Wechsler Intelligence Scale For Adults – Fourth Edition Profiles Of Adults With Autism Spectrum Disorder

Abstract

Aim. 

In this study, we have compared 229 Wechsler Adults Intelligence Scale – Fourth Edition (WAIS-IV) cognitive profiles of different severity adults with autism spectrum disorder to verify the impact of several variables including sex, age, level of education, and autism severity level in an Italian sample. Moreover, we wanted to find out the optimal cut points for the major intelligence quotients to discriminate autism severity levels. 

Methods. 

The Wechsler Adult Intelligence Scale is a tool used to assess an individual's intelligence level, including multiple dimensions such as vocabulary, detailed understanding, pattern recognition, reasoning, and numerical memory. Memory is an important aspect. There is a certain correlation between them.

Research shows that individuals with higher levels of intelligence tend to have better memories. This can be verified from multiple angles: First, individuals with high IQs tend to have stronger abilities to learn and remember quickly and apply knowledge when solving problems; secondly, individuals with high IQs are better able to learn and memorize knowledge. Understand and analyze knowledge and extract key information; in addition, high-IQ individuals can more effectively integrate and connect knowledge during the long-term accumulation of knowledge to promote long-term storage of memory.

However, it is worth noting that memory is not the only factor that determines intelligence level. In addition to memory, intelligence level also includes many other aspects, such as reasoning, creativity, etc. Therefore, it is not accurate enough to rely solely on memory to estimate intelligence level. The Wechsler Adult Intelligence Scale is designed to comprehensively assess multiple indicators of an individual's intelligence level for a more accurate assessment.

In general, memory is an important aspect of intelligence level. The Wechsler Adult Intelligence Scale is a tool for comprehensively assessing an individual's intelligence level. It can help us understand multiple aspects of an individual's intelligence level more accurately so that we can better understand the individual's intelligence level. Develop reasonable training plans and educational directions. It can be seen that we need to improve memory, and Cistanche deserticola can significantly improve memory because Cistanche deserticola is a traditional Chinese medicinal material that has many unique effects, one of which is to improve memory. The efficacy of minced meat comes from the various active ingredients it contains, including acid, polysaccharides, flavonoids, etc. These ingredients can promote brain health in various ways.

Click Know Short-term Memory how to improve

Participants were recruited from two National Health System Centers in two different Italian regions and were assessed with gold-standard instruments as a part of their clinical evaluation. According to DSM-5, cognitive domains were also measured with multi-componential tests. We used the Italian adaptation of WAIS-IV. We checked our hypotheses using linear regression models and receiver operating characteristics (ROC) curves. 

Results. 

Our results showed that age and level of education have a strong impact on Verbal Comprehension (VCI) and Working Memory Indexes (WMI). Gender differences are relevant when considering the VCI and Processing Speed Index (PSI) in which women obtained the best performance. These differences are still relevant when considering cut points of ROC because 69 resulted in being the optimal cut point for women, and 65 for men. 

Conclusions. 

Few conclusions can be assumed only by examining Full Scale Intelligence Quotient (FSIQ) scores as they include different information about broader cognitive abilities. Looking deeper at main indexes and their subtests findings are consistent with previous research on the disorder (moderate correlations of FSIQ, Perceptual Reasoning index, WMI, and PSI with the participants’ age), while other results are unforeseen (no effect of sex found on FSIQ score) or novel (significant effect of education on VCI and WMI). Using an algorithm predicting optimal cut points for discriminating through autism severity levels can help clinicians to better label and quantify the required help a person may need, a test cannot replace diagnostic and clinical evaluation by experienced clinicians.

Introduction

Autism spectrum disorder (ASD) is a neurodevelopmental disorder with an early onset and a genetic component. ASD is characterized by deficits in socio-emotional reciprocity, impaired verbal and non-verbal communication skills, and an inability to develop and maintain adequate social relationships with peers. ASD core symptoms are associated with the presence of repetitive verbal and motor behaviors, restricted patterns of interest, need for an unchanging environment (or in any case predictable and stable), and hypo- or hypersensitivity to sensory inputs. The onset of clinical symptoms occurs during the early years of life (APA, 2013). Specifiers consider the possibility of several comorbidities, such as a cognitive deficit, language impairment, catatonia, medical or environmental factors, or other neurodevelopmental disorders.

