2.1 Research questions

This scoping review was guided by the following questions: (1) What symptom assessment tools have been used in symptom network studies among patients with gastric cancer? (2) Which core symptoms and bridge symptoms have been identified in gastric cancer symptom networks? (3) What key symptom associations and temporal relationships have been reported among symptoms? (4) What factors are associated with the identification of core symptoms in gastric cancer symptom networks?

2.2 Inclusion and exclusion criteria

The inclusion criteria were developed according to the PCC framework (Participants, Concept, and Context) [9]. Studies were eligible if they met the following criteria: (1) Participants: patients diagnosed with gastric cancer, with no restrictions on age, sex, or ethnicity; (2) Concept: studies focusing on symptom network analysis in patients with gastric cancer, including the identification of core symptoms, bridge symptoms, symptom pathways, or network structures; and (3) Context: studies conducted during the perioperative period, chemotherapy period, or postoperative chemotherapy period. Given the limited evidence base, studies of mixed cancer populations were also eligible when gastric cancer patients were explicitly represented in the sample and the study examined symptom networks in a treatment context relevant to gastric cancer. Because gastric-cancer-specific networks could not be isolated in such studies, their findings were treated as indirect evidence and interpreted separately. Studies were excluded if they met any of the following criteria: (1) publications written in languages other than Chinese or English; (2) duplicate publications or studies for which the full text was unavailable; and (3) non-original research articles, including reviews, guidelines, case reports, conference abstracts, and conference proceedings.

2.3 Search strategy

A comprehensive literature search was conducted in PubMed, Web of Science, CINAHL (Cumulative Index to Nursing and Allied Health Literature), the Cochrane Library, Embase, China National Knowledge Infrastructure, Wanfang Data, VIP Database, and CBM (Chinese Biomedical Literature Database). The search covered the period from database inception to April 1, 2026. A combination of controlled vocabulary/subject headings and free-text terms was used. Within each concept, synonymous terms were combined with the Boolean operator OR, and the gastric cancer concept was combined with the symptom network analysis concept using AND. Truncation was used where applicable, and the search syntax was adapted to the indexing rules and retrieval functions of each database. In addition, a snowballing approach was applied by manually screening the reference lists of the included studies and relevant articles to identify additional eligible records. The search terms were developed around two main concepts: gastric cancer and symptom network analysis. Chinese-language search terms corresponded to “gastric tumour”, “malignant gastric tumour”, “gastric cancer”, “symptom network”, “network analysis”, and “core symptoms”. English search terms included “gastric cancer*”, “stomach cancer*”, “gastric neoplasm*”, “stomach neoplasm*”, “malignant gastric neoplasm*”, “malignant stomach neoplasm*”, “cancer of the stomach”, “familial diffuse gastric cancer”, “symptom network*”, “network of symptoms”, “symptom network structure”, “network analysis”, “symptom network analysis”, “core symptom*”, and “key symptom*”. The complete PubMed strategy, including MeSH and title/abstract terms and Boolean combinations, is presented in Table 1.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. PubMed search strategy.

https://doi.org/10.1371/journal.pone.0358940.t001

2.4 Literature screening and data extraction

Duplicate records were removed using Zotero 7.0. Two researchers, the first and second authors, independently screened all retrieved studies based on the predefined inclusion and exclusion criteria. In the first stage, titles and abstracts were reviewed to exclude clearly irrelevant studies. In the second stage, the full texts of potentially eligible studies were assessed. Disagreements were discussed with a third researcher, the corresponding author, until consensus was reached. The final included studies were then confirmed. Data were extracted using a standardized Microsoft Excel form. The extracted items included first author, publication year, country, study design, sample size, study population, assessment time points, assessment tools, and main findings. Methodological characteristics of the network analyses were also charted, including network model, software/packages, centrality indicators, bridge indicators, temporal indicators, and reporting of network stability and estimation accuracy.

2.5 Reporting guideline and methodological quality appraisal

This review was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR), and the completed checklist is provided in S1 File. A formal risk-of-bias or methodological quality appraisal was intentionally not undertaken because the aim of this scoping review was to map the breadth, characteristics, methods, and findings of the available evidence rather than to estimate pooled intervention effects or grade certainty of evidence. Instead, methodological features directly relevant to the interpretation of symptom network studies, including network estimation approaches, centrality and bridge indicators, temporal indicators, and stability/accuracy reporting, were systematically charted.

