Consensus among experts on healthy eating and diet quality index Consenso entre especialistas sobre alimentação saudável e índice de qualidade da dieta

Resumo O objetivo do artigo é elaborar um consenso sobre Alimentação Saudável e Índice de Qualidade da Dieta (IQD) para subsidiar estudo de validação do IQD. Os especialistas foram identificados entre autores de artigos publicados entre 2010 e 2015 com os unitermos healthy eating index e diet quality index. A consulta foi feita por meio da técnica Delphi e do método Likert. Para definição do consenso se considerou o atendimento a pelo menos três critérios: pontuação mínima em cada asserção (≥ 3,00); desvio padrão (≤ 1,5); frequência das respostas concordantes (≥ 51%) e diferença de intervalo interquartil (≤ 1,0). Os temas Alimentos altamente palatáveis, Oleaginosas e Carnes e ovos não alcançaram o consenso na primeira rodada. Os especialistas propuseram novos temas: Glúten, Fracionamento das refeições, Consumo de álcool e Inserção de nutrientes no IQD. Embora os marcadores de qualidade e de risco na alimentação sejam estudados periodicamente, só foi possível estabelecer consenso sobre temas como frutas, hortaliças, leite e derivados, leguminosas e oleaginosas como marcadores de qualidade após fundamentação teórica. Os alimentos processados e prontos para consumo, alimentos altamente palatáveis, doces e gorduras em excesso, e álcool foram prontamente identificados como de risco. Palavras-chave Alimentação saudável, Índice, Índice de qualidade da dieta, Consenso Abstract The article aims to achieve a consensus about Healthy Eating and Diet Quality Index to enable a validation study on the Diet Quality Index. Experts were identified among authors of articles published from 2010 to 2015 that presented the key worlds healthy eating index and diet quality index. The query was carried out by combining the Delphi technique with the Likert method. To determine a consensus, at least three of the following criteria had to be met: minimum score in each statement (≥ 3,00); standard deviation (< 1,5); frequency of agreement (≥ 51%) and differences between interquartile ranges (< 1,0). Topics regarding Highly palatable foods, oilseeds, and Meat and eggs did not arrived at a consensus in the first round. Experts proposed new themes: Gluten, Meal frequency, Alcohol consumption, and Including nutrients in the diet quality index. Although quality and risk markers in diet are periodically studied, it was only possible to reach consensus on subjects such as fruits, vegetables, milk and dairy products, legumes, and oilseeds as quality markers after theoretical justification. Processed and ready-to-eat foods, highly palatable foods, excessive sweets and fats, and alcohol were readily identified as risk factors.

introduction Non-communicable chronic diseases (NCD) are in the fifth place of the major death risks in the world 1 . About 30% of the global population is overweight, and it is possible this number will grow to up to 50% in 2030 2 . In Brazil, 50.1% of the population is overweight and obesity reaches 12.4% of the adults 3 . These data arise from lifestyles that are out of balance with health and eating habits, being one of the main factors collaborating to the increase and prevention of NCDs 1,4 .
Food guides offer orientations to promote healthy changes in the dietary intake of the population 5 . In 2006, the Food Guide for the Brazilian Population (GAPB -Guia Alimentar para a População Brasileira) was published, which classified foods into groups according to the adapted food pyramid of 1999 6 . Its new edition prescinds from such logic and conceptualizes quality according to the foods' processing 5 .
Indexes to assess diet quality have been proposed, in line with Food Guides, to enable preventing NCD and care situations 7 . The Healthy Eating Index (HEI) 8 was built from the recommendations for the American population and influenced, in Brazil, the proposition of the Revised Diet Quality Index (DQI-R) 9 , adjusted for the food groups in the Adapted Food Pyramid 6 .
In 2013, the Diet Quality index in association with the Digital Food Guide (DQI-DFG) was published 10 , based on guidelines of the Public Health School of the University of Harvard 11 and adjusted for the eating habits of the Brazilian population. It was devised in a design that is in line with the mains principles of the current GAPB 5 : it is based on food, not in nutrients, without disregarding the relevance of nutritional quotas; prioritizes fresh or minimally processed foods and indicates the importance of consuming less processed and ultra-processed foods; aliments are classified into groups by similarity of nutritional composition and according to the evidence of implications of their regular intake over health 10 .
Considering the existence of particular characteristics on the Brazilian diet, including the importance of beans as a diet quality marker, this study aims to establish a consensus on Healthy Eating and Diet Quality Index in a national perspective by querying researchers with experience on the subject. The result of the consensus described in this work will base a validation study on the DQI-DFG content 10 .

