The Role of Dysphagia in Depression, Anxiety, and Symptom Clusters Among Nasopharyngeal Carcinoma Survivors After Intensity‐Modulated Radiotherapy: A Latent Profile and Mediation Analysis Based on a Cross‐Sectional Study

ABSTRACT

Objective

To identify distinct dysphagia profiles among nasopharyngeal carcinoma (NPC) survivors and to examine the mediating roles of depression and anxiety in the relationship between these profiles and broader symptom clusters. This study hypothesized that subgroup membership would have differential effects on symptom clusters through psychological mediating pathways. The study aimed to provide evidence for the development of precision symptom management interventions.

Methods

In this cross‐sectional study, 271 survivors recruited from a tertiary cancer hospital in Guangzhou (January 2023–December 2024) completed standardized assessments of dysphagia, psychological status, and symptoms. Latent profile analysis (LPA) identified subgroups of survivors with distinct dysphagia profiles; exploratory factor analysis extracted symptom clusters, multivariable linear regression examined associated factors, and parallel mediation analysis with bootstrap resampling tested the mediating effects.

Results

Dysphagia prevalence among NPC survivors was 57.2%, with the highest prevalence (72.7%) observed among survivors within 6 months post‐treatment. LPA identified two survivor subgroups with distinct dysphagia profiles: a high dysphagia burden subgroup (51.0%) and a low dysphagia burden subgroup (49.0%). Three symptom clusters were extracted: Systemic/Psychological, Oral/Swallowing, and Gastrointestinal/Neuropathic. Membership in the high dysphagia burden subgroup and anxiety significantly predicted all clusters (p < 0.05). Mediation analyses revealed that dysphagia subgroup membership had significant total effects on all clusters (p < 0.001). Depression showed statistical mediation for the Systemic/Psychological and Oral/Swallowing clusters, whereas anxiety showed statistical mediation for the Systemic/Psychological and Gastrointestinal/Neuropathic clusters.

Conclusions

NPC survivors exhibit distinct dysphagia burden profiles, including high‐ and low‐burden subgroups. A high dysphagia burden profile is directly associated with a greater overall symptom burden and indirectly contributes to more severe symptom clusters via depression and anxiety. These findings highlight the need for early identification of survivors within this high‐risk subgroup, and the integration of swallowing rehabilitation with psychological interventions for comprehensive symptom management.

1. Introduction

Nasopharyngeal carcinoma (NPC) is highly prevalent in Central and Southern China, with an incidence rate ranging from 25 to 30 cases per 100,000 population, ranking the highest among head and neck cancers [1]. The combination of intensity‐modulated radiotherapy (IMRT) and chemotherapy has increased the 5‐year survival rate of NPC patients to over 85.5% [2]. However, high‐dose chemoradiotherapy may lead to long‐term pathological changes, including soft tissue fibrosis, laryngeal dysfunction, and esophageal stenosis [3, 4]. Among these complications, dysphagia is one of the most common and severe late adverse effects, with a reported prevalence ranging from 13% to 93.5% [5]. Dysphagia may result in physical complications such as aspiration, malnutrition, and aspiration pneumonia, while also substantially impairing psychological well‐being and quality of life [6].

Existing studies have primarily focused on the short‐term assessment of swallowing function, leaving survivors in the extended post‐treatment phases insufficiently evaluated. More importantly, substantial individual heterogeneity exists in the severity of dysphagia among long‐term NPC survivors. Previous studies have largely relied on mean‐level comparisons, which may obscure underlying profile‐specific characteristics. Latent Profile Analysis (LPA) [7], a person‐centered analytical approach, can identify latent classes of individuals with distinct symptom characteristics, thereby providing methodological support for understanding variations in survivors' symptom severity and developing targeted interventions. This study applies a person‐centered approach to identify distinct dysphagia burden profiles among NPC survivors, providing insights for precision symptom stratification.

The concept of symptom clusters has evolved over the past two decades. Early studies defined symptom clusters as groups of concurrent and interrelated symptoms [8], with subsequent discussions focusing on the optimal number of symptoms required to constitute a cluster [9]. More recent expert recommendations from Miaskowski and colleagues emphasize that symptom clusters should be conceptualized as groups of symptoms that occur together, are related to each other, and may share common underlying mechanisms [10]. Importantly, symptom clusters should be examined within specific populations and clinical contexts, considering their potential impact on patient outcomes.

In clinical practice, dysphagia frequently coexists with multiple symptoms such as xerostomia, taste alterations, and pain; however, its central role within multisymptom clusters remains unclear. In addition, depression (44.2%) and anxiety (35.3%) are highly prevalent among long‐term NPC survivors and are associated with significantly reduced quality of life [11]. Based on the biopsychosocial model, bidirectional interactions exist between physical and psychological symptoms [12, 13]. Psychological distress may not only be triggered by physical symptoms, but may also exacerbate symptom burden by reducing adherence to rehabilitation, thereby creating a vicious cycle. Building upon this biopsychosocial framework, we constructed an a priori theoretical model for the current study. Specifically, we hypothesized that: (1) a high‐burden dysphagia profile is directly associated with greater symptom cluster severity; and (2) depression and anxiety serve as parallel mediators in the pathways from dysphagia subgroup membership to these specific symptom clusters. To our knowledge, this hypothesized mediation framework has not yet been empirically tested.

Therefore, this cross‐sectional study used LPA followed by parallel mediation analysis to identify latent classes of long‐term NPC survivors with distinct dysphagia profiles, compare differences in overall symptom cluster severity across subgroups, and, for the first time, examine the mediating roles of depression and anxiety in these associations. The study proposed the following hypotheses: (i) distinct latent classes of NPC survivors based on dysphagia profiles exist; (ii) class membership is significantly associated with symptom cluster severity; and (iii) depression and anxiety mediate the relationships between subgroup membership and symptom clusters. The findings of this study may provide a basis for developing targeted symptom management strategies that integrate psychological interventions. The hypothesized mediation model is presented in Figure 1.

