Positive Screening Results on the Modified Checklist for Autism in Toddlers: Implications for Very Preterm Populations

Although the earliest reports of long term outcomes in preterm survivors focused on the identification of neurosensory morbidity and cerebral palsy, improved multidisciplinary follow-up for these populations increased the awareness of the high prevalence of cognitive sequelae and a range of psychopathology associated with very preterm birth.1 Most notably, studies of extremely preterm and extremely low birthweight survivors have documented a high prevalence of behavioral disorders, emotional problems, and attention deficits across the lifespan. Population denominators and the selection of outcome measures may differ in studies, but the high prevalence of social difficulties and deficits in social competence remains a robust finding.2

There also is a growing concern about a high prevalence of social and communication difficulties in these children. Many clinicians can provide anecdotal evidence of extremely preterm children who have difficulties integrating with their peers and who exhibit atypical behavioral features characteristic of the autism spectrum. Epidemiological studies provide supporting evidence for these observations identifying preterm birth as a risk factor for autism. Yet, there have been few systematic studies of the prevalence of autism spectrum disorders (ASD) in preterm populations.

Perhaps the relative paucity of research not only belies the recent increasing awareness of ASD in this population, but also reflects the difficulty in assessing such symptomatology in large-scale studies. Psychiatric interviews yielding clinical diagnoses are arguably the gold standard, but are costly and time consuming and typically comprise only 1 part of a wider battery of neuropsychological and clinical assessments that must be administered. The few large-scale studies of preterm cohorts conducted with these measures have reported a low but importantly raised prevalence of ASD in low birthweight children at school age of 1% to 2%.3, 4

Parent-report questionnaires are useful tools for assessing symptomatology and provide a more cost- and time-efficient measure for large population-based studies. Newly validated tools, such as the Modified Checklist for Autism in Toddlers (M-CHAT),5 have been developed to enable the early identification of infants and toddlers at risk for autism. With the M-CHAT, 25% of infants born <30 weeks gestational age screened positive for the risk of an ASD,6 and Kuban et al report a similarly high positive screening rate of 21% in infants born <28 weeks with the same measure.7

Although these studies undoubtedly confirm a true increase in the risk for autism, other confounding factors may inflate the positive screening rate in this population. The issue of false-positive results arises not only for the M-CHAT,5 but for all screening measures because cutoff points are selected to maximize identification of individuals at greatest risk while providing a compromise between false-negative and false-positive screening results. However, the specificity of behavioral screening tools is further compromised in extremely preterm populations. The high frequency of neurological, cognitive, and sensory sequelae may give rise to false-positive classifications as the functional consequences of such disorders, which may be equated with the socio-communicative deficits commonly seen in ASD.

This issue is both highlighted and addressed by Kuban et al in this issue of The Journal.7 The authors report findings from the ELGAN Study—a large multi-center study of babies born <28 weeks gestational age. Although the high prevalence of positive screening results in this population is noteworthy, the primary thrust of this paper lies in the authors' attempt to quantify the impact of neurocognitive disabilities on false-positive screening results in this population.

The authors report that the risk for positive screening results on the M-CHAT was increased 23-fold in children with severe neuromotor impairment, 8-fold in children with visual and hearing impairments, and 13-fold for children with severe cognitive impairment. The positive screening rate was reduced to 16% when children with neuromotor and sensory impairments were excluded and to 10% when children with cognitive impairment were also excluded. Because cognitive impairment is frequent in children diagnosed with an ASD, it may be argued that adjustment for this over-compensates for extreme prematurity, but perhaps these adjusted rates better reflect the true prevalence of autism spectrum symptoms in this population.

However, the lack of clinical diagnostic information precludes verification of false-positive results in this study, as the authors acknowledge. Although false-positive screening results may be important clinically to identify a group of children in whom diagnostic assessment may also be beneficial,8 the identification of new cutoff scores may be required for refining diagnostic usefulness in this and other impaired populations. Therefore it will be crucial to follow-up with these children to measure the prevalence of those who continue to manifest autism spectrum symptoms and who receive clinical diagnoses of ASD later in life. This is likely to be most fruitful in such a large-scale study that has a rich dataset of perinatal variables and thus has the potential to identify possible biomarkers for such symptoms. This study also underscores the need for applying standard diagnostic criteria, because screening measures cannot fully substitute for the experience and rigor of clinical raters in diagnosing autism and isolating ASD symptomatology within the context of other significant and varied sequelae.

Finally, the authors also draw attention to a further issue of the etiology of ASD in preterm populations. A discussion of the correlates of positive screening led the authors to suggest that autism in this population may be more characteristic of children who have a syndromic or medically defined ASD-type, and thus highlights the contemporary issue of whether autism is a primary deficit in this population or perhaps represents socio-communicative deficits that are part of an “extremely preterm phenotype.” Language and learning difficulties have been accounted for by the general cognitive ability in an extremely preterm population,9 and thus it is conceivable that autism spectrum symptomatology may be caused by cognitive dysfunction. Indeed, in the EPICure study we recently have shown that adjustment for IQ accounts for the excess of repetitive and stereotyped behavior symptoms in extremely preterm children compared with their term peers.10 The remaining excess of reciprocal social interaction and communication problems, although reduced after adjustment for IQ, may potentially be accounted for by other neuropsychological difficulties, such as deficits in attention and executive function that have been shown to confer additional morbidity in this cohort.11, 12 Further systematic investigations are now needed to determine the population prevalence and correlates of extremely preterm children with ASD and to further elucidate etiological factors.

The paper by Kuban et al is timely, highlighting the potential prevalence of autism spectrum symptoms and the pitfalls of screening for autism in impaired populations. The findings do not negate the need for early screening because this remains crucial for identifying children in need of diagnostic assessment and referral. Rather, it raises our awareness of issues involved in interpreting screening results both clinically and for research purposes. Ultimately, we will need to look longer term and to clinical diagnoses to enable us to understand better the significance of such symptoms in preterm populations.