We appreciate Reem M. Ghandour highlighting the changes in sampling approach between 2011–12 and 2016 for the National Survey of Children’s Health and sharing the concern that these changes make trend analyses impossible. These sampling changes are acknowledged in appendix A1 to our article, available online.

However, despite the changes made to the survey, we maintain that our selected study period and interrupted time series analytic approach were reasonable for two key reasons. First, each survey was representative of state and national populations and was conducted by highly respected survey organizations. Second, Ghandour notes that when investigating possible systematic biases resulting from differential nonresponse and mode effects for key indicators, “no consistent mode-specific patterns were identified.” This suggests that estimates of obesity prevalence derived from the surveys were unlikely to be systematically different as a result of the sampling approach change. It seems implausible that non-systematic errors resulting from the survey redesign would be large enough to account for increased obesity prevalence from 2003 to 2011–12 for children in poverty (which is consistent with prior research using other nationally representative data)¹ and then produce a consistent, continuing decline in obesity prevalence for children in poverty from 2016 onward (and no trend among children not in poverty).

Therefore, we respectfully submit that although we acknowledge the National Survey of Children’s Health sampling changes and the potential for unknown biases in our analysis of observational data, the obesity prevalence estimates from the survey over this time frame can still be leveraged to examine population-level shifts in obesity that followed the Healthy, Hunger-Free Kids Act of 2010.