Question Is moderate to late preterm (MLP) birth (32 to 36 weeks' gestation) associated with neurodevelopment at school age?
Findings In this cohort study of 159 children born MLP and 137 born early term or later (≥37 weeks' gestation) recruited at birth, children born MLP had poorer scores in cognitive and academic performance and more behavior difficulties at age 9 years than children born early term or later.
Meaning The findings suggest that developmental surveillance to school age is important for children born MLP.
Importance Although children born moderate to late preterm (MLP; 32-36 weeks' gestation) have more neurodevelopmental problems compared with children born early term or later (≥37 weeks' gestation), detailed understanding of affected domains at school age is lacking. Little is known of risk factors for poorer development.
Objective To examine whether being born MLP compared with being born early term or later is associated with neurodevelopmental outcomes at age 9 years and to describe factors associated with poorer neurodevelopment in children born MLP.
Design, Setting, and Participants This prospective, longitudinal cohort study recruited children born MLP and children born early term or later with healthy birth weight (≥2500 g) at a single tertiary hospital in Melbourne, Victoria, Australia, between December 7, 2009, and March 26, 2014. Nine-year follow-up occurred between June 20, 2019, and February 27, 2024.
Exposure Moderate to late preterm birth.
Main Outcomes and Measures Cognitive ability, academic performance, motor function, behavior, and social communication skills, assessed at 9-year follow-up. Group differences were estimated using linear, logistic, or quantile regression adjusted for multiple birth and socioeconomic risk. Multiple imputation was used to account for missing data. Associations of antenatal and neonatal factors and developmental delay at 2 years with poorer 9-year neurodevelopment were explored using univariable regression.
Results Of 201 recruited children born MLP and 201 born early term or later, 159 born MLP (79.1%; 72 [45.3%] male) and 137 born early term or later (68.2%; 75 [54.7%] male) were assessed. Compared with children born early term or later, children born MLP had lower mean (SD) full-scale IQ scores (105.2 [13.6] vs 110.1 [13.0]; adjusted mean difference, -4.4 [95% CI, -7.7 to -1.0]) and poorer performance for cognitive domains, including verbal comprehension, visuospatial, and working memory. They also had poorer academic performance: pseudoword decoding (mean [SD] score, 103.0 [11.3] vs 107.3 [10.5]; adjusted mean difference, -4.0 [95% CI, -7.0 to -1.1]) and mathematics (mean [SD] score, 96.6 [14.7] vs 101.5 [14.5]; adjusted mean difference, -5.0 [95% CI, -8.8 to -1.2]). Children born MLP had similar manual dexterity to those born early term or later (mean [SD] score, 8.4 [3.5] vs 9.1 [3.4]; adjusted mean difference, -0.9 [95% CI, -1.8 to 0.04]) but more behavioral difficulties (50 of 158 [31.7%] vs 29 of 135 [21.5%]; adjusted risk ratio, 1.57 [95% CI, 1.06-2.33]). Developmental delay at 2 years was associated with poorer 9-year neurodevelopment across multiple domains.
Conclusions and Relevance In this longitudinal cohort study of children born MLP, neurodevelopmental challenges persisted into school age. An assessment at age 2 years may assist in identifying children born MLP who are at risk of school-age impairments.
Moderate to late preterm (MLP) birth (32-36 weeks' gestation) accounts for 85% of preterm birth (<37 weeks' gestation), totaling 11.4 million births worldwide in 2020. Recent research has refuted past misconceptions that MLP birth was associated with little to no increase in developmental delays compared with full-term birth. Given the large number of children born MLP, even small increases in developmental delays could have substantial effects on health and educational resources.
There are limitations of current research. Most studies report developmental outcomes in the first few years after birth, which are only moderately predictive of outcomes at school age, a critical time in a child's development during which specific cognitive and academic skills begin to emerge. The few studies reporting outcomes at school age have focused on broad summary scales without detailed characterization of the underlying neurodevelopmental skills. Knowledge of the specific skills affected is critical for designing interventions to improve outcomes for children born MLP. Furthermore, information on early-life risk factors that predict poorer outcomes at school age for children born MLP is lacking. Risk factors identified for children born at less than 32 weeks' gestation, such as neonatal brain injury, bronchopulmonary dysplasia, and neonatal surgery, rarely occur in children born MLP. Identifying early-life risk factors for poorer development at school age will help prioritize surveillance programs and access to early intervention.
