Perinatal & Pediatric
Screen device use and body mass development in children: A causal inference approach to sensitive periods Yuna Koyama* Yuna Koyama Ryan Keen Yena Kyeong Jonathan Bernard Natarajan Padmapriya Falk Muller-Riemenschneider Fabian Yap Yung Seng Lee Law Evelyn Chung Ning Peipei Setoh Henning Tiemeier
Background: “Sensitive periods” are commonly suggested in epidemiological studies but rarely embedded in a causal framework. We examined causal associations of screen time with body mass index (BMI) and possible sensitive periods, leveraging 10 repeated exposure and outcome assessments from ages 1 to 10 years.
Methods: We analyzed data on 1,030 children (546 boys) from a prospective cohort in Singapore. Parents reported children’s screen time. Children’s BMI was derived from anthropometric measures of height and weight. We contrasted a mixed model approach with g-estimation of structural nested mean modeling that estimates average treatment effects while accounting for feedback between chained exposures and time-varying confounders. Average effects at any given age, latent effects, and sensitive period effects were explored.
Results: A 1 standard deviation (SD) increase in screen time at 2 years, but not at other ages, was associated with persistently higher BMI by 0.117 SD (95% confidence interval [CI]: 0.004; 0.230) between ages 2-10 years in boys in mixed modeling. G-estimation showed no sensitive period nor latent effects. However, an average effect across ages exists; if boys increased their screen time by 1 SD at any given age, their BMI at any subsequent age would increase by 0.045 SD (95% CI: 0.002; 0.087). No associations were indicated in girls in both models.
Conclusions: Using g-estimation, we showed an average treatment effect of screen time on subsequent BMI in boys but found no evidence for sensitive periods. Time-varying confounders, chained exposure, and latent effects are key for identifying sensitive period effects.