FIGURE 1.

Hypothesized mediation model of dysphagia burden profiles, depression, anxiety, and symptom clusters.

2. Methods

2.1. Research Design

This study employed a cross‐sectional quantitative design.

2.2. Sample and Setting in Research

Convenience sampling was used to recruit NPC survivors who attended a tertiary hospital in Guangdong Province between January 2023 and December 2024. The inclusion criteria were as follows: (1) histologically confirmed NPC diagnosed by imaging and endoscopic biopsy according to the American Joint Committee on Cancer (AJCC) 8th edition [14]; (2) currently in the post‐treatment follow‐up phase (survivorship); (3) age ≥ 18 years; (4) completion of IMRT; (5) informed consent provided by the survivors; and (6) adequate comprehension and communication abilities. The exclusion criteria were as follows: (1) severe cognitive impairment or psychiatric disorders; and (2) critical illness or the presence of other major comorbidities.

2.3. Data Collection

At the time of assessment, all survivors had completed their definitive radiotherapy for NPC. During follow‐up visits, two specially trained nurses provided face‐to‐face explanations to the survivors regarding the study objectives, questionnaire instructions, and important considerations. Written informed consent was obtained before the questionnaires were administered. Participants were assured that refusal to participate would not affect their access to medical care. Each survey session lasted approximately 15–30 min and was conducted while participants waited for their clinic appointments. All participants received a printed health education booklet as a token of appreciation, and standardized instructions were used to explain the content. Participants completed the questionnaires independently; for those who experienced difficulties, the research staff assisted in completing the questionnaires by oral administration with participant consent. Disease‐related information was supplemented by reviewing electronic medical records. Questionnaires were collected on‐site and checked for completeness, and those with > 10% missing data or patterned responses were excluded. All data were double‐checked by two researchers before entry into the database.

2.4. Ethical Considerations

This study adhered to the principles outlined in the Declaration of Helsinki and followed applicable ethical guidelines. Ethical approval for this study was obtained from the Medical Ethics Committee of the authors' institution (Approval No. SL‐B2021‐173‐02).

2.5. Instruments

2.5.1. Sociodemographic and Clinical Characteristics

Based on a literature review and expert consultation, a 13‐item self‐designed questionnaire was used to collect survivors' sociodemographic and disease‐related information. Sociodemographic variables included age, sex, residence, marital status, educational level, monthly per capita household income, return to work status, and medical expense payment method. Disease‐related variables encompassed the presence of chronic conditions, time since treatment completion, receipt of chemotherapy, AJCC staging, and weight loss exceeding 10%.

2.5.2. Assessment of Dysphagia

Dysphagia was assessed using the M.D. Anderson Dysphagia Inventory (MDADI) [15]. It consists of 20 items across four domains (global status, emotional, functional, and physical) rated on a 5‐point Likert scale, with scores ranging from 1 to 5. Higher scores indicate better swallowing function and quality of life. A validated Chinese version of the MDADI was utilized in this study [16]. A composite MDADI score of 60, corresponding to an average item score of 3.0, was used as the cutoff for clinically significant moderate‐to‐severe dysphagia [17]. In the present study, the Cronbach's α coefficient was 0.961.

2.5.3. Assessment of Anxiety

Anxiety symptoms were assessed using the Generalized Anxiety Disorder‐7 (GAD‐7) [18]. The scale consists of seven items rated on a 4‐point Likert scale ranging from “not at all” to “nearly every day,” yielding a total score from 0 to 21. Cutoff scores of 5, 10, and 15 indicate mild, moderate, and severe anxiety, respectively [19]. In the present study, the Cronbach's α coefficient was 0.949.

2.5.4. Assessment of Depression

Depressive symptoms were measured using the Patient Health Questionnaire‐9 (PHQ‐9) [20]. It contains nine items rated on a 4‐point Likert scale, with total scores ranging from 0 to 27. Cutoff scores of 5, 10, and 15 represent mild, moderate, and severe depression, respectively, and a score ≥ 5 was used to indicate depressive symptoms [21]. In this study, the Cronbach's α coefficient was 0.917.

2.5.5. Assessment of Symptom Clusters

Symptom burden was evaluated using the M.D. Anderson Symptom Inventory‐Head and Neck Module (MDASI‐H&N) [22]. It includes two subscales: symptom severity (22 items) and symptom interference (6 items). Each item is rated on an 11‐point scale from 0 to 10, with higher scores indicating greater symptom severity or greater interference with daily life. A validated Chinese version was utilized [23], and the Cronbach's α coefficient in the present study was 0.877.

2.6. Sample Size Determination

According to the sample size estimation method based on LPA and regression analysis [24], methodological guidelines for LPA suggest a minimum sample size of 50 per subgroup. As this study intended to explore up to four potential subgroups, a minimum of 200 valid cases was necessary. Factoring in an estimated 10%–20% invalid response rate, the final recruitment target was set at 220 to 240 participants.

2.7. Data Analysis

Statistical analyses were performed using IBM SPSS Statistics version 26.0 for multivariable linear regression and R (version 4.4.3) for LPA, exploratory factor analysis (EFA), and parallel mediation analyses.

2.7.1. Descriptive Analysis

Descriptive statistics summarized demographic and clinical characteristics. Differences in dysphagia status across demographic and clinical characteristics were examined using independent‐samples t‐tests, Mann–Whitney U tests, or chi‐square (χ ²) tests, as appropriate.

