ASD Validity

Findings for ASD global and regional brain sizes have been varied. Riddle et al. (2016) found that 443 individuals with ASD had a 2.17 % increase in gray matter compared to 390 typically developing (TD) individuals, but found no group difference in white matter. However, when Riddle et al. (2016) compared matched subsamples of 300 ASD and TD children, no differences between TD and ASD white matter or gray matter were found. Vasa et al. (2012) reported that only 8 of 73 individuals diagnosed with ASD had any atypical brain features. Lenroot and Yeung (2013) reported that the majority of individuals diagnosed with ASD had typical brain and head size. However, in other studies, approximately 10 to 24 % of individuals diagnosed with ASD had persisting macrocephaly (Gillberg and de Souza 2002; Lainhart 2015; Nebel et al. 2015; Sacco et al. 2015; Tammimies et al. 2015). In addition, a number of studies have reported that 3 to 15 % of individuals with ASD had persisting microcephaly (Gillberg and de Souza 2002; Nebel et al. 2015; Roullet et al. 2013; Stevens et al. 2013).

Although Riddle et al. (2016) found no regional brain differences in 443 individuals with ASD compared to 390 TD individuals, Lefebvre et al. (2015) noted that many studies had found significantly smaller corpus callosum in ASD than TD controls. By contrast, Lefebvre et al. (2015) found no difference in corpus callosum size in their sample of 694 individuals with ASD compared to TD controls. Studies of the amygdala in ASD have reported significantly “increased, decreased and preserved volumes” of the amygdala (Bellani et al. 2013, p. 3). Similarly, Stigler et al. (2011) noted that studies have reported increased, decreased, and typical (preserved) volumes for the fusiform gyrus in ASD.

D’Mello et al. (2015) reported that ASD was characterized by reduced gray matter in the cerebellum in lobule VII. However, a consensus paper on the cerebellum in ASD (Fatemi et al. 2012) concluded that only a subgroup of ASD had atypical cerebellar anatomy, and this was a smaller cerebellar vermis volume and fewer Purkinje cells. Studies of the brainstem in ASD have reported both typical and atypically smaller volumes of brainstem gray matter (Jou et al. 2013). Despite evidence for abnormal auditory brainstem response (ABR) in ASD (Rosenhall et al. 2003; Lukose et al. 2015), Jou et al. (2013) concluded that ASD brainstem studies were conflicting and inconclusive.

To date, no unitary atypical brain size or volume has been found for ASD, and no unitary atypical ASD regional brain structure has been found. Equally important, varied ASD electrophysiology findings (Billeci et al. 2013) and varied ASD molecular neurochemistry findings (Zürcher et al. 2015) stand against the idea that there is any shared single ASD brain impairment in electrophysiology or molecular neurochemistry.

Of course, research has not uncovered the full complexity of human brain development, functions, and networks (Sporns and Betzel 2016). When much more is known about the emergence of regional functions and connectivity of the brain, it may be possible in the future to establish a unitary model of ASD brain impairment.

Hundreds of unitary models of ASD brain impairment have been theorized and studied. For example, Baron-Cohen et al. (2015) proposed that ASD stemmed from fetal steroidogenic abnormalities. Fishman et al. (2014) proposed that ASD resulted from atypical overconnectivity of the brain’s mirror neuron regions and theory of mind regions. By contrast, Khan et al. (2015) proposed that ASD resulted from underconnectivity in local brain regions when activated by a “functionally relevant task” (p. 1407). Mullins et al. (2016) posited that mutations in the FMR1 gene and TSC genes disrupt long-term depression (LTD) and long-term potentiation (LTP), and therefore, these mutations in ASD cause atypical LTP and LTD processes that impair the excitation-inhibition balance in brain development and function which determines ASD (Mullins et al. 2016).

Robertson et al. (2015) claimed that the “prime suspect” single cause for ASD was disrupted GABAergic signaling that impaired neurodevelopment and cortical computations. However, Estes and McAllister (2015) theorized that ASD was caused by atypical immune system factors that converged on the MEF2 and mTOR signaling hubs and thus disrupted the brain’s developmental synaptic function and plasticity. Irimia et al. (2014) theorized a basis for ASD brain impairment in atypically greater misregulation of nSR100-dependent microexons. Focusing on early visual attention in ASD, Klin et al. (2015) posited that the developmental failure of the reward-based interactional eye fixation to co-opt the brain’s innate reflexive eye fixation caused ASD.

Because these eight theories explain different aspects of ASD brain function, they are not necessarily mutually exclusive. However, there has been no attempt to synthesize any subset of these eight theories, or synthesize any subset of prior unitary ASD brain theories (Waterhouse 2008). Most importantly, no unitary brain impairment theory to date has been replicated to become a standard explanation of ASD brain disruption.

The theory with the most active support argues that ASD results from impaired brain connectivity that is likely to have been preceded by early brain overgrowth (Solso et al. 2015; Stoner et al. 2014; Venkataraman et al. 2015). Although the underconnectivity theory has been studied for more than 15 years (Just et al. 2004) and has many supporters (Anderson 2013; Dawson 2016; Ecker et al. 2015; Green et al. 2015; Shen et al. 2013), nonetheless, this theory has not become standard because there is as much evidence against the theory as there is for it.

Supporting the existence of early ASD brain overgrowth, Ecker et al. (2015) asserted that, on average, toddlers with ASD have “a larger brain volume than typically developing children” (p. 1), and Shen et al. (2013) and Anderson (2013) reported that ASD was characterized by atypical early brain overgrowth. Sussman et al. (2015) reported that early brain overgrowth was limited to males with ASD, and Chaste et al. (2013) found that the subgroup of ASD with an atypically larger head circumference expressed greater symptom severity and lower IQ.

