In this study, we describe a high prevalence of autism spectrum features among survivors of extremely preterm birth as detected by a positive score on a widely used screening instrument, the M-CHAT. In addition, we describe a strong correlation between positive M-CHAT scores in this population and the detection of internalizing behavioral problems and socialization and communication deficits using other widely used instruments, the CBCL and the VABS, respectively. Together, these findings provide evidence for a high prevalence of positive initial screenings for autism spectrum behaviors in survivors of extreme prematurity. To our knowledge, this is the first description of this phenomenon in expreterm infants. In addition, we have identified several factors that seem to increase the risk for positive M-CHAT scores in these children, including lower birth weight and gestational age, male gender, prenatal infection, greater illness acuity based on SNAP-II scores, and abnormal MRI studies.
No previous reports of autism screening testing in ex-preterm infants are available for comparison with our findings. Recent population-based studies identified pre-maturity or low birth weight,29,43,44
as well as a history of neonatal intensive care,39,40,45,46
as important perinatal risk factors for subsequent autism spectrum disorders. Badawi et al40
found a sixfold increased risk for autism spectrum disorders among children with a history of moderate to severe neonatal encephalopathy. Others described the association between autism and severe medical conditions such as epilepsy, cerebral palsy, fragile X syndrome, tuberous sclerosis, Down syndrome, congenital rubella, and significant hearing and visual impairments.41
As an explanation for these associations, it has been speculated that pregnancy, delivery, and/or neonatal complications act through independent etiologic pathways to increase the risk for autism or interact with a genetic predisposition by interfering in the developmental process at critical times.39,42
In this population, we identified several independent risk factors that were associated with a greater likelihood for abnormal M-CHAT scores. These included lower birth weight and gestational age, male gender, chorioamnionitis, and significantly greater illness severity (by the SNAP-II scores). Our finding of a high rate of chorioamnionitis by placental pathology in infants with positive M-CHAT scores is consistent with previous reports of recurrent maternal infections during pregnancy in children with autism.44
The presence of a higher acuity of illness suggests that exposure to factors that are associated with preterm birth and the hazards of prematurity itself are associated with a greater risk for a positive autism screening. Finally, our finding of a higher prevalence of positive autism screenings among boys corroborates previous studies that reported a significant association between autism spectrum disorders and male gender.38,43,44
The children in our study were tested at a mean age of 21 months’ adjusted age. To date, no studies have addressed the presence of neuroimaging abnormalities in children with autism spectrum disorders at such a young age, because autism is not typically recognized before 2 years of age47
; however, a broad spectrum of neuroanatomic abnormalities have been reported in children with autism.47–52
Recent studies suggest a role for disturbed connectivity between specific brain regions in autism spectrum disorders.53,54
Cerebellar lesions, particularly in the vermis, have been a common finding in autopsy and neuroimaging studies of children with autism.47,55–58
In our study, the presence of neonatal MRI abnormalities was a significant univariate predictor of positive scores on the M-CHAT test at follow-up. Although the number of infants with cerebellar hemorrhage (either isolated or combined with supratentorial parenchymal injury) was relatively small, this subgroup was significantly more likely to have positive M-CHAT scores than infants without cerebellar injury on MRI. These findings corroborate our results from a previous study59
in which more than one third of ex-preterm infants with isolated cerebellar hemorrhagic injury had positive autism screening tests at follow-up; however, in the current study, this specific relationship could not be tested in a multivariate model given the small size of the subgroup with cerebellar injury. Consequently, cerebellar injury was incorporated into an overall abnormal MRI category, which failed to reach statistical significance in its relationship with positive M-CHAT scores. Larger studies will be needed to address the potential importance of this relationship. Furthermore, it is important to note that in this study, a subset of ex-preterm infants with normal conventional MRI studies had positive screening scores for autism spectrum disorders. It is possible that these children have brain injury below the resolution of current conventional MRI techniques, and more sophisticated quantitative MRI techniques may be needed to address this issue. Using such techniques, we previously showed a significant impairment of cerebellar growth in preterm infants as early as term corrected age, even in the absence of obvious cerebellar injury by conventional MRI studies.60
Children with global developmental delay or specific language impairments and those with autism spectrum disorders may share a number of common features, particularly at a young age.61–63
Although we have not yet tested our population with diagnostic instruments for autism, when controlling for the potential confounding effect of language delays and age at testing, we found no significant effect of these factors on positive M-CHAT scores. Furthermore, there was no significant association between functional motor delays on the VABS and an abnormal M-CHAT. Although we acknowledge that the presence of developmental delay in our cohort may have contributed to the high prevalence of positive M-CHAT scores, our findings suggest that any such contribution was modest at best.
We acknowledge several important limitations to our study. First, it is important to recognize that the M-CHAT is specifically designed as a screening instrument to identify toddlers at risk and is by no means diagnostic for autism spectrum disorders. More comprehensive and standardized diagnostic tests for autism (eg, Autism Diagnostic Interview,64
Autism Diagnostic Observation Schedule65
) are available and are specifically designed to be applied to children who have positive results on screening tests such as the M-CHAT. Because the M-CHAT is designed to screen toddlers at approximately 18 months of age, it is possible that the sociobehavioral deficits identified in this study are transient or, conversely, may emerge or increase over time. Furthermore, although we correlated autism with behavioral and adaptive measures, we did not use standardized cognitive assessment. For these reasons, we are performing standardized diagnostic tests for autism and cognitive testing in this study population. Another limitation of our study is the lack of socioeconomic data on our cohort, which may have influenced both child development and maternal reporting. Finally, our study cohort represents a selected high-risk population of preterm infants31
and therefore may not be generalizable to healthier preterm populations.