Food consumption is associated with frailty in edentulous older adults: evidence from the ELSI-Brazil study

Abstract This cross-sectional study aimed to evaluate the association between food consumption (meat, fish, and fruits and vegetables), anthropometric indicators (body mass index, waist circumference, and waist-to-height ratio), and frailty; and to verify whether these associations vary with edentulism. We used data from 8,629 participants of the Brazilian Longitudinal Study of Aging (ELSI-Brazil) (2015-16). Frailty was defined by unintentional weight loss, weakness, slow walking speed, exhaustion, and low physical activity. Statistical analyses included multinomial logistic regression. Of the participants, 9% were frail and 54% pre-frail. Non-regular meat consumption was positively associated with pre-frailty and frailty. Non-regular fish consumption, and underweight were associated only with frailty. Models with interactions reveled a marginal interaction between meat consumption and edentulism (p-value = 0.051). After stratification, non-regular meat consumption remained associated with frailty only in edentulous individuals (OR = 1.97; 95%CI 1.27-3.04). Our results highlight the importance of nutritional assessment, oral health, and public health-promoting policies to avoid, delay and/or reverse frailty in older adults.


Introduction
Frailty is defined as a biologic syndrome of decreased reserve and resistance to stressors derived from cumulative declines in physiologic systems that might increase adverse outcomes 1 . As a multidimensional biologic syndrome, frailty involves the interaction of physiological, nutritional, psychological, cognitive, and social factors 2 . However, it is commonly assessed through physical components named frailty phenotype, including unintentional weight loss, weakness, slow walking speed, self-reported exhaustion, and low physical activity 1 . Considering the nutritional status, the association with frailty is probably bidirectional, either deriving 3 or leading to inadequate food consumption. Indeed, food and nutritional intake depend on intrinsic and extrinsic factors. Within the intrinsic factors, we highlight the edentulism, i.e., the state of being without natural teeth, a high-prevalent oral health problem leading to a poorer eating ability that might cause frailty in older adults [4][5][6] .
A balanced diet pattern is fundamental to delay frailty onset among older adults 7 . It includes adequate energy, protein, and other nutrient intakes, associated with a wide range of food groups and low fast-food consumption. Regarding protein intake, the beneficial effect on frailty might be linked to maintaining muscle mass and preventing sarcopenia during the aging process 8 . However, the literature reports contradictory results regarding the role of either energy or protein intake on frailty 9 . Recently published international works have suggested that total protein intake and protein source are crucial for preventing frailty 10,11 . Proteins from animal sources, such as meat, eggs, and dairy, have shown stronger associations with frailty 10 since they present a higher quantity of essential amino acids (i.e., leucine) linked to the muscle protein anabolism in older adults 11 .
Findings from a previous study showed that dietary patterns with elevated protein intake, including higher fish consumption, were associated with a lower frailty prevalence 12 . Fish consumption has been recommending as part of a healthier dietary because it is a protein source, omega-3 Poly-unsaturated Fatty Acids (PUFAs), suggesting a protective effect on frailty 13 . Moreover, omega-3 PUFAs intake has the potential to prevent both muscle mass and strength losses, considering its anti-inflammatory effects 14 .
A systematic review showed consistent evidence that a high-quality general dietary pattern is inversely associated with the frailty risk 15 . This effect might be mediated by an elevated fruit and vegetable intake, supported by recent results from cohort studies demonstrating a protective effect on frailty onset 16,17 . Fruit and vegetable intake might decrease frailty risk due to different mechanisms. They are important sources of dietary fibers, micronutrients, and antioxidants 18 negatively associated with frailty 17 . Moreover, fruit and vegetable intake stimulate the immunologic system 19 and decreases inflammatory response 20 , both described as linked to frailty among older adults 21 .
The nutritional status goes beyond the food consumption indicators and also includes the anthropometric indicators. The adequate anthropometric indicators have been described as protective against frailty due to their link with sarcopenia 22 . Cross-sectional studies have shown a U-shaped association of Body Mass Index (BMI) with frailty in older adults 23,24 , in which both low and high BMI increases frailty risk. Even though suitable BMI varies across these studies, their findings highlight the importance of keeping adequate BMI to decrease frailty prevalence. Nevertheless, considering BMI limitations on sensitivity and distinction of body composition, other studies also investigated waist circumference and its association with frailty. These studies evidenced an association between frailty and elevated waist circumference 25,26 .
In Brazil, a population-based study named Frailty in Brazilian Older People Study (in Portuguese, FIBRA) demonstrated a significantly higher prevalence of frailty among older adults with low BMI and elevated waist circumference 26 . Despite being a multicenter and representative study of the older adult population from seventeen cities, it is not nationally representative. Therefore, nationally representative studies are needed to a broader comprehension of nutritional status on frailty. According to data from the Brazilian Longitudinal Study of Aging (EL-SI-Brazil), frailty prevalence was 9% in older adults and increased with age, summing nearly 21% among those aged 70 years and over 27 .
Thus, the aims of the current study were: (1) to evaluate the association between the food consumption indicators (meat, fish, and fruits and vegetables consumption) and frailty; (2) to evaluate the association between anthropometric indicators (BMI, waist circumference, and waistto-height ratio) and frailty; and (3) to assess whether the association of food consumption indicators with frailty vary with edentulism. For these purposes, we used a nationally representative sample of older Brazilian adults derived from the ELSI-Brazil.

