The association between radiation dermatitis severity and social integration among patients with head and neck cancer undergoing radiotherapy: Statistical mediation by mood states

Abstract

Objectives

This study aimed to examine whether positive and negative mood composites statistically mediate the association between radiation dermatitis severity and social integration among patients with head and neck cancer undergoing radiotherapy.

Methods

Employing a cross-sectional design, this study recruited 223 eligible patients with head and neck cancer undergoing radiotherapy through consecutive sampling at a tertiary-level hospital in Liaoning Province, China, from January to December 2024. Research instruments included a demographic questionnaire, the Radiation-Induced Skin Reaction Assessment Scale, the Profile of Mood States-Short Form, and the Social Cohesion Scale. The mediation analysis focused on the subsample of 183 patients with dermatitis. A parallel mediation analysis was performed using the SPSS PROCESS macro (Model 4), controlling for potential confounders including age, sex, marital status, education, radiation dose, and concurrent chemotherapy.

Results

The severity of radiation dermatitis was significantly negatively correlated with social integration and positive emotions, while being significantly positively correlated with negative emotions. After controlling for covariates, the direct association between dermatitis severity and social integration was not statistically significant. The bootstrap-based mediation analysis revealed statistically significant indirect associations through both the positive and negative mood composites. Specifically, greater dermatitis severity was associated with higher negative emotions and lower positive emotions, which in turn were associated with reduced social integration.

Conclusions

These cross-sectional findings suggest that mood states may represent relevant pathways linking radiation dermatitis severity to social integration in patients with head and neck cancer undergoing radiotherapy. As the present design cannot establish causality, the observed associations are hypothesis-generating. Psychological screening and supportive care warrant further evaluation alongside dermatitis management. Longitudinal and interventional studies are required to confirm the directionality of these pathways and to determine whether targeting mood states improves social integration outcomes in this population.

Introduction

Head and neck cancer (HNC) refers to a heterogeneous group of malignant tumors arising from the mucosal surfaces of the oral cavity, pharynx, larynx, nasal cavity, and paranasal sinuses.¹ Radiation therapy, often in combination with chemotherapy, is the primary treatment modality for this condition.² Due to the anatomical location and radiosensitivity of surrounding tissues, patients with head and neck cancer (HNC) are particularly prone to developing radiation-induced dermatitis (RID), an acute skin reaction to ionizing radiation. In this patient population, the incidence of RID approaches 100%, presenting a significant clinical challenge.³ The clinical manifestations associated with this condition range from mild erythema and pruritus to more severe reactions, such as dry or wet desquamation.⁴ Unlike dermatological side effects in other body regions that can be easily concealed by clothing, RID in patients with HNC occurs on the face and neck—areas that are highly visible and critical for social interaction and self-identity. This unique visibility makes concealment difficult, thereby exacerbating the patient's vulnerability to profound psychosocial consequences, including body image disturbance, psychological distress, and avoidance of social activities, ultimately impairing quality of life.⁵ In severe cases, RID may necessitate unanticipated interruption of radiotherapy, which could potentially compromise treatment efficacy.⁶

The physical changes associated with RID may be associated with reduced social functioning. Given the exposed nature of the symptoms, patients may experience feelings of shame or heightened self-consciousness, which may in turn be associated with social withdrawal and reduced social functioning.⁷ Social integration, defined as the extent to which an individual participates in social networks and maintains a sense of connectedness and belonging to their community, is a crucial component of psychosocial well-being and recovery.⁸ A body of research, particularly among survivors of breast cancer,^5,9 has identified a correlation between treatment-related appearance changes and social avoidance, depression, and diminished self-esteem. This finding indicates that the relationship between physical symptoms and social outcomes is not direct, but rather is likely influenced by underlying psychological processes. Guided by the Biopsychosocial Model, we hypothesize that physical symptoms (RID) do not occur in isolation but elicit psychological responses (mood states) that are, in turn, associated with behavioral outcomes (social integration). Investigating these interrelationships is essential to identify whether psychological support can mitigate social impairment even when physical symptoms persist. Mood states, encompassing negative emotions (e.g., tension, depression, fatigue) and positive emotions (e.g., vitality, self-esteem), appear to be an important factor. This distinction is grounded in the Two-Factor Theory of Affect,¹⁰ which posits that positive and negative emotions are distinct dimensions that can coexist and independently influence psychological functioning. Furthermore, grouping subscales into these two broad dimensions allows for a parsimonious examination of whether social impairment is associated with the presence of distress, the absence of vitality, or both. This distinction is potentially relevant for future intervention research, as the Broaden-and-Build Theory suggests that positive and negative emotions may benefit from different strategies.¹¹ Previous research has identified mediators such as body image disturbance and stigma.¹² However, these studies predominantly focus on negative psychological outcomes. Mood state represents a more immediate and proximal psychological reaction to physical symptoms than complex cognitive constructs like stigma. Furthermore, focusing solely on distress overlooks the potential role of positive psychological resources (e.g., vigor), which are critical for social engagement. Therefore, this study selects mood states to capture both the accumulation of negative affect and the depletion of positive affect as dual pathways.

The present study is primarily guided by the Theory of Unpleasant Symptoms (TOUS).¹³ This theoretical framework posits that symptom experience comprises three interrelated components: the symptom itself, influencing factors, and functional outcomes. Within the proposed model, RID is conceptualized as the primary symptom, and social integration is operationalized as the functional outcome. While TOUS traditionally categorizes psychological variables as influencing factors that moderate outcomes, recent adaptations of the model in cancer research suggest that psychological responses can also act as proximal mediators.¹⁴ Based on the TOUS framework and the theoretical distinctions described above, the following hypotheses were specified prior to data collection to guide the planned mediation analysis: mood state functions as a proximal psychological response to the physical burden of dermatitis, which is in turn hypothesized to be associated with social integration. This hypothesized “Symptom → Psychological Response → Functional Outcome” pathway represents a statistical model to be examined cross-sectionally; it does not imply established temporal or causal ordering. To isolate this specific psychological pathway, physiological (e.g., radiation dose) and situational factors (e.g., demographics) are treated as covariates.

