BACKGROUND: This study examined mental health outcomes in extremely preterm children (EPT) born at 23 to 25 weeks of gestation between 1992 and 1998 at 2 Swedish tertiary care centers that offered regional and active perinatal care to all live-born EPT infants.
METHODS: We assessed 132 (98%) of the 134 EPT survivors at 10 to 15 years of age alongside term-born controls. Behavioral and emotional problems were evaluated by using Achenbach’s Child Behavior Checklist and Teacher Report Form and Conners’ Parent and Teacher scales for attention-deficit/hyperactivity disorder.
RESULTS: Parents and teachers reported significantly more problems with internalizing behaviors as well as attention, social, and thought problems in EPT children than in controls, even when those with major neurodevelopmental disabilities (NDDs) were excluded. Multivariate analysis of covariance of the behavioral problems reported by parents and teachers revealed no interactions, but significant main effects emerged for group status (EPT versus control) and sex, with all effect sizes being medium to large and accounting for 8% to 14% of the variance. Compared with the controls, EPT children without NDDs had significantly increased rates of ≥90th percentile for total Conners’ attention-deficit/hyperactivity disorder problem scores (parents: 40% vs 15%, odds ratio: 3.7, P < .001) (teachers: 24% vs 9%, odds ratio: 3.3, P = .005). The corresponding rates were higher in the total population.
CONCLUSIONS: EPT children with or without NDDs had behavioral problems characterized by a higher risk for anxiety and attention, social, and thought problems. These findings further strengthen the proposition that a preterm behavioral phenotype is recognizable in adolescents born EPT.
- ADHD —
- attention-deficit/hyperactivity disorder
- ANCOVA —
- analysis of covariance
- ASD —
- autism spectrum disorders
- CBCL —
- Achenbach’s Child Behavior Checklist
- CPRS —
- Conners’ Parent Rating Scale
- CTRS —
- Conners’ Teacher Rating Scale
- DSM-IV-TR —
- Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision
- ELBW —
- extremely low birth weight
- EPT —
- extremely preterm
- ES —
- effect size
- MANCOVA —
- multivariate analysis of covariance
- NDD —
- neurodevelopmental disability
- NSI —
- neurosensory impairment
- OR —
- odds ratio
- TRF —
- Achenbach’s Teacher Report Form
- VLBW —
- very low birth weight
What’s Known on This Subject:
There is ample evidence that extremely low birth weight children face an increased risk of mental health issues throughout early childhood (namely attention, hyperactivity, emotional, and social problems), but few studies have focused on adolescents born extremely preterm, especially after active perinatal care.
What This Study Adds:
This study adds important evidence about a behavioral phenotype in adolescents born extremely preterm, which is characterized by specific risks for attention, anxiety, social problems, and autism-related symptoms.
Behavioral and emotional problems such as inattention, hyperactivity, anxiety, and poor social and peer relations are increasingly being reported in extremely low birth weight (ELBW)1–5 and extremely preterm (EPT) children throughout childhood.6–10 However, there is a lack of studies that cover adolescence, including EPT adolescents born after the advent of modern neonatal intensive care and specifically after active perinatal care was introduced in the 1990s.
This study’s primary aim was to examine the nature, frequency, and severity of mental health problems in adolescents aged 10 to 15 years who were born at ≤25 weeks of gestation in the 1990s, after active perinatal care, and compare them with matched term-born controls. The secondary aim was to assess whether any observed differences in mental health problems persisted among EPT children without major neurodevelopmental disabilities (NDDs). We hypothesized that EPT adolescents would demonstrate a preterm behavioral phenotype11 composed of a risk of attention problems, internalizing behaviors such as anxiety and depression, and social problems.
