Horse

Background: Previous observational studies have consistently shown slower weight and length gains in infants with prolonged breastfeeding than in those who were formula-fed from birth or breastfed for a shorter duration. These studies have inferred that prolonged breastfeeding causes slower growth in infancy. Objective: We compared infant growth associated with ≥12 months vs a shorter duration of breastfeeding based on three different analytic approaches to the same data from a randomized trial: intention-to-treat (ITT, “as randomized”), observational (“as fed”), and instrumental variable (using randomization as an “instrument” to achieve ≥12 months of breastfeeding). Design: This is a cluster-randomized trial of a breastfeeding promotion intervention. Anthropometric measurements were obtained at birth and 1, 2, 3, 6, 9, and 12 months. Results: The three analytic approaches yielded different results. The ITT approach showed more rapid growth in the first 2 months among infants randomized to the breastfeeding promotion intervention vs the control, with a decreasing difference over the ensuing months and nearly identical weight, length, and BMI by 12 months. The observational analysis revealed a different trend: higher weight and length in infants breastfed ≥12 months vs <12 months during the first 3 months, no difference by 6 months, and infants breastfed <12 months with increasingly higher weight and length from 6 to 12 months. The IV analysis showed a temporal pattern that was similar to that seen in the ITT analysis, but with larger (and less precise) differences among infants breastfed for ≥12 vs <12 months. Conclusions: We observed major differences in experimental (ITT and IV) vs observational approaches to analyzing data obtained from the same children. These approaches lead to opposite causal inferences about the relationship between infant feeding and growth and underline the importance of ensuring that the postulated cause (feeding) temporally precedes its hypothesized effect (growth).