To operationalize the mediator, we draw on the Two-Factor Theory of Affect,¹⁰ which posits that positive and negative emotions are distinct dimensions that can coexist and independently influence psychological functioning. This theoretical distinction provides the basis for decomposing mood states into two composite scores—negative emotions (tension, depression, anger, fatigue, confusion) and positive emotions (energy, self-esteem)—and for specifying a parallel mediation model with two distinct indirect pathways. Additionally, the Broaden-and-Build Theory offers interpretive support for the positive emotion pathway, suggesting that positive emotions broaden thought-action repertoires and build personal resources necessary for social engagement.¹¹ This theory is not used as a guiding framework for the study design but rather to contextualize why the depletion of positive affect may be associated with reduced social functioning.

TOUS serves as the primary structural framework guiding the hypothesized model, the Two-Factor Theory of Affect provides the rationale for variable operationalization, and the Broaden-and-Build Theory supplements the interpretation of one specific pathway. The biopsychosocial perspective is consistent with this integrated approach but is not treated as a separate guiding framework.

Method

Study design and participants

This study adhered to the STROBE checklist standards for cross-sectional research. From January to December 2024, patients undergoing radiotherapy for head and neck cancer were recruited using consecutive sampling in the inpatient and outpatient departments of the Radiotherapy Department at a tertiary hospital in Liaoning Province. Patients' psychosocial status was assessed via face-to-face distribution, completion, and collection of questionnaires.

The inclusion criteria for participants were as follows: (1) aged 18 years or older; (2) pathologically confirmed diagnosis of primary head and neck cancer (including oral cavity, pharynx, larynx, nasal cavity, or paranasal sinuses); (3) currently undergoing radiotherapy to the head and neck region (patients were assessed after receiving 2 weeks of treatment to allow for symptom manifestation); (4) expected survival of more than 6 months, to exclude the confounding psychological impact of imminent end-of-life care. This was determined based on: (a) clinical staging confirming M0 status (absence of distant metastasis), and (b) the attending radiation oncologist's professional assessment, informed by the patient's performance status, disease extent, and planned treatment protocol. No validated prognostic scoring tool was used; this judgment-based approach is acknowledged as a limitation; (5) able to comprehend, speak, and read Chinese and capable of completing the questionnaires independently; and (6) voluntarily participated in the study and provided written informed consent.

Exclusion criteria included: (1) diagnosis of a severe, pre-existing psychiatric disorder (e.g., schizophrenia, bipolar disorder, or major depressive disorder diagnosed prior to the cancer diagnosis), as documented in the medical record; (2) presence of distant metastasis at the time of enrollment; (3) previous history of radiotherapy to the head and neck region; (4) concurrent participation in other clinical trials, especially those involving psychological or dermatological interventions; and (5) presence of pre-existing skin diseases in the radiation field (e.g., psoriasis, eczema, or lupus erythematosus).

The initial sample size was estimated using a prevalence-based formula for cross-sectional studies:

$$n=\frac{Z_{\alpha /2}^2\times p\times \left(1-p\right)}{{\delta }^2}$$

where α = 0.05, Z₀.₀₅/₂ = 1.96, P = 0.275 (estimated based on prior literature reporting impaired social functioning among patients with head and neck cancer undergoing radiotherapy).¹⁵ As no published study during the design phase reported a precise proportion of HNC patients with good social integration using the Social Cohesion Scale, this estimate was derived from related social functioning data in HNC populations, and δ = 0.05. This yielded a minimum of 306 participants. With an additional 10% allowance for incomplete data, the target was 337 cases.

We acknowledge that this prevalence-based formula was designed to estimate a population proportion with a specified precision, rather than to power a mediation analysis. At the study design stage, this formula was selected because: (a) no prior effect size estimates for the specific mediation pathways (RID → mood → social integration) were available in the head and neck cancer literature to inform an a priori power analysis for mediation; and (b) the parallel mediation model, the construction of two mood composites (positive and negative emotions), and the three directional hypotheses were specified during the study design phase based on the TOUS framework and existing literature available at that time.

During the recruitment period, 294 patients were approached for eligibility screening. Of these, 38 did not meet the inclusion criteria (principal reasons: expected survival < 6 months (n = 12), presence of distant metastasis (n = 9), documented pre-existing psychiatric disorder (n = 7), previous radiotherapy to the head and neck region (n = 5), concurrent participation in other clinical trials (n = 3), and pre-existing skin disease in the radiation field (n = 2)). Of the 256 eligible patients invited to participate, 24 declined. The remaining 232 patients provided written informed consent and completed the questionnaire package. Of these, nine returned questionnaires with incomplete responses that could not be resolved through on-site verification, and were excluded. Ultimately, 223 valid cases were included in the full sample (participation rate: 87.1%, 223/256). Among these, 183 patients (82.1%) who developed radiation dermatitis constituted the analytical sample for the mediation analysis.

We acknowledge that post-hoc power calculations are inherently conditional on observed results and do not substitute for prospective power planning. The post-hoc analysis is presented solely to provide context regarding the statistical adequacy of the analytical sample. Additionally, following Fritz and MacKinnon,¹⁶ who recommend a minimum sample of approximately 148 to detect medium-sized indirect effects (αβ) with 0.80 power using bias-corrected bootstrap methods, our analytical sample of 183 exceeds this threshold. Nevertheless, the discrepancy between the planned enrollment target (337) and the final analytical sample (183) is acknowledged as a limitation, and replication in larger, multi-center cohorts with a priori power analyses specifically designed for mediation is warranted.

Instrument

General information questionnaire

A general information questionnaire was based on a review of relevant literature identifying key sociodemographic predictors of quality of life in patients with cancer,¹⁵ to collect patients' sociodemographic characteristics, such as gender, age, educational attainment, and marital status. Clinical disease-related data were referenced from medical records, including radiotherapy modality (e.g., Intensity-Modulated Radiation Therapy (IMRT)), radiation dose, and concurrent chemotherapy status.

Radiation-Induced Skin Reaction Assessment Scale

RISRAS was developed by Noble–Adams in 1999 to assess the severity of radiation dermatitis.¹⁷ We selected RISRAS over CTCAE because it incorporates a patient-reported component (e.g., pain, itching), which is crucial for assessing the subjective psychological burden of the symptom. The scale consists of two parts. The first part is a patient-reported symptom assessment form, comprising four items: skin tightness, pain, itching, and burning sensation in the radiation field. Each item is scored on a 0–3 scale using a four-point Likert scale. The second part comprises a clinician's professional assessment form with four items: erythema, dry desquamation, wet desquamation, and necrosis. Each item is scored on a 0–4 scale. The total score is the sum of both forms (range 0–28), with higher scores indicating more severe symptoms. This study employed the Chinese version of RISRAS.¹⁸ Although previous studies reported test-retest reliability coefficients ranging from 0.64 to 0.75, the scale demonstrated excellent internal consistency in the current sample, with a Cronbach's α coefficient of 0.956.