2.6 Ethical review

As this review was conducted solely on the basis of published studies, it did not involve human participants, individual-level data, or animal experiments. Ethical approval was therefore not required. The review protocol has been registered with the Open Science Framework under DOI: https://doi.org/10.17605/OSF.IO/MJAZ3

3.1 Results of literature screening

A total of 669 records were identified through the initial search. Following eligibility screening based on the predefined inclusion and exclusion criteria, 9 studies [10–18] were ultimately included in this review. The literature screening process is shown in S1 Fig.

3.2 Basic characteristics of the included studies

Among the 9 included studies, 5 were cross-sectional studies [11–14,16] and 4 were longitudinal studies [10,15,17,18]. Seven studies were conducted in China [10–16], and two were conducted in South Korea [17,18]. The study populations mainly comprised patients receiving chemotherapy for gastric cancer, patients undergoing postoperative chemotherapy, and patients in the perioperative period of gastric cancer treatment. In addition, two studies [16,18] included mixed cancer populations with gastric cancer patients. These studies were retained to map potentially relevant evidence but were treated as indirect evidence because gastric-cancer-specific symptom networks were not reported. The detailed characteristics of the included studies are presented in Table 2.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Basic characteristics of the included studies (n = 9).

https://doi.org/10.1371/journal.pone.0358940.t002

3.3 Main findings

A total of 7 studies [11–14,16–18] explicitly identified core symptoms across different treatment stages, while 3 studies [11,15,17] reported bridge symptoms. Three studies [11–13] explored factors associated with core symptoms, and 2 studies [10,15] examined temporal relationships among symptoms. The detailed characteristics are presented in Table 3. All 9 studies [10–18] used R software for network estimation and visualization. Among them, 5 studies [11,12,14,17,18] used strength centrality to identify core symptoms, and 3 studies [12,15,16] reported the stability and accuracy of their network results. The detailed methodological characteristics are presented in Table 4.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Main findings of the included studies.

https://doi.org/10.1371/journal.pone.0358940.t003

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Network analysis methods and reporting characteristics.

https://doi.org/10.1371/journal.pone.0358940.t004

4.1 Heterogeneity in the identification of core symptoms among patients with gastric cancer

Core symptoms are symptom nodes with relatively high centrality and strong connectivity to other nodes in the network; identifying these symptoms may help prioritize candidates for targeted symptom management [5]. However, core symptoms identified among patients with gastric cancer varied across studies, likely due to differences in treatment stages, assessment instruments, time points, and symptom nodes analyzed. Although some overlap existed among studies, no universally applicable core symptoms have been established. For example, among patients receiving postoperative chemotherapy, Zou et al. [13] identified altered taste as a core symptom, whereas Li et al. [14] identified sadness as the most prominent symptom. Differences in assessment tools may also contribute to these variations, as the MDASI-GI-C emphasizes gastrointestinal symptoms, while the MDASI incorporates psychological symptoms and quality-of-life domains. In addition, core symptoms changed across treatment phases; vomiting, fatigue, and pain were prominent after surgery [12], whereas altered taste, vomiting, and loss of appetite were more common during chemotherapy [13]. Overall, fatigue, vomiting, altered taste, loss of appetite, and pain warrant attention in gastric cancer symptom networks; however, no core symptom has been consistently identified across all treatment phases. Future studies should standardize assessment tools, time points, inclusion criteria, and network metrics, and conduct stage-specific and stratified analyses to improve the stability and comparability of core symptom identification.