The study population comprised survivors born EPT at the university hospitals of Uppsala and Umea in Sweden between January 1992 and December 1998. The NICUs of these perinatal referral centers serve the Uppsala region and northern Sweden. During the study period, 261 infants were born at 23 to 25 weeks of gestation, 213 were born alive, 140 survived long enough to be discharged from the hospital, and 6 died during their first year. All EPT infants born alive at 23 to 25 weeks during the study period were given life support at birth.12 We identified the remaining 134 of 140 infants at 10 to 15 years of age, and 132 (98%) were assessed at a mean age of 12 years (range: 10.1–15.9 years). Previous articles have included the characteristics of the cohort, definitions of neonatal characteristics, overall mortality and cognitive assessments, neurosensory impairment (NSI), neuropsychological outcomes, and special health care needs at 10 to 15 years of age.12–15 A major NDD was defined as one or more major NSIs or a major cognitive impairment,14 namely a full-scale intelligence quotient of <−2 SD (≤70) on the Wechsler Intelligence Scale for Children, Third Edition, revised.16
The control group was recruited from the national Swedish Medical Birth Register by selecting a term child of normal birth weight who had the same sex, was born at the same hospital, and at the closest time (7 days before or after) as the EPT child. Six matched control participants were identified for every EPT child. The parents of the first child on the list were approached. If participation was declined, the parents of the second child were approached, and, if necessary, additional parents were approached until a control participant agreed to participate or the pool of 6 children was depleted. The recruitment of controls was a difficult process, which is why only 103 (78%) control children were recruited as opposed to 1 control for each EPT child as initially planned. Of the 103 control participants, 66%, 19%, 10%, and 5% were first, second, third, and fourth invitees, respectively, from a pool of 6 matched control participants for every index child.
The recruitment procedures have been described previously.14,15 The contact details for the study families were obtained from the Swedish National Tax Board, and with their permission, they were contacted by the research nurse, who explained the procedure and study protocol. The study was approved by the regional Ethical Review Board of Umea, Sweden.
Assessment of Mental Health
The parents and teachers completed Achenbach’s Child Behavior Checklist (CBCL) and Teacher Report Form (TRF), respectively, for 6- to 18-year-olds.17 Both provide a total problem score consisting of 8 narrow-band scales. The principal components analyses reveal 8 sets of behaviors: withdrawn, somatic complaints, anxious and/or depressed, social problems, thought problems, attention problems, delinquency, and aggression. Two of the narrow-band scales (aggression and delinquency) provide a broad-band externalizing score, 3 (anxious, somatic, and withdrawn) provide a broad-band internalizing score, and 3 (social, thought, and attention problems) indicate difficulties that do not fit either broad-band dimension.
The CBCL and TRF scales measure 6 sets of behavioral syndromes on the basis of the criteria in the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR),18 namely affective disorders, anxiety, somatic problems, attention-deficit/hyperactivity disorder (ADHD), conduct disorder, and oppositional defiant disorder.
The scores for all CBCL and TRF problem subscales are in the abnormal range if they are >93rd percentile for the narrow bands and >90th percentile for the broad bands of the control subjects of the same sex.17 Respondents based their answers on the preceding 6 months.
We also used the Conners’ Parent Rating Scale (CPRS)19 and Teacher Rating Scale (CTRS),20 which are 10-item ADHD questionnaires that have been well validated in many countries, including Sweden.21 Scores range from 0 to 30, with higher scores indicating more problems associated with ADHD symptoms on the basis of the DSM-IV-TR.19,20 We clustered the 2 versions into 2 domains, namely impulsive or restless behavior (items 1–6) and emotional ability (items 7–10), which have been shown to reflect 2 distinct types of behaviors associated with ADHD.22
Information on socioeconomic variables was obtained from the Nordic Health and Family Questionnaire.23 The composite social risk index included maternal education (high school or greater or ≤9 years of schooling), family structure (2-parent or single-parent families), and family income (high or low). Responses were scored 0 or 1, respectively, and the composite scale ranged from 0 to 3 for the lowest to highest social risk.