Profile of Mood States-short form

The POMS scale was first developed by McNair et al., in 1971 and shortened by Shacham et al.^19,20 This study utilizes the Chinese localized short form,²¹ which comprises 40 items across seven dimensions: Tension, Anger, Fatigue, Depression, Confusion, Energy, and Self-Esteem. Each item employs a five-point Likert scale (0–4). To examine distinct mediating pathways, we created two composite scores based on the theoretical structure of affect:²²

• a)Negative Mood Composite: Sum of scores from Tension, Anger, Fatigue, Depression, and Confusion subscales. Higher scores indicate greater negative mood. We note that this composite encompasses both affective distress (tension, depression, anger, confusion) and somatic-affective experience (fatigue), consistent with the POMS conceptualization of mood as a multidimensional construct.
• b)Positive Mood Composite: Sum of scores from Energy and Self-Esteem subscales. Higher scores indicate greater positive mood. This composite captures both positive activated affect (energy/vigor) and positive self-evaluation (self-esteem). We use the term “mood composites” rather than “emotions” to reflect the broader nature of these aggregate scores.

The Cronbach's alpha coefficient for the overall scale in the present study was 0.813. The Cronbach's alpha for the negative emotion composite score was 0.916, and for the positive emotion composite score was 0.945, indicating good internal consistency.

A CFA was conducted at the subscale level using maximum likelihood estimation (n = 183). The a priori two-factor model demonstrated acceptable to good fit: χ² (13) = 23.45, P = 0.037, χ²/df = 1.80, CFI = 0.967, TLI = 0.951, RMSEA = 0.066 (90% CI: 0.018–0.107), SRMR = 0.042. Standardized factor loadings ranged from 0.71 (Fatigue) to 0.85 (Depression) for the negative mood factor, and from 0.89 (Self-Esteem) to 0.92 (Energy) for the positive mood factor. The correlation between the two latent factors was −0.41, supporting their treatment as distinct but related constructs. No post-hoc model modifications were introduced. Full CFA results are presented in Supplementary File 1.

Social cohesion scale

SCS was developed by Fone et al. and adapted into Chinese.^23,24 Although originally designed to measure social cohesion, the scale's two dimensions—Routine Integration (reflecting behavioral participation in social networks and activities) and Intimate Integration (reflecting emotional belongingness and sense of connectedness)—are conceptually aligned with the definition of social integration adopted in this study: “the extent to which an individual participates in social networks and maintains a sense of connectedness and belonging to their community".⁸ The Chinese version has demonstrated adequate construct validity and has been used to assess social integration levels in prior health research.^15,24 This dual structure makes it suitable for capturing both behavioral and affective facets of social functioning in patients with cancer. The scale comprises two dimensions: Routine Integration (reflecting behavioral participation) and Intimate Integration (reflecting emotional belongingness). Each dimension contains four items rated on a five-point Likert scale. Total scores range from 8 to 40 points, with higher scores indicating better social integration. While originally developed for community populations, the items (e.g., sense of belonging, social interaction) are highly relevant to the social challenges faced by patients with cancer. Previous studies have demonstrated that social cohesion constructs are valid and reliable predictors of quality of life in patients with chronic illnesses and cancer.¹⁵ In this study, the Cronbach's α coefficient for the scale was 0.710.

Data collection

Data were collected by two researchers, both oncology nurses with more than 5 years of clinical experience, who received standardized training on the study protocol and ethical principles. Prior to formal data collection, inter-rater reliability for the clinician-rated component of RISRAS was established through a calibration exercise. Both researchers independently assessed 20 patients using the clinician portion of RISRAS (erythema, dry desquamation, wet desquamation, necrosis). Ratings were completed independently without consultation. Agreement was quantified using Cohen's weighted kappa, which accounts for the ordinal nature of the rating scale. The weighted kappa was 0.87 (95% CI: 0.79–0.95), indicating excellent agreement. Discrepancies (observed in 3 of 20 cases, all involving one-grade differences in erythema scoring) were resolved through consensus discussion, and a standardized rating protocol with photographic anchors was adopted for the study period. Researchers provided detailed explanations of the study content to eligible patients. Following written informed consent, a self-administered questionnaire package was provided. Completion took approximately 15–20 minutes. To ensure data quality, researchers implemented strict quality control measures: (1) on-site checks for missing items immediately upon return; (2) logic verification (e.g., ensuring age matched birth year); and (3) double-entry of 20% of the data to check for entry errors.

Ultimately, 223 valid questionnaires were recovered. As reported in"Study design and participants" section Section 2.1, the participation rate was 87.1% (223 of 256 eligible invited participants).

All assessments (clinician-rated RISRAS and self-administered POMS-SF and SCS) were completed on the same day for each participant, during a scheduled treatment visit. Participants were assessed at varying points in their radiotherapy course, with a minimum of 2 weeks completed.

Data analysis

Data were analyzed using SPSS 27.0. Descriptive statistics (frequencies, percentages, means, and standard deviations) were used to summarize demographic and clinical characteristics. One-way ANOVA was used to compare social integration scores across different dermatitis severity groups. Pearson correlation analysis was employed to examine pairwise relationships among primary study variables. Patients with a RISRAS score of 0 (no dermatitis, n = 40) were excluded from the mediation analysis to focus specifically on the associations within the symptomatic range. Including asymptomatic patients could artificially inflate correlations and skew the distribution (zero-inflation).

Prior to conducting the mediation analysis, the key assumptions underlying ordinary least squares (OLS) regression were systematically evaluated.

