Predictive Factors of Crucial Nutrition Impact Symptom Clusters in Patients With Head and Neck Cancer Undergoing Radiotherapy

Yujie Wang Peking University Health Science Centre https://orcid.org/0000-0002-7299-875X Lichuan Zhang Peking University School of Nursing Bing Zhuang Peking University School of Nursing Tong Zhang Peking University School of Nursing Sanli Jin Peking University School of Nursing Zhou Huang Peking University Cancer Hospital: Beijing Cancer Hospital Dan Zhao Peking University Cancer Hospital: Beijing Cancer Hospital Baomin Zheng Peking University Cancer Hospital: Beijing Cancer Hospital Shaowen Xiao Peking University Cancer Hospital: Beijing Cancer Hospital Liqing Gong Peking University Cancer Hospital: Beijing Cancer Hospital Yan Sun Peking University Cancer Hospital: Beijing Cancer Hospital Qian Lu (  luqian@bjmu.edu.cn ) Peking University Cancer Hospital: Beijing Cancer Hospital https://orcid.org/0000-0003-2611-3284


Introduction
According to the estimation from the Globocan 2020, more than seven hundred thousand patients got head and neck cancer (including lip cancer, salivary glands cancer, oropharynx cancer nasopharynx, and hypopharynx), which contributed to 3.9% of the new cases of all cancer [1]. Radiotherapy (RT) is a dominating therapy for HNC and nutritional problems are prevalent in patients receiving RT because of acute symptoms. About 62.3-71.8% of patients with HNC lose more than 5% weight at the end of RT [2,3], and 74.0-92.7% of them get moderate or severe malnutrition [4,5]. Critical weight loss and malnutrition have predictive value for patients' prognosis [6,7]. Studies have suggested that nutrition impact symptoms (NISs, e.g., dysphagia, mucositis, xerostomia) are key in uential factors of weight loss and malnutrition [3,8,9]. In addition, NIS could directly impact patients' quality of life (QoL) negatively [8]. Based on this, the management of NIS needs a high priority for improving HNC patients' nutritional status and QoL during RT.
Previous studies indicated that patients always experience multiple NISs during RT [3,10]. They may have similar occurrence mechanism and may be members of a symptom cluster which is de ned as " consisting of two or more symptoms that are related to each other and occur together" [11]. For example, dysphagia is related to edema and odynophagia induced by mucosal radiation injury [12]. Analysis of NISs clusters could help understand the intercorrelation between them and integrate interventions to promote the effectiveness of symptom management in HNC patients with RT. Several studies have explored symptom clusters in patients with HNC. They adopted the National Cancer Institute Common Toxicity Criteria or the M. D. Anderson Symptom Inventory, and identi ed clusters like HNC-speci c cluster (radiodermatitis, dysphagia, etc.), gastrointestinal cluster (dehydration, nausea, etc.), general cluster (numbness or tingling, etc.) and so on [13][14][15]. Nevertheless, these studies did not give speci c attention to NIS, which may have a more signi cant impact on patients' outcomes. We explored the NIS clusters in HNC patients undergoing concurrent chemoradiotherapy with the head and neck symptom checklist (HNSC)-a NIS speci c evaluation tool [16] and explored relationships between clusters and weight loss. But patients with RT with or without other therapy may have different NIS pro les. Also, QoL is an important patient-reported outcome and needs evaluation. Hence, it is necessary to identify the symptom pattern and explore the crucial NIS cluster that could in uence both patients' weight and QoL in HNC patients receiving RT, so that we can take holistic and targeted intervention and improve patients' outcomes with limited medical resources.
Exploring predictive factors of symptoms could help identify at-risk patients and perform an early intervention to alleviate the severity of NIS, improve patients' nutritional status, QoL and prognosis. However, few studies explore the predictive factors of symptom clusters in HNC patients. Xiao et al. [17] found that patients who were White and accepted a higher level of education were more likely to get severe severity of HNC-speci c symptom cluster. In contrast, female and never smoked patients could get more severe gastrointestinal symptom cluster. However, they only focused on patients' sociodemographic and clinical variables. Studies suggested that nutritional risk and nutritional status before treatment may also be associated with the severity of symptoms during treatment [18,19]. At present, there lacks evidence about predictive factors of NIS clusters in patients with HNC receiving RT.
Therefore, this study aimed to analyze the NIS clusters in HNC patients receiving RT using a prospective longitudinal design. Then, we identi ed crucial clusters correlated with weight loss and QoL. Also, we attempted to explore predictive factors of the crucial NIS clusters, including sociodemographic factors, disease-related characteristics, nutritional risk identi ed by the Nutritional Risk Screening 2002 (NRS2002), and nutritional status identi ed by the Global Leadership Initiative on Malnutrition (GLIM), to direct future management of HNC patients.