Recent prevalence estimates indicate 1: 44 children in the USA and 1: 77 children in Italy (Maenner et al., 2016). Adults’ prevalence is around 1: 68 revealing a significant increase in the population of adults with ASD (Christensen et al., 2016). Alongside this factor, another relevant element to be considered is the gender ratio across autistic people (Loomis et al., 2017) which is still debated and evidence mixed results. Sex-linked genetic factors and male vulnerability to brain insult may account for some of the gender differences (APA, 2013). Recent epidemiological studies revealed a 2–3: 1 male predominance compared to the widely cited 4–5: 1 ratio from earlier studies (Mattila et al., 2011; Idring et al., 2012; Baxter et al., 2015; Zablotsky et al., 2015; Keller et al., 2020) although this ratio may depend upon intellectual abilities and it appears as low as 2: 1 when ASD is associated with intellectual disability, and as high as 6–8: 1 in high-functioning autism (HFA; Fombonne, 2005, 2009). It is supposed that this higher male prevalence is due to autistic females’ ability to mask their social difficulties, to cultural factors, and a smaller number of studies on ASD in the female population (Attwood, 2007; Lai et al., 2011; Kirkovski et al., 2013) and different ASD phenotypes (Mandy et al., 2012; Van Wijngaarden-Cremers et al., 2014; Howe et al., 2015). A recent study by Wilson et al. (2016) involving 1244 adults (935 males and 309 females) referred for ASD assessment reported sex differences in clinical outcome. Results concluded that 639 males and 188 female people were diagnosed with ASD of any subtype. Indeed, in the study, no significant effect of sex (male IQ > female IQ; F(2) = 2.47, p = 0.09, η2 p = 0.02) on IQ was found. Regarding intelligence outcomes, their results confirmed previous research reporting lower IQ scores in females with a diagnosis of ASD compared to male participants (Fombonne, 2005). Indeed, Halpern and LaMay (2000) found no significant sex difference for the g-factor whereas sex differences play a role regarding achievements on subtests and indexes level using the Wechsler Intelligence Scale for Adults – 4th Edition (WAIS-IV; Wechsler, 2013).

Studies on the typical developmental (TD) population examining the gender differences using subtests and derived indices from WAIS-IV highlighted better performances of men in IQ, Verbal Comprehension (VC), Perceptual Reasoning (PR), and Working Memory (WM) indices (Longman et al., 2007; Irwing, 2012; Daseking et al., 2017). Instead, the Processing Speed (PS) index was the only one in which women had better outcomes. These results were in line with an Italian study by Pezzuti et al. (2020) that found that men performed significantly better than women in the Arithmetic subtest and the WMI of the WAIS-IV. In their study comparing performances of TD on WAIS-R and WAIS-IV, gender differences appeared broader and more extensive in the WAIS-R sample, as other previous authors mentioned using WAIS-III (Dolan et al., 2006; Van der Sluis et al., 2006). A factor analysis study from Colom and Garcia-Lopez (2002) outlined that there are no sex differences in general ability (g) on the Spanish standardization of the WAIS-III. The authors stated that the average sex differences favoring males have to be attributed to specific group factors and test specificity. Likewise, results obtained by Van der Sluis et al. (2006) using Dutch WAIS-III indicate differences between men and women in performance regarding specific cognitive abilities, but not in general intelligence (g). In contrast, for the US standardization sample of the WAIS-III, Irwing (2012) reported sex differences not only regarding specific abilities but also in g. Men outperformed women in general intelligence [Full Scale Intelligence Quotient (FSIQ)] and on subtests like Information, Arithmetics, and Symbol-Search, whereas women outperformed men on the Processing Speed Index (PSI).

Educational level (Ceci and Williams, 1997; Gustafsson, 2001) and age also contribute to the understanding of differences in IQ outcomes. Ceci (1991) suggests that the more years of education the better cognitive skills. This phenomenon is due to the exposition of contexts that allow people to learn relevant information, to concentrate on problems, and it teaches approaches of cognition on which the majority of intelligence tests are based. Results from an Italian study (Tommasi et al., 2015) showed that the WAIS-R detects individual differences in intelligence properly measured by IQ scores at different educational levels. Indeed, there is an average increase equal to 1.9 IQ points in the IQ global composite score per year of education. As previously hinted, age needs to be considered when accounting for IQ differences and efficiency across time (Baltes et al., 1998; Schaie and Willis, 2010). Most of the studies focused on the key role of Working Memory and its connection to general abilities. It has been argued that in TD a significant detrimental effect of age on Working Memory resources is played (Craik and Salthouse, 2008; Robert et al., 2009).