• (1)Linearity: Scatter plots of standardized residuals against predicted values were examined for each regression model. No systematic curvilinear patterns were observed, supporting the assumption of linear relationships between predictors and outcomes.
• (2)Normality of residuals: The normality of residuals was assessed using the Shapiro–Wilk test and visual inspection of Q–Q plots. Although the Shapiro–Wilk test indicated slight deviations from normality for some models (P < 0.05), the Q–Q plots showed approximate normality. Importantly, the bias-corrected bootstrap method (5000 resamples) employed in the PROCESS macro does not rely on the assumption of normally distributed residuals, rendering the analysis robust to minor violations of this assumption.
• (3)Homoscedasticity: Plots of standardized residuals versus predicted values were inspected for each model. No obvious fan-shaped or funnel-shaped patterns were detected, suggesting that the homoscedasticity assumption was reasonably met.
• (4)Independence of errors: The Durbin–Watson statistic was calculated for each regression model. Values ranged from 1.82 to 2.14, all falling within the acceptable range (1.5–2.5), indicating no serious autocorrelation of residuals.
• (5)Absence of influential outliers: Cook's distance was computed for all cases. No observations exceeded the conventional threshold of 1.0 (maximum Cook's D = 0.087), suggesting that no single case exerted undue influence on the regression estimates.

Rigorous quality control measures were implemented during data collection to minimize missing data. As a result of on-site verification and immediate follow-up for incomplete responses, the final dataset of 223 valid questionnaires contained no missing values on the key study variables (RISRAS, POMS-SF, and SCS). Nine questionnaires with incomplete responses that could not be resolved on-site were excluded prior to analysis. Therefore, complete case analysis was conducted, and no imputation methods were required. We acknowledge that the exclusion of these nine incomplete cases (3.0% of approached eligible participants) represents a potential, though minimal, source of bias. A sensitivity analysis retaining all 223 participants (including those with RISRAS = 0) was conducted to assess the robustness of the findings to the exclusion of asymptomatic patients (Supplementary File 2).

To test the hypothesized indirect associations, we employed the SPSS PROCESS macro (Version 4.1, Model 4) developed by Hayes.²⁵ A parallel mediation model was specified as follows. Independent variable (X): Radiation dermatitis severity (RISRAS total score, continuous); Dependent variable (Y): Social integration (SCS total score, continuous); Mediator 1 (M1): Positive emotions (POMS-SF positive composite score, continuous); Mediator 2 (M2): Negative emotions (POMS-SF negative composite score, continuous). The parallel mediation model estimates the specific indirect effect of each mediator while simultaneously controlling for the other, thereby isolating the unique contribution of each emotional pathway. Three regression equations were estimated: (1) X predicting M1, (2) X predicting M2, and (3) X, M1, and M2 jointly predicting Y. All covariates were entered into all three equations simultaneously.

Statistical inference for indirect effects was based on the bias-corrected bootstrap method with 5000 resamples, generating 95% confidence intervals (CI). An indirect effect was considered statistically significant when the 95% CI did not contain zero. This approach was selected because it does not assume normality of the sampling distribution of the indirect effect and provides more accurate Type I error rates and statistical power compared to the Sobel test, particularly in small-to-medium samples.²⁵ A pairwise contrast of the two specific indirect effects was also computed to determine whether the two pathways differed significantly in magnitude.

Several potential confounding variables were included as covariates in the model based on both theoretical considerations and empirical evidence.

• (a)Demographic variables (age, gender, marital status, education level): These were selected based on prior literature demonstrating their associations with mood states and social functioning in cancer populations.^15,26 Age and gender are established predictors of emotional distress in oncology patients. Marital status serves as a proxy for structural social support, which may independently influence both mood and social integration. Education level is associated with coping resources and health literacy, potentially affecting emotional regulation and social engagement.
• (b)Clinical variables (radiation dose, concurrent chemotherapy): These were included because they represent physiological factors within the TOUS framework that may independently influence both dermatitis severity and patient outcomes. Higher radiation doses are associated with more severe skin reactions,³ and concurrent chemotherapy may exacerbate both physical symptoms and psychological distress.² Including these variables helps isolate the specific indirect association between dermatitis severity and social integration through mood states, beyond the confounding influence of treatment intensity.

We note that the prescribed dose to GTV (dichotomized at ≥ 65 Gy vs. < 65 Gy) reflects planned treatment intensity rather than cumulative radiation exposure at the time of assessment. Ideally, cumulative dose received or number of fractions completed at assessment would serve as more precise covariates; however, cumulative dose at assessment was not systematically recorded in accessible medical records. We conducted a supplementary analysis including radiotherapy week at assessment as an additional covariate, which did not substantively alter the mediation findings.

The selection of covariates was further guided by the principle of parsimony: only variables with established theoretical or empirical links to both the predictor (dermatitis severity) and the outcome (social integration) were included to avoid overfitting the model given the sample size.

Variance Inflation Factors (VIF) and tolerance values were calculated for all predictors in each of the three regression models (Model 1: Positive Emotion; Model 2: Negative Emotion; Model 3: Social Integration). Across all models, VIF values ranged from 1.05 to 2.38, well below the conventional threshold of 5.0 (or the more conservative threshold of 3.0), and all tolerance values exceeded 0.40. Additionally, the condition indices for all models were below 15, further confirming the absence of problematic multicollinearity. These results indicate that the regression coefficients were not unduly inflated by collinearity among predictors, supporting the stability and interpretability of the estimated effects. All statistical tests were two-tailed, with a significance level (α) set at 0.05. The normality of continuous variables was assessed using the Shapiro–Wilk test. Although some variables deviated from a strict normal distribution, the bias-corrected bootstrap method (5000 samples) was employed. This method does not rely on the assumption of normality for the sampling distribution of the indirect effect, ensuring robust results. Given that the mediator (POMS-SF) and outcome (SCS) variables were assessed via self-report at a single time point, common method bias may inflate their inter-correlation. However, the predictor variable (RISRAS) combines patient-reported symptoms with clinician-rated skin findings, which partially mitigates CMB for the X→M and X→Y paths. Nevertheless, CMB remains a concern for the M→Y association. First, Harman's single-factor test was conducted: an exploratory factor analysis with all scale items entered simultaneously (unrotated) revealed that the first factor accounted for 32.5% of the total variance, below the conventional 40% threshold. Second, we examined the correlation matrix for evidence of excessively high correlations (r > 0.90) among study variables, which would suggest CMB; the highest observed correlation was r = 0.59, well below this threshold. Together, these results suggest that CMB was not a pervasive issue in the present dataset, though we acknowledge that these methods provide only approximate assessments and cannot entirely rule out shared method variance.

