Serum Carotenoids and Cancer-Related Fatigue: An Analysis of the 2005–2006 National Health and Nutrition Examination Survey

Cancer-related fatigue is a prevalent, debilitating condition, and preliminary evidence suggests a relationship between higher diet quality and lower fatigue. Serum-based carotenoids, Vitamin A, and Vitamin E are biomarkers of fruit and vegetable intake and therefore diet quality. To further elucidate the link between diet quality and cancer-related fatigue, associations were assessed between these serum-based nutrients and fatigue among American adults with special attention to cancer history. Data were analyzed from the United States 2005–2006 National Health and Nutrition Examination Survey dataset. Ten carotenoids, vitamin A, vitamin E, and γ-tocopherol were measured from fasting blood samples and fatigue was patient-reported. Associations between carotenoid concentration and fatigue were estimated using ordinal logistic regression models. Adjusted models included a diagnosis of cancer (with the exception on nonmelanoma skin cancer, yes/no), age, body mass index, race/ethnicity, education, and exercise habits as covariates, and additional models included a cancer × nutrient interaction. Of 4,091 participants, 272 (8.0%) reported a history of cancer. Greater fatigue was associated with lower serum trans-lycopene, retinyl palmitate, and retinyl stearate (all P < 0.05) in separate models adjusting for potential confounders. For these nutrients, a one-SD increase in nutrient was associated with a 6.8%–9.9% lower risk of greater fatigue. Among cancer survivors only (n = 272), statistically significant associations were not observed between any of the nutrients and fatigue. In conclusion, greater serum concentrations of carotenoid biomarkers were associated with less fatigue. These results support further exploration into relationships between carotenoid intake, diet quality, and persistent fatigue. Significance: Cancer-related fatigue often persists for years into survivorship, reduces quality of life, and prevents people from returning to their lives before cancer. Interventions to address cancer-related fatigue are much needed. Herein, serum carotenoids were associated with lower fatigue, thereby supporting further development of nutritional interventions to address fatigue in survivorship.


Introduction
Cancer-related fatigue affects at least 30%-90% of patients with cancer, depending on the type of cancer and the treatment (surgery, chemotherapy, and/or radiation; refs. [1][2][3][4][5][6]. Cancer-related fatigue is one of the most prevalent and treatment, although early nutritional intervention and dietary counseling can significantly improve nutritional status and quality of life during treatment (9,(12)(13)(14) and improve fatigue in survivorship (15). For example, in a randomized controlled trial among breast cancer survivors, a 3-month diet high in fruit, vegetables, whole grains, and omega-3 fatty acids reduced cancer-related fatigue compared with a general health curriculum (15). Moreover, diet and obesity are closely related, and obesity is a risk factor for exacerbated side effects in survivorship, including cancer-related fatigue (16,17).
Dietary assessment by means of patient-report (e.g., food frequency questionnaires, food diaries, food recall) tends to have a high error rate due to recall bias, difficulty in estimating serving sizes, social desirability bias, and other factors (18). Blood-based biomarkers have emerged as objective surrogates for specific foods and nutrients and may serve as a proxy for diet quality, specifically, serum carotenoid concentrations (18,19).
Serum carotenoids are surrogates for diet quality because they are plentiful in many fruits and vegetables (19,20). Carotenoids are high in fruits and vegetables that have a yellow, red, or orange hue, including carrots, cantaloupe, sweet potatoes, tomatoes, and apricots, as well as leafy green vegetables in which the chlorophyll masks the carotenoid pigments, including kale, broccoli, and collard greens. Carotenoids are relatively specific to fruits and vegetables, with the exception of egg yolks, which are one of the only animal products that contain carotenoids. Consequently, carotenoids are established as reliable biomarkers of fruit and vegetable intake (21). Carotenoids include lycopene, β-carotene, lutein, zeaxanthin, and others, and some are provitamins that are converted into vitamin A in the body. Because people tend to eat a fairly consistent diet in adulthood (22,23), carotenoid biomarkers from fasting blood samples are considered reliable for epidemiology studies, with less than 8% variation in blood concentrations of lycopene, β-cryptoxanthin, and lutein over four years (24).
Several studies have suggested that serum carotenoids may correlate with cancer-related fatigue or fatigue in general. For example, an inverse association was observed between carotenoid intake and fatigue among 40 adult cancer survivors in a cross-sectional study in the United States (25). In addition, in a study exploring the effects of nutritional biomarkers on sleep factors in the 2005-2006 National Health and Nutrition Examination Survey (NHANES) dataset, an inverse association was observed between carotenoids [retinyl esters and lutein + zeaxanthin (combined predictor)] and poor sleep-related daytime dysfunction (not necessarily cancer-related fatigue) adjusting for history of cancer among other comorbidities (26).
However, there have been no studies to our knowledge assessing serum carotenoids specifically on fatigue in the context of cancer survivorship. Thus, we evaluated the associations between serum carotenoid concentrations and fatigue, with special attention to cancer survivors compared with individuals without a prior cancer diagnosis. These data are hypothesis generating, yet shed light on associations between serum nutrient biomarkers and cancer-related fatigue and can help guide intervention development (nutritional supplements, meal replacements, and dietary interventions) to alleviate cancer-related fatigue in survivorship.