Counter to the early brain overgrowth model, though, Chaste et al. (2013), Raznahan et al. (2013), and Cederlund et al. (2014) all concluded that very few children with ASD had early brain overgrowth, and Raznahan et al. (2013) reported that larger ASD head circumference was not found when local TD controls were used. Lainhart (2015) reviewed research and noted that only “a very small subgroup of ASD children” (p. 79) had larger brain volumes. In a meta-analysis, Sacco et al. (2015) found only 5.7 % of individuals with ASD had macrocephaly and just 9.1 % had brain overgrowth. Tammimies et al. (2015) also found that only 63 of 258 children with ASD had macrocephaly. Also countering the ASD early brain overgrowth model, 3 to 15 % of individuals diagnosed with ASD have been born with microcephalic brains (Nebel et al. 2015; Roullet et al. 2013; Stevens et al. 2013). Most importantly, the generalizability of the ASD early brain overgrowth model is limited by the evidence that the majority of individuals diagnosed with ASD have typically developing head and brain growth (Lenroot and Yeung 2013; Nebel et al. 2015; Riddle et al. 2016; Vasa et al. 2012).

Supporting the malconnectivity component of the currently dominant model, Dawson (2016) asserted that “long-range connections between different brain regions are weaker in people with ASD” (para. 4). Venkataraman et al. (2015) determined that ASD brain malconnectivity occurred in two networks: a language network including temporal lobe areas and a “social-person” network including temporal and parietal areas. Venkataraman et al. (2015), however, found no impairment in ASD frontal lobe connectivity. Conversely, Kitzbichler et al. (2015) reported that ASD malconnectivity was centered in frontal lobe-linked connections.

Counter to both Kitzbichler et al. (2015) and Venkataraman et al. (2015), though, Redcay et al. (2013) reported finding no evidence for atypical connectivity in ASD. Tyszka et al. (2014) also stated that their ASD sample showed, “no evidence at all for altered connectivity at the whole-brain level” (pp. 7–8). Kirkovski et al. (2015) found no differences between high functioning individuals with ASD and typical controls in white matter in major tract bundles determined by any method: fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), or axial diffusivity (AD). Koldewyn et al. (2014) reported finding no general impaired white matter in ASD. Lefebvre et al. (2015) found no differences between 694 individuals diagnosed with ASD and typical controls in the largest white matter tract in the brain—the corpus callosum. The researchers questioned whether any regional ASD malconnectivity model could be meaningful because most of the brain is involved in connectivity (Lefebvre et al. 2015).

Another problem for standardizing the overgrowth-malconnectivity model is that there are so many different versions of this model (Fishman et al. 2014; Kennedy et al. 2015; Khan et al. 2015; Kitzbichler et al. 2015; Venkataraman et al. 2015). As noted by Kennedy et al. (2015), “the connectivity hypothesis has been vague ever since its inception… slowly morphing from a theory about underconnectivity in ASD, to one about distal underconnectivity paired with local overconnectivity, to one about atypical connectivity in either direction (or both)” (p. 81).

Despite intense research efforts, the malconnectivity model has not been successfully replicated (Ecker et al. 2015; Kennedy et al. 2015; Khan et al. 2015; Lefebvre et al. 2015), and neither has any of the myriad other unitary ASD brain impairment models. Moreover, none of these models has accounted for the variation in ASD global and regional brain structures, or the variation in ASD electrophysiology findings (Billeci et al. 2013), or the variation in neurochemistry findings (Zürcher et al. 2015).

Many studies have looked for the biological validity of individual ASD diagnostic and associated nondiagnostic symptoms (Anderson et al. 2014; Boucher 2011; Brunsdon and Happé 2014; Chen et al. 2015; Harrop et al. 2013; Hormozdiari et al. 2015; Jason et al. 2015; Pina-Camacho et al. 2012; Shuster et al. 2014).

Syndromic ASD is found with a known genetic or environmental syndrome, such as fragile X syndrome (FXS) or valproate syndrome. Syndromic ASD occurs with Mendelian single gene syndromes, CNV syndromes, and environmental syndromes. Although syndromic ASD has been estimated to account for only 5 to 25 % of all cases of ASD (Adviento et al. 2014; de la Torre-Ubieta et al. 2016; Jeste and Geschwind 2014), more forms of syndromic ASD continue to be identified (Adviento et al. 2014; Richards et al. 2015; Roullet et al. 2013; Smile et al. 2013; Yu and Berry-Kravis 2014; Yuen et al. 2015). In addition, syndromic and idiopathic ASD brain impairments have been reported that are similar (Adviento et al. 2014; Blackmon 2015; D’Angelo et al. 2015; Guilmatre et al. 2014; Klusek et al. 2015; Sala et al. 2015).

The key question is whether syndromic ASD diagnostic symptoms are caused by an independent brain impairment unique to ASD. For example, when ASD occurs with cerebral palsy (Smile et al. 2013), are ASD diagnostic symptoms caused by an ASD-specific brain connectivity problem? Or are ASD diagnostic symptoms caused by characteristic cerebral palsy gray matter injuries, brain malformations, and focal vascular insults? If syndromic ASD symptoms do not result from an independent ASD-specific brain impairment, then ASD diagnostic symptoms in syndromic ASD must result from widely varied brain impairments.

Peters et al. (2013) and Tye and Bolton (2013) argued for an independent unique ASD brain impairment in syndromic ASD. Peters et al. (2013) asserted that in syndromic ASD with tuberous sclerosis complex (TSC), the TSC brain tubers and TSC malorganization of the brain caused TSC symptoms, but that brain underconnectivity was independently and uniquely causal for ASD alone. The researchers found that underconnectivity occurred in syndromic ASD with TSC, and in idiopathic ASD, but not in TSC alone. Similarly, Lainhart (2015) claimed that syndromic ASD with TSC and idiopathic ASD both expressed the ASD-specific reduced long-range connectivity with increased short-range connectivity.

