Childhood overweight and obesity, which are strongly associated with adult overweight and increased risk of non-communicable disease [1], are prevalent and increasing in lower- and middle-income countries in Latin America including Peru [2, 3]. This increase is attributed to unhealthy diets and physical inactivity [4–6] and current prevention strategies focus on these drivers [4], but early rapid postnatal weight gain is also associated with risk of overweight and obesity [7, 8].

The evidence that early rapid weight gain is a factor in childhood and adult obesity is robust [7–9] but most reported findings are from high-income countries, where child undernutrition is uncommon, or from longitudinal cohorts in low and middle income countries (LMIC) that included children born before the recent increase in obesogenic changes in lifestyle associated with the nutrition transition, and/or concurrent behavioural risk factors were not reported [10].

Breastfeeding has been thought to protect against obesity [11–13] but the one randomized trial of breastfeeding did not show any effect of increased breastfeeding on obesity [14, 15] and another study [16] reported that a longer duration of breastfeeding was not associated with reduced adult adiposity. This last study included populations from LMIC countries with high rates and long duration of breastfeeding. Thus, the possible effect of breastfeeding remains unclear [17].

In addition there are few, if any, longitudinal studies that include rural populations whose lifestyle and growth trajectories are distinct from urban areas but that still experience adult obesity, as in Peru [18].

In Peru, 98.6% of babies are breastfed, with a median duration of 21 months and a median of 4.4 months exclusive breastfeeding [19]. This study presents results from a cohort study of children almost all of whom were breastfed; contrasts rural and urban households; and includes weight, height and waist circumference measurements, as well as concurrent obesogenic risk factors such as physical activity and diet [5].

We hypothesize that rapid weight gain from birth to 1 year (referred to as infant weight gain) and from 1 to 5 years (referred to as early childhood weight gain) would predict overweight and obesity (henceforth referred to as overweight), obesity, BMI for age, and waist circumference at 8 years even when concurrent obesogenic practices are allowed for, but that growth trajectories vary between rural and urban households.

Young Lives (YL) is a longitudinal study of children growing up in poverty in four countries [20]. This analysis is confined to 2052 children from a representative sample including all but the 5% richest districts of Peru, due to the study’s focus on poverty. Further details of the sampling strategy have been described [21]. Children aged 6 to 18 months (1 year) were enrolled and household visits conducted between June and December 2001; (Round 2, 2006) when the children were aged 4.5–5.5 years (5 years); and (Round 3, 2009) when the children were aged 7.5–8.5 years (8 years). There was considerable household migration, but families were tracked and 1963 (95.7%) of children were interviewed at 5 years and 1942 (94.6%) at 8 years.

The study collected data on many aspects of the household and the index child; instruments are available online [22]. Relevant to this analysis, information was collected when children were 1 year on breastfeeding (duration and whether the child was currently breastfed) and educational level of parents. Birth weight was taken from the child’s health card (66.7%), or mother’s report. At 1 year children were weighed nude or in light clothing on a calibrated digital balance with 100 g precision (Soehnle: Backnang, Germany) and length was measured using locally-made rigid stadiometers with 1 mm precision by 6 trained supervisors who had passed a standardization exercise. At ages 5 and 8 years all fieldworkers were trained and standardized in anthropometry. Standing height without shoes was measured using locally made instruments accurate to 1 mm and weight was measured in light clothing using digital platform balances as above. No deduction was made for clothing. At 8 years waist circumference was measured to the nearest 1 mm using non-extensible tape placed against the skin midway between the lower rib and superior iliac crest with the tape horizontal to the ground [23]. Data were collected on migration and household location, socioeconomic indicators, the child’s diet including consumption of high calorie purchased food (fast food), hours of sleep and daily activities, and maternal anthropometry.

Information was checked for completion and inconsistencies in the field, and data were entered using range and consistency checks. YL data for rounds 1–3 can be accessed on request online [22]. This analysis is based on data from 1521 children with complete data on birth weight, anthropometry and selected co-variables at all three survey rounds. Of the 1942 children measured in all 3 rounds, 421 were excluded: 224 without birth weight; 14 with a disability that interfered with growth measurement; 47 children were not measured either in round 2 or round 3; 106 mothers were pregnant; and 30 had missing data on covariables.

