Prevalence of Autism Spectrum Disorder in Adolescents Born Weighing <2000 Grams

Abstract

OBJECTIVE:

To estimate the diagnostic prevalence of autism spectrum disorders (ASDs) in a low birth weight (LBW) cohort.

METHODS:

Participants belonged to a regional birth cohort of infants (N = 1105) born weighing <2000 g between October 1, 1984, and July 3, 1989, and followed up by periodic assessments to 21 years of age. At 16 years (n = 623), adolescents were screened for ASD using a wide net (previous professional diagnosis of an ASD or a score above a liberal cutoff on the Social Communication Questionnaire or the Autism Spectrum Symptoms Questionnaire). At 21 years (n = 189), 60% of screen positives and 24% of screen negatives were assessed for diagnoses of ASD by the Autism Diagnostic Observation Schedule or the Autism Diagnostic Interview–Revised.

RESULTS:

Samples retained at ages 16 and 21 years were representative of samples assessed at earlier ages except for lower levels of social risk. Of positive screens, 11 of 70 had ASD; of negative screens, 3 of 119 had ASD. The fractions of the 2 screening groups with ASD (14.3% in screen-positives and 2.5% in screen negatives) were weighted by fractions of screen-positives and screen-negatives among the adolescents (18.8% and 81.2%, respectively). This calculation produced an estimated prevalence rate of ASD in the entire cohort of 5% (31 of 623).

CONCLUSIONS:

The diagnostic prevalence of ASD in this LBW preterm cohort was higher than that reported by the Centers for Disease Control and Prevention for 8-year-olds in the general US population in 2006.

WHAT'S KNOWN ON THIS SUBJECT:

Low birth weight and/or prematurity are considered to be risk factors for autism spectrum disorders, largely on the basis of retrospective studies or prospective studies that have used screening rather than diagnostic instruments.

WHAT THIS STUDY ADDS:

This is the first study of the prevalence of autism spectrum disorders in a prospectively followed low birth weight population using validated diagnostic instruments. The 5% estimated prevalence rate found was ∼5 times the rate reported in the general population.

Low birth weight (LBW) and prematurity are established risk factors for cognitive and motor disability.^1,2 The prevalence of neurodevelopmental impairment increases with diminishing birth weight and gestational age.³ As survival of the smallest and most immature infants improves, impaired survivors represent an increasing public health challenge.

Emerging evidence suggests that LBW may also be a risk factor for autism spectrum disorders (ASDs). Retrospective studies have suggested an association between LBW and/or preterm birth and risk for ASD.^4–10 Prospective studies in children^11–13 and adolescents^3,14–16 have supported this suggestion. Most prospective studies, however, have determined risk for ASD on the basis of screening instruments rather than a definitive diagnosis. Only 1 prospective study to date has assessed the actual diagnostic prevalence of ASD in children born prematurely.^17,18 In that study, 219 parents of 11-year-olds born at the extremely low gestational age of <26 weeks in the United Kingdom and Ireland were interviewed by telephone or responded to an online survey of a multipurpose structured psychiatric interview, the Development and Well-being Assessment (DAWBA).¹⁹ Using this instrument, the prevalence of ASD in births <26 weeks was reported to be 8%. Such extreme prematurity, however, affects <0.25% of all US births.²⁰

This report presents the screening and diagnostic prevalence of ASD in a large population-representative cohort of adolescents born weighing between 500 and 2000 g, a weight interval that represents ∼3% of US births.²¹

METHODS

Neonatal Brain Hemorrhage Study Cohort

The study sample was drawn from the Neonatal Brain Hemorrhage Study (NBHS) cohort, consisting of all infants admitted to 3 hospitals that together cared for 85% of births <2000 g in 3 central New Jersey counties from October 1, 1984, to July 30, 1987.²² At that time, these 3 counties were demographically comparable to the nation as a whole, except for having a slightly higher per capita income and slightly lower proportion of minorities.²² This cohort was prospectively enrolled at birth and reassessed at ages 2,^23,24 6,^25,26 9,^27,28 16,^29,30 and 21 years. All enrolled families spoke English.

