Measuring "Camouflage" in Males and Females with Autism: Clinical, Cognitive, and Neuroanatomical Associations

Background:  Clinicians have observed that some high-functioning adults with autism, particularly females, ‘camouflage’ their social-communication difficulties, often leading to anxiety and/or depression. This phenomenon has not yet been measured, and its cognitive-biological bases remain unclear.

Objectives:  To (1) quantify ‘developmental change’ (DC) and ‘camouflage’ (CF) in high-functioning adults with autism and compare that between males and females; (2) test the hypotheses that higher CF is associated with more severe anxiety/depression, and cognitively better signal detection from background events and more conservative responses; and (3) explore the neuroanatomical bases of CF.

Methods:  Data from age- and IQ-matched high-functioning adult males and females with autism (N=30/group) were analyzed. Childhood ADI-R and current ADOS social-communication, Autism Spectrum Quotient (AQ), and ‘Reading the Mind in the Eyes’ Task (RMET) scores were all standardized (mean-centered and uniformly scaled [divided by maximum possible score of each]). DC was operationalized as the difference between ADI-R and ADOS (S_ADI-R－S_ADOS, higher scores reflect more ‘decline’ in autistic features). ‘Camouflage’ was operationalized as the discrepancy between internal ‘actual’ states and external behavioral presentations in interpersonal contexts, quantified as the difference between self-report autistic traits and explicit behaviors (S_AQ－S_ADOS), and between mentalizing ability and explicit behaviors (－S_RMET－S_ADOS). Their first principle component score was taken as a parsimonious CF measure (higher scores reflect more ‘masking’). Depression/anxiety were measured by the Beck Depression/Anxiety Inventories (BDI/BAI). Signal-sensitivity and response bias were measured using a Go/No-Go task. Non-linearly normalized gray and white matter segments estimated from structural MRI (3T-system) were used for behavior-neuroanatomy analyses by mass-univariate general linear models and multivariate pattern recognition approach (Gaussian process regression, leave-one-out-cross-validation, 5000 permutations assessing significance).

Results: Both DC and CF were normally distributed and unrelated to age or IQ. Females scored significantly higher than males (DC: p=.030; CF: p<.001). CF was positively correlated with BDI (but not BAI) (Pearson’s r=.301, p=.019) and Go/No-Go sensitivity (but not response bias) (r=.311, p=.017). However when sex-stratified, the CF-BDI correlation was significant in males (r=.533, p=.002) but not females (r=.030, p=.876), and the CF-sensitivity correlation was significant in females (r=.432, p=.017) but not males (r=.233, p=.223); sex differences in correlations (testing the significance of CF-by-sex interactions) were both at trend-level (p=.068, p=.072). In the whole sample CF was negatively correlated with gray matter volume at right middle superior temporal gyrus and rolandic operculum, and correlations were sex-dependent at left medial temporal lobe and cerebellum (males: positive; females: negative) and left dorsolateral prefrontal cortex (males: negative; females: non-significant). Total gray matter voxels had modest predictive power for CF in males (correlation=.27, p=.029; mean-squared-error=.14, p=.030) but not in females; white matter had no predictive power.

Conclusions:  On average, high-functioning adult females with autism showed more ‘camouflage’ than high-functioning adult males with autism, with substantial variability in both sexes. Higher camouflage was associated with more depressive symptoms in males, and better signal-detection sensitivity in females. The neuroanatomical associations of camouflage were partly sex-dependent. The measurement, mechanisms, consequences, and heterogeneity of ‘camouflage’ in autism warrant further investigation.