Mediterranean Diet and Fatigue among Community-Dwelling Postmenopausal Women

Abstract

We investigated cross-sectional relationships between the Mediterranean diet and overall fatigue, energy, and weariness scores among 4563 women aged 65+ from the Women’s Health Initiative study. We also used the Isocaloric Substitution approach to explore whether the substitution of fish for red and processed meat, whole for non-whole grains, and whole fruit for fruit juice relate to RAND-36 measured overall fatigue and its subdomains. The alternate Mediterranean Diet (aMED) Index quintiles (Q1-Q5) and selected Mediterranean foods available on a Food Frequency Questionnaire were exposure measures. Results showed aMED Q5 was associated with 2.99 (95% CI: 0.88, 5.11), 4.01 (95% CI: 1.51, 6.53), and 2.47 (95% CI: 0.24, 4.70) point improvements in fatigue, energy, and weariness scores, respectively, compared with aMED Q1. Substituting fish for red and processed meat and whole for non-whole grains was associated with more favorable fatigue scores, whereas substituting whole fruit for juice was not.

INTRODUCTION

Fatigue, the subjective report of weariness and lack of energy, is a common and bothersome symptom in older adults. Estimates show that about 27% to 50% of community-dwelling adults 65 years and older and 75% of adults 85 and older report fatigue,¹ and studies showed fatigue is more prevalent among older women than older men.^2,3 In older adults, fatigue significantly relates to a decreased function and performance, including physical,^4,5 cognitive functioning,⁶ and specific task performance such as chair rise.⁷ Studies also found about twice the risk of hospitalization and using home help five years later⁸ and a higher risk of mortality among fatigued older adults than unfatigued older adults.^4,9 Besides, for fatigued older individuals, decreased functioning leads to an increased need for further effort, and thus, fatigue worsens. This spiral helps to explain the unremitting and escalation of fatigue.¹⁰ A need to address unfolding escalation in fatigue in a rapidly growing number of older adults led to a concerted effort by researchers to understand modifiable factors that prevent fatigue and improve outcomes in those who experience fatigue. Nutritional factors are especially pertinent in that regard because they are modifiable and have been implicated in experiments and randomized clinical trials (RCTs) in fatigue occurrence and severity. ^11,12,13 Despite the promising findings that involve individual nutrients, the role of the dietary pattern has yet to be examined in relation to fatigue, especially while leveraging well-powered cohorts with characterized dietary measures. Dietary patterns represent an integrated picture of frequency, proportion, and variety of food, drinks, and nutrient consumption in a habitual diet.¹⁴ Compared with individual nutrients, dietary patterns are more advantageous for research and clinical use because they consider the substitution effect andsynergic and antagonistic interactions between nutrients.¹⁵

RCTs and Epidemiological Evidence

Although many dietary patterns exist, the Mediterranean diet – a way of eating based on the traditional cuisine of Mediterranean countries and typically high in vegetables, fruits, whole grains, fish, beans, nuts and seeds, and olive oil – might be relevant to studying fatigue in old age as the diet has been studied in people aged 65 years and older.^16–20 These studies, however, were limited by ad hoc measures of fatigue, which were commonly operationalized as one or two items from the Center for Epidemiologic Studies Depression Scale (CES-D) scale.^18–20 Some RCTs were identified, which showed adhering to a Mediterranean diet significantly improved fatigue. ^21.22 Although informative, one important caveat is that all these previous studies were focused on clinical populations (e.g., people with diabetes or people with arthritis) and missed the opportunity to examine the dietary pattern-fatigue relationship among a more general population of community-dwelling older adults. Further, most studies examining the Mediterranean diet and fatigue relationship considered only the role of nutrition, but not physical activity.^18–20 Individuals who adhere to a healthy diet are commonly engaged in other healthy behaviors such as physical activity, and physical activity relates to fatigue;²³ as such, it is crucial to include physical activity in the model to adjust for the potentially confounding relationship.

Potential Physiological Mechanisms of the Mediterranean Diet Related to Fatigue

The exact mechanism by which a Mediterranean diet exerts its beneficial effects in improving fatigue is not known. A Mediterranean-style diet might influence fatigue through the pathways of reduced saturated fatty acid intake, reduced amino acid, reduced calorie intake, increased phytochemical intake, and gut microbiota.²⁴ For people with high adiposity, the Mediterranean diet might not only affect fatigue via anti-inflammation and anti-oxidation but also trigger other metabolic pathways involved in lipid processing.²⁴ Thus, the association between the Mediterranean diet and fatigue might differ among those with high versus low adiposity. Similarly, fatigue is a common symptom in people with diabetes,²⁵ and as the Mediterranean diet has been involved in insulin regulation,²⁴ there might be an additional benefit for those with diabetes compared with those without diabetes.

