Significant Increase in Power in Clinical Trials with Use of the Newly Available Vineland-3 Growth Scale Values

Background: Although adaptive behavior is relevant due to its clinical significance, it is used rarely as a primary outcome measure in clinical trials. This may be because adaptive behavior scores are often expressed as scaled or standard scores. Scaled scores, which provide information about a child relative to their peers, may remain unchanged even as a child’s ability level increases or decreases. Further, scaled scores may decrease even if ability increases, if the rate is slower than chronological-age expectations. Thus, studies adequately powered to detect changes in scaled scored require relatively large and potentially unobtainable sample sizes. The third edition of the Vineland Adaptive Behavior Scales (VABS-3) includes growth scale values (GSV), which are an index of ability derived through item response theory. GSV quantities change within individuals, including small changes. Given its recent publication, the VABS-3 GSV should be explored as a potential outcome measure in studies of neurodevelopmental disorders.

Objectives: The goal of this study was to demonstrate the advantage of VABS-3 GSV for use as an outcome measure in clinical trials of neurodevelopmental disorders. Using data from an ongoing natural history study of creatine transporter deficiency (a rare, X-linked, metabolic condition associated with autism spectrum disorder and other neurodevelopmental disorders), we quantified change over 6 months in three types of scores (scaled score, age equivalent, and GSV). Post-hoc power calculations yielded estimates for sample sizes (per group) required to detect the observed effects with 80% power in future studies.

Methods: The interview form of the VABS-3 was administered to parents during study visits spaced at 6-month intervals. Here, we used data from the 14 participants with histories of global developmental delay and varying degrees of autism symptomatology who completed baseline and 6 month visits (age M=7.1, SD=3.6, range 3 – 13 years; VABS-3 Adaptive Behavior Composite Score M=47.9, SD=15.6, range 20-69). Linear mixed models were used to estimate mean change and the standard error of that estimate.

Results: Preliminary data indicate that for most subdomains, effect sizes were largest for GSV scores (see Figure 1). In some cases, the V-Scale effect size was the largest; however, V-scale in these cases indicated a worsening of relative standing, while the GSV indicated slight improvement or stability (e.g., Playtime). Where both the V-Scale and GSV indicated improvement over 6 months, the decrease in required sample size associated with use of GSV ranged from 43% to 83%. For example, the required sample size for Receptive Language GSV was n=26, compared to n=42 and n=46 for AE and V-scale, respectively.

Conclusions: Scaled or standard scores are poorly suited for use as outcomes in clinical trials because detectable changes in relative standing require proportionally larger changes in ability. The VABS-3 contains GSV, which index ability and are therefore explicitly designed for the assessment of within-person change. Using data from a natural history study of CTD, we demonstrated that studies which use GSV rather than scaled scores will require significantly fewer participants and will be less likely to mistake slower-than-expected growth as decline.