Study design and participants
This was a prospective longitudinal study. Subjects who were pathologically diagnosed with HNC and scheduled for RT from March 2017 to December 2019 were recruited from one cancer hospital in Beijing. The other recruitment criteria were: (1) age ≥ 18 years; (2) entirely voluntary participation; (3) ability to communicate clearly. Considering factor analysis, the number of subjects should be ve to ten times the number of items on the scale. A total of 17 symptoms were assessed in this study, so 85 to 170 subjects were necessary [20].
All subjects were informed about the study, including the purpose and content of this study prior to the RT. This study had got approval from the Ethics Committee of the author's University (No. IRB00001052-17002), and the procedure adhered to the principles in the Declaration of Helsinki.
De ne the crucial NIS cluster In this study, we de ned a crucial NIS cluster as the cluster that had signi cantly correlation with both weight loss rate (WLR) and QoL.

General and clinical information
We used a questionnaire to record the demographic, sociology, and clinical information at T 1 (age, gender, marital status, tumor site, tumor stage, and RT type).
Nutritional risk and nutritional status before RT Nutritional risk was assessed by the NRS 2002 [21]. This tool comprises of three parts: undernutrition, disease severity, and age. The undernutrition assesses three aspects: patients' body mass index (BMI), recent weight change, and food intake change. The disease severity is determined by the nutritional requirements. Each aspect will be classi ed as absent (Score = 0), mild (Score = 1), moderate (Score = 2), and severe (Score = 3).
Patients with score ≥ 3 had nutritional risk.
Nutritional status was evaluated by the GLIM which is newly proposed for diagnosing malnutrition [22]. Based on identi ed nutritional risk, the diagnosis criteria include three phenotypic criteria (non-volitional weight loss, low body mass index, and reduced muscle mass) and two etiologic criteria (reduced food intake or assimilation, and in ammation or disease burden). The muscle mass was measured by the InBody 120 (Biospace Co., Ltd, Seoul, South Korea) based on the bioelectrical impedance analysis. Appendicular skeletal muscle index (ASMI) and fat free mass index (FFMI) were used to de ne the low body mass index (male, ASMI 7 kg/m 2 or FFMI 17 kg/m 2 ; female, ASMI 5.7 kg/m 2 or FFMI 15 kg/m 2 for Asian) [22]. Patients with at least one phenotypic criterion and one etiologic criterion would be diagnosed with undernutrition.

Occurrence of nutrition impact symptoms
We adopted the Head and neck patient symptom checklist (HNSC) to evaluate 17 NIS (oral mucositis and dry mouth, etc.) [23]. This checklist assesses two dimensions: the severity and interference with dietary of symptoms using a ve-point Likert scale (1 = not at all to 5 = a lot). Jin et al. [24] had validated it in Chinese patients with HNC receiving RT. It had good psychometric performance and could be used to assess the NIS of patients with HNC.

Weight loss rate
Patients' weight was measured by the InBody120 device (Biospace Co., Ltd, Seoul, South Korea) when they wore light clothes and removed shoes. WLR was calculated using the equation: WLR= (baseline weight-present weight)/baseline weight×100%.

Global quality of life
The global QoL was evaluated with the global QoL subscales from the European Organization for Research and Treatment of Cancer (EORTC) Questionnaire-Core 30 (QLQ-C30) [25]. The global QoL dimension includes two items on global health and global life quality with a seven-point Likert scale. Higher scores indicated better global life quality.

Data collection
Trained research fellows collected data by face-to-face interviews with uniform instruction at three follow-up visits in the outpatient. Before RT (baseline, T 1 ), the general information was reported by subjects. Information on disease and treatment characteristics was determined from the medical records. The occurrence of NIS, weight, QoL, nutritional risk, and nutritional status were evaluated. At the third week (T 2 ) and the end of RT (T 3 ), occurrence of NIS, present weight and QoL were reassessed.

Data analysis
All data analysis was conducted using the IBM SPSS Statistics for Windows, version 24.0 (IBM Corp., Armonk, N.Y., USA). The NIS's interference had several missing values and was lled up by the mean of the nearby two points [26]. Counts and percentages were used to report categorical variables, while mean and standard deviation were used to report continuous variables. Exploratory factor analysis (EFA) with orthogonal rotation and principal component analysis was used to extracted symptom clusters. The individual measure of sampling adequacy (MSA) for each symptom was calculated to evaluate correlations between symptoms. We deleted symptoms with low MSA (< 0.5) as they could result in too many factors [27]. The average score of severity of NIS at T 2 and T 3 was used to conduct EFA. The number of factors was determined by the principle of eigenvalue > 1. Symptoms with weak factor loadings (< 0.40) were deleted. If symptoms had fair loadings on more than one factor, it was placed with the factor that it was mostly related to conceptually. [27] Symptom clusters were named and interpreted based on analysis results and theoretical considerations. Cronbach's α coe cient was adopted to assess clusters' internal consistency. The unweighted mean of symptoms was calculated to get the composite scores of clusters at T 2 and T 3 [27]. Generalized estimating equations (GEE) were adopted to explore the association between clusters and WLR or global QoL, and the predictive factors of symptom clusters. Variables with P < 0.1 in univariable models were entered into the multivariable models. As the nutritional risk and nutritional status before RT had a closed correlation with each other, so they would be placed in two multivariable GEE models separately (Model 1 and Model 2).

General information
There were 600 subjects receiving RT in the day-care ward. At baseline, a total of 537 eligible patients accepted the assessment. Finally, 334 subjects completed three follow-up visits. During the follow-up, 172 (32.0%) patients declined because of time con ict, severe toxicities, or exhausting visits. Besides, several subjects were excluded because of incomplete data or interrupted treatment (Fig. 1). The age of the 334 subjects was 53.5 ± 12.9 (18.0-88.0) years. They accepted one fraction of intensity-modulated radiation (IMRT) in each weekday. The planned RT schedule included 31.0 ± 2.2 (17-35) fractions with a total of 65.3 ± 5.0 (36-75) Gy. At T 2 , they received 12.38 ± 2.14 fractions RT, and 28.78 ± 2.68 fractions at T 3 . Table 1 displayed subjects' general information.

Occurrence of NIS and symptom clusters
We found that pain, taste change, dry mouth, thick saliva, and di culty swallowing were the top-ve frequent and serious symptoms during RT, while pain, oral mucositis, di culty swallowing, taste change, loss of appetite had the most impact on dietary intake in our population ( Table 2).
Before EFA analysis, we dropped diarrhea (MSA = 0.485) as weak correlation with other symptoms. After conducting EFA with the other 16 symptoms, we deleted the symptom of constipation as it had low factor loadings on all factors (< 0.40). As a result, 15 symptoms were retained in the next EFA analysis. The KMO test value was 0.770 and Bartelett's test of sphericity has statistical signi cance (χ 2 = 1727.947, P < 0.001). However, dry mouth and altered smell had loadings > 0.40 on two factors. Considering the factor loading and conceptual meaning, the dry mouth was placed with the rst factor, and the altered smell was placed with the third factor. Finally, we extracted four clusters that explained 60.536% of the variance, as shown in Table 3. The RT-speci c symptom cluster consisted of pain, di culty swallowing, oral mucositis, thick saliva, di culty chewing, and dry mouth (Cronbach's α = 0.813). The psychological status cluster consisted of anxiety, depression, and lack of energy (Cronbach's α = 0.745). The eating experience cluster consisted of altered smell, loss of appetite, taste change, and feeling full (Cronbach's α = 0.496). The upper gastrointestinal symptom cluster consisted of vomiting and nausea (Cronbach's α = 0.698). The composite scores of four symptom clusters were shown in Table 4.

Crucial NIS clusters
In the GEE models, the time point and four clusters were independent variables, while the WLR and global QoL were set as the dependent variable, respectively. The univariable model indicated that all the four clusters had a signi cant correlation with the WLR and global QoL. The multivariable GEE models were presented in Table 5. The RT-speci c symptom and upper gastrointestinal symptom clusters were signi cantly correlated with both WLR and global QoL, so they were de ned as crucial symptom clusters.

Predictive factors of crucial NIS clusters
As the had a signi cant correlation with WLR and global QoL; we analyzed their baseline predictive factors. In the GEE model, the independent variables included the time point, general information, nutritional risk, or nutritional status before RT. The RT-speci c symptom cluster and upper gastrointestinal symptom cluster were used as dependent variables separately. The univariable model showed that the RT-speci c symptom cluster was related to the time point, gender, age, baseline nutritional risk, and nutritional status; the upper gastrointestinal symptom cluster was associated with the time point, gender, age, tumor site, tumor stage, and RT type. Table 6

Discussion
NISs are symptoms that can signi cantly impact patients' dietary intake and metabolic absorption, which is related to patients' nutritional status and clinical outcomes. However, previous research attached limited importance to it. This prospective longitudinal study using the HNSC, the NIS-speci c tool, showed that patients with HNC experienced multiple aggravating NIS during RT, consistent with other investigations [9,28].
As a result, the management of NISs should be brought to the forefront.
Four NIS clusters were extracted with the HNSC: RT-speci c symptom cluster, psychological status cluster, eating experience symptom cluster and upper gastrointestinal symptom cluster. The RT-speci c symptom cluster comprised pain, di culty swallowing, oral mucositis, thick saliva, di culty chewing, and dry mouth. Direct and indirect radiation injury resulted in mucosa and tissue damage. Thus, patients commonly experience acute toxicities like mucositis, dry mouth, and soft tissue edema. Studies showed that mucositis and its pain were key determinants of dysphagia in patients with HNC accepting RT [12,29]. Also, xerostomia and thick saliva caused by the damaged salivary gland could impact swallowing function [30,31]. The eating experience cluster included altered smell, loss of appetite, taste change, and feeling full. Smell and taste function are important for avor awareness and food perception, so altered smell and taste could in uence patients' food enjoyment and interest in eating [32]. However, its Cronbach α coe cient (0.496) was relatively low. One possible explanation is the different anatomical positions of smell and taste receptors and the different tumor sites. A prospective study indicated that HNC patients receiving RT all experienced taste change, but only 30% of them got altered smell. [33] Besides, chemotherapy can damage taste and smell function. In our follow-up subjects, some received chemotherapy and they might have different symptoms pro les. [32] The RT-speci c symptom and gastrointestinal symptom clusters were de ned as crucial NIS clusters as they had signi cant relationships with both WLR and QoL. NISs within the former cluster were related to swallowing function and were prerequisite for good dietary intake and nutritional status [34,35]. For QoL, the RT-speci c symptom, gastrointestinal symptom, and psychological status cluster had a negative impact on it. On the one hand, the RT-speci c symptom and gastrointestinal symptom clusters are related to more WLR, while more WLR is related to worse QoL [36]. On the other hand, severe symptom burden could directly in uence patients' QoL [34,36]. Anxiety and depression are also prevalent problems in cancer patients. The study conducted by Hortense et al. indicated that anxiety and depression also had correlations with participants' QoL [37]. The RTspeci c symptom cluster and gastrointestinal symptom cluster were crucial clusters for managing HNC patients with RT, so healthcare professionals should focus limited medical resources on these NISs to improve patients' nutritional status and QoL. It is essential to explore their predictive factors for better symptom cluster management.
For the RT-speci c symptom cluster, patients who were older, female, diagnosed with oral cavity cancer, and had nutritional risk or malnutrition at T 1 had more probability to experience severe symptoms. Previous research showed that female patients tended to get more severe dysphagia and mucositis. [38, 39] But the study of Xiao et al. [17] in HNC patients with concurrent chemoradiotherapy did not nd the correlation between gender and severity of HNC-speci c symptom cluster. This may be due to different populations, evaluation instruments, symptoms within a cluster, and alternative predictive factors. A systematic review suggested that age, tumor stage, tumor site, therapy, and pretreatment oral function were related to oral toxicities [35]. A crosssectional study with 2248 cancer patients showed that patients with nutritional risk got signi cantly more severe adverse events. [19] In addition, a retrospective study indicated that pretreatment nutritional status could predict severe adverse events in HNC patients with RT [40]. Therefore, healthcare professionals should identify patients with higher risk of severe RT-speci c symptom cluster early and pay more attention to their management before RT. Especially, the pretreatment nutritional intervention should be put in the priority queue to reduce patients' nutritional risk and improve their nutritional status at T 1 . So that we could mitigate their NIS cluster burdens, improve nutritional status and QoL during RT.
For the upper gastrointestinal symptom cluster, patients who were female, accepted intensive therapy (concurrent chemoradiotherapy with or without surgery) were more likely to get severe upper gastrointestinal symptom cluster. Xiao et al. [17] also found that women were more likely to get severe gastrointestinal symptom clusters. RT and chemotherapy drugs could both conduct an impact on the central nervous system (dorsal vagal complex) and elevate levels of 5-hydroxytryptamine that involves in the generation of nausea and vomiting [41,42], so participants accepted concurrent chemoradiotherapy experienced severe upper gastrointestinal symptom clusters. As a result, clinical practitioners could identify patients at risk of severe upper gastrointestinal symptom cluster, and then provide early prevention measures for nausea and vomiting.
This study also had several limitations. Firstly, this was a monocentric study with a relatively high attrition rate, so this may limit the generalization of our ndings. In the future, investigators could perform a multicenter study with larger sample to validate the NIS clusters and their predictors. Secondly, this study only recruited part of alternative predictive factors and researchers can take more factors into account that may relate to NIS burden like chemotherapy regimens or pretreatment NIS. Thirdly, we excluded patients with tube feeding and total parental nutrition. Futural studies could explore their relationship and effect on NIS clusters, nutritional status, and QoL.

Conclusions
In conclusion, this study identi ed two crucial NIS clusters in HNC patients receiving RT. The RT-speci c symptom cluster and upper gastrointestinal symptom cluster both had a negative impact on WLR and QoL, so healthcare professionals should pay more attention to their management. For sociodemographic and clinical characteristics, female patients tended to report more severe symptom cluster burden. Patients at an older age and with oral cavity cancer were more likely to get severe RT-speci c symptom cluster, while patients accepted intensive therapy would experience more severe upper gastrointestinal symptom cluster. Thus, we should recognize patients at risk and intervene early. In addition, patients with nutritional risk and malnutrition had a higher risk of getting severe RT-speci c symptom cluster. Clinical practitioners should give early nutritional management before RT to improve HNC patients' NIS severity, nutritional status and QoL during treatment.
Declarations Fundings: This work was supported by the National Key Research and Development Project (2017YFC1309204) and the Nursing Research Project of Chinese Medical Association Publishing House (CMAPH-NRP2019002).

Con ict of interest:
None declares.
Availability of data and material: The corresponding author should be contacted directly for any queries related to the availability of data and material.
Code availability: Not relevant to this study. This study had got approval from the Ethics Committee of the author's University (No. IRB00001052-17002).

Consent to participate:
All subjects were informed about the study.

Consent for publication:
Not relevant to this study.      Figure 1 Patient eligibility within the studied cohort