The discontinuation of psychostimulants is followed by accelerated growth [
58]. Conversely, as noted earlier, prior treatment status is a strong correlate of AAP-induced weight gain. Nonetheless, we found no significant association between change in weight prior to the onset of risperidone and the change afterwards.
Weight gain during the first few weeks of olanzapine treatment is a strong predictor of future change in weight [
22,
59]. We found only a statistical trend suggesting that an increase in BMI z score during the first three months of treatment is associated with further weight gain during more extended treatment, after taking into effect the other “primary” covariates. However, the sample size in this analysis was too small and requires future replications.
The results reported here are subject to limitations. First, it would be ideal if a randomized trial involving various AAPs explored both efficacy and tolerability during extended exposure. However, such a study is challenging to conduct for ethical, budgetary, and practical reasons. Alternatively, studies such as this one can help fill the gap by providing sorely-needed information about the long-term safety of AAPs using a more pragmatic design. Nonetheless, this remains a retrospective study with missing data, particularly regarding patients who discontinue risperidone early on and, as a result, are not enrolled. It is somewhat reassuring that the amplitude of weight gain we found is comparable to that reported in shorter-term studies [
12]. In addition, nearly a third of our participants (53/163) were excluded from the analysis due to missing data. They differed from the included participants on enrollment weight, but not BMI, z score. Thus, this finding could be due to a type I error since it was the only variable that was significantly different between the two groups across 39 tested. Secondly, the specific aim of the parent study was to investigate the metabolic and skeletal effects of risperidone specifically in youths who have been receiving it over extended periods of time. It is in these children that one would be most concerned about long-term AAP safety, making the results both relevant to common clinical practice and generalizable to the broad population of children treated with AAPs. Third, except for those obtained upon study enrollment, all anthropometric measurements were not necessarily collected following standard procedures. However, like others [
60], we found weight and height measurements extracted from the medical record to be highly correlated with the corresponding research measurements, obtained within a month. Further, we contend that it is the trajectory of change in weight (or BMI) during the course of treatment, as opposed to single measurements, that is of most clinical relevance. Modeling the weight (BMI) trajectory requires serial measurements, making it less susceptible to measurement errors. Fourth, we used the prescribed dose of risperidone as a predictor rather than risperidone serum concentration. This is particularly important as the serum concentration, had it been obtained repeatedly over the treatment course, would have better captured medication adherence. As discussed earlier, long-term adherence is challenging to estimate accurately and tends to decrease over time. In fact, this phenomenon could potentially explain the apparent attenuation of the effect of risperidone dose on weight that we found. Fifth, while comorbidity was common, the prevalence of various disorders was somewhat bimodal with ADHD and disruptive behavior disorders being almost universal with the other disorders afflicting a relative minority. This precluded our ability to explore whether the underlying psychiatric disorders moderated risperidone-related weight gain. Short-term randomized pediatric trials in disruptive behavior, bipolar, psychotic, tic, and pervasive developmental disorders have all revealed a somewhat comparable weight gain with AAPs [
22,
36,
61‐
64]. However, none of these studies directly compared weight gain across different disorders, leaving unanswered the question of whether psychopathology moderates one’s propensity to gain weight during AAP treatment. Also, with the heritability of AAP-induced weight gain estimated at 60 to 80% [
65], it is likely that genetic factors contribute to this adverse event as we and others have shown [
66,
67]. These were not considered in the present analysis but should be in future ones. Finally, it is important to replicate our findings in a larger sample where females and individuals from a diverse racial/ethnic background are better represented.