Study design and population
This cross-sectional study included baseline data of the ELSI-Brazil, conducted in 2015-16. ELSI-Brazil is a nationally representative, population-based longitudinal study of 9,412 community-dwelling older adults aged 50 years and over from 70 municipalities of the five great regions in Brazil. The ELSI-Brazil sampling used a design with selection stages, combining stratification of primary sampling units (municipalities), census tracts, and households. All older adults aged 50 years and over residents in the selected households were eligible for interview, including 328 participants who needed a proxy. Further details of sampling design and procedures are described elsewhere 28 .
ELSI-Brazil followed the standards set by the Declaration of Helsinki and was approved by the ethics board of the Oswaldo Cruz Foundation, Minas Gerais (protocol 34649814.3.0000.5091). All participants signed an informed consent form.

Frailty
Frailty was defined according to the number of positive components of the frailty phenotype 1 . Each component was assessed as follows 27 : (1) unintentional weight loss, assessed through self-reported unintentional weight loss higher than > 4.5 kg in the past three months; (2) weakness, directly assessed through the best of three grip strength measures obtained in the dominant hand, using a hand dynamometer. Weakness was defined according to the lowest quintile, after adjusting for sex and BMI quartiles, being bedridden during the interview, and the inability to perform the test (i.e., those who tried but were not able to complete the test); (3) slow walking speed, directly assessed through the 3-meter timed walk test (in seconds), considering the best of two attempts. Slow walking speed was defined according to the highest quintile of time, stratified by sex, height, and the inability to perform the test 29 ; (4) self-reported exhaustion, assessed through two questions from the Center for Epidemiologi-cal Studies (CES-D): "How often in the last week did you feel that everything did was an effort?" and "How often in the last week did you feel that could not get going?" 30 . Exhaustion was defined when reporting frequencies greater than 3-4 days in at least one question; and (5) low physical activity, calculated through a physical activity score considering the metabolic equivalents per week in kilocalories (kcal) based on the short-form of the International Physical Activity Questionaire (IPAQ) 31 . Low physical activity was defined according to the lowest quintile, stratified by sex.
The dependent variable was categorized into "frailty" when the participant showed three or more positive components, "pre-frailty" when the participant showed one or two, and "non-frailty" when the participant did not show any positive component of the frailty phenotype 1 .

Food consumption indicators
Consumption was assessed according to the following food consumption indicators: (1) meat consumption, considering regular consumption as eating red (beef, pork, lamb) or white (poultry) meats in five or more days per week 32 ; (2) fish consumption, considering regular consumption as eating fish in one or two days per week 33 ; and (3) fruit and vegetable consumption, considering regular consumption as eating fruits or vegetables in five or more days per week, excluding fruit juice 32 .

