Are males and females with autism spectrum disorder more similar than we thought?

Background

The predominance of autism spectrum disorder (ASD) diagnoses among males, with average estimates suggesting a 4:1 ratio (Christensen et al., 2016), is one of the most consistent features of the disorder. Previous research examining gender differences among individuals with ASD has suggested that females with this disorder have fewer restricted and repetitive behaviors (Frazier et al., 2014; Wilson et al., 2016), but lower intelligence quotient (IQ) associated with greater symptom impairment overall (Fombonne, 2009). Furthermore, some research indicates that females with ASD may be less likely to show the typical ASD cognitive profile of greater nonverbal skills compared to verbal skills (Ankenman et al., 2014). It has been suggested that these differences in behavioral and cognitive profiles may impact the presentation of social and communication symptoms in females with ASD (Black et al., 2009). However, research examining gender differences in clinical profiles of core social communication symptomatology among individuals with ASD has reported inconsistent findings; indeed, females with ASD have been reported to have superior, equivalent, or poor social communication skills compared to males (for reviews see Kirkovski et al., 2013; Van Wijngaarden-Cremers et al., 2014).

Objective

The purpose of this study was to examine gender differences in age of diagnosis, cognitive profiles, social communication symptomatology, and ASD symptom severity in a community-based sample of males and females with ASD. This study attempted to overcome past methodological limitations using a large sample of females and a combination of clinician report measures that include data from direct observation as well as parent report information contained within the clinical record.

Methods

Evaluations were conducted across statewide outpatient clinics operated by the University of North Carolina TEACCH Autism Program between January 2001 and March 2013. A total of 679 participants (males = 566, females = 113) who provided consent for their clinical data to be used for research purposes were selected for analysis (See participant characteristics in Table 1). These participants were administered an Autism Diagnostic Observation Schedule–Generic (ADOS-G; Lord et al., 2000), Childhood Autism Rating Scale (CARS; Schopler et al., 1988), and a developmental/IQ measure, such as the Wechsler scales, as part of their assessment. The ADOS-G is a standardized tool based on behavioral ratings conducted during direct observation, while the CARS standardized ratings incorporate all clinical observations, including those from the ADOS-G, parent report, and data from clinical records with possible scores ranging from 15 to 60. Higher scores on both the ADOS-G and CARS indicate the presence of more ASD symptoms. Both the ADOS-G and CARS were completed by trained clinicians. Participants were included if they received a Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM-IV) clinical diagnosis of autistic disorder, Asperger’s syndrome, or pervasive developmental disorder-not otherwise specified (PDD-NOS). Use of a sample of individuals referred to community-based clinics for diagnostic evaluations is unique as much of what is published on ASD uses a research sample rather than a clinical sample. Thus, the present sample may be representative of a broader community-based population rather than representative of those selected based on specific research inclusion criteria.

OPEN IN VIEWER

Table 1.

Sample characteristics.

	FemalesM (SD)	MalesM (SD)
N	113	566
Race (% White)	77%	71%
CARS total	32.05 (4.8)	31.12 (4.3)
ADOS-G total	12.85 (4.5)	13.58 (4.5)
Age at diagnosis (years)	10.55 (6.6)	10.13 (6.7)
Full-scale IQ	85.59 (22.1)	85.98 (21.8)
Verbal IQ	90.68 (21.4)	92.27 (20.8)
Nonverbal IQ	89.65 (20.8)	94.74 (19.8)

SD: standard deviation; CARS: Childhood Autism Rating Scale; ADOS-G: Autism Diagnostic Observation Schedule–Generic; IQ: intelligence quotient.

Age at the time of initial ASD diagnosis ranged from 1 year, 9 months to 56 years, 4 months with a median age of 8 years, 7 months. The mean full-scale intelligence quotient (FSIQ) score for the sample was 85 (standard deviation (SD) = 22; range = 40–165). Approximately 24% of the sample had an FSIQ score below 70. While the rates of co-occurring intellectual disability were lower than is reported by recent prevalence estimate of 31% based on 8-year-old children from the Centers for Disease Control and Prevention (CDC; Christensen et al., 2016), this could be due to the wider age range in this sample which may be identifying individuals toward the higher end of the spectrum that are identified in later childhood, adolescence, or adulthood.

Results

ADOS-G module scores

No gender differences were found in the overall sample’s ADOS-G Communication Total, t(667) = −1.74, p = 0.08, d = 0.18, Social Interaction Total, t(668) = −1.50, p = 0.13, d = 0.15, or Total scores, t(677) = −1.59, p = 0.11, d = 0.16. However, when specific modules were examined, gender differences on the ADOS-G were found on some modules (see Table 2). No significant gender differences were found on modules 1 or 4. On module 2, females obtained lower scores on Social Interaction, t(96) = −2.26, p = 0.03, d = 0.57, and Total, t(98) = −2.08, p = 0.04, d = 0.53. While mean scores for females exceeded the ADOS-G module 2 suggested cutoffs for autism spectrum across domains, females’ scores were approximately 2 points lower than males. Females also obtained lower scores by approximately 1–2 points for all domains of the module 3 ADOS-G, Wilks’ λ = 0.978, F (3, 378) = 2.9, p = 0.04, η_p², and all t’s > −1.94 and all p’s < 0.05. While these differences on modules 2 and 3 are below the significance level of 0.05, adjusting for the large number of analyses, these differences are approaching the significance level of 0.01. Correlations between the CARS Total score and ADOS-G social, communication, and total algorithm scores were highly positively correlated, all r’s > 0.40 (range = 0.40–0.46) and all p’s < 0.001, with no differences across gender.