We sought to address these knowledge gaps by estimating whether being born MLP compared with being born early term or later (≥37 weeks' gestation) is associated with cognitive ability, academic performance, motor function, behavior, and social communication skills at 9 years of age. We also aimed to identify factors in the newborn period and infancy that were associated with poorer neurodevelopment at 9 years of age. We hypothesized that compared with children born early term or later, those born MLP would have worse outcomes in all reported domains of development at 9 years of age. We also hypothesized that earlier gestational age, male sex, higher socioeconomic risk, and developmental delay at 2 years of age would be associated with poorer development at 9 years of age.
LaPrem is a longitudinal cohort study of children born MLP and children born early term or later with healthy birth weight (≥2500 g) who were recruited after birth from the Royal Women's Hospital, a tertiary hospital in Melbourne, Victoria, Australia, between December 7, 2009, and March 26, 2014. Children with congenital abnormalities or genetic syndromes known to affect development were excluded. In addition, infants born early term or later were excluded if they were unwell at birth, received resuscitation, or were admitted to the neonatal nursery. Children were previously assessed at age 2 years. At 9 years' corrected age, children underwent a 2-day assessment, between June 20, 2019, and February 27, 2024. Follow-up of participants was limited during 2020 and 2021 due to extensive COVID-19 pandemic lockdowns in Victoria, Australia. Ethical approval was provided by the human research ethics committees of the Royal Women's Hospital and the Royal Children's Hospital Melbourne. All parents gave written informed consent. The study protocol has been published elsewhere. This report followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.
Perinatal, neonatal, and maternal data were collected prospectively. Socioeconomic risk was assessed at the 9-year follow-up using 6 variables: family structure, educational level of the primary caregiver, employment status and occupation of the primary income earner, language spoken at home, and maternal age at birth of the child. Each variable was scored on a 3-point scale (from 0 for lowest risk to 2 for highest risk), summed to give a total score, and dichotomized to higher (total score ≥2) or lower (total score <2) socioeconomic risk.
Children were assessed by trained assessors (R.E., T.L.F., K.L.C., L.R) unaware of birth status or clinical history. Age was corrected for prematurity to account for a known bias in cognitive test scores. In addition to the formal assessments, parents were asked whether their child had a developmental disability, including cerebral palsy, autism spectrum disorder (ASD), and attention-deficit/hyperactivity disorder (ADHD).
Cognitive ability was measured using the Wechsler Intelligence Scale for Children (WISC), Fifth Edition: Australian and New Zealand Standardized Edition. Full-scale IQ (FSIQ) was used as a measure of general intelligence, with age-standardized index scores from 5 domains of cognitive functioning: verbal comprehension, visuospatial, fluid reasoning, working memory, and processing speed. Intellectual impairment was defined as a score less than -1 SD (15 points) relative to the mean for the children born early term or later.
Reading (single-word reading and pseudoword decoding), spelling, and mathematics skills were assessed using the Wechsler Individual Achievement Test, Third Edition: Australian and New Zealand Standardized Edition. Impairments in reading, spelling, and mathematics were defined as less than -1 SD (15 points) relative to the mean for the group born early term or later.
Motor function was assessed using the Movement Assessment Battery for Children-Second Edition (MABC-2). The overall MABC-2 standard score and 3 subscales (balance, aiming and catching, and manual dexterity) have a mean (SD) score of 10 (3), with higher scores indicating better performance. Motor impairment was defined as either cerebral palsy or MABC-2 score less than or equal to the fifth centile.
Behavior was assessed using the Strengths and Difficulties Questionnaire (SDQ), a parent- or primary caregiver-reported measure that assesses 5 scales relating to emotional symptoms, conduct problems, hyperactivity and inattention, peer relationship problems, and prosocial behavior. Higher scores from the first 4 scales indicate more behavioral problems and were summed to a total difficulties score (range, 0-40). Any behavioral difficulty was defined as having a total difficulties score in the 80th percentile or higher based on Australian norms.
Social communication was assessed using the Lifetime form of the Social Communication Questionnaire (SCQ), a parent- or primary caregiver-reported measure evaluating reciprocal social interaction, communication, and repetitive and/or stereotypic behavior. Higher scores indicate more behaviors associated with autism. Scores greater than the cutoff of 15 points indicate a higher likelihood of ASD, with recommendations for further diagnostic assessment.
