Design and study population

This is a cross-sectional assessment of the Young Lives Cohort Study, a longitudinal study conducted in four developing countries: Ethiopia, India, Peru and Vietnam¹⁸. This study started in 2002 with two base groups: infants from 6 to 18 months (younger cohort) and children from 7 to 8 years (older cohort) who were chosen at random. At present, the follow-ups have been carried out for a period of almost 15 years, 2006-2007 (second round), 2009-2010 (third round), 2012-2013 (fourth round) and 2016-2017 (fifth round) for both cohorts.

Selection of participants

The original study was undertaken in three stages. In the first stage, the country was split into equal geographical regions, based on the national poverty map of Peru developed in 2000 by the National Compensation and Social Development Fund (FONCODES in Spanish)¹⁹. Aimed at over-representing the poorest areas, the 5% of the wealthiest districts (i.e. those with the highest socioeconomic status) were excluded and a systematic sample of the remaining districts was carried out. In the second stage, 20 sentinel sites were chosen and an area within each sentinel site was selected using table of random numbers. Once the census areas were selected, a random sample of blocks was selected. In the third stage, the children who were within the selection parameters (6 to 18 months) and those who were in the chosen blocks were selected. The objective was to survey 100 children for each sentinel site.

For the present manuscript, Peruvian children from the younger cohort were included, using data from the third round (2009-2010) of the Young Lives Study. The information was taken from the household survey, available at the official webpage of Young Lives. The household questionnaire was completed by the head of the family and includes questions about household composition, characteristics of the child caregiver, socioeconomic status, child health (including eating habits), anthropometrics, etc. Only those with complete information on the variables of interest (socioeconomic level, consumption of fast food and soft drinks) were analysed.

Sample size

Regarding the sample size, estimations were conducted assuming a significance level of 5% and a power of 80%, a total of 948 individuals were needed to find a difference of 5% in the proportion of children consuming fast food in the highest and lowest quintile of socioeconomic status, based on results of a previous study²⁰. Since 2052 children were enrolled at the study baseline, and 92 were lost to follow up until the third assessment, 1940 were available for analyses. However, of them, only 1901 had complete data with the variables of interest and were included in the manuscript. Nevertheless, this sample size was enough to find the association of interest.

Definition of variables

Two outcome variables were used. The first one was the consumption of fast food, defined based on the question “During the past 30 days, how many times have the child eaten salty and fatty foods?” Peruvian examples were used to clarify this question including well-known products such as fries, burgers, pizza, grilled chicken, etc. On the other hand, the second outcome was the consumption of soft drinks, evaluated through the question “During the past 30 days, have the child taken soft drinks?” Similarly, some examples of sodas were used to clarify this question. Both questions had six response options: daily, two to three times a week, once a week, every two weeks, less than every two weeks, and never. For analyses purposes, these options were categorised into three groups: always (including the options “daily” and “two to three times a day”), sometimes (“once a week”, “every 2 weeks” and “less than every fortnight”) and never.

The exposure variable was the household socioeconomic level measured by using a wealth index, a variable constructed by the Young Lives study based on three measures: quality of housing, access to services, and assets possession at home. These measures were obtained from the questions asked to the family head using the household survey and carried out face to face. The indicator of the quality of housing was obtained through the simple average of the quality of the material of the floor, ceiling and walls, and also included household density. Access to basic services was obtained through the simple average of possession of electricity, water, sanitation and cooking fuel; while, for the calculation of the assets index, the simple average of the list of possession of domestic articles was estimated. Finally, the calculation of the wealth index was the result of the simple average of the three indexes, obtaining a value between 0 and 1, where a number closer to one indicates a higher socioeconomic status. For analyses purposes, this variable was divided into quintiles, with the first quintile being the poorest and the fifth quintile the wealthiest.

Other variables were taken into account to evaluate the association of interest including: sex (male vs. female), age (7 and 8 years), geographical area (urban vs. rural), mother’s education level (<7 years, 7-11 years, and ≥12 years), father’s educational level (<7 years, 7-11 years, and ≥12 years), physical activity of the child in the last 7 days, based on the mother self-report of physical activity of at least one hour per day (<7 days a week and 7 days a week), number of meals in the last 24 hours (5 times a day, <5 times a day, and >5 times a day), presence of excess of weight according to the child’s body mass index (yes or no). Body mass index (BMI) in the child was calculated based on the Z-score taken from the World Health Organization (WHO) reference tables. Finally, the BMI of the mother (normal, overweight and obese) was also used, obtained from the weight and height data and the use of international standards.