So the profile of intelligence level is one of the relevant factors to be considered when diagnosing people with ASD, alongside other cognitive, neuropsychological, socio-demographic, and core-symptoms measures (Happé et al., 2016). Recognizing how people with ASD may vary on this construct it may be crucial for identifying ASD subtypes (Grzadzinski et al., 2013). Therefore, ASD subtypes change according to different cognitive ability patterns (Grzadzinski et al., 2013). Nonetheless, there are no distinctive IQ profiles of individuals with ASD (Siegel et al., 1996; Ghaziuddin and Mountain-Kimchi, 2004; Goldstein et al., 2008; Williams et al., 2008; Charman et al., 2011). Intellectual abilities have been more challenging to assess in individuals with ASD because of their characteristics and assessment tools. Many researchers focused on children, but few authors studied cognitive performance patterns in adults with ASD and how these patterns can differentiate severity levels and typical performance configurations. WAIS-IV (Wechsler, 2013) is the most widely used and renewed cognitive performance test for the assessment of verbal adults with ASD. Other standardized measures of intelligence include the Stanford–Binet (e.g. Roid, 2003), Raven’s Progressive Matrices (RPM; Raven et al., 1998), and Leiter-3 (Roid et al., 2013). The use of Wechsler scales has been supported by several studies (Filipek et al., 1999; Mottron, 2004). Nevertheless, previous research has highlighted how the RPM (Raven et al., 1998) could be more adequate for describing the cognitive profile of people with ASD (Dawson et al., 2007; Hayashi et al., 2008; Soulières et al., 2011). Indeed, as pointed out by Dawson et al. (2007) the Wechsler scale may underestimate the intelligence of individuals with ASD mainly because of its emphasis on verbal instruction and tasks. However, the structure and the characteristics of the RPM, suitable for fluid reasoning tasks, maybe a more appropriate measure of the intelligence of people with ASD. Results of the comparison between performances of Wechsler and RPM scores of adults with and without ASD highlighted a significantly higher performance of the ASD group on RPM compared to the TD group, whose performances across the scales were without significant differences. However, the IQ discrepancy between people with ASD and TD made the in-depth comprehension of the differences in the cognitive performances of ASD people using RPM and the Wechsler scale. Results of a separate but related study suggest that the higher performance on the RPM as compared to the Wechsler measures primarily occurs for individuals with ASD with cognitive impairment (Bölte et al., 2009). Holdnack et al. (2011) compared performances between the control group, HFA, and Asperger disorder (AS) in the WAIS-IV subtests. No statistically significant differences between AS and control groups were found whereas the HFA group had the lowest scores. However, both ASD and control groups’ performances on Matrix Reasoning and Digits Forward revealed no significant differences. Regarding Coding subtests, all three groups differed significantly from each other. Eventually, in Visual Puzzles where the HFA group performed significantly more poorly than the control group, the AS group did not differ from either the HFA or the control group.

Summing up, several demographic variables are associated with different cognitive level abilities in TD. However, based on our knowledge, no study evaluated together the effects of age, sex, level of education, and level of autism on the cognitive performances of people with ASD measured with the Italian WAIS-IV in a large sample. So, in the present study, we tested several hypotheses:

(1) Test the association between the demographic variables and level of autism with FSIQ, main indexes, and subtests, as a preliminary step for further and in-depth analyses. A moderate correlation between age and level of education and FSIQ and the main indexes was expected.

(2) Assuming the FSIQ could not thoroughly explain the strengths and weaknesses of people with ASD assessed with the WAIS-IV, we wanted to identify if like TD, significant effects of the independent variables were found on the four indexes together (VCI, WMI, PRI, PSI) and the underlying subtests. Specifically, we expected no sex differences in FSIQ in both levels of autism; significant effects of age and level of education on VCI, WMI, and PSI; and ASD female participants’ better performances on PSI.
(3) Eventually, we wanted to test the hypothesis that better performances on the four indexes can predict less severe autistic symptoms. Indeed, optimal cut-off scores for discriminating autism severity levels using WAIS-IV were investigated.

