Symptom clusters and influencing factors in pancreatic cancer patients with adjuvant chemotherapy after surgery: a cross-sectional study

Pancreatic cancer (PC) is an aggressive gastrointestinal malignancy with an extremely poor prognosis. Although its 5-year survival rate has increased from 3% to 13%, it remains the lowest among all cancers^[1]. Radical surgery is the only potentially curative treatment and the means to long-term survival. However, only 15%–20% of patients are eligible for resection^[2]. This low resectability rate is largely due to the lack of specific early symptoms and reliable biomarkers, underscoring the urgent need for early diagnostic strategies^[3]. Emerging molecular imaging techniques further highlight that early tumor identification is key to improving prognosis^[4]. Adjuvant chemotherapy significantly reduces postoperative recurrence risk and prolongs survival. Regimens such as gemcitabine plus nab-paclitaxel (AG) and modified FOLFIRINOX are widely used but often cause toxic side effects^[5] and, combined with surgical trauma and psychological distress, can trigger complex multiple symptoms that severely impair patients' quality of life.

Current evidence confirms that PC patients undergoing postoperative adjuvant chemotherapy bear a heavy symptom burden. Kristensen et al.^[6] identified pain as a core chemotherapy-related symptom, with progressive increases in severity, accompanied by gastrointestinal symptoms such as nausea and vomiting, which severely undermine patients' ability to carry out daily activities. Similarly, a longitudinal assessment by Burrell et al.^[7] of 143 post-operative patients receiving chemotherapy found that fatigue, sleep disturbance, appetite loss, and weight loss were widespread and severe. Expanding on this, Seelen et al.^[8] conducted a multicenter prospective study and demonstrated that beyond physical complaints (e.g., pain, fatigue) and gastrointestinal symptoms (e.g., loss of appetite, diarrhea), patients commonly suffer from psychological complications including depression and anxiety throughout chemotherapy, resulting in a complex, multi-layered burden. Statistics show that PC patients on chemotherapy experience an average of 13.39 ± 5.83 symptoms, a burden markedly heavier than that of patients with other solid tumors such as breast or lung cancer^[9,10]. Notably, addressing this complex symptom profile during chemotherapy represents a key clinical hurdle. Suboptimal symptom control often triggers treatment delays or early cessation, diminishes therapy adherence, and ultimately compromises survival outcomes^[11].

In clinical practice, symptoms frequently do not occur in isolation but coexist as 'Symptom Clusters' (SCs)—defined as two or more concurrent, interrelated, relatively stable symptoms that exhibit statistical independence^[12]. Although studies have explored SCs in pancreatic cancer patients (e.g., Burrell et al.^[7] identified 16 SCs via exploratory and confirmatory factor analysis in stage II PC patients who underwent surgical resection or with subsequent chemoradiotherapy, with severity significantly associated with reduced overall quality of life). Yang^[13] used the Memorial Symptom Assessment Scale (MSAS) to evaluate patients across chemotherapy cycles and identified seven SCs (pancreatic cancer-specific, treatment-related, psychological, gastrointestinal, body image disturbance, fatigue, and excretory SCs). Research focusing specifically on the postoperative adjuvant chemotherapy phase remains insufficient. The framework of SCs helps comprehensively assess patients' symptom burden and develop integrated intervention strategies. Unlike single symptom interventions, SC-oriented models not only improve intervention efficiency and rational allocation of medical resources but also reduce symptom severity and enhance patient-reported outcomes^[14].

The Theory of Unpleasant Symptoms (TOUS) offers a framework for symptom experiences by describing four symptom dimensions—timing, severity, distress, and quality—emphasizing their dynamic, holistic nature^[15]. Currently, SCs are identified based on occurrence, severity, and distress. Among these, severity is a core dimension that is directly associated with key outcomes, including physiological functional status, treatment adherence, and quality of life^[16]. Occurrence offers epidemiological description, distress reflects psychological impact, whereas severity directly captures symptom intensity and burden. For instance, Lin et al.^[17] found that the severity dimension showed excellent model fit in AML chemotherapy patients. Thus, identifying SCs via severity can precisely reveal the physiological state of patients during chemotherapy, providing clear targets for clinical intervention.