These analyses show that rapid weight gain in infancy and in early childhood in both urban and rural children of both sexes in this predominately poor, breastfed population in Peru is associated with overweight and obesity at age 8 years even when other determinants of childhood obesity are taken into account. The association is seen with both weight gain in infancy and in early childhood and is evident in urban and rural children although, while in rural children the greatest effect (OR 2.6) was seen in the adjusted model for rapid weight gain in infancy, in urban children the greatest effect (OR 2.4) was for rapid weight gain in early childhood.

The inference is supported by the association with waist circumference, in urban children but the association is weaker in rural children and becomes non-significant in fully adjusted models at both ages of rapid weight gain. In the case of the rural children concurrent behaviours have a larger impact on waist circumference in the adjusted models. This may be the result of different growth trajectories of rural and urban children or may be due to the lack of discrimination in our indicators of concurrent behaviours in urban children, for instance more than 70% of children drank soda drinks and ate “fast food” at least once a week. In both cases greater height is associated with greater risk of overweight and obesity so it is unlikely that the discrepancy between rural and urban is due to overestimation of overweight due to the fact that the calculation of BMI includes height squared in the denominator.

These findings support evidence that the increasing problem of overweight and obesity in childhood is not only mediated through concurrent behavioural determinants of overweight such as physical activity and diet but that factors such as birthweight, the perinatal hormonal milieu, complimentary feeding practices, and the gut microbiota present in early years including infancy and early childhood also play a significant role [28].

These findings are in keeping with other studies using more accurate techniques such as deuterium [29] and ultrasound measurement of visceral and subcutaneous fat compartments [30], that showed an association of rapid weight gain in early childhood (1–4 years) (although not infancy) with fat mass in adolescence.

These findings contribute to the body of evidence, mainly from developed countries, of the association between early rapid weight gain and overweight/obesity in childhood, adolescence and adulthood [8, 11, 31, 32]. In keeping with other studies these analyses showed a sex difference, but in this case the effect of rapid weight gain was not limited to girls and occurred during infancy, as well as early childhood. This study extends the observations to a recently studied LMIC population with high prevalence and long duration of breastfeeding. Although there is no detailed information on total duration of breastfeeding, 85.8% were still breastfed at R1 (aged 6–18 m). Furthermore, this study also includes indicators of concurrent risk factors for overweight including diet, sleeping patterns and physical activity that were not available for other analyses of rapid weight gain in LMICs [10]. It has also been possible to contrast rural and urban children and show similar results in relation to overweight and obesity measured by BMI, although there are differences with respect to waist circumference. The longitudinal nature of this study will provide further data on longer-term weight trajectories.

There are some weaknesses with our study. Since Young Lives is primarily a study of the effects of poverty it was necessary to include the geographical and social diversity of Peru. A birth cohort was not feasible for this design because of cost and logistics, and thus children were enrolled aged 6–17.9 months and birth weight was based on reported data. Inaccuracy in the birth weights might have weakened the associations and may have added uncertainty to the importance of intrauterine growth restriction (IUGR) in our population. The data on hours of sleep, consumption of soda drinks, fast food consumption and physical exercise is based on self-report using a limited number of questions rather than direct observation or detailed dietary recall. This limits our ability to attribute the observed weight gain to dietary or other factors.

The findings are important as the 8 year old children in our study had a prevalence of overweight 28.1% and 8.6% obesity, consistent with combined prevalence of overweight and obesity in school-age children from Brazil (33.5%) and Mexico (29.3%) [2]. This shows that Peru, along with other countries in South America is in the midst of an increasing problem of overweight and obesity, despite also having a burden of under nutrition, with stunting affecting 32.3% rural and 10.3% urban children under 5 years [19]. Evidence showing that early rapid weight gain in infancy and early childhood can lead to later childhood overweight and obesity, even when almost all babies are breastfed, means that programs and policies need to evolve to ensure that improvements in the quality of the diet improve linear growth without adding additional calories that may induce excessive weight gain.

In keeping with other studies, we have shown that in a breastfed population from Peru rapid weight gain in infancy and in early childhood in both urban and rural children is associated with overweight and obesity at age 8 years.