Methods

Participants

In total, 270 adults with ASD (Mage = 26.3 S.D. = 9.35) were evaluated at the Regional Center for Autism Spectrum Disorder in Turin and the Regional Centre for Autism in L’Aquila (Italy). The Regional Center of ASL Citta di Torino is a national mental health system department providing services for people with ASD. The center provides clinical assessment, and psychological and educational interventions for people with autism (Keller et al., 2020). The Regional Reference Center for Autism – a structure of the Abruzzo Region Health System – performs diagnostic, clinical, and consulting activities and provides treatments for individuals with ASD. Most of the patients were referred by the general psychiatrist for an ASD assessment and came to either center for the first time or returned for a follow-up evaluation. All the diagnoses were made according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) (APA, 2013) criteria considering clinical anamnesis, clinical interview, cognitive assessment with WAIS-IV (Orsini and Pezzuti, 2013), diagnostic evaluation with ADI-r (Rutter et al., 2003) and ADOS module 4 (Lord et al., 2002) or RAADS (Ritvo et al., 2011), following structured diagnostic pathway (multistep network model, Keller et al., 2020). Of the entire sample, 169 people received diagnosis of ASD with level 1 (male = 75%, Medu = 12.4, S.D. = 2.64; female = 25%, Medu = 13.6, S.D. = 2.91), 60 with ASD level 2 (male = 75%, Medu = 10.9, S.D. = 2.18; female = 25%, Medu = 11.3, S.D. = 2.47) and 39 with ASD level 3 (male = 79%, Medu = 10.9, S.D. = 1.96; female = 21%, Medu = 11.5, S.D. = 1.60). To be included in the study, all the patients received a formal clinical diagnosis of ASD according to DSM-5 (APA, 2013) criteria. People with comorbid psychopathology (n = 42) were included only if they were either in remission or of minimal impact on daily functioning. In total, 3.9% with ASD level 1 and comorbid depressive disorders (male = 3%, female = 0.9%), 3.49% with ASD level 1 and personality disorders (male = 2.18%, female = 1.31%), 2.18% with ASD level 1 and specific learning disorders (male = 1.31%, female = 0.87%), 1.31% people with ASD level 1 (male = 0.43%, female = 0.86%) and 0.43% males with ASD level 2 and obsessive-compulsive disorder, 1.31% with ASD level 1 and epilepsy (male = 0.87%, female = 0.43%), 1.31% with ASD level 1 and anxiety disorder (male = 0.43%, female = 0.87%), 1.31% with ASD level 1 and schizophrenia (male = 0.87%, female = 0.43%), 0.87% with ASD level 1 and attention-deficit/hyperactivity disorder (male = 0.43%, female = 0.43%), 0.87% with ASD level 1 and developmental coordination disorder (male = 0.43%, female = 0.43%), 0.43% females with ASD level 1 and Turner syndrome, 0.43% males with ASD level 2 and Tourette syndrome, 0.43% with ASD level 1 and gender dysphoria were included.

In total, 39 participants with level 3 and two participants with level 2 were excluded from the original sample because they were not suitable for a verbal cognitive evaluation with WAIS-IV since their communication was through gestures or other alternative communication systems.

All demographic variables and characteristics of the final sample are presented in Table 1.

Measures
Data about cognitive abilities were collected using the WAIS-IV (Wechsler, 2013). The WAIS-IV is used to assess the intellectual profile of people between 16 and 90 years old. It is composed of four scores and a general intelligence index. The four indexes are VCI, PRI, WMI, and PSI. Every index is composed of two or three subtests that are required to obtain the total IQ score. The ten core subtests are Vocabulary, Information, Similarities, Digit Span, Arithmetic, Block Design, Matrix Reasoning, Visual Puzzles, Coding, and Symbol Search. It also contains five additional subtests: Comprehension, Letter–Letter-number sequencing, Figure Weights, Picture Completion, and Cancellation. In our sample, we used the ten-core subtests for all ASD people and levels. We calculated the subtest scores, the indexes’ scores, and the full-scale IQ index. Every raw score was corrected with Italian standardization scores of the WAIS-IV (Orsini and Pezzuti, 2013).

The WAIS-IV and the entire psychological evaluation were administered by a licensed psychologist in a large and bright room in one session from 45 min to 1.5 hours.

The structure of the WAIS-IV and its indexes and subtests is represented in Table 2.

The age of each participant was calculated at the moment of the WAIS-IV administration and expressed in integers.

Level of autism was classified into three different levels as stated in DSM-5 (APA, 2013), so that level 1 was the less severe while level 3 was the most severe. The level of severity was assessed through clinical interviews made by two independent psychologists and a psychiatrist with participants and caregivers. Eventually, in a final reunion, the entire professional team discussed and agreed to one of the three levels of support required by the person.

Years of education were collected considering each school cycle years entirely completed. Any interrupted instruction years were not added to the number. Thus, considering the Italian compulsory education system, 5 years were assigned if a person completed the first school cycle. Other 3 years were given if a person completed the second school cycle. Finally, 5 years were considered if a person completed the last compulsory education cycle. Moreover, 3 to 5 years of additional educational years were given if a person completed a bachelor's or a master’s degree.

Psychopathological comorbidity was considered a dichotomous variable in terms of the presence or absence of any disorder.

Data analysis

An analytical approach was used to better describe and understand the data collected. At first, descriptive and correlational analyses were run to explore data and the distribution of the variables across ASD levels and to determine whether there was a relationship between the variables of interest. A moderate association between variables represents one of the conditions for exploring cause–effect phenomena through in-depth subsequent analysis.

Indeed, to better understand the effects of socio-demographic and ASD-related variables on cognitive performance indexes, linear regression models were used to analyze the impact of age, education, ASD level, sex, and comorbidity on WAIS-IV indexes. Linear regression is a predictive analysis used to determine if a set of predictor variables (independent variables) predict an outcome (dependent variables). Through analysis of the variance test, we evaluated an ‘overall’ effect considering the differences between means. Instead, the p-value for each mean in the regression models was used to easily understand which mean is different from the reference one.