Some children did not have impairment in any of the reported domains. For these children, we computed a composite outcome of no impairment.
Data were analyzed using Stata, version 18 (StataCorp LLC). Participant characteristics were summarized as mean (SD) or number (percentage) and compared between those who were assessed at 9 years and those who were not. Continuous outcomes at 9 years were compared between groups using mean or median differences estimated using linear or quantile regression, respectively. Binary outcomes were compared between groups using risk ratios (RRs) and risk differences (RDs) estimated using generalized linear models with a binomial family and either a log (RRs) or an identity (RDs) link function. A separate model was used for each developmental outcome. All models were adjusted for multiple birth and socioeconomic risk at 9 years, and except for quantile regressions, they were fitted using generalized estimating equations with an exchangeable correlation structure to account for clustering of multiple births within the same family and robust SEs. The directed acyclic graph in eFigure 1 in Supplement 1 depicts the assumptions made for this analysis.
To account for missing data, estimates were obtained using multiple imputation by chained equations; assumptions and a justification for the use of multiple imputation are provided in eFigure 2 in Supplement 1. A separate multiple imputation procedure was conducted for each outcome. All imputation models included gestational age group, multiple birth, socioeconomic risk, perinatal and neonatal variables, disability status at 2 years (cerebral palsy and/or developmental delay, defined as a score <-1 SD below the mean for children born early term or later on either the cognitive, language, or motor domains on the Bayley Scales of Infant and Toddler Development, Third Edition [Bayley-III]), and the relevant outcome. Forty imputed datasets were created, and the resulting inference was combined using Rubin rules. We also reported results from a complete case analysis.
Within the group born MLP, associations between early-life variables and 9-year outcomes were explored using univariable logistic or linear regression fitted using generalized estimating equations with an exchangeable correlation structure to account for clustering of multiple births within the same family and robust SEs. The variables were selected a priori and included assisted conception, antenatal corticosteroids, gestational age, multiple birth, male sex, birth weight z score, respiratory support, higher socioeconomic risk at birth, and developmental delay at 2 years of age (defined as a score <-1 SD below the mean for children born early term or later on either the cognitive, language, or motor domains on the Bayley-III). Given the multiple comparisons, we interpreted our findings by focusing on overall patterns and magnitude of differences rather than on individual P values.
Of the 201 children born MLP and 201 born early term or later who were recruited at birth, data at 9 years were available for 159 born MLP (79.1%; 87 [54.7%] female and 72 [45.3%] male) and 137 born early term or later (68.2%; 62 [45.3%] female and 75 [54.7%] male) (Figure 1). There were group differences in perinatal characteristics related to prematurity (Table 1). Compared with children born early term or later, children born MLP were more likely to be born following assisted conception, in a multiple birth, and/or via cesarean delivery; have lower birth weight z scores and higher socioeconomic risk; and have received respiratory support after birth. There were no children in either group diagnosed with cerebral palsy, blindness, or deafness at 9 years. In both groups, mothers of participants who were assessed at 9 years were older, and the proportion of participants with higher socioeconomic risk was lower compared with those who were not assessed at 9 years (eTable 1 in Supplement 1). In the group born MLP, the proportions of participants exposed to antenatal corticosteroids, male participants, and those from a multiple birth were lower among those who were assessed at 9 years than among those who were not. In the group born early term or later, the rate of assisted conception was almost double among those who were assessed at 9 years compared with those who were not.
Children born MLP had lower mean (SD) FSIQ scores than children born early term or later (105.2 [13.6] vs 110.1 [13.0]), with an adjusted mean difference of -4.4 points (95% CI, -7.7 to -1.0 points), which is equivalent to -0.3 SD (Table 2 and eFigure 3 in Supplement 1). Similar results were found for most other cognitive domains, with the greatest differences in verbal comprehension, visuospatial, and working memory indices. Children born MLP performed less well in reading (mean [SD] pseudoword decoding score, 103.0 [11.3] vs 107.3 [10.5]) and mathematics (mean [SD] score, 96.6 [14.7] vs 101.5 [14.5]) compared with children born early term or later, with adjusted mean differences of -4.0 points (95% CI, -7.0 to -1.1 points) and -5.0 points (95% CI, -8.8 to -1.2 points), respectively, which are in the range of -0.3 SD. The overall standard scores for the MABC-2 were similar in the group born MLP and in children born early term or later although among those born MLP, and performance in the manual dexterity scale was similar (mean [SD] score, 8.4 [3.5] vs 9.1 [3.4]; adjusted mean difference, -0.9 [95% CI, -1.8 to 0.04]). Hyperactivity and inattention scores were slightly higher among the children born MLP compared with children born early term or later.