Anthropometric indicators
Anthropometric indicators included BMI, waist circumference, and waist-to-height ratio: -BMI: calculated dividing weight in kilograms by height in meters squared (kg/m 2 ). Weight and height were objectively measured using standard protocols. The participants were asked to wear soft clothes, be barefoot, and remove all accessories and objects from their pockets. Both feet were positioned on a calibrated portable digital scale platform, brand Seca, model 813, equally distributing their weight on both feet, maintaining the arms extended along the body side, and looking at a horizontal line. Height was measured by a portable vertical stadiometer of the brand Nutri-Vida, without shoes and in the stand position. The interviewer placed the interviewee's head in the Frankfurt plane to record the measure. BMI cutoffs were according to the participant's age. For those aged up to 60 years, "underweight" (< 18.5 kg/m 2 ), "adequate weight" (18.5 to 24.9 kg/m 2 ), and "overweight" (> 24.9 kg/m 2 ) groups were based on the World Health Organization (WHO)'s criteria 34 . For those aged 60 years and over, "underweight" (< 22.0 kg/m 2 ), "adequate weight" (22.0 to 27.0 kg/ m 2 ), and "overweight" (> 27.0 kg/m 2 ) groups were based on the Lipschitz's criteria 35 .
-Waist circumference: measured with an inextensible metric tape, brand Seca, positioned at the midpoint between the 10 th rib and iliac crest with the participant standing, barefoot, feet apart, arms alongside the body, with the raised shirt, and during the expiratory phase. Among those aged up to 60 years, waist circumference was dichotomized as recommended by WHO 34 , in "adequate" (< 80 cm for women and < 94 cm for men) or "elevated" (≥ 80 cm for women and ≥ 94 cm for men). For those aged 60 years and over, we used a recent cutoff reported for the older Brazilian adults: "adequate" (< 88.7 cm for women and < 96 cm for men) or "elevated" (≥ 88.7 cm for women and ≥ 96 cm for men) 36 .
Anthropometric measures were performed twice, using the mean of both measures. Further details about procedures and protocols can be consulted at "Interview Handbook" from EL-SI-Brazil 38 (available at the ELSI-Brazil homepage: http://elsi.cpqrr.fiocruz.br/en/).

Edentulism
Edentulism was defined as the absence of any natural teeth ("no" or "yes"), collected using the following self-reported question: "How many teeth do you have?".

Potential confounding variables
Potential confounding variables were based on statistically variables associated with frailty in previously published studies 16,27,39 : -Sociodemographic characteristics: sex (female or male); years of age ("50-59", "60-69", "70-79", or "80 years and over"); marital status (living with a partner or not, i.e. single, widowed or divorced); education, considering the complete years of schooling ("< 8", "8-11" or "≥ 12"); and household income per capita, based on the total monthly gross household income divided by the number of residents, categorized into terciles, from the poorest to the richest 40 .
-Health-related characteristics: current smoking status investigated through the question "Do you currently smoke?" ("no" or "yes", i.e., whether the individual positively answered the question, independently of the number of smoking cigarettes and the duration of the smoking habit); self-rated health ("excellent/very good or good", "fair", or "bad/very bad"); the number of chronic conditions diagnosed by a physician based on self-report, including hypertension, diabetes, depression, cancer, arthritis or rheumatism, high cholesterol, stroke, and cardiovascular disease ("none", "one" or "two or more"); and activity limitations, measured by participants' self-reports of any difficulty to carry out at least one out of six basic Activities of Daily Living (ADLs), including walking in the same floor, transferring, toileting, bathing, dressing, and eating 41 ("no" or "yes").

Statistical analyses
Initially, we described differences in the variables' distribution according to frailty using the Pearson chi-square test with Rao-Scott correction. We used multinomial logistic regression to estimate odds ratio (OR) and their 95% confidence intervals (95%CI) to examine the strength of the association between the independent variables (food consumption indicators [meat, fish, and fruit and vegetable consumption], and anthropometric indicators [BMI, waist circumference, and waist-to-height ratio]) and frailty, using the non-frailty category as the reference. Adjusted multivariate analysis was performed separately for each of the three anthropometric indicators because of collinearity, including all food consumption indicators in each model. We made the adjustments in the sequential models as follows: (1) sociodemographic characteristics (Model 1); (2) health-related characteristics (Model 2) and, finally, (3) Models 1 and 2 together (fully adjusted model -Model 3). We also tested the interaction of edentulism with the food consumption indicators previously associated with frailty in Model 3 and plotted the results in charts.
All analyses were performed using STATA software (Stata Corp., College Station, United States), version 14.0, using the svy command, which allows us to consider the complex design and surveys weights. The significant level was set at 5%.