Moreover, in a cascade approach model, we performed a more in-depth analysis considering each index as a dependent variable and socio-demographic and ASD-related variables as covariates. For the subsequent analyses, we performed a multivariate analysis of covariance (MANCOVA) to assess for statistical differences on multiple continuous dependent variables – the four WAIS-IV indexes – by two independent grouping variables, while controlling for one or more variables called the covariates. Through MANCOVA we created a model with four dependent variables (the four WAIS-IV indexes), sex, ASD level, and comorbidity as independent variables and age and education as covariates. Eventually, we repeated the same analysis using each index’s subtests as the dependent variables, sex, ASD level, and comorbidity as independent variables and age and education as covariates.

Likewise, consistent with the third aim of the research, we wanted to discriminate among ASD severity levels. The area under the curve (AUC) and receiver operating characteristics (ROC) (Metz, 1978; Zweig and Campbell, 1993) were used to inspect the performance of the two ASD-level groups on WAIS-IV composite indexes. ROC–AUC reveals how much the five WAIS-IV composite scores are capable of distinguishing between ASD severity levels. The higher the AUC, the better the model is at distinguishing between participants with 1 and 2 severity levels. A ROC is a plot of the true-positive rate (sensitivity) v. false-positive rate (1-specificity) associated with every possible cut-off value for a measure. The AUC is a measure of diagnostic accuracy and predictive validity that can be used to compare the predictive value of different measures. The AUC can range between 0.5 (random discrimination) and 1 (perfect discrimination)

For the analysis, we used R Studio (R Studio Team, 2020) and Jamovi (The Jamovi Project, 2021) software.

Results

For statistical analysis, two adults with level 2 and 39 adults with level 3 were excluded because they could not be assessed with the WAIS-IV. So, the final sample was composed of 229 people of levels 1 and 2. The descriptive statistics of the sample and the four indexes are presented in Table 3. For a better understanding of data distribution across the levels and indexes, we presented histograms with a density of the FSIQ and the four indexes in Fig. 1.

In simple correlation analysis (see Table 4), age was significantly correlated with FSIQ (r = 0.300, p < 0.001), VCI (r = 0.323, p = 0.01), PRI (r = 0.214, p = 0.001), WMI (r = 0.247, p < 0.001) and PSI (r = 0.235, p < 0.001). A relevant result was the absence of significance between block design and age (r = 0.084, p = 0.207). A similar result was found between Arithmetic and age (r = 0.206; p = 0.002). Level of education was significantly correlated with FSIQ (r = 0.376, p < 0.001), while the stronger association was only with the VCI (r = 0.264, p < 0.001) and its subtests, Similarities (r = 0.346, p < 0.001), Vocabulary (r = 0.387, p < 0.001) and Information (r = 0.366, p < 0.001). Although no significant correlation between the level of education and WMI was found, Arithmetic was moderately correlated with the level of education (r = 0.301; p < 0.001).

All the associations between the main indexes and subtests were significant ( p < 0.001).

In linear regression models, we considered the joint effects of sex, level of education, level of autism, age, and comorbidity on FSIQ. In model 1, age (β = 0.371; t = 2.779; p = 0.006), level of autism (β = −35.205; t = −12.636; p < 0.001) and level of education (β = 1.530; t = 3.268; p < 0.001) were significant, suggesting that the higher the age, the level of autism and education, the better the FSIQ score. Model 1 explained 54.3% of the variance in FSQI scores (R2 adjusted = 0.512, F(4, 224) = 60.9, p < 0.001). No significant effects of comorbidity were found on FSIQ (β = 0.479; t = 0.153; p = 0.87).

Using multivariate multiple regression models with MANCOVA we tested different hypotheses. In model 2 we considered the joint effects of the previous model independent variables separately on the four indexes (VCI, PRI, WMI, PSI). Sex (F = 8.23; p < 0.001), age (F = 4.54; p = 0.002), level of education (F = 3.53; p = 0.008) and level of autism (F = 63.80; p < 0.001) have a significant impact on the four indexes when considering them together. No significant effects were found considering the joint effects of sex and level of autism on the four indexes (F = 1.95; p = 0.103) nor of comorbidities (F = 1.77; p = 0.135). Therefore, model 2 suggests that male patients perform better than females and the higher the level of education and age the better the four indexes’ scores. Indeed, considering the direct effect of the variables on every single index we found that the effect of sex was statistically significant on VCI (F = 4.429; p = 0.036) and PSI (F = 10.835; p = 0.001) and remained significant when the joint effect with level is considered on PSI (F = 6.788; p = 0.010). Education has a statistically significant effect on VCI (F = 12.374; p ⩽ 0.001) and WMI (F = 8.288; p = 0.004).