For dichotomous outcomes, the number of children with no impairment was similar between groups, with an adjusted RD of -8.9% (95% CI, -21.7% to 3.9%). The rate of any academic impairment in the group born MLP was similar to that among children born early term or later (adjusted RD, 6.4%; 95% CI, -6.8% to 19.7%) (Table 3). Rates of impairment in cognition, motor, or communication problems were similar between children born MLP and those born early term or later. However, the group born MLP had more behavioral difficulties than children born early term or later (50 of 158 [31.7%] vs 29 of 135 [21.5%]), with an overall adjusted RR of 1.57 (95% CI, 1.06-2.33) and adjusted RD of 12.8% (95% CI, 2.1%-23.5%). Of the 16 children identified as at risk of autism on the SCQ, all but 2 (12.5%; 1 in each group) had received a diagnosis of ASD. The complete case analyses had similar conclusions (eTables 2 and 3 in Supplement 1).
Developmental delay at 2 years of age was associated with impairments in all developmental domains at age 9 years (Figure 2 and eTable 4 in Supplement 1). Receiving antenatal corticosteroids was associated with motor impairment and behavioral difficulties. Higher socioeconomic risk was associated with cognitive impairment and poorer social communication. However, a higher gestational age and multiple birth were associated with better social communication.
The major findings of this cohort study are that children born MLP had lower scores in multiple neurodevelopmental domains and more behavioral difficulties at 9 years of age compared with their peers born early term or later. Differences were greatest for general cognitive ability, reading (pseudoword decoding) and mathematics, manual dexterity, and hyperactivity and inattention. The findings at 9 years are consistent with cognitive, language, and motor delays reported at 2 years' corrected age in the group born MLP compared with children born early term or later. We chose a cutoff of less than -1 SD because even mild developmental delay affects function and children with mild delay may benefit from intervention. The current findings add to the growing evidence that the developmental challenges associated with MLP birth persist to school age. We also identified that developmental delay at 2 years was associated with a range of adverse outcomes at 9 years in children born MLP. This suggests that developmental assessment at 2 years of age is important to identify the children born MLP who may have high risk of neurodevelopmental and behavioral problems at school age.
Nonetheless, most children born MLP scored within -2 SD of the reference mean score of 100 for FSIQ and all academic domains (eFigure 3 in Supplement 1). We reported mean differences in the range of -0.3 SD for general cognition, which is considered to be a clinically important difference and is of similar magnitude to the standardized mean differences pooled from 4 studies of children born MLP with ages ranging from 6 to 13 years. The group born early term or later in our study had a mean FSIQ of 110.1, higher than published norms, which may reflect selection of a higher-functioning group or a Flynn effect (the tendency for IQ scores to drift upward over time) given that the version of the WISC used in our study was published in 2016. As differences of similar magnitude were found in most cognitive domains, including verbal comprehension, visuospatial, fluid reasoning, and working memory, the association of MLP birth with cognition is unlikely to be specific to selective cognitive skills.
The group born MLP had lower academic performance scores than children born early term or later in pseudoword decoding and mathematics in the order of a clinically important difference of -0.3 SD. The rates of any impairment in academic performance were similar in the group born MLP than children born early term or later. Lower educational achievement at 6 to 11 years of age has been reported in children born MLP compared with those born early term or later, with pooled relative risk estimates of 1.96 (95% CI, 1.11-3.43) for those born at 32 to 33 weeks' gestation and 1.21 (95% CI, 1.10-1.32) for those born at 34 to 36 weeks' gestation. Difficulties in mathematics have been reported in children born very preterm compared with children born early term or later, although this has been less studied in groups born MLP. A linkage study in California reported poorer mathematics proficiency from grades 3 to 8, which included children born MLP of similar ages to those in our study. Knowing which domains of academic performance are most affected may help tailor interventions for children born MLP.
There were lower MABC-2 scores in the manual dexterity subscale among children born MLP. However, rates of motor impairment were similar between groups. Of interest, the group born early term or later had mean MABC-2 standard scores that were lower than test norms, and a relatively high rate scored at or below the fifth centile. The reasons for this are unclear and warrant further investigation. For other developmental domains, the children born early term or later had higher means than the test norms, similar to previous findings in which Australian groups born early term or later had higher means than published test norms.
