Stability of early risk assessment for autism spectrum disorder in preterm infants

Abstract

Stability and change in early autism spectrum disorder risk were examined in a cohort of 99 preterm infants (⩽34 weeks of gestation) using the Autism Observation Scale for Infants at 8 and 12 months and the Autism Diagnostic Observation Schedule—Toddler Module at 18 months. A total of 21 infants were identified at risk by the Autism Observation Scale for Infants at 8 months, and 9 were identified at risk at 12 months, including 4 children who were not previously identified. At 18 months, eight children were identified at risk for autism spectrum disorder using the Autism Diagnostic Observation Schedule—Toddler Module, only half of whom had been identified using the original Autism Observation Scale for Infants cutoffs. Results are discussed in relation to early trajectories of autism spectrum disorder risk among preterm infants as well as identifying social-communication deficiencies associated with the early preterm behavioral phenotype.

Keywords

autism spectrum disorders development early diagnosis preterm screening

Introduction

In this study, we assessed early autism markers in a sample of preterm born infants with the aim of bridging between the two fields of study: early markers of autism and developmental outcomes of preterm infants.

Early markers of autism in high-risk populations

Most children with autism spectrum disorders (ASD) are not usually diagnosed before the second or third year of life. Early identification of ASD risk, even without meeting full criteria for a clinical diagnosis, may promote early intervention programs focusing on cognitive, social, and daily living skills (Dawson et al., 2010). Earlier rather than later intervention is more beneficial due to early brain plasticity and potentially modifiable abnormalities in early reward circuitry (Dawson, 2008; Helt et al., 2008), thus strengthening the need for early identification of ASD and its prodrome (Yirmiya and Charman, 2010).

Evidence for early markers of ASD has been obtained from retroactive parental reports, analysis of early home videos of children later diagnosed with ASD, prospective screening studies in which infants who score positive on early ASD screeners are followed longitudinally, and prospective studies focusing on young siblings of children with ASD, who are followed longitudinally from birth on. Several predictive markers of ASD were reported in the first two years of life, including atypical eye gaze, reduced orienting to name and social smiling, delays in play skills, atypical regulation of attention and emotion, repetitive interests and behaviors, and motor delays (Yirmiya and Charman, 2010; Zwaigenbaum et al., 2013).

Bryson et al. (2008) developed the Autism Observation Scale for Infants (AOSI) to monitor the signs of ASD in high-risk infants between the ages of 6 and 18 months. The AOSI successfully differentiated siblings who were subsequently diagnosed with ASD from siblings who were not subsequently diagnosed with ASD. The AOSI also identified elevated levels of autistic traits in unaffected siblings (with no ASD diagnosis) compared to low-risk infants with no family history of ASD (Brian et al., 2008; Georgiades et al., 2013; Zwaigenbaum et al., 2005). Most prior studies employing the AOSI included high-risk siblings of children with ASD; therefore, its generalization to other at-risk populations remains unclear (Bryson and Zwaigenbaum, 2014). Recently, Ben-Sasson and Carter (2012) demonstrated the utility and the validity of the AOSI in 12-month-old infants from the general population who were identified as at risk by a screening questionnaire—the First Year Inventory (FYI; Reznick et al., 2007). Infants who were identified as at a high risk by the FYI scored significantly higher on the AOSI compared to a group identified as at a low risk by the FYI.

The etiology, developmental trajectories, and the phenotypic characteristics of ASD and related symptomatology may be different among pre- and full-term infants (Bowers et al., 2015), yet little is known about the developmental trajectories of ASD among preterm infants. Furthermore, despite the higher risk for ASD among infants born preterm, screening studies and sibling studies aimed at identifying ASD usually do not include data regarding preterm birth among the participants nor examine prematurity as an independent variable, and the assessment instruments for ASD have not yet been validated for the preterm subpopulation. Thus, the aim of this study was to narrow this gap and examine early markers of ASD risk in a sample of preterm infants.

Prematurity and ASD risk

Advances in neonatal intensive care have dramatically increased survival of preterm infants, especially among the most prematurely born (Msall, 2010). However, this decrease in mortality has not been followed by a decrease in long-term neuro-developmental comorbidity, resulting in an increasing number of children with significant developmental difficulties and disabilities (Johnson and Marlow, 2014; Rushing and Ment, 2004). Johnson and Marlow (2011) conceptualized what they termed the “preterm behavioral phenotype,” which has overlapping behavioral manifestations with the ASD phenotype. The preterm phenotype is characterized by difficulties in regulation of social interaction, language, attention, sensory processing, and motor control (Hofheimer et al., 2014; Johnson and Marlow, 2011). To the best of our knowledge, this association has not yet been addressed systematically with gold-standard measures, and there are paucity of data regarding clinical assessment of ASD in preterm children. Furthermore, ASD assessment instruments, including the AOSI and the Autism Diagnostic Observation Schedule—Toddler Module (ADOS-T), have not been yet specifically validated for the preterm subpopulation.

Nonetheless, several lines of research suggest that preterm born children are at an elevated risk for ASD. Prematurity and low birthweight have been identified in population studies as risk factors for ASD (Larsson et al., 2005; Schendel and Bhasin, 2008; Sugie et al., 2005), with increasing risk associated with shorter gestation (Kuzniewicz et al., 2014; Leavey et al., 2013). As seen in Table 1, researchers examined that the prevalence of ASD diagnoses in cohorts of children born prematurely used different inclusion criteria and different diagnostic instruments with participants of a wide age range (Elgen et al., 2002; Hack et al., 2009; Indredavik et al., 2004; Johnson et al., 2010, 2011; Pinto-Martin et al., 2011; Treyvaud et al., 2013) and reported a 1%–8% rate of ASD diagnosis among school-aged children, adolescents, and young adults (aged 7–21 years). These reported rates were significantly higher than rates of ASD in full-term counterparts.

Table 1.

Studies on ASD prevalence among preterm and low birthweight children and adolescents.