In the following multivariate multiple regression models, we evaluated the effects of sex, age, education, autism levels, and comorbidities on the core subtests of the four indexes. Digit Span and Arithmetic were considered as the core subtests of WMI. The results highlighted significant effect of level of autism (F = 73.036; p < 0.001), age (F = 3.832; p = 0.023) and education (F = 4.244; p = 0.016) on both subtests. No effects of comorbidities were found on WMI subtests (F = 0.121; p = 0.886).

Considering the core subtests of VCI, sex (F = 2.859; p = 0.038), level of education (F = 4.822; p = 0.003), level of autism (F = 73.258; p < 0.001) and age (F = 5.932; p < 0.001) had a statistically significant impact on Similarities, Vocabulary and Information. If we look at the univariate tests’ results sex has a significant impact only on Vocabulary (F = 7.337; p = 0.007) with no significance on Similarities and Information. No effects of comorbidities were found on VCI subtests (F = 0.623; p = 0.601).

Indeed, for the effects on Block Design, Matrix Reasoning, and Visual Puzzles, the level of autism was the only covariate with a strong impact on the three subtests (F = 44.375; p < 0.001). No other relevant results were found except for a small significant effect of sex and autism levels on VP (F = 4.433; p = 0.036).
The last model considered the effects of variables on Symbol Search and Coding and revealed a significant effect of sex (F = 5.21; p = 0.006), level of autism (F = 60.29; p < 0.001), and the interaction between sex and level of autism (F = 3.22; p = 0.042) on the two subtests. However, the effect of the variables isolated on each subtest age has a statistically significant impact on Symbol Search. 

ROC results are presented in Table 5. According to the previous analysis, sex was statistically different on several indexes and subtests, and because of the small female sample size, we decided to treat males and females separately. In Table 5 we used ROC on female (n = 57) and male (n = 172) samples. Different cut-points were found to be discriminative between levels 1 and 2 considering FSIQ. Each index differed statistically significantly from the chance level (α = 0.05).

In the female sample, a score of 69 differentiates between levels while a range varying from 65 to 69 scores can distinguish between males with different autism levels. VCI distinguishes between levels 1 and 2 at a score of 74 in female participants. Whereas, in male participants, the clinical range to consider varies from 67 to 76. PRI’s best score for a female sample is 79 while for the male sample, a score of 77 is the best compromise considering sensitivity and specificity. Regarding WMI, a cut-point of 69 resulted in a strong parameter for distinguishing level 1 and 2 autism in women. For the male population, an adequate cut-point is 72 with good sensitivity and specificity. Finally, for the PSI, in the female sample, 81 was a good cut-point, while for the male sample, the good cut-point was 70.

Discussion

Limited researchers focused on an in-depth study of the cognitive profile of adults with autism in the international context and no research in the Italian context (Fombonne, 2005; Wilson et al., 2016). To our knowledge, the majority of authors focused on the cognitive and social performances of children or adolescents with ASD (Bodner et al., 2014). Several studies focused on comparing the cognitive performance of adults with ASD with TD or HFA with AS and TD (Holdnack et al., 2011). None of them explored the effect of socio-demographic variables on the cognitive performances of people with ASD. So, in our research, we explored the cognitive profile of adults with ASD who reached a clinical diagnosis. After exploring data with descriptive analyses, we performed a correlation of Full Scale, Primary Index Scales, and main subtest and socio-demographic variables. The results showed that FSIQ, PRI, WMI, and PSI moderately correlate with the age of the participants. More specifically, it is supposed that the level of education has a significant impact on cognitive skills measured by WAIS-IV indexes (Ceci, 1991; Baltes et al., 1998; Schaie and Willis, 2010; Pezzuti et al., 2019; Borella et al., 2020). Instead, an interesting result is the almost independence of the subtest Block Design from age and education which can be considered as a culturally and age-independent subtest in our sample.

Subsequently, we used a cascade approach, analyzing at first the Full-Scale Index, then the four fundamental indexes, and eventually the subtests that form the four main indexes. The decision for this choice was made to reduce the impact of two errors: the errors made during the transformation of the weighted scores into composite scores and when the difference between the indexes or the subtests was such as to invalidate the score of the index itself. In the first linear regression model, we evaluated the impact of age, level of education, sex, and level of autism on the FSIQ. The results showed a high level of significance for both age and education, indicating that each score in the FSIQ is correlated with an increase of 0.37 years and, for each year of education there is an increase of approximately 1.5 points in the FSIQ. These results are in line with a study on TD by Tommasi et al. (2015) that evidenced an average increase of 1.9 IQ points in the IQ global composite score per year of education. Contrary to our expectations and previous results that evidenced autistic females' disadvantage in IQ scores compared to autistic males, no sex effects were found on the FSIQ score in our sample. As previously mentioned, few conclusions can be assumed only by examining FSIQ scores as they include different information about broader cognitive abilities.

