Rapid Infant Weight Gain: Critique of the Gilley et al Paper and Thoughts on How to Move the Field Forward

One must question the use of the Healthy Start Study for this research. The sample size is 414 and only 77 infants had RWG. These numbers can be halved for sex-stratified analyses. Despite these low numbers, the aim was to examine whether the associations of RWG with child body size and composition differ according to maternal ppBMI. The problem is that the numbers are not anywhere near large enough to compare how the associations differ between women with ppBMI at different points of the distribution (eg, 30 vs 23 kg/m2). A model will provide estimates but whether they can be trusted is questionable. Other studies have similar variables in samples that are at least 10-fold larger (eg, Millennium Cohort Study),^1,2  and it would be good to address the same research question using those data.

The paper includes no theory as to why the associations might be stronger at higher levels of ppBMI. This was the stated hypothesis and what was observed in girls for child BMI (z-score) and FFMI(kg/m^2). Like the authors, I have no explanation why the interaction estimates for child BMI and FFMI were so different in boys than girls. For percent FM, estimates are only shown for sexes combined. The interaction estimate was 0.28 (−0.01 to 0.57); it is a shame that this finding is largely dismissed because of a p value of .056 > .05. In the discussion, the authors write: “studies at older ages may clarify whether there is truly an interactive effect on percent FM” because the adiposity rebound “may make detection of percent FM differences more challenging.” I do not agree with this reasoning, particularly given that the study found strong relationships of RWG and ppBMI with percent FM. Unfortunately, the use of different outcome scales (ie, z-score, kg/m^2, %) and a mixture of sex-stratified versus sex-specific results makes comparison (eg, between outcomes) challenging. Adjustment of the models for mediators is also potentially problematic.³ 

The association of RWG with child BMI is self-fulfilling in the sense that it partly reflects tracking of weight (eg, between 5 months and 5 years). The authors defined RWG as a change in infant weight >+0.67 Z-scores (equivalent to one centile band on UK but not USA charts) because “this cut-off has been previously correlated with increased risk for childhood obesity.” But if they had used a higher cut-off (eg, >+1.34 Z-scores) they would have found stronger associations of RWG with the outcomes.¹  The use of >+0.67 Z-scores is common practice, but this cut-off is arbitrary and does not identify extreme cases. We have long known that growth assortment occurs following birth, with many infants shifting up 1 centile band on a growth chart.⁴  Perhaps what is more important is whether linear growth also follows the same pattern. The highest BMI values and obesity risk will occur in infants with RWG whose length does not also increase by >+0.67 Z-scores. Appropriate consideration of birth size is also important given some evidence that fetal growth modifies the relationship of RWG with child adiposity.⁵