Standardized forms were used for the data collection and SPSS Statistics version 24 (IBM, Armonk, NY) was used for the data analysis. Differences in dichotomous outcomes between the groups were analyzed with the χ2 test or Fisher’s exact test with odds ratios (ORs) and 95% confidence intervals (CIs) as appropriate. Continuous outcome measures were compared with unpaired Student’s t tests.
The multivariate analysis of covariance (MANCOVA) was used to test the differences between the groups and controlled for known sex and social risks on outcomes. The independent variables were the EPT versus control group and girls versus boys, and the dependent variables were the CBCL and TRF, DSM-IV-TR–oriented behavioral syndrome subscales. The composite social risk index was a covariate. Analysis of covariance (ANCOVA) was performed on the 6 DSM-IV-TR syndrome subscales of the CBCL and TRF. The independent variables and covariates were the same as in the MANCOVA. All MANCOVAs and ANCOVAs were repeated in the groups without NDDs. All significant MANCOVA and follow-up ANCOVA effects were interpreted by using Cohen’s criteria for effect size (ES), in which effects are deemed small (1%–5.8%), medium (5.9%–13.8%), and large (>13.8%).24 P values < .05 were considered significant.
Two of the 134 eligible EPT families refused to participate, and 132 EPT children were assessed. The CBCL was completed by 130 of the 132 EPT group parents and all 103 control group parents, and the TRF was completed by 128 of the 132 EPT group teachers and all 103 control group teachers. The CPRS was completed by 127 of the 132 EPT group parents and all 103 control parents, and the CTRS was completed by 124 of the 132 EPT group parents and 102 of the 103 control group teachers.
The sociodemographic characteristics were similar in the 2 groups apart from significant differences in mother’s education ≤9 years (EPT 14% versus control 4%, P = .02), and mothers with university education (EPT 35% versus control 53%, P = .003) (Table 1). A major NSI was reported in 17 (13%) EPT children but no controls (P = .001).14 The major NDD rate in the EPT and control groups was 31% and 3%, respectively (P < .001) (Table 1).14
Mental Health Measures
A high level of reliability was demonstrated in the psychometric analyses of the behavioral subscale scores for the CBCL, TRF, CPRS, and CTRS (Cronbach’s α = 0.94, 0.93, 0.94, and 0.87, respectively). The correlations between the CBCL and TRF subdomains ranged from r = 0.26 to 0.75 and from r = 0.2 to 0.67 (all r’s: P < .05), respectively.
Behavioral Scores According to Parent (CBCL) and Teacher (TRF) Reports
Parents (CBCL) and teachers (TRF) reported significantly higher behavioral scores for EPT children than for control subjects for internalizing problems (anxious and/or depressed, withdrawn, or somatic problems) and attention, thought, and social problem scales (Tables 2 and 3). The corresponding values in single births compared with multiple births in the EPT cohort were not significantly different (data not shown). The mean raw total problem scores increased significantly for EPT children compared with control children and were not altered when children without NDDs were studied (Tables 2 and 3).
Parents and teachers gave more EPT children than controls abnormal range scores for a number of behaviors (Table 4). The adjusted OR for parent-reported abnormal behavior among EPT children ranged from 2.8 to 3.7 for internalizing problems, the social problem subscale, and attention and thought problems (Table 4). Teachers also reported significantly more abnormal-range scores in EPT children for internalizing, attention, and thought problems (Table 4).
The Conners’ Parent and Teacher Rating Scales
EPT children scored significantly higher in total problem scores on the Conners’ Parent and Teacher Rating scales, and these were not altered when children with NDDs were excluded (Table 5). Both parents and teachers were more likely to give EPT children than controls abnormal range scores in both restless and impulsive behaviors and emotional lability. Similar patterns were seen when children with major NDDs were excluded (Table 6).
