Pretreatment Dietary Patterns Are Associated with the Presence of Nutrition Impact Symptoms 1 Year after Diagnosis in Patients with Head and Neck Cancer

Head and neck cancer accounted for an estimated 65,000 new diagnoses in men and women in the United States in 2019, resulting in roughly 14,260 deaths (https://www.cancer.net/cancer-types/head-and-neck-cancer/statistics). Head and neck cancer is a heterogeneous disease typically including epithelial malignancies of the oral cavity, oropharynx, hypopharynx, and larynx of the squamous cell histologic type (1). Head and neck cancer was historically associated with extensive exposure to tobacco and alcohol consumption. However, high risk human papillomavirus (HPV) has emerged as the primary etiologic factor for a subset of oropharyngeal tumors (2–4). Patients with head and neck cancer develop severe morbidities before and/or during treatment as a result of tumor location, treatment with radiotherapy, or surgical resection of the tumor (5). Many of these morbidities compromise food intake, and thus are termed nutrition impact symptoms (NIS).

Notably, at least 90% of head and neck cancer survivors develop NIS during the acute phase of treatment (6, 7). However, NIS that persist chronically (>6 months posttreatment) are understudied (5, 6, 8). Common NIS experienced by patients with head and neck cancer include trismus, xerostomia, dysphagia, difficulty chewing, taste alterations, and mucositis (5). One study conducted among patients with head and neck cancer reported that aggregate burden of NIS was a significant independent predicator of reduced food intake, weight loss, and survival (9). Other consequences of NIS include poor oral hygiene, prolonged eating time, disruption of relationships and social isolation, depression, and decreased quality of life (10). Because NIS burden can result in significantly reduced dietary intake and quality of life there is an urgent need for early and effective NIS prevention and intervention.

While previous work has established that the presence of NIS is associated with decreased food intake and weight loss (8, 11–14), no studies have examined how pretreatment dietary intake may influence the presence of NIS later in the disease trajectory. The pathogenesis of NISs are complex and differ depending on the symptom, but generally share one common mechanism—cell damage due to inflammation (15, 16). Our research team has previously reported a whole foods pretreatment dietary pattern, characterized by high intakes of vegetables, fruits, poultry, legumes, fish, wine, and whole grains, to be associated with lower head and neck cancer recurrence and mortality (17), as well as decreased markers of systemic inflammation, specifically TNFα, IL6, and IFNγ (18). Diet has the potential to reduce inflammation and affect biological processes involved in the pathogenesis of symptoms common in patients with head and neck cancer (18). Many nutrients and phytochemicals found in foods have been long known to have anti- or proinflammatory properties (19, 20). It is possible consumption of foods abundant in nutrients that modulate inflammation prior to treatment may influence the development and/or severity of NIS throughout the disease trajectory. Thus, the objective of this secondary analysis of a longitudinal cohort was to determine whether pretreatment dietary patterns are associated with the presence of self-reported NIS 1 year after diagnosis. The hypothesis was that a dietary pattern characterized by foods with anti-inflammatory properties (e.g., fruits, vegetables, whole-grains, low-fat dairy, and less saturated fat) before treatment would be associated with lower risk of self-reported chronic NIS. On the other hand, we hypothesized that a dietary pattern characterized by foods with proinflammatory properties (e.g., red and processed meats, fried foods) before treatment would be associated with higher risk of self-reported chronic NIS.

This secondary analysis included 336 patients with head and neck cancer enrolled in the University of Michigan Head and Neck Specialized Program of Research Excellence (UM HN-SPORE) prospective cohort study. The independent variable of interest included dietary patterns at diagnosis. Variables controlled for (covariates) were age, tumor site, cancer stage, smoking status, body mass index (BMI), calories, and HPV status. The dependent variables were individual and aggregated NIS 1 year postdiagnosis.

Between November 2008 and July 2013, all patients who were newly diagnosed with a previously untreated, primary squamous cell carcinoma of the oral cavity, oropharynx, hypopharynx, or larynx were invited to participate in the UM HN-SPORE prospective cohort study. Patients were recruited from UM otolaryngology, radiation oncology, and dental clinics. Institutional review board approval was obtained from the UM Health System (Ann Arbor, MI). Exclusion criteria included: (i) less than 18 years of age; (ii) pregnancy; (iii) non-English speaking; (iv) diagnosis of mental instability; (v) diagnosis of another non-upper aero-digestive tract cancer, or (vi) non-squamous cell carcinoma. All participants of the original cohort who had completed a pretreatment (i.e., after diagnosis but prior to starting treatment) food frequency questionnaire (FFQ) matched with 1-year self-reported NIS data were included in this analysis. No exclusions were made based on primary treatment goal or modality because any instance of being bothered by NIS is important to consider for palliative and supportive care, regardless of whether or not a patient is being treated with curative intent.

Of the original 1,031 participants enrolled, 489 had complete pretreatment FFQ data. Participants were excluded if they were missing 1-year NIS data (n = 137) or data on key covariates planned to be used in multivariable models (n = 3 missing BMI). Participants were also excluded if they reported and estimated daily energy intake of <500 (kcals)/day or >5,000 kcals/day (n = 13). Any reported energy intake <500 or >5,000 kcals/day is considered biologically implausible and thus these observations are likely unreliable (21, 22). The final sample size included 336 participants.

Participants completed a self-administered epidemiologic health questionnaire at baseline (i.e., after diagnosis but prior to starting treatment) that included data on demographic, clinical, and behavioral characteristics including tobacco, alcohol, physical activity, sleep, comorbidities, depression, and quality of life. Dietary data were obtained at diagnosis using the self-administered 2007 Harvard FFQ (23). An electronic medical record review was conducted for each participant to collect data on tumor site, cancer stage, and treatment modalities.

Usual dietary intake over the past year was estimated using the 131-item self-administered, semi-quantitative 2007 Harvard FFQ, a valid and reproducible measure of usual dietary intake (21, 24, 25). The Harvard FFQ allows participants to choose the average frequency of consumption of food items over the past year on a Likert scale with choices ranging by individual questions. The FFQ includes standard portion sizes for each item (e.g., one apple, three oz. chicken, two slices bacon). Total energy and nutrient intake was estimated by summing intakes from each food on the basis of the selected standard portion size, reported frequency of consumption, and nutrient content of each food item (17). Daily food group servings were estimated by summing the frequency weights of each food item on the basis of reported daily frequencies of consumption (17, 21). FFQs were classified a priori into 39 food/food groups using methods described in similar studies (17, 26, 27). Principal component analysis (PCA) was used to derive pretreatment dietary patterns as in previous studies of dietary patterns and cancer outcomes in this head and neck cancer cohort (2, 17).

Age was modeled as a continuous variable. Smoking and drinking status was categorized as current/former versus never smoked, where “current” status reflects use in the 12 months prior to cancer diagnosis. BMI (kg/m²) at diagnosis was calculated on the basis of self-reported height and weight measures, which were previously reported to be well correlated (r = 0.98) with clinically measured height and weight in this patient population (17). BMI was categorized into four groups: (i) underweight (<18.5 kg/m²); (ii) normal weight (18.5–24.9 kg/m²); (iii) overweight (25.0–29.9 kg/m²); and (iv) obese (≥30 kg/m²). Tumor site was recorded from operative notes and surgical pathology forms and categorized into four groups: (i) oral cavity; (ii) oropharynx; (iii) hypopharynx; and (iv) larynx. Tumor–node–metastases cancer stages were classified according to the American Joint Committee on Cancer VII edition and converted to stage I–IV groupings. To increase the statistical power of stage-wise comparisons, stage was categorized a priori for analyses as I–III versus IV. Because NIS burden is likely to be greatest among those who receive radiotherapy, treatment modality was categorized into two groups—radiation versus no radiation. A total of 170 participants had tumor tissue available for HPV testing from biopsy or surgical resection. Validated PCR methods were used to determine HPV status, as described previously (28). Participants with equivocal or missing HPV status were given a status of “unknown”. HPV status was categorized into three groups: (i) HPV positive; (ii) HPV negative; or (iii) unknown for statistical analyses.

NIS were measured using the UM Head and Neck Quality of Life questionnaire, a validated and multidimensional instrument to assess head and neck cancer–related functional status, and well-being (29). Self-reported NIS were assessed at the pretreatment timepoint and again 1 year postdiagnosis using a Likert scale ranging from “1: not at all bothered” by symptom to “5: extremely bothered”. Symptom scores were dichotomized as “not at all” versus “slightly–extremely” bothered. The research team agreed it was essential to dichotomize symptoms in this manner as sensitivity to symptoms likely varies among individuals and thus any degree of being bothered by symptoms should be considered significant. Data on seven NIS were reported, including trismus, xerostomia, dysphagia of liquids, dysphagia of solid foods, difficulty chewing, taste, and mucositis. A study-specific overall NIS summary score (sum of seven symptoms, range 7–34) was derived. The continuous NIS summary score was dichotomized as <13 versus ≥13, the median and mean in the dataset. The research team selected all six symptoms from the head and neck quality of life eating domain in addition to one symptom in the pain domain (mucositis) to create the NIS summary score, as these symptoms are most likely to impact dietary intake.

Descriptive statistics (means and frequencies) were generated for all demographic, epidemiologic, and clinical variables. Multivariable logistic regression models were used to examine the associations between derived dietary patterns (fit by quartiles of exposure) and each of the individual seven symptoms, as well as the dichotomous NIS summary score. Covariates were chosen a priori based on variables known or hypothesized to be associated with dietary intake and NIS. Covariates considered for inclusion in the final models were age, sex, pretreatment NIS, tumor site, cancer stage, pretreatment smoking status, pretreatment drinking status, treatment modality, total calories, HPV status, and pretreatment BMI. Tests for collinearity were performed among all potential covariates. The final multivariable models were adjusted for age, smoking status, BMI, tumor site, cancer stage, HPV status, and total calories. As sex, drinking status, and treatment modality were found to be highly correlated with other covariates, these variables were excluded from the final model to prevent issues of collinearity. ORs and 95% confidence intervals (CI) were estimated for each quartile (Q) of dietary pattern score compared with the lowest, Q1. In addition, a test for trend across increasing quartile of intake was performed by setting each individual's dietary pattern score to the median for that quartile and treating it as a continuous variable.

When performing subanalyses, simple models with fewer covariates (age, smoking status, cancer stage, and tumor site) were used because of statistical power considerations. To assess the potential for effect measure modification, stratification analyses by smoking status (current/former and never smoked), cancer stage (I–III and IV), treatment modality (radiation and no radiation), and BMI (underweight/normal weight and overweight/obese) were considered. To examine the robustness of results, sensitivity analyses were conducted, in which NIS burden was modeled as a continuous variable and also dichotomized as “not at all–slightly bothered” versus “moderately–extremely bothered”. In an effort to examine the potential for reverse causation (i.e., that participants with higher NIS at diagnosis would prevent patients from eating normally), sensitivity analyses were conducted, in which the associations between pretreatment dietary patterns and pretreatment NIS (as opposed to 1-year NIS) were examined in multivariable models. The addition of pretreatment NIS as a covariate in all primary multivariable models was also tested. All statistical analyses were performed in SAS 9.4 (SAS Institute Inc.; ref. 30). P < 0.05 was considered statistically significant.

Overall epidemiologic characteristics of the study population are shown in Table 1. The mean age for study participants was 60 years. The vast majority of participants were white males. The most common tumor location was the oropharynx. HPV-positive tumors were confirmed in 21% of the population, in which 83% were tumors of the oropharynx. More than half the tumors were stage IV. Approximately 70% and 93% of study participants were current or formers smokers and alcohol users, respectively. Roughly 67% of the population was overweight or obese at diagnosis. Select characteristics of the study participants, according to self-reported NIS burden are shown in Table 2. Participants with a lower NIS summary score were more likely to be diagnosed with stage I–III cancers, have tumors located in the oral cavity or larynx, treated without radiation, and never smokers.

Two major dietary patterns emerged from PCA. The first pattern, termed the prudent dietary patterns, was characterized by high intakes of fruit, vegetables, whole-grains, low-fat dairy, legumes, and less saturated fat. The second pattern, termed the Western dietary pattern, was characterized by high intakes of red and processed meats, refined grains, potatoes, French fries, high-fat dairy, condiments, desserts, snacks, and sugar-sweetened beverages. The factor-loading matrix for the two dietary patterns is presented in Supplementary Table S1.

ORs and 95% CI corresponding to the magnitude of associations for the pretreatment prudent dietary pattern score and self-reported 1 year postdiagnosis NIS burden dichotomized as “not at all” versus “slightly–extremely bothered” are reported in Table 3. After adjusting for age, tumor site, cancer stage, smoking status, BMI, calories, and HPV status, significant inverse associations were observed between pretreatment prudent pattern score and dysphagia of liquids, dysphagia of solid foods, difficulty chewing, and mucositis at 1 year postdiagnosis. A statistically significant inverse association was observed between the dichotomized NIS summary score and the prudent pattern. No significant associations were observed between the Western pattern and NIS burden.

Results of subanalyses stratified by smoking status, cancer stage, and BMI for the NIS summary score are displayed in Table 4. A significant inverse association between prudent dietary pattern score and NIS was observed in never smokers, but the association for current/former smokers was not significant. A significant inverse association between prudent dietary pattern score and NIS was observed in those who were underweight or normal weight at diagnosis, but the association for those overweight or obese at diagnosis was not significant. Parameter estimates for subanalyses stratified for treatment modality and cancer stage were not significantly different from the estimates for the overall population.

Results of sensitivity analyses, where NIS was modeled as a continuous variable and individual NIS were dichotomized as “not at all–slightly bothered” versus “moderately–extremely bothered” were consistent with all primary results: difficulty chewing (OR 0.44; 95% CI, 0.21–0.95; P = 0.03), dysphagia of liquids (OR 0.19; 95% CI, 0.05–0.67; P = 0.009), dysphagia of solid foods (OR 0.22; 95% CI, 0.09–0.50; P = 0.0004), dichotomized NIS summary score (OR 0.37; 95% CI, 0.17–0.76; P = 0.006), and continuous NIS summary score (OR 2.52; 95% CI, 1.40–4.54; P = 0.002). The exception was mucositis, which was no longer statistically significant (Supplementary Table S2). The addition of pretreatment NIS to the original multivariable models as a covariate did not significantly alter parameter estimates: difficulty chewing (OR 0.42; 95% CI, 0.18–0.97; P = 0.04), dysphagia of liquids (OR 0.35; 95% CI, 0.15–0.79; P = 0.01), dysphagia of solid foods (OR 0.35; 95% CI, 0.15–0.80; P = 0.01), dichotomized NIS summary score (OR 0.24; 95% CI, 0.11–0.53; P = 0.0004), and continuous NIS summary score (OR 2.97; 95% CI, 1.58–5.56; P = 0.0007; Supplementary Table S3). Sensitivity analyses examining associations between pretreatment prudent dietary score and individual pretreatment NIS (as opposed to 1-year NIS) yielded null results, with the exception of trismus (OR 0.33; 95% CI, 0.12–0.89; P = 0.03; Supplementary Table S4).

In this prospective cohort study of patients with newly diagnosed head and neck cancer, high intake of a pretreatment prudent dietary pattern was associated with lower risk of self-reported NIS at 1 year postdiagnosis. Stratified analyses suggest possible effect modification by smoking status and BMI. While previous studies have assessed the acute relationship of NIS burden on general dietary intake (31) in patients with head and neck cancer, this is the first study to prospectively examine associations between pretreatment dietary patterns and self-reported NIS burden beyond the acute phase of treatment.

In the early 2000′s, it was hypothesized that antioxidant supplementation during radiotherapy may protect normal cells from reactive oxygen species (ROS) damage, allowing for better tolerance of treatment and higher dosage without adverse toxicities (32). On the basis of this hypothesis, two previous double-blind, placebo-controlled randomized clinical trials (RCT) were conducted to test high-dose antioxidant supplementation with β-carotene and α-tocopherol in patients with head and neck cancer during radiation (33, 34). Both RCTs resulted in reduced treatment toxicity but one was discontinued early because of increased recurrence and mortality with supplementation, while the other showed a nonsignificant increase in mortality with supplementation (33, 34). The authors concluded that the high-dose antioxidant supplements may have reduced the therapeutic efficacy of radiotherapy by quenching radiation-induced ROS intended to damage cancer cells (33, 34). On the other hand, the results of this study suggest that a dietary pattern consisting of whole foods abundant in antioxidants and phytochemicals may offer a promising strategy for reducing treatment-related toxicities without also reducing overall prognosis. In fact, previous research from the UM HN-SPORE cohort provided evidence that the prudent pattern may improve recurrence and survival (17, 35). Future RCTs should be developed that test interventions focusing on soft/cooked vegetables, smoothies, and other foods characterizing the prudent pattern prepared in a way that is easier for this population to chew and swallow.

No associations were found between NIS and the Western dietary pattern score. This was surprising as previous research has suggested the Western dietary pattern to be proinflammatory and inflammation is a common shared etiologic factor involved in the pathogenesis of these symptoms (18). A possible explanation may be that after diagnosis, patients are motivated to change their diet to a seemingly “healthier” one consisting of fruits, vegetables, low-fat dairy, and plant-based proteins, counteracting potential proinflammatory effects of Western pattern foods. Future research should focus on how dietary patterns may change after head and neck cancer diagnosis and the associated outcomes.

In stratified analyses, there was a suggestion of effect measure modification by smoking status and BMI. For those who have never smoked, the prudent dietary pattern was statistically associated with decreased NIS summary score burden but the association was diminished in current and former smokers. It is possible that a high prudent dietary pattern in current and former smokers may not offer the protective potential needed to prevent chronic NIS. Previous research suggests that cigarette smoke may result in increased metabolic turnover, with antioxidant micronutrients expended in response to increased oxidative stress. Alternatively, smoking may decrease micronutrient absorption. Regardless, ever smokers have lower levels of circulating antioxidant micronutrients and may require additional micronutrient intake than never smokers prior to observing protective benefits (36, 37).

The prudent dietary pattern was significantly inversely associated with the NIS summary score for under/normal weight patients but not for overweight or obese patients. Patients with a lower BMI at diagnosis may experience greater NIS burden and thus be more likely to benefit from a prudent dietary pattern. This hypothesis is consistent with Table 2, which shows that the mean NIS summary score was lower among patients diagnosed who were overweight/obesity at diagnosis as compared with those who were under/normal weight. In subanalyses stratified by treatment modality, it was surprising that radiation status did not modify the association between prudent dietary pattern and NIS especially considering chronic radiation associated dysphagia is often a complication following head and neck cancer radiotherapy (38).

Our study is not without fault. Dietary patterns and NIS burden relied on self-report and may be vulnerable to measurement error and systematic biases. For instance, the potential presence of NIS at diagnosis may influence pretreatment dietary intake, leading to recall bias when reporting usual diet from the past year. While there is a high prevalence of acute toxicity in patients with head and neck cancer, our research team was unable to assess acute associations of dietary patterns with NIS burden; these data were only collected at pretreatment and 1 year in this survival cohort. Finally, while the NIS examined were from a validated quality of life questionnaire, the NIS summary score was created for this specific analysis and is not validated.

It is important to note that the observational study design does not prove causality and thus reverse causality cannot be ruled out. However, if reverse causality was present, we hypothesize a significant inverse association would have been observed for the Western dietary pattern and NIS. Observing significant inverse association with both dietary patterns might provide stronger evidence for reverse causation, that is, a lack of NIS leads to higher reported dietary pattern scores, in general, simply because the patient is able to consume more food. Sensitivity analysis assessing pretreatment diet and pretreatment NIS were null. Further, results of sensitivity analyses modeling NIS burden in different ways (i.e., continuous summary score and individual NIS dichotomized as “not at all–slightly” vs. “moderately–extremely bothered”) remained statistically significant, supporting the robustness of the observed associations.

To our knowledge, this was the only study in head and neck cancer survivors to date examining the associations between pretreatment dietary patterns and chronic NIS burden. Strengths of this analysis include the prospective, longitudinal design and ability to control for multiple confounding factors. Results of this analysis may be generalizable to other predominately non-Hispanic white head and neck cancer survivors.

In conclusion, consumption of a pretreatment prudent diet classified by high intakes of fruit, vegetables, whole grains, low-fat dairy, and legumes may help reduce the risk of chronic NIS such as difficulty chewing, dysphagia of liquids, dysphagia of solids foods, and mucositis 1 year after diagnosis in head and neck cancer survivors. The results may be modified by smoking status and BMI. This study provides evidence that consuming whole foods abundant in antioxidants may be an efficacious and safe alternative to reducing treatment toxicities compared with the high-dose antioxidant supplements that were tested in the early 2000′s. Future research should utilized RCT designs to test whether increasing consumption of foods characterizing the prudent pattern before and during head and neck cancer treatment results in reduced NIS and improved survival.

K.R. Zarins reports receiving a commercial research grant from Brooklyn ImmunoTherapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S.L. Crowder, A.E. Arthur

Development of methodology: S.L. Crowder, K.P. Sarma, L.S. Rozek, A.E. Arthur

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.P. Sarma, K.R. Zarins, G.T. Wolf, L.S. Rozek

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.L. Crowder, Y.T. Chen, Z. Li, G.T. Wolf, A.E. Arthur

Writing, review, and/or revision of the manuscript: S.L. Crowder, A.M. Mondul, M.Y. Pepino, K.R. Zarins, G.T. Wolf, L.S. Rozek, A.E. Arthur

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.L. Crowder, Z. Li, G.T. Wolf

Study supervision: L.S. Rozek

The authors thank the patients, clinicians, and principal investigators of the individual projects at the UM Head and Neck SPORE program who provided access to the longitudinal clinical database and were responsible for the recruitment, treatment, and follow-up of patients included in this article. These investigators included Avraham Eisbruch, Theodore Lawrence, Mark Prince, Jeffrey Terrell, Shaomeng Wang, and Frank Worden. NIH/NCI P50CA097248 was awarded to G.T. Wolf. USDA National Institute of Food and Agriculture, Hatch project 1011487 was awarded to A.E. Arthur. S.L. Crowder was supported by a Carle-Illinois Cancer Scholars for Translational and Applied Research Fellowship and an Academy of Nutrition and Dietetics Colgate Palmolive Fellowship in Nutrition and Oral Health.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.