Two-Stage Design and Sampling Methods

The prevalence of ASD in the NBHS cohort was estimated by using a 2-stage design. In the first stage, a screening for ASD using parent questionnaires was conducted at the 16-year follow-up. In the second stage, at the 21-year follow-up, a diagnostic assessment for ASD was conducted with young adults recruited from all those who screened positive in the first stage and from a systematically obtained sample of those who screened negative in the first stage. Lists of screened negative potential recruits were initially selected at random in batches on the basis of year of birth (to ensure minimum appropriate age of participants), stratified according to gender to match the male:female ratio of screen-positives for that birth year. When numbers of screen-negative recruits fell short for a given birth year, additional recruits were drawn from all remaining screen-negatives of the appropriate gender without respect to birth year.

First Stage: Screening Procedure for ASD (at 16 Years of Age)

As part of a larger psychiatric follow-up,²⁹ adolescents were assessed for symptoms of ASD by using the Social Communication Questionnaire (SCQ), previously known as the Autism Screening Questionnaire,³¹ and the Autism Spectrum Screening Questionnaire (ASSQ).³² Adolescents' parents were also asked whether a professional had ever diagnosed their child with an ASD (autism, Asperger syndrome, or pervasive developmental disorder [not otherwise specified]). Screening data were obtained primarily by questionnaires collected at a home visit (499 [80%]) and the remainder by mail and/or telephone (127 [20%]).

A participant in the 16-year follow-up was considered “screen-positive” if he or she met any of the following criteria: (1) a score of ≥9 on the SCQ; (2) a score of ≥12 on the ASSQ; or (3) a history of professional diagnosis of ASD. The screen was designed to cast as wide a net as possible, so the cutoff points used were lower than the customary cutoff points of 15 for the SCQ³¹ and 22 for the ASSQ.³²

Second Stage: Recruitment and Diagnostic Procedure

Young adults were located by using tracking information available from earlier follow-ups; when necessary, Accurint, a national tracking system, was used. In some cases, parents facilitated contact.

The diagnostic instruments used were: (1) the Autism Diagnostic Interview–Revised (ADI-R)³³ (administered to parents); and (2) the Autism Diagnostic Observation Schedule (ADOS, module 4) (administered to the young adults).³⁴ Both the ADIR and the ADOS are well-validated instruments widely used in research. The validity of these tools may be compromised in the circumstance of severe to profound disability (eg, mental age < 2 years).^35,36 In this circumstance, additional information obtained from clinical history can be combined with information from the ADI-R and/or ADOS to inform a best-estimate diagnosis using the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition criteria.³⁷ Two research associates, who were blind to screening status and research reliable in the administration and scoring of the diagnostic instruments, completed the assessments under the supervision of the director of the Regional Autism Center (Dr Levy). In-person assessments took place at the Children's Hospital of Philadelphia or at the Children's Hospital of Philadelphia Specialty Care Center in Princeton, New Jersey. For the ADIR, scores needed to be above published threshold scores on each of 4 domains: (1) reciprocal social interaction; (2) communication; (3) restricted, repetitive, stereotyped patterns of behavior; and (4) abnormality of development evident at or before 36 months.³⁸ For the ADOS, the total score on the communication and social domains needed to be above the published threshold for an ASD.³⁹

Of the total 189 patients evaluated at 21 years of age, data were available from both the ADOS and the ADI-R on 144 (76%). An additional 75 had data from either the ADOS alone (n = 10) or the ADI-R alone (n = 65). Most of those evaluated with the ADIR alone (n = 62) had distance or scheduling issues that precluded an in-person ADOS evaluation; the remaining 3 patients were evaluated only by parental interview using the ADI-R because they were severely disabled.

Human Subjects Consideration

This study was approved by the institutional review board of the Children's Hospital of Philadelphia and the New York State Psychiatric Institute. Consent was obtained from the participating young adults (or their legal guardians, as applicable) and their families.

Statistical Analysis

Two sets of analyses addressed attrition issues. The first set compared those who were screened as adolescents to the remaining eligible adolescents on basic sample characteristics shown in Table 1 using t tests for continuous measures and χ² tests for categorical ones. The second set determined whether sample loss between adolescence and young adulthood was conditional on ASD screening status. These analyses were conducted using 2 (positive versus negative screen status) by 2 (retained in sample versus not retained) analysis of variances (for continuous dependent measures), logistic regressions (for dichotomies), and multinomial regressions (for categorical variables having >2 categories). Significant interactions and main effects are reported; consistent with general statistical practice, follow-up comparisons between retention groups within screen status strata were only undertaken where there were significant interactions.