Isocaloric Substitution Method (ISM)

Beyond the association and interaction analysis in examining dietary patterns and fatigue, an increasing number of studies have started using isocaloric substitution methods (ISM) to compare the relative health contributions of different foods and macronutrients.^26–30 ISM is, in nature, a linear regression.^31,32 Substitution is achieved via including both the nutrients of interest and the total energy intake as independent variables. As such, the interpretation of each coefficient of individual food group/macronutrient refers to the change in outcome when substituting the food group/macronutrient for the food group/macronutrient excluded in the model. The application of ISM removes confounding effects from total energy intake, reduces extraneous variation, and predicts or mimics the effect of dietary interventions.³¹ ISM is relevant to study dietary patterns in general and the Mediterranean diet in particular because the health impact of a specific food item cannot be isolated from that of other foods it replaces; further, individuals usually alter the intake of specific foods primarily by changing the dietary composition rather than by changing the total energy intake unless physical activity or body composition change considerably.³¹ Since the Mediterranean diet includes a high intake of vegetables, fruits, fish, and whole-grain and a relatively low intake of red and processed meat and non-whole grains; it is, therefore, important to examine how substituting certain favorable for unfavorable foods under the same food category relates to fatigue. Furthermore, the Mediterranean diet recommends a high intake of fruits; however, it is still unclear whether both whole fruits and fruit juices confer similar health benefits concerning fatigue. Some researchers considered juice as a mere source of sugar,³³ which might increase hypertension risks,³⁴ but others suggested that fruit juice is a good source of vitamin C, folate, potassium, and polyphenols.³⁵

In this study, we examined three questions. 1) Is there an association between the Mediterranean diet and fatigue among community-dwelling older women while adjusting for physical activity and relevant confounders? 2) To what extent do diabetes and body mass index (BMI) individually moderate the relationship between the Mediterranean diet and fatigue, adjusting for physical activity and relevant confounders? 3) Does the substitution of fish for red and processed meat, whole grain for refined grain, and whole fruit for fruit juice relate to improved fatigue separately, adjusting for physical activity and relevant confounders? We hypothesized that higher adherence to the Mediterranean diet would be associated with an improvement in the scores of fatigue and its subdomains in community-dwelling older women; diabetes and BMI would individually moderate the relationship between the Mediterranean diet and fatigue; and the substitution of fish for red and processed meat, whole grain for refined grain, and whole fruit for fruit juice would relate to improved fatigue separately.

METHODS

Sample

The study is an observational cross-sectional design with participants from two ancillary studies of the Women’s Health Initiative Long Life Study (WHILLS) : Objective Physical Activity and Cardiovascular Disease Health (OPACH) Study and WHI-Food Intake (WHI-FI). Details about the main WHI study have been previously described.³⁶ The WHILLS was conducted to collect data on healthy aging and cardiovascular risk factors during the second Extension Study of the WHI among a subcohort of 7,875 WHI participants from March 2012 to May 2013 from all 40 original US clinical centers.³⁷ Among WHILLS, 7,048 women consented to participate in the OPACH study, and they were provided with an ActiGraph GT3X+ (Pensacola, Florida) triaxial accelerometer, a sleep log, and an OPACH Physical Activity (PA) Questionnaire between March 2012 and April 2014 either during a home visit or via express mail afterward. Participants were asked to wear the triaxial accelerometer on their hip for seven consecutive days during waking and sleeping hours except when bathing or swimming.³⁸ The WHI-FI ancillary study also recruited participants from WHILLS. Within four weeks of completing WHILLS measures, WHILLS participants received an invitation to participate in the WHI-FI Study in which they received a Food Frequency Questionnaire (FFQ).³⁹ A total of 6,094 FFQs were returned by WHILLS participants. For this current study, we retained 6,489 OPACH participants who returned their accelerometer, provided usable data, and were older than 65 years. Among the 6,489 women, 4,812 completed FFQs and 4,563 reporting plausible average energy intakes, defined as >= 600 kcal and <= 5,000 kcal.⁴⁰ These 4,563 women composed our final analytical sample (see Figure 1). Approval was obtained from the Fred Hutchinson Cancer Research Center Institutional Review Board as well as from each clinical center's Institutional Review Board, and all WHI participants provided written informed consent.

Figure 1.

Among 7875 Women’s Health Initiative Long Life Study (WHILLS) participants,7048 women consented to participate in the Objective Physical Activity and Cardiovascular Disease Health (OPACH) study, and 6489 OPACH participants returned their accelerometer with usable data and were older than 65 years. Meanwhile, a total of 6094 WHI-Food Intake (WHI-FI) study Food Frequency Questionnaires (FFQs) were returned by WHI LLS participants. Finally, 4812 participants completed both OPACH and FFQs measures with 4563 reporting plausible energy intakes. Thus, the final analytical sample size is 4563.

Measures

Dietary variables.

Diet was measured at WHILLS baseline via FFQ. The FFQ included 140 composite and single-food line items asking about frequencies and approximate portion sizes consumed during the past three months, among which thirteen adjustment questions asked about fat intake, fortification of juices consumed, and type of cold cereal consumed. Three summary questions asked about the usual intake of fruits, vegetables, and added fats for comparison with information gathered from the line items. Nutrient calculations based on reported intake were performed using the Nutrition Data System for Research software version 2012, developed by the University of Minnesota Nutrition Coordinating Center.⁴¹ Total energy intake was calculated as: 4 (kcal/ gram)* carbohydrate (gram/day) + 4 (kcal/ gram)* protein (gram/day) + 9 (kcal/ gram)* fat (gram/day) + 7 (kcal/ gram)* alcohol (gram/day).²⁶

The food groups involved in the substitution analysis are as follows. Fish intake included fried fish, fish sandwich, fried shellfish, not fried shellfish (i.e., shrimp, lobster, crab, and oysters), canned tuna, tuna salad, tuna casserole, broiled or baked white fish, and broiled or baked dark fish (i.e., salmon, mackerel, bluefish). Red and processed meat included frankfurters, sausage, processed, luncheon meats, cooked lean meat from all types of organ meats, including that from beef, pork, veal, lamb, game, poultry, and fish. Whole grain options included whole-grain bread and rolls, whole grain crackers, whole grains as a side dish such as brown rice, and whole-wheat pasta. The following fruit items were included as whole fruits: apples, applesauce, peaches, nectarines, pears, bananas, plums, apricots (fresh, canned, or dried); oranges, grapefruit, and tangerines (not juice); cantaloupe, orange melon, and mango, watermelon and red melon; berries such as strawberries and blueberries; dried fruit (other than apricots) such as raisins and prunes; any other fruit such as grapes, fruit cocktail, pineapple, and cherries. Fruit juice was composed of two separate items asking about orange and grapefruit juices and all other 100% fruit juice types.⁴²

The alternate Mediterranean diet (aMED) score.