2.7.2. Latent Profile Analysis

LPA was performed using item scores from the dysphagia scale as observed indicators. Models with 1 to 5 latent profiles were sequentially fitted. The number of profiles was not predetermined; rather, the final solution was selected in a data‐driven manner using prespecified statistical and interpretive criteria. Model fit was evaluated using three types of indices: (1) Information criteria: Akaike information criterion (AIC), Bayesian information criterion (BIC), and sample‐size‐adjusted BIC (aBIC); lower values indicate better fit. (2) Classification quality: Entropy values closer to 1 indicate better classification accuracy, with entropy ≥ 0.8 representing > 90% accuracy. (3) Likelihood ratio tests: the Lo–Mendell–Rubin likelihood ratio test (LMRT) and the bootstrapped likelihood ratio test (BLRT), with p < 0.05 indicating that the model with k profiles fits significantly better than the model with k − 1 profiles.

2.7.3. Exploratory Factor Analysis

To identify symptom clusters, a single exploratory factor analysis with varimax rotation was conducted, including all 22 symptom severity items from the MDASI‐H&N. The final factor solution was determined based on a set of pre‐established criteria: (1) factors were retained if they had an eigenvalue ≥ 1 and contained at least two symptoms; (2) individual symptom items were retained within a factor if their factor loading was ≥ 0.35; and (3) if a symptom cross‐loaded at ≥ 0.35 on multiple factors, it was assigned to the factor where it exhibited the highest loading. Once the final symptom clusters were identified, a total score for each cluster was created by calculating the mean score of its symptom items. This approach was selected to retain the original response scale metric, thereby facilitating the clinical interpretation of symptom severity across different clusters.

2.7.4. Multivariable Linear Regression Analysis

Multivariable linear regression models were conducted to identify independent predictors for the severity of each symptom cluster, allowing for the control of potential confounders. First, preliminary bivariate analyses (independent samples t‐tests, ANOVA, or Pearson correlations, as appropriate) were performed. The variables evaluated in these preliminary analyses included all collected sociodemographic characteristics (age, sex, residence, marital status, etc.), clinical/disease‐related variables (presence of chronic conditions, time since treatment completion, receipt of chemotherapy, AJCC staging, etc.), as well as dysphagia, anxiety, and depression scores. Only variables that achieved statistical significance (p < 0.05) in these preliminary bivariate analyses were strictly selected and incorporated into the multivariate models using the stepwise method. Prior to final modeling, multicollinearity diagnostics were performed, confirming that all Variance Inflation Factor (VIF) values were well below the threshold of 10, indicating no significant collinearity among the predictors. To account for potential criterion contamination due to symptom overlap across different instruments (e.g., “sadness” on both the MDASI and PHQ‐9), a sensitivity analysis was conducted. Overlapping items were temporarily excluded from the cluster scores, and the regression models were re‐run to verify the stability of the primary findings.

2.7.5. Mediation Analysis

To test the a priori theoretical model proposed in the Introduction, parallel mediation analyses were conducted after identifying dysphagia burden profiles through LPA. In this specified model, dysphagia subgroup membership (derived from the LPA) was entered as the independent variable representing NPC survivors' dysphagia severity. Rather than using a continuous score, this dysphagia variable was treated as a binary categorical variable and dummy‐coded (1 = low dysphagia burden subgroup, 0 = high dysphagia burden subgroup). The three identified symptom clusters (Systemic/Psychological, Oral/Swallowing, and Gastrointestinal/Neuropathic) served as the dependent variables, with depression and anxiety acting as parallel mediating variables. Using the lavaan package in R, parallel two‐mediator models were constructed to estimate direct, indirect, and total effects. Model parameters were estimated using maximum likelihood estimation, and 95% confidence intervals (CI) for indirect effects were obtained via nonparametric bootstrap (2000 resamples).

3. Results

3.1. LPA Identifies Survivor Subgroups Based on Dysphagia Profiles

Of 291 distributed questionnaires, 271 valid responses were collected (93.12% response rate); 20 were excluded primarily due to concurrent medical exams. To identify survivor subgroups based on their dysphagia profiles, five latent profile models were evaluated. Although AIC, BIC, and aBIC values declined continuously with additional profiles, the two‐profile model demonstrated excellent classification accuracy (entropy = 0.975) and a significantly better fit than the one‐profile model (LMRT and BLRT, p < 0.001). This two‐profile solution was selected as optimal because it avoids the overlapping, less clinically actionable profiles seen in models with ≥ 3 profiles and clearly distinguishes “mild” from “moderate‐to‐severe” survivors, aligning with intervention thresholds. Among the survivors, 138 (51.0%) were classified into the high dysphagia burden group and 133 (49.0%) into the low dysphagia burden group (Table 1). The high dysphagia burden group exhibited substantially lower scores across all MDADI items than the low dysphagia burden group (Figure 2), supporting the clinical labeling of the two latent profiles, consistent with previous research [25].

TABLE 1.

Model fit indices for LPA identifying survivor subgroups (N = 271).

Number of categories	AIC	BIC	aBIC	LMRT (p)	BLRT (p)	Entropy	Category (n)
1	16,742.577	16,886.662	16,759.834
2	141,71.939	14,391.668	14,198.255	< 0.001	< 0.001	0.975	138/133
3	13,275.191	13,570.565	13,310.568	0.2503	< 0.001	0.962	51/121/99
4	12,647.234	13,018.252	12,691.670	0.0907	< 0.001	0.972	94/45/96/36
5	12,485.449	12,932.112	12,538.945	0.2977	< 0.001	0.963	42/87/39/68/35

FIGURE 2.