4.2 Limited evidence on bridge symptoms and factors associated with core symptoms

Bridge symptoms connect different symptom clusters or dimensions and may help identify potential targets for limiting cross-domain symptom connectivity within the network [6]. However, among the nine included studies, only three reported bridge symptoms, and the identified bridge symptoms varied due to differences in treatment stages, assessment tools, and inclusion criteria. Zou Yanling et al. [11] found that sleep disturbances, loss of appetite, and somnolence acted as bridge symptoms in patients undergoing chemotherapy for gastric cancer, whereas Li et al. [15] reported that dry mouth was a bridge symptom during early postoperative chemotherapy and transitioned to abdominal distension during middle and late stages. These findings suggest that bridge symptoms may show stage-specific changes, although current evidence remains limited. Future studies should standardize symptom assessment tools, inclusion criteria, and assessment time points, and conduct longitudinal network analyses across treatment phases to identify relatively stable bridge symptoms and guide phased symptom interventions. Research on factors associated with core symptoms also remains limited, with current evidence mainly focusing on vomiting, fatigue, distress, and altered taste. Ge Jiaoyu et al. [12] found that age, sex, and surgical approach were associated with postoperative vomiting; hemoglobin and prealbumin levels with postoperative fatigue; and education level and postoperative complications with distress. Studies by Zou Yanling et al. [11], Zou et al. [13], and Li Siyu et al. [10] suggested an association between Phlegm-dampness constitution and altered taste, with Li Siyu et al. [10] further identifying a temporal association between Phlegm-dampness constitution and subsequent taste changes. According to Traditional Chinese Medicine theory, Phlegm-dampness constitution is associated with impaired spleen-stomach transformation and transportation, abnormal fluid metabolism, and internal accumulation of phlegm and dampness. Clinically, this may manifest as a sticky sensation in the mouth, thick and greasy tongue coating, and poor appetite [19]. A thick and greasy tongue coating has been proposed as a possible correlate of altered taste perception, while chemotherapy-related injury to taste and olfactory sensory cells provides a separate biomedical explanation for taste changes [20]. These mechanisms should be regarded as theoretical explanations rather than causal interpretations of the observed network associations. Accordingly, taste changes may warrant closer monitoring in patients with gastric cancer undergoing chemotherapy, including those classified as having Phlegm-dampness constitution. Within culturally appropriate TCM-informed care, strategies such as dietary guidance, oral care, moderate exercise, acupoint massage, and auricular seed therapy [21] may be considered as supportive approaches; however, the network studies included in this review do not establish that modifying TCM constitution improves taste perception or eating status. Prospective intervention studies are needed to test these hypotheses.

TCM constitution also raises a distinct issue of cross-cultural transferability. It is a culturally embedded diagnostic and classification construct without direct conceptual or measurement equivalents in many non-Chinese healthcare systems. Therefore, associations involving TCM constitution should not be assumed to generalize to other cultural or clinical settings simply because the symptom nodes themselves are common across populations. Testing transferability would require explicit cross-cultural operationalization and validation of the constitution construct, followed by replication of the observed symptom-network associations in more diverse populations.

4.3 Limited evidence on longitudinal symptom relationships and potential value of strongly connected symptom pairs for intervention

Temporal relationships among symptoms provide information on time-ordered statistical associations between earlier and later symptoms, helping to characterize symptom evolution patterns and generate hypotheses for proactive symptom management [22]. Among the four longitudinal studies included in this review, only two applied cross-lagged network analysis to examine temporal symptom relationships, indicating that current evidence remains limited. Although Li Siyu et al. [10] and Li et al. [15] both explored cross-temporal symptom associations, differences in assessment time points and included symptoms make it difficult to establish consistent conclusions. Future studies should standardize longitudinal assessment schedules, conduct continuous follow-up across treatment phases, and replicate temporal symptom associations through multicenter, large-sample studies. Despite limited evidence regarding temporal relationships, several strongly connected symptom pairs have been repeatedly identified and may represent potential intervention targets. Altered taste and loss of appetite showed consistent associations across multiple studies [11,13–15,18], and cross-lagged network analyses further showed that altered taste was temporally associated with subsequent loss of appetite. Therefore, early assessment of taste changes during chemotherapy, combined with oral care, dietary adjustment, and nutritional support, could be evaluated prospectively as a strategy for reducing later appetite loss and malnutrition risk. Nausea–vomiting was also repeatedly identified as a stable gastrointestinal symptom association [11,13,14], highlighting the importance of early prevention and continuous management of chemotherapy-induced nausea and vomiting. In addition, psychological symptom associations, particularly grief–distress, were repeatedly observed [11,14,16], suggesting that symptom management should incorporate psychological screening and emotional and social support alongside physical symptom control. Overall, symptom management for gastric cancer should move beyond individual symptom control toward interventions targeting key symptom interactions, with comprehensive and continuous management focusing on gastrointestinal symptoms, nutritional status, and psychological well-being.