OPEN IN VIEWER

Table 2.

Mean raw scores on the ADOS-G algorithm by module for males and females.

		Module 1	Module 2	Module 3	Module 4
N	Female	18	13	64	18
	Male	67	90	329	80
Communication	Female	6.06	5.77	3.84*	3.83
	Male	5.85	6.48	4.35*	3.75
Social	Female	10.83	7.15*	7.83*	8.61
	Male	9.74	9.08*	8.67*	8.36
Total	Female	16.83	12.92*	11.78*	12.44
	Male	15.60	15.61*	12.96*	12.12

*p < 0.05.

Conclusion

Overall, the results of this study suggest either very small or no gender differences in age at diagnosis, IQ score, cognitive profiles, or ASD symptom severity in this community-based sample. This is a departure from clinical lore that females with ASD are more likely to have lower IQ and more severe impairments. There is a slight difference in symptom severity with females having higher average total scores on the CARS, but this difference is likely of minimal clinical significance based on its small effect size and the magnitude of the difference (i.e. less than 1 point). In contrast, on the ADOS-G, females were found to receive lower scores than males particularly on modules 2 and 3. Similar to the profile obtained by the CARS, while females received lower ADOS-G scores, this may be of minimal clinical significance based on the magnitude of the difference. However, as the scores on these tools help to inform clinical judgment, this trend may be worth consideration when interpreting scores for males and females.

Other recent studies with relatively large samples have reported a wide range of findings, such as females presenting with greater social communication difficulties and poorer cognitive functioning during childhood and adolescence (Frazier et al., 2014) to females presenting with milder social communication difficulties as rated by the ADOS in adulthood (Lai et al., 2011). While our study’s results are consistent with more similarities than differences among males and females with ASD in this community clinic sample, some subtle differences were noted, particularly in females with communication skills necessary for module 2 and module 3 (i.e. phrase and sentence speech). This is similar to research by Rynkiewicz et al. (2016) who hypothesized that females with ASD who have better communication skills, including nonverbal gesture use, may demonstrate “camouflaging effects” of ASD symptoms. Similarly, Howe et al. (2015) suggest that females with fluent speech may show less social impairment. Taken together with other recent research in this area, the current results suggest that there may be periods of development, particularly in early-middle childhood and adolescence, when ASD social communication symptoms may present in different or more subtle ways in females.

This difference in social communication abilities as measured by the ADOS-G appears particularly true for individuals with higher verbal cognitive abilities on module 3 in the current sample. Given the increased language demands of fluent speech required to receive a module 3 ADOS-G, it may not be surprising that a profile of more developed verbal skills is related to scores received on this module. This may suggest that the historical notion of a stronger nonverbal than verbal ability profile may not be a universal feature of ASD particularly as this profile appears to shift throughout development and may vary by gender (Ankenman et al., 2014). Our finding that those with a higher verbal score had a lower ADOS-G total score on module 3 also suggests that those with higher verbal abilities may be more difficult to identify using current diagnostic measures.

As this is a large sample of a clinical population of individuals diagnosed with ASD in community-based clinics, there was substantial power to find potential differences for many analyses. While sample sizes for some of the ADOS-G modules and for some of the age groups are smaller in size, this limited the extent to which certain research questions could be asked of those smaller groups. Unsurprisingly, significant results were found where sample sizes were large enough to provide sufficient power. While this is a feature of this clinic-based sample, future research will need to continue to include participants across all of these groups. Another strength of this study is that it includes direct observation from the ADOS-G as well as parent report information contained in the clinical record that factors into the CARS score. Given that this study represents one of the few large studies of females with ASD across a wide age range, it may provide insight into the nature of some potential developmental changes between childhood and adulthood that fits with previous research findings.

Future research examining the possibility of more subtle gender differences, particularly in verbally fluent children and adolescents, is needed. It is possible that gender differences and these subtle social communication symptom differences may not be fully captured by our current quantitative measures. As we learn more about the social communication differences present during this time period (e.g. improved nonverbal gesture use in females, presence of more socially appropriate special interests), this may have important clinical implications in developing gender inclusive diagnostic tools and norms. Interestingly, some studies have pointed toward females displaying fewer restrictive/repetitive behaviors (RRB) as a possibility for lower diagnostic rates (Frazier et al., 2014). Because data for this study were collected using the ADOS-G as the second edition of the ADOS (ADOS-2) was not yet available, RRBs were not included in the algorithm scores and is thus a limitation of this study. Future research using the ADOS-2, CARS-2, and Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5) criteria to examine gender differences in RRBs is needed.

This study suggests that clinical lore about gender differences between males and females with ASD may not be valid. This may be especially true in community-based settings. While some subtle differences in ASD symptomatology were observed, particularly, in verbally fluent children and adolescents, in general age of diagnosis, IQ, and symptom severity were similar across genders.