Risk of Developmental Disorders in Children Born at 32 to 38 Weeks’ Gestation: A Meta-Analysis

This systematic review was registered with PROSPERO (CRD42021298773) and reported according to PRISMA guidelines.¹⁶  The population of interest was children born between 32 and 38 weeks, aged 2 to 17 years at assessment, born after 1996, when antenatal steroids administration in preterm labor became routine practice.¹⁷  Comparator groups were (where possible) children born at full term (39–40/41 weeks), or if this data were not presented, at term (≥37 weeks) . The outcomes were developmental disorders as per the National Institute for Health and Care Excellence,¹⁸  which in some cases were subdivided further to allow meaningful comparison: cerebral palsy (CP); developmental coordination disorder (DCD); visual impairment; hearing impairment; sleep apnea; oro-motor feeding problems; social, emotional, and behavioral problems subdivided into: global social, emotional, and behavioral problems, internalizing behaviors, externalizing behaviors, and social problems; attention deficit hyperactivity disorder (ADHD); autism spectrum disorder (ASD); developmental delay subdivided into: global developmental delay, language delay, motor developmental delay, and cognitive developmental delay; cognitive impairment; executive function problems; low educational achievement, subdivided into: not “school ready” aged ≤5 years, low educational achievement aged 6 to 11 years, and low educational achievement aged 12 to 17 years; and special educational needs (including physical disabilities which affect learning as well as learning impairments).¹⁹ 

Searches were conducted in Medline, Embase, Psychinfo, Cumulative Index of Nursing and Allied Health Literature on February 10, 2022 (updated November 22, 2022). The search terms are listed in Supplemental Table 4. In addition, a Google scholar search was conducted on February 14, 2022. Initially the advanced search feature of Google Scholar was used, but the resulting studies were not relevant. Therefore, a basic search was performed using the question: “What is the risk of neurodevelopmental disorder in children born between 32 and 38 weeks gestation?”. The top 50 results were exported. Reference lists of included studies were searched for additional studies that met the inclusion criteria, as follows: at least 1 developmental disorder identified on a validated questionnaire, standardized test, or physician diagnosis; provided estimates for children born moderately or late preterm, or early term.

One author (K.P.) reviewed all titles, abstracts, and full texts using Covidence software.²⁰  Twenty percent were independently reviewed by a second author (C.C.). There was substantial agreement; 93.6% and Cohen’s κ 0.62 at abstract screening; 98.4% and Cohen’s κ 0.85 at full text review.²¹  Disagreements were resolved by discussion.

Where key data were missing, authors were contacted by e-mail; if no response was received, the study was excluded. Where studies used the same cohort, the most recent publication was included, unless an older study provided the results in a binary format, enabling prevalence calculation (Supplemental Table 5 for summary). The full list of exclusion criteria can be found in the study protocol (CRD42021298773).

Data were extracted by K.P. and a 20% sample was cross-checked by C.C. for accuracy. See Supplemental Table 3 for data extraction form. Where only odds ratios were presented, the study authors were contacted to request the raw data.

The Newcastle-Ottawa scale was used to assess quality with scores categorized as poor (0–2), fair (3–5), or good (6–9) (Supplemental Fig 11).²² 

An unadjusted pooled relative risk of each outcome by gestational age was calculated using random effects meta-analysis because of expected heterogeneity, quantified using the I² statistic. The number of children with each disorder, the number without, and the total number of participants (split into gestational subgroups) was used in the calculations. If raw numbers were not presented (or supplied on request), the paper’s unadjusted calculated effect sizes (relative risk) were used. When there were 0 cases, a continuity correction of 0.5 was applied. For continuous measures, studies were pooled using Hedge’s G standardized mean difference. Signs were changed where necessary.²³  Prevalence was pooled by gestational age with a fixed effect model using the inverse-variance method. The Freeman Tukey double arcsine transformation was specified where necessary to stabilize variances.^24,25  Publication bias was assessed using Egger’s test and funnel plots.

