Association between breast feeding and food consumption according to the degree of processing in Brazil: a cohort study

Study design and sample

This is a cohort study based on data from birth cohorts that are part of the RPS Birth Cohort Consortium (Ribeirão Preto, Pelotas and São Luís). This consortium is a research network, involving three research groups from three municipalities in different regions of Brazil, which has monitored nine birth cohorts over decades.16

The study includes data from four of these birth cohorts from the RPS Consortium, namely: those born in 1982, 1993 and 2004 in Pelotas and in 2010, in São Luís. Pelotas is located in southernmost Brazil and, according to the last census, carried out in 2010, the municipality had 328 275 inhabitants and a Human Development Index (HDI) of 0.739.17 São Luís is the capital of the state of Maranhão, northeast region and, according to data from the last census, in 2010, its population was 1 014 837 inhabitants and its HDI was 0.768.18

In the Pelotas birth cohorts, all live births to mothers living in the urban area of Pelotas, delivered in public and private maternity hospitals, were eligible for the perinatal study. In the São Luís cohort (2010), one in three live births to mothers living in the city in 2010, delivered in public and private maternity hospitals with 100 or more births per year, were eligible for the first phase of the study. Details on the number of participants included in each cohort and monitored at each follow-up are available elsewhere.1 16 19 20

In this study, participant data collected at birth and at the following follow-ups were used: Pelotas cohort (1982)—follow-ups at 24 and 48 months, 22 and 30 years; Pelotas cohort (1993)—follow-ups at 6, 12 and 48 months, 18 and 22 years; Pelotas cohort (2004)—follow-ups at 12, 24, 48 months and 11 years; and São Luís cohort (2010)—follow-up at 12 and 13–36 months. All individuals with available data on breast feeding in childhood and food consumption in the segments mentioned below were included in this study. Thus, this study evaluated 4027 participants from the Pelotas cohort (1982) aged 22 and 3410 aged 30; 1165 participants from the Pelotas cohort (1993) aged 18 and 1042 aged 22; 3338 participants from the Pelotas cohort (2004) and 1093 participants from the São Luís cohort (2010).

Exposure variable

The study’s exposure variable was the duration of breast feeding, which was reported by the mothers. Definitions of the WHO were used to categorise the type and duration of breast feeding. EBF was considered as the time during which the child received only breast milk, straight from the breast or milked, or human milk from another source, without the addition of other liquids or solids, with the exception of vitamin drops or syrups, oral rehydration salts, mineral supplements or medicines. Predominant breast feeding (PBF) was considered as the time during which the child received, in addition to breast milk, water or water-based drinks (sweetened water, teas, infusions), fruit juices and ritual fluids. Total breast feeding was considered the total time the child received breast milk (directly from the breast or milked), regardless of whether or not they received other foods.21

To construct the total breastfeeding variables data collected at 24 and 48 months in the 1982 cohort, at 6, 12 and 48 months in the 1993, and at 12, 24 and 48 months in the 2004 cohort were used. To create the variables relating to EBF, in turn, data collected during follow-up at 24 months in the 1982 cohort, at 12 months in the 2004 cohort, and at 6 and 12 months in the 1993 cohort were used. In both cohorts, we used the earliest available information to minimise recall bias. In the São Luís cohort, data to construct total breastfeeding variables were obtained in the follow-ups at 13–36 months, and to create EBF variables data from follow-up at 12 months were used.

In the 1982 Pelotas cohort, in the follow-ups at 24 and 48 months, mothers were interviewed and answered the following question: ‘For how long did the child breastfeed?’ Based on the response to this question in months, variables for total breast feeding were constructed. In the follow-up at 24 months, mothers were also asked about when the child started (in months) consuming other types of food besides breast milk (cow’s milk, powdered milk, mashed fruit, vegetable puree, family food) and whether the child had consumed any food other than breast milk (cow’s milk, powdered milk, mashed fruit, vegetable puree, family food, yoghurt, porridge, bread or biscuits, others) on the previous day. From the answers to these questions, the age (in months) at which the child started ingesting foods and other milks besides breast milk was determined, considering the previous month as the duration of EBF.