Notably, merely identifying SCs is insufficient to fully guide clinical practice, as their severity varies substantially among individuals and may be influenced by demographic characteristics, disease stage, treatment regimens, and other factors^[18]. In-depth exploration of the factors influencing SCs is essential for developing individualized management strategies. Current research predominantly focuses on breast and lung cancer, having confirmed that factors such as BMI, age, self-efficacy, social support, and financial burden are associated with SCs burden^[19,20]. However, research on SCs determinants in PC remains limited: Reyes-Gibby et al.^[21] identified comorbidities, treatment phase, and regimens as influencing factors in PC patients undergoing chemoradiotherapy, while Burrell et al.^[7] conducted a longitudinal study on surgical patients, highlighting preoperative anxiety and pain as predictors of postoperative SCs severity. Collectively, existing evidence primarily addresses patients receiving chemoradiotherapy or surgery alone, with a lack of systematic investigation into SCs determinants during adjuvant chemotherapy following PC surgery. Notably, this population endures dual burdens of surgical trauma and chemotherapy toxicity, potentially resulting in unique symptom patterns and influencing factors. Therefore, guided by the TOUS framework, this study aims to systematically describe postoperative PC patients' symptom experiences during adjuvant chemotherapy. It will focus on the severity dimension to identify SCs via exploratory factor analysis and to analyze factors influencing their severity, with the ultimate goal of developing timely, precise interventions based on SCs' peak severity periods.

This was a single-center study conducted at the Pancreatic Center of a Grade A tertiary hospital in Nanjing, Jiangsu Province, China. This study was approved by the Institutional Review Board of the participating hospital and strictly adhered to the Declaration of Helsinki. All participants provided written informed consent for secondary data use. A pilot study (n = 20), completed in July 2024, indicated that symptom occurrence was significantly lower in patients undergoing their first chemotherapy cycle (n = 5) than in those who had received subsequent cycles (n = 15). Based on this finding, the main study protocol optimized the inclusion criteria to 'having completed at least one full cycle of chemotherapy' to focus on the phase with a more significant symptom burden. Recruitment for the main study (n = 224) officially commenced on January 20, 2025. Data from the pilot study were used solely for design optimization and were not included in the final analysis.

Study participants

Inclusion criteria: adult patients (≥ 18 years) with pathologically confirmed pancreatic cancer who had undergone surgery, were receiving adjuvant AG chemotherapy, and had completed at least one full chemotherapy cycle. Participants were required to be conscious, able to communicate verbally, and capable of providing signed informed consent.

All included stage IV patients were oligometastatic cases who underwent resection of both the primary tumor and metastatic lesions.

Exclusion criteria: patients who had received neoadjuvant therapy; were undergoing concurrent radiotherapy, immunotherapy, or targeted therapy; had a history of other malignancies; or suffered from severe cardiovascular or cerebrovascular diseases, organ dysfunction, or psychiatric/cognitive disorders; and those lacking key baseline laboratory data required for the study.

The sample size met two empirical criteria^[22]: absolute N ≥ 100, and a sample-to-variable ratio (N:K) of 5–10:1. Of the 32 symptoms assessed by the Memorial Symptom Assessment Scale, only those with occurrence rates between 20% and 80% were retained for factor analysis, yielding 21 symptoms (K = 21). The final sample comprised 224 valid cases, satisfying N ≥ 100 and exceeding the recommended N:K upper limit (224/21 ≈ 10.7:1 > 10:1).

Study instruments

General information questionnaire

Self-designed by the research group, the questionnaire collected demographic data (e.g., age, sex, education level, primary caregiver, smoking/alcohol use, comorbid chronic diseases) and clinical characteristics (e.g., body mass index [BMI], tumor location, American Joint Committee on Cancer [AJCC] stage, metastasis, time since diagnosis, chemotherapy cycles, HbA1c, complete blood count parameters, and aspartate aminotransferase [AST]). All data were retrieved from the hospital's electronic medical record system.