Based on nutrient and food item intakes estimated from the FFQ, an aMED index was used to assess the extent of adherence to a Mediterranean Diet. The aMED score was adapted from a traditional Mediterranean diet to accommodate for use among people in non-Mediterranean countries.⁴³ The aMed index has nine components: whole grains, vegetables (excluding potatoes), fruits (including juices), legumes, nuts, fish, ratio of monounsaturated fat to saturated fat, red and processed meats, and alcohol.^43,44 For fruits, vegetables, nuts, legumes, whole grains, fish, or ratio of monounsaturated to saturated fat, when intake was above the median of each component, participants received one point. For red and processed meat, when consumption was below the median, one point was counted. For alcohol, the participant received one point only when the intake was between 5 and 15 g/d. Otherwise, participants received zero points. The total aMED index score spans from 0 (minimal adherence to the diet) to 9 (maximal adherence to the diet).

Fatigue measurement.

The RAND-36 Vitality subscale⁴⁵ was used to measure overall fatigue and its two subdomains: energy and weariness. The Vitality subscale comprises four questions that ask about the frequency of having a lot of energy and feeling tired, worn out, and full of pep over the previous four weeks, with item responses ranging from 1 (all the time) to 6 (none of the time). Energy was calculated as the mean of feeling full of pep and having a lot of energy; weariness was calculated as the mean of feeling worn out and feeling tired; and overall fatigue was calculated as the mean score of the four questions. Overall fatigue, energy, and weariness were then recoded to range from 0 to 100, with a higher score representing a more favorable health status.^45–48

Accelerometer measurement.

Accelerometer data from the three planes (i.e., vertical plane and the two other axes) were processed using ActiLife software version 6 (Pensacola, FL) using 15-second epochs and the normal filter and then integrated with a vector magnitude, an indicator of total physical activity volume. The criteria for physical activity classification were low light-intensity physical activity (LLPA, 19–225 counts/15 s), high light-intensity physical activity (HLPA, 226–518 counts/15 s), and moderate to vigorous physical activity (MVPA, ≥519 counts/15 s).⁴⁹

Body Mass Index (BMI) and Diabetes.

BMI and diabetes served as moderator variables. BMI was collected by self-report of weight and height and calculated by dividing the weight (kg) by the square of the height (m). BMI was categorized as underweight (≤ 18.5), normal weight (18.5–25), overweight/obese (≥25). Diabetes was measured by a self-report question “Has a doctor ever told you that you had sugar diabetes or high blood sugar when you were not pregnant?” Self-reported diabetes has correctly classified diabetes cases and non-cases and is sufficiently accurate to allow use in epidemiologic studies.^50,51

Covariates.

The list of covariates included demographics (age, race/ethnicity, education, Neighborhood Social Economic Status [NSES]), health-related variables (self-rated health, depression, diabetes, BMI, number of chronic diseases), and health behavior variables (sleeping hours, social support, low light physical activity time, high light physical activity time, moderate to vigorous physical activity time, total energy). Depression was measured via a shortened version of the Center for Epidemiological Studies Depression Scale (CES-D) and considered a score greater or equal to 0.06 as evidence of depressive symptoms.⁵² The number of self-reported chronic diseases was measured as the sum of chronic obstructive pulmonary disease, cardiovascular disease, cerebrovascular disease, cancer, osteoarthritis, and cognitive impairment. To evaluate social support, a questionnaire from the Medical Outcomes Study (MOS) was included. The questionnaire was designed to assess the amount of social support the patient has available, including nine questions (four subscales: emotional/informal support, affection, tangible support, and positive social interaction) about how often each different type of support is available to participants. Responses were scored on a five-point scale ranging from “none of the time” to “all of the time.” ⁵³ Sleeping time variable was taken from OPACH Form 321, and it referred to the number of hours spent sleeping during a usual day and night. Responses ranged from 1=<4 hours to 8=16+ hours. Race/ethnicity and the highest level of education were from the WHI baseline (1993–1998). Age, self-rated health, depression, and the number of chronic diseases were from WHI or LLS questionnaires proximal to the OPACH study baseline. Data from the 2000 census were used to assess NSES at the census tract level. NSES was an index of the six census tract variables: (1) percent of those older than 25 with the education less than high school; (2) percent of male unemployment; (3) percent of households whose income are below the poverty line; (4) percent of households receiving public assistance; (5) percentage of female-headed households with children; and (6) the median household income. The variables for the index were identified through confirmatory factor analysis. This NSES index was shown from prior studies as a crucial neighborhood-level predictor of a variety of health outcomes.^54–56 The NSES index was considered a continuous variable ranging from 0 to 100 for US census tracts; higher scores indicated more affluent tracts.