Therefore, in model 2 we ran a MANCOVA using the four indexes as dependent variables, sex and severity levels as factors, and age and education as covariates. The results showed a statistically significant difference in all the variables except when the interaction between sex and level of autism is considered. Looking deeper at results and the impact of the variables on indexes, results highlight a significant sex difference in Verbal Comprehension and Processing Speed indexes in the female participants which perform better than the male peers. This latter result is not surprising since even TD female adults outperformed males in processing speed tasks (Daseking et al., 2017). However, unexpectedly, and never outlined before, female autistic adults had better performances in vocabulary compared to autistic males. Although these results are surprising and new, further studies need to be conducted to counterbalance the number of female and male ASD participants. The effect of female advantage on PSI remains significant when the interaction with ASD level is considered. Indeed, the performance of female participants on PSI is better both in ASD levels 1 and 2. Another not surprising result is the effect of education on the Verbal Comprehension index suggesting that people with higher education perform better in verbal acquired knowledge and verbal reasoning, as previous literature pointed out (Tommasi et al., 2015). However, the effects of education on Working Memory are partly new and remain significant when both subtests are considered for the analysis. However, further studies need to be conducted to better understand the direction of this effect. It can be postulated that years of education contribute to better Digit Span and Arithmetic performances as better WMI performances increase the likelihood of a higher level of education. Unpredictably no statistical effect of sex on WM was found, revealing a similar way for both male and female participants to perform in this cognitive domain. This result is in contrast to a recent Italian study on TD by Pezzuti et al. (2020) in which there was an outperformance of men in WMI composite scores and its Arithmetic subtest. The absence of effects of sex on this index in our autistic sample could be interpreted in light of extreme male brain theory (Baron-Cohen, 2002) whereby autism can be considered as an extreme of the normal male profile.

In model 4 subtests of the VCI (Similarities, Vocabulary, and Information) are considered and the results showed a significant effect on all the variables except when the interaction between sex and ASD level is taken into account. Looking deeper at the univariate analyses, the significant effects of education, age, and level of autism on individual subtests are confirmed on each subtest. The literature supports these findings, showing that the level of education is a predictor of greater verbal competence (Abad et al., 2015). However, the previous sex differences found considering the VCI composite scores disappeared when each subtest was considered for the analysis, except for Vocabulary. Even this result is in contrast to previous research (Longman et al., 2007; Irwing, 2012; Daseking et al., 2017) that outlined the superiority of men with TD in the Verbal Comprehension Index. Conversely, in our sample females with ASD outperformed males with ASD when the Vocabulary subtest was considered in the analysis. However, this difference is considered statistically significant only at ASD level 1, no sex differences in VCI subtests were detected when ASD level 2 is considered.

In model 5 we used the subtests Block Design, Matrix Reasoning, and Visual Puzzles as dependent variables. The results showed only a significant effect of the level of ASD on the subtests considered. The superiority of males with TD in PRI composite score (Longman et al., 2007; Irwing, 2012; Daseking et al., 2017) was not confirmed in our autistic sample, indicating that subtests of PRI are more sensitive to ASD severity level in our sample.

In model 6, Symbol Search and Coding were used as dependent variables. The results revealed a statistically significant effect of sex and levels of autism on both subtests, confirming the previous results when the PSI composite score was analyzed. Even when the joint effect of sex and level of autism is controlled, the result remains statistically significant on each subtest. This result is in line with the previous studies on TD considering the female superiority in the Processing Speed Index (Pezzuti et al., 2020); hence the same pattern seems to occur in the ASD population.

Using WAIS-IV main indexes or subtest cut-off scores to better discriminate between autism levels’ can be controversial but useful for clinicians who must describe one person's functioning according to DSM-5 (APA, 2013) classification. For the Full-Scale Index, the best cut-points revealed were 69 for females and 65 for males using Youden’s indexes. In VCI, the optimal cutpoints were 74 and 69 for females and males, respectively; regarding the PRI, the best cut-points were 79 for females and 73 for males; in WMI 69 for females and 72 for males; finally, for PSI the optimal cut-points were 81 for females and 70 for males.

Although all these predictive results can help clinicians to better discriminate between different levels of severity, a test cannot replace diagnostic assessment by experienced clinicians. However, cut-off scores are taken together with the previous findings about the almost independence of PRI from age, level of education and sex can partly direct clinical evaluation to visuospatial abilities when assessing people with ASD across levels.