Memorial Symptom Assessment Scale (MSAS)

Developed by Portenoy et al. in 1994^[23], this scale is designed to capture the multifaceted nature of the symptom experience in oncology populations. It has proven reliable, valid, and psychometrically robust across cancers—including pancreatic cancer—via extensive validation. It assesses the occurrence, severity, and distress of 32 symptoms, as well as the frequency of 24 of these symptoms. Occurrence was indicated as 'Yes' or 'No'. A four-point Likert scale was used to assess frequency and severity, while distress is rated on a scale from 0 to 4 points. The Chinese version of the MSAS (MSAS-Ch) used in this study was officially validated by Cheng et al. in 2009^[24]. Cronbach's α ranged from 0.79 to 0.87, and it demonstrated a high content validity index (CVI) of 0.94, showing good reliability and validity.

Nutritional Risk Screening 2002 (NRS-2002)

Developed by Kondrup et al. in 2002^[25], it is endorsed by the European Society for Clinical Nutrition and Metabolism (ESPEN) for identifying malnutrition risk among adult patients in hospital settings. The scoring stage covers nutritional status, disease severity, and age adjustment (≥ 70 years). Nutritional risk is assessed on a scale from 0 to 7 points, with a score of ≥ 3 considered indicative of risk. Studies show satisfactory diagnostic accuracy (sensitivity 0.98, specificity 0.77). Consequently, this study used the Chinese version of NRS-2002, translated by Chen et al. in 2005^[26], widely promoted and adopted in nutritional risk screening for Chinese cancer patients.

Karnofsky Performance Status (KPS) scale

First proposed by Karnofsky in 1948^[27], it was used to assess patients' daily living abilities and quality of independent life. The scoring ranges from 0 (death) to 100 (normal). Higher scores indicate better functional status.

Procedures

All research personnel underwent uniform standardized training to familiarize themselves with the questionnaire content and master standardized investigation methods. All questionnaires were checked on-site and corrected as needed; no missing data were present in the final 224-patient sample.

Baseline data collection: at the time of enrollment, trained researchers completed the General Information Questionnaire by reviewing electronic medical records and conducting face-to-face interviews with patients or their family members. Nutritional status was assessed using the Nutritional Risk Screening 2002 (NRS-2002) tool.

Symptom and functional status assessment: on day 7 of the chemotherapy cycle, the MSAS and the KPS scale were administered to evaluate the patients' symptom experiences (with a focus on severity) and their functional status, respectively. To minimize subjective bias and ensure data accuracy, the researchers read each item in a neutral tone and recorded responses objectively based strictly on patient feedback. Questionnaires were collected on-site and reviewed immediately; any data identified as questionable were promptly verified and corrected.

Data analysis

Data were analyzed using SPSS software (v27.0). Categorical variables were presented as frequencies and percentages; continuous variables with normal distribution were described using mean ± standard deviation, while skewed variables were described using median and interquartile range. Correlation analysis was employed to examine the associations between nutritional indicators, biochemical indicators, and symptom clusters. Univariate analysis was performed using independent samples t-tests or analysis of variance (ANOVA). Bonferroni correction was applied for conservative interpretation of the univariate comparisons, but the univariate analysis itself was exploratory screening to avoid missing potential variables. Hence, all predictors with unadjusted p < 0.05 were entered into the multivariable linear regression to identify factors influencing symptom cluster severity, with p < 0.05 considered statistically significant for the final model. Bonferroni-adjusted significance level is provided for reference.

Exploratory Factor Analysis (EFA) with principal component analysis and varimax rotation was performed based on symptom severity to identify high-burden SCs. Following methodological recommendations^[28], we excluded symptoms with an occurrence of < 20% or > 80% to ensure adequate variability and stable factor solutions, as symptoms with very low or very high occurrence provide limited variance for correlation estimation and may compromise the stability and interpretability of the factor structure. Prior to factor extraction, the Kaiser–Meyer–Olkin (KMO) and Bartlett's test of sphericity confirmed suitability for factor analysis (KMO > 0.5; Bartlett's test p < 0.05). Factors were extracted based on the criteria of eigenvalue ≥ 1, minimum number of items ≥ 2, and factor loading ≥ 0.4. Items displaying cross-loadings were assigned to the factor with the highest loading. Subsequently, Cronbach's α coefficient was calculated post-extraction to verify the reliability of each symptom cluster. Finally, clusters were named based on their constituent items, relevant literature, and clinical significance.