Methods
Research tool. Topics considered in the construction of the expert query instrument correspond to components and food groups included in the DQI-DFG (Table 1) 10 . The theoretical foundation for devising the statements was gathered from a systematic study on field publications from 2007 to the construction of the first version of DQI-DFG, published in 2009. Conclusions and recommendations of studies on the relationship between diet and disease prevention [12][13][14][15][16] , national and international research on Diet Quality Indexes (DQI) and healthy eating 7,10,16,17 , and official documents of international and national organizations with the same purpose 5,18-21 were embodied into this study.
The instrument comprised two analysis dimensions consisting of 21  The query statements were organized in a Likert-type scale 22 with four possible answers. Numeric values were attributed to the answers, distributed over constant intervals, from full agreement to full disagreement, through the intermediate terms inclined to agree and inclined to disagree. The numerical order could be whether descending (4, 3, 2, 1) or ascending (1, 2, 3, 4) depending on the statement being favorable or unfavorable, thus enabling the calculation of mean values and standard deviation.
Arriving at a consensus. Delphi Technique 23,24 was used to implement the research instrument, and at least three of the four criteria established should be met 25 : (a) average score for each statement and for both dimensions examined, with the following interpretation: from 1.00 to 1.99 points: consensus for disagreement; from 2.00 to 2.99 points: lack of consensus; from 3.00 to 4.00 points: consensus for agreement.
Experts selection. For retrieving research published from April 2010 to April 2015, the database Web of Science was consulted using the search terms healthy eating index and diet quality index, as well as Brazil as research location. Data available at the Lattes Platform were used to identify the authors with academic training in Nutrition and stricto sensu graduate studies.
Pre-test. The instrument was pre-tested by a group of five researchers, to analyze the statements' clarity and phraseology.
Instrument application. For starting the first round, the experts received an email invitation with the study casuistry and a link to access the query instrument, hosted by the QuestionPro tool. After reading an introductory text for understanding the method and agreeing with the Informed Consent Form, the invited experts had access to the instrument statements.
In starting round 2, they received the first results' consolidation. Statements for which consensus was not achieved and the experts' considerations about the open question were organized within the same Likert-type scale, with the same answer options. For both situations, each topic was accompanied by a theoretical foundation. Statements that resulted in consensus in the first round were reintroduced to the experts, so they could ratify or not their previous opinions. At the end of round 2, they received the final result of the consensus. Both the application of the two rounds and consensus emergence occurred between April and May 2015.
Study on the internal consistency of the instrument. To measure the internal consistency of the instrument, Cronbach's alpha coefficient was applied with acceptance criteria R > 0.80 8 .
Statistical analysis. Statistical analysis was conducted from data processing using using Statistical Package for Social Sciences -SPSS software version 18 for Windows.
Ethical aspects. This study was approved by the Research Ethics Committee of the Federal University of São Paulo.

Results and discussion
Construct and theoretical reference. Theoretical foundation and construct used in the instrument design were organized into themes, as shown below: Fresh foods: eating fresh and little processed foods is associated with a lower risk of developing cardiovascular diseases and prevention of most chronic diseases 18 . The application of this knowledge favors a balanced, tasty diet, which promotes a sustainable food system 5 .
Meat and eggs: these are important sources of protein and the recommendation is varying their  26 . Meat is the main source of vitamin B12, an essential nutrient for the functioning of brain and nervous systems and formation of blood cells 19 . However, the intake of red and processed meat is associated with the risk of developing colorectal cancer: 50g of meat/day raises the risk of developing cardiovascular diseases to 30% 21 .
Considering the possible risks associated with the excessive consumption of these foods, quotas must be established for their intake 10 .
Milk and dairy products: these feature a good concentration of nutrients (proteins, vitamin A, and especially calcium), in addition to offering satiety 5 . Special care should be given to sources with higher concentrations of fat (fat cheeses) 11,17 .
Glycemic index: a diet rich in carbohydrates, especially those with high glycemic index, is associated with obesity, type-2 diabetes, metabolic syndrome 17 and cardiovascular diseases 4 . Therefore, the classification of potatoes in the "moderation components" 11,17,26 , in addition to refined cereals 4,16 , is justified. Excessive intake of sugar raises the energy value of the diet, thus favoring weight gain, contributes to the emergence of these diseases and introduces factors related to the dependency of its consumption, increasing its intake 15 .