Analyses were undertaken by gestational age groupings: moderately preterm (32–33 weeks); late preterm (34–36 weeks); early term (37–38 weeks)^5,26,27 ; moderate-to-late preterm (32–36 weeks), but only if results were presented in this format without further breakdown.

Where results were not presented as above, broader groupings were used: 32 weeks was coded as 32–33 weeks; 35–36 weeks was coded as 34–36 weeks; 33–36 weeks, 32–34 weeks, 32–35 weeks, or 32–37 weeks were all coded as 32–36 weeks; and 32–34 weeks and 35–36 weeks were combined into 32–36 weeks.

Where data were available for children born at 32–33 weeks and 34–36 weeks, the 32–36 week category was not presented.

Sensitivity analysis was undertaken for relative risk of CP, global developmental delay, and educational outcomes aged 6 to 11 years by comparing results when studies using a ≥37 week term comparison group (as opposed to a 39–40/41 weeks comparison group) were included and excluded.

The initial search identified 11 630 records with 17 further studies from reference screening and updated searches. As shown in Fig 1, 9028 studies were excluded at the first stage and 1462 studies were full text screened; 1386 were excluded, resulting in 76 studies included. There was full agreement between the 2 researchers conducting data extraction.

Key characteristics of the included studies are summarized in Table 1. Several studies reported multiple outcomes. All were cohort or cross sectional studies. Sample sizes ranged from 83 to 1 390 601 and covered the full age range (2–17 years). The majority were from economically developed countries. There were 59 studies rated as “good” quality and 15 as “fair’” (Supplemental Table 7). Two studies were cross sectional, therefore 2 of the Newcastle Ottawa Scale questions did not apply and so were not classified as good, fair, or poor quality. Twenty four meta-analyses were conducted, summarized in Table 2. No papers reported oro-motor feeding problems.

The relative risk of screening positive for, or a diagnosis of, a developmental disorder compared with children born at term, was highest in the moderately preterm group; there was a statistically significant increased relative risk of CP, social problems, ASD, global DD, cognitive impairment, low educational attainment, and special educational needs (Figs 2–9). Forest plots for the remaining disorders are presented in the supplemental information (Supplemental Figs 12–37).

The relative risk of CP compared with children born at term was, for 32 to 33 weeks: 14.1 (95% confidence intervals [CI]: 12.3–16.0), 34 to 36 weeks: 3.52 (95% CI: 3.16–3.92) and 37 to 38 weeks: 1.44 (95% CI: 1.32–1.58). The prevalence of CP per 1000 children was, for 32 to 33 weeks: 17.1 (95% CI: 15.1–19.3), 34 to 36 weeks: 2.95 (95% CI: 2.53–3.39), 37 to 38 weeks: 2.05 (95% CI: 1.91–2.21), and ≥37 weeks: 0.53 (95% CI: 0.50–0.57).

The relative risk of global developmental delay compared with children born at term was, for 32 to 33 weeks: 2.89 (95% CI: 2.77–3.02), 34 to 36 weeks: 1.61 (95% CI: 1.25–2.08) and 37 to 38 weeks: 1.14 (95% CI: 1.12–1.16). The prevalence of global developmental delay per 1000 children was, for 32 to 33 weeks: 350 (95% CI: 335–365), 34 to 36 weeks: 132 (95% CI: 128–136), 37 to 38 weeks: 138 (95% CI: 136–140), and ≥37 weeks: 65.5 (95% CI: 64.7–66.3).

The relative risk of ADHD compared with children born at term was, for 32 to 36 weeks: 1.25 (95% CI: 1.14–1.38), 34 to 36 weeks: 1.62 (95% CI: 1.38–1.90), 37 to 38 weeks: 1.19 (95% CI: 1.00–1.42). The prevalence of ADHD per 1000 children was, for 32 to 36 weeks: 34.9 (95% CI: 33.6–36.3), 34 to 36 weeks: 75.7 (95% CI: 66.1–86.0), 37 to 38 weeks: 37.2 (95% CI: 32.4–42.8), and ≥37 weeks 26.6 (95% CI: 26.3–26.8).