Missing Data

The primary missingness came from the outcome variable (missing rate: 7.2%) and covariates which mainly included NSES (missing rate: 10.7%), social support (missing rate: 9.6%), depression (missing rate: 9%), sleep (missing rate: 3.4%), and BMI (missing rate: 0.7%). The exposure variables had almost no missingness as we chose people with plausible nutritional data in analysis as depicted in Figure 1. We conducted multiple imputations based on the maximizing expectation method to account for the missing data in both dependent and independent variables.⁵⁷ A list of food groups and nutrients was excluded in our analytical model but served as auxiliary variables to generate our ten imputed datasets, thus optimizing the approximation. Models were run on each of the 10 imputed datasets, and then parameter estimates and variances were calculated, based on Rubin’s rule.^58,59

Statistical Analysis

First, we ran descriptive analytics for demographic and health characteristics in the imputed dataset by four quartiles of overall fatigue scores. Comparison of the variables across outcome quartiles was conducted by Chi-square test for categorical variables, ANOVA for continuous variables, and Kruskal-Wallis tests for non-normally continuous variables. Next, multiple linear regression models were conducted with aMED quintiles as predictor variables (i.e., Quintile 1= aMED ranged between [0, 2); Quintile 2 = [2, 4); Quintile 3 = [4, 5), Quintile 4 = [5, 6); Quintile 5 = [6, 9]) and overall fatigue/energy/weariness as dependent variables. A list of covariates was selected based on previously established knowledge of factors relating to fatigue and Bayesian model averaging (BMA), an application of Bayesian inference to model selection. Then, the cross-product terms of aMED levels and BMI levels (aMED Quintile 1& overweight/obese as the reference group) and aMED levels and diabetes (aMED Quintile 1& without diabetes as the reference group) were added to the regression models to explore whether the relationship of aMED on fatigue depended on body size and diabetes. Isocaloric substitution models were estimated to quantify the trade-off in substituting fish for red and processed meat, whole grain for non-whole grain, and whole fruit for fruit juices while adjusting for all covariates, the total energy intake, and all other energy sources. Fish and red and processed meat were in the unit of ounces of cooked lean meat; grains were in the unit of ounce equivalents, fruits and juice were in the unit of cup equivalents. The p values for statistical tests were 2 tailed and considered statistically significant at an alpha level of 0.05. All analyses were performed with RStudio (version 3.6.2).

RESULTS

The demographic and nutrition characteristics of the analytical sample are shown in Table 1. As shown, the average age (SD) was 78.99 (6.6) and 53% were White. The number of older women in underweight, normal weight, and overweight/obese categories were 67, 1482, and 3014, respectively. The average (SD) of aMED score was 4.24 (1.87), and the average daily intake of whole grains, fruits, and vegetable consumption for the overall participants was 1.54 ounce equivalents, 1.53 cups, and 1.49 cups, respectively. Participants with more favorable fatigue scores were healthier (Quartile 1: 25% self-reported excellent health; Quartile 4: 80% self-reported excellent health), leaner (Quartile 1: BMI of 29.03; Quartile 4: BMI of 26.81), and more active (Quartile 1: MVPA of 40.38 min; Quartile 4: MVPA of 65.06 min); and fewer reported sleep for more than 10 hours (Quartile 1: 34% sleep >= 10 hours; Quartile 4: 17% sleep >= 10 hours, p <.001). Participants with more favorable fatigue scores reported similar intakes of fruits (Quartile 1: 1.48 cup equivalents; Quartile 4: 1.59 cup equivalents, p = .075) but tended to have a higher intake of vegetables (Quartile 1: 1.37 cup equivalents; Quartile 4: 1.62 cup equivalents, p <.001) while having less intake of total energy (Quartile 1: 1645.6 kcal/day; Quartile 4: 1528.45 kcal/day, p <.001), red and processed meat (Quartile 1: 1.77 ounce; Quartile 4: 1.36 ounce, p <.001), fat (Quartile 1: 58.2 gram/day; Quartile 4: 51.12 gram/day, p <.001), carbohydrate (Quartile 1: 207.2 gram/day; Quartile 4: 194.12 gram/day, p <.001), and animal protein (Quartile 1: 42.82 gram/day; Quartile 4: 38.86 gram/day, p <.001).

Table 1:

Baseline characteristics by overall fatigue score quartiles for 4563 Women’s Health Initiative participants age 65 and older