Cluster severity scores were calculated as the arithmetic mean of the symptom severities within each cluster. These raw scores (1–4 scale) were directly entered into regression models without standardization, given the common metric and confirmed normality (Shapiro–Wilk test, all p > 0.05).

This study ultimately included 224 pancreatic cancer patients. The mean age was 64.22 ± 7.93 years, and 132 patients (58.9%) were male. The majority of patients had an education level of junior high school or below (68.3%). Additionally, 58.9% of the patients had comorbid chronic diseases. Tumors were predominantly located in the head and neck of the pancreas (65.6%), with a mean time since diagnosis of 11.28 ± 11.28 months. AJCC stage was primarily stage II (62.5%). Detailed information is presented in Table 1.

Symptom occurrence and severity

Pancreatic cancer patients experienced an average of 14.88 ± 4.40 symptoms during chemotherapy. A total of 21 symptoms had an occurrence rate ≥ 20%. The five most prevalent symptoms were: nervousness (76.3%), anxiety (75.4%), feeling sad (75.0%), lack of energy (70.5%), and lack of appetite (67.0%). Nervousness was the most severe (2.05 ± 1.55) and the most distressing (2.02 ± 1.57) symptom. Detailed data are shown in Table 2.

Symptom clusters based on severity

EFA was performed on 21 symptoms, with the KMO measure of 0.761 and a significant Bartlett′s test of sphericity (p < 0.001). Five common factors were extracted, accounting for a cumulative variance of 72.13%: Factor 1 (Gastrointestinal Symptom Cluster) included six symptoms. Factor 2 (Psychological Symptom Cluster) included four symptoms. Factor 3 (Physiological Energy Depletion Symptom Cluster) included five symptoms. Factor 4 (Neuromuscular Symptom Cluster) included four symptoms. Factor 5 (Body Image Disturbance Symptom Cluster) included two symptoms. Detailed results are presented in Table 3.

Univariate analysis of symptom cluster severity

The results revealed that the severity of the gastrointestinal symptom cluster was significantly associated with time since diagnosis, chemotherapy cycles, NRS-2002, and AST levels. The psychological symptom cluster was influenced by age, smoking history, comorbid chronic disease history, tumor location, AJCC stage, KPS score, and neutrophil count. The physiological energy depletion symptom cluster was associated with smoking history, comorbid chronic disease history, KPS score, and chemotherapy cycles. The neuromuscular symptom cluster was associated with alcohol consumption history, chemotherapy cycles, NRS-2002, and AST levels. Furthermore, education level, primary caregiver, and time since diagnosis were associated with the body image disturbance symptom cluster (all p < 0.05; for details, see Table 4).

Multiple linear regression analysis of symptom clusters

Variables significant in the univariate analysis were entered into multiple linear regression models. All five regression models were statistically significant (p < 0.001). The Durbin–Watson statistics (1.857–2.082) indicated no residual autocorrelation. Model diagnostics showed no multicollinearity (all VIF < 10), and residuals were approximately normally distributed (all p > 0.05). As detailed in Table 5, the severity of the gastrointestinal symptom cluster was independently associated with chemotherapy cycles, time since diagnosis, NRS-2002, and AST level (explaining 35.0% of variance). The psychological symptom cluster severity was associated with AJCC stage, tumor location, comorbid chronic disease history, and KPS score (explaining 9.0% of variance). Factors influencing the physiological energy depletion symptom cluster were comorbid chronic disease history, KPS score, and smoking history (explaining 16.5% of variance). The neuromuscular symptom cluster was associated with alcohol consumption history and chemotherapy cycles (explaining 16.6% of variance). The body image disturbance symptom cluster was associated with education level, primary caregiver, and time since diagnosis (explaining 13.4% of variance).

Sensitivity analysis

A sensitivity analysis excluding the 17 stage IV patients (Supplementary Tables S1, S2) confirmed the stability of the five symptom clusters and the majority of regression findings, with only minor changes in a few predictors (e.g., AST and nutritional risk score became non-significant). The cumulative variance explained was 72.19%.