Distribution of survivor subgroups. The red dashed line at a mean item score of 3.0 corresponds to the MDADI composite score cutoff of 60 for moderate‐to‐severe dysphagia, as described in the Methods section [17].

3.2. General Characteristics of NPC Survivors

Notably, the highest prevalence of dysphagia was observed among survivors within 6 months post‐treatment completion (72.7%) (see Table S1 for details). The 271 survivors were categorized into high dysphagia burden (N = 138) and low dysphagia burden (N = 133) subgroups based on their dysphagia profiles. All survivors had completed IMRT. Among them, 84.9% had also received chemotherapy, whereas none underwent primary surgical treatment. Compared to the low dysphagia burden subgroup, the high dysphagia burden subgroup was significantly older (p = 0.005), predominantly male (p = 0.046), and differed in residential distribution (p = 0.008) and time since treatment completion (p < 0.001). Socioeconomically, they had lower education (p < 0.001) and income (p < 0.001), affecting their return‐to‐work rate (39.9% vs. 58.6%, p = 0.002). Furthermore, they reported significantly higher scores for depression, anxiety, and head‐and‐neck symptom burden (all p < 0.001). Detailed characteristics are presented in Table 2.

TABLE 2.

Demographic, behavioral, clinical, and psychosocial characteristics of survivors (N = 271).

Variables	Categories	N (%)/mean ± SD	High dysphagia burden subgroup (N = 138)	Low dysphagia burden subgroup (N = 133)	t/χ 2 /Z	p
Age (years)		45.37 ± 10.33	47.10 ± 9.59	43.56 ± 10.79	2.855 a	0.005
Sex	Male	173 (63.8)	96 (69.6)	77 (57.9)	3.996 b	0.046
	Female	98 (36.2)	42 (30.4)	56 (42.1)
Residence	Urban	114 (42.1)	47 (34.1)	67 (50.4)	9.563 b	0.008
	Town	63 (23.2)	32 (23.2)	31 (23.3)
	Rural	94 (34.7)	59 (42.7)	35 (26.3)
Marital status	Unmarried	31 (11.4)	13 (9.4)	18 (13.5)	1.135 c	0.567
	Married	238 (87.8)	124 (89.9)	114 (85.7)
	Divorced	2 (0.8)	1 (0.7)	1 (0.8)
Education level	Primary or below	23 (8.5)	18 (13.0)	5 (3.8)	−4.586 d	< 0.001
	Middle school	72 (26.5)	47 (34.1)	25 (18.8)
	High school/technical school	65 (24.0)	32 (23.2)	33 (24.8)
	College or above	111 (41.0)	41 (29.7)	70 (52.6)
Monthly income	< 3000 CNY	112 (41.3)	71 (51.4)	41 (30.8)	−3.987 d	< 0.001
	3001–5000 CNY	68 (25.1)	34 (24.6)	34 (25.6)
	5001–10,000 CNY	53 (19.6)	22 (15.9)	31 (23.3)
	> 10,000 CNY	38 (14.0)	11 (8.1)	27 (20.3)
Medical payment method	Public expense	7 (2.6)	5 (3.6)	2 (1.5)	11.241 c	0.081
	Employee insurance	131 (48.3)	55 (39.9)	76 (57.1)
	Resident insurance	39 (14.4)	23 (16.7)	16 (12.0)
	New rural cooperative	72 (26.6)	45 (32.6)	27 (20.3)
	Self‐pay	7 (2.6)	4 (2.9)	3 (2.3)
	Commercial insurance	9 (3.3)	4 (2.9)	5 (3.8)
	Others	6 (2.2)	2 (1.4)	4 (3.0)
Chronic disease	No	186 (68.6)	88 (63.8)	98 (73.7)	3.093 b	0.079
	Yes	85 (31.4)	50 (36.2)	35 (26.3)
Time since treatment completion	Within 6 months after treatment	77 (28.4)	52 (37.7)	25 (18.8)	−3.489 d	< 0.001
	6 months–1 year after treatment	45 (16.6)	23 (16.7)	22 (16.5)
	1–3 years after treatment	65 (24.0)	29 (21.0)	36 (27.1)
	3–5 years after treatment	34 (12.5)	15 (10.9)	19 (14.3)
	> 5 years after treatment	50 (18.5)	19 (13.7)	31 (23.3)
AJCC stage	I/II	18 (6.6)	11 (8.0)	7 (5.3)	3.723 c	0.155
	III/IV	244 (90.0)	125 (90.6)	119 (89.4)
	Unclear	9 (3.4)	2 (1.4)	7 (5.3)
Chemotherapy	No	41 (15.1)	17 (12.3)	24 (18.0)	1.730 b	0.188
	Yes	230 (84.9)	121 (87.7)	109 (82.0)
Weight loss ≥ 10%	No	129 (47.6)	61 (44.2)	68 (51.1)	1.302 b	0.254
	Yes	142 (52.4)	77 (55.8)	65 (48.9)
Return to work	Yes	133 (49.1)	55 (39.9)	78 (58.6)	9.570 b	0.002
	No	138 (50.9)	83 (60.1)	55 (41.4)
PHQ‐9		4.68 ± 5.00	6.46 ± 5.78	2.82 ± 3.08	6.506 a	< 0.001
GAD‐7		3.44 ± 4.15	4.67 ± 4.85	2.17 ± 2.75	5.245 a	< 0.001
MDADI		65.06 ± 17.17	—	—	—	—
MDASI‐H&N		2.25 ± 1.82	2.90 ± 1.94	1.57 ± 1.40	6.505 a	< 0.001

Note: Bold values indicate statistically significant p values (p < 0.05), highlighting significant differences between the high‐ and low‐dysphagia burden subgroups.

^a The Independent‐samples t‐test.