Variable	Quartile 1 (0, 50] (n=1319)	Quartile 2 (50, 65] (n=1290)	Quartile 3 (65,75] (n=883)	Quartile 4 (75, 100] (n=1071)	Total participants (n=4563)	p for Comparison
Age (mean ± SD)	79.85±6.73	79.61±6.41	78.69±6.52	77.45±6.38	78.99± 6.6	p<.001
Education, n (%)						p<.001
High school or less	293(22)	269(21)	163(18)	178(17)	902(20)
Some college	564(43)	487(38)	329(37)	392(37)	1772(39)
College or more	462(35)	534(41)	391(44)	501(47)	1889(41)
Race/Ethnicity, n (%)						p<.001
White	771(58)	705(55)	464(53)	500(47)	2440(53)
Black	360(27)	378(29)	267(30)	345(32)	1350(30)
Hispanic	188(14)	207(16)	152(17)	226(21)	773(17)
NSES (mean ± SD)	72.95±9.5	73.51±9.5	73.54±9.36	73.85±9.09	73.46±9.37	0.201
Social Support* (mean ± SD)	34.79±8.14	36.97±7.18	38.72±6.82	39.78±6.45	37.34±7.5	p<.001
Sleeping hours, n (%)						p<.001
<= 7 hours	353(27)	400(31)	313(35)	378(35)	1444(32)
8~9 hours	512(39)	578(45)	382(43)	507(47)	1979(43)
>= 10 hours	454(34)	312(24)	188(21)	186(17)	1140(25)
Diabetes, n (%)	344(26)	248(19.22)	133(15)	138(13)	866(19)	p<.001
BMI, mean ± SD	29.03±6.18	28.33±5.81	27.21±5.04	26.81±4.88	27.95±5.64	p<.001
Depression, n (%)	189(14)	43(3.3)	16(1.8)	11(1)	259(5.7)	p<.001
Number of Chronic Diseases, mean ± SD	1.15±0.89	0.99±0.81	0.84±0.78	0.72±0.73	0.95±0.83	p<.001
Self-Rated Health, n (%)						p<.001
Excellent/very good	330(25)	598(46)	595(67)	853(80)	2371(52)
Good	675(51)	618(48)	266(30)	213(20)	1776(39)
Fair/poor	314(24)	77(6)	22(2)	5(0.5)	416(9)
LLPA, min/d (mean ± SD)	180.4±50.62	187.9±48.17	192.68±48.81	197.42±47.95	188.84±49.39	p<.001
HLPA, min/d (mean ± SD)	88.66±34.76	97.66±34.29	105.42±35.63	108.51±32.88	99.17±35.19	p<.001
MVPA, min/d (mean ± SD)	40.38±30.34	48.13±32.28	57.92±34.56	65.06±37.07	51.83±34.68	p<.001
Energy, kcal/day (mean ± SD)	1645.6±669.63	1630.23±645.91	1611.39±643.45	1528.45±612.3	1607.53±646.7	p<.001
aMED (mean ± SD)	3.99±1.85	4.18±1.83	4.43±1.83	4.46±1.91	4.24±1.87	p<.001
Fruits, cup equivalents (mean ± SD)	1.48±1.15	1.5±1.22	1.6±1.19	1.59±1.2	1.53±1.18	0.075
Vegetables, cup equivalents mean ± SD	1.37±0.85	1.44±0.9	1.6±0.99	1.62±1.05	1.49±0.95	p<.001
Nuts, ounce equivalents of lean meat mean ± SD	0.87±1.02	0.89±0.97	0.99±1.16	1.03±1.17	0.94±1.07	p<.01
Legumes, ounce equivalents of lean meat mean ± SD	0.36±0.52	0.4±0.6	0.4±0.58	0.45±0.69	0.4±0.6	p<.05
Whole grains, ounce equivalents mean ± SD	1.52±1.3	1.59±1.38	1.53±1.27	1.51±1.31	1.54±1.32	0.273
Fish, ounce cooked lean meat (mean ± SD)	0.67±0.7	0.74±0.79	0.8±0.79	0.78±0.84	0.74±0.78	p<.01
Ratio of monounsaturated to saturated fat	1.18±0.25	1.19±0.25	1.19±0.25	1.23±0.3	1.2±0.26	p<.001
Red and processed meat, ounce cooked lean meat (mean ± SD)	1.77±1.54	1.67±1.38	1.55±1.3	1.36±1.28	1.61±1.4	p<.001
Alcohol, gram/day (mean ± SD)	4.43±11.88	5.22±10.91	5.68±10.52	6.44±13.12	5.39±11.71	p<.01
Carbohydrate, gram/day (mean ± SD)	207.2±88.19	205.22±86.42	204.55±85.69	194.12±85.18	203.06±86.63	p<.001
Fat, gram/day (mean ± SD)	58.2±28.2	56.66±26.93	55.13±27.84	51.12±24.43	55.53±27.03	p<.001
Protein, gram/day (mean ± SD)	65.6±28.98	65.46±27.97	64.83±27.86	61.46±26.49	64.44±27.95	p<.001
Vegetable protein, gram/day (mean ± SD)	22.59±10.39	22.98±10.51	23.08±10.6	22.65±11.51	22.81±10.73	0.598
Animal protein, gram/day (mean ± SD)	42.82±22.33	42.57±21.33	41.74±21.69	38.86±20.65	41.63±21.61	p<.001

Note. aMED: Alternate Mediterranean diet. NSES: Neighborhood Social Economic Status. BMI: Body Mass Index; LLPA: low light physical activity; HLPA: high light physical activity; MVPA: moderate to vigorous physical activity.

^*Social support total score ranged from 9 to 45. Depression: items were a shortened version of the Center for Epidemiological Studies Depression Scale (CES-D), a score greater or equal to 0.06 as depressive. The number of self-reported chronic diseases was measured as the sum of chronic obstructive pulmonary disease, cardiovascular disease, cerebrovascular disease, cancer, osteoarthritis, and cognitive impairment.

Regression models showed that Quintile 5 (i.e., aMED score ranging between [6, 9]) was favorably associated with overall fatigue and energy in minimally, partially, and fully adjusted models, compared with aMED Quintile 1 (Table 2). As shown, after controlling for various levels of physical activities, the magnitude of Quintile 5 coefficient slightly diminished; however, it was significantly associated with a 2.99 (95%CI: 0.88, 5.11) improvement in overall fatigue score, 4.01 (95%CI: 1.51, 6.53) improvement in energy score, and 2.47 (95%CI: 0.24, 4.70) improvement in weariness score, compared with aMED Quintile 1.