Results

Participant characteristics and correlation analysis

A total of 294 patients were screened, of whom 38 were ineligible and 24 declined participation. After excluding 9 incomplete questionnaires, 223 valid cases were included (participation rate: 87.1% of eligible invitees). Among the total sample, 183 patients (82.1%) developed radiation dermatitis and constituted the analytical sample for mediation analysis. The participant flow is presented in Fig. 1. The majority of participants received IMRT or VMAT, while the remainder received TOMO. The demographic and clinical characteristics of the full enrolled sample (n = 223) are presented in the upper section of Table 1. Descriptive statistics for key study variables (dermatitis severity, mood composites, and social integration) are reported in the lower section of Table 1 for the analytical subsample with radiation dermatitis (n = 183). A comparison of baseline characteristics between patients with and without dermatitis is presented in Supplementary File 3. No statistically significant differences were observed in sociodemographic characteristics between the two groups. However, clinical differences may exist, and the absence of demographic differences does not establish full representativeness of the analytical subsample.

Fig. 1.

Participant flowchart.

Table 1.

Socio-demographic and clinical characteristics of the full enrolled sample (N = 223) and key study variable scores for the analytical subsample with radiation dermatitis (N = 183).

Characteristics	Category	n (%)
Age (year), mean ± SD		56.43 ± 11.84
Sex	Male	116 (52.01)
	Female	107 (47.99)
Marital status	Married	140 (62.78)
	Single/Divorced/Widowed	83 (37.22)
Educational level	College degree or above	66 (29.60)
	High school and vocational school	74 (33.18)
	Junior high school and below	83 (37.22)
Residence	City	142 (63.68)
	Rural	81 (36.32)
Occupational status	In-service	86 (38.57)
	Retirement	50 (22.42)
	Unemployed	87 (39.01)
Smoking	Yes	166 (74.44)
	No	57 (25.56)
Diabetes	Yes	65 (29.15)
	No	158 (70.85)
Radiotherapy method	IMRT	87 (39.01)
	VMAT	80 (35.87)
	TOMO	56 (25.11)
Radiotherapy dose (Prescribed dose to GTV)	< 65 Gy	147 (65.92)
	≥ 65 Gy	76 (34.08)
Concurrent chemotherapy	Yes	136 (60.99)
	No	87 (39.01)
Dermatitis	Yes (included in mediation analysis)	183 (82.06)
	No	40 (17.94)
Primary tumor site
Nasopharynx		89 (39.91)
Oral cavity		38 (17.04)
Larynx		42 (18.83)
Oropharynx		27 (12.11)
Hypopharynx		15 (6.73)
Nasal cavity/paranasal sinuses		12 (5.38)
Clinical stage (AJCC 8th ed.)
Stage I		24 (10.76)
Stage II		58 (26.01)
Stage III		87 (39.01)
Stage IVA/B		54 (24.22)
Previous surgery
Yes		76 (34.08)
No		147 (65.92)
Treatment intent
Curative (definitive/adjuvant)		223 (100)
Radiotherapy week at assessment, median (IQR)		4 (3–5)
Fractions completed at assessment, median (IQR)		18 (14–24)
Key variables (score)		Mean ± SD
Dermatitis severity (RISRAS)		12.73 ± 5.05
Negative mood composite (POMS-SF)		59.33 ± 5.47
Positive mood composite (POMS-SF)		23.64 ± 4.44
Social integration (SCS)		22.48 ± 4.05

Notes: IMRT = Intensity-Modulated Radiation Therapy; VMAT = Volumetric Modulated Arc Therapy; TOMO = Tomotherapy; RISRAS = Radiation-Induced Skin Reaction Assessment Scale; POMS-SF = Profile of Mood States-Short Form; SCS = Social Cohesion Scale. Key variable scores are reported for the analytical subsample (n = 183) only. Socio-demographic and clinical characteristics (upper section) are reported for the full enrolled sample (n = 223). Key variable scores (lower section: RISRAS, POMS-SF composites, and SCS) are reported exclusively for the analytical subsample of patients with radiation dermatitis (n = 183) who were included in the mediation analysis.

To descriptively illustrate the association between dermatitis severity and social integration, patients with radiation dermatitis (n = 183) were categorized into three groups based on RISRAS total scores: mild (< 8, n = 35), moderate (8–14, n = 78), and severe (> 14, n = 70). These thresholds were based on approximate tertile divisions of the RISRAS clinical score range (0–28) to create clinically distinguishable severity levels and have not been formally validated for the Chinese version of RISRAS or for HNC patients specifically. This analysis serves as a supplementary descriptive illustration of dose–response trends and was not used to test the study hypotheses.

Levene's test indicated that the assumption of homogeneity of variances was satisfied [F (2, 180) = 0.84, P = 0.432]. A one-way ANOVA revealed statistically significant differences in social integration scores among the three groups [F (2, 180) = 15.42, P < 0.001, η² = 0.146]. Post-hoc pairwise comparisons with Bonferroni correction yielded the following results: patients with severe dermatitis reported significantly lower social integration scores (20.7 ± 3.9) compared to those with mild dermatitis (25.3 ± 4.4; mean difference = 4.6, 95% CI: 2.58 to 6.62, P < 0.001) and moderate dermatitis (22.7 ± 4.0; mean difference = 2.0, 95% CI: 0.39 to 3.61, P = 0.009). The difference between the mild and moderate groups was also statistically significant (mean difference = 2.6, 95% CI: 0.61 to 4.59, P = 0.005).

Correlation analysis

Descriptive statistics and Pearson correlation analyses were conducted for all key variables (n = 183). The severity of radiation dermatitis was significantly negatively correlated with social integration (r = −0.41, P < 0.01) and the positive mood composite (r = −0.59, P < 0.01), while being significantly positively correlated with the negative mood composite (r = 0.41, P < 0.01). Furthermore, both the negative mood composite (r = −0.52, P < 0.01) and the positive mood composite (r = 0.49, P < 0.01) were significantly correlated with social integration. Additionally, the positive and negative mood composites showed a moderate negative correlation (r = −0.38, P < 0.01), supporting their treatment as distinct but related constructs. Variance Inflation Factors (VIF) for all predictors across the three regression models ranged from 1.05 to 2.38 (all < 3.0), and condition indices were below 15, confirming the absence of problematic multicollinearity (detailed diagnostics reported in Methods section 2.4).

Mediation analysis

All regression assumptions were assessed prior to analysis and were satisfactorily met. The results of the parallel mediation analysis are presented below. To examine the mediating roles of positive and negative emotions simultaneously, a parallel mediation analysis was conducted using the SPSS PROCESS macro (Model 4). Several potential confounding variables, including age, gender, marital status, education level, radiation dose, and concurrent chemotherapy, were controlled for in the model. As recommended by Hayes, 5000 bootstrap samples were used to generate stable 95% confidence intervals (CI) for the indirect effects.