In the 1993 Pelotas cohort, in the follow-ups at 6, 12 and 48 months, mothers were asked: ‘Does the child breastfeed?’ In case of a negative response, they were further questioned: ‘Until what age did the child breastfeed?’. Based on the response to either or both of these questions, variables for total breast feeding in months were constructed. In the follow-up at 6 months, mothers were asked if they had started offering any items from a list of liquids and foods (boxed milk, powdered milk, tea, juice, fruit puree, savoury puree, porridge, yoghurt, bread/biscuits, egg yolk, egg white, meat, bean broth, bean grain, rice, pasta, others), and if so, when they started offering them. In the follow-ups at 6 and 12 months, mothers were also asked to report the last day the child ate as usual, and whether the child had consumed the listed foods (cow’s milk, powdered milk, coffee, water/tea, juice, bread, biscuits, yoghurt, fruits, egg, rice, beans, vegetables, pasta, potatoes/cassava, meat, others) at any time during the day (on waking up, in the morning, at lunch, in the afternoon, at dinner, before sleeping, overnight). From the responses about dietary intake, the age (in months) at which the child started ingesting foods and other milks besides breast milk was determined, considering the previous month as the duration of EBF.

In the 2004 Pelotas cohort, the collection of information regarding child feeding, as well as the construction of variables for total and EBF, was conducted in the same manner as in the Pelotas cohort of 1993. However, breastfeeding data were obtained during the follow-ups at 12, 24 and 48 months, while data on the introduction of complementary foods were collected during the follow-up at 12 months.

In the 2010 São Luís cohort, in the follow-up at 13–36 months, mothers were asked: ‘Did the child breastfeed yesterday?’. In case of a negative response, they were further questioned: ‘Until what age did the child breastfeed?’. If the mother’s answer was ‘yes’ to the first question, they were further questioned: ‘Until what age did the child breastfeed exclusively?’. Based on the response to these questions, variables for exclusive and total breast feeding in months were constructed.

In the analyses, the breastfeeding duration was assessed numerically and categorically, operationalised as follows: EBF duration (months), EBF up to 4 months (yes, no), EBF up to 6 months (yes, no), total breastfeeding duration (months) and total breastfeeding duration (<12 months, ≥12 months). For the Pelotas cohorts of 1982 and 1993, however, PBF was assessed since EBF was rarely practised at the time of these studies.

Outcome variables

The outcome variables of the study were the consumption of fresh or minimally processed foods (FMPF) and UPF, evaluated as a percentage of total daily energy intake (%kcal).

Food consumption data from participants in the Pelotas cohorts were obtained through semiquantitative Food Frequency Questionnaires (FFQs), referring to food consumption in the last 12 months. In the 1982 Pelotas cohort, at 22 and 30 years of age, the FFQs contained 85 and 88 food items, respectively, and were administered by trained interviewers. These FFQs were derived from an instrument previously submitted to validation22 and modified only to meet regional food habits. In the 1993 Pelotas cohort, at age 18, the FFQ consisted of 81 food items and was administered by a trained interviewer, while at age 22, the FFQ had 94 items and was self-administered. In the 2004 Pelotas cohort, at the 11-year follow-up, the participant’s mother answered a self-administered FFQ consisting of 88 food items. Data on food consumption of participants in the 2010 São Luís cohort were obtained through a 24-hour recall administered by a trained interviewer.

The intake of each food item was converted into grams per day. Afterwards, daily intakes of macronutrients and energy (kcal) were estimated using food composition tables.23 24

Foods were grouped according to the NOVA classification into FMPF, processed and UPF.25 Finally, the proportion of energy (%kcal) of each food group in relation to the total energy intake of the diet was calculated.

Data analysis

Statistical analysis was conducted using the software Stata, V.14.0 (StataCorp). The numerical variables were described using measures of central tendency and dispersion and the categorical variables were described using relative and absolute frequencies.

The Student’s t-test was used to compare the differences in the mean consumption of FMPF and UPF between the categories of breastfeeding duration variables, considering a significance level of 5%.

Crude and adjusted linear regression analysis was used to evaluate the association between the breastfeeding duration variables and FMPF and UPF consumption.

To control confounding, the independent variables included in the adjusted model were identified by building a directed acyclic graph (DAG) (online supplemental figure 1) in the software Daggity, V.3.0. Using the backdoor criterion,26 the need for minimum adjustment was identified for the following variables: maternal education at birth (>12, 9–11, 8–5, <5 years of schooling), family income at birth (first—lowest income, second, third tertile), pregestational body mass index (kg/m2) and gestational age (weeks of gestation).

Supplemental material

Supplemental material

In the 1993 cohort, the samples at 6, 12 and 48 months were subsamples composed of all low birthweight infants and an additional 20% of the cohort. Because of this weighting was carried out by low birth weight.

This post was originally published on https://bmjopen.bmj.com