Factors associated with exclusive breastfeeding in the first six months of life in Brazil: a systematic review

ABSTRACT OBJECTIVE To identify factors associated with exclusive breastfeeding in the first six months of life in Brazil. METHODS Systematic review of epidemiological studies conducted in Brazil with exclusive breastfeeding as outcome. Medline and LILACS databases were used. After the selection of articles, a hierarchical theoretical model was proposed according to the proximity of the variable to the outcome. RESULTS Of the 67 articles identified, we selected 20 cross-sectional studies and seven cohort studies, conducted between 1998 and 2010, comprising 77,866 children. We identified 36 factors associated with exclusive breastfeeding, being more often associated the distal factors: place of residence, maternal age and education, and the proximal factors: maternal labor, age of the child, use of a pacifier, and financing of primary health care. CONCLUSIONS The theoretical model developed may contribute to future research, and factors associated with exclusive breastfeeding may subsidize public policies on health and nutrition.

Breastfeeding is a crucial issue for public health, for it directly affects the standards of health and mortality of populations. 4,14,20,39,53a,b Part of this trend can be attributed to national policies of breastfeeding promotion, protection, and support. 35fferent social and cultural contexts may influence the practice of exclusive breastfeeding and its determinants.A study conducted in cities of three countries noted that higher levels of maternal education were related both with higher prevalence of exclusive breastfeeding in Santos, SP, Southeastern Brazil, and with lower prevalence in Mexico City, Mexico, and in Sula and Tegucigalpa, Honduras. 32eking greater population homogeneity, this review was restricted to the Brazilian context, since the determinants of exclusive breastfeeding may behave differently in diverse cultures.
The aim of this study was to identify the factors associated with exclusive breastfeeding in the first six months of life in Brazil.

Publications of epidemiological studies conducted in
Brazil about factors associated with exclusive breastfeeding were analized.Bibliographical research was carried out on Medline (via PubMed) and LILACS databases.Delimitation by period or by language were not considered.A manual search of the references included in the bibliography of each article was carried out.
Observational analytical epidemiological studies were included, in which exclusive breastfeeding was treated as outcome, with adjustment of the factors studied together and for possible confounding factors, which have adopted the definition of the World Health Organization (WHO) c for exclusive breastfeeding (children receive only human milk, directly from their mother or extracted, and no longer receive any other liquid or solid, except vitamin drops or syrups, vitamin supplements or medication) and whose sampling procedure has generated a representative population of infants from maternities, cities, states, or of the Brazilian nation.
In articles which more than one age group was evaluated (more than a statistical model to evaluate two or more different age groups), the age group of older age was chosen (with a limit of six months), since the objective was to evaluate the outcome closer to the WHO c recommendation of exclusive breastfeeding until six months of age.
The following studies were excluded: those with results subject to selection bias (such as losses exceeding 20.0%) or with possible information bias (such as interviews with mothers of children over one year old); that presented only the p value (without presenting the association measures), which considered only the population born with low weight; and bibliographic reviews (systematic or unsystematic).
In the case of studies using the same database and published in more than one article (different journals and years), we included those that used different age groups or different analytical variables and methods.
The selected articles were stored under the Portable Document Format (PDF) into a directory shared in the cloud, separated according to databases of origin (Medline and LILACS), and classified into different folders between included and excluded.The evaluation of the methodological quality of the selected studies was obtained by adjusting the scale "Effective Public Health Practice Project: Quality Assessment Tool for Quantitative Studies -QATQS" (http://www.ephpp.ca/tools.html).Of this scale, five questions were evaluated (classified as "strong", "moderate", or "weak"): 1) selection bias; 2) study design; 3) confounding factors; 4) data collection methods; and 5) type of analysis employed for the outcome.Blinding issues of QATQS were not used in any study (since no clinical trial was included), and issues of loss of follow-up were not applied in sectional studies.In the "study design" question, sectional studies had lower score than the cohort studies, because in sectional studies the temporality between exposition variables and the outcome may not always be established.
Considering the final score of the QATQS scale of each article selected, the articles were considered strong if none of the questions were rated as weak; moderate, in the case of studies that showed one of the questions rated as weak; and weak, studies with one or more questions thus evaluated.
Data extraction was independently performed by two reviewers by a structured form, in which were recorded: the last name of the first author; publishing journal and year; location(s) of performance; year and period of performance; study design; population of the study; sample design used; strategy for selection of subjects of research; inclusion and exclusion criteria; total sample number; sample number evaluated; total losses and reason of losses; age group of the studied children; type of outcome; type of statistical analysis; factors of control or adjusting the statistical model; results of the model with association measure and statistical significance; exclusive breastfeeding prevalence or median; limitations of the study; and observations.In the event of disagreement among peers, a third reviewer was consulted.
Data tabulation included: reference of the article (with the last name of the first author, journal, and year of publication); location of the study and data collection; sample number evaluated (and data source); outcome of the study (exclusive breastfeeding or its interruption); statistical analysis employed; prevalence (or median) of exclusive breastfeeding found and the age group of this prevalence (expressed in months); factors associated with exclusive breastfeeding in a statistically significant manner (obtained from the results of the statistical models), as well as its association measure and other factors evaluated without statistically significant association with exclusive breastfeeding.Two tables were made, one for cross-sectional studies and another for cohort studies.As the revised studies measured the prevalence or the duration of exclusive breastfeeding in different age groups, the summary tables of this outcome contain this information.The next step consisted in individually analyzing the association found between the factors investigated and exclusive breastfeeding, highlighting and quantifying the following aspects: in how many studies these factors were investigated, in how many studies an association with exclusive breastfeeding was identified in statistical models and what is its direction.
The last step of the study consisted in creating a hierarchical theoretical model (using the assumptions established by Víctora et al) 49 organizing all factors found according to the proximity to the outcome.The selection of the allocation levels of variables followed the logic of chronological classification between factors present before pregnancy, during pregnancy, immediate postpartum and at the time of discharge until six months of life.
Four levels of variables were proposed, grouped in hierarchical blocks: 1) distal characteristics (contextual, domestic, household, and maternal); 2) distal intermediate (from pregnancy and prenatal care); 3) proximal intermediate (childbirth care, maternal characteristics during hospitalization, and characteristics of the newborn); 4) proximal (characteristics of the nursing mothers and the family, of the babies, and of health services).
To provide parsimony to the summary of variables identified and also for the creation of the theoretical model, the terminology used in each article for each variable was standardized.
Most of the studies were conducted in towns and covered 77,866 children.Regarding Brazilian regions, 14 studies were carried out in the southeastern, six in the south, five in the Northeast, one in the Midwest, and one in the Northern region.The systematic review included studies conducted between 1998 and 2010 (Tables 1 and 2).
The punctual prevalence of exclusive breastfeeding at six months ranged from 3.9% in Bauru, 30 SP, Southeastern Brazil, to 8.5% in Pernambuco, Northeastern Brazil, both in 2006. 6Regarding the WHO indicator, the prevalence of exclusive breastfeeding in children younger than six months of life, resulting from the survey of this population, ranged from 0% in 10 cities in the state of Sao Paulo, Southeastern Brazil, in 1998 46 to 58.1% in the city of Rio de Janeiro, Southeastern Brazil, in 2007 31 (Tables 1 and 2).
Factors associated with exclusive breastfeeding were organized into hierarchical levels (Table 3), being the following the most often exploited (more than a fifth of the 27 studies): place of residence, skin color, maternal age and education, parity, marital status, number of prenatal visits, birth in Baby-Friendly Hospital (BFH), childbirth type, birth weight, sex of the newborn, maternal work, age of the child, financing of the primary health care of the child unit, and the use of a pacifier.
The factors most frequently associated with exclusive breastfeeding (factors investigated in at least six studies and which have showed association in at least one-third of the studies in which they were investigated) were (according to the category positively associated with the outcome): place of residence (residence in the capital, in the metropolitan area, or in rural areas), intermediate maternal age, maternal education, lack of maternal work, age of the child (descending), the nonuse of a pacifier, and financing of primary health care (private) (Tables 1, 2, and 3).3).
Of the selected studies, eight used hierarchial theoretical model to identify factors associated with exclusive breastfeeding before starting the statistical modeling, and only one considered contextual variables 47 (Tables 1 and 2).
Based on the factors listed in the analysis of the 27 selected studies, was created a hierarchial theoretical model of the factors associated with exclusive breastfeeding.Some were constituted of the group of similar factors, such as "difficulties in breastfeeding", which grouped the following variables: nipple fissure, pre-set schedule to breastfeed, and difficulties in latching or positioning.Similarly, were grouped as "emotional indicators" the variables: maternal self-worth and psychological distress (Figure 2).

DISCUSSION
A systematic review of Brazilian epidemiologic studies showed a relevant production of studies from the late 1990s having exclusive breastfeeding as outcome, which were conducted mainly in the Southeast region of Brazil.Most of the studies selected for this review showed moderate quality, and only a quarter of the articles had a longitudinal design.Therefore, the evidence of the factors associated with exclusive breastfeeding in children under six months of age in Brazil found in this review can be considered as moderate.
The number of variables listed in epidemiological studies, and used to explain the duration of exclusive breastfeeding, was high, and the discussion of the findings of this systematic review was marked by the organization of the variables in hierarchical levels.
Because of the diversity of backgrounds and factors investigated, the use of summary measures of association derived from meta-analysis techniques was considered invalid.
Among distal factors, the place of residence was the contextual variable investigated the most, and the results were discordant, and sometimes the urban environment, 29 sometimes the rural 11 were associated with exclusive breastfeeding.Most of distal factors seem to represent maternal socioeconomic factors.Maternal education was the factor most widely investigated, with almost half of the studies having observed an association between maternal education and exclusive breastfeeding, and the findings were unanimous: the low education level was associated with the interruption of exclusive breastfeeding.In epidemiological  studies, the socioeconomic gradient is reproduced, in general, in a health gradient. 22The national research on breastfeeding also report these differences, in which mothers with higher education exclusively breastfeed for more time.d The variable "skin color or race", in turn, may represent customs, social norms and traditions, 27 besides income 8 and social relations.e Considering the national surveys, white mothers breastfed exclusively for more time, b but only one study found an association between skin color and higher prevalence of exclusive breastfeeding. 31ternal age and parity may represent the experience with breastfeeding. 31,51All studies that investigated them observed an association between greater parity and exclusive breastfeeding. 16,18,23,44,46,47Regarding maternal age, the intermediate ages seem to be protective for exclusive breastfeeding, because both teenage mothers 5,29,40,46,47 and those with 35 years of age or more 6,10,16 interrupt it prematurely.
Considering the distal intermediate factors concerning pregnancy, the number of prenatal visits was the variable most frequently investigated.The three studies that found an association between this variable and the outcome indicated the low number of prenatal visits as a risk factor for exclusive breastfeeding.Santo et al 40 and Vieira et al 52 consider that low compliance to prenatal care may represent women who are less careful with their health; on the other hand, Demétrio et al 11 consider that this low compliance may reflect low access to sources of information on breastfeeding.
The public or private service pervades all hierarchical levels evaluated: prenatal care (distal intermediate level), childbirth (proximal intermediate), and childcare (proximal).The private primary care was associated with the outcome in three of 10 studies. 44,47,52his variable may represent both the access to health services and the maternal socioeconomic status, for the access to health services may be determined by more distal variables such as skin color, sex, education, and income. 36ong the proximal intermediate factors, the birth weight was the one most widely used, finding a positive association between children with adequate birth weight and exclusive breastfeeding in three of the 21 studies that investigated it. 9,44,47This may be explained because children with low birth weight are more likely to spend more time hospitalized in neonatal unit, thus spending more time separated from their mothers. 41In addition, these children may have more difficulties in initiating or maintaining breastfeeding, since both the frequency and the pressure of the suction increase as the gestational age and weight of newborn increase. 25stational age, in turn, was an indicator hardly used in the studies, for differences or biases in the rankings of this variable that may occur. 42Although no studies have found an association between this variable and the outcome, it is suggested to maintain it in the studies.
Another factor widely used in the studies was the childbirth type, however, only two studies have found an association between the vaginal childbirth and higher prevalence of exclusive breastfeeding. 1,2The vaginal childbirth contributes to the timely initiation of breastfeeding, 3 being possible to assume that it can also provide its maintenance in exclusive mode.Another hypothesis would be the possible relationship between socioeconomic characteristics and access to public health services, 17 since both the vaginal childbirth 13 and timely initiation of breastfeeding are more practiced at these services. 3ong the proximal intermediate factors studied, those that assess the guidelines received in hospital 23,52 (positive association with the outcome), the difficulties to breastfeed during hospitalization 7 (negative association), and exclusive breastfeeding on discharge 31 (positive association) may be the most adequate to evaluate peri-partum related aspects which may determine the duration of exclusive breastfeeding.
The variable "sex of the baby" was used in 14 studies, considering that two 6,47 found a positive association between female and one between male 5 and exclusive breastfeeding.The prevalence of breastfeeding among girls was higher in the capitals of the entire Brazil; b however, it is unclear whether this increased prevalence is due to some cultural aspect such as the belief that boys need greater nutritional intake by other foods in addition to breast milk. 32,47garding the proximal factors considered, the use of a pacifier was the factor most strongly associated with the interruption of exclusive breastfeeding. 2,7,9,16,21,24,28,30,43,44,51,52The use of a pacifier may lead to the reduction in breastfeeding frequency, interfering in breast demand, and possibly changing the baby's oral dynamics. 50A Brazilian study concluded that, in addition to the causal relationship between the use of a pacifier and breastfeeding interruption be unclear (it is unknown whether the use of a pacifier is a marker of the interruption of breastfeeding, or if it is a cause of the same), the process of using pacifiers is dynamic, with children starting or stopping the use of a pacifier throughout the period. 50In a randomized study conducted in Canada, 19 the authors observed that the use of a pacifier can be a marker of the interruption of breastfeeding or of low motivation to breastfeed rather than be the cause of the interruption of breastfeeding.
Maternal work was a variable widely used in studies, 1,2,6,7,[9][10][11]16,21,24,28,30,31,34,38,[43][44][45][46][47]52 and in the six studies that found statistically significant association, it showed a negative association with the outcome. However, this variable must be ivestigated considering if the mother is or is not on maternity leave.51 Mothers that work outside the home with maternity leave would have better conditions to maintain exclusive breastfeeding during the maternity leave period.
Most of the studies based on surveys did not consider the child's age, but the probability to be exclusively breastfed decreases as the age of the child increases.All the studies that used this variable found an association between the descending age (or early age) of the child and exclusive breastfeeding. 1,10,18,21,28,29,31,44ong all the variables considered proximal, those that evaluate the access to information or guidance on breastfeeding that women receive in primary health care services could be those more directly associated with exclusive breastfeeding.However, only Pereira et al 31 used this variable, noting that guidelines in group and on positioning and latching of the baby were associated with a higher prevalence of exclusive breastfeeding.
Evaluating statistical modeling strategies, less than a quarter of the studies included in this review adopted a theoretical model prior to analysis, organizing the variables in hierarchical levels. 6,7,18,24,31,43,52Although expendable, to create this conceptual model is important because it requires prior knowledge about the social and biological factors associated with outcome, assisting to establish an order of logic input variables in the model based on hierarchy of factors and not considering only purely statistical criteria. 49Thus, a hierarchical theoretical model was proposed, including factors identified in  the studies of this systematic review, which can assist in planning data collection and statistical modeling strategy of epidemiological studies related to exclusive breastfeeding.
Public policies to promote, protect, and support breastfeeding adopted in Brazil since the 1980s have contributed to the increase in the median duration of breastfeeding and its exclusive mode across the Country. 35hese policies, however, cannot be considered as an individual attribute: having a childbirth in Baby-Friendly Hospital or maternity with Human Milk Bank may depend on the context in which the woman lives as well as her access to these services.
In addition, the local contexts within each city (districts, neighborhoods, surroundings) may vary: in the city of Rio de Janeiro, e.g., a great variation in the adoption of the 10 steps to successful breastfeeding (recommended by the Breastfeeding-Friendly Primary Care Unit Initiative) are found between the units of primary health network. 37In addition, disparities between public and private health units must be considered, such as those observed in the adoption of breastfeeding in the first hour of life in hospitals (recommended by the Baby-Friendly Hospital Initiative). 3nsidering this possible context effect, it is plausible that nursing mothers who are residing in the same regions or municipalities (including districts, neighborhoods, or census units) share social and economic factors (contextual factors) that influence the duration of exclusive breastfeeding, e.g., standards and attitudes toward breastfeeding; the organization and access to primary health services in their neighborhood; and the level of action and policies for promoting, protecting, and supporting breastfeeding.In fact, variables such as socioeconomic indexes and the number of pro-breastfeeding actions existent in certain regions have already been used for the evaluation of factors associated with exclusive breastfeeding. 47ny factors used in analytical epidemiological studies and its directionality in association with exclusive breastfeeding were identified and described, noting the frequency with which they are used and the heterogeneity of categories and cutoff points.Instead of defining the effect of each of the factors identified in the systematic review by meta-analysis, we decided to discuss them according to a hierarchical theoretical model.Some recommendations concerning the findings of this study include the completion of further studies in the North and Midwest of the Country, as well as the encouragement to academic work on little explored factors in association with exclusive breastfeeding.
The use of a conceptual theoretical model prior to statistical analysis, preferring the hierarchical organization of variables in relation to the proximity to the outcome, may help the choice of variables to be included in the studies and to evaluate the intermediation of more proximal variable blocks in relation to the more distal ones.
It is suggested that future studies consider context variables to investigate the association with exclusive breastfeeding, since inclusion of contextual variables with concomitant multilevel models is a useful strategy for the adequacy of these models. 12Studies covering the triangulation of qualitative and quantitative methods 26 to the understanding of the relation of some factors with exclusive breastfeeding could also contribute to a better understanding of the subject.
The main limitation of this systematic review was the selection bias, since abstracts published in conference proceedings were not included, which is called the "grey literature". 33Another limitation is the possibility that relevant studies have not been found by the search strategy used.The possible subjectivity of the authors in the evaluation and selection of articles was minimized by the independent search of the literature, by the standardized form-filling, and by assessing the quality of the articles selected for the review.
In conclusion, the study of determinants of exclusive breastfeeding is of vital importance for public health, and epidemiological studies have an important role for the understanding of this theme in Brazil.However, the emergence of new and more sophisticated statistical tools, as well as the growing complexity of explanatory models and the context effects of the factors associated with exclusive breastfeeding, brings a new challenge to scholars of the topic: the careful use of these resources and the dissemination of the results in a clear and purposeful way, directed to the development and improvement of public policies for promoting, protecting, and supporting breastfeeding which are reflected in the health and well-being of the population.

Figure 1 .
Figure 1.Descriptive flowchart of steps of systematic review in Medline and LILACS databases.

Figure 2 .
Figure 2. Hierarchical theoretical model of factors associated with exclusive breastfeeding.

Table 1 .
Sectional studies on factors associated with exclusive breastfeeding.

Table 2 .
Included sectional studies on factors associated with exclusive breastfeeding.

Table 3 .
Factors investigated regarding the association with exclusive breastfeeding, organized by hierarchical level, frequency of use, and number of times they were associated with exclusive breastfeeding in a statistically significant way.
Ecological or contextual factors Actions in exclusive breastfeeding in the municipality or region Size of the population or socioeconomic indexes of population compoundsPlace of residence (urban or rural; slum or not; countryside or capital)