Letter to the Editor: Reply to Johnson

We appreciate the thoughtful comments by Dr Johnson and address points of concern below.

First, there are certainly larger cohort studies in which to ask questions about maternal body size, early-life growth, and future obesity – including the Millennium Cohort Study. However, these large studies frequently rely on self-reported maternal prepregnancy BMI, which is subject to bias.¹  Larger studies may also lack directly-assessed body composition, a main outcome of interest and major strength of our study. We point out that with n = 414, we were able to detect significant and biologically relevant differences in adiposity among children who experienced rapid infant weight gain versus those who did not.

Second, the stronger associations observed among children of women with higher prepregnancy BMI may reflect that maternal adiposity and rapid infant growth operate jointly to promote child adiposity. For example, maternal prepregnancy BMI may influence offspring adiposity through in utero programming of adipogenic potential,²  whereas rapid infant growth may result from early-life microbiota composition,³  among other factors. The possibility of distinct pathways gives rise to a biological interaction and provides the impetus for the study at hand.

Third, our interest in sex differences is based on literature demonstrating differences in obesity-related outcomes following in utero exposure to developmental overnutrition for males versus females.⁴  We used a data-driven approach by first testing for a statistical interaction and then conducting sex-stratified analyses where appropriate. Although the model of percent fat mass was not stratified by sex, we presented sex-specific estimates in Figure 2 of the original manuscript.

Fourth, Dr Johnson inquired on criteria for statistical significance and scales for each outcome. We agree that P values are arbitrary, but we remained true to the α-level selected for significance before modeling to avoid false positive findings. Regarding outcome scales, we assessed age- and sex-specific offspring BMI z-score as it is an internationally accepted proxy for excess adiposity⁵  and kept body composition in their native units for interpretability as there are currently no established thresholds or external references.

Fifth, regarding concern about inclusion of mediators as covariates, we agree that small for gestational age may be a mediator, but we chose to include it given the importance of birth size to infant growth, which Dr Johnson acknowledges in the comment. Even with SGA in the model, we still observed significant associations.

Finally, there are multiple approaches for defining rapid infant weight gain. We selected >+0.67 z-scores as this cutoff is widely used in the literature, thereby allowing for comparison of our findings to previous studies. Further, pediatric medical providers are familiar with z-scores, facilitating easy identification of rapid infant weight gain in a clinical setting. We chose to not focus exclusively on “extreme cases” of rapid infant weight gain as this may exclude some children with increased risk of obesity and further limit our sample size. We did not study linear growth in the current study, but agree it is an interesting area for future research.