themes
Assertions -Construct In natura foods 1) Healthy eating is mostly composed by food in natura. a 7) Consuming fruits, vegetable, greenery, cereals, meat, milk and milk derivative, vegetable oil and oilseeds are regularly compatible with healthy eating. a Beef and eggs 2) Beef and processed meat must be classified in distinctive groups to the other meat under the comprehension of a healthy eating. a 15) A healthy eating requires the presence of either meat or eggs. a Milk and dairy products 3) Milk and milk derivative must compose the healthy food choosing repertoire. a Glycemic Index 4) Taking into consideration the food sugar level, potatoes are classified in the same group as bread and refined cereal. a Highly palatable foods 5) Highly palatable food shouldn't be part of daily eating. b 11) Sugar rich food, like soft drinks, chocolates, ice-creams or biscuits, might belong to a healthy eating if their ingestion isn't usual and in small amounts. a 12) The build criterion of a diet quality index must be stringent with highly palatable food, due to the high risk of its ingestion. a Processed fats 6) The industrial mayonnaise production characteristics justify its classification may fit better in processed fat group than in vegetable oils. a 8) The margarine production characteristics justify that its classification may fit better in vegetable oils group than in processed fat group. b Source of fat in food 9) Food rich in fat, when consumed moderately, might be part of eating. a Legumes intake (beans) 10) It's important to consider the ingestion of beans and leguminous at Brazilian eating pattern. a Nuts intake 13) In order to keep a healthy eating, it is necessary to ingest oilseeds on a daily basis. b Purpose of the diet quality index 14) A diet quality index, besides been an analysis tool of eating ingestion, might also be used as an additional instrument to actions of food and nutritional education. a 20) It's suitable to build a diet quality index from food groups. a Moderation components 16) When building a diet quality index, the moderation components are sugars, beef and processed meat, processed or animal fat, refined cereals and potato. a Sugar and sweet taste 17) Sweet beverage, even diet or light, must not compose healthy eating. a Olis 18) The vegetable oils role about healthy justifies its inclusion among the adequacy components. a Vitamins and minerals supplements 19) The use of vitamins supplements and minerals is a necessary strategy to ensure a good nutrition. b Alchool intake 21) The evidence about the alcohol effects in health contraindicate its ingestion. a Numerical values that indicates the answer possibilities: a (4) Completely agree, (3) Agree in most cases, (2) Disagree in most cases, (1) Completely disagree b (1) Completely disagree, (2) Disagree in most cases, (3) Agree in most cases, (4) Completely agree Highly palatable foods: are those with a high content of fat, sugar, and salt, combined or not, poor in fibers and with little or no nutritional contribution 27,28 . Examples include soft drinks, sugary drinks, snacks like "chips" and candy, in addition to sausages and processed meats 27 . These foods can also bring health risks when excessively consumed; hence, with consumption quotas and a healthy eating pattern there is the possibility of eating these foods eventually, to satisfy the palate 17,29 .
Processed fats: mayo and margarine are foods that provide trans fats and sodium, which, when consumed in excess, are associated with the further development of cardiovascular diseases 5 . For this reason, they make up the group of processed fats (moderation component), being recommended their reduced intake.
Sources of fat in the diet: a high intake of foods that are source of fats is associated with the further development of cardiovascular diseases 5 . However, the moderate intake of fat balances the distribution of other macronutrients 30 , in addition to contributing as a source of vitamins, essential fatty acids, and minerals 5 . The safe intake of foods from this group can also offer taste satisfaction without increased risk 17 .
Legumes intake (beans):rice and beans represent about a quarter of the energy consumed in the Brazilian diet 5 . Legumes are excellent sources of proteins, fibers, vitamins, mono-unsaturated fatty acids, and minerals 11,17 .
Oilseeds intake: these are excellent sources of protein, fiber, vitamins, minerals, and mono-unsaturated fatty acids 11,17 . Their intake may help vary the repertoire of proteins' sources to compose a diet with a lower frequency of red and processed meats, which contributes to lowering the risk of heart disease and diabetes 14 .
Purpose of the Diet Quality Index: monitoring and assessment of food intake can concur to healthy choices and strengthening of subject autonomy 31 . Indexes are auxiliary tools to assess diet quality and, thus, to facilitate the analysis of dietary intake 7 . To devise this Index, foods are classified into groups according to their similarity in nutritional composition and the evidence about the implications of their regular intake on health. This criterion was established aiming to emphasize foods and not nutrients 10 .
Sugar and sweet taste: studies on the properties that lead to sugar and sweet addiction show that the excessive intake of these foods decreases the production of ghrelin and interferes with leptin transport and signaling, thus reducing dopamine and stimulating food-related pleasure and increased intake 15 .
Oils: by providing essential fatty acids, oils represent a protective factor regarding cardiovascular diseases. However, their inflammatory potential should be considered for their dosage in food preparation 11,18,19 .
Vitamin and mineral supplements: the proper intake of fresh foods prevents most chronic diseases 18 . Such effect comes from the food itself, not merely from the isolated nutrients 5 . Therefore, an adequate diet meets the subject nutritional needs without the need for supplements.
Alcohol intake: drinking wine is widely recognized as part of the prevention of cardiovascular diseases 20 ; for cancer, however, even wine is an important risk factor 12 . Alcohol is the most used drug worldwide; its misuse accounts for 3.2% of deaths. Another justification for not including alcohol in the Index stems from the fact that alcoholism has become both a social and public health problem in Brazil 13