TABLE 1.

Sample Characteristics: LBW Adolescents Screened for ASD at 16 Years of Age and Others Eligible but Not Screened for ASD at 16 Years of Age

	16-y-Old Sample Screened for ASD (n = 623)	Others Eligible at 16 y of Age (n = 239)a
Maternal social risk, %b,c	43.9	65.1
Minority status, %b	27.0	40.6
Less than high school education, %b,d	12.0	33.5
Age younger than 19 y, %b,e	5.6	16.0
Unmarried, %b,f	22.9	46.4
Receiving public assistance, %b,g	21.6	41.7
Male gender, %	49.3	52.3
Birth weight, mean ± SD, g	1475.3 ± 353.3	1492.7 ± 377.9
≥1500, %	52.5	54.4
1000–1499, %	35.3	32.2
<1000, %	12.2	13.4
Gestational age, mean ± SD, completed wk	31.2 ± 3.1	31.3 ± 3.3
>36 wk, %	3.7	4.2
34–36 wk, %	19.3	18.9
32–33 wk, %	22.3	24.4
28–31 wk, %	42.1	40.3
<28 wk, %	12.7	13.2
Small for gestational age, %h	32.3	34.7
Bronchopulmonary dysplasia, %i	5.3	4.2

^aSurviving and nonadopted LBW cohort members at age 16 years who were lost to follow-up (n = 151), who refused participation (n = 83), or who were assessed at age 16 years but were missing all 3 components of the ASD screen (n = 5). Of these last 5 patients, 3 were disabled mentally (IQ < 55) and/or could not walk without assistance. Thirty-one survivors were not eligible to be seen at age 16 years because they had been adopted.

^bAge 16 years sample screened for ASD versus others eligible at age 16 years: P ≤ .001, unadjusted for multiple tests.

^cMother's social risk at time of participant's birth. One or more of the components listed in the 5 successive rows.⁴⁷ When ≥1 components of risk were missing, the total was pro-rated by multiplying 5 by the proportion of nonmissing risk factors that were positive for risk.

^dNot present for all cases; column n values are 559 and 182, respectively.

^eNot present for all cases; column n values are 622 and 238, respectively.

^fNot present for all cases; column n values are 538 and 168, respectively.

^gNot present for all cases; column n values are 509 and 168, respectively.

^hLowest 10th percentile of weight for weeks of postconception age, using norms of Yudkin et al.⁴⁸

ⁱDefined as the need for oxygen and/or ventilatory support at 36 weeks' postmenstrual age. Not present for all cases; column n values are 620 and 238, respectively.

Prevalence in the entire cohort was calculated by weighting the proportions of screen-positives and screen-negatives found to have ASD as young adults by the proportion of screen-positives and screen-negatives in the adolescent sample.

The relation of the ASD diagnosis to basic sample characteristics was examined by using χ² statistics. SPSS software⁴⁰ was used for all analyses; the criterion for statistical significance was set at a 2-tailed α value of 0.05.

RESULTS

Sample Attrition

Figure 1 shows the attrition of the sample from birth to young adulthood. By 16 years of age, 862 adolescents were considered eligible for follow-up; of these, 628 participated. All but 5 participating families completed the autism screening instruments at that time, leaving a final sample of 623 (72%; Fig 1). Of these, 80.7% completed all 3 parts of the screen and very few (8.0%) completed only 1 part. At 21 years of age, 60% of the screen-positive group (n = 70) was evaluated for an ASD diagnosis at 21 years of age (Fig 1), whereas 40% (n = 47) were lost to follow-up because of inability to contact (n = 11), inability to schedule because of time/distance constraints (n = 27), or refusal (n = 9).

FIGURE 1.

Sample attrition.

Table 1 shows the characteristics of adolescents screened for ASD and those eligible but not screened at 16 years of age. The 2 groups differed significantly only with respect to lower maternal social risk at birth in those screened.

Table 2 shows the characteristics of those who were and were not retained in the sample from 16 to 21 years of age according to screen status. These data address the possibility that characteristics of those retained (versus not retained) were dependent on screen status. With the exception of 1 component of maternal birth social risk (ie, receipt of public assistance), this was not the case. For this 1 variable, there was a significant interaction between screen status and sample retention (Wald χ₂¹ = 4.92, P = .027) such that there was a greater discrepancy between retained and not-retained among the screen-positives, although those retained had lower percentages of those receiving public assistance in both groups (Table 2 legend).