Study Population
NHANES is an initiative by the U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, and National Center for Health Statistics to collect objective and patient-reported data regarding the health and nutritional status of nationally representative civilian noninstitutionalized children and adults living in the United States. It is a continuous program, and the 2005-2006 NHANES oversampled adolescents, older Americans, African Americans, and Mexican Americans to address specific needs of these underrepresented populations. NHANES is conducted in accordance with ethical guidelines-the protocol was approved by the National Center for Health Statistics Research Ethics Review Board and all participants provided written informed consent (27,28). Trained NHANES professionals administered cross-sectional surveys in a multistage probability cluster sampling design using at-home interviews and health examinations using a mobile examination center. The details of recruitment and data collection have been previously described (27,28). The 2005-2006 NHANES dataset was selected for this analysis because it included serum carotenoid measures, cancer status, and a question on daytime fatigue.
The University of Rochester Research Subjects Review Board reviewed the protocol for this analysis and declared it exempt (STUDY00006373). Participants were excluded from this analysis if they had missing data for the question regarding a history of a cancer diagnosis (yes/no; n = 565), the Depression Screener Questionnaire (the source of the fatigue outcome variable; DPQ; n = 503), and/or carotenoid laboratory values (n = 391). Because nonmelanoma skin cancer is usually treated with surgery under a local anesthetic and not chemotherapy or radiation, we included participants with a single cancer diagnosis of nonmelanoma skin cancer in the group without a recent cancer diagnosis (n = 57). We then performed a sensitivity analysis with these participants in the group with a recent cancer diagnosis; estimates for the association between nutrient and fatigue changed less than 10% in all models except those for cisand trans-beta carotene and vitamin A, all of which had estimates <0.002 units.

Fatigue Measurement
The DPQ, also referred to as the Patient Health Questionnaire-9, is a nine-item questionnaire that is based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) and has been validated in the general population (29,30). Daytime fatigue was assessed via one question from the DPQ (DPQ040): "Over the last 2 weeks, how often have you been bothered by the following problem[s]: feeling tired or having little energy?" Response options were "not at all" (0 or none), "several days" (1 or light), "more than half the days" (2 or moderate), and "nearly every day" (3 or severe). While NHANES did not include a question specific to cancer-related fatigue, a single-item question such as this one is often used to assess patient-reported fatigue (31,32).