Table 2.

Linear regression analysis for the associations of aMED levels as independent variables with the overall fatigue, energy and weariness outcomes (n=4563)

		M1 β(95%CI)	M2 β(95%CI)	M3 β(95%CI)
Overall Fatigue	Quintile 1	Reference	Reference	Reference
	Quintile 2	0.32 (−2.07, 2.71)	0.43(−1.58, 2.44)	0.39(−1.6, 2.39)
	Quintile 3	2.81 (0.36, 5.24)*	2.15(0.09, 4.22)*	1.99(−0.06, 4.04)
	Quintile 4	2.91 (0.40, 5.42)*	2.19(0, 4.37)	1.92(−0.24, 4.09)
	Quintile 5	5.32 (2.94, 7.70)*	3.35(1.22, 5.48)*	2.99(0.88, 5.11)*
Energy	Quintile 1	Reference	Reference	Reference
	Quintile 2	1.09(−1.79, 3.97)	1.16(−1.24, 3.56)	1.13(−1.24, 3.50)
	Quintile 3	4.40(1.36, 7.45)*	3.43(0.87, 5.99)*	3.18(0.65, 5.71)*
	Quintile 4	4.60(1.53, 7.66)*	3.40(0.78, 6.03)*	2.98(0.38, 5.59)*
	Quintile 5	7.34(4.46, 10.22)*	4.61(2.08, 7.14)*	4.01(1.51, 6.53)*
Weariness	Quintile 1	Reference	Reference	Reference
	Quintile 2	−0.17(−2.5, 2.16)	−0.01(−2.11, 2.10)	−0.06(−2.16, 2.05)
	Quintile 3	1.63(−0.81, 4.08)	1.39(−0.87, 3.65)	1.30(−0.95, 3.56)
	Quintile 4	1.48(−0.98, 3.94)	1.36(−0.93, 3.65)	1.26 (−1.02, 3.54)
	Quintile 5	3.76(1.44, 6.08)*	2.59(0.35, 4.82)*	2.47 (0.24, 4.70)*

Note. Quintile 1. aMED ranged between [0, 2).

Quintile 2. aMED ranged between [2, 4).

Quintile 3. aMED ranged between [4, 5).

Quintile 4. aMED ranged between [5, 6).

Quintile 5. aMED ranged between [6, 9].

M1: unadjusted model

M2: adjusted for age, race/ethnicity, education, NSES, self-rated health, depression, BMI, number of chronic diseases, sleeping hours, social support, total energy, diabetes

M3: adjusted for age, race/ethnicity, education, NSES, self-rated health, depression, BMI, number of chronic diseases, sleeping hours, social support, total energy, diabetes, low light physical activity time, high light physical activity time, moderate to vigorous physical activity time.

^*Significant at p<0.05

The number of chronic diseases: the sum of chronic obstructive pulmonary disease, cardiovascular disease, cerebrovascular disease, cancer, osteoarthritis, and cognitive impairment

NSES: Neighborhood Social Economic Status

The interaction between aMED quintiles and BMI levels or aMED quintiles and diabetes were non-significant for all three fatigue outcomes. For all three outcomes, the interaction showed underweight:aMED Q2, p > .05; underweight:aMED Q3, p > .05; underweight:aMED Q4, p > .05; underweight:aMED Q5, p > .05; normalweight: aMED Q2, p > .05; normalweight: aMED Q3, p > .05; normalweight: aMED Q4, p > .05; normalweight: aMED Q5, p > .05 & aMED Q2:Diabets, p > .05; aMED Q3:Diabets, p > .05; aMED Q4:Diabets, p > .05; aMED Q5:Diabets, p > .05. For ease of interpretation, we simulated the adjusted mean values across the regression models with interaction terms. See Supplementary Figure 1.

Figure 2 showed that regressions estimates, namely the mean score change in overall fatigue, energy, and weariness from substituting one-ounce fish for the same number of ounces of red and processed meat were 1.69 (95% CI: 0.99, 2.39), 2.22 (95% CI: 1.36, 3.07), and 1.11 (95% CI: 0.34, 1.88) points on overall fatigue, energy, and weariness scores. Of note, one ounce red and processed meat represented 62% of red and processed meat that WHI participants consumed on average; one ounce equivalent of fish approximated 135% of average daily consumption of fish in the WHI data. The regressions estimates, namely the mean score change in overall fatigue, energy, and weariness from substituting one-ounce equivalents of whole grain for the same amount of non-whole grain were 0.68 (95% CI: 0.20, 1.16) and 0.90 (95% CI: 0.39, 1.4) points in overall fatigue and weariness scores, respectively. One-ounce equivalents represented approximately 37% of non-whole grain and 65% of whole-grain consumption for WHI participants per day on average. The regressions estimates from the substituting of one-cup equivalents of whole fruit for the same amount of fruit juice were not significant. The regression estimates were 0.2, (95% CI: −0.66, 1.06), 0.28 (95% CI: −0.78, 1.35), and 0.14 (95% CI: −0.75, 1.02) for overall fatigue, energy, and weariness, respectively.

Figure 2.