^b The Pearson's χ ² test.

^c Fisher's exact test was used.

^d The Mann–Whitney U test.

3.3. Exploratory Factor Analysis of Symptom Clusters Among NPC Survivors

Among the survivors, the three most prevalent symptoms were dry mouth, fatigue, and mucus in the mouth and throat. To determine the symptom clusters, a single EFA was performed on all 22 symptom items. The data demonstrated excellent suitability for factor analysis (KMO = 0.96; Bartlett's test χ ² = 5205.65, p < 0.001). The final solution extracted three factors, accounting for 48.5% of the total variance. The remaining variance may reflect symptom‐specific and individual variation, measurement error, or additional symptom dimensions not captured by the current factor structure. Based on our retention criteria, all 22 items successfully loaded at > 0.35 and were assigned to factors. These three factors were labeled as the Systemic/Psychological, Oral/Swallowing, and Gastrointestinal/Neuropathic Symptom Clusters (Table 3).

TABLE 3.

Factor loadings of symptom items by symptom clusters (N = 271).

Item/symptom cluster	Factor loading
Systemic/Psychological Symptom Cluster
Difficulty with remembering	0.861
Shortness of breath	0.845
Distress/feeling upset	0.790
Fatigue	0.687
Drowsiness	0.714
Sadness	0.718
Disturbed sleep	0.688
Pain	0.526
Choking/coughing	0.561
Difficulty with voice/speech	0.418
Problems with teeth/gums	0.392
Oral/Swallowing Symptom Cluster
Mucus in mouth/throat	0.858
Problems with tasting food	0.752
Dry mouth	0.686
Mouth/throat sores	0.613
Difficulty swallowing/chewing	0.576
Numbness/tingling	0.434
Lack of appetite	0.455
Constipation	0.391
Gastrointestinal/Neuropathic Symptom Clusters
Nausea	0.743
Vomiting	0.637
Skin pain/burning/rash	0.464

3.4. Univariate Analysis of Demographic Characteristics and Symptom Clusters Among NPC Survivors

Univariate analyses demonstrated that monthly income and chronic comorbidities were significantly associated with the Systemic/Psychological Symptom Cluster. Residential area, educational level, monthly income, chronic comorbidities, medical payment method, time since treatment completion, weight loss ≥ 10%, and return‐to‐work status were significantly associated with the Oral/Swallowing Symptom Cluster. In addition, sex, residential area, educational level, monthly income, time since treatment completion, and return‐to‐work status were significantly associated with the Gastrointestinal/Neuropathic Symptom Cluster. Detailed results are presented in Table S2.

3.5. Multivariable Linear Regression Analysis of Factors Associated With Symptom Clusters

Multiple linear regression revealed that subgroup membership and anxiety independently predicted all three symptom clusters (p < 0.05). The final models explained 44.6%, 38.6%, and 29.6% of the variance in the Systemic/Psychological, Oral/Swallowing, and Gastrointestinal/Neuropathic Symptom Clusters, respectively (adjusted R ² = 0.440, 0.370, and 0.280; all model p < 0.001). For the Oral/Swallowing cluster, symptom burden decreased as time after treatment increased, with the 3‐ to 5‐year group exhibiting the lowest scores compared to those within 6 months after treatment (β = −0.254, p < 0.001). Within the Systemic/Psychological cluster, depression (β = 0.374) and anxiety (β = 0.277) emerged as strong positive predictors (both p < 0.001). For the Gastrointestinal/Neuropathic cluster, anxiety was the strongest predictor (β = 0.373, p < 0.001), while female sex (β = 0.146, p = 0.007) and rural residence (β = 0.137, p = 0.012) were also associated with greater symptom burden. Detailed results are presented in Table 4. Sensitivity analyses excluding potentially overlapping symptom items yielded results consistent with the primary analyses (Table S3).

TABLE 4.

Multiple linear regression analysis of symptom clusters in NPC survivors (N = 271).

Variable	B	SE	β	t	p
Systemic/Psychological
Dysphagia
High dysphagia burden subgroupref.
Low dysphagia burden subgroup	−0.453	0.182	−0.122	−2.489	0.013
Depression	0.139	0.028	0.374	5.023	< 0.001
Anxiety	0.124	0.033	0.277	3.812	< 0.001
Constant	1.927	0.325		5.929	< 0.001
Oral/Swallowing
Time since treatment completion
Within 6 months after treatmentref.
6 months–1 year after treatment	−0.689	0.325	−0.119	−2.124	0.035
1–3 years after treatment	−1.140	0.293	−0.226	−3.891	< 0.001
3–5 years after treatment	−1.648	0.357	−0.254	−4.611	< 0.001
> 5 years after treatment	−1.057	0.319	−0.191	−3.315	0.001
Dysphagia
High dysphagia burden subgroupref.
Low dysphagia burden subgroup	−0.502	0.227	−0.117	−2.209	0.028
Depression	0.132	0.034	0.305	3.840	< 0.001
Anxiety	0.102	0.040	0.196	2.527	0.012
Constant	3.350	0.419		7.999	< 0.001
Gastrointestinal/Neuropathic
Time since treatment completion
Within 6 months after treatmentref.
3–5 years after treatment	−0.712	0.319	−0.116	−2.229	0.027
Sex
Maleref.
Female	0.618	0.226	0.146	2.742	0.007
Residence
Urbanref.
Rural	0.585	0.230	0.137	2.543	0.012
Dysphagia
High dysphagia burden subgroupref.
Low dysphagia burden subgroup	−0.624	0.229	−0.153	−2.729	0.007
Return to work	0.409	0.221	0.101	1.854	0.065
Anxiety	0.183	0.027	0.373	6.800	< 0.001
Constant	0.367	0.577		0.636	0.526

Note: ref. = the reference category of the categorical variables in the model.

