Autism questionnaire scores do not only rise because of autism.

Abstract

EDITOR—We wish to respond to Morris et al.’s letter regarding our study and methodology.1, 2 In our study, Social Responsiveness Scale (SRS) mean differences between neurofibromatosis type 1 (NF1) and controls became negligible when they were adjusted on confounding variables measuring attention-deficit/hyperactivity (ADHD) and internalizing symptoms, and communication skills. Interpreting unadjusted elevated SRS scores as reflecting a specific elevation of autism symptomatology among participants with NF1 was not supported. SRS variance was explained by the combination of internalizing symptoms and communication delays (18.3%), and not only by ADHD (15.3%). A covariance analysis framework suited our a priori objectives by estimating the variance explained with each sequential adjustment. Similar results would have been obtained with multiple linear regression or semi-partial correlations. Four other findings reinforced the internal validity of our results. First, mean scores of NF1 participants on autism symptomatology measures other than SRS were very low. Second, Vineland Adaptive Behavior Scales, Second Edition scores of participants with NF1 did not show the trough in Socialization relative to the Communication and Daily Living Skills domains typically seen in autism spectrum disorder (ASD). Third, alleviating concern about low participation and selection bias, the electronic medical records survey showed increased frequency of ADHD, and not ASD, diagnoses among children with NF1. Fourth, ADHD and SRS scores were highly correlated within Simons Simplex Collection-typically developing controls (ρ=0.540; p<0.001), demonstrating that this correlation did not arise from the presence of autism. This pattern of results is not compatible with interpreting the elevation of SRS scores as a true elevation in autistic symptomatology. Prior investigations3-5 also showed that parent-reported autism symptom ratings either on questionnaires or diagnostic interviews can be confounded by the co-occurrence of behavioral problems and cognitive/developmental delays. Morris et al. claim that presumed overlapping ADHD/ASD genetic liability exempts investigators from adjusting SRS scores on covariates. On this, we add three comments. First, testing hypotheses about associations requires that confounding is reduced in relevant measures. In this argument, it is useful to consider that the terminology of confounding refers to two separate concerns. Concern 1: systematic error in measuring one specific health/disease construct or variable, a question of construct validity. Here, the issue is whether a raised SRS score means true elevation of autistic symptomatology or whether an SRS score can rise due to non-autistic phenomena that interfere with the measurement process and contaminate SRS scores. Concern 2: estimating the association between two separate health/disease constructs. Whatever measure of association is chosen, the association between the two variables of interest may be confounded by a third variable leading to bias in the association estimate. The extent of confounding, and the methods to remedy it, will vary according to study context and design; yet, generating unbiased estimates of associations (concern 2) requires, as a pre-requisite, that each construct is validly measured (concern 1). Our study dealt mostly with concern 1, pointing at construct validity limitations of SRS scores. Second, the need to remove confounding in estimating associations still exists when dealing with correlated variables reflecting shared etiologies or mechanisms, such as the need to adjust for height, age, or anxiety levels to evaluate the unique effects of body weight, head circumference, or depression respectively. Third, in the INFACT study,6 Morris et al. also adjusted SRS scores on ADHD covariates but methodological limitations prevented them from identifying other confounding effects observed in independent studies (e.g. behavioral problems, language and intellectual level)3-5 (Appendix S1, online supporting information). We warn SRS users against oversimplified interpretations of SRS scores. Misuse of terminology such as ‘pathological shift’ or elevated ‘autistic trait burden’ to describe elevated SRS scores forces a prefabricated interpretation that our, and other studies,3-5 caution against. The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.