However, Jeste and Geschwind (2014), Hall et al. (2010), and Adviento et al. (2014) proposed that syndromic ASD arose from the syndrome’s brain impairments. Specifically, for the case of syndromic ASD with TSC, Jeste and Geschwind (2014) argued that the TSC brain tubers and TSC malorganization of the brain that caused TSC symptoms also simultaneously caused ASD symptoms. Jeste and Geschwind (2014) argued that both ASD and TSC symptoms resulted from TSC tubers that occurred in the temporal, frontal, and occipital cortex and in the cerebellum, because these tubers cause many brain impairments, including disorganized axonal tracts, increased axonal growth, abnormal myelination, aberrant synapse formation, and aberrant white matter organization. Similarly, for syndromic ASD with FXS, Hall et al. (2010) argued that an individual with FXS who expressed ASD symptoms did not have two disorders, but had one genetic disorder, FXS, that caused one brain impairment, the lack of fragile X mental retardation protein (FMRP) that, in turn, caused both FXS symptoms and ASD symptoms. Gabis and Pomeroy (2014) also supported the single brain impairment model of syndromic ASD with evidence suggesting that ASD symptoms expressed in syndromic ASD likely resulted from a unitary “biological or neural pathway” (p. 297).

Some researchers have theorized that a network or networks of multiple ASD risk genes disrupt just a few impaired brain circuits or pathways in ASD (Chen et al. 2015; de la Torre-Ubieta et al. 2016; Ecker et al. 2015; Geschwind and State 2015; Parikshak et al. 2013; Ruggeri et al. 2014; Willsey et al. 2013). For example, de la Torre-Ubieta et al. (2016) stated “evidence from known mutations does suggest significant convergence in the pathways in which the mutations are found” (p. 349). Willsey et al. (2013) predicted that ASD risk genes together would yield, “a much smaller set of underlying pathophysiological mechanisms” (p. 1004), and Geschwind and State (2015) also proposed that ASD genetic risk factors would converge on only a few “specific molecular pathways or brain circuits” (p. 9).

There is evidence that ASD risk gene variants do converge on specific brain tissue in development (Parikshak et al. 2013; Willsey et al. 2013). Willsey et al. (2013) reported that nine ASD risk genes were expressed in one brain development layer at a shared time of mouse fetal brain growth. Parikshak et al. (2013) reported that ASD risk genes were linked to “laminae containing postmitotic neurons during early fetal development…(and) upper-layer glutamatergic neurons in adult cortex” (p. 1118). Ziats and Rennert (2016) lauded the findings of Willsey et al. (2013) and Parikshak et al. (2013) for providing evidence “that a few common mechanisms may ultimately relate the heterogeneous set of ASD candidate genes to one another” (p. 4).

However, there are limits to the explanatory power of the “multiple genes-few brain circuits” model because “the function of most known genes is not fully understood; the grouping of affected genes is often arbitrary; and the concepts of pathways and networks are based on biochemistry, which may not be appropriately capturing the complex scenarios of the true biological system” (Vissers et al. 2016, p. 5). Most of the nearly 800 identified ASD risk genes (Butler et al. 2015) are not yet well understood, and the full extent of polygenicity for ASD is not yet known. It could be that ASD polygenicity is as great as that found for schizophrenia (SCZ). Loh et al. (2015) reported SCZ was so polygenic that most of the human genome contained SCZ-linked loci, raising the concern that a future more powerful analysis could link SCZ to the entire human genome, making genetic analysis effectively uninformative.

Yet another difficulty for the ASD multiple genes-few brain circuits model is that there are many known functional gene pathway groups. For example, Wen et al. (2016) found ASD risk genes in pathways for cell signaling, metabolism, neuroactive ligand-receptor interaction, and nervous system development. Wen et al. (2016) noted yet another ASD genetics inference problem. The researchers reported that the MAPK signaling pathway and the calcium signaling pathway that were central to the ASD pathway network were integrated by “ASD genes that encode proteins functioning in multiple steps” (Wen et al. 2016, p. 16). Because these proteins affect multiple body systems and thus can cause problems throughout the body, Wen et al. (2016) concluded that ASD symptoms were likely to be caused “by underlying more pervasive processes that are not specific to ASD brain or behavior features” (p. 13).

Another concern is that some “multiple genes yielding few disrupted brain circuits” models have not addressed known variation in ASD brain impairments. For example, Parikshak et al. (2013) stated that gene variants disrupted laminae in fetal development that resulted in upper-layer glutamatergic neuron malconnectivity that was the malconnectivity that defined ASD. However, Parikshak et al. (2013) did not discuss the evidence that many with ASD have no brain malconnectivity (Redcay et al. 2013; Tyszka et al. 2014). Thus, the Parikshak et al. (2013) model cannot apply to ASD in general. Parikshak et al. (2013) also proposed that the chromatin remodeling ASD risk gene ARID1B caused the corpus callosum abnormalities that characterized ASD. Here again, Parikshak et al. (2013) did not discuss evidence that many with ASD have no corpus callosum abnormalities (Lefebvre et al. 2015). Consequently, the role of the ARID1B risk gene in ASD cannot be generalized.