To sum up, some authors evidenced an underestimating effect of the cognitive abilities of ASD people when assessed with WAIS-IV compared to RPM (Dawson et al., 2007; Hayashi et al., 2008; Soulières et al., 2011). However, this phenomenon seems to be better applied to ASD people with cognitive impairment and not to AS (Bölte et al., 2009; Holdnack et al., 2011) or average cognitive abilities. So, cognitive impairment should be of concern when selecting any assessment tool to use with people with ASD and when interpreting the results of their achievement on that measure. Alongside cognitive impairment, language delay plays a significant impact on IQ outcome, as Bodner et al. (2014) evidenced in their study that resulted in better WAIS-IV IQ than RPM scores in verbally able adults. Thus, multiple factors need to be considered before assessing people with ASD (context, situation, abilities assessed, different methods) prioritizing a multi-method multi-informant approach. Therefore, predicting the academic or adaptive functioning of people with ASD across the lifespan based on cognitive assessment tools should be done with caution since neither the Wechsler nor the RPM fully gathers all the information needed to assess cognitive functioning in people with ASD.

Limitations and directions for future research

A possible limitation of the study is the small number of female participants compared to the male participants, which may preclude the generalization of results. Besides, the reduced female ASD sample and the results of no sex differences on IQ general composite scores can be partly due to the female sample size. However, the sample was composed of different numbers of males and females according to the ASD prevalence.

Only the presence or absence of comorbidities in findings has been investigated in the research. Although a limited number of participants had clinical diagnoses that could have a strong effect on WAIS-IV subtests, such as Psychotic Disorders or ADHD, further studies are needed to evaluate the single effect of comorbidities on outcomes.

Availability of data and materials

The anonymized datasets analyzed in the current study are available from the corresponding author on request.

Acknowledgements.

We thank all the people who took part in this study. We appreciate the participation of autistic participants and their relatives who, with their interest and dedication, make autism research possible.

Financial support.

No financial support was received for the research.

Conflict of interest.

No conflict of interest was reported by the authors.

Ethical standards.

All procedures performed in studies involving human participants were by the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

References

1. Abad F, Sorrel M, Román F and Colom R (2015) The relationships between WAIS-IV factor index scores and educational level: a bifactor model approach. Psychological Assessment 28, 987–1000.

2. American Psychiatric Association (2013) Diagnostic and Statistical Manual of Mental Disorders, 5th Edn. Arlington, VA: Author.

3. Attwood T (2007) The Complete Guide to Asperger’s Syndrome. London: Jessica Kingsley Publishers.

4. Baltes PB, Lindenberger U and Staudinger UM (1998) Life span theory in developmental psychology. In Damon W and Lerner RM (eds), Handbook of Child Psychology: Vol. 1. Theoretical Models of Human Development, 5th Edn. Hoboken, NJ: Wiley, pp. 1029–1143.

5. Baron-Cohen S (2002) The extreme male brain theory of autism. Trends in Cognitive Sciences 6, 248–254.

6. Baxter AJ, Brugha TS, Erskine HE, Scheurer RW, Vos T and Scott JG (2015) The epidemiology and global burden of autism spectrum disorders. Psychological Medicine 45, 601–613.

7. Bodner KE, Williams DL, Engelhardt CR and Minshew NJ (2014) A comparison of measures for assessing the level and nature of intelligence in verbal children and adults with autism spectrum disorder. Research in Autism Spectrum Disorders 8, 1434–1442.

8. Bölte S, Dziobek I and Poustka F (2009) Brief report: the level and nature of autistic intelligence revisited. Journal of Autism and Developmental Disorders 39, 678–682.

9. Borella E, Pezzuti L, De Beni R and Cornoldi C (2020) Intelligence and working memory: evidence from administering the WAIS-IV to Italian adults and elderly. Psychological Research 84, 1622–1634.

10. Ceci SJ (1991) How much does schooling influence general intelligence and its cognitive components? A reassessment of the evidence. Developmental Psychology 27, 703–722.

11. Ceci SJ and Williams WM (1997) Schooling, intelligence, and income. American Psychologist 52, 1051.

12. Charman T, Pickles A, Simonoff E, Chandler S, Loucas T, and Baird G (2011) IQ in children with autism spectrum disorders: data from the Special Needs and Autism Project (SNAP). Psychological Medicine 41, 619–627.

13. Christensen DL, Baio J, Van Naarden Braun K, Bilder D, Charles J, Constantino JN, Daniels J, Durkin MS, Fitzgerald RT, Kurzius-Spencer M, Lee LC, Pettygrove S, Robinson C, Schulz E, Wells C, Wingate MS, Zahorodny W, Yeargin-Allsopp M and Centers for Disease Control and Prevention (CDC) (2016) Prevalence and characteristics of autism spectrum disorder among children aged 8 years – autism and developmental disabilities monitoring network, 11 sites, United States, 2012. MMWR Surveillance Summaries 65, 1–23.

For more information:1950477648nn@gmail.com