Authors and year of publication	N	Inclusion criteria	Assessment age (years)	ASD measure	ASD diagnosis (%)	Other measures applied	Additional findings
Elgen et al. (2002)	130 preterm	<2000 g with no major disability	11	Asperger Syndrome Diagnostic Interview	1%	Child Behavior Checklist; Children Assessment Schedule
Elgen et al. (2002)	131 full-term	<2000 g with no major disability		Asperger Syndrome Diagnostic Interview		Child Behavior Checklist; Children Assessment Schedule
Indredavik et al. (2004)	56 preterm	<1500 g	14	Autism Spectrum Screening Questionnaire (ASSQ)	2% Asperger diagnosis, 8% Asperger symptoms	Schedule for Affective Disorders, ADHD Rating Scale, Children’s Global Assessment Scale
	60 full-term SGA
	80 full-term AGA
Hack et al. (2009)	219 preterm	<1000 g	8	Parent Child Symptom Inventory (CSI-4)	4%: 4 autism, 3 Asperger, 1 PDD-NOS		All children with ASD had comorbid behavioral diagnoses
Hack et al. (2009)	176 full-term			Parent Child Symptom Inventory (CSI-4)	4%: 4 autism, 3 Asperger, 1 PDD-NOS		All children with ASD had comorbid behavioral diagnoses
Johnson et al. (2010, 2011)	219 preterm	<26 weeks	11	Screening with SCQ followed by clinical interview (DAWBA)	16% screened positive, 8% assigned an ASD diagnosis	IQ Test; Child Behavior Checklist	SCQ false-positive was associated with cognitive and behavioral impairment
Johnson et al. (2010, 2011)	153 full-term			Screening with SCQ followed by clinical interview (DAWBA)	16% screened positive, 8% assigned an ASD diagnosis	IQ Test; Child Behavior Checklist
Pinto-Martin et al. (2011)	623 at time 1	<2000 g	16	SCQ, ASSQ, parent report on child’s diagnoses	18.8% on one or more measure
	189 at time 2	Time 2 included 60% of screen positives and 24% of screen negatives at time 1	21	Autism Diagnostic Observation Schedule, Autism Diagnostic Interview	5%
Treyvaud et al. (2013)	177 preterm	<30 weeks or <1250 g	7	DAWBA	4.5%
Treyvaud et al. (2013)	65 full-term

ASD: autism spectrum disorder; SGA: small for gestational age; AGA: appropriate for gestational age; ADHD: attention deficit hyperactivity disorder; PDD-NOS: pervasive developmental disorder—not otherwise specified; DAWBA: development and well-being assessment; SCQ: Social Communication Questionnaire.

Addressing the need for earlier identification of ASD risk in preterm cohorts, researchers have administered the Modified Checklist for Autism in Toddlers (M-CHAT; Robins et al., 2001)—which is the recommended screening instrument for ASD (Johnson and Myers, 2007) at the corrected age of 18–24 months (Dudova et al., 2014; Guy et al., 2015; Kuban et al., 2009; Limperopoulos et al., 2008; Luyster et al., 2011; Moore et al., 2012; Stephens et al., 2012; Wong et al., 2014; see Table 2). As seen in Table 2, the rates of positive screening of 10%–40% among these younger samples are alarmingly higher than the rates of 1%–8% found among the older samples of preterm children. Major sensory, motor, and cognitive impairments partially accounted for the positive screening among the younger samples (Johnson and Marlow, 2009), yet, among subgroups of extremely preterm children who were free of motor, visual, and hearing impairments, the prevalence of positive screening still surpassed the rates among older preterm children and children from the general population.

Table 2.

Screening for ASD in preterm toddlers.

Authors and year of publication	N	Inclusion criteria	Average age (months)	ASD screening measure	Positive scores (%)	Other measures applied	Additional findings
Limperopoulos et al. (2008)	91	<1500 g	22	M-CHAT	26%	Vineland Adaptive Behavior Scales and Child Behavior Checklist	Positive screening was associated with internalizing problems and socialization deficits
Kuban et al. (2009); Luyster et al. (2011)	988	<28 weeks	24	M-CHAT	21%	Neurological examination, gross motor assessment, Bayley Scales, and parental interview	After excluding children with sensory-motor difficulties, 16% screened positive. After excluding children with cognitive difficulties, 10% screened positive
Moore et al. (2012)	523	<26 weeks	24	M-CHAT	41%	Parent report of child abilities	After excluding children with sensory or motor difficulties, 16% screened positive. Almost all children with sensory-motor difficulties screened positive
Stephens et al. (2012)	554	<27 weeks without neuro-sensory impairment	18–22	PDDST-II, response to name, and response to joint attention	20% on at least one tool, 10% on PDDST-II, 6% on response to name, and 9% on response to joint attention	Gross motor assessment, Bayley Scales, and Brief Infant-Toddler Social and Emotional Assessment	Positive screen was associated with cognitive and language delay and behavioral problems
Dudova et al. (2014)	75	<1500 g without neuro-sensory impairment	24	M-CHAT, CSBS-DP-ITC, and ITSP	42.7% on at least one tool, 34.7% on CSBS-DP-ITC, 25.3% on M-CHAT, and 14.7% on ITSP	n = 19 with positive screen had diagnostic assessment including ADOS	8 of 19 with positive screening were assigned ASD diagnosis
Guy et al. (2015)	634 preterm761 full-term	32–36 weeks	24	M-CHAT	14.5% on questionnaire only, 2.4% after follow-up interview, and 2% after exclusion of children with neuro-sensory impairments	Parent report of child abilities	Preterm infants had significantly higher risk of positive screen, even after follow-up. False-positive screens were associated with cognitive and behavioral problems
Wong et al. (2014)	141	<30 weeks without neuro-sensory impairment	24	Q-CHAT)	Q-CHAT scores of preterm cohort were significantly higher than general population scores	Bayley-III	Lower Bayley language scores were associated with higher Q-CHAT scores

ASD: autism spectrum disorder; M-CHAT: Modified Checklist for Autism in Toddlers; PDDST-II: Developmental Disorders Screening Test—Second Edition, Stage 2; CSBS-DP-ITC: Communication and Symbolic Behavior Scales Developmental Profile Infant—Toddler Checklist; ITSP: Infant/Toddler Sensory Profile; ADOS: Autism Diagnostic Observation Schedule; Q-CHAT: Quantitative Checklist for Autism in Toddlers.

Screening positive on the M-CHAT or on other screening instruments that rely mostly on parental reports may not necessarily identify markers for ASD in the preterm population, but rather may serve as an indication of developmental difficulties associated with the preterm phenotype (Hofheimer et al., 2014). Most of the aforementioned screening studies did not include follow-up assessments with clinical observational measures. Thus, there is paucity of data regarding clinical, observer-based assessment of ASD risk among the preterm population.

Behavioral abnormalities and deficits in social interactions in preterm children may be present at very early ages, from birth throughout the first years of life (Korja et al., 2012; Vanderbilt and Gleason, 2011). Preterm infants’ interactions with caregivers have shown that preterm infants displayed lower social responsiveness than full-term infants (Gerner, 1999; Singer et al., 2003), more withdrawal behaviors (Keren et al., 2003), lower levels of social gaze and gaze synchrony (Feldman and Eidelman, 2007), and more difficulties in the regulation of the social interaction (Montirosso et al., 2010). Researchers utilizing observer-based measures have also shown that preterm infants have more difficulties in social initiation and responding to joint attention (De Groote et al., 2006; De Schuymer et al., 2011a, 2011b). Thus, early social-communication symptomatology may be detected in preterm infants as part of the early preterm phenotype during the first years. However, the stability of these behaviors over time and their association with ASD symptoms were not yet examined.

Given the overlap between ASD symptoms and the preterm phenotype, research and clinical practice calls for additional instruments, which allow systematic observation and evaluation of the emerging atypical behaviors in preterm infants, by experts who conduct the evaluation independently of parental reports. This study addresses this by utilizing the AOSI as an observer-based screener for ASD risk in preterm infants during the first year of life and the ADOS-T (Luyster et al., 2009) for ASD risk assessment at 18 months.

Study rationale

The importance of repeated assessments when screening for ASD in the first years of life was recently emphasized (Zwaigenbaum et al., 2009). Thus, in our prospective study of preterm infants, we employed the AOSI at 8 months, re-administered it at 12 months, and administered the ADOS-T at 18 months, therefore assessing the stability of ASD risk from 8 to 18 months. Follow-up assessments of children who screen positive are important in order to examine the stability of high-risk status over time. Nonetheless, follow-up assessments of children who screen negative are no less important to examine the stability of low-risk status over time. This study included preterm born infants, born between 24 and 34 weeks of gestational age. General developmental abilities were assessed at each time point with the Mullen Scales of Early Learning (MSEL; Mullen, 1995) and the Vineland Adaptive Behavior Scales-II (Sparrow et al., 2005). The AOSI was employed as a screener for ASD risk twice, at the ages of 8 and 12 months, followed by the ADOS-T at 18 months. The ADOS-T at 18 months yields a tentative indication of ASD risk, yet regardless of whether and ASD diagnosis will or will not follow, risk as identified by the ADOS-T indicates significant and meaningful difficulties in current functioning and may suggest that intervention could be beneficial (Luyster et al., 2009).

In sum, the goals of this study were to estimate the prevalence of positive screening for ASD risk using observational instruments employed at two time points during the first year of life (8 and 12 months) and then at 18 months, to examine stability and changes over time in early ASD risk

Methods

Participants and procedure

Participants included 110 children (45% girls), born at Hadassah University Hospital between 2009 and 2013. The study was approved by the hospital’s institutional review board (IRB) committee (249-09). Inclusion criteria included singleton birth at 24–34 weeks of gestation and no genetic disease or major congenital anomaly. All eligible families were approached during the infant’s neonatal intensive care unit (NICU) hospitalization, and parents who agreed to participate signed an informed consent form and completed a demographic questionnaire. Of the 200 families who were approached, 110 consented to participate (52% enrollment rate). Refusals were usually due to time and/or distance constraints or unwillingness to commit to multiple developmental assessments. No significant differences emerged between those who declined participation and those who agreed regarding infant birthweight or gestational age at birth.

Maternal age at birth ranged from 19 to 51 years (M = 31.70, standard deviation (SD) = 6.01), with 100 mothers (91%) living with a spouse. A total of 21 mothers (19%) completed high-school education, 19 mothers (13%) had some non-academic professional education, 46 mothers (42%) completed an undergraduate degree, and 24 mothers (22%) completed a Master and/or Doctoral degree. Regarding household income, 21% of the families’ income was below the median national household income (1st and 2nd deciles), 64% were within the median household income (3rd–7th decile), and 15% were above the median household income (8th–10th decile). Gestational age at birth ranged from 24 to 34 weeks (M = 31.28, SD = 2.57) and birthweight ranged from 490 to 2400 g (M = 1541.38, SD = 474.32). See Table 3 for a detailed description of neonatal medical complications in the sample.

Table 3.

Neonatal medical complications.

Characteristics	M	SD	Range
Gestational age at birth (weeks)	31.22	2.58	24–34
Birthweight (g)	1541	474.32	490–2400
Apgar score at 1 min	7.74	1.84	1–9
Apgar score at 5 min	8.99	1.21	2–10
No. of days of mechanical ventilation	3.19	8.02	0–44
No. of days in the NICU	42.27	34.70	9–205
Characteristics	N		%
Small for gestational age	8		7.8
Respiratory distress syndrome	28		27.46
Bronco pulmonary dysplasia	17		16
Sepsis	14		13.7
Intraventricular hemorrhage	9		8.8
Retinopathy of prematurity	9		8.8

SD: standard deviation.

Assessments were held at the research laboratory at the age of 8 months (M = 8.39, SD = 0.34, range: 7.50–8.89), 12 months (M = 12.06, SD = 0.40, range: 11.50–12.92), and 18 months (M = 18.19, SD = 0.36, range: 17.15–18.96). All ages were corrected for prematurity. Families were reimbursed for travel costs and received a videotape and a report of the assessments. Three infants who were diagnosed with severe sensory or motor deficits (two with cortical blindness and one with severe cerebral palsy) were excluded from the analyses, and three were lost between recruitment and the 8-month assessment. One infant missed only the 8-month assessment and another one missed only the 12-month assessment, thus data were available for 103 infants at 8 and 12 months. Three more infants did not complete the 18-month assessment, thus 18-month data were available for 101 infants. Altogether, 99 infants had a complete data set for all three assessments.

Measures

MSEL

The MSEL (Mullen, 1995) assess developmental functioning of children from birth through 68 months of age. The MSEL composite score offers a standardized general score (M = 100, SD = 15) based on four standardized scales (M = 50, SD = 10): fine motor, visual reception, expressive language, and receptive language and includes an additional gross motor scale. Previous studies showed high test–retest reliability (0.71–0.96) and inter-rater reliability (0.91–0.99). The MSEL revealed good concurrent validity in prior samples, with high correlations emerging for the expressive language (0.72–0.85) and receptive language (0.72–0.80) scales with the Preschool Language Assessment (Blank et al., 1978) and with high correlations (0.65–0.82) emerging between the MSEL fine motor scale and the Peabody Fine Motor Scale (Folio and Fewell, 1974).

Vineland Adaptive Behavior Scales-II

The Vineland-II (Sparrow et al., 2005) is a structured interview administered to caregivers to assess the child’s daily living skills on four domains: communication, daily living skills, socialization, and motor skills. Caregivers are asked to rate whether the child currently exhibits each described behavior or not. The four domain scores are standardized and compose the Adaptive Behavior Composite Score (M = 100, SD = 15 in domain and composite scores).

AOSI

The AOSI (Bryson et al., 2008) is an 18-item direct observational measure designed to detect and monitor putative signs of autism in infants between the ages of 6–18 months. The AOSI items assess visual attention, social-communication, play, and sensory-motor development. This set of semi-structured activities is administered by a trained examiner who is both skilled in interacting with infants and knowledgeable about ASD. The activities provide an interactive context in which the examiner engages the infant in play while conducting a set of systematic presses to elicit particular target behaviors. The relative presence or absence of key AOSI behaviors is rated on a scale from 0 to 3, with 0 implying normal functioning and higher values indicating increasing deviation from the norm. Bryson et al. (2008) found excellent inter-rater agreement on the total score: intra-class correlation coefficients were 0.71, 0.90, and 0.92 at 6, 12, and 18 months, respectively. Test–retest reliability at 12 months was also good (intra-class correlation coefficient = 0.63). Based on the data from Bryson et al.’s initial AOSI validation sample, counting the number of items coded one or above, a total score of ⩾7 markers and a total summary score of AOSI items of ⩾9 were considered to be possible indicators of ASD risk. However, these cutoffs have not yet been validated in additional samples, and the AOSI cutoffs are currently recommended for use as a research rather than a clinical instrument for screening (Bryson and Zwaigenbaum, 2014). The current research team was trained and obtained a minimum of 80% agreement with the creators of the measure as well as within the research team prior to administration. A total of 15% of the procedures were double-coded, yielding Cohen’s kappa of 0.77% and 90.1% agreement (range: 88%–92%).

ADOS-T

The Autism Diagnostic Observation Schedule (ADOS; Lord et al., 2002) is a semi-structured, standardized observational assessment designed to assess behaviors related to ASD. It provides a number of opportunities for communication, social interaction, and play or imaginative use of materials, and it measures social and communicative behaviors diagnostic of autism. Inter-rater and test–retest reliability, as well as internal validity, have been demonstrated for the ADOS. A new Toddler Module (ADOS-T; Luyster et al., 2009) was designed for use with children aged <30 months. A subset of 14 items comprises the diagnostic algorithm of the ADOS-T for 12- to 30-month-old infants, structured in two domains: Social Affect and Restricted Repetitive Behaviors. Algorithm scores of 0–9 indicate little-or-no concern, suggesting that the child displays no more ASD-related behaviors than children at the same age who do not have ASD; algorithm scores of 10–13 indicate mild-to-moderate concern and the presence of behaviors that are likely to be consistent with ASD and therefore require close monitoring; and algorithm scores of 14 and above indicate moderate-to-severe concern, strongly consistent with an ASD diagnosis and also require extensive monitoring.

The research team was trained and obtained a minimum of 80% reliability prior to administration. A total of 15% of the procedures were double-coded, yielding an averaged 0.73 kappa and 85.4% agreement (range: 83%–91%). Children whose ADOS-T algorithm scores at 18 months were in the mild-to-moderate (10–13) or moderate-to-severe (14 and above) concern ranges were considered as at risk for ASD. Children with ADOS-T algorithm scores in the no-to-little concern range were considered as not at risk for ASD.

Clinical judgment

The research team (including MY, AH, EF, IG, and NY) reviewed the results of the various observational instruments and parental reports, and according to their best clinical judgment gave the families feedback and recommendations for further evaluation and treatment if needed. Children who received ADOS-T scores in the concern range and/or who presented clinically significant deficiencies were referred for further assessments/interventions.

Results

Assessments at 8 and 12 months

At 8 months, the MSEL composite scores ranged from 59 to 115 (M = 91.59, SD = 12.46) and the Vineland-II total scores ranged from 70 to 120 (M = 94.29, SD = 9.40; see Table 4). The number of AOSI markers at 8 months ranged from 0 to 11, and the total scores ranged from 0 to 22 (see Table 5). The 8-month AOSI total scores were significantly correlated with gestational age at birth (r = −0.32, p < 0.001), birthweight (r = −0.28, p < 0.001), the MSEL composite score (r = −0.43, p < 0.001), and the Vineland-II total score (r = −0.27, p < 0.01).

Table 4.

MSEL and Vineland-II domain scores at 8-, 12-, and 18-month CA.

	8 months			12 months			18 months
	M	SD	Range	M	SD	Range	M	SD	Range
MSEL subscales
Gross motor	38.58	10.86	20–61	42.21	11.32	20–62	41.05	11.39	20–62
Visual reception	48.49	10.85	23–69	51.36	7.74	38–70	50.92	8.45	25–65
Fine motor	45.64	10.60	20–68	59.06	7.87	39–80	50.85	8.14	20–72
Receptive language	44.51	9.21	22–68	46.50	7.84	28–71	44.83	11.92	20–72
Expressive language	43.73	8.32	26–60	47.52	7.99	20–66	42.98	8.12	27–70
Vineland subscales
Communication	103.68	9.92	62–120	100.79	10.06	66–127	98.65	7.68	84–124
Daily living	96.83	9.90	72–122	91.92	10.97	75–122	104.97	10.46	80–137
Social	95.13	7.90	73–120	96.55	9.29	77–126	99.69	9.48	84–125
Motor	86.32	14.66	49–116	94.36	11.15	67–113	99.29	11.27	71–129

MSEL: Mullen Scales of Early Learning; CA: corrected age; SD: standard deviation.

Table 5.

AOSI scores and positive screens at 8- and 12-month CA.

Assessment age (months)	Score			Screening criterion (cutoff)	Who scoredabove cutoff (n)	Who screened positive^a (n)
Assessment age (months)	M	SD	Range	Screening criterion (cutoff)	Who scoredabove cutoff (n)	Who screened positive^a (n)
8 (N = 103)	3.80	2.37	0–11	No. of markers ⩾7	17	21
	5.22	3.82	0–22	Total score ⩾9	16
12 (N = 103)	2.81	2.03	0–10	No. of markers ⩾7	6	9
	3.50	2.68	0–13	Total score ⩾9	5

AOSI: Autism Observation Scale for Infants; CA: corrected age; SD: standard deviation.

Positive screening: either number of markers ⩾7 or total score ⩾9.

At 12 months, the MSEL composite scores ranged from 73 to 127 (M = 102.40, SD = 11.6) and the Vineland-II total scores ranged from 72 to 126 (M = 94.84, SD = 8.83; see Table 4). The number of AOSI markers at 12 months ranged from 0 to 10 and the total scores ranged from 0 to 13 (see Table 5). The 12-month AOSI total scores were significantly correlated with gestational age at birth (r = −0.22, p < 0.05), the MSEL composite score (r = −0.37, p < 0.001), and the Vineland-II total score (r = −0.38, p < 0.001).

In the current sample, the AOSI scores were lower overall (less abnormal) at 12 months compared to the AOSI scores at 8 months (see Table 5). When defining positive screening as either the number of AOSI markers ⩾7 or a total AOSI score ⩾9, 21 children screened positive at the age of 8 months, whereas only 9 children screened positive at the age of 12 months. As seen in Table 6, five of the nine infants who screened positive at 12 months (55.5%) also screened positive at 8 months, whereas the remaining four infants (44.4%) who screened positive at the age of 12 months were not identified as at risk in the 8-month screening. A total of 15 of the 20 (76%) infants who screened positive at 8 months no longer screened positive at 12 months (one infant who screened positive at 8 months did not attend the 12-month assessment).

Table 6.

Stability of AOSI screening^a between assessments at 8- and 12-month CA (N = 103).^b

	12-month assessment interval		Total at 8 months
	Who screened positive (n)	Who screened negative (n)
8-month assessment interval
Who screened positive (n)	5	15	20^b
Who screened negative (n)	4	79	83
Total at 12 months	9	94^b	103

AOSI: Autism Observation Scale for Infants; CA: corrected age.

Positive screening: either number of markers ⩾7 or total score ⩾9.

Missing data for one infant at 8 months, who screened negative at 12 months and another infant at 12 months who screened positive at 8 months.

Assessment at 18 months

At 18 months, the MSEL composite scores ranged from 69 to 126 (M = 95.20, SD = 12.60) and the Vineland-II total scores ranged from 81 to 131 (M = 100.50, SD = 9.12; see Table 4). At 18 months, the ADOS-T algorithm scores ranged from 0 to 15 (M = 3.74, SD = 3.40). ADOS-T algorithm scores were significantly correlated with gestational age at birth (r = −0.26, p < 0.01), birthweight (r = −0.25, p < 0.01), the MSEL composite score at 18 months (r = −0.45, p < 0.001), and the Vineland-II total score at 18 months (r = −0.45, p < 0.001).

As seen in Table 7, eight children at the age of 18 months (8% of the sample) had ADOS-T algorithm scores indicating elevated ASD concern, all of whom were males: six with scores in the mild-to-moderate concern range (10–13) and two (participants 1 and 2 in Table 7) with scores in the moderate-to-severe concern range (15). Based on clinical judgment, all these children had clinically significant developmental difficulties and were referred for intervention. Among those eight children whose ADOS-T scores indicated ASD concern, three received scores in the average range (88–98) and five received scores in the developmental delay range (71–82) of the MSEL composite scores. With regard to the Vineland-II total scores at the age of 18 months, seven children scored in the average range (89–104) and one scored in the developmental delay range (81). Two of these eight children were enrolled in a special day care program addressing their special needs due to extreme prematurity, whereas three others were receiving weekly intervention in the community (e.g. physiotherapy, occupational therapy). The remaining three of these eight children were not receiving any intervention at the time of the 18-month assessment and were referred by the research team for further evaluation and intervention in Child Development Centers in the community. For those who were already receiving intervention, our recommendations were to continue.

Table 7.

Characteristics of 18 -month-old boys identified with ASD risk by ADOS-T (n = 8).

Case no.	GA at birth (weeks + days)	8-month CA				12-month CA				18-month CA
		AOSI		MSEL composite	VABS total score	AOSI		MSEL composite	VABS total score	ADOS-T algorithm	MSEL composite	VABS total score
		Total	Markers	MSEL composite	VABS total score	Total	Markers	MSEL composite	VABS total score	ADOS-T algorithm	MSEL composite	VABS total score
1	33 + 1	10	8	74	95	11	8	84	87	15	98	93
2	27 + 5	10	7	88	94	8	7	94	80	15	72	81
3	33 + 3	7	6	90	95	2	1	115	95	10	88	97
4	31 + 1	2	1	91	98	0	0	99	98	10	80	95
5	26 + 4	7	6	67	72	3	3	86	81	10	74	91
6	24 + 5	14	9	74	74	10	6	84	84	11	73	91
7	29 + 0	15	9	61	84	10	6	83	81	10	82	89
8	28 + 1	5	4	96	104	7	5	112	96	10	94	104

ASD: autism spectrum disorder; ADOS-T: Autism Diagnostic Observation Schedule—Toddler Module; CA: corrected age; GA: gestational age; AOSI: Autism Observation Scale for Infants; MSEL: Mullen Scales of Early Learning; VABS: Vineland Adaptive Behavior Scales.

On the AOSI, cutoffs for positive screening were number of markers ⩾7 or total score ⩾9. On the MSEL, normative scores ranged from 88 to 98 and developmental delay scores ranged from 61 to 82. On the ADOS-T, concern for ASD was moderate for scores of 10–13 and was severe for scores of 14 and above.

Stability of ASD risk at 18 months

As seen in Table 8, four of the eight children who were classified as at risk for ASD at 18 months screened positive on the AOSI at both the 8- and 12-month assessments using the cutoffs for positive screening (i.e. either AOSI markers ⩾7 or total AOSI score ⩾9). However, the remaining four children who were classified as at risk at 18 months screened negative on both earlier AOSI screenings.

Table 8.

Stability of ASD risk between the AOSI at 8 and 12 months and the ADOS-T at 18 months.

AOSI screening	18-month ADOS-T risk
	High ASD risk (n = 8)		Low ASD risk (n = 91)
	Predicted at 8 months	Predicted at 12 months	Predicted at 8 months	Predicted at 12 months
Screened positive (n)	4	4	15	5
Screened negative (n)	4	4	76	86

ASD: autism spectrum disorder; AOSI: Autism Observation Scale for Infants; ADOS-T: Autism Diagnostic Observation Schedule—Toddler Module.

Discussion

To the best of our knowledge, this study is one of the first to employ systematic observational instruments for ASD risk assessment over time, in preterm infants—the AOSI at 8 and 12 months and the ADOS-T at 18 months. Previous studies with high-risk siblings of children with ASD utilized two cutoff criteria of the AOSI (number of AOSI markers ⩾7 and total AOSI score ⩾9) to identify most of the siblings who were later diagnosed with ASD (Brian et al., 2006; Zwaigenbaum et al., 2005). However, the AOSI has not been carefully examined in other high-risk samples such as preterm infants.

Using the aforementioned cutoffs, 21% of the preterm infants screened positive at 8 months on at least one cutoff criterion. At 12 months, the AOSI scores were lower, and rates of positive screens decreased from 21% to 9%. The ADOS-T, which provides a standardized quantitative measure of ASD symptoms, was used as a preliminary assessment of ASD risk at 18 months. Eight infants (8.1%) had ADOS-T algorithm scores that indicated elevated ASD risk—similar to the 9% rate of positive screens at 12 months.

Prevalence of ASD risk in preterm infants

The 21% prevalence of ASD risk as assessed at 8 months with the AOSI is similar to previous reports of 20% positive screens on the M-CHAT in extremely preterm toddlers (Kuban et al., 2009; Limperopoulos et al., 2008). However, 75% of the infants in our study who screened positive at 8 months were no longer identified as at risk at 12 and 18 months, thus screening at 8 months of age is most likely too early as it may yield an overestimation of the ASD risk in this sample. The more stable rate of 8%–9% of children identified at risk at 12 and 18 months with observational instruments is lower than the rates of ASD risk when these are collected from parental reports. The result of higher rates of positive score on the M-CHAT compared with the ADOS-T is not surprising, given that the M-CHAT is intended to be used as a screener to determine whether further diagnostic assessment using more direct instruments such as the ADOS-T is warranted. Moreover, as our study was not limited to extremely preterm infants as most of the M-CHAT studies, the infants in the current sample overall had fewer medical complications and were at a lower medical risk compared to previous samples (Kuban et al., 2009; Limperopoulos et al., 2008). Indeed, these findings are comparable to those of prior samples that included both extremely and moderately preterm children as compared to full-terms (Guy et al., 2015; Pinto-Martin et al., 2011).

Nonetheless, the rate of positive cases in this study may reflect early manifestations of the preterm phenotype, including developmental and language delays, behavior problems, and difficulties in regulation of social interactions that overlap with early ASD markers. Considering the preterm behavioral phenotype, symptoms of social and communicative difficulties are likely to range from full-blown ASD to broader autism phenotype features, as reflected in the elevated scores in the AOSI and the ADOS-T. It is thus essential to identify and treat children who are likely to exhibit marked deficits in the second year, regardless of what the eventual diagnostic classification might be, as well as to longitudinally follow and examine who does, or does not, develop the full clinical picture of ASD.

Stability and change of ASD risk among preterm infants

We next examined the stability and individual trajectories of ASD risk over time. The rates of positive screens decreased from 21% to 9% between 8 and 12 months. As the AOSI does not have age norms, it is not surprising that infants are indeed exhibiting more advanced social-communicative abilities when reevaluated 4 months later (Inada et al., 2010). Most of the children who scored above the cutoff at 8 months showed improvement and thus screened negative at 12 months. Although ASD risk decreased over time at the group level, individual trajectories revealed instability and inconsistency over time with some infants’ difficulties appearing for the first time at 12 months. The AOSI was designed to detect lack of or delay in attainment of certain skills as well as the presence of atypical behavioral characteristics in infants. Thus, it may be that the improvement over time reflects “catching up” delays and overcoming of difficulties with time, whereas certain abnormal characteristics may have appeared for some infants only at 12 months.

A similar pattern emerged between 12 and 18 months. Half of the children who had screened positive at 12 months were not classified 6 months later as at risk by the ADOS-T. Even more important, half of the 18-month-olds who were classified as at risk had not screened positive earlier. These results, with infants screening positive at different ages, demonstrate the challenge of early screening for ASD, specifically in a sample of preterm infants, and also emphasize the importance of repeated screenings for detecting early ASD markers (Pierce et al., 2011; Zwaigenbaum et al., 2009). Diagnostic assessment throughout childhood and adolescence may allow further examination of the trajectories of ASD risk in preterm infants. It may be the case that some early features of ASD risk are transient, appearing early in life and then later on disappearing with the development or vice versa with some difficulties and atypical behaviors emerging only later in the development.

Clinical implications

Sibling studies within the context of ASD have already shown the large heterogeneity of ASD symptomatology and the disorder’s different trajectories and onsets. This study contributes important information on an additional subgroup of preterm infants who are at risk for ASD and indeed show social-communication difficulties. The clinical manifestation of ASD in preterm infants may be different from other populations (Bowers et al., 2015). Thus, professionals who work with infants and toddlers at risk for ASD should take into consideration the unique preterm phenotype and account for perinatal risk factors and their associated behavioral phenotypes. For example, there may be characteristics that in the general population or among siblings of children with ASD are considered as indicative of an ASD risk (e.g. motor control, posture abnormalities, and regulatory difficulties) that are more common among preterm infants and may not necessarily be specific ASD markers in this population. In addition, professionals following the development of preterm infants should take into consideration the early manifestations of ASD risk and consider assessing social-communication and related behaviors in routine follow-up of preterm infants. The AOSI is rather easy to apply in various settings and is coded in real-time and thus addresses the need for a systematic measure of early social-communication risk.

Limitations and further directions

In this study, we examined early ASD risk in preterm infants and found instability in ASD risk and heterogeneity in developmental trajectories. ASD risk was assessed at the relatively early age of 18 months with the ADOS-T, and follow-up is thus necessary to determine the diagnostic outcome. These preliminary results should be interpreted with caution, and validation and modification of the assessment instruments for preterm infants are warranted.

In future studies examining ASD risk in preterm infants, it may be useful to examine more homogeneous samples by separating extremely preterm from moderately preterm infants. Additionally, studies utilizing both observational screening and parental reports instruments such as the M-CHAT may enable comparison of different measures and the potential of using a combination of instruments to improve the validity of screening. Finally, larger samples, with possibly larger outcome groups, can enable analyses of the items and identify specific behavioral patterns that are unique for prediction of ASD in preterm population.

Footnotes

Acknowledgements

We are grateful to the families who took part in this study for their cooperation.

Funding

This study was supported by the Shalem Fund, Grant No. 90 (Nurit Yirmiya), Harris Foundation (Nurit Yirmiya), Milton Rosenbaum Foundation for Psychiatric Research (S.E.F.), and Teva Pharmaceutical Industries Ltd as part of the Israeli National Network of Excellence in Neuroscience NNE (M.Y.).

References

Ben-Sasson

Carter

(2012) The application of the first year inventory for ASD screening in Israel. Journal of Autism and Developmental Disorders 42: 1906–1916.

Blank

Rose

Berlin

(1978) Preschool Language Assessment Instrument. New York: Grune & Stratton.

Bowers

Wink

Pottenger

. (2015) Phenotypic differences in individuals with autism spectrum disorder born preterm and at term gestation. Autism 19(6): 758–763.

Brian

Bryson

Garon

. (2008) Clinical assessment of autism in high-risk 18-month-olds. Autism 12: 433–456.

Brian

Zwaigenbaum

Bryson

SE, et al.

(2006) Predictive validity of the autism observation scale for infants. In: International Meeting for Autism Research (IMFAR), Montreal, QC, Canada, 1–3 June.

Bryson

Zwaigenbaum

(2014) Autism observation scale for infants. In: Patel

Preedy

Martin

(eds) Comprehensive Guide to Autism. New York: Springer, pp. 299–310.

Bryson

Zwaigenbaum

McDermott

. (2008) The autism observation scale for infants: scale development and reliability data. Journal of Autism and Developmental Disorders 38: 731–738.

Dawson

(2008) Early behavioral intervention, brain plasticity, and the prevention of autism spectrum disorder. Development and Psychopathology 20: 775–803.

Dawson

Rogers

Munson

. (2010) Randomized, controlled trial of an intervention for toddlers with autism: the Early Start Denver Model. Pediatrics 125: e17–e23.

10.

De Groote

Roeyers

Warreyn

(2006) Social-communicative abilities in young high-risk preterm children. Journal of Developmental and Physical Disabilities 18: 183–200.

11.

De Schuymer

De Groote

Beyers

. (2011a) Preverbal skills as mediators for language outcome in preterm and full term children. Early Human Development 87: 265–272.

12.

De Schuymer

De Groote

Striano

. (2011b) Dyadic and triadic skills in preterm and full term infants: a longitudinal study in the first year. Infant Behavior and Development 34: 179–188.

13.

Dudova

Kasparova

Markova

. (2014) Screening for autism in preterm children with extremely low and very low birth weight. Neuropsychiatric Disease and Treatment 10: 277–282.

14.

Elgen

Sommerfelt

Markestad

(2002) Population based, controlled study of behavioural problems and psychiatric disorders in low birthweight children at 11 years of age. Archives of Diseases in Childhood—Fetal Neonatal Edition 87: F128–F132.

15.

Feldman

Eidelman

(2007) Maternal postpartum behavior and the emergence of infant–mother and infant–father synchrony in preterm and fullterm infants: the role of neonatal vagal tone. Developmental Psychobiology 49: 290–302.

16.

Folio

Fewell

(1974) Peabody Developmental Motor Scales. 2nd ed. Austin, TX: Pro-Ed.

17.

Georgiades

Szatmari

Zwaigenbaum

. (2013) A prospective study of autistic-like traits in unaffected siblings of probands with autism spectrum disorder. JAMA Psychiatry 70: 42–48.

18.

Gerner

(1999) Emotional interaction in a group of preterm infants at 3 and 6 months of corrected age. Infant and Child Development 8: 117–128.

19.

Guy

Seaton

Boyle

. (2015) Infants born late/moderately preterm are at increased risk for a positive autism screen at 2 years of age. The Journal of Pediatrics 166(2): 269–275.e3.

20.

Hack

Taylor

Schluchter

. (2009) Behavioral outcomes of extremely low birth weight children at age 8 years. Journal of Developmental and Behavioral Pediatrics 30: 122–130.

21.

Helt

Kelley

Kinsbourne

. (2008) Can children with autism recover? If so, how? Neuropsychology Review 18: 339–366.

22.

Hofheimer

Sheinkopf

Eyler

(2014) Autism risk in very preterm infants—new Answers, more questions. The Journal of Pediatrics 164: 6–8.

23.

Inada

Kamio

Koyama

(2010) Developmental chronology of preverbal social behaviors in infancy using the M-CHAT: baseline for early detection of atypical social development. Research in Autism Spectrum Disorders 4: 605–611.

24.

Indredavik

Vik

Heyerdahl

. (2004) Psychiatric symptoms and disorders in adolescents with low birth weight. Archives of Disease in Childhood—Fetal Neonatal Edition 89: F445–F450.

25.

Johnson

Myers

(2007) Identification and evaluation of children with autism spectrum disorders. Pediatrics 120: 1183–1215.

26.

Johnson

Marlow

(2009) Positive screening results on the Modified Checklist for Autism in Toddlers: implications for very preterm populations. The Journal of Pediatrics 154: 478–480.

27.

Johnson

Marlow

(2011) Preterm birth and childhood psychiatric disorders. Pediatric Research 69: 11R–18R.

28.

Johnson

Marlow

(2014) Growing up after extremely preterm birth: lifespan mental health outcomes. Seminars in Fetal and Neonatal Medicine 19: 97–104.

29.

Johnson

Hollis

Hennessy

. (2011) Screening for autism in preterm children: diagnostic utility of the Social Communication Questionnaire. Archives of Disease in Childhood 96: 73–77.

30.

Johnson

Hollis

Kochhar

. (2010) Autism spectrum disorders in extremely preterm children. The Journal of Pediatrics 156: 525–531.e2.

31.

Keren

Feldman

Eidelman

. (2003) Clinical interview for high risk parents of premature infants (CLIP) as a predictor of early disruptions in the mother–infant relationship at the nursery. Infant Mental Health Journal 24: 93–110.

32.

Korja

Latva

Lehtonen

(2012) The effects of preterm birth on mother-infant interaction and attachment during the infant’s first two years. Acta Obstetricia et Gynecologica Scandinavica 91: 164–173.

33.

Kuban

KCK

O’Shea

Allred

. (2009) Positive screening on the Modified Checklist for Autism in Toddlers (M-CHAT) in extremely low gestational age newborns. The Journal of Pediatrics 154: 535–540.

34.

Kuzniewicz

Qian

. (2014) Prevalence and neonatal factors associated with autism spectrum disorders in preterm infants. The Journal of Pediatrics 164: 20–25.

35.

Larsson

Eaton

Madsen

. (2005) Risk factors for autism: perinatal factors, parental psychiatric history, and socioeconomic status. American Journal of Epidemiology 161: 916–925.

36.

Leavey

Zwaigenbaum

Heavner

. (2013) Gestational age at birth and risk of autism spectrum disorders in Alberta, Canada. The Journal of Pediatrics 162: 361–368.

37.

Limperopoulos

Bassan

Sullivan

. (2008) Positive screening for autism in ex-preterm infants: prevalence and risk factors. Pediatrics 121: 758–765.

38.

Lord

Rutter

DiLavore

. (2002) Autism Diagnostic Observation Schedule. 2nd ed. Los Angeles, CA: Western Psychological Services.

39.

Luyster

Gotham

Guthrie

. (2009) The autism diagnostic observation schedule-toddler module: a new module of a standardized diagnostic measure for autism spectrum disorders. Journal of Autism and Developmental Disorders 39: 1305–1320.

40.

Luyster

Kuban

KCK

O’Shea

. (2011) The Modified Checklist for Autism in Toddlers in extremely low gestational age newborns: individual items associated with motor, cognitive, vision and hearing limitations. Paediatric and Perinatal Epidemiology 25: 366–376.

41.

Montirosso

Borgatti

Trojan

. (2010) A comparison of dyadic interactions and coping with still-face in healthy pre-term and full-term infants. British Journal of Developmental Psychology 28: 347–368.

42.

Moore

Johnson

Hennessy

. (2012) Screening for autism in extremely preterm infants: problems in interpretation. Developmental Medicine and Child Neurology 54: 514–520.

43.

Msall

(2010) Central nervous system connectivity after extreme prematurity: understanding autistic spectrum disorder. The Journal of Pediatrics 156: 519–521.

44.

Mullen

(1995) Mullen Scales of Early Learning. San Antonio, TX: Pearson.

45.

Pierce

Carter

Weinfeld

. (2011) Detecting, studying, and treating autism early: the one-year well-baby check-up approach. The Journal of Pediatrics 159: 458–465.e6.

46.

Pinto-Martin

Levy

Feldman

. (2011) Prevalence of autism spectrum disorder in adolescents born weighing <2000 grams. Pediatrics 128: 883–891.

47.

Reznick

Baranek

Reavis

. (2007) A parent-report instrument for identifying one-year-olds at risk for an eventual diagnosis of autism: the first year inventory. Journal of Autism and Developmental Disorders 37: 1691–1710.

48.

Robins

Fein

Barton

. (2001) The Modified Checklist for Autism in Toddlers: an initial study investigating the early detection of autism and pervasive developmental disorders. Journal of Autism and Developmental Disorders 31: 131–144.

49.

Rushing

Ment

(2004) Preterm birth: a cost benefit analysis. Seminars in Perinatology 28: 444–450.

50.

Schendel

Bhasin

(2008) Birth weight and gestational age characteristics of children with autism, including a comparison with other developmental disabilities. Pediatrics 121: 1155–1164.

51.

Singer

Fulton

Davillier

. (2003) Effects of infant risk status and maternal psychological distress on maternal-infant interactions during the first year of life. Journal of Developmental & Behavioral Pediatrics 24: 233–241.

52.

Sparrow

Cicchetti

Balla

(2005) Vineland Adaptive Behavior Scales: Vineland-II. 2nd ed. Circle Pines, MN: AGS Publishing.

53.

Stephens

Bann

Watson

. (2012) Screening for autism spectrum disorders in extremely preterm infants. Journal of Developmental and Behavioral Pediatrics 33: 535–541.

54.

Sugie

Fukuda

. (2005) Neonatal factors in infants with autistic disorder and typically developing infants. Autism 9: 487–494.

55.

Treyvaud

Ure

Doyle

. (2013) Psychiatric outcomes at age seven for very preterm children: rates and predictors. Journal of Child Psychology and Psychiatry 54: 772–779.

56.

Vanderbilt

Gleason

(2011) Mental health concerns of the premature infant through the lifespan. Pediatric Clinics of North America 58: 815–832.

57.

Wong

Huertas-Ceballos

Cowan

. (2014) Evaluation of early childhood social-communication difficulties in children born preterm using the Quantitative Checklist for Autism in Toddlers. The Journal of Pediatrics 164: 26–33.

58.

Yirmiya

Charman

(2010) The prodrome of autism: early behavioral and biological signs, regression, peri- and post-natal development and genetics. Journal of Child Psychology and Psychiatry 51: 432–458.

59.

Zwaigenbaum

Bryson

Garon

(2013) Early identification of autism spectrum disorders. Behavioural Brain Research 251: 133–146.

60.

Zwaigenbaum

Bryson

Lord

. (2009) Clinical assessment and management of toddlers with suspected autism spectrum disorder: insights from studies of high-risk infants. Pediatrics 123: 1383–1391.

61.

Zwaigenbaum

Bryson

Rogers

. (2005) Behavioral manifestations of autism in the first year of life. International Journal of Developmental Neuroscience 23: 143–152.