TABLE 2.

Birth Risk Factors and 16-Year Neurodevelopmental Status in Screen-Positives and Screen-Negatives According to Sample Retention From 16 to 21 Years

	Screened Positive at 16 y of Age		Screened Negative at 16 y of Age
	Retained in Sample (n = 70)	Not Retained in Sample (n = 47)	Retained in Sample (n = 119)	Not Retained in Sample (n = 387)
Perinatal risk factors
HUS status
NA, % (reference category)	58.6	66.0	84.0	81.1
Germinal matrix and/or intraventricular hemorrhage, %	18.6	14.9	14.3	14.0
Parenchymal lesion and/or ventricular enlargement, %	22.9	19.1	1.7	4.9
Birth social risk: ≥1 of 5 risk conditions, %a	38.6	73.9	26.1	46.8
Minority status, %	18.6	53.2	11.8	30.0
Less than high school education, %b	11.9	26.2	6.5	12.0
Age < 19 y, %	2.9	10.9	1.7	6.7
Unmarried, %c	15.4	42.5	8.2	26.3
Receiving public assistance, %d	16.1	51.3	14.7	21.1
Male gender, %	54.3	59.6	56.3	45.0
Birth weight, mean ± SD, g	1420 ± 395	1404 ± 370	1507 ± 348	1484 ± 344
≥1500 g, % (reference category)	50.0	44.7	55.5	53.0
1000–1499 g, %	30.0	40.4	33.6	36.2
<1000 g, %	20.0	14.9	10.9	10.9
Gestational age, mean ± SD, completed wk	31 ± 3	31 ± 3	31 ± 3	31 ± 3
>36 wk, % (reference category)	5.7	0.0	1.7	4.4
34–36 wk, %	20.0	17.0	21.0	18.9
28–31 wk, %	48.6	38.3	42.0	41.3
<28 wk, %	15.7	14.9	16.0	10.9
Small for gestational age, %e	34.3	31.9	22.7	34.9
Bronchopulmonary dysplasia (%)f	11.6	6.4	5.0	4.2
Neurodevelopment at age 16 y
Disabled cognitively or motorically, %g	18.6	23.4	0.8	1.3
Inability to walk without assistance, %h	12.9	12.8	0.8	1.3
Severe intellectual impairment, %i	14.3	23.4	0.0	0.8

Significant interaction between screen status and retention status. The only significant interaction involved a greater disparity between retained and not retained in the percentage of mothers who received public assistance at the time of their child's birth among the screen-positives than among the screen-negatives (Wald χ₂¹ = 4.92, P = .027). In the screen-positives, the percentage of mothers receiving assistance in those retained was 35.2 percentage points lower than in those not retained (χ₂¹ = 14.17, P < .001), whereas among the screen-negatives this percentage was 6.4 percentage points lower in those retained than in those not retained (χ₂¹ = 1.86, P = .172).

Significant main effects for retention in sample, irrespective of screen status. Those not retained (n = 434) compared with those retained (n = 189) were more likely to be male and small for gestational age (55.6% vs 46.5%, and 34.6% vs 27.0%, respectively; both P < .05). They were also more likely to have ≥1 social risk factors (49.7% vs 30.7%; P < .001). Of the social risk factors, those not retained were more likely to be of minority status (32.5% vs 14.3%; P < .01) and to be unmarried (28.0% vs 11.0%; P < .001).

Significant main effects for screen status, irrespective of retention in sample. Those who screened positive at age 16 years (n = 117) compared with those who screened negative (n = 506) had slightly lower mean birth weights (1413.6 vs 1489.6 g; P < .05) and were more likely to have parenchymal lesions or ventricular enlargement as determined by neonatal head ultrasound (21.4% vs 4.2%; P < .001). At the time of the adolescents' birth, the mothers of those who screened positive were more likely than the mothers of those who screened negative to have ≥1 social risk factors (52.6% vs 41.9%; P < .001). They were more likely to be a member of a minority (32.5% vs 25.7%; P < .01), to have less than a high school education (17.4% vs 10.7%; P < .05), to be unmarried (25.7% vs 22.2%; P < .05), and to be receiving public assistance (29.7% vs 19.6%; P < .001). The adolescents were also more likely to be disabled (either cognitively or motorically) and more likely to be nonambulatory and severely mentally disabled (20.5% vs 1.2%, 12.8% vs 1.2%, and 18.8% vs 0.6%, respectively; all, P <0.001) at age 16 years. Males were only marginally more common among the screen-positives (56.4%) than among the screen-negatives (47.6%) (P < .10).

NA indicates not available.

^aComponents are listed in the 5 succeeding rows. When ≥1 components of risk were missing, the total was pro-rated by multiplying 5 by the proportion of nonmissing risk factors that were positive for risk.

^bNot present for all cases; column n values are 67, 42, 107, and 343, respectively.

^cNot present for all cases; column n values are 65, 40, 93, and 335, respectively.

^dNot present for all cases; column n values are 62, 39, 95, and 313, respectively

^eIn the lowest decile of the distribution of weight for gestational age using the gender-specific norms of Yudkin et al.⁴⁸

^fDefined as the need for oxygen and/or ventilatory support at 36 weeks postmenstrual age. Not present for all cases; column n values are 69, 47, 119, and 385, respectively.

^gSevere intellectual impairment or inability to walk without assistance.

^hItem 5 of the Activities Limitations Questionnaire.⁴⁹

ⁱFor those seen in person at age 16 years, Wechsler Abbreviated Scale of Intelligence, Full-Scale IQ (FSIQ)⁵⁰ <55 or untestable and ≥2 domains of the Vineland Adaptive Behavior Scale (VABS)⁵¹ scored as low or moderately low; for those assessed by telephone interview at that age, average of FSIQ scores from testing at 6 and/or 9 years <55 and ≥2 subdomains of the VABS scored as low or moderately low, or when FSIQ data were unavailable from earlier waves of testing, a composite score of the VABS from the telephone interview <55.

Significant main effects of retention, irrespective of screen status, were all in the expected direction, namely that those not retained in the sample at 21 years of age (n = 434) were more likely than those retained in the sample (n = 189) to be at greater risk for suboptimal neurodevelopmental outcomes (Table 2 legend).

Significant main effects of screen status, irrespective of sample retention, were similarly all in the same direction. Those who screened positive at 16 years of age (n = 117) compared with those who screened negative (n = 506) were more likely to be at greater risk for suboptimal neurodevelopmental outcomes (Table 2 legend). Notably, however, boys were only marginally more common among the screen-positives (56.4%) than among the screen-negatives (47.6%; P < .10), indicating that the gender matching of screen-negatives to screen-positives was fairly successful. Another important characteristic of the screen positive and screen-negative samples (data not shown in Table 2)—namely, the percentage of families in each group who were interviewed by telephone because of scheduling and distance issues—also did not differ across the groups (32.9% and 34.5%, respectively).

Screening Prevalence at 16 Years of Age

Of the 623 adolescents screened for ASD, 117 (18.8%) were screen-positive by 1 or more of the 3 screening criteria. Of the 117 screen-positives, 62 were positive on the SCQ alone, 26 were positive on the ASSQ alone, and 6 were positive solely by parent report of a professional diagnosis. Sixteen were positive on 2 parts, and 7 were positive on all 3 parts of the screen.

Diagnostic Prevalence of ASD

Of the 70 young adults examined who were screen-positive as adolescents, 11 were found to have ASD (Table 3). Of the 119 screen-negatives, 3 were found to have ASD.

TABLE 3.

Estimated Prevalence of ASD in Total Cohort (N = 623) Based on Findings in 198 Participants With Diagnostic Assessment at 21 Years of Age

	ASD Prevalence in Young Adults by Screening Status, % (n/N)	Screening Status Prevalence in Total Cohort, % (n/N)	Estimated n and Prevalence of ASD in Total Cohort
Screen-positive	15.7 (11/70)	18.8 (117/623)	18 (15.7% × 117)
Screen-negative	2.5 (3/119)	81.2 (506/623)	13 (2.5% × 506)
Total	7.0 (14/189)	—	5.0% (31/623)

To estimate the prevalence of ASD cases in the original sample of 623 adolescents, the numbers of young adults found to have ASD in the 2 screening groups (14.3% among the screen-positives and 2.5% among the screen-negatives) were weighted by the fractions of screen-positives and negatives among the adolescents (18.8% and 81.2%, respectively) as shown in Table 3. Thus, the best estimate of the prevalence of ASD in the entire adolescent LBW cohort is 5.0% (31 of 623).

Characteristics of the Cases

Of the 14 cases, 3 were screen-negative as adolescents and 11 were screen-positive (Table 4). All 14 cases had an ADI-R completed by a parent. Six did not have ADOS scores; 5 of these could only be interviewed by telephone, precluding an ADOS assessment, and 1 was assessed using the ADOS, but the results were considered invalid because of physical disability. Of the 8 cases with a completed ADOS, 2 met the threshold on both the ADI-R and ADOS, 1 on only the ADI-R, and the rest on only the ADOS.

TABLE 4.

Characteristics of the ASD Cases

Case No.	21-y Assessment				Gender	16-y Assessment
	ADI-R	ADOS	Age, moa			FSIQ	VABSC	Riley Scoreb	Other DSM-IV Diagnosesc
			First Word	First Phrase
Screen-negatives
1	No diagnosis	ASD	<24	<33	Male	118	92	0	None
2	No diagnosis	ASD	21	21	Male	115	85	8	ODD
3	No diagnosis	ASD	17	19	Male	127	95	2	None
Screen-positives
4	ASD	No dx	<24	Unknown	Female	112	65	11	None
5	No diagnosis	Autism	18	24	Female	99	81	12	Motor-vocal tics
6	ASD	Autism	30	45	Male	97	64	9	Tourette's syndrome
7	ASD	Not testedd	36	60	Male	77	45	7	Transient tic disorder; specific phobia
8	ASD	Not testedd	≥24	36	Male	76	56	7	ADHD; ODD; CD; social phobia
9	ASD	ASD	15	36	Female	72	64	7	ADHD
10	No diagnosis	ASD	36	48	Male	68	74	6	none
11	ASD	Not testedd,e	60	60	Male	Not testedd,e	39	—d,e	Not assessedd,e
12	ASD	Not testede	36	36	Male	Not testede	32	—e	Not assessede
13	ASD	Not testedd	18	30	Male	Not testede	54	—d	Not assessedd
14	ASD	Not testedd	36	45	Male	Not testede	50	—d	Not assessedd

FSIQ indicates full-scale IQ (from the WASI); VABSC, Vineland Adaptive Behavior Scale Composite⁵¹; DSM-IV, Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition; ODD, oppositional/defiant disorder; ADHD, attention deficit hyperactivity disorder; CD, conduct disorder.

^aBased on ADIR scoring for items 9 (first word) and 10 (first phrase) where “normal” is <24 and <33 months, respectively.

^bScores ≥5 are in the top 10% of the Riley Motor Problems⁵² distribution; they indicate some motor impairment.

^cAscertained at age 16 years with the Diagnostic Interview Schedule for Children/Parent Version–IV, using the threshold algorithm,³⁰

^dAssessed by telephone; no in-person assessments at this age.

^eSevere motor impairment; nonambulatory.

The overall male to female gender ratio was 3.7:1 (11 boys, 3 girls). Sixty-four percent (9 of 14) were relatively high functioning and verbal (ie, phrase speech and IQ ≥70). Of the 11 cases who were screen-positive in adolescence, 6 were reported to have been diagnosed with ASD by a professional and 5 had not. For these 11 subjects, scores on the Weschler Abbreviated Scale of Intelligence at 16 years of age ranged widely, from 50 (the default score for those too cognitively impaired to be tested) to 112; their Vineland scores also varied widely, ranging from 32 to 81. By contrast, of the 3 cases who screened negative in adolescence, all IQ scores were at least 1 SD above the mean (≥115), whereas Vineland composite scores ranged from 85 to 95. None of these 3 cases met threshold on the ADI-R. Six of the 14 cases (42.9%) met criteria for at least 1 other psychiatric disorder at 16 years of age on the basis of a validated psychiatric interview.³⁰

None of the basic sample characteristics in Table 1 were significantly related to a diagnosis of ASD, although there were marginal relations for gender and birth weight (P < .10). The percentage of boys having a diagnosis of ASD (9.9%) was larger than that of girls having a diagnosis (3.3%; χ₂¹ = 3.32, P = .07). In addition,10.6% of those born weighing <1500 g had an ASD diagnosis compared with only 3.7% of those born weighing between 1500 and 2000 g (χ₂¹ = 3.68, P = .055).

DISCUSSION

This is the first study to have estimated the prevalence of ASD in a prospectively followed LBW population using research-validated diagnostic instruments. ASD prevalence in this US population-representative cohort of LBW (<2000 g) adolescents was estimated to be 5%, a fivefold increase over that reported by the Centers for Disease Control and Prevention for 8-year-olds in the US general population in 2006 (0.9%).⁴¹ In addition, we found that 18.8% of the adolescents screened positive for ASD. Screening criteria were designed to maximize sensitivity, contributing to the high false-positive rate encountered. Nonetheless, 2.5% of those with ASD, notably high-functioning boys, were not picked up on screening.

As a group, the 14 subjects with ASD diagnosed in young adulthood seemed to be relatively high functioning in terms of IQ and spoken language. The proportion (63%) with an IQ of ≥70 (at least borderline intelligence) was higher than that of a recent population estimate of 45% for school-aged cases of ASD.⁴² All of the cases here had developed phrase speech, which has recently been reported as absent in 10% to 20% of persons with ASD.⁴³ One possible reason for these apparent differences is that ASD associated with prematurity has a distinctive neurodevelopmental profile. Alternatively, a number of additional low-functioning ASD cases might have been diagnosed had the retention rate in the screen-positive group been greater. The greater retention of those with less overall social risk is consistent with this possibility. Addressing these alternative possibilities is beyond the scope of this report.

Prospective studies have reported high screening prevalence in LBW and/or premature toddlers ranging from 21%^13,44 to 37%¹² using the Modified Checklist for Autism in Toddlers. In prospective cohorts of older children (11 and 14–15 years), the screening prevalence was not quite so high (<1% to 3.7%), but some studies did not use autism-specific screens.^16,45,46 Johnson et al¹⁷ recently reported screening and diagnostic prevalence of ASD of their prospective cohort of all births <26 weeks in 1995 in the United Kingdom and Ireland. Of the survivors, 71% were assessed at 11 years using the SCQ screening instrument and a clinical best estimate diagnostic package, the DAWBA. Screening prevalence was 15.8%, comparable to that found in our study. Diagnostic prevalence was 8%. It is not clear whether restriction of the Johnson et al sample to extremely low gestational age survivors and/or features of the DAWBA account for the higher diagnostic prevalence rate in their study.

To date, the only epidemiologic study to estimate prevalence rates of ASD in adults born prematurely is that of Moster et al.² Using linked compulsory national registries in Norway, they found that the rates of disability payments for ASD among adults 20 to 36 years old in 2003 (born between 1967 and 1983) rose significantly as gestational ages declined. Rates in that study rose from 0.05% among those born at term (gestational age ≥ 37 weeks) to 0.6% among those born at gestational ages of 23 to 27 weeks. Although these rates were considerably lower than the 5% found here, the inverse relation of gestational age to ASD of sufficient severity to warrant disability payments in adulthood highlights the long-term public health impact of prematurity on ASD-associated disability.

Strengths

In this study, rigorous diagnostic procedures were used to estimate the diagnostic prevalence of ASD in AN LBW population. Given that the present cohort was not restricted to those at the lower extremes of birth weight and gestational age, generalizability of our findings is broader than is often the case in LBW/preterm follow-up studies. Moreover, because this cohort is population based, findings are likely to have greater generalizability than those from hospital-based cohorts.

Limitations

The proportion lost to follow-up over the 5-year period between the adolescent and young-adult follow-up was relatively high. However, the differences in characteristics between those retained in the young adult sample versus those not retained were, for the most part, the same among those who screened positively and negatively in adolescence. Some adolescents did not receive all parts of the 3-part screen. In addition, not all young adults were assessed using both the ADIR and ADOS. ASD diagnosis at an earlier follow-up age might have increased participant availability for the in-person ADOS evaluations. Finally, the lack of a normal birth weight control group allowed comparison only with general population estimates of ASD prevalence.

CONCLUSIONS

The estimated prevalence of ASD diagnoses in this LBW cohort was 5 times the prevalence reported by the Centers for Disease Control and Prevention for 8-year-olds in the general US population in 2006. This prospective study, using rigorous diagnostic procedures, confirms that the rate of ASD is elevated among LBW/preterm survivors.