The relative risk of low educational achievement aged 6 to 11 years compared with children born at term was, for 32 to 33 weeks: 1.96 (95% CI: 1.11–3.43), 34 to 36 weeks: 1.21 (95% CI: 1.10–1.32), and 37 to 38 weeks: 1.13 (95% CI: 1.08–1.19). The prevalence of low educational achievement aged 6–11 years per 1000 children was, for 32 to 33 weeks: 304 (95% CI: 285–324), 34 to 36 weeks: 199 (95% CI: 195–203), 37 to 38 weeks: 224 (95% CI: 221–227), and ≥37 weeks: 163 (95% CI: 162–164). A similar pattern is seen across the other developmental disorders, as shown in Supplemental Figs 12–39 (in the supplemental information) and in the prevalence summary bar charts (Fig 10).

Visual or hearing impairment were slightly more prevalent in some of the groups born before full term. The relative risk of visual or hearing impairment compared with children born at term was increased in all gestational subgroups, although this only reached statistical significance in the 37 to 38 week gestation group (Supplemental Figs 15–18). There was a small increased risk and prevalence of sleep apnea but its prevalence is low compared with other disorders (Fig 10).

There is an association with reduced gestational age and increased risk and prevalence of all forms of social, emotional, and behavioral problems, except externalizing disorders. The most prevalent problem of this type is social problems (eg, showing empathy and playing with other children²⁸ ). There was a small increased risk and prevalence of ASD in the 34 to 36 week group. The only disorder found with neither an increased relative risk nor prevalence among the 32 to 36 week children was externalizing behaviors; this was based on 1 study, as shown in Table 3. There was no increased relative risk of executive function disorders (there is no prevalence estimation since it was reported as a continuous outcome).

Among early term children, prevalence and relative risk of developmental disorder was increased for several disorders: CP, DCD, visual impairment, ADHD, global developmental delay, cognitive impairment, not school ready, low educational achievement, and special education needs (Table 2). There was no data for early term children for: sleep apnea, internalizing behaviors, externalizing behaviors, social problems, motor developmental delay, or cognitive developmental delay.

When only the papers with a full term (39–40/41 weeks) comparison group were included, only 1 paper remained in each gestational subgroup for CP and global developmental delay. The overall interpretation was unchanged in all cases (Supplemental Figs 70–73).

In general, the heterogeneity statistics (I²) were 0 (Figs 2–9). Some subgroups had a higher heterogeneity, for example in the educational achievement aged 6 to 11 years meta-analysis (Fig 8). A possible source of heterogeneity was different measurements for developmental disorder; in the CP meta-analysis Larroque 2008, You 2019a, and You 2019b examined participants rather than using linked records, possibly explaining why these studies had a higher prevalence of CP. The different methods used to identify each developmental disorder are shown in Supplemental Table 8. Heterogeneity between subgroups was low.

The majority of funnel plots (16 of 24) were symmetrical with nonsignificant Egger’s tests (14 of 24) (Supplemental Table 9 and Supplemental Figs 48–69 in the supplemental information).

Children born between 32 and 38 weeks are at increased risk of screening positive for, or receiving a diagnosis of, a developmental disorder compared with children born at term. In most cases an inverse gradient association with gestational age was demonstrated. The highest increased relative risk compared with children born at term was for children born 32 to 33 weeks for CP, but CP has a low prevalence compared with other developmental disorders.

The association between increased risk and prevalence of global developmental delay or language delay compared with children born at term is evident in all gestational age groups between 32 and 38 weeks. Interpreting the cognitive developmental delay meta-analysis is challenging; prevalence was relatively high in the term group, with confidence intervals overlapping the 32 and 36 week group. However, there was an increased relative risk of cognitive developmental delay in the 32 to 36 week group compared with term. It is likely that the picture seen with prevalence is a result of which studies reported different gestational age groups.