Regression estimates from Isocaloric Substitution Models of the mean score change in overall fatigue from substituting one-ounce cooked lean meat from fish for red and processed meat, substituting one-ounce equivalents whole-grain for non-whole grain, substituting one cup equivalent whole fruit for fruit juice.

Regression estimates from Isocaloric Substitution Models of the mean score change in energy from substituting one-ounce cooked lean meat from fish for red and processed meat, substituting one-ounce equivalents whole-grain for non-whole grain, substituting one cup equivalent whole fruit for fruit juice.

Regression estimates from Isocaloric Substitution Models of the mean score change in weariness from substituting one-ounce cooked lean meat from fish for red and processed meat, substituting one-ounce equivalents whole-grain for non-whole grain, substituting one cup equivalent whole fruit for fruit juice.

DISCUSSION

This paper described the associations between the Mediterranean diet and overall fatigue/energy/weariness in a general population of women aged 65 years and older. The mean aMED score of those older women was 4.24 (1.87) on a range between 0 and 9, similar to the score of 4.3 reported in a previously published WHI study of a younger (mean age = 64) but larger cohort (sample size > 90000).⁶⁰ The aMED score of 4.3 out of 9 indicated a moderate adherence to the Mediterranean diet among community-dwelling older women. Due to a lack of aMED classification, the moderate category was based on traditional Mediterranean diet scores, ^44,61 whose tertile classification was similar for tertiles of aMED across samples.⁶² Meanwhile, the average daily consumption of 1.53 cups,1.49 cups, and 1.54 ounce equivalent of fruits, vegetables, and whole grains, respectively, indicated a relatively low consumption. According to the Healthy Mediterranean-diet recommendation offered by 2020–2025 Dietary Guidelines for Americans, the recommended amount for vegetables, fruits, and whole-grain were 2.5 cups, 2.5 cups, and 3 ounces for a person who eats 2000 calories per day.⁶³ Due to the wide variability, a distribution analysis showed in the current sample only 16% met the fruit recommendation, 13%met the vegetable recommendation, and 10.5% met the whole grain recommendation.

After adjusting for physical activity levels, we found that the magnitude of the relationship between aMED and fatigue declined, but a higher adherence to the Mediterranean diet was still significantly associated with more favorable fatigue/energy/weariness scores, indicating the unique contribution of the Mediterranean diet in addressing fatigue. Our finding was similar to a previous prospective study among older women with type 2 diabetes, where the Mediterranean diet significantly reduced fatigue, but the associations were attenuated when leisure-time physical activity was considered.²² In our previous work, we found the importance of moderate and vigorous physical activity in relation to fatigue/energy/weariness.⁶⁴ Specifically, we found 34 minutes of moderate and vigorous physical activity was associated with a 1- to 2-point improvement in fatigue/energy/weariness outcomes. In the current work, we found that optimal adherence to a Mediterranean diet was related to 2.5- to 4-point improvement in fatigue/energy/weariness compared to the lowest adherence. As such, it might be reasonable to assume that the combination of physical activity and the Mediterranean diet will act synergistically in mitigating fatigue. This assertion echoes prior research that suggested diet and physical activity have complementary and interactive effects on many physiological parameters such as energy, lipid, glucose, and metabolic balance.⁶⁵ Noticeably, physical activity is part of the Mediterranean lifestyle and is consistently recommended with the Med diet pyramids.⁶⁶ Thus, based on our findings, adherence to the Mediterranean lifestyle might relate to better fatigue scores in older women.

A minimal clinically important difference is essential to detect the smallest difference in fatigue score that these older women perceive as beneficial/improved fatigue feeling.. Prior research showed for the SF-36 vitality subscale, the minimal clinical difference ranged from 7.3 to 11.3 for improvement.⁶⁷ However, these numbers were drawn from people with rheumatoid arthritis and systemic lupus erythematosus, which might not be relevant to generally healthy community-dwelling older women. Another study revealed an average change of 5–10 points in SF-36 vitality as the minimal clinically important difference to predict job loss, hospitalizations, and mortality.⁶⁸ The combined effect of physical activity and the Mediterranean diet may or may not yield a clinically important difference; however, these lifestyle modalities may be suggested as health recommendations for women complementary to other evidence-based treatments of fatigue.

Non-significant interactions indicated a consistent diet-fatigue relationship across BMI levels and diabetes status. The similar Mediterranean diet-fatigue relationship in women with high and low BMI and persons with and without diabetes may show that salutary effects of the Mediterranean diet on fatigue transcend several metabolic factors. That is to say, physiological processes by which the Mediterranean diet affects fatigue may work through the routes that are not involved in fat and glucose metabolism. Unfortunately, we did not find previous publications that support this assertion; as such, further research is needed.

The ISM showed that one ounce fish substituted for red and processed meat relates to about a 2-point improvement in fatigue/energy. One ounce is about 62% of the red and processed meat one participant consumed on average. After the one ounce substitution, the average daily consumption of fish is estimated at 1.74 ounce (12.18 ounce per week), which still lags behind the recommended by Dietary Guidelines for Americans 2020–2025 15 ounce target.⁶³ This result may be especially useful for people loyal to animal protein consumption. Both fish and red meat are from animal sources; however, fish has a higher level of omega-3 fatty acids, so the beneficial effect of fish over red and processed meat may lie in levels of omega-3 fatty acids. Previous studies suggested omega-3 fatty acids were related to the anti-inflammatory process and the occurrence of fatigue.⁶⁹ Future researchers may consider testing the mechanism of omega-3 fatty acids by checking if the magnitude of the association alters after including omega-3 fatty acids in the model. However, there are other dietary differences between the two forms of animal protein which might serve as the potential mechanism.