Results
Of the 9,412 ELSI-Brazil participants, 8,629 had complete information on frailty and were included in the current analyses. Those included mainly were female (53.7%), with a mean age of 62.2 years (± 9.6). They tended to be younger and showed a lower prevalence of low-educated (p < 0.05) than those excluded. The prevalence of frailty was 9.1%, pre-frailty was 53.5%, and non-frailty was 37.4%. Non-regular meat and fruit and vegetable consumption were reported by 23.9% and 20% of the participants, respectively. More than half participants (55.6%) have non-regular fish consumption. Regarding anthropometric indicators, 7.4% of the participants had underweight, and 61.3% were overweight, according to BMI. Prevalence rates of elevated waist circumference and waistto-height ratio were 62.6% and 70.2%, respectively. The characteristics with significant different distribution across frailty categories were: meat consumption, fish consumption, fruit and vegetable consumption, BMI, waist circumference, waist-to-height ratio, sex, age, marital status, education, per capita household income, self-rated health, number of chronic conditions, activity limitations, and edentulism. Table 1 shows these descriptive results.
Considering that the estimates in the fully adjusted models were similar when entering the anthropometric indicators (BMI, waist circumference, and waist-to-height ratio), the results from models with BMI, the most commonly used anthropometric indicator, were shown separately in Table 2. The results from models of the association between frailty and nutritional status adjusted and for waist circumference and waist-to-height ratio were presented in Table 3. According to Table 2, after adjustments, pre-frail were more likely to non-regularly consume meat (OR = 1.22, 95%CI 1.03-1.44) than non-frail participants, as well as frail older adults (OR = 1.44, 95%CI 1.10-1.89). Also, frail older adults were more likely to show non-regular fish consumption (OR = 1.38, 95%CI 1.11-1.72) and being underweighted (OR = 1.74, 95%CI 1.17-2.58) than non-frail older adults.
Additionally, we fitted models with the interaction terms between food consumption indicators previously associated with frailty in Model 3 (meat and fish consumption) and edentulism, finding a marginal interaction between meat consumption and edentulism (p-value = 0.051). After splitting the odds of frailty by categories of edentulism, we found that non-regular meat consumption remained positively associated with frailty only among those who were edentulous (OR = 1.97; 95%CI 1.27-3.04) (data not shown). As plotted in Figure 1 (A), it means that the probability of frailty increased from 7.5% (95%CI 6.2-8.9%) in those with regular meat consumption to 10.3% (95%CI 7.4-13.2%) in those with non-regular meat consumption, while the probability of frailty remained around 8% among non-edentulous, independently of meat consumption. The different pattern among edentulous and non-edentulous was not observed for fish consumption (B).

