Methodological and Reporting Issues in Child Weight Research in ASD

Thursday, May 14, 2015: 11:30 AM-1:30 PM
Imperial Ballroom (Grand America Hotel)
S. N. Grondhuis, The Ohio State University, Jackson, MS
Background: Researchers have been increasingly interested in studying weight in particular groups, such as those with autism spectrum disorder (ASD). They are specifically attempting to identify risk factors and correlates of abnormal weight, possibly for use in prevention or intervention programs, as well as to increase understanding of how prescribing medications could have long-term effects on body mass. While there is literature available that can offer insight, often times the methodology or reporting practices are inconsistent, which can lead to confusion and prevent future results from being compared to existing data. Part of this confusion in weight reporting may stem from authors choosing to report p values and specific results (e.g., odds rations or F tests) in the absence of reporting the amount of variance for which these risk factors were able to account.

Objectives: Identify the amount of variances different studies are able to account for and compare that to the results that are reported in their publications. Evaluate whether the way current data are presented is useful for general consumption, and if not, construct recommendations that will further understanding and allow comparisons to be drawn between studies.

Methods: Search for weight literature using keyterms such as “child,” “children,” “weight,” “overweight,” “obese,” and “obesity” that were published no earlier than the year 2000. All abstracts were evaluated to see if the paper included (a) original research, (b) identification of a risk factor or correlate of weight, and (c) included statistical outcomes that could be manipulated into amount of variance accounted if it was not explicated stated. If a paper appeared to include these criteria, it was evaluated further. Special attention was given to papers directly including children with ASD.

Results: Preliminary results at the time of abstract submission identified 23 articles that met the criteria mentioned above. Four of the articles related to children with ASD and an additional four related to children with intellectual or developmental disabilities. Although it was expected that multiple statistical techniques would be used to evaluate weight (multiple regression, logistic regression, etc.), the results produced highly variable outcomes, with amount of variance ranging from .01%-41%.

Conclusions: Most studies where investigators reported statistic  that were able to be converted into amount of variance accounted for produced relatively small amounts of variance (less than 8%). If studies accounted for substantially greater amounts of variance, they normally used BMI that was unadjusted for age, which is not recommended for children. Eleven studies out of the 23 preliminarily included were not able to be converted, as they did not include all needed necessary output (e.g., explicit degrees of freedom). Future researchers should be mindful that accounting for considerable variance is difficult, and possibly unlikely in weight studies, and that small numbers should be reported so it is common knowledge. The ASD field is investigating abnormal weight issues with increasing frequency, and as such, it is in our interest to quantify specific results (especially variance) for increased clarity by reporting more than significant p values.