We observed small benefits to substituting whole grain for the same amount of non-whole grain. However, it is challenging to study the mechanism of the substitution benefit because whole grain and non-whole grain are markedly different.⁷⁰ For example, compared with non-whole grain, whole-grain foods tend to be higher in phytochemicals, fiber, and some B vitamins.⁷¹ Fiber and phytochemicals were reported to be related to fatigue improvement through the cholesterol-lowering effect, protection against oxidative stress, inflammation, and beneficial modifications in gut microbiome taxonomy and metabolites.²⁴ Thus, pending further research, one cautions interpretation might be that whole grains may be more appropriate in the context of fatigue than non-whole grains. Our findings from ISM (i.e., fish in replace for red and processed meat and whole-grain for non-whole grain) supported that the beneficial effect of the Mediterranean diet might not be due to the action of a single food but rather the accumulated or synergic impact of several featured foods from Mediterranean diet or the overall Mediterranean diet lifestyle.

Our finding that the impact on fatigue from the substitution of one-cup equivalent of whole fruit for the same amount of 100% fruit juice was not significant provided preliminary evidence that 100% juice is as healthy as whole fruits concerning fatigue in this population. Although, there is currently no description about the role of 100% fruit juice in the Mediterranean diet, our finding resonates with the general healthy eating recommendations by 2020 – 2025 Dietary Guidelines for Americans, where both whole fruits and 100% fruit juices are promoted.⁶³ Our finding also resonates with experts participating in a roundtable discussion who stated a lack of science-based reason to restrict access to 100% fruit juice in public health nutrition policy.³⁵ The expert panel also estimated that reducing access to 100% fruit juice could lead to some unintended consequences such as reduced daily fruit intake and increased consumption of sugar-sweetened beverages.³⁵The strengths of this study include dietary pattern rather than individual foods as main exposure variables, the use of the relatively novel ISMs to examine the reallocation of food and dietary variables, the inclusion of physical activities, and finally the generalized older population. First, we adopted Mediterranean dietary pattern as our exposure variable, especially we used the aMED score, which was tailored for people in non-Mediterranean countries but was reliable and valid measure in assessing adherence to Mediterranean diet.⁴³ Such that we were able to assess an integrated picture of frequency, proportion, and variety of food, drinks, and consider the substitution effect, synergic and antagonistic interactions between foods and nutrients. Second, by applying ISM, we were able to examine calories allocation among various foods groups to achieve improved fatigue scores. ISM removes confounding effects from total energy intake and allows for the study of multiple food groups and also the macronutrient composition while holding total energy constant to identify a better dietary pattern,⁴⁰ thus ISM can serve as another method to triangulate results in instances when RCT is not available. Third, we controlled physical activities and were able to find that our conclusion of Mediterranean diet remained after adjusting for physical activities. Besides, our physical activities were accelerometer measured and were calibrated specifically for older women. These objective physical activities are more nuanced and reliable than self-reports. Finally, we examined the relationship between dietary variables and fatigue in community-dwelling older women rather than a specific clinical population, thus, our findings are relevant to the majority of older women.

The limitations of this study include the sample, measurements, and cross-sectional design. First, the sample studied here, while racially diverse, was limited to older women. Previous research, however, did find differences between men and women concerning the metabolic effects of the Mediterranean diet.^72,73 Second, prior researchers has confirmed the existence of systematic bias in dietary self-reports and provided methods of correcting for measurement error.⁷⁴ However, because the FFQ was updated from the original WHI FFQ to reflect the changing food supply, calibration equations developed within the WHI to account for measurement errors were not applied. Further, because an FFQ is composed of a pre-specified food list, it may not reflect the eating patterns of any given population.⁷⁵ Third, the NSES was collected around the year 2000 census, so it might not be a reliable representative of the true NSES. Fourth, despite the possibility of examining a longitudinal relationship between the Mediterranean diet and fatigue using WHI data, the sparseness of collection interval and out-of-date information may not represent a contemporary diet pattern, thus limiting our choice of analysis to the cross-section. The ISMs in the cross-sectional study should not be considered equivalent to a RCT in which real behavior changes occur over time and cause future outcomes.

Conclusions

We showed that the Mediterranean diet was uniquely associated with fatigue, even after controlling physical activity levels. The Mediterranean dieťs beneficial effect on fatigue seems similar among people with or without diabetes and obesity. Substituting fish for red and processed meat and whole grains for non-whole grains was associated with improvements in overall fatigue, energy, and weariness scores. Similar substitution of fruit for fruit juice was not associated with fatigue. These findings indicate that Mediterranean diet might have the potential to break the escalation of fatigue. Future researchers may consider including energy-fatigue measures in RCTs of older adults in which increasing adherence to the Mediterranean diet and reducing fatigue are intervention targets so that improvements in fatigue can be directly tested.

Take away points.

• Higher adherence to Mediterranean Diet was associated with 2.5-to-4-point improvements in fatigue, energy, and weariness scores, compared with the lowest adherence to Mediterranean Diet.

• The Mediterranean diet was independently associated with more favorable self-reported fatigue scores, even after controlling health and lifestyle factors.

• Substituting fish for red and processed meat and whole for non-whole grains was associated with more favorable fatigue scores, whereas substituting whole fruit for juice was not.