infância à adolescência Tracking of dietary patterns from childhood to adolescence

MÉTODOS: Foram realizadas buscas nas bases de dados MEDLINE/PubMed, Lilacs e SciELO utilizando os seguintes termos: “tracking”, “dietary patterns” e “childhood/adolescence” e sinônimos. Foram encontrados 45 resumos e, após aplicação dos critérios de inclusão, 13 artigos foram incluídos. A continuidade do padrão foi avaliada por três principais análises estatísticas: coefi cientes de correlação (Pearson ou Spearman), coefi ciente kappa e análise de probabilidades.


INTRODUCTION
A demographic, epidemiological and nutritional transition is occurring worldwide. 1Changes in global patterns of food sources, forms of processing and distribution have led to the predominance of highly processed foods and beverages. 14The consequences of changes in the frequency of intake and in food and beverage preparation are the general imbalance in energy intake and the increase in the prevalence of obesity.These changes fi rst appeared in highincome countries and now affect billions of individuals. 14rdiovascular diseases, largely associated with overweight, are one of the main causes of death in the adult Brazilian population. 23Nearly 1/5 of Brazilian adolescents are overweight, according to the 2008 Household Budget Survey.a An adequate diet and regular physical activity are proven protective factors against overweight and chronic diseases. 23Maintaining healthy eating habits since childhood and throughout life is one of the requirements for a healthy life.Studies show low tracking of dietary patterns throughout life.In epidemiology, the term "tracking" is defi ned as the stability of a given variable during a certain period of time. 6,21Diet tracking represents the maintenance of eating habits, nutrient intake or food intake throughout time.The continuity of these habits can be understood as positive or negative behavior, depending on the eating habits.
Studies can identify a negative variation in patterns when there is a change from a pattern considered to be healthy to another with a lower nutritional quality. 12he literature includes a wide variety of methods used to assess tracking. 4,7,11,13e percentage of individuals who maintain their positive or negative dietary pattern throughout childhood (zero to nine years) and adolescence (ten to 19 years) varies according to the population group.
Eating habits, whether healthy or unhealthy, acquired and established during adolescence, are very likely to continue throughout adult life. 9,10However, between childhood and adolescence, the tracking of dietary habits or patterns has not been clearly identifi ed yet.This assessment is important, considering the fact that healthy food intake promotion measures should be prioritized from the fi rst years of childhood if dietary patterns are in fact stable from childhood to adolescence, so that healthy eating habits can be acquired and maintained throughout one's life cycle.
The present study aimed to review the scientifi c literature on the tracking of dietary patterns from childhood to adolescence.

METHODS
A scientific literature review of the MEDLINE/ PubMed, Lilacs and SciELO databases was performed.In addition, the bibliographical references of the articles found were searched.
The following were adopted as inclusion criteria: a longitudinal design with an analysis of tracking from primary data; a period of follow-up longer than one year; and date of publication until July 2010.Studies were included regardless of an analysis adjusted for confounding factors being performed.
The following were adopted as exclusion criteria: ill individuals; individuals aged 20 years or more; assessment of time trends in food intake without an analysis of the continuity/stability of dietary patterns; different observational cohort study designs; review articles, theses and dissertations; and relevant methodological problems, such as an excessive number of losses/refusals during follow-up without an analysis being presented.
A search was performed from a syntax of the following title and abstract words (search limits): "tracking, change, continuing, stability" to find articles that assessed continuity; "nutrient intake, dietary patterns, dietary intake, diet" to search for the dietary patterns; and "childhood, infants, adolescence, adolescent" to identify the age group.In addition to words used in groups, other synonyms were included.The syntax corresponded to the search for articles that had at least one word from each of the groups in their title or abstract.These terms were used alone and in a combined form in the Lilacs and SciELO databases, as these do not have the option to input syntax.