Finally, models of multiple genes disrupting a few brain circuits have ignored ASD environmental risk factors. For example, de la Torre-Ubieta et al. (2016) stated that ASD genetic risk factors caused “deficits in neurogenesis, cell fate, neuronal migration and morphogenesis during fetal development and dysregulated synaptic function” (p. 349). This inventory does not include intracranial hemorrhage effects found in ASD diagnosed with extreme prematurity (Kuzniewicz et al. 2014), or the brain cysts and apoptosis of neurons infected with cytomegalovirus (CMV) found for ASD occurring with congenital CMV infection (Engman et al. 2015).

While evidence suggests that gene variants are the dominant cause of the pathophysiology of ASD symptoms, just as gene variants have been found to be the dominant cause for most human traits and disorders (Polderman et al. 2015), there are significant findings for many ASD environmental risk factors (Boukhris et al. 2015; Grabrucker 2013; Lyall et al. 2014; Maramara et al. 2014; Ornoy et al. 2015; Rossignol et al. 2014; Schieve et al. 2014). Moreover, heritability studies have variably calculated ASD environmental risk factor influence at 7 to 35 % (Tick et al. 2016), 5 to 44 % (Colvert et al. 2015), 50 % (Sandin et al. 2014), and 55 % (Hallmayer et al. 2011). Huguet et al. (2016) reviewed ASD findings and proposed a specific distribution of ASD risk factors: 49.8 % common inherited variants, 2.6 % rare inherited CNVs and SNVs, 6.6 % de novo SNVs, 2.9 % de novo CNVs, and 38.1 % environmental risk factors. Huguet et al. (2016) considered the de novo SNVs and CNVs to be environmental and thus 52.4 % of ASD had genetic causes, and 47.6 % had environmental causes. Most importantly, gene-environment interactions have been identified for ASD (LaSalle 2013; Jeste and Geschwind 2014; Volk et al. 2014).

The “multiple gene-few brain circuits” models are premature and insufficiently explanatory because ASD risk gene functions in networks are complex and not yet fully understood. There are hundreds of ASD risk genes, and there is evidence for gene-gene and gene-environment interaction in ASD etiology. It is also possible that ASD polygenicity is so extreme as to hinder specific causal inferences, and possible that ASD symptoms result from gene variants causing systemic processes that are “not specific to ASD” (Wen et al. 2016, p. 13).

Kanner (1943, 1951) asserted that autism social withdrawal always occurred with the obsessive desire for the preservation of sameness. Although Kanner’s claim of an invariant bond between these core symptoms provided construct validity for the infantile autism diagnosis, subsequent research has not supported Kanner’s claim. No invariant link between ASD social impairment and ASD RRBs has been discovered (Frazier et al. 2014; Harrop et al. 2013; Pina-Camacho et al. 2012; Shuster et al. 2014). Many who express ASD diagnostic social impairment do not express the need for the preservation of sameness or any of the RRBs (Brennan et al. 2015; Kim et al. 2014; Kulage et al. 2014; Mandy et al. 2011; McPartland et al. 2012; Ventola et al. 2006). For example, in comparing toddler diagnostic instruments, Ventola et al. (2006) found that when a diagnostic instrument required RRBs, a majority of the children who had clinical autism diagnoses did not meet DSM-IV-TR autistic disorder (AD) diagnostic criteria (American Psychiatric Association 2000). However, all clinically diagnosed children were correctly diagnosed with AD when assessed by instruments that did not require the expression of RRBs (Ventola et al. 2006).

In a study of 256 children with DSM-IV-TR pervasive developmental disorder-not otherwise specified (PDD-NOS), Mandy et al. (2011) found that 97 % of those with PDD-NOS expressed AD diagnostic social impairment, but only 3 % expressed AD RRBs. Comparing DSM-IV-TR and DSM-5 diagnostic categories, McPartland et al. (2012) reported that most individuals previously diagnosed with Asperger’s syndrome or PDD-NOS did not meet DSM-5 ASD diagnostic criteria, but would meet DSM-5 criteria if the requirement for RRBs was eliminated. Similarly, Kim et al. (2014) reported that nearly all individuals with a prior diagnosis of AD, Asperger’s disorder, or PDD-NOS diagnosis who did not meet DSM-5 ASD criteria did express ASD social impairment, but did not express two of the four RRBs and/or atypical responsiveness.

Individuals who express ASD diagnostic social impairment but no RRBs or one RRB are in diagnostic limbo (Happé et al. 2006; Hus et al. 2007; Lam et al. 2008; Mandy and Skuse 2008; Szatmari et al. 2006; Watt et al. 2008). These individuals might meet criteria for the new DSM-5 social communication disorder (SCD) diagnosis (Smith et al. 2015). However, Bishop (2014) claimed that the DSM-5 category of language disorder was flawed, and Norbury (2014) warned that the SCD diagnosis would confusingly overlap with the ASD diagnosis because many diagnosed with SCD express full ASD social impairment and also express one of the two required RRBs and/or atypical sensory responsiveness.

As reviewed in the section above, ASD social impairment and ASD RRBs do not have an invariant link, and many individuals exhibit ASD diagnostic social impairment but express none or only one of the RRBs. Consequently, Happé et al. (2006), Boucher (2011), Shuster et al. (2014), and Brunsdon and Happé (2014) all raised the concern that no brain impairment had been found that caused only the ASD social impairment and ASD RRBs to occur together, and no theory had explained why the core ASD diagnostic symptoms did occur together. Boucher (2011) and Shuster et al. (2014) argued for studying core symptoms separately, and Waterhouse (2013) recommended studying ASD social impairment alone.