Discussion
Our results showed a prevalence of frailty and pre-frailty of 9.1% and 53.5%, respectively, among older Brazilian adults. After adjustments, pre-frailty was associated with meat consumption, whereas frailty was associated with meat and fish consumption and being underweight. Models considering interaction terms reveled that non-regular meat consumption increased the odds of frailty only among those who were edentulous, despite their marginal statistical association.
Studies encompassing the role of food consumption on frailty show that some food and nutrient consumptions, including protein and overall diet quality, are essential to avoid this syndrome 7,15,42 . In the current study, meat consumption was associated with frailty, and pre-frailty, corroborating longitudinal studies 7,15,42 . Overall, meat is a good protein source, which is fundamental to muscle mass anabolism and maintenance 11 , and carries out a diversity of functions: structural, enzymatic, hormonal, protection, and transportation. Moreover, meat is a lipid and complex B vitamins source. Lipides, along with proteins, contribute to achieving an adequate energy intake 42 . Complex B vitamins, mainly vitamin B 12 or cyanocobalamin, acting as a central nervous system metabolism cofactor. Lower vitamin B 12 may cause motor and sensorial impairments, weakness, numbness in the distal limb, balance problems, and gait ataxia 43 .
However, our results demonstrated that non-regular meat consumption was statistically associated with frailty only when having edentulism, corroborating the evidence of a poor nutrition pathway between worse oral health and frailty, although increasing edentulism with age in our sample (from 15.3% of the participants aged 50-59 to 63.1% of those aged 80 years and over, data not shown). Edentulism is associated with overall lower consumption of essential and high-er of ultra-processed food, independent of dental prosthesis use. These associations are probably explained by the poorer eating ability and mas- ticatory efficiency among edentulous individuals, difficulty the consumption of hard-to-crew foods such as meat 5 . Therefore, other easy-to-crew protein sources intake should be encouraged, such as beans, traditional in Brazilian cuisine, and the expansion of oral health teams among primary care.
In the current study, fish was measured separately from red and white meat. Fish is a vitamin D source and lower vitamin D concentration determines frailty onset within three years 44 . Moreover, most vitamin D sources are also rich in omega-3 PUFAs, an essential nutrient to reduce inflammation 14 . Thus, less fish consumption also raises frailty risk 45 .
Despite prior evidences of higher fruit and vegetable consumption leading to a lower frail- ty risk 16,17 , our findings do not demonstrate an association. These food groups include a wide range of vegetables with high nutrients content. They are energy, dietary fibers, vitamin A and C, minerals sources, and more recently described as antioxidants, phytoestrogens, and anti-inflammatory agents due to phytochemicals 18 . One hypothesis for the absence of an association with frailty in the ELSI-Brazil sample is the few years of education of older Brazilian adults: 63.6% of the participants have less than eight years of education. This group concentrates the higher prevalence of the non-regular fruit and vegetable consumption (76.6%) and has limited income resources (mean per capita household income of R$ 832.00 corresponding to 323 dollars at the beginning of 2015) to maintain a more appropriate food variety intake 46 . Historically, inadequate food ingestion was suggested as the initial part of the frailty development cycle in aging, because it contributes to malnutrition and weight loss 1 . Therefore, the relationship between malnutrition and frailty has been studying for years. Our results showed that being underweight was associated with frailty. Underweight has been related with skeletal muscle mass decrease due to an imbalance between muscle protein degradation and synthesis related to sarcopenia, cachexia, and sedentarism. All those are described as frailty risk factors 47 , including evidence in Brazilian studies 23,24,26 .
On the other hand, although international and Brazilian growing evidence [23][24][25][26]48,49 have linked overweight to frailty, we did not observe this association in the current study, independently of the anthropometric indicator used (elevated BMI, elevated waist circumference, or elevated waist-toheight ratio). A 22-year follow-up study showed that middle-aged and obese adults were five times more likely to develop frailty than individuals with adequate weight 50 . Obesity has been associated with a wide range of inflammatory markers probably implicated in the genesis of frailty 51 and the decrease of anabolic hormones necessary to maintain physiologic systems' integrity and function 50 . Lower levels of anabolic hormones could lead to worse physical functioning and reduced muscle strength, contributing to frailty 52 . The absence of association in the current study might be partially explained by different cutoffs used in the above-mentioned studies. However, we also used waist circumference and waist-to-height ratio to measure overweight, which are described as more accurate abdominal obesity indicators in older ages 51 . Again, we did not find any association with frailty. Another explanation is that we did not use weight estimates among those participants whose objectively measured weight was not possible, such as bedridden participants. This approach might have excluded a higher proportion of frailty participants from the analysis. And finally, survival bias might occur in cross-sectional studies, once the probability of survival is lower among overweight and frail individuals.
Older-adult long-term care is a worldwide concern linked to aging due to common health problems at older ages, including frailty. Frail and pre-frail older adults tend to overload health care services and contribute to significant health costs' expansion. Therefore, efforts to enlarge independence and maintain adequate nutritional status are necessary. The association of underweight and non-regular meat consumption with frailty is highlighted from our results, which demand However, several gaps related to adequate nutritional status in older adults remain, and more researches are necessary. We suggest that researchers establish cutoffs for frailty components, such as weakness and slow walking speed, to facilitate their use in routine nutritional practice.
To the best of our knowledge, this was the first study encompassing an older-adult nationally representative sample in Brazil about nutritional status indicators and frailty. Our results highlight the comprehensiveness of care, in which oral health should be along with nutrition to improve health in aging. It permits the diagnosis of underweight and food consumption indicators inadequacies associated with frailty. Therefore, using a large nationally representative sample of older adults living in Brazil and including a methodological rigor, we premise that oral health interventions might improve nutritional status, and consequently avoid, delay or reverse frailty. Moreover, we included in our analyses participants who were unable to perform the objective tests to classify frailty as having the worst performance stead of missing data, to avoid differential losses in frailty group.
However, some limitations should be raised. First, the cross-sectional design precludes us from establishing causality between anthropo-metric/food consumption indicators/edentulism and frailty among older adults. Longitudinal studies concerning food consumption [15][16][17] , and edentulism [4][5][6] , corroborate the direction of our findings (i.e., these variables leading to frailty), although prior evidences of bidirectionality 3,53 . Second, the ELSI-Brazil also does not include any muscle mass indicator to include sarcopenia in the analyses. Third, the exclusion of older and low-educated older adults tended to attenuate the coefficients leading to non-observation of marginally associated variables.

conclusion
Our results showed that non-regular meat and fish consumption and underweight were higher among frailty. However, non-regular meat consumption seems to increase the odds of frailty only among those who were edentulous. These results highlight the importance of the oral health and nutritional assessment in older adults to avoid, delay, and/or reverse frailty. Therefore, primary care strategies targeting food and nutrition education should be considered to promote healthier food consumption and to keep adequate weight across the life span, as well as expanding oral health teams.

collaborations
MF Lima-Costa designed research; NTM Ygnatios, JL Torres, and MF Lima-Costa conducted research, analyzed data, wrote the paper, read and approved the final manuscript.