To the best of our knowledge, this is the first study to explore symptom clusters and the factors that influence them in PC patients receiving postoperative adjuvant AG chemotherapy. This study investigated the symptom experience of PC patients. Patients reported a mean of 15 symptoms. Nervousness showed the highest occurrence, severity, and distress. In contrast to lymphoma patients, who primarily experience fever and fatigue^[29], this difference may be related to the poor prognosis, high recurrence rate, and chemotherapy side effects of pancreatic cancer^[30]. Additionally, drowsiness and abdominal distension were also highly prevalent. These findings support TOUS, in which symptoms are multidimensional and linked by shared biopsychosocial mechanisms.

Note that this cross-sectional study did not measure inflammatory cytokines, hormones, or metabolic biomarkers. Therefore, the following discussion of potential mechanisms is based on the literature and should be considered hypothesis-generating rather than conclusive.

Composition and influencing factors of symptom clusters

Based on the severity dimension, this study identified five symptom clusters: gastrointestinal, psychological, physiological energy depletion, neuromuscular, and body image disturbance. Compared to prior symptom-cluster studies in gastric^[31] and breast cancer^[20], our findings reveal both overlapping domains and distinct configurations specific to pancreatic cancer. This population-level heterogeneity underscores the necessity of tailoring clinical interventions to disease-specific symptom severity profiles rather than applying generic management protocols.

The gastrointestinal symptom cluster is centered on lack of appetite, accompanied by nausea, vomiting, dry mouth, and abdominal distension. This contrasts with prior studies where nausea and diarrhea predominated^[32,33]. We hypothesize this reflects PC's unique physiology: postoperative reductions in digestive enzyme secretion and glucose dysregulation may delay gastric emptying and suppress hypothalamic appetite regulation. Notably, greater severity of appetite loss may be associated with poorer nutritional status (malnutrition prevalence: 42.4%–88.0%)^[34]. We found that longer chemotherapy cycles were associated with higher cluster scores, consistent with the cycle-dependent toxic effects of chemotherapy^[35]. However, Li et al.^[36] found more severe GI symptoms in children with leukemia at initial chemotherapy, likely due to the high-intensity 'induction remission' regimen. The AG regimen (gemcitabine plus nab-paclitaxel) is speculated to cause symptom exacerbation via gastrointestinal mucosal injury, chronic inflammation, and neuroregulatory imbalance. This inflammatory link is indirectly supported by glioma research showing FBXL16 mediates cytokine release to regulate macrophage recruitment^[37,38]. Additionally, a longer time since diagnosis positively correlates with GI symptom severity, consistent with the conclusions of Wei et al.^[39], possibly due to higher tumor burden aggravating digestive tract compression or infiltration and mucosal damage from multiple chemotherapies. Secondly, a higher NRS-2002 score is linked to more severe GI symptoms. This differs from the report by Shi^[40] showing no such correlation in glioma patients, possibly because the pancreas (a core digestive organ) is intimately linked to nutrition, and the AG regimen's significant damage to GI mucosa usually forms a malnutrition-related vicious cycle^[41]. However, this hypothesis remains to be supported by direct evidence. Furthermore, higher AST levels are associated with worse GI symptoms, matching^[42] findings in brain tumors, potentially due to chemotherapy-induced hepatocellular injury (manifested as elevated AST), reducing bile secretion (impairing fat digestion), and decreasing protein synthesis (weakening digestive enzyme function).