Measurement of Serum Carotenoids
Blood was collected via venipuncture after an overnight (>9 hour) fast by a trained phlebotomist into a red-top or royal blue-top Vacutainer blood tube (33). Serum was harvested using standard procedures and stored at −70°C until analysis (up to 2 years). Carotenoids-total (cis-and trans-) lycopene, α-carotene, trans-β-carotene, cis-β-carotene, β-cryptoxanthin, lutein + zeaxanthin, trans-lycopene, retinyl palmitate, retinyl stearate; vitamin A; and vitamin E were quantified at one of 28 laboratories across the United States using highperformance liquid chromatography (HPLC) with photodiode array detection; methods are detailed in the NHANES documentation (28).

Statistical Analyses
All statistical analyses were performed in SAS, version 9.4 (SAS Institute).
Descriptive statistics for demographics and clinical characteristics (means, proportions) were calculated for the overall population and in relation to cancer history (yes/no). Age was considered a continuous variable and BMI was coded as a categorical variable with traditional cut-off points (underweight, <18.5 kg/m 2 ; normal weight, 18.5-<25 kg/m 2 ; overweight, 25-<30 kg/m 2 ; or obese, ≥30 kg/m 2 ; ref. 16). Race/ethnicity was categorized into Mexican American, Other Hispanic, non-Hispanic White, non-Hispanic Black, and other non-Hispanic race including multiracial. Education was categorized into less than a high school education, less than a college degree, or at least a fouryear college degree. Habitual exercise was estimated using patient-reported metabolic equivalents (MET) hours per week and was treated as a continuous variable. To adjust for the complex sampling design, oversampling of select populations, and nonresponse, the data were weighted so that our results represent the total U.S. population. Specifically, appropriate adjustments were applied for strata (sdmvstra), cluster (sdmvpsu), and Mobile Examination Center weight (wtmec2yr). Using ordinal logistic regression (PROC SURVEYFREQ) and appropriate sample weights, we assessed the relationship between fatigue [question DPQ040 as an ordinal variable (0, 1, 2, 3)] and each of the nutrient concentrations in independent models. We performed a crude analysis as well as one adjusting for cancer diagnosis (excluding nonmelanoma skin cancer, yes or no), age, BMI, race/ethnicity, education, and exercise. To assess effect modification between serum carotenoid concentration and a history of a cancer diagnosis, an interaction term was included in the adjusted models and stratum-specific estimates are reported for those with and without a history of cancer (excluding nonmelanoma skin cancer). P < 0.05 was considered statistically significant.

Data Availability
The data analyzed in this study are publicly available from NHANES at https://www.cdc.gov/nchs/nhanes/.

Results
A total of 4,091 participants were included in this analysis, of which 272 (8.0%) had a history of a cancer diagnosis (with the exception of nonmelanoma skin cancer; Table 1). Individuals with a history of cancer were, on average, 15.5 years older than those without a cancer diagnosis (P < 0.001) and there was a slightly greater proportion of females among those with a cancer diagnosis (P < 0.001). Interestingly, there were notable differences in the distribution of races and ethnicities among those with and without a history of cancer-9.9% of non-Hispanic Whites had had a cancer diagnosis while only 2.1% of Mexican Americans, 4.1% of other Hispanic Americans, 4.2% of Black Americans, and 2.7% of those of mixed race or other races had had a cancer diagnosis (P < 0.001). In line with typically reported statistics (39), approximately two-thirds of all participants were overweight or obese, with no significant between-group differences (P = 0.85). Among those with a history of cancer compared to those without a history of cancer, serum concentrations of some carotenoids and vitamins tended to be higher (e.g., trans-β-carotene, vitamin A, both P < 0.01), some tended to be lower (e.g., total lycopene, γ-tocopherol, both P < 0.01), and some were similar (e.g., lutein and zeaxanthin, P = 0.56). Those with a history of cancer reported a greater daytime fatigue score than those without a history of cancer (scale from 0-4, mean ± SE, 0.87 ± 0.08 vs. 0.69 ± 0.02, two-sided t test P = 0.03).
We modeled the associations between nutrient concentrations and fatigue. Table 2 and Supplementary Table S1 illustrate the results of crude and fully adjusted models for each of the measured carotenoids, vitamin A, and vitamin E. In models adjusted for age, BMI, race/ethnicity, education, exercise habits, and cancer history, higher serum carotenoid concentrations were associated with lower risk for fatigue for trans-lycopene, retinyl palmitate, and retinyl stearate (P < 0.05). These ORs indicate that, assuming a linear association between nutrients and fatigue, a one-SD increase in trans-lycopene (11.4 μg/dL) is associated with a 9.9% lower risk of more severe fatigue (95% CI, 1.0%-1-8.0%; Fig. 1).
With regard to covariates (Supplementary  Table S4).