Brunsdon and Happé (2014) argued that there was no unitary basis for ASD diagnostic symptoms because each ASD diagnostic symptom was linked to “different genes, neural patterns and cognitive components that influence distinct behavioral symptoms” (p. 27). The claim of different neural patterns for each symptom is supported by the evidence discussed above that each ASD diagnostic symptom has been found with a different set of varied brain impairments (Harrop et al. 2013; Hormozdiari et al. 2015; Jason et al. 2015; Pina-Camacho et al. 2012; Shuster et al. 2014; Zilbovicius et al. 2013). However, existing evidence argues that ASD social impairment and the RRBs are very unlikely to be generated by separate genes in one individual. Although studies of TD twins found social impairment and RRBs to be independently heritable (Robinson et al. 2011; Ronald et al. 2011), this is not true for syndromic ASD with FXS, RTT, TSC, the RASopathies, the Shankopathies, FAS, or ASD with extreme prematurity. In these and other cases of syndromic ASD, a specific gene mutation or prenatal event appears to cause both ASD social impairment and RRBs as well as other symptoms. In addition, individual SNPs, such as those that disrupt the Ras/MAPK pathway (Adviento et al. 2014), also appear to cause both core diagnostic symptoms in idiopathic ASD. Finally, where multiple risk genes combine to cause ASD, it has not yet been determined that one gene or set of genes yields social impairment while another gene or set of genes yields the RRBs and/or atypical sensory responsivity (Jiang et al. 2013; Lim et al. 2013; Murdoch and State 2013; Sanders et al. 2015).

ASD diagnostic symptoms are expressed together because ASD genetic and nongenetic risk factors cause brain impairments that yield both ASD diagnostic symptoms (Chen et al. 2015; Ornoy et al. 2015) and at rates above chance (Happé et al. 2006). However, ASD risk factors do not cause only ASD social impairment and RRBs to be expressed together at rates above chance. In fact, fewer than 5 % of individuals with ASD have been found to express ASD diagnostic social impairment and RRBs together without any non-ASD symptoms (Lundström et al. 2015b), and/or minor physical anomalies (MPAs) (Tammimies et al. 2015), and/or MCAs (Timonen-Soivio et al. 2015). More than 95 % of individuals with ASD express ASD diagnostic symptoms along with ADHD, ID, epilepsy, language impairment, motor dysfunctions, attention deficits, anxiety, varied medical conditions, and many other symptoms (Coleman and Gillberg 2012; Lundström et al. 2015b; Pine et al. 2008; Richards et al. 2015; Waterhouse 2013). Because the vast majority of those diagnosed with ASD also express one or more non-ASD symptoms, ASD lacks the construct validity that would be provided by diagnostic symptom co-expression in the absence of other symptoms.

Counter to this, however, Rutter (2014) argued that “there is a problem in defining autism on the basis of particular features without considering a broader pattern” (p. 55) of non-ASD symptoms found with ASD. Rutter (2014) argued that ADHD, ID, epilepsy, language impairment, and other nondiagnostic symptoms provided ASD with a validating pattern of ASD-specific nondiagnostic symptoms. However, this validating pattern of non-ASD symptoms is not one consistent pattern, but instead, non-ASD symptoms vary widely from one diagnosed individual to another. Moreover, these associated symptoms stand against the construct validity of a unique co-expression of ASD symptoms. Most importantly, ADHD, ID, epilepsy, and language impairment occurring with ASD symptoms are linked to varied brain impairments that stand against the biological validity of ASD (Gabis and Pomeroy 2014; Jeste and Geschwind 2014; Klusek et al. 2015; Peters et al. 2013).

ASD core diagnostic symptoms occur alone together without other nondiagnostic symptoms in vanishingly few individuals. Consequently, there is insufficient coverage for ASD construct validity based on unique ASD symptom co-expression. Some researchers have proposed to make ASD more homogeneous by refining ASD diagnostic criteria (Bishop et al. 2016; Sonuga-Barke 2016) and by developing more sensitive ASD diagnostic screening instruments (Bone et al. 2016). However, it is unlikely that the many nondiagnostic symptoms such as ADHD, ID, epilepsy, and language impairment that occur with ASD that stem from varied ASD brain impairments caused by varied ASD risk factors (Kida and Kato 2015) will be eliminated by refinement of the ASD criteria or refinement of ASD screening measures.

Many very young TD children express RRBs, including restricted interests, repetitive motor behaviors, change resistance, and/or atypical sensory responsiveness (Camarata 2014; Harrop et al. 2013; Van Hulle et al. 2012). In fact, Harrop et al. (2013) noted that young TD children and young children diagnosed with ASD express similar rates of RRBs. However, RRBs disappear in typically developing children after they “serve the purpose of developmental mastery” (Harrop et al. 2013, p. 3). Thus, RRBs in older children with ASD may be evidence for developmental delay or evidence of atypical limits to development (Camarata 2014; Harrop et al. 2013). However, only extremely shy but otherwise typically developing children show severe social withdrawal in that they “rarely initiate social contacts with available playmates, tend to withdraw from social interactions” (Coplan et al. 2013, p. 862).

Happé et al. (2006) found modest or weak correlations between ASD social impairment symptoms and RRBs in typically developing 7- and 8-year-old twins. Posserud et al. (2013) reported separate factors for social impairment and RRBs in a sample of 10,220 typically developing adolescents. Conversely, Constantino and Charman (2015) concluded that ASD “characteristic traits and symptoms…are as highly inter-related in the general population as they are in ASD syndromes” (p. 7), and asserted that ASD traits were continuous; therefore, any boundary between TD and ASD was arbitrary.

RRBs are commonly expressed in young typically developing children (Harrop et al. 2013), but ASD social withdrawal is extremely rare in typical development (Coplan et al. 2013). These findings and TD twin study findings (Robinson et al. 2011; Ronald et al. 2011) argue against a general invariant coupling of ASD social impairment and ASD RRBs in TD.