Study on the internal consistency of the instrument. The instrument presented good internal consistency (R = 0.94).
Consultations with experts. Although the dimensions "Healthy Eating" and "Diet Quality Index" deal with frequent topics of the participants' expertise, and even though the GAPB 5 has just recently re-stimulated the discussion on "Healthy Eating", two rounds were required to achieve a consensus.
Of the 54 experts selected and invited into the study, 30% agreed to participate, being 88% of them affiliated with public institutions, regionally distributed as follows: 6% in the Northeast, 18% in the Midwest, 38% from Southeast and 38% in the South. All the researchers work on the theme of Healthy Eating for at least four years (56% between four and 10 years, 19% from 11 to 20 years, and 25% between 20 and 30 years); 41% of them classified their own experience on the subject as proper qualification and 59% as highly qualified. Regarding the dimension Diet Quality Index, 50% reported having experience between 6 and 10 years, 38% had from 2 to 5 years of experience, 6% had only 1 year and 6% did not respond; 13% identified their condition as highly qualified, 75% considered they had proper qualification and 12% declared to have inappropriate qualification. These data show homogeneity among the research participants regarding domain over the subjects under study. Of the 16 experts who responded to round 1, 10 (63%) continued to round 2. According to Giannarou & Zervas 25 , a sample between 10 and 15 participants with these characteristics can produce sufficient results.
1st round. In a comprehensive analysis, both Dimensions examined reached a consensus with an average of 3.02 points (SD = 0.67) and 3.43 points (SD = 0.31) for Dimensions 1 and 2, respectively. Table 3 shows the results of the analysis used to arrive at a consensus for Dimensions 1 and 2, respectively, at each round. All statements of D2 reached consensus, unlike D1, where three statements did not complete at least three of the four criteria adopted in this study (statements 5, 13, and 15). Assertions that have not reached consensus were reintroduced to the experts in the 2nd round, with the corresponding theoretical foundation.
2nd round. Both dimensions examined reached a consensus with an average of 3.18 points (SD = 0.62) for D1 and 3,27 points (SD = 0.9) for D2. As shown in Table 3, the statement "Highly palatable foods cannot be part of the daily diet" (statement 5) did not achieve a consensus even after the theoretical foundation was presented. Despite the understanding that they lead to health risks when excessively eaten, the possibility to use this foods for eventual satisfaction may have motivated such result. Effects such as the stimulus to addiction behaviors 28 and the displacement of high nutritional density foods, leading to low diet quality 27 , justify the proposition of this statement. However, even the proposal "Foods high in sugar -sodas, chocolates, ice cream or cookies, may be part of a healthy diet if their intake is not frequent and in small amounts" (statement 11) did not achieve consensus. This topic, however, is resumed in the open question, which indicates it caused doubts among the experts. Thus, a theoretical foundation was elaborated to respond the doubts posed, so it was possible to arrive at a consensus on the theme.
Meats, eggs and oilseeds are relevant protein sources and provide mineral micronutrients and vitamin B12 17,26 . This understanding justifies the statement "a healthy diet requires the presence of meats or eggs" (statement 15) 19 . Nevertheless, considering that the consumption of red and processed meats is associated with the risk of developing colorectal cancer 21 , the approach expressed in "a healthy diet requires daily intake of oilseeds" (statement 13) becomes relevant due to its nutritional composition and the possibility of diversifying protein sources in the diet 17 . Thus, consensus was achieved for both statements, which is also represented by the descriptive level of response patterns comparison between first and second rounds.
Open question. Suggestions from experts in response to the open question generated 11 discussion topics. Table 4 presents the results of analyses employed for these topics -all reached consensus in accordance with the criteria established in this study.
Among the topics brought by experts, we considered that "Classification of potato according to glycemic index", "Alcohol intake", and "Purpose of Diet Quality Index" are included in the design of the analysis instrument. Although consensus was reached in the first round for all these topics, when they were reintroduced in the open question a theoretical foundation was elaborated to reinforce the response to these manifestations. For the other themes proposed by the experts, the following arguments were presented (emphasis added): Processed foods (must be considered) as components of the diet quality index Index components that include meats and fats are classified as "moderation" and are built with a subclass for processed foods. Other foods such as cookies, sugary drinks, snacks like chips and candy are also classified as "moderation components", for which the recommended intake is limited to small quantities 4 . For the food group "milk and dairy products", am adjustment to the Index is required to include sugary yogurt and dairy drinks. Such design is in line with the guidelines proposed by the GAPB, which supports the identification of foods according to their degree of processing, especially the message "avoid the consumption of ultra-processed foods".