Abbreviations: β, standardized regression coefficient; B, regression coefficient; SE, standard error.

3.6. Mediating Effects of Depression and Anxiety on Symptom Clusters Among NPC Survivors

Parallel mediation analyses revealed the pathways through which subgroup membership was associated with symptom clusters (Table 5, Figure 3). Negative path coefficients indicate a higher symptom burden among survivors in the high dysphagia burden subgroup.

TABLE 5.

Direct, indirect, and total effects of dysphagia severity on symptom clusters mediated by depression and anxiety.

Outcome variable	Effect type	Path	β	SE	z	p	95% CI (LL–UL)
Systemic/Psychological Symptom Cluster	Direct	Dysphagia → Symptoms	−0.453	0.177	−2.57	0.010	−0.80, −0.11
	Indirect	Dysphagia → Depression → Symptoms	−0.508	0.147	−3.46	< 0.001	−0.83, −0.26
		Dysphagia → Anxiety → Symptoms	−0.311	0.113	−2.76	0.006	−0.54, −0.10
	Total indirect	—	−0.819	0.130	−6.28	< 0.001	−1.08, −0.57
	Total effect	—	−1.272	0.211	−6.02	< 0.001	−1.67, −0.85
Oral/Swallowing Symptom Cluster	Direct	Dysphagia → Symptoms	−0.699	0.233	−3.00	0.003	−1.17, −0.25
	Indirect	Dysphagia → Depression → Symptoms	−0.548	0.174	−3.15	0.002	−0.92, −0.24
		Dysphagia → Anxiety → Symptoms	−0.214	0.141	−1.52	0.129	−0.50, 0.09
	Total indirect	—	−0.762	0.139	−5.49	< 0.001	−1.06, −0.50
	Total effect	—	−1.461	0.244	−6.00	< 0.001	−1.93, −1.00
Gastrointestinal/Neuropathic Symptom Cluster	Direct	Dysphagia → Symptoms	−0.610	0.220	−2.77	0.006	−1.05, −0.16
	Indirect	Dysphagia → Depression → Symptoms	−0.297	0.166	−1.79	0.074	−0.65, 0.01
		Dysphagia → Anxiety → Symptoms	−0.300	0.147	−2.04	0.042	−0.62, −0.02
	Total indirect	—	−0.597	0.135	−4.43	< 0.001	−0.88, −0.35
	Total effect	—	−1.207	0.237	−5.08	< 0.001	−1.67, −0.75

FIGURE 3.

Mediation model depicting the associations between dysphagia burden profiles and symptom clusters through depression and anxiety. (A) Systemic/Psychological Symptom Cluster. (B) Oral/Swallowing Symptom Cluster. (C) Gastrointestinal/Neuropathic Symptom Cluster. Three separate mediation analyses were conducted, one for each symptom cluster outcome. β = standardized coefficient. Values in brackets = 95% bootstrap CI. Dashed arrows = direct effect (c″). N = 271. Bootstrap = 2000.

3.6.1. Systemic/Psychological Symptom Cluster

Membership in the high dysphagia burden subgroup was significantly associated with greater severity of these symptoms (total effect β = −1.272, p < 0.001). This association was partially mediated by both depression and anxiety, with depression demonstrating a stronger mediating effect (indirect β = −0.508 vs. −0.311, both p < 0.05) (Figure 3A).

3.6.2. Oral/Swallowing Symptom Cluster

Subgroup membership showed a profound overall association with this cluster (total effect β = −1.461, p < 0.001). Clinically, depression served as a significant partial mediator (indirect β = −0.548, p = 0.002), whereas the mediating role of anxiety was not statistically significant (Figure 3B).

3.6.3. Gastrointestinal/Neuropathic Symptom Cluster

Similarly, membership in the high dysphagia burden subgroup was significantly associated with worse symptoms (total effect β = −1.207, p < 0.001). Interestingly, for this specific cluster, anxiety was the significant partial mediator (indirect β = −0.300, p = 0.042), while depression did not show a significant mediating effect (Figure 3C).

4. Discussion

This study employed LPA and mediation analysis to characterize clinically meaningful heterogeneity in dysphagia burden among NPC survivors, identifying two latent subgroups (high‐ and low‐dysphagia burden subgroups). Moreover, it clarified the statistical mediating effects of depression and anxiety in the associations between subgroup membership and three symptom clusters (Systemic/Psychological Symptom Cluster, Oral/Swallowing Symptom Cluster, and Gastrointestinal/Neuropathic Symptom Cluster). The main findings were as follows: (1) subgroup membership and anxiety were significant independent predictors of all three symptom clusters; and (2) depression and anxiety partially or selectively mediated the associations between subgroup membership and symptom clusters, with their roles varying across different clusters. These findings provide a basis for symptom management strategies according to dysphagia burden profiles and symptom cluster characteristics.

4.1. Long‐Term Dysphagia Burden: From Prevalence to Clinical Subgrouping

This cross‐sectional analysis revealed a high prevalence of dysphagia among survivors at different times after treatment completion. Notably, the prevalence was 72.7% among survivors within 6 months after treatment and remained 56.0% among those more than 5 years after treatment. Although the cross‐sectional design precludes conclusions regarding changes in dysphagia over time, these findings suggest that dysphagia remains common even years after treatment. This observation is consistent with the recommendation by Doi et al. [26] that NPC survivors should receive long‐term follow‐up after IMRT. Previous studies have suggested that persistent dysphagia after radiotherapy may be related to late treatment‐related toxicities, including muscular fibrosis, lymphatic edema, neural injury, and salivary gland dysfunction [27, 28].