Discussion
Herein, higher serum concentrations of carotenoids, vitamin A, and vitamin E, which correlate with fruit and vegetable intake, were collectively associated with lower odds of greater fatigue ( Fig. 1; Table 2). These data were derived from a nationally representative sample of Americans, with racial and ethnic minorities oversampled and, although these data are correlational and hypothesis-generating by nature, they warrant further investigation of dietary interventions that increase serum carotenoid concentrations to address cancer-related fatigue in survivorship.
Carotenoids are high in many fruits and vegetables and are established as reliable biomarkers of fruit and vegetable intake (21). Thus, our results that higher serum carotenoid concentrations were associated with less fatigue (Table 2) suggest that higher fruit and vegetable intake are associated with less fatigue. These results are consistent with Beydoun and colleagues (26)   cohort (controlling for cancer diagnosis). In addition, a recent meta-analysis concluded that a diet rich in fruits and vegetables may alleviate cancer-related fatigue (40). However, Zick and colleagues (15) saw no associations between change in carotenoids and change in fatigue in a randomized controlled trial assessing the effects of a 3-month dietary intervention on fatigue compared to an attention control among 30 breast cancer survivors; they did observe improvements in fatigue.
We interpret the implications of our observed serum carotenoids using total lycopene as an example. We observed that a 21-μg/dL greater concentration of total lycopene (one SD) was associated with an 8.8% lower risk of more severe fatigue (Fig. 1). With directed approaches, this amount of increase in lycopene is feasible with dietary intervention. For example, Zick and colleagues (15) implemented a custom "Fatigue Reduction Diet" that was rich in fruits and vegetables among fatigued breast cancer survivors who consumed <5.5 servings of fruits   We observed a statistically significant interaction between cancer and lutein + zeaxanthin; however, among cancer survivors only, there were no statistically significant relationships between cancer and any of the nutrients (Table 3). While this study may not be adequately powered to reveal interactions between cancer and any nutrients, these data suggest that the mechanisms by which nutritional status and fatigue are related may not be specific to the cancer experience. Nevertheless, the prevalence and severity of fatigue is higher among cancer survivors compared with people who have not had a cancer diagnosis (Table 1), and the development of nutritional interventions are warranted to alleviate fatigue among this high-risk population. Mechanisms underlying other chronic fatigue disorders include chronic inflammation, oxidative stress, mitochondrial dysfunction, circadian rhythm disruption, and others (42,43). Adverse pathophysiology can be initiated or exacerbated by the cancer experience, and cancer-related fatigue can be derived from the pathology of the cancer itself; the biological effects of surgery, radiation, and/or chemotherapy; and the mental burden of a cancer diagnosis and treatment. In addition, after a cancer diagnosis, fatigue, cognitive impairment, negative affect, and sleep disturbances tend to co-occur-the "psychoneurologic symptom cluster;" these symptoms likely share underlying pathologies in the brain and in the periphery (44).
Nutritional interventions can effectively reduce inflammation and oxidative stress (45) and are promising in the prevention and treatment of cancer-related fatigue (40).
Because serum carotenoids are biomarkers for higher fruit and vegetable intake, it is unknown if associations between serum carotenoids and fatigue stem from the carotenoids themselves, other components of carotenoid-rich foods (e.g., dietary fiber), and/or synergistic effects between carotenoids and other components of the food matrix (46). While there have been no studies to our knowledge testing the effects of supplementation of carotenoids from dietary supplements on supportive care outcomes, randomized controlled trials testing the effects of β-carotene supplementation on the development or progression of cancer have been null or suggest that supplementation can increase the risk of poor outcomes, especially in current smokers (47). Moving forward, it will be important to determine who an intervention may benefit (e.g., survivors of which type(s) of cancer, which treatments survivors are undergoing/underwent, baseline demographics or metabolic clinical characteristics), components of an optimized intervention (e.g., length in months, which dietary patterns), and whether carotenoids are useful biomarkers for intervention adherence and/or risk for persistent cancer-related fatigue.