Procedures

All the interviewers were trained to comply with the protocol of collection, data entry, data cleaning, and storage of the information. Each team collecting data in each sentinel site was comprised by a researcher, a fieldworker and a psychologist¹⁷. For the anthropometric data collection, an anthropometrist was trained according to the techniques of the WHO. For the evaluation of height, a stadiometer was used, and standing height information was collected; whereas calibrated digital scales were utilised for weight (in kg). Data was entry and processed using Microsoft Access.

Statistical analysis

Data was analysed using STATA 13.0 (Stata Corp, College Station, TX, US). Initially, the description of quantitative variables was conducted using means and standard deviations (SD) and, in the case of categorical variables, proportions. Then, a descriptive analysis of the characteristics of both child and parents according to the household socioeconomic level, and subsequently, according to fast food and soft drinks consumption was conducted. Chi square test was used to evaluate these comparisons. Finally, crude and adjusted models were created to assess the association between household socioeconomic level and fast food and soft drinks, using Poisson regression models with robust variance, reporting prevalence ratios (PR) and 95% confidence intervals (95% CI). In these latter models, the first (lower) quintile of socioeconomic level was used as the reference group as our hypothesis expected to have an increasing consumption of unhealthy food with higher socioeconomic status. Finally, the variance inflation factor was used to avoid collinearity in the final model.

Ethics

The initial study was approved by the Social Sciences Ethics Division of the University of Oxford ¹⁸. In addition, the study was approved by each country where it was developed. Thus, in Peru, the original study was approved by the ethics committee of Instituto de Investigación Nutricional. This research protocol and analysis were approved by the Ethics Committee of the Universidad Peruana de Ciencias Aplicadas - UPC, in Lima, Peru.

Characteristics of the study population

From 2052 subjects originally enrolled at the baseline of the study, 112 (5.5%) children were lost to follow-up and 39 (1.9%) children did not have complete data. Thus, data of 1901 children, of which 942 (49.6%) were girls, with a mean age of 7.5 (SD: 0.5) were further analyzed.

The description of participants according to the household socioeconomic level is shown in Table 1. Of note, as household socioeconomic level increases, mother’s and father’s education level also do so. On the contrary, at higher household socioeconomic level, children tend to perform less physical activity and have a higher prevalence of excess of weight.

Fast food and soft drink consumption

Overall, 24.1% of children reported having always consumed fast food, while only 20.1% reported never having to do so. The prevalence of fast food consumption increases with higher household socioeconomic level, reaching a maximum in the fourth quintile and decreasing after that (Table 2). In addition, children who always consumed fast food were concentrated in the urban area, had less physical activity levels, ate more than 5 meals a day, had overweight mothers, and had greater excess of weight.

In the case of soft drinks, 22.3% of children reported having always consumed this kind of beverages, while only 13.0% reported never having to do so. The prevalence of soft drink consumption increases with higher household socioeconomic level, and similar to fast food, reached a maximum in the fourth quintile and decreasing after that (Table 2). Children who always consumed soft drinks were concentrated in the urban area, ate more than 5 meals a day, and had overweight mothers.

Association between socioeconomic status and consumption of fast food and soft drinks

In multivariable model and compared with the first socioeconomic quintile, those in the fourth (p=0.001) and fifth (p=0.01) quintile were more probable to always consume fast food after controlling for different confounders (Table 3). In addition, an increase in the strength of association was reached up to the fourth socioeconomic quintile and then decreased after that. The same pattern was found in those who reported eating fast food sometimes, though results were not significant.

On the other hand, compared to the first socioeconomic quintile, those in the third (p=0.001), fourth (p<0.001) and fifth (p<0.001) quintile were more probable to always consume soft drinks after controlling for different confounders (Table 3). Similarly, an increase in the strength of association was evident up to the fourth quintile and then reduced in the highest quintile. The same pattern was found in those who reported consuming soft drinks sometimes, although they were only significant in the fourth and fifth quintiles.