The psychological symptom cluster includes sadness, anxiety, nervousness, and irritability, with sadness symptoms being the most severe. The findings are highly consistent with the cluster (worry, anxiety, sadness, irritability) summarized by Harris et al.^[43], suggesting that psychological symptom clusters with high negative emotional burden are common across cancer types. However, its presentation exhibits notable cancer-type heterogeneity, related to cancer site, treatment modality, and psychosocial stressors. For example, in lung cancer, clusters tend to center on sadness and distress linked to prognosis-related worry and dyspnea^[44]. We found that AJCC stage III patients exhibited more psychological symptoms than stage I, consistent with Tie et al.^[45], likely due to the advanced stage's increased tumor burden, complex treatments, and prognostic uncertainty exacerbating psychological distress. Secondly, head/neck cancer patients experienced a greater psychological burden than those with body/tail, echoing Wei et al.^[39], potentially due to head/neck tumors inducing more visible physical symptoms (e.g., jaundice, pain) as well as appearance changes and impaired social functioning, amplifying stress. A KPS score was negatively correlated with psychological severity, corroborating the conclusions of Zhang et al.^[18] in lung cancer patients. Reduced functional status (e.g., inability to work or perform self-care) often triggers helplessness and negative emotions. Additionally, patients with chronic diseases had more severe psychological symptoms, consistent with the report by Burger & Okkenhaug^[46] in hematological cancers. Prior models hypothesize that chronic disease–tumor coexistence drives persistent inflammation, which disrupts hypothalamic–pituitary–adrenal (HPA) axis function through cytokines, thereby inducing sickness behaviors such as depression and anxiety^[38]. Notably, although psychological symptoms formed a distinct cluster, reporting bias warrants consideration. Grounded in the symptom perception hypothesis, elevated negative affect heightens hypervigilance, amplifying somatic complaints. It also correlates strongly with self-reported severity but only weakly with clinician-assessed severity^[47]. Thus, the high psychological symptom burden observed may reflect both genuine distress and an affective amplifier of gastrointestinal and fatigue symptoms. This does not invalidate the cluster structure; rather, it implies that targeting negative affect may yield cross-domain benefits. Future studies require objective measures to disentangle amplification from true burden.

The physiological energy depletion symptom cluster is characterized by lack of energy, difficulty concentrating, and drowsiness. The prevalence of cancer-related fatigue (CRF) in this study (70.54%) was lower than that detected by disease-specific tools like FACT-F^[7], yet slightly higher than the 60.1% in Chen et al.^[48], where low loading of lack of energy in an 'emotional-illness' cluster suggests that methodological and cultural differences may lead to underestimation. Mechanistically, the high severity of this cluster is associated with platinum-based chemotherapy-induced myelosuppression, inflammation, and reduced oxygen-carrying capacity of red blood cells (leading to tissue hypoxia)^[49]. This same inflammatory mechanism is also observed in CAR-T cell therapy, where cytokine release syndrome induces fatigue and cognitive impairment, supporting immune activation as a common pathway underlying energy depletion across treatment modalities^[50]. Given its high clinical relevance, the more severe the CRF, the faster the decline in patients' physical function and quality of life, and concurrent gastrointestinal and psychological symptoms further amplify fatigue. We found that patients with comorbid chronic diseases had lower energy levels, which aligns with Ma et al.^[51]. Chronic diseases are speculated to not only exacerbate fatigue by depleting physiological reserves, superimposing chemotherapy toxicity, and mediating inflammation, but also to contribute to symptom development via immune-metabolic dysregulation^[52]. Additionally, lower KPS scores correlated with more severe physiological symptoms, aligning with reports by Dodd et al.^[12] in adult cancer patients, and linked to reduced exercise tolerance and increased energy burden from chemotherapy. Furthermore, smokers exhibited greater physiological symptom severity. Smoking may impair cardiopulmonary function and metabolism, amplifying chemotherapy-related energy depletion^[53].

The neuromuscular symptom cluster is characterized by hand/foot swelling, numbness, and tingling. It is closely associated with AG chemotherapy-induced peripheral neuropathy (CIPN) caused by drugs like paclitaxel^[49], severely affecting patients' daily activities and quality of life. However, its composition and severity ranking vary across cancer types: lung cancer exhibits a psycho-neurological cluster with psychological symptoms more severe than somatic ones^[54], whereas breast cancer shows a fatigue-pain dominant cluster^[20]. This variation may reflect the significantly different chemotherapy regimen used for pancreatic cancer. Analysis revealed alcohol consumption history positively correlated with neuromuscular symptoms, differing from the findings by Zhang et al.^[55] (milder symptoms in drinkers)—possibly due to varied definitions of drinking history (our study did not limit drinking duration to examine long-term effects). Notably, Mellion et al.^[56] demonstrated that alcohol and its metabolites can directly damage neuromuscular tissues, causing myasthenia and pain, which explains our result. We also found chemotherapy cycles positively correlated with neuromuscular symptom severity, consistent with the previous study^[39] on gastric cancer patients—attributed to drug toxicity, tissue damage, and reduced body repair capacity in association with higher cycle counts.