Network analysis provides intervention-relevant information that differs structurally from conventional symptom cluster analysis. Conventional symptom cluster methods primarily identify groups of symptoms that co-occur and therefore support coordinated management of symptom sets [23], whereas network analysis additionally distinguishes high-connectivity central symptoms, bridge symptoms linking different symptom communities, and, in longitudinal models, time-ordered associations between symptoms. These features can generate more specific hypotheses about intervention prioritization and sequencing: central symptoms may be considered candidate high-connectivity targets, whereas bridge symptoms may be considered when the goal is to limit propagation of symptom burden across domains. For example, Li et al. [15] found that altered taste was temporally associated with subsequent loss of appetite. A conventional cluster analysis would support co-management of these eating-related symptoms, whereas the cross-lagged network suggests that early management of taste change could be tested as an upstream strategy before appetite deterioration occurs. The same study identified dry mouth as a bridge symptom in early chemotherapy and abdominal distension as a bridge symptom in later cycles, suggesting that candidate intervention targets may change across treatment phases. These implications remain hypothesis-generating because centrality, bridge status, and cross-lagged associations do not by themselves establish causal intervention effects.

4.4 Similarities and differences between mixed cancer studies and gastric cancer-specific studies

Mixed cancer studies and gastric cancer-specific studies have demonstrated both shared symptom patterns and disease-specific differences. Wang et al. [16] identified distress as a core symptom in a mixed gastrointestinal cancer population, with psychological and emotional symptoms forming a central component of the network. This finding was consistent with the importance of grief, distress, anxiety, and depression reported in gastric cancer-specific studies [11,14,17]. Rha et al. [18] identified fatigue as the most central symptom among patients with various cancer types undergoing chemotherapy and reported strong associations between anxiety–depression, fatigue–somnolence, and loss of appetite–altered taste. These findings partially correspond with the central role of fatigue and the altered taste–loss of appetite association observed in gastric cancer studies [11–15,17]. These similarities may reflect shared symptom patterns related to cancer and chemotherapy, but they do not establish that the underlying symptom-network structure is identical across cancer types. However, gastric cancer-specific studies placed greater emphasis on gastrointestinal and disease-specific symptoms, including vomiting, altered taste, loss of appetite, abdominal distension, and gastric cancer-related concerns. In the two mixed cancer studies, gastric cancer patients accounted for only 62/202 and 83/249 participants, respectively, and no separate symptom networks were constructed for gastric cancer subgroups. Therefore, findings from mixed cancer studies should be interpreted as indirect evidence with limited applicability to gastric cancer populations.

4.5 Network analysis modeling and heterogeneity in reporting

In addition to differences in treatment phases, assessment tools, and study populations, variations in network modeling approaches represent an important source of heterogeneity. Cross-sectional networks estimate conditional associations among symptoms at a single time point but cannot determine temporal sequences. In contrast, cross-lagged panel networks can estimate time-ordered associations but may be influenced by assessment intervals, regularization parameters, and node selection; such associations do not establish causality. Multi-time-point network analyses used by Shim et al. [17] and Rha et al. [18] captured changes in network structures across treatment stages but could not fully account for individual-level temporal effects. Methodological differences may also affect the identification of core and bridge symptoms. While Wang et al. [16] used Gephi for clustering and network analysis, most studies applied R-based regularized network models; differences in algorithms, centrality metrics, and standardization procedures may contribute to inconsistent symptom rankings. Therefore, repeated identification of symptoms such as fatigue and distress should not be interpreted solely as evidence of cross-study consistency. The interpretation of centrality results also depends on network stability and estimation accuracy. Although some studies reported strength centrality stability coefficients (CS-coefficients) above 0.50, indicating relatively stable rankings, lower stability estimates for closeness or betweenness centrality at certain time points suggest that these results should be interpreted cautiously. In addition, incomplete reporting of bootstrap confidence intervals for edge weights limited evaluation of estimation precision. Future studies should standardize reporting of network models, regularization parameters, edge-weight confidence intervals, centrality stability coefficients, and difference tests, and prioritize interpretation of indicators supported by adequate stability to improve transparency, reproducibility, and comparability across studies.