RESULTS
A total of 4,441 titles were identifi ed.Of these, 4,396 were excluded when the titles were read, based on the inclusion and exclusion criteria.Subsequently, 45 abstracts were read and 15 articles were selected.Among these, three were excluded because of the high number of losses in the study and because of the lack of an analysis, and one was included from the bibliographical references as it dealt with the term "food group intake", thus totaling 13 articles (Figure ).
A total of three main types of statistical analyses were identifi ed in the assessment of tracking: correlation coeffi cients (Pearson or Spearman), kappa coeffi cient and probability analysis (Table 1).
There were six studies that assessed tracking during childhood, of which five performed a correlation analysis; one, the kappa coeffi cient agreement analysis; two, the paired t-test; and four, probability analyses, aiming to assess the continuity of the dietary pattern identifi ed in the beginning of the follow-up.One article was mentioned more than once as more than one type of analysis was performed (Table 2).-Percentage of agreement between quartiles of intake: chi-square goodness-of-fi t (considers the value observed and that randomly expected) Energy, protein, cholesterol, vitamins D and E, phosphorus, magnesium, iron and zinc showed the highest level of tracking.
Among the food groups, fruits and vegetables showed the highest level of tracking A total of fi ve articles that performed the correlation analysis were consistent and showed coeffi cients equal to or higher than 0.40, regardless of the dietary pattern assessed and the period between assessments.
One study used the kappa coeffi cient and found three dietary patterns, obtained with principal component analysis (PCA), known as "healthy", "traditional" and "processed".Paired t-test was also used to assess the difference in means of dietary pattern scores between periods.In addition, another study used the t-test and found statistically signifi cant differences between mean intakes between the fi rst and third years, except for polyunsaturated fat.The remaining tests were used to estimate the proportion of children who remained in the same percentiles of intake during the periods analyzed.
The studies that assessed tracking during childhood found a moderate level of tracking in this stage and reasonable stability of eating habits in the cohort of children.All of them showed correlation coeffi cients indicative of the presence of continuity of dietary patterns and the majority showed percentages higher than 50% of children who remained in the same percentiles of intake during a certain period.
A total of three studies assessed the continuity of dietary patterns during adolescence and one of them assessed tracking with only one test (kappa coeffi cient).Other two studies used correlation analysis and probability analysis and/or kappa coeffi cient.
The results obtained from correlation showed coeffi cients between 0.3 and 0.6 (fair to moderate) in both studies that performed this type of assessment (Table 3).
A total of two studies assessed the continuity of food intake using kappa coefficient and one study also performed a correlation analysis, obtaining similar results in both assessments; these two tests revealed poor tracking.Among those that performed probability analyses, the results of the fi rst study in Table 3 showed an absence of tracking during six years of follow-up.However, the study conducted by Li & Wang 7 found that approximately 40% of adolescents remained in the same quartile of intake of the nutrients assessed.
A total of four studies assessed tracking between childhood and adolescence (Table 4).There were three studies using the correlation test that showed similar results.Authors considered the coeffi cient values to be from weak to moderate (0.18 ≤ r ≤ 0.68) during different assessment periods: three, six and 15 years of follow-up.
Finally, there were two studies that used the kappa coeffi cient to assess the presence of tracking of nutrient intake and found coeffi cients associated with weak tracking (k ≤ 0.38).

DISCUSSION
The results found point to the presence of weak to moderate tracking during the three periods.However, such results derived from different statistical approaches.
With regard to the statistical analyses used to assess tracking (Pearson and/or Spearman correlation analysis, kappa coeffi cient and probability analysis), relevant points must be considered.In the assessment performed using correlation, the intake is analyzed in terms of continuous numerical values in grams or proportion of calories between both assessments.Perfect tracking would be represented by a correlation coeffi cient equal to one.Thus, the amount of food or nutrient consumed in the fi rst assessment should be the same or change proportionately in the same direction for all individuals.Studies that used this type of analysis adopted the following classifi cation: absence of tracking (r < 0.2), weak tracking (0.2 < r ≤ 0.4) and moderate tracking (r > 0.4).
The kappa coeffi cient assesses the agreement of food intake (usually divided into percentiles) among two or more follow-up periods.The kappa coefficient value excludes the agreement that would be randomly expected.This method was used by fi ve 7,13,15,17,22 out of the 13 studies of this review and it showed weak tracking, regardless of the period assessed.
The probability analysis assesses the continuity of food intake in a position or grade after a certain period.In this assessment, individuals placed in the highest or lowest percentiles in the initial assessment are taken into consideration.
The results found during childhood point to moderate tracking, suggesting a continuity of intake throughout this age group.This is because children's diet is closely monitored and defi ned by the family, especially the purchase and preparation of foods. 16,18e results obtained during childhood were consistent, considering the magnitude of coeffi cients and the probability of individuals remaining in the same percentile of intake throughout time.Studies showed the presence of continuity of intake between two and six years of follow-up.In the analysis with correlation, coeffi cients varied from 0.40 to 0.60 (moderate continuity).
The same consistency is not observed during adolescence.One of the studies 4 did not show continuity of intake of the nutrients assessed.Another study, conducted in Northern Ireland, 17 showed weak to moderate tracking with kappa coeffi cient and a third study 7 evidenced tracking with the analyses performed (correlation, kappa and probabilities), although its assessments had been conducted during a period of one year.Thus, the results pointed to weak tracking of food intake for periods equal to or lower than three years between assessments.This is probably the period when there is the greatest diversity of foods.The continuity of dietary patterns from childhood to adolescence, assessed with correlation analysis in three studies, showed weak to moderate tracking.For this type of analysis, the period between assessments was not a determinant, as the coeffi cients found in the three studies were similar, even among very different periods such as two, fi ve and 14 years between assessments.
Based on the kappa coeffi cient, two studies showed weak tracking and only one of them, with a different analysis, revealed a discontinuity of dietary patterns.
The results found are consistent with the study conducted by Wang et al (2003), 22 which emphasized higher percentages of continuity among those who were in the extreme percentiles.As an example, participants who were in the highest tertile of intake of a pattern of fruits and vegetables had a higher probability of remaining on this level of intake than others who were initially in intermediate percentiles.
There were differences in sample size.Studies included between 200 and 400 individuals on average, three had samples with less than 100 children/adolescents and one large English cohort assessed nearly 6,200 children.However, researchers considered that the representativeness of cohorts was maintained even with smaller samples, because studies that did not perform an assessment of losses (when these are excessive) were not included.The present study did not intend to describe or compare dietary patterns per se, nor to diagnose nutritional problems, but rather to analyze whether individuals characterized by a certain dietary pattern, regardless of their being positive or negative, continued to show this pattern in subsequent years.
The follow-up period varied among studies and probably infl uenced the results.The highest level of tracking appeared among the outcomes assessed in the shortest period.
Relevant differences were observed in the form of analysis of the dietary consumption.Studies that assessed aspects ranging from the intake of specifi c micronutrients to the assessment of dietary patterns produced by factorial analysis (based on food groups) were included.In all cases, the assessment of continuity or stability of intake was performed with the same methodology between different points in time.
The defi nition of the majority of outcomes derived from the assessment of the dietary intake by quantitative methods and food surveys such as daily records and 24-hour recalls were the ones most frequently used.The food frequency questionnaire was used in two studies that assessed food intake by factorial analysis, identifying dietary patterns and working with scores (percentiles) of intake of their populations.
The stability of intake of micronutrients was found in few studies. 3,4,8,17,19,20,24Thus, most of the evidence of tracking in the literature is based on the intake of macronutrients (carbohydrates, proteins and fats) of the diet.
Eating habits during childhood are considered to be stable, as are those from childhood to adolescence.However, there seems to be greater variability of dietary patterns during adolescence.This should be viewed with caution, considering the reduced number of studies assessed.The fi ndings suggest that dietary patterns during childhood continue until adolescence, although they may be changed or discontinued during adolescence.
Eating habits developed during childhood are important, because the introduction of new foods occurs in this initial stage of life, which can determine the dietary patterns in subsequent years.The development of the sense of taste in children has a genetic component and an environmental/social component, which may have an early infl uence on the dietary pattern that they will repeat throughout the years. 2e establishment of healthy eating habits must be promoted early in life.Eating habits can substantially change during growth, but the imprint and relevance of early learning and certain types of social behavior acquired remain throughout the life cycle.

Figure .
Figure.Flow chart showing the results of the bibliographical review.

Table 1 .
Summary of the main methodological aspects of articles included from the systematic review.

Table 2 .
Tracking of dietary patterns during childhood, stratifi ed according to type of analysis.

Table 3 .
Tracking of dietary patterns during adolescence, stratifi ed according to type of analysis.

Table 4 .
Tracking of dietary patterns from childhood to adolescence, stratifi ed according to type of analysis.