Examination of Sex Differences on the Autism Diagnostic Observation Schedule (ADOS) and Autism Diagnostic Interview-Revised (ADI-R) in a Clinic-Referred Sample

Background: Current studies examining sex differences in autism spectrum disorder (ASD) have yielded conflicting results. While some evidence suggests that females exhibit fewer symptoms in social communication (Lai et al., 2011; Zwaigenbaum et al, 2012) and restricted and repetitive behaviors (Frazier et al., 2014; Hartley et al., 2009; Mandy et al., 2011; Tillmann et al., 2018), others have found similar levels of symptom presence and severity across males and females (Anderson et al., 2013; Baron-Cohen et al., 2003; Lord et al., 1982; Reinhardt et al, 2014). Further exploration of sex-specific behavioral manifestations of ASD is critical given that females are more likely to be under-identified and diagnosed later than males (Begeer et al.; 2013, Giarelli et al., 2010; Wilson et al., 2016).

Objectives: To examine potential sex differences on gold-standard assessments for ASD (i.e., the ADOS and ADI-R) in a clinic-referred sample using symptom domain scores, overall severity level, and cutoffs for optimal sensitivity and specificity.

Methods: Analyses were conducted using a sample of 257 participants, comprised of 199 males (77.4%) and 58 females (22.6%), who received a comprehensive evaluation at a university-based clinic specializing in ASD and other neurodevelopmental disorders. Calibrated severity scores (CSS) were used as a measure of severity, both overall and within each symptom domain (i.e., Social Affect and Restricted and Repetitive Behaviors). Subscale scores from the ADI-R were included to examine caregiver-reported symptom severity across Social Interaction, Communication, and Repetitive Behavior domains. MANOVA analyses were completed to assess main effects for sex and diagnostic group (i.e., ASD and non-ASD) on ADOS and ADI-R scores; sensitivity, specificity, and cutoff scores were examined for CSS scores using ROC curve analyses for males and females separately.

Results: No significant differences were found on ADOS CSS or ADI-R subscale scores between male and female groups with ASD. However, ROC curve results suggested that a lower cutoff on the ADOS CSS maximized sensitivity and specificity for accurately detecting ASD in females (cutoff=4, area under the curve (AUC)=.90, sensitivity=.94, specificity=.84), compared to males (cutoff=6, AUC=.86, sensitivity=.89, specificity=.72). While ideal for males, a cutoff of 6 decreased sensitivity (.83) for females without improving specificity. The optimal cutoff for females (i.e., 4) fell within the “ASD” ADOS classification range, while a cutoff within the more severe “autism” classification range (i.e., 6) best predicted ASD in males.

Conclusions: Results support previous findings that sex differences are often not readily apparent when comparing scores on gold-standard assessments, despite prevalent clinical observations that girls with ASD seem to present differently than their male counterparts. However, variability across groups in optimal CSS cutoffs suggests that subtler symptom severity and scores may be expected when using the ADOS as a diagnostic tool for girls. This could be especially relevant in high acuity clinical populations with comorbid neurodevelopmental and psychiatric disorders in order to reduce false negatives in girls with complex presentations. Future analyses should consider sex differences by age to examine implications of developmental trajectory on male and female symptom presentation.