In 2 disorders (hearing impairment and ASD), 1 group born 32 to 38 weeks had a lower prevalence of developmental disorder than the term group (Table 2). In both cases this was where the gestational subgroup only included results from 1 study (Table 3). It is likely that this is a peculiarity of the tools used to assess the disorder in that study, since when the relative risks are considered, this effect was not seen.

These findings are largely consistent with previous research that has shown increased risk of developmental disorders among children born moderately preterm, late preterm, and early term.^3,8,10,19  This review demonstrates that difficulties faced by children born 32 to 38 weeks persist through childhood, with evidence of increased risk and prevalence of cognitive impairment and low educational achievement aged 6 to 11 years, in contrast to previous research suggesting developmental delay in preterm infants may be transient.^29,30 

The proportion of children affected by a developmental disorder is generally lower among children born between 32 to 38 weeks compared with extremely preterm children. However, late preterm and early term birth are common; in the United States in 2020, 7.4% of children were born late preterm, whereas only 2.7% were born under 34 weeks.¹¹  Therefore, small increases in relative risk (compared with full term children) may have a considerable affect, both clinically and economically, at a population level.^30–33 

Developmental disorders affect 35% to 52% of children born extremely or very preterm^31,34  (which is not substantially higher than the prevalence of DCD, social problems, and low educational achievement found in children born 32–38 weeks in this meta-analysis). CP affects 8% to 9% of very preterm children,²⁹  compared with 1.71% (95% CI: 1.51–1.93) born between 32 and 33 weeks in this review. Children born between 32 and 38 weeks may experience a different profile of disorders to extremely preterm children, possibly mediated through different pathways.^13,29,33,35  Birth before full term may be the result of maternal ill health (eg, preeclampsia, gestational diabetes, infection) and a suboptimal intrauterine environment.³³  These antenatal issues may be driving the increased risk and prevalence of developmental disorders, as opposed to the prematurity per se.^7,11,36  Conversely, babies born just a few weeks early have markedly different brain maturation to full term children.^10,35,37  It is possible that birth between 32 and 38 weeks’ gestation may disrupt evolution of neural connections, potentially resulting in developmental disorder.^19,33  Children born before full term are more likely to have medical complications in the immediate neonatal period, in some cases leading directly to a developmental disorder.^19,38  After the neonatal period, children born late preterm are 2 to 3 times more likely to attend the emergency department or be admitted to hospital.^6,39  The increased medical needs of children born before full term affect both the child and family. Parents of late preterm infants have been shown to have high emotional distress and anxiety levels.^6,40  Furthermore, admission to the neonatal unit is associated with both acute stress and post-traumatic stress disorder among parents.⁴¹  It is plausible that early complications, prolonged admission, or readmission to hospital indirectly affects child development via the negative effect on the whole family.

This was a broad ranging, comprehensive review. The search strategy identified a large number of studies; there was a total of over 8 million children in the meta-analyses. Including a full range of developmental disorders enabled comparison of prevalence of different developmental disorders across gestational ages. Calculating prevalence meant the increased risk could be contextualized. Although there were a large number of children in total, because each developmental disorder was considered separately according to gestational age subgroups, there were sometimes small numbers in each subgroup (Table 3). There were relatively few studies reporting outcomes for early term children; the subgroup meta-analysis often only contained data from 1 study, although this often represented more children than in the other gestational subgroups combined.

Developmental disorders are, by definition, a heterogeneous group. To determine which conditions to investigate and maximize this review’s applicability, we used the NICE guideline, “Developmental follow-up of children and young people born preterm,” as a reference.¹⁸  The NICE guideline considers “problems with inattention, impulsivity, or hyperactivity” separately from “executive function problems.” However, ADHD is closely associated with impaired executive function, and some have argued that “executive function problems” do not represent a diagnosis as such.^42,43  The NICE guidance also describes increased risk of low educational achievement (“educational attainment” in the UK) among children born preterm. Although there are probably many children who have