MANCOVA Effects According to Parents’ Reports (CBCL)
The 2 (group status) times 2 (sex) MANCOVA of 6 DSM-IV-TR–oriented CBCL syndrome scales revealed significant, multivariate main effects for the EPT versus control group status ([Wilks’λ = 0.85], F6, 218 = 6.2; P < .001; ES = 4.7%) and sex ([Wilks’λ = 0.93], F6, 218 = 2.8; P = .01; ES = 7.0%). When children with NDDs were excluded, significant multivariate main effects remained for both prematurity ([Wilks’λ = 0.88], F6, 185 = 4.2; P = .001; ES = 12.0%) and sex ([Wilks’λ = 0.94], F6, 185 = 2.2; P = .04; ES = 6.5%). No interaction emerged, which indicated that prematurity and sex were independently associated with the multivariate measure of the CBCL syndrome scales.
MANCOVA Effects According to Teacher Reports (TRF)
Similar to the CBCL, the MANCOVA of the DSM-IV-TR–oriented, analogous TRF syndrome scales revealed significant multivariate main effects for prematurity ([Wilks’λ = 0.81], F6, 217 = 8.7; P < .001; ES = 19.5%) and sex ([Wilks’λ = 0.88], F6, 217 = 5.0; P < .001; ES = 12.2%). When children with NDDs were excluded, the multivariate effects remained significant for prematurity ([Wilks’λ = 0.85], F6,185 = 5.3; P < .001; ES = 14.7%) and sex ([Wilks’λ = 0.88], F6, 185 = 4.3; P < .001; ES = 12.2%). Furthermore, the MANCOVA of the NDD-free population also revealed significant multivariate effects for the social risk index ([Wilks’λ = 0.93], F6,185 = 2.4; P = .03; ES = 7.1%). No interaction emerged, which indicated that prematurity, sex, and social risk were independently associated with the multivariate measure of the TRF syndrome scales.
Univariate Effects by Parent Report (CBCL)
The CBCL for the EPT children demonstrated significant effects on 3 of the 6 DSM-IV-TR–oriented syndrome scales (affective, anxiety, and attention), and this did not alter when NDD-free children were analyzed. The analyses also demonstrated the effect of social risk on ADHD problems (total population: P = .001, ES = 4.6%; NDD-free group: P = .002, ES = 5.1%) and sex on somatic problems (total population: P = .004, ES = 3.5%; NDD-free children: P = .01, ES = 3.4%), indicating that boys had higher ADHD scores and girls had higher somatic problem scores. However, there was no interaction between status and sex, which suggested that EPT boys and girls differed from their respective control subjects to similar extents.
Univariate Effects by Teacher Report (TRF)
The univariate analysis of the TRF revealed significant effects for prematurity, similar to the parental reports, in 3 of the 6 DSM-IV-TR–oriented syndrome scales in both total population and NDD-free children, namely affective problems, anxiety, and attention problems (Table 3). The analyses also demonstrated that in addition to prematurity, both social risk and sex were significantly associated with the number of behavior problems. The effect of social risk emerged for both the total population and the NDD-free group, respectively, with regard to affective problems (P = .02, ES = 2.6% and P = .03, ES = 2.4%), anxiety problems (P = .04, ES = 2% and P = .007, ES = 3.8%), ADHD problems (P = .03, ES = 2% and P = .015, ES = 3.1%), and conduct problems (P = .008, ES = 3.1% and P = .007, ES = 3.7%).
When we looked at the sex differences in the total population and the NDD-free groups, we found that boys had higher ADHD (P = .006, ES = 3.3% and P = .01, ES = 3.2%, respectively) and conduct problems (P = .01, ES = 3.1% and P = .02, ES = 2.8%, respectively) than girls, whereas girls had higher scores for anxiety problems (P = .003, ES = 4.4% and P = .04, ES = 3.3%, respectively) than boys. None of the ANCOVAs showed an interaction between sex and status, indicating that EPT boys and girls differed from their respective control subjects to similar extents.
