Infant feeding and adiposity: scientific challenges in life-course epidemiology.

The article by Durmus x et al (1) published in this issue of the Journal is yet another important contribution from the Generation R Study in Rotterdam and shows the value of birth cohort studies in life-course epidemiology. The focus of this article is the relation between infant feeding (the duration and exclusivity of breastfeeding and the age at introduction of solid foods) and adiposity at 6 y of age. The major value of the study is its use of dual-energy X-ray absorptiometry and abdominal ultrasound to estimate percentage body fat and its distribution, ie, abdominal compared with general body fat. Because a great deal of information was collected during pregnancy and infancy, many important potentially confounding factors were measured and controlled for statistically. These features, along with a large sample size (n 1⁄4 5063), improve the validity and precision of estimates of the associations between infant feeding and later general and abdominal adiposity, which are risk factors for later type 2 diabetes, hypertension, and coronary heart disease in adulthood. The results show that crude associations in the expected direction (lesser adiposity with more prolonged breastfeeding and later introduction of solids) attenuate and become nonsignificant after adjustment for key confounding variables. Along with these strengths come several minor weaknesses. Although 30% of the cohort provided no information on breastfeeding, it seems unlikely that those losses led to an important selection bias. Another potential problem, which we discuss in greater detail below, relates to the authors’ adjustment for the child’s height at the time of the adiposity measurements. First, however, we discuss why confounding by factors that influence both infant feeding and later adiposity is so likely to bias studies of the effects of the former on the latter.