ASSOCIATION OF BREAST-FEEDING AND STUNTING IN PERUVIAN TODDLERS - AN EXAMPLE OF REVERSE CAUSALITY

Background. Child feeding recommendations include breastfeeding beyond 12 months, however, some researchers have reported increased rates of malnutrition in breastfed toddlers. A negative association between gr owth and breastfeeding may reflect reverse causality; that is, the out come (growth) is a determinant of the predictor (breastfeeding), and n ot vice versa. We examined this question with data from 134 Peruvian t oddlers. Methods. A linear regression analysis predicted length at the age of 15 months by length at 12 months, study interval, and 12-14.9- month breastfeeding, complementary food intake, and diarrhoeal inciden ce. This analysis defined the association between breastfeeding and li near growth. To elucidate the direction of the effect between breastfe eding and linear growth, logistic regression was used to predict the p robability of weaning by the end of 14 months. Determinants included w eight-for-age (W/A) at 12 months, complementary food intake at 9-11.9 months, and change in diarrhoeal incidence between 9 and 14.9 months. Results. There was a significant (P < 0.01) interaction of breastfeedi ng, diarrhoeal incidence, and complementary food intake on length at 1 5 months. Increased breastfeeding was associated with a 1.0 cm decreas e in length gain when dietary intake was low and diarrhoeal morbidity was high, implying that breastfeeding is harmful. The logistic analysi s, however, demonstrated that the risk of weaning decreased only when W/A and dietary intake were low and diarrhoeal morbidity was high. Con clusions. The negative association between breastfeeding and linear gr owth reflected reverse causality. Increased breastfeeding did not lead to poor growth; children's poor growth and health led to increased br eastfeeding. Children's health must be considered when evaluating the association of breastfeeding with anthropometric outcomes.