Total and direct effects

The results revealed a significant total effect of radiation dermatitis severity on social integration (B = −0.329, SE = 0.054, P < 0.001). However, after introducing the mediating variables (positive and negative emotions), the direct effect of dermatitis severity on social integration was no longer significant (B = −0.099, SE = 0.060, P = 0.101). This pattern of results is consistent with a statistical model in which the association between dermatitis severity and social integration is statistically accounted for by the inclusion of mood state variables in the model. However, given the cross-sectional design, this finding should be interpreted as evidence of significant indirect associations rather than confirmation of a causal mediation mechanism. The overall model (including dermatitis severity, both mediators, and covariates) explained 42.5% of the variance in social integration (R² = 0.425, F = 12.68, P < 0.001). This R² reflects the combined contribution of all predictors (dermatitis severity, both mood composites, and six covariates) and should not be attributed to the indirect pathways alone. As a descriptive summary, the total indirect effect constituted approximately 69.9% of the total association. However, this proportion should be interpreted with caution: proportion-mediated estimates are known to be unstable in cross-sectional observational models, no confidence interval is available for this ratio, and its magnitude may be inflated by unmeasured confounding or model misspecification. The regression results are summarized in Table 2.

Table 2.

Regression results for the parallel mediation model (N = 183).

Predictors	Model 1: Positive mood composite (M1)				Model 2: Negative mood composite (M2)				Model 3: Social integration (Y)
	B (SE)	β	t	P value	B (SE)	β	t	P value	B (SE)	β	t	P value
Constant	28.95 (2.29)∗∗∗	–	12.65	< 0.001	55.97 (3.20)∗∗∗	–	17.48	< 0.001	32.08 (4.35)∗∗∗	–	7.37	< 0.001
Dermatitis severity (X)	−0.53 (0.05)∗∗∗	−0.60	−9.77	< 0.001	0.43 (0.08)∗∗∗	0.39	5.69	< 0.001	−0.10 (0.06)	−0.12	−1.64	0.101
Positive mood composite (M1)	–	–	–	–	–	–	–	–	0.23 (0.07)∗∗	0.25	3.27	0.001
Negative mood composite (M2)	–	–	–	–	–	–	–	–	−0.25 (0.05)∗∗∗	−0.34	−5.11	< 0.001
Covariates
Age	0.03 (0.03)	0.07	1.01	0.313	−0.00 (0.04)	−0.01	−0.01	0.993	−0.03 (0.02)	−0.08	−1.34	0.181
Sex	−0.40 (0.69)	−0.04	−0.58	0.561	−0.31 (0.96)	−0.02	−0.32	0.747	−0.90 (0.61)	−0.09	−1.49	0.138
Marital status	0.90 (0.56)	0.09	1.61	0.109	−1.55 (0.79)∗	−0.14	−1.97	0.049	1.50 (0.50)∗∗	0.17	3.01	0.003
Educational High school vs. Ref	0.35 (0.62)	0.04	0.56	0.574	−0.15 (0.87)	−0.01	−0.17	0.864	0.48 (0.55)	0.06	0.87	0.384
Education: College + vs. Ref	0.68 (0.66)	0.07	1.03	0.305	0.32 (0.92)	0.03	0.35	0.729	1.02 (0.58)	0.12	1.76	0.080
Radiotherapy dose	−0.06 (0.55)	−0.01	−0.11	0.912	0.45 (0.77)	0.04	0.57	0.563	0.49 (0.49)	0.06	1.01	0.313
Concurrent chemotherapy	−0.88 (0.57)	−0.09	−1.56	0.121	−0.37 (0.79)	−0.03	−0.46	0.643	0.11 (0.50)	0.01	0.21	0.835
R2	0.376				0.192				0.425
F	13.11∗∗∗				5.17∗∗∗				12.71∗∗∗

Notes: B = Unstandardized Coefficient. SE = Standard Error. β = Standardized Coefficient. Covariates were coded as: Sex (0 = Male, 1 = Female), Marital status (0 = Single/Divorced/Widowed, 1 = Married), Concurrent chemotherapy (0 = No, 1 = Yes), Radiotherapy dose (0 = < 65 Gy, 1 = ≥ 65 Gy). Age was entered as a continuous variable. Educational level was entered using two indicator variables: (1) High school/vocational school vs. reference, and (2) College degree or above vs. reference (reference category: Junior high school and below). Variance Inflation Factors (VIF) for all predictors ranged from 1.05 to 2.38 (all < 3.0). Durbin–Watson statistics: Model 1 = 1.92, Model 2 = 2.14, Model 3 = 1.82. Indirect effects were estimated using 5000 bias-corrected bootstrap resamples. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001.

Indirect effects

Dermatitis severity was significantly negatively associated with the positive mood composite (B = −0.529, P < 0.001), which in turn was positively associated with social integration (B = 0.228, P = 0.001). The indirect effect through the positive mood composite was significant (indirect effect = −0.120, 95% CI: −0.221 to −0.031). Dermatitis severity was significantly positively associated with the negative mood composite (B = 0.432, P < 0.001), which in turn was negatively associated with social integration (B = −0.253, P < 0.001). The indirect effect through the negative mood composite was also significant (indirect effect = −0.109, 95% CI: −0.200 to −0.047). Results are presented in Table 3 and Fig. 2.

Table 3.

Total, direct, and indirect effects of dermatitis severity on social integration.

Effect pathway	Effect	BootSE	Boot 95% CI (LL, UL)	Descriptive proportion of total association (%)
Total Effect(c)	−0.329	0.054	(−0.436, −0.223)	–
Direct Effect(c')	−0.099	0.06	(−0.219, 0.020)	30.10% (ns)
Total indirect Effect(ab)	−0.23	0.057	(−0.352, −0.131)	69.90%
Specific indirect effects
Via positive Emotion (a1b1)	−0.12	0.049	(−0.221, −0.031)	36.50%
Via negative Emotion (a2b2)	−0.109	0.039	(−0.200, −0.047)	33.10%
Pairwise contrast
Positive minus negative (a1b1-a2b2)	−0.011	0.068	(−0.141, 0.127)	–

Notes: Bootstrap sample size = 5000. LL = Lower Limit; UL = Upper Limit; ns = not significant. The “Descriptive Proportion of Total Association” is calculated as (specific indirect effect / total effect) × 100% and is reported for descriptive purposes only. This proportion is not a formal effect size, has no associated confidence interval, and should not be used for statistical inference. Proportion-mediated estimates are known to be unstable in cross-sectional observational models and may be influenced by unmeasured confounding or model misspecification. The confidence interval for the pairwise contrast (Positive minus Negative) includes zero, indicating that no statistically significant difference was detected between the two estimated indirect effects; this does not establish equivalence.