Interestingly, we observed that Hispanic participants, especially Mexican Americans, and Black participants had a lower incidence of cancer than White participants (Table 1), and Mexican Americans reported less fatigue (Supplementary Whole-food nutritional interventions are in their infancy and remain promising to minimize cancer-related fatigue. Exercise and psychosocial behavioral interventions are unequivocally beneficial in the treatment of cancer-related fatigue and, in fact, more potent than current pharmaceutical interventions (4), though nutritional interventions remain understudied (40). During chemotherapy, patients tend to consume less total energy, fat, and protein (50), which leads to a large prevalence of malnutrition, especially among older patients (51,52 (55,56). In addition, it is believed that nutrients from whole foods have synergistic benefits on health that cannot be harnessed from extracts of bioactive compounds (55).
This study has a unique strength in that we were able to assess patient-reported fatigue among thousands of Americans and investigate associations between serum carotenoid concentration, a proxy for diet quality, and fatigue among people with and without a cancer diagnosis. Using ordinal logistic regression modeling, we were able to investigate effect modification of a history of cancer and nutrients on the experience of fatigue in survivorship.
This study, however, is not without limitations. Herein, fatigue was measured using a single-item question regarding feeling tired and having little energy; it was not measured using a standard, validated uni-or multi-dimensional questionnaire (57). However, single-item questions that capture fatigue are very common and have been validated and deemed useful for screening, diagnosis, and research purposes (32,58). For example, Wolfe observed that the single question, "How much of a problem has fatigue or tiredness been for you in the past week?" with response options 0-10, performed at least as well as longer fatigue questionnaires in regard to sensitivity to change and correlation with clinical measures (31). Also, this study was cross-sectional and therefore we cannot declare causation. Causation could occur in either direction, where good nutritional status could help alleviate fatigue, or having more energy could help facilitate healthy dietary behaviors and/or healthy metabolism. More research is needed to investigate whether interventions that improve diet quality could help alleviate cancer-related fatigue post-diagnosis and post-primary treatment. Moreover, a history of a cancer diagnosis is associated with a plethora of psychological (e.g., stress), logistical (e.g., interference in a person's job and personal life), and biological factors (e.g., chemotherapy, radiation, hormone therapy) that could all contribute to fatigue, and these factors could not be distinguished herein. However, this broad analysis among a heterogeneous survivor population contributes information regarding relationships between carotenoids and cancer-related fatigue that apply to the whole population. Finally, this was a secondary analysis and is therefore hypothesis-generating rather than hypothesis-testing. These results should be confirmed in future studies where assessments of the correlation between carotenoid concentration and fatigue are the primary outcome.

Conclusions
Associations were observed between circulating carotenoid concentrations (which reflect carotenoid and therefore fruit and vegetable intake) and fatigue in a large population that is generalizable to the American population, though there was no evidence of an association between serum carotenoids and fatigue that was specific to the cancer experience. The high prevalence and severity of fatigue among survivors, in combination with the observation that higher serum carotenoid concentrations were associated with less fatigue, warrants future randomized controlled trials that aim to increase fruit and vegetable intake to reduce the susceptibility to fatigue among cancer survivors.