The DSM-III nosology (American Psychiatric Association 1980) triggered an increase in the comorbidity of diagnoses because DSM-III divided complex psychiatric phenotypes into fixed diagnostic categories with specific required symptoms. As a result, often more than one diagnosis was needed to cover the full range of an individual’s symptoms (Maj 2005).

Skokauskas and Frodl (2015) found moderately high levels of comorbidity of ASD and bipolar disorder (BPD). Levy et al. (2010) found that 39 % of those with ASD expressed anxiety and mood disorder symptoms, compared with only 4 % of controls. Pine et al. (2008) reported that 57 % of children with BPD, 38 % of children with major depressive disorder (MDD), and 25 % of children with anxiety disorder expressed ASD symptoms. Croen et al. (2015) reported that 54 % of adults with ASD were diagnosed with an additional psychiatric disorder, including anxiety (29 %), BPD (11 %), depression (26 %), SCZ (8 %), and OCD (8 %).

OCD and ASD symptoms have been reported to be comorbid in frequencies ranging from 1.5 to 81 % and OCD compulsions have parallels in ASD insistence on sameness and ASD repetitive behaviors (Meier et al. 2015; Stone and Chen 2015). Surprisingly, although OCD is often found with increased gray matter volumes in the caudate nuclei (Meier et al. 2015), the OCD-like insistence on sameness in ASD without OCD has been linked to every subcortical region except the caudate (Eisenberg et al. 2015).

Half of those diagnosed with SCZ have expressed ASD diagnostic social impairment (Matsuo et al. 2015), and social cognitive impairment was found to be the same in Asperger’s disorder and SCZ (Lugnegård et al. 2013), as well as the same in ASD and SCZ (Eack et al. 2013). Moreover, SCZ was diagnosed in 2.4 % of those with ASD (Kohane et al. 2012). Evidence also has indicated that if one parent was diagnosed with SCZ, there was an increased risk for having a child with ASD (Larsson et al. 2005).

Multiple complex developmental disorder (MCDD) is a psychosis prodrome disorder leading to overt psychosis or SCZ that is often found with ASD or PDD symptoms, as well as with panic, explosive emotional behaviors, magical thinking, easy confusability, and paranoid preoccupations (Ad-Dab’bagh and Greenfield 2001; de Bruin et al. 2007; Kyriakopoulos et al. 2015; Sprong et al. 2008; Oranje et al. 2013; Ziermans et al. 2009). MCDD is not a DSM-5 diagnosis, and the brain basis of MCDD remains unknown. Oranje et al. (2013) reported no deficits in P50 wave suppression and prepulse inhibition (PPI) of the startle reflex in MCDD with ASD suggesting that there were no SCZ-like sensory gating problems. Ziermans et al. (2009) also found no abnormalities in brain gray matter or white matter in MCDD with PDD.

The comorbidity of SCZ, BPD, MDD, OCD, MCDD, and anxiety symptoms in ASD (along with symptoms of ID, epilepsy, language impairment, motor dysfunctions, and others) stands against the construct validity of the ASD diagnosis by demonstrating that most complete ASD phenotypes are inadequately covered by the two core ASD diagnostic symptoms. One likely contributor to such high ASD comorbidity is that ASD shares risk genes with other disorders (Brandler and Sebat 2015).

Yirmiya and Charman (2010) observed that little is known about “the prodrome of ASDs” (p. 450), and Reeb-Sutherland and Fox (2015) noted that studies attempting to predict which infant siblings would go on to develop ASD “had little success reliably identifying behavioral markers during infancy that predict the later manifestation of ASD” (p. 390). Volkmar and Reichow (2014) suggested that one difficulty for early ASD diagnosis was “the broad range of severity and associated communicative and cognitive disability” (p. 11). Barbaro and Dissanayake (2013) reported that at 12 months, deficits in pointing, waving, imitation, eye contact, and name response were significant markers for an ASD diagnosis, but that by 24 months, only deficits in eye contact remained a reliable index of ASD.

Elsabbagh and Johnson (2016) reported that social impairment was not a key early characteristic of ASD, which instead included five features: (1) head lag when an infant is pulled up to sitting, (2) atypically high sensitivity to sensory experience, (3) trouble with consonants in earliest language, (4) atypically slower ability to shift away attention, and (5) general atypically lower level of activity. Similarly, in a study of young siblings of individuals with ASD, Sutera et al. (2007) reported that ASD diagnostic social and communicative skills were not predictive but that motor skills were predictive of later diagnostic outcome.

A consensus panel of ASD researchers reported there was no “single behavioral sign or a single developmental trajectory that is predictive of all diagnoses of ASD” (Zwaigenbaum et al. 2015, p. S37), and the panel asserted that ASD heterogeneity made it unlikely that any defining early ASD behavioral marker will ever be found.

As discussed in the first section of this paper, a dominant theory has argued that ASD results from early brain overgrowth with later impaired brain connectivity (Solso et al. 2015). Consequently, early brain overgrowth has been proposed as the primary biomarker for ASD (Anderson 2013; Ecker et al. 2015; Shen et al. 2013). However, as previously outlined, early brain overgrowth is rare in ASD. The meta-analysis of Sacco et al. (2015) found macrocephaly in less than 6 % of ASD, and also, as reported earlier, Chaste et al. (2013), Raznahan et al. (2013), and Cederlund et al. (2014) found no evidence that ASD was characterized by early brain overgrowth.

Jones and Klin (2013) argued that early decline in eye fixation would be the best single biomarker of ASD. The researchers claimed this biomarker reflected that a disrupted reward-based interactional eye fixation regulatory system was the core ASD brain impairment (Klin et al. 2015). Reeb-Sutherland and Fox (2015), however, suggested that atypical eyeblink conditioning was a possible single ASD biomarker, and Jeste et al. (2015) proposed that EEG patterns might be a possible biomarker for ASD.

