrepancies indicate that variations in assessment tools may result in significant differences in the structure and interpretation of symptom networks, even among similar patient populations. Such inconsistencies are not conducive to clinical practice, particularly when formulating effective symptom management strategies based on these assessments.

Therefore, the development of breast cancer–specific, sensitive, and standardized symptom assessment instruments is urgently required. A unified approach to tool selection and application will enhance the comparability and accuracy of symptom network analyses, and ultimately improve symptom management strategies in both research and clinical settings.

Combining latent variable analysis with network analysis enhances understanding of symptom network heterogeneity across subgroups

Given the substantial individual variability in symptom expression among patients with breast cancer, identifying heterogeneous subgroups based on symptom profiles may be the key to optimizing symptom management strategies [28]. Latent variable analysis enables the identification of unobserved subpopulations within a larger cohort that share similar observable characteristics, but differ in their underlying symptom structures [29], offering valuable insights into symptom heterogeneity.

Several studies included in this review employed integrated approaches that combine latent variable modeling, specifically latent class analysis (LCA) or latent profile analysis (LPA), with network analysis. In these studies, LCA or LPA was used to classify patients into symptom-based subgroups, followed by the construction of separate symptom networks for each subgroup. For example, Cai et al. [17] utilized LCA in combination with network analysis and found that core symptoms varied significantly across subgroups of patients undergoing chemotherapy. Similarly, He et al. [21] and Teng et al. [24] applied LPA and observed differences in core symptoms between subgroups. It is evident that the symptom networks of subgroups exhibit greater heterogeneity. Future research should aim to enhance both the depth and breadth of latent variable analysis combined with network analysis in studying symptom networks among breast cancer patients. This integrated analytical approach can be employed to more accurately delineate the classification criteria for symptom subgroups in breast cancer patients with varying clinical characteristics, thereby providing methodological support for the refined identification of these subgroups. Furthermore, the subgroup-specific characteristics of symptom networks revealed through this method can offer a scientific foundation for developing targeted intervention strategies.

Specifically, differentiated intervention strategies should be formulated based on the unique symptom profiles of various subgroups. The core symptoms identified within each subgroup ought to serve as primary targets for intervention, prioritizing those key symptoms that exert significant influence on patients’ overall symptom burden and are likely to trigger the transmission and exacerbation of other symptoms. Concurrently, by leveraging the connection characteristics associated with bridge symptoms within these subgroups, it is essential to regulate the strength of associations between bridge symptoms and other related symptoms to disrupt cyclical interactions among them, ultimately leading to a reduction in the overall symptom burden experienced by patients within these subgroups.

Pain should be prioritized as the primary intervention target for patients experiencing a significant symptom burden in breast cancer

Chang et al [20] and He et al [21] demonstrated that, following the sub-group classification of symptom burden in breast cancer patients, pain emerges as the core symptom within the high symptom burden group. This finding indicates that pain in breast cancer patients does not occur in isolation; rather, it is significantly intertwined with other distressing symptoms such as fatigue, sleep disturbances, and emotional abnormalities. Pain serves as a “core node” linking these symptoms and acts as a critical marker for distinguishing between high distress groups and low/moderate distress groups. Consequently, if pain is not effectively managed, alleviating other symptoms will be considerably hindered, making it challenging to reduce the overall symptom burden experienced by patients. This underscores the importance of targeting pain management as a key intervention strategy for improving outcomes among breast cancer patients with high symptom distress. A comprehensive and accurate assessment represents both an essential initial step and a fundamental guarantee for effective cancer pain management [30]. Medical professionals should perform thorough evaluations regarding the type, severity, and duration of patients’ pain experiences. In addition to standardized pharmacological treatments, non-pharmacological interventions—such as music therapy and mindfulness practices—should also be integrated into care strategies [31]. Furthermore, traditional Chinese medicine approaches like acupuncture and moxibustion hold significant application value in this context. The study conducted by Li et al. [32] validated the efficacy of these methods in managing pain among liver cancer patients; thus, providing valuable insights for implementing similar interventions aimed at addressing pain in breast cancer populations.