Importantly, LPA identified two subgroups of survivors: a high dysphagia burden subgroup (51.0%) and a low dysphagia burden subgroup (49.0%), demonstrating clinically meaningful differences in dysphagia severity among cancer survivors [29]. Previous studies have suggested that factors such as age, weight, tumor stage, and radiotherapy field design may contribute to differences in dysphagia severity [30].

Therefore, dysphagia management should move beyond simple symptom identification toward clinical risk stratification based on dysphagia burden. On the one hand, structured long‐term follow‐up systems should be established, with radiotherapy optimization to spare swallowing‐related structures integrated throughout treatment [31]. On the other hand, combined with nutritional indicators and weight change [32], as well as gold‐standard assessments such as videofluoroscopic swallowing study and fiberoptic endoscopic evaluation of swallowing [33, 34], high dysphagia burden survivors may warrant closer clinical monitoring.

4.2. Structural Features of Symptom Clusters and Their Impact on Quality of Life

Exploratory factor analysis extracted three symptom clusters: Systemic/Psychological Symptom Cluster, Oral/Swallowing Symptom Cluster, and Gastrointestinal/Neuropathic Symptom Cluster, with a cumulative variance explanation of 48.5%. This supports the concept that symptoms occur in clustered patterns [35].

The Systemic/Psychological Symptom Cluster included fatigue, sadness, distress/feeling upset, difficulty with voice/speech, shortness of breath, and choking/coughing, collectively reflecting the combined physiological and psychological burdens experienced by NPC survivors. Psychologically, radiotherapy‐induced changes in physical appearance serve as an important trigger for negative emotions, whereas the cumulative effects of treatment‐related toxicities are more likely to precipitate anxiety and panic reactions. Studies have shown that the prevalence of cancer‐related fatigue among NPC patients is as high as 82.5%, which is closely associated with psychological distress and poor sleep quality [36], and may further form a vicious cycle through neuroendocrine‐immune pathways [37].

Although our study did not directly assess neurocognitive function indicators, it is important to acknowledge that the literature has well documented the potential long‐term complications faced by head and neck cancer survivors, such as radiotherapy‐induced brain injury. Such injuries can manifest as persistent cognitive impairment, including memory decline and reduced attention [38], with the severity of injury closely linked to the radiation dose delivered to the hippocampus, temporal lobe, and cerebellum [39]. The symptom clusters observed in our clinical practice (e.g., fatigue, psychological distress) may exacerbate or interact with these underrecognized cognitive deficits. Therefore, future research should integrate symptom assessment with neuropsychological testing and neuroimaging evaluations for comprehensive analysis.

The Oral/Swallowing Symptom Cluster primarily included xerostomia, mucus abnormalities, swallowing and chewing difficulties, as well as taste disturbance, anorexia, constipation, and local numbness/tingling. These symptoms are largely attributable to radiation‐induced salivary gland acinar atrophy and fibrosis. Reduced secretion of saliva and protective proteins leads to viscous oral mucus, increasing the risk of oral infections and dental caries [40]. Tam et al. [41] further demonstrated that anorexia and taste disturbance decrease over time after treatment, indicating a dynamic symptom trajectory.

The Gastrointestinal/Neuropathic Symptom Cluster reflected cumulative treatment toxicity, including nausea, vomiting, and skin pain/burning/rash. According to the 2016 MASCC/ESMO guideline update, chemotherapy‐induced nausea and vomiting are classically categorized into three temporal patterns: acute (occurring within 24 h post‐chemotherapy), delayed (24 h to several days post‐chemotherapy), and anticipatory (a conditioned response triggered by treatment‐related cues) [42]. For NPC survivors, anticipatory nausea and vomiting may persist as a conditioned response even after active treatment has been completed [43]. These persistent complaints can contribute to ongoing nutritional impairment and reduced long‐term treatment adherence [42]. Radiation dermatitis occurs in 85%–95% of head and neck cancer patients and represents one of the most common radiotherapy‐related toxicities [44].

Multiple regression analyses further demonstrated that dysphagia subgroup membership was an independent predictor of all three symptom clusters (p < 0.05), indicating its central role within the symptom. A high dysphagia burden profile is not only directly associated with a more severe Oral/Swallowing Symptom Cluster but also indirectly linked to greater Systemic/Psychological symptoms through functional limitations. Therefore, symptom management should shift from single‐symptom control to cluster‐based integrated interventions.

4.3. Dysphagia and Negative Emotions: Core Drivers Within the Symptom Cluster

The relatively higher levels of depression and anxiety observed among survivors with a high dysphagia burden were consistent with the findings of Wang et al. [11], who reported that psychological distress remains common among NPC survivors after treatment. These findings provide additional context for the mediating roles of depression and anxiety observed in the present study.

Anxiety was an independent predictor of all three symptom clusters, consistent with previous evidence suggesting links between psychological distress and somatic symptom burden. This finding is consistent with autonomic dysregulation and central nervous system hyperactivation under anxiety states. Depression significantly predicted both the Systemic/Psychological Symptom Cluster (β = 0.374) and the Oral/Swallowing Symptom Cluster (β = 0.305), highlighting its central role in psychological and swallowing‐related symptom burden.

Previous studies have indicated that negative emotions play a central and highly interactive role in the overall symptom burden among head and neck cancer patients [45]. In the present study, the association between dysphagia subgroup membership and the Systemic/Psychological Symptom Cluster was partially mediated by depression and anxiety, further supporting the critical role of these psychological factors in symptom amplification.

This study suggests a close association between dysphagia and negative emotions. Previous studies have shown that dysphagia‐related eating difficulties, nutritional risk, and social withdrawal may contribute to depression and anxiety [46, 47], whereas psychological distress may increase symptom perception and overall symptom burden [48, 49]. These findings suggest that swallowing rehabilitation combined with psychological support may help improve symptom management.