Small and Pelphrey (2015) proposed, “Innate olfactory behaviors may provide a link between early emerging sensory motor behaviors and the social deficits that characterize ASD” (p. R675). Rozenkrantz et al. (2015) reported that longer time sniffing an unpleasant odor was 81 % accurate in differentiating children with ASD from typically developing children, and Rozenkrantz et al. (2015) and Small and Pelphrey (2015) suggested that a sniff test could serve as an effective single biomarker for ASD.

However, Varcin and Nelson (2016) argued that “The heterogeneity inherent to ASD necessitates… sets of markers, rather than a single marker” (p. 127). Glatt et al. (2012) identified a blood-based gene expression profile of 48 genes that reliably identified young children with ASD. Taylor et al. (2014) and Ruggeri et al. (2014) proposed large panels of biomarkers to include markers such as head circumference above the 97th percentile, long-range functional hypoconnectivity and short-range hyperconnectivity, ERP-measured speed of response to human faces, elevated blood serotonin (5-HT) levels, and autoantibodies against a range of brain antigens localized in GABAergic neurons. However, Varcin and Nelson (2016) concluded that “Biomarkers with sufficient sensitivity and specificity for clinical application are yet to be identified in ASD” (p. 124).

Ideally, clinicians use early symptoms to determine prognosis and assign individuals to treatment. However, ASD research has discovered many varied early behaviors and varied brain markers, and no consistent predictor pattern has been found. Panels of multiple biomarkers have been proposed to help net ASD heterogeneity.

Understanding the developmental course of ASD is crucially important for managing life care and planning treatment. However, ASD occurs with many varied developmental paths (Fountain et al. 2012; Lord et al. 2015), and there is no one consistent life outcome (Fein et al. 2013; Helles et al. 2015; Steinhausen et al. 2016). Variation in ASD developmental course has ranged from typical development in infancy that becomes atypical in early childhood, to marked infant impairment in social and cognitive skills that changes to become optimal adaptive functioning in adulthood (Anderson et al. 2014; Fein et al. 2013; Fountain et al. 2012; Helles et al. 2015; Howlin et al. 2013; Levy and Perry 2011; Lord et al. 2015; Yirmiya and Charman 2010).

The assumption that ASD was a unitary entity led to efforts to establish a single recurrence risk rate for ASD. Recurrence risk is the likelihood that a second child with ASD would be born in a family where a child has already been diagnosed with ASD. No consistent recurrence risk for ASD has been established, and rates have varied from 6 to 19 % (Grønborg et al. 2013; Ozonoff et al. 2011; Ronemus et al. 2014; Rosti et al. 2014; Werling and Geschwind 2013).

However, Matsunami et al. (2014) asserted that no single inheritance model for ASD could be correct, because many varied forms of genetic transmission occur with ASD, and Ronemus et al. (2014) noted that ASD genetic transmission recurrence risk must vary for families with one child with ASD (simplex) and families with more than one child with ASD (multiplex). An example of varied genetic transmission was reported by Jiang et al. (2013), who found multiple varied ASD inheritance patterns in just 16 families: 12 rare X-linked deleterious variants, 7 rare deleterious autosomal-dominant mutations, 13 deleterious missense mutations, and 15 de novo deleterious mutations. Another problem for recurrence rate determination was discovered by Grønborg et al. (2013), who found increased ASD risk in maternal half-siblings, indicating that environmental factors unique to the mother’s pregnancy history may contribute to ASD recurrence risk.

Researchers have identified many varied ASD genetic risk factors with varied transmission patterns, and because there are also many varied ASD environmental risk factors, no single recurrence risk can ever be determined for the ASD diagnosis.

The assumption that ASD was a unitary disorder led researchers to look for a unitary BAP in families of individuals diagnosed with ASD. The BAP is theorized to be the expression of attenuated ASD symptoms in family members. However, BAP symptoms have been found to be as heterogeneous as ASD symptoms and have included cognitive, language, and social skill impairments, as well as attenuated and nonattenuated psychiatric symptoms (Sucksmith et al. 2011).

Researchers have proposed that ASD includes biologically valid subgroups (Aldinger et al. 2015; Brandler and Sebat 2015; Ellegood et al. 2015; Georgiades et al. 2013; Williams et al. 2014; Whitehouse and Stanley 2013; Williams and Bowler 2014). Williams et al. (2014) predicted that ASD would eventually become a set of biologically valid subgroups defined “by genetic and/or neurological…findings” (p. 339), and Brandler and Sebat (2015) predicted that ASD would naturally disband into valid “quanta of separate genetic disorders” (p. 502).

Researchers have subtyped many aspects of ASD. There has been ASD gender subgrouping (Halladay et al. 2015; Ypma et al. 2016), ASD brain feature subgrouping (Piras et al. 2014; Hahamy et al. 2015), ASD behavior subgrouping (Chaste et al. 2015; Kim et al. 2016; Veatch et al. 2014), ASD developmental course subgrouping (Fountain et al. 2012; Lord et al. 2015), ASD behavior and brain feature subgrouping (Lombardo et al. 2015), ASD MPAs subgrouping (Tammimies et al. 2015), ASD MCAs subgrouping (Timonen-Soivio et al. 2015), and ASD risk gene variant subgrouping (Hormozdiari et al. 2015; Krumm et al. 2014; Noh et al. 2013).

