Aim: Although reverse causation is well known in the breastfeeding-allergy arena, only a few studies have used a valid statistical method to test its occurrence. This study aimed to illustrate reverse causation between eczema and breastfeeding and to examine their effect estimate under this occurrence. Methods: Time-to-event data from the Infant Feeding Practices Study II were used. Model 1: early signs of eczema were modeled as a time-varying covariate to assess reverse causation using survival analysis approach. Model 2 examined the crude and adjusted relative risk between breastfeeding and early signs of eczema using generalized linear model. Results: Overall, 2,350 mother-infant pairs were included. On average, infants with early signs of eczema versus without, breastfed for 5.2 ± 0.2 months compared with 4.10 ± 0.1, respectively. Model 1, an early sign of eczema, was associated with approximately a 16% (95% confidence interval 2-28) reduction in the risk of ceasing breastfeeding indicating reverse causation. Model 2, each month of breastfeeding increased the risk for early signs of eczema by 3%. Conclusion: Modeling early signs of eczema as a time-varying covariate using survival analysis approach detects reverse causation and children with early signs of eczema were breastfed for longer duration. Failure to account for reverse causation may create biased result and a conclusion where breastfeeding increases the risk for early signs for eczema. Future studies should perceive reverse causation by using time-varying covariate in survival analysis approach.
Keywords: a time-varying covariate; breastfeeding; infant eczema; reverse causation; time-to-event data.