4.4. Differential Mediating Roles of Depression and Anxiety in the Relationship Between Dysphagia and Symptom Clusters

Mediation analyses revealed distinct roles of depression and anxiety across symptom clusters (Table 5). For the Systemic/Psychological Symptom Cluster, subgroup membership showed a significant total effect (β = −1.272, p < 0.001), with both direct and indirect effects (via depression and anxiety) being significant. This suggests that a high dysphagia burden is associated with more severe systemic and psychological symptoms both directly and indirectly through emotional distress. Consistent with previous research findings [50, 51], patients with dysphagia exhibited significantly higher scores for anxiety and depression compared to those without dysphagia, and feelings of anxiety and panic during eating were more prevalent.

For the Oral/Swallowing Symptom Cluster, subgroup membership had a significant direct effect (β = −0.699, p = 0.003), and was mainly mediated by depression, whereas the anxiety pathway was not significant. The direct effect reflects the physiological damage and functional limitations experienced by survivors after radiotherapy (e.g., dry mouth, swallowing pain, and difficulty chewing). In their study of head and neck cancer patients, Zebralla et al. [52] demonstrated a significant correlation among dysphagia, voice disorders, and pain. The indirect effect suggests that depression may be associated with greater symptom burden through psychological processes reported in previous studies and with functional deterioration beyond physiological impairment [53, 54]. It is noteworthy that the indirect pathway of anxiety within the s showed no significant effect, which may be attributed to sample size limitations or the sensitivity of the measurement tool and warrants further validation in future studies.

For the Gastrointestinal/Neuropathic Symptom Cluster, subgroup membership demonstrated both direct and indirect effects, with anxiety acting as the primary mediator. Previous studies have suggested that anxiety is associated with chemotherapy‐induced nausea and vomiting [55]. The mediating role of depression in this cluster was not statistically significant, possibly due to stronger anxiety‐driven somatic amplification.

Overall, depression and anxiety represent important psychological factors associated with symptom burden and reduced quality of life [56]. These findings suggest that symptom management strategies for NPC survivors may need to consider symptom cluster profiles. Specifically, for survivors with predominant Systemic/Psychological symptoms, psychological interventions addressing emotional distress may be particularly relevant. For those with prominent Oral/Swallowing symptoms, depression‐focused assessment and support warrant clinical attention. For survivors with Gastrointestinal/Neuropathic symptom burden, anxiety‐targeted approaches could be explored to interrupt conditioned symptom responses. Future intervention studies are needed to test these hypotheses.

5. Limitations

This study has several limitations. First, the cross‐sectional design precludes the inference of causal relationships and temporal dynamics among variables. Future research should adopt longitudinal or experimental designs to further validate the mediating pathways and causal effects. Secondly, all key variables were assessed via self‐reported measures. While patient‐reported outcomes are essential for capturing subjective symptoms and psychological distress, they may introduce recall bias. For dysphagia, future studies should combine patient self‐reports with objective instrumental evaluations, such as imaging findings and physiological swallowing assessments, for a more comprehensive assessment. Thirdly, the sample was drawn from a single medical center. Although this center treats NPC survivors from a relatively wide geographic area, selection bias may still be present, limiting the generalizability of the findings. Future studies should consider multicenter designs with larger sample sizes and include additional potential mediators, such as self‐efficacy and social support, to further clarify the comprehensive mechanisms underlying symptom cluster formation. Building on the subgroup characteristics identified in this study, targeted interventions integrating psychological support and swallowing rehabilitation could be developed and tested, ultimately facilitating the development of targeted rehabilitation strategies. Additionally, while we conducted sensitivity analyses to account for item overlap across the instruments (e.g., MDASI, and PHQ‐9), the inherent conceptual redundancy among head and neck‐specific measures remains a limitation that should be considered when interpreting the strength of the correlational findings.

6. Conclusion

Dysphagia is one of the most severe and distressing symptoms among NPC survivors [57]. The present study demonstrates that NPC survivors can be stratified into distinct dysphagia burden profiles, reflecting clinically meaningful variations in swallowing‐related symptom severity. Belonging to the high dysphagia burden subgroup was associated with more severe physical symptoms and functional limitations, and this association was statistically mediated by depression and anxiety. Notably, psychological mechanisms play a central mediating role in the Systemic/Psychological Symptom Cluster, highlighting the critical importance of integrating psychological interventions in this population. Clinically, early identification of survivors with a high dysphagia burden, together with routine psychological assessment, may facilitate more comprehensive symptom management, and ultimately enhance long‐term quality of life.

Author Contributions

Hui Zhao: conceptualization, writing – original draft, methodology. Yongjiao Kang: conceptualization, writing – original draft, methodology. Jiajie Xu: conceptualization, methodology, writing – original draft. Yanling Wen: investigation, data curation. Wen Hu: investigation, data curation. Yuying Fan: supervision, writing – review and editing, methodology. Yan He: writing – review and editing, supervision, methodology.

Funding

This work was supported by the Yue‐Qin Nursing Research and Innovation Fund of Sun Yat‐sen University Cancer Center (Grant No. YQ2025007‐B).

Ethics Statement

This study was approved by the Medical Ethics Committee of Sun Yat‐sen University Cancer Center (Approval No. SL‐B2021‐173‐02).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Prevalence of dysphagia in nasopharyngeal carcinoma survivors by time since treatment completion.

Table S2: Comparison of general characteristics among different symptom clusters (N = 271).

Table S3: Sensitivity analyses after excluding potentially overlapping symptom items.

Data Availability Statement

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.