We reported higher rates of behavioral difficulties using the SDQ in the group born MLP compared with children born early term or later. Pettinger et al reported ADHD diagnosis or ADHD symptom prevalence of 34.9 (95% CI, 33.6-36.3) per 1000 children in the group born MLP compared with 26.6 (95% CI, 26.3-26.8) per 1000 children among those born early term or later. There was a slight increase in relative risk for ASD in children born MLP compared with children born early term or later. Our study did not formally assess for ASD but instead classified children at risk of ASD based on an assessment of social communication. All but 2 children who were at risk of ASD had a previous diagnosis of ASD. The sample size in our study was not powered to detect group differences in either ADHD or ASD, although the sample had relatively high rates for both diagnoses.
We identified few early-life factors that were associated with 9-year neurodevelopment, apart from developmental delay at 2 years of age. Antenatal corticosteroid administration was associated with motor impairment and behavior difficulties. This concurs with recent concerns about the potential negative developmental association between antenatal corticosteroids and late preterm birth. The lack of associations between gestational age and many outcomes at 9 years suggests that there is little difference in risk for children born within the 32- to 36-week gestational age range. Our study supports the importance of a developmental assessment at 2 years of age to identify children born MLP who may have high risk of later neurodevelopmental problems. As the absolute numbers of children born MLP were large, screening tools may be appropriate as a first-line approach to stratify risk.
Our study's strengths include recruitment of participants at birth, with detailed perinatal, neonatal, and early infancy data. We conducted a broad suite of direct assessments by assessors blinded to participant group and clinical history, rather than relying on diagnostic categories; this enabled deep phenotyping of deficits that contributed to neurodevelopmental problems reported by other studies in a meta-analysis. This information may be valuable to clinicians caring for children born MLP and also for development of targeted interventions.
This study also has limitations. Our retention rates at 9 years were lower than at the 2-year follow-up, in part due to disruptions with COVID-19 lockdowns in Victoria, Australia. We used multiple imputation informed by directed acyclic graphs for estimation in the presence of missing data. The single tertiary center cohort may represent children who were sicker at birth and thus at higher risk of developmental delays compared with the whole population of children born MLP.
The findings of this cohort study suggest that neurodevelopmental problems in a broad range of domains in children born MLP may persist at school age. Neurodevelopmental status at 2 years was associated with longer-term outcomes. Thus, consideration should be given to developmental screening at 2 years of age for children born MLP. Further research for an in-depth understanding of affected developmental domains is important to develop targeted interventions for this large and growing group of children.
Corresponding Author: Jeanie L. Y. Cheong, MD, Neonatal Services, Royal Women's Hospital, 20 Flemington Rd, Parkville, Level 7, Newborn Research, VIC 3052, Australia (jeanie.cheong@thewomens.org.au).
Author Contributions: Dr Cheong had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Profs Anderson and Spittle were joint senior authors.
Critical review of the manuscript for important intellectual content: All authors.
Administrative, technical, or material support: Olsen, Ellis, FitzGerald, Cameron, Spittle.
Supervision: Cheong, Anderson, Spittle.
Conflict of Interest Disclosures: Dr Cheong reported receiving grants from the Australian National Health and Medical Research Council (NHMRC) during the conduct of the study and grants from the Medical Research Future Fund of Australia, Royal Children's Hospital Foundation, and Thrasher Foundation outside the submitted work. Dr Doyle reported receiving grants from the NHMRC during the conduct of the study. Dr FitzGerald reported receiving grants from the NHMRC during the conduct of the study. Dr Cameron reported receiving grants from the NHMRC during the conduct of the study. Dr Anderson reported receiving grants from the NHMRC during the conduct of the study. Prof Spittle reported receiving grants from the NHMRC during the conduct of the study and outside the submitted work. No other disclosures were reported.
Funding/Support: This work was supported by project grants 1028822 and 1161304 (Prof Cheong, Dr Doyle, and Profs Anderson and Spittle) and Leadership Level 1 investigator grants 2016390 (Dr Cheong) and 1176077 (Dr Anderson) from the Australian NHMRC and by the Victorian Government's Operational Infrastructure Support Program.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Data Sharing Statement: See Supplement 2.
Additional Contributions: We thank the research coordinators, families, and children for their participation in this study.