Among children who were born at an extremely low gestational age, those who have motor, vision, or hearing impairments are much more likely than others to screen positive on the M-CHAT. Because we have yet to assess these children for an autism-spectrum disorder at an older age, we do not know if these children with motor, vision, or hearing impairments really are at increased risk of an autism-spectrum disorder, or if their visual, hearing, and motor deficits are equated with characteristics commonly seen in autism, such as visual avoidance, inconsistent response to voice, and failure to point or play with toys.
Although some of the risk of screening positive on the M-CHAT appears to be related to motor and special sensory impairments, among children without such impairments, the rate of screening positive was still 16%, nearly three times higher than expected among unselected populations. Even among children without cognitive impairment, 10% screened positive, nearly twice the rate expected.
The M-CHAT was developed in the late 1990s as a first-stage screen for ASD in toddlers aged 18 to 24 months with sensitivity of 0.87 and specificity of 0.99 in American children [2
]. More recent reports indicate that sensitivity, specificity, and positive predictive value (PPV) might be lower [3
]. In unselected children, the PPV is 0.11, and in high-risk children it is 0.6 [3
]. When adding telephone interview confirmation for the M-CHAT screens, something that was not done in the ELGAN study, PPV increases to 0.65 in unselected children and to 0.76 in high-risk children. If the M-CHAT were an ideal screening tool, it would identify all who should be given an autism-spectrum diagnosis (high sensitivity), and a low number of others who should not be given this diagnosis (high specificity). Only 0.6% of children in the general population are given an autism-spectrum diagnosis [19
], yet almost ten times that many (5.7%) screen positive [3
]. In contrast, however, 21% of our sample screened positive.. We have not yet verified ASD in our sample, and so cannot provide any information about sensitivity, specificity and predictive values of the MCHAT in our sample.
Children whose mental age was below 16 months equivalent were much more likely to be M-CHAT positive than the remainder of the cohort. Inclusion of children with mental age less than 16 months should not diminish the importance of our findings because two-thirds of children destined to have ASD have cognitive impairment [21
Approximately 70% of children with autism have no identifiable medical or genetic cause and are morphologically normal [22
]. Among such children with “idiopathic” autism, the ratio of boys to girls is close to 4 rather than the ratio of between 1 and 2 seen among children who have a medical or genetic cause or are morphologically abnormal. Fully 15% of children with idiopathic autism have macrocephaly, whereas microcephaly appears to be especially common among the children who have a syndrome or a medical- genetic basis for their ASD [22
]. Finally, severe cognitive impairment is considerably more common among children with syndromic and otherwise explained autism than among children with idiopathic autism [22
]. With a sex ratio of 1.4, an elevated rate of microcephaly (data not shown), and a high rate of severe cognitive impairment, the children who screened positive in our sample more closely resemble the pattern seen in children who have syndromic/medical-genetic disorder-explained autism than in those with idiopathic autism.
If an appreciable proportion of the children in our sample are documented to have ASD, some of the increased rate might be attributable to prematurity-related antecedents. One study found that among adolescents with normal IQ and vision, those diagnosed with periventricular leukomalacia were more likely than controls to have impairments in perceiving and understanding the actions of others, features seen also in those with ASD [23
]. These children had diminished right temporal lobe white matter size on MRI volumetric studies.
We raise the possibility that in its present form, the specificity of the M-CHAT may be low among children born extremely prematurely because of associated developmental impairments and other unrecognized characteristics Because this cohort did not undergo a diagnostic evaluation for autism, we cannot assess the false positive rate for the M-CHAT screen. If the verified rate of ASD in our population is only minimally or moderately increased among children who screen positive, however, then the M-CHAT might require modification for use in children born at extremely low gestational ages and for children who have physical and special sensory impairments.
Our cohort is at high risk for developmental impairments, whether or not they ultimately carry the diagnosis of ASD, because 23% of children with a false positive M-CHAT have developmental language disabilities and/or global developmental disorders [18
]. Indeed, our data indicate that only 10% of ELGAN children who screen positive on the M-CHAT have a normal MDI and PDI.
Strengths of the current study include the large sample based on gestational age rather than birth weight, efforts to minimize inter-observer disagreements about motor findings, standardized administration of the BSID-II, and completion of the M-CHAT screen. A limitation of the M-CHAT screen in this study is that telephone confirmation of results was not sought.