Main findings

Our cross-sectional analysis shows a positive relationship between household socioeconomic level and the consumption of soft drinks, a finding that is almost similar to that of fast food. In addition, the strength of the association increases up to the fourth socioeconomic quintile, and after that, drops in the fifth socioeconomic quintile. Finally, about a quarter of the children evaluated reported always consuming fast food and soft drinks.

Comparison with previous studies

Up to our knowledge, no previous studies in Peru have comprehensively reported the association between household socioeconomic status and the consumption of fast food and soft drinks. Moreover, no studies were found that have studied the prevalence of fast food and soft drinks consumption in our context. A longitudinal study using data of the younger cohort of the Young Lives reported that daily consumption of soft drinks was 22%, 30%, and 48% for the low, medium and high socioeconomic level, respectively²¹. For this latter report, the authors used information from the third and fourth round of assessments of the cohort, but the models were not controlled for potential confounders. On the other hand, other studies assessing the relationship between socioeconomic status and the consumption of soft drinks and fast food were very heterogeneous. A study conducted during the economic transition in South Korea showed a positive association between the consumption of soft drinks and fast food²²; thus, as the population showed a higher wealth index, the consumption of soft drinks and fast food also increased. The increase in consumption was consistent up to the last quartile of wealth, a result very similar to our study. This finding is, however, contrasted with subsequent studies in the same country, in which as the local economy improved, the eating behavior of the adolescent population also improved, measured as higher consumption of fruits and vegetables and a decrease in the consumption of fast food and soft drinks^23-24.

Regarding the consumption of soft drinks, a study conducted in Colombia aimed at finding socioeconomic differences associated with the consumption of soft drinks among children and adolescents²⁵. This report found that children coming from homes of medium and high socioeconomic status had more probability of consuming soft beverages compared to those who came from lower socioeconomic levels. Of note, our findings as well as other from scientific literature, reflects how the trend of consumption of fast foods and soft drinks varies according to the socioeconomic gradient in each particular country. Thus, as the income of the population increased, also increases the consumption of fast food and soft drinks in developing countries; but, on the contrary, in developed countries, the poorest present a greater consumption of this type of products. For example, in Canada, children who belonged to a lower socioeconomic level were more likely to consume soft drinks compared to those who were from the higher socioeconomic status²⁶.

Relevance in public health

As the economy of developing countries improves, individuals will adopt unhealthy behaviours belonging to developed countries, including their lifestyles and eating habits, many of which lead to problems of excess of weight^27-28. This study reports that children from higher socioeconomic status tend to consume more fast food and soft drinks. Fast foods and soft drinks may have deleterious effect due to the additional caloric intake provided, which could trigger an excess of weight in children²⁹. This process has been previously described by Popkin^10,16 as nutritional transition. As countries are in different stage of this nutritional transition, findings of the association between socioeconomic position, including but not limited to income, socioeconomic status, educational level, assets, etc., and unhealthy diet may vary. This is the main reason why this association needs further scrutiny at the country and regional level. Thus, our results suggest the need to implement interventions aimed at reducing the consumption of fast food and soft drinks. Interventions may have a greater impact if they adopt education strategies^30-33, as well as government policies such as the control of the advertising of these products³⁴, increase in the tax on soft drinks and fast food^35-36 in this way regulate their consumption and address the problem of obesity and other comorbidities.

Strengths and limitations

To our knowledge, this is one of the first studies conducted in Peru to assess a potential association between household socioeconomic level and the consumption of fast food and soft drinks. However, this study has also limitations. First, the study design, a cross-sectional study in nature, cannot determine causality but the only association between the variables of interest. Second, some information bias may arise as questions were responded by mothers instead of children. On the other hand, only one question was used to collect information regarding fast food and soft drinks consumption; as a result information bias may arise as a concern. Similarly, recall bias may be present since the questions were based on the last 30 days before the interview, and moreover, 30 days may not be enough time to detect all the unhealthy food consumption as this can change according to seasons of the year. Finally, the instrument evaluated the consumption of soft drinks and fast food as frequency and not quantity (volume in the case of drinks).