Ratings by parents and teachers showed that EPT children had significantly higher problem scores in anxious and/or depressed, withdrawn, attention, social, and thought problems scales and ADHD-related symptoms. These problems persisted significantly even when children with major NDDs were excluded, although the magnitude of the differences narrowed between the 2 groups. No concomitant differences in externalizing problems were indicated by parent or teacher reports. Studies of very preterm or very low birth weight (VLBW) adolescents11,25–27 and the few reports on EPT adolescents6,9,10 have shown greater consistency in the pattern of behavioral problems, although prevalence estimates have varied. This suggests that there is a preterm behavioral phenotype, which is characterized by attention, anxiety, and social problems.11 We and others11 believe that ELBW or EPT birth appears to confer a specific risk for a triad of behavioral disorders and symptoms. This was clearly demonstrated by our group6 and others2 more than a decade ago by using the CBCL in 5 countries. Our findings demonstrated striking similarities in behavior problems in EPT or ELBW children and term-born controls, with significantly increased scores on attention, social, and thought scales. Few diagnostic studies in adolescents born LBW, VLBW, or EPT further support the existence of a preterm behavioral phenotype with a specific risk for autism spectrum disorders (ASD), ADHD, and emotional disorders.9,28–30 A national study of 11-year-old EPT and ELBW children without major NDDs showed a similar risk for increased disorders and symptoms of anxiety, inattention, and autism.10 Importantly, none of these studies reported a concomitant increase in the prevalence of conduct disorders.
The evidence of increased anxiety scores in childhood have been notably inconsistent in ELBW or more mature preterm cohorts.2,3 ELBW children did not exhibit increased anxiety scores at 8 to 10 years of age compared with cohort-specific, term-born controls,2 whereas others have shown increased emotional problem scores in very preterm children.31 Conversely, the 3 population-based national studies of EPT children at 6 to 11 years of age consistently reported increased scores in anxiety and/or depression by using parent and teacher reports, which is in common with the findings of this study.6,7,10 Internalizing behaviors, such as withdrawn and anxious and/or depressed behavior, have mirrored the increased rate of anxiety disorders,9,30,32 anxiety and depression disorders, and phobias33 in diagnostic evaluations of children born EPT, ELBW, and VLBW.
The clinical interpretation of the increased scores for thought problems according to parent and teacher reports, as seen in our EPT cohort, have gradually become slightly clearer over the past decade. The CBCL and TRF thought problem and social problem scales are relatively short subscales that reflect a heterogeneous group of DSM-IV-TR disorders.34 Many of the items in the thought scale, when combined with social problem scales, reflect some of the symptoms seen in ASD. These include items in the thought scales such as “can’t get mind off,” “repeats acts,” “sees things,” “hears things,” “strange behavior,” “strange ideas,” and items like “lonely,” “dependent,” and “not get along well” in the social problems scales. Evidence from studies that use screening measures and psychiatric diagnostic evaluations indicate that significantly more preterm adolescents have clinical symptoms and screen positive for ASD than the confirmed diagnoses suggest.32,35,36 The only population-based study of EPT children born <26 weeks’ gestation so far reported an 8% prevalence of ASD at 11 years of age.36 We have previously reported a 7% rate of ASD at 10 to 15 years of age for the present cohort.13 There seem to be far more EPT children with ASD symptoms who do not meet the diagnosis criteria and so may not receive help.
Our finding of increased inattention and/or hyperactivity problems, but not conduct problems, was in line with studies on behavioral outcomes for EPT, ELBW, or VLBW children investigated at early school age, during adolescence,2,6,9,25,29,30,33 or as young adults.37,38 Furthermore, the greater risk for inattention relative to hyperactivity has been found in studies in which significantly higher mean scores have been reported for inattention, but not hyperactivity, compared with term-born controls.4,9,10,32,33 Studies using DSM-IV-TR–based criteria, which allows for the diagnosis of ADHD subtypes, have demonstrated a greater risk of disorders that are, in some cases, specific to ADHD, predominantly the inattentive subtype.9,30,33,39 Recent research has revealed that poor attention and working-memory problems, which are important elements of executive functioning, strongly predict academic achievement in EPT adolescents with or without significant intellectual difficulties.15,39,40 These findings stress the importance of identifying children who require referrals, diagnostic procedures, and interventions at an early stage to moderate or prevent drawn-out, negative developmental cascades in learning.
The strengths of this study included the use of well-validated instruments to assess behavioral problems in children. The CBCL, TRF, CPRS, and CTRS have been used for both clinical and research purposes for many years across different cultures.41 These instruments provide cost-efficient, timely measures for large-scale studies with widespread geographical populations such as ours. Different raters and settings contribute to a wider picture of children’s behaviors because raters can provide important and different information.42,43 Other strengths of this study included an excellent follow-up rate (98%) among eligible EPT survivors and the control group. One of the few limitations was the lack of a blinded, in-depth psychiatric interview to categorize DSM-IV-TR diagnoses. Adding observational assessments would have strengthened the validity and reliability of the behavioral and emotional evaluations. Another possible weakness of our study could be the relatively small sample size. Random statistical significance (type I errors) as a result of multiple testing may have occurred. However, the main results in the behavioral outcomes of interest (anxiety, social and thought problems, and internalizing behaviors) were strongly significant, with robust P values, and therefore, we believe that our findings are valid and robust.
This study further strengthens the proposition that a preterm behavioral phenotype is recognizable in adolescents born EPT with or without NDDs and is characterized by an increased risk for inattention, anxiety, and social problems. It highlights the importance of detecting problems in this growing group of children to assess their needs for special support and facilitate their everyday life at home and in school.
We thank the study subjects for completing the questionnaires and taking part in the examinations, the invaluable contribution of the late Prof Gunnar Sedin of Uppsala University for initiating and designing this follow-up investigation, and the dedicated work of our research nurse Margareta Backman (Umea) and project coordinator Nighat Farooqi (Umea).
- Accepted April 4, 2017.
- Address correspondence to Aijaz Farooqi, MD, PhD, Unit of Pediatrics, Institute of Clinical Sciences, University of Umeå, Umeå 901 85, Sweden. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: The work was funded by the ALF Västerbotten, Umea University, Västerbotten County Council, Sweden (grant ALF -VLL 67971) and Jerringfonden. The funders had no role in any aspect of the study or manuscript.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Shum D,
- Neulinger K,
- O’Callaghan M,
- Mohay H
- Farooqi A,
- Hägglöf B,
- Sedin G,
- Gothefors L,
- Serenius F
- Samara M,
- Marlow N,
- Wolke D; EPICure Study Group
- Anderson PJ,
- De Luca CR,
- Hutchinson E,
- Spencer-Smith MM,
- Roberts G,
- Doyle LW; Victorian Infant Collaborative Study Group
- Fevang SKE,
- Hysing M,
- Markestad T,
- Sommerfelt K
- Serenius F,
- Ewald U,
- Farooqi A,
- Holmgren P-Å,
- Håkansson S,
- Sedin G
- Holsti A,
- Adamsson M,
- Hägglöf B,
- Farooqi A,
- Serenius F
- Wechsler D
- Achenbach TM,
- Rescorla LA
- Achenbach TM,
- Dumenci L
- Berntsson L,
- Köhler L
- Cohen J
- Gardner F,
- Johnson A,
- Yudkin P, et al; Extremely Low Gestational Age Steering Group
- Elgen I,
- Sommerfelt K,
- Markestad T
- Delobel-Ayoub M,
- Arnaud C,
- White-Koning M, et al; EPIPAGE Study Group
- Indredavik MS,
- Vik T,
- Heyerdahl S,
- Kulseng S,
- Fayers P,
- Brubakk AM
- Pinto-Martin JA,
- Levy SE,
- Feldman JF,
- Lorenz JM,
- Paneth N,
- Whitaker AH
- Hack M,
- Youngstrom EA,
- Cartar L, et al
- Copyright © 2017 by the American Academy of Pediatrics