This commentary refers to ‘Lifecourse trajectories of body mass index and adult cardiometabolic risk: is childhood a sensitive period?’, by L.N. Anderson et al., 2019;40:2920.

As the authors point out, other group-based trajectory model (GBTM) analyses of body mass index (BMI) identified up to five BMI trajectories, most of which were consistent with those identified in our sample.¹ However, these studies did not report the small trajectory group of ‘BMI improvers’. There are several explanations for this discrepancy. First, compared to our study sample that spans 30 years, previous studies followed participants over more restricted lifecourse periods (childhood/adolescence or adolescence/adulthood). Since the ‘improvement’ in BMI for age curves does not manifest until age ∼25 years in our ‘improvers’ group, it is not surprising that this group was not identified in previous studies. Second, the specification of the GBTMs used in previous studies was different from ours in several respects preventing direct comparison. Estimating models within the GBTMs framework involves making various decisions throughout the process [i.e. functional form for the measurement model, specification of the random-effect and variance-covariance structure (and between-class differences, if any), and the statistical criteria used for model comparison]. Along with considerations around sample size, missing data, and number of repeated measures within-subject, such decisions all likely affect class enumeration, model identifiability, and ultimately, results.² With the increasing popularity of GBTMs, inadequate or incomplete reporting of the results tends to impede their interpretation and critical appraisal, and make it difficult to compare results between studies. In response to this, a group of authors have recently developed a set of Guidelines (GRoLTS),² which we adhered to.

Our findings need to be replicated in larger external populations, and understanding the characteristics of participants classified in this small ‘resolving’ trajectory is paramount. We are currently examining long-term BMI trajectories in four studies that form part of the International Childhood Cardiovascular Cohort Consortium.³ We also agree that our cohort began before the more recent epidemic of paediatric obesity, which might impact the generalizability of our findings to current/future cohorts. In contemporary population samples, the proportion of children with more adverse BMI trajectories is likely higher, with the identified trajectories shifting towards higher average BMI levels in each group. This further highlights the need for replication in aging cohorts where paediatric obesity levels more closely reflect contemporary levels. However, these studies might have the same limitation as ours if paediatric obesity levels continue to shift over time. Nevertheless, it is likely our observed association between suboptimal lifecourse BMI trajectories and adverse adult cardiometabolic outcomes will still be relevant, and perhaps more so, for current/future generations where the incidence of overweight and obesity is predicted to keep rising.

Conflict of interest: none declared.