Males and females are valid biological subgroups. Males with ASD show more aggression and repetitive behaviors, while females with ASD express more mood disorders, and that females with ASD have greater cognitive impairment than males with ASD (Jeste and Geschwind 2014). There is a sex ratio of four or five males for every female with ASD (Halladay et al. 2015; Jeste and Geschwind 2014; Romano et al. 2016; Ypma et al. 2016). Romano et al. (2016) asserted that male genetic variants make them more vulnerable to ASD risk factors than are females. Jeste and Geschwind (2014) proposed that a possible reduced female vulnerability, termed the female protective effect (FPE), arises from genes in females that are found on the sex chromosomes, and the effects of sex hormones themselves. Halladay et al. (2015) noted support for the FPE in that females with ASD carry a higher number of dnCNVs and de novo loss of function point mutations than do males with ASD.

However, the sex ratio of males and females in ASD has not been adequately explained by either increased male vulnerability or a special female protective factor (Halladay et al. 2015; Ypma et al. 2016). For example, at the lowest level of ID, the male-female ASD sex ratio is 1:1. In addition, different etiologies have different sex ratios. Syndromic ASD with RTT includes only females, and syndromic ASD with FXS includes only males. Taking ASD apart by etiologies will likely reveal that the current overall ASD sex ratio is composed of many different separate sex ratios that vary by etiology, sex chromosome gene effects, and sex hormone effects.

Tammimies et al. (2015) divided children diagnosed with ASD into three groups: those with many MPAs, those with few MPAs, and those with some MPAs. Children with few MPAs had fewer molecular diagnoses than the children with many MPAs (Tammimies et al. 2015). However, there was genetic heterogeneity in each of the three ASD MPA subgroups, including varied de novo mutations and other varied molecular diagnoses (Tammimies et al. 2015).

Similarly, Chaste et al. (2015) found that subgrouping ASD by distinct phenotypes did not result in finding sets of genetic variants that were unique to distinct clinical phenotypes. In fact, fewer and greater numbers of overlapping genetic variants were reported across the ASD phenotypic groups. Chaste et al. (2015) nonetheless argued that clinical phenotypes must have some “power for discovering genetic associations” (p. 781).

Despite many efforts to form valid subgroups of ASD, no ASD subgroups have become standard, and no ASD subgroup has yet achieved adequate internal homogeneity (Chen et al. 2015). Thus, to date, ASD has not been subdivided into valid subgroups that together would provide construct validity for the ASD spectrum. Males and females are valid biological subgroups, but the sex ratio of males and females in ASD has not been adequately explained, and different ASD etiologies have different sex ratios.

ASD should be disbanded in research because it lacks validity, and finding effective medical treatments requires the analysis of individual variation in valid pathophysiological mechanisms linked to specific risk factors in complete neurodevelopmental phenotypes (Kennedy et al. 2015; London 2014; Volkmar and McPartland 2015; Waterhouse and Gillberg 2014). Because the ASD diagnosis is defined by just two symptoms from an individual’s complete set of symptoms, an ASD-defined sample is widely over-inclusive, capturing many different “lifelong conditions that can arise from a complex combination of multiple genetic and environmental factors” (Dawson 2016, para. 1), in which “the genetic architecture of ASD seems to be different from one individual to another” (Huguet et al. 2016, p. 118), and there are many individually varied “connections between brain and behavior” (Kennedy et al. 2015, p. 81). In sum, the heterogeneity of risk factors, brain impairments, and non-diagnostic symptoms in an ASD-defined sample blocks valid scientific inference.

There is no catalog of brain impairments for neurodevelopmental disorders because much past research was based on the belief that each neurodevelopmental disorder had its own specific brain impairment (Bourgeron 2015a). Jeste and Geschwind (2014) highlighted the need for a catalog of brain impairments in genetic research by inserting a question mark where (ASD) brain impairments should have been listed (Jeste and Geschwind 2014, Fig. 1, p. 75). Individual variation in brain impairments reflects variation in risk factors, and as argued for RDoC, pathophysiology variation is not aligned with diagnostic categories but is a focus for specific medical treatment.

Disbanding the ASD diagnosis will require new screening, evaluations, and health records for all neurodevelopmental disorders to include social impairment and RRBs and sensory atypical responsivity independent of the ASD diagnosis. Many forms of electronic health records (EHRs) have been outlined (Castillo et al. 2015; McCoy et al. 2015; Simon et al. 2014).

Gillberg (2010) proposed “Early Symptomatic Syndromes Eliciting Neurodevelopmental Clinical Examinations” or ESSENCE, to include early appearing difficulties in general development, communication and language, social interrelatedness, motor coordination, attention, activity, behavior, mood, and/or sleep. ESSENCE disorders include ASD, ADHD, ID, BPD, oppositional defiant disorder (ODD), specific language impairment (SLI), learning disabilities (LD), nonverbal learning disabilities (NVLD), tic disorders, behavioral phenotype syndromes, rare epilepsy syndromes, and reactive attachment disorder. Gillberg (2010) found that comorbidity resulted in most children having two or more diagnoses. Empirical support for ESSENCE (Bourgeron 2015b; Hatakenaka et al. 2016) has demonstrated that it effectively documents the variation and complexity of neurodevelopmental disorder symptoms.

London (2014) similarly recommended that research should study DSM-5 neurodevelopmental disorders as a whole, but proposed a multiaxial model that would allow for the identification of specific symptoms such as sensory overresponsiveness and ID, allow for possible etiologic factors such as genetic findings or infection during pregnancy, allow for functional assessment, and also allow for treatment outcome history. Most importantly, London (2014) noted that this reformulated nosology would direct researchers and clinicians to attend to an individual’s specific symptoms and not force researchers to assign a categorical diagnosis. Instead an inventory of individual symptoms and risk factors would provide a template suitable for clinical use in lieu of a diagnostic category.