Sodium and foods rich in trans fats and/or cholesterol (must be considered) as components of the diet quality index
The option for focusing on foods -and not nutrients -guided the organization of the Index into groups divided by nutritional composition similarity according to evidence about impacts of its regular intake on health. Thus, aliments with high sodium concentration are included in the subclass of processed foods. Foods that are source of trans fat and/or cholesterol are covered by groups "red and processed meats", "animal and processed fats", and "milk and dairy products". Trans fats also include formulations of products under the group "sugar and sweets" in aliments such as cookies and ice cream, among others. Thus, the experts' suggestions are satisfied.

Gluten as a relevant topic
Eliminating gluten from the diet is indicated when confirmed the diagnosis of coeliac disease, dermatitis herpetiformis, allergy or sensitivity to gluten. Considering that this Index is meant for the healthy population, a component about gluten was not proposed 32 .
(How to include in the IQD) food and nutritional appeals: fortified, diet, and light The Index elaboration was based on the emphasis to foods and not nutrients, and the groups were generated by nutritional composition similarity and implications of the regular intake on health. Thus, the diversity of modified foods (enriched or with functional claims, diet and light) makes the grouping complex.
High intake of sweet aliments leads to action that create dependency through the reduction of dopamine and consequent pleasure stimulation 15,16 . Therefore, these foods are grouped with sugars and candies, included into moderation components, so that the intake is reduced.

Meal frequency
There is no evidence about any benefit of food intake fractionation to create a guideline on the appropriate number of meals per day 5,14 .

Nutritional density of the foods
The organization of "adequacy components" is based on the nutritional contribution of the foods it comprises. Such components ensure the supply of nutrients to meet nutritional needs, as well as bioactive compounds. They are, therefore, high nutritional density foods 10 .
"Alcohol" is addressed for in the statement 21, being a consensus on both rounds, including in the theoretical foundation presented for the open question. The justification that drinking wine favors the prevention of cardiovascular diseases 20 , as well as the absence of this information in the current GAPB 5 could have hindered the arrival at this consensus. Nonetheless, the consensus on not including alcohol as a component to a healthy diet was achieve due to the fact this drug is a risk factor for cancer 12 and its misuse has become a social and public health problem 13 .

Conclusion
The combined application of Likert method and Delphi technique enables the arrival at a consensus on Healthy Eating, with the recommendation of the following markers of diet quality: "fresh foods", "fruits", "vegetables", "milk and dairy products" and "legumes". "Processed and readyto-eat foods", "excessive sweets and fats", "supplements" and "alcohol" were identified as risk factors. The importance given to legumes and the construct that supports the non-inclusion of alcohol as part of a healthy diet are original components of the DQI-DFG and are among the main contributions of this study, unparalleled for querying Brazilian experts for this purpose. In spite of this fact, analysis on the Index application results in different contexts may point improvements that overcome any study limitations. Among the weaknesses of the Index devising process, the most important are the absence of the indication of weekly consumption components -which can by suitable for bovine and porcine meats, and the lack of cooking techniques as risk markers for diet quality.
Each of the variables of analysis that contribute for the construction of healthy food profile in the form of a Diet Quality Index should express the best understanding possible from evidence of food and nutrition fields, incorporating dimensions associated with the eating ritual and sustainable food systems. Thus, renewing consensus about major issues concerning healthy eating may help improving Diet Quality Indexes.