The body image disturbance symptom cluster centers on two core symptoms: 'I do not look like myself' and hair loss, consistent with previous findings^[57]. In our study, 58% of patients experienced hair loss, similar to the 58.1% reported by Nozawa et al.^[58] as the leading treatment-related appearance change. Although not directly life-threatening, such changes can blur identity and disrupt self-continuity, leading to self-perception distress and reduced self-esteem. This study found that patients with lower education experienced more severe image disturbance, consistent with the lung cancer study by Ma et al.^[59], potentially due to weak cognitive restructuring ability and the inability to actively utilize social resources through multiple channels to participate in body image management. Patients with spouses as primary caregivers scored higher in distress, similar to the study by Fang et al.^[60], likely because intimate relationships intensify attractiveness anxiety and fear of being disliked or a burden^[51]. Additionally, patients diagnosed for over 12 months reported more disturbance, aligning with previous studies^[35,39], linked to disrupted body integrity and impaired dignity from disease burden and chemotherapy-related damage.

Clinical implications

Leveraging symptom cluster profiling, we developed a risk-stratified, proactive nursing model. For the gastrointestinal cluster, NRS-2002 screening was intensified every two cycles; oral nutritional supplements (ONS) were triggered preemptively for scores ≥ 3, coupled with AST-guided hepatoprotection and pancreatic enzyme replacement therapy (PERT). Psychological and neuromuscular clusters utilized targeted surveillance: MSAS screening for AJCC stage III or head/neck tumors, and intensified neurotoxicity monitoring for those with a drinking history or ≥ 4 cycles. Screen-positive patients received immediate nurse-led cognitive restructuring and prophylactic neuroprotection (e.g., gabapentin). Physiological energy depletion management featured KPS-tailored exercise and nocturnal hypoxemia screening for smokers/COPD patients, optimizing medication timing and glycemic control. For body image disturbance, culturally adapted interventions—including visual aids, spousal involvement, and narrative therapy—facilitated identity reconstruction. This integrated approach shifts the paradigm from reactive management to proactive, multi-cluster governance.

Limitations

This study provides valuable empirical evidence on symptom clusters and influencing factors in pancreatic cancer patients receiving AG adjuvant chemotherapy. However, several limitations should be noted. First, the single-center design and convenience sampling may limit the generalizability of the findings and introduce selection bias; although sensitivity analysis supported robustness, future studies in non-metastatic or stage IV cohorts are needed and should also conduct sensitivity analyses to assess selection bias. Second, symptom assessment was only conducted on day 7 post-chemotherapy, failing to capture the dynamic trajectory of symptom clusters across chemotherapy cycles, particularly the differences between acute and delayed toxicities. Thus, no causal or temporal interpretations can be drawn from these data. Third, while the 20%–80% prevalence threshold for symptom inclusion is methodologically sound, it may exclude rare but clinically significant symptoms that could impact patient outcomes. Fourth, the adjusted R² values of the regression models were relatively low; future research should expand the range of predictor variables to improve the explanatory power of the model. Last, due to the lack of an independent validation sample, confirmatory factor analysis could not be performed in a strict sense. Future multicenter, longitudinal studies should track the trajectories of symptom progression, validate the identified symptom cluster structure, and assess its stability and replicability across populations and regimens (e.g., modified FOLFIRINOX).

This study identified a distinct symptom severity profile in post-operative pancreatic cancer patients receiving AG chemotherapy, with nervousness being the most severe symptom. Five major symptom clusters were recognized. The severity of these clusters was associated with factors such as chemotherapy cycles, nutritional risk, and tumor stage, highlighting the need for targeted interventions including nutritional support, psychological care, and functional exercise. Future research should focus on identifying potential predictors of symptom severity and exploring both common mechanisms and tailored management strategies across symptom clusters to advance symptom science.