Fig. 2.

Parallel Mediation Model. X = Radiation dermatitis severity (RISRAS total score); M1 = Positive Mood Composite (POMS-SF positive composite); M2 = Negative Mood Composite (POMS-SF negative composite); Y = Social Integration (SCS total score). The analysis controlled for age, gender, marital status, education level, radiation dose, and concurrent chemotherapy. Solid lines indicate significant paths. Dashed lines indicate non-significant paths. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ns = not significant.

Comparison of mediation effects

A pairwise contrast analysis was performed to compare the strength of the two indirect effects. The difference between the indirect effect of positive emotions and negative emotions was not statistically significant (contrast = −0.011, 95% CI: −0.141 to 0.127), indicating that no statistically significant difference was detected between the magnitudes of the two specific indirect effects. This does not establish equivalence, as the analysis was not designed as an equivalence test. Among the covariates, marital status was a statistically significant predictor, with married patients reporting higher levels of social integration (B = 1.50, P = 0.003) and lower negative emotions (B = −1.55, P = 0.049). However, it is important to note that marital status served only as a proxy for structural social support in this study. It does not capture the quality of the marital relationship, which can vary significantly and influence emotional outcomes.

Sensitivity analysis

To evaluate the robustness of the primary findings, a sensitivity analysis was conducted including all 223 participants (i.e., retaining the 40 patients with RISRAS = 0). The parallel mediation model was re-estimated with the same covariates. The results were consistent with the primary analysis: dermatitis severity was significantly negatively associated with positive emotions (B = −0.468, P < 0.001) and significantly positively associated with negative emotions (B = 0.389, P < 0.001). The indirect effects via both positive emotions (Effect = −0.105, 95% CI: −0.198 to −0.027) and negative emotions (Effect = −0.098, 95% CI: −0.178 to −0.039) remained statistically significant. The direct effect of dermatitis severity on social integration remained non-significant after accounting for the mediators (B = −0.085, P = 0.128). The total indirect effect was −0.203 (95% CI: −0.312 to −0.110), and the overall model explained 39.8% of the variance in social integration (R² = 0.398, F = 14.02, P < 0.001). The pairwise contrast between the two indirect effects was not significant (Contrast = −0.007, 95% CI: −0.125 to 0.118), consistent with the primary analysis. The consistency of these results with the primary analysis (n = 183) suggests that the exclusion of asymptomatic patients did not substantively alter the pattern of findings (see Supplementary Document 2 for full regression results).

Discussion

Main findings

The objective of this study was to examine the psychological pathways through which radiation dermatitis severity may be associated with social integration among patients with head and neck cancer. Using a parallel mediation model controlling for key physiological and situational covariates, this study identified statistically significant indirect associations through mood states. The findings suggest that the statistical association between dermatitis severity and social integration is accounted for by two distinct mood-related pathways: the intensification of negative emotions and the attenuation of positive emotions. No statistically significant difference was detected between the magnitudes of the two indirect effects.

It is important to distinguish statistical mediation from causal mediation. Statistical mediation demonstrates that the inclusion of mood state variables accounts for the previously significant association between dermatitis severity and social integration—a necessary but not sufficient condition for causal inference.²⁵ The cross-sectional design precludes establishing temporal ordering or excluding all confounding variables. Therefore, the observed indirect associations should be interpreted as plausible pathways warranting longitudinal confirmation rather than established causal mechanisms. Bidirectional relationships—for example, social isolation exacerbating negative mood, which in turn heightens symptom perception—cannot be excluded.

The confirmation of the two-factor structure in our CFA results reinforces the theoretical premise that dermatitis is associated with social integration through two distinct mood-related pathways simultaneously. By aggregating specific subscales into broad positive and negative dimensions based on the Two-Factor Theory of Affect,¹⁰ this study highlights that the depletion of positive affect is as relevant as the elevation of negative affect—a finding that might have been obscured if analyzing individual subscales separately.

Our findings are consistent with recent systematic reviews highlighting the impact of cancer treatment on social isolation and social connectedness among survivors of cancer.²⁷ The negative emotion pathway aligns with prior psycho-oncology literature demonstrating that emotional distress mediates symptom-functioning relationships.^26,28,29 This pathway may be particularly pronounced in HNC, where dermatitis occurs on visible areas integral to identity and communication, making patients vulnerable to heightened self-consciousness and avoidance behaviors.^7,30 However, alternative explanations warrant consideration: the association may partially reflect internalized stigma or anticipated social rejection,¹² concurrent symptoms (mucositis, dysphagia) that restrict social participation through functional pathways independent of mood,^2,4 personality traits predisposing both heightened symptom perception and negative mood, or the cumulative psychological burden of cancer treatment. Although we controlled for radiation dose and chemotherapy status, these covariates may not fully capture the totality of symptom burden. Future studies should incorporate measures of stigma, overall symptom burden, personality traits, and treatment-related distress to disentangle these potential confounders.

The positive emotion pathway aligns with the Broaden-and-Build Theory,¹¹ which posits that positive emotions broaden thought-action repertoires and build resources for social engagement. Growing evidence supports the applicability of this theory in cancer survivorship, suggesting that positive affect is essential for resilience and social re-engagement during chronic illness.³¹ Similarly, the depletion of positive emotions may reflect cumulative physical debilitation coinciding with worsening dermatitis, or reduced functional independence during treatment, rather than a specific response to visible skin changes alone. These possibilities underscore the need for longitudinal designs that can track the temporal evolution of dermatitis, positive affect, and social engagement across the radiotherapy trajectory.

These findings extend the Theory of Unpleasant Symptoms (TOUS) by suggesting that psychological “influencing factors” may function as intermediary variables comprising two distinct pathways—the accumulation of distress and the depletion of vitality—rather than as a monolithic moderating construct.¹³ The parallel mediation structure suggests that clinical approaches may need to address both pathways concurrently to support social functioning. However, alternative models deserve consideration, including a common cause model (unmeasured third variables driving both mood and social outcomes), a feedback loop model (social withdrawal reducing access to support, worsening mood), and body image disturbance as a more proximal mediator.¹² These alternatives cannot be adjudicated with cross-sectional data alone and represent important directions for future longitudinal, experimental, or qualitative research.

Implications for nursing practice and research

The observed statistical associations suggest—but do not confirm—that mood states may represent a relevant pathway linking dermatitis severity to social functioning. All clinical implications below are hypothesis-generating and require validation through prospective intervention studies.

First, these findings suggest that routine brief mood screening (e.g., the POMS-SF or Distress Thermometer)^19,20 beginning at the onset of visible dermatitis (typically week 2–3 of treatment) may help identify patients at risk of social withdrawal. Patients presenting with elevated negative emotions or depleted positive emotions may warrant further psychosocial assessment, though optimal screening thresholds require empirical validation.

Second, the dual-pathway findings suggest a conceptual framework for future intervention development. For the negative mood pathway, potential approaches warranting investigation include brief cognitive-behavioral therapy adapted for appearance-related distress and relaxation techniques that could be delivered by trained oncology nurses.³² For the positive mood pathway, behavioral activation, peer support programs, and mindfulness-based stress reduction merit future evaluation.³³ Whether such approaches improve social integration in patients with HNC remains to be determined through randomized controlled trials.²⁸

Third, our findings do not diminish the importance of evidence-based dermatitis management. Rather, they suggest that optimal care may benefit from integrating brief psychosocial assessment into existing radiotherapy nursing workflows alongside dermatological care,^4,6 without substantial additional time burden. Practically, this means that nursing assessment of radiation dermatitis could extend beyond grading skin reactions to include brief inquiry about emotional well-being and social participation.

These recommendations should be considered hypotheses for future research rather than evidence-based guidelines.

Limitations

Several limitations should be noted. First, the sample size was calculated using a prevalence-based formula¹⁵ rather than a method designed for mediation analysis, as no prior effect size estimates for the specific indirect pathways were available at the design stage. Although the analytical sample (n = 183) exceeds the threshold recommended by Fritz and MacKinnon for detecting medium indirect effects via bias-corrected bootstrap (n ≈ 148),¹⁶ replication in larger, multi-center cohorts with a priori mediation-specific power analyses is warranted.

Second, the exclusion of patients without dermatitis (n = 40) may introduce interval truncation bias, though the sensitivity analysis including all 223 participants yielded consistent results (Supplementary File 2).

Third, the cross-sectional design precludes establishing temporal precedence, controlling for all potential confounders, or ruling out reverse causation. The directional pathways proposed should be interpreted strictly as associative patterns consistent with our theoretical framework (TOUS),¹³ not as confirmed causal mechanisms. Future research should employ prospective longitudinal designs with at least three measurement points to establish temporal precedence, using cross-lagged panel models or latent growth curve models.²⁵

Fourth, as a single-center study at a tertiary hospital in northeastern China, generalizability is limited by institutional, regional, cultural, and health care system factors. Chinese cultural values emphasizing collectivism and “face” (mianzi) may uniquely influence how patients experience and respond to visible disfigurement. The gender distribution (52% male) differs from global HNC registries where male-to-female ratios typically range from 2:1 to 4:1, possibly reflecting institutional case-mix.¹

Fifth, reliance on self-report measures may introduce common method bias (CMB), though Harman's single-factor test (first factor: 32.5% of variance) and the mixed-method nature of RISRAS¹⁷ partially mitigate this concern.

Sixth, several clinically important confounders were not adjusted for, including mucositis severity, dysphagia, speech impairment, nutritional status, performance status, and baseline social support quality.^2,4 These symptoms may independently impair social functioning through functional pathways that do not operate through mood states. The observed associations may partially reflect broader treatment burden rather than the specific pathway from visible dermatitis through mood to social outcomes. Future studies should incorporate comprehensive symptom assessments and social support measures.³⁰

Finally, the Social Cohesion Scale yielded modest internal consistency (α = 0.710) and was developed for community populations.^23,24 Future studies should consider disease-specific measures and objective indicators of social functioning (e.g., frequency of social outings, return-to-work status).

Conclusions

This cross-sectional study found that the statistical association between radiation dermatitis severity and social integration among patients with head and neck cancer was no longer significant after accounting for two co-occurring mood-related pathways: the intensification of negative emotions and the depletion of positive emotions. Both indirect associations were statistically significant, and no detectable difference in their magnitudes was observed, though this does not confirm equivalence between the two pathways. This indirect-only pattern is consistent with the hypothesized model derived from the Theory of Unpleasant Symptoms but does not establish that mood states are the sole or causal pathway, given the cross-sectional design. These findings suggest that routine screening for emotional distress using validated tools may be valuable for early identification of patients at risk of social withdrawal, and that clinical approaches may benefit from addressing both negative and positive emotional dimensions alongside physical symptom management. Potential intervention strategies—such as cognitive-behavioral techniques for negative emotions and behavioral activation or mindfulness-based approaches for positive emotions—represent directions for future feasibility and intervention research rather than evidence-based recommendations from the present study. Prospective longitudinal studies and randomized controlled trials are needed to confirm the directionality of these associations and to evaluate whether targeting mood states during radiotherapy causally improves social integration outcomes in this population.

CRediT authorship contribution statement

Yike Wang: Conceptualization, Data curation, Investigation, Methodology, Project administration, Writing - original draft, Writing - review & editing. Yanzhongcheng Xu: Conceptualization, Data curation, Investigation, Methodology, Project administration, Writing - original draft, Writing - review & editing. Meiyan Ji: Conceptualization, Formal analysis, Methodology, Project administration, Resources, Software, Supervision, Validation. Xinghua Bai: Conceptualization, Supervision, Writing - review & editing. All authors have read and approved the final manuscript.

Ethics statement

This study was approved by the Ethics Committee of The First Hospital of China Medical University (Approval No. : 202374). The procedures were conducted in accordance with the Declaration of Helsinki, and written informed consent was obtained from all participants.

Data availability statement

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

Declaration of generative AI and AI-assisted technologies in the writing process

No AI tools/services were used during the preparation of this work.

Funding

This study received no external funding.

Declaration of competing interest

The authors declare no conflict of interest.

Appendix. Supplementary materials

The following are the Supplementary data to this article: