Breastfeeding: what changed after a decade?

ABSTRACT Objective: to analyze the changes in prevalence, median duration and correlates of breastfeeding in a small city in São Paulo state, Brazil. Method: analysis of two cross-sectional studies, conducted at intervals of one decade, with 261 and 302 children younger than two years, respectively. We used Kaplan-Meier survival analysis for calculation of the median duration of breastfeeding, and Cox regression for correlates analysis, with significance level of 5%. Results: an increase of 33.4% in the prevalence of exclusive breastfeeding and 20.9% in breastfeeding was identified. Regarding the latter, the median duration increased from 7.2 to 12 months. In the most recent study, the median duration was lower in first-born children who used pacifiers, and it was not associated with breastfeeding incentive actions. Conclusions: advances in the prevalence and duration of breastfeeding were observed during the 10 year-period, however, pacifier use still remains associated to a shorter median duration of breastfeeding. Our findings contribute to highlighting the need for intensification of nursing actions in the promotion of breastfeeding, and discouragement regarding the use of pacifiers.

A study developed in the beginning of the 2000s, in a small city in the State of São Paulo, Brazil, showed a prevalence of breastfeeding in 41% of infants under 24 months, and exclusive breastfeeding in 13% of those under six months, with a median duration of 7.2 months of breastfeeding and only 28 days of exclusive breastfeeding. The correlates of lower median breastfeeding were: lactation, first order of birth, and pacifier use (12) . In view of the implementation of public policies to promote, protect, and support breastfeeding and the expansion of primary health care in the last decade in the country, as well as the conduct of a new study on health and nutrition conditions in the same city, we concluded it would be pertinent and timely to analyze the changes in prevalence, median duration, and correlates of breastfeeding between 2001 and 2013, in this small city of São Paulo state, which was the objective in this study.
The study was conducted based on the hypothesis that the prevalence and median duration of breastfeeding increased during the period under review; the correlates of the median duration of breastfeeding did not change; and, finally, the actions to promote breastfeeding

Methods
This is a cross-sectional study, conducted during 2013, in a small city in the State of São Paulo, Brazil, that compares data to another methodologically similar study conducted in in the same city, in 2001 (12) .
In 2001, the city population was about 26 thousand inhabitants, and the local health care system consisted of a health center, seven Primary Health Care Facilities (PHCF), a Psychosocial Care Center (PCC) and a small size municipal hospital (50 beds). In 2013, the population increased to 48 thousand inhabitants, and the health system is now composed of 12 BHUs, three outpatient clinics, and has maintained the PCC and the municipal hospital (13) .
The first population-based study was conducted with a representative sample of 261 children under two years of age (12) . The second study, performed in 2013, was part of a larger project approved by the Research Ethics Committee (Process nº 193.468), which evaluated the health and nutrition of children less than three years of age enrolled in the PHCFs, in a representative sample proportional to the number of children enrolled in every PHCF. The sample calculation, performed with the Epiinfo 6.04 software, indicated a sample of 350 children, with 95% confidence level and a 5% margin of error.
Toriyama ATM, Fujimori E, Palombo CNT, Duarte LS, Borges ALV, Chofakian CBN. Inclusion criteria were: child enrolled in one of the 12 PHCFs, and attendance at the health service with the mother during the period of data collection. Among the 399 mothers invited, 35 refused, one did not meet the inclusion criteria, and five were excluded (three twins, one adopted, and one with neurological disease).
Of the 358 children from zero to three years old who composed the sample, 302 children under two years of age (84.4%) were analyzed in this study, and compared with the previous study. other studies on breastfeeding, as it gives the advantage of analyzing both information from weaned and still breastfed children at the time of the interview (15)(16) . For the child who still received breast milk at the time of the interview (censored child), the time of breastfeeding referred to the current age at the interview. To evaluate the difference between breastfeeding medians in the univariate analysis, the log-rank test was used.
The identification of correlates of duration of breastfeeding was performed using multiple Cox analysis with duration of breastfeeding were considered for model fit: marital status and maternal level of education (8,17) .
Both models were developed by inserting the variables simultaneously. In the Cox regression model, the measure of association in Hazard Ratio is similar to relative risk, and indicates the likelihood of a subject who did not have the event, having it at that time. The significance level of 5% was considered statistically significant in the final model. The Schoenfeld test was used to verify the adequacy of the Cox models, and all variables met the risk proportionality assumptions. Table 1 shows the changes in the profile of the sample studied in 2013 (n=302), in relation to the population of the 2001 study (n=261) (12) . A statistically higher percentage of families with per capita income greater than or equal to 1.8 minimum wages was found, along with a reduction in the proportion of mothers with three or more years of education, and an increase in working mothers (p<0.001).

Results
There was a 20% increase in guidance on breastfeeding in prenatal and childbirth periods (p<0.001), and a reduction in the pacifier use (p=0.05).   In the univariate analysis, the variables statistically associated with the median duration of breastfeeding, in 2013, were: marital status, birth order, and pacifier use ( Table 3). The median duration of breastfeeding was lower in children whose mothers has no partners, first born infants, and those who used pacifiers. In the analysis of the Toriyama ATM, Fujimori E, Palombo CNT, Duarte LS, Borges ALV, Chofakian CBN.   (Table 4).

Discussion
Our findings corroborate the findings in national surveys (18)(19) and trend analysis of the breastfeeding performed in the last decade (7) , which shows an increase in the prevalence and median duration of breastfeeding, and also confirms the hypothesis of this investigation.  (19) . Even so, this prevalence is still considered poor by the WHO, which classifies as good only a prevalence that reaches a rate of 50% (6) .
On the other hand, the practice of breastfeeding not only increased from 2001 to 2013, but also was prolonged; a more expressive increase of breastfeeding was found among children from six to nine months and from 12 to 15 months, the last age group recommended by WHO (14) as an indicator of breastfeeding continuity.
Despite the increase, the percentages found are still slightly lower than the prevalence of 87.6% of breastfeeding in children under six months, and of 47.5% in the range of 12 to 15 months, verified in Brazil in 2006 (18) . good, from 21 to 22 months; and, finally, very good, from 23 to 24 months (6) . Even so, breastfeeding and exclusive breastfeeding rates both in Brazil and in this small city are better when compared to countries such as China and the United States (11) .

The 4.8 months increase observed between 2001
and 2013 was very similar to the rate of increase of 4.6 months observed in the 1996 to 2006 decade (4) .
The median duration of breastfeeding found in the city also remained similar to that observed in Brazil in 2008 (11.3 months), although slightly above the median of www.eerp.usp.br/rlae Rev. Latino-Am. Enfermagem 2017;25:e2941. the Southeast Region (10.1 months), and in the capital of the State of São Paulo (9.8 months) (7) .
In addition to the implementation of public policies to promote, protect and support breastfeeding, and the expansion of basic health care, the population profile may have influenced the breastfeeding indicators.
The improvement in family income, an increase in the proportion of mothers who received counseling on breastfeeding during prenatal and childbirth care, and reduction in pacifier use may have contributed positively, as they have been shown to be associated with increased prevalence and duration of breastfeeding (8,10,17) , which highlights the relevance of nursing care in primary health care (the gateway to the health system), with the prioritization of maternal and child health care through appointments, educational groups, and home visits.
On the other hand, the increased proportion of mothers who were in the labor market and the decreased proportion of women with three years or more of education from 2001 to 2013 may have interfered negatively on the breastfeeding patterns observed in this study. Since there is evidence that mothers with better schooling present longer duration of breastfeeding (8) and lower maternal schooling, as well as integration into work, are determinant for interruption of exclusive breastfeeding (17) , breastfeeding indicators could have been better if maternal profile was more favorable.
In 2013, the same correlates observed in 2001 (12) remained associated with the median duration of breastfeeding, confirming the pre-established hypothesis.
Thus, the median duration of breastfeeding was lower among first-born children and among those who used pacifiers.
Birth order has been studied in terms of women's parity and previous experience in breastfeeding.
Primiparous women breastfeed for less time, either because of insecurity, younger age, less education, less knowledge about the benefits of breastfeeding, or less willingness to cope with social and cultural difficulties. However, it is necessary to consider that the intention of pregnant woman to breastfeed is another factor strongly related to the duration of the breastfeeding. In addition, each birth occurs in different family contexts, so the influence of this variable is difficult to analyze (20) . The approach to breastfeeding in the prenatal period, considering the situation of primiparous women, can reduce anxiety and even increase the number of pregnant women with the intention to breastfeed. On the other hand, the experience of breastfeeding from mothers of children born second or later in birth order must be valued, and can be considered as a protective factor for breastfeeding.
There is evidence that pacifier use is associated with maternal work outside the home, primiparity, and lack of breastfeeding within the first hour of birth (21) . Classically, pacifiers have been associated with lower frequency and duration of breastfeeding, especially exclusive breastfeeding (22) . Pacifier use has been discouraged since the beginning of the 2000s, however, it continues to be highly prevalent and influences negatively the maintenance of breastfeeding in our study, probably because it is a culturally accepted practice in Brazil, and also because the mechanisms involved in the relationship with weaning have not yet been fully elucidated (16,23) .
It is important to emphasize that the effectiveness of breastfeeding actions depends on the health professionals considering the complexity of the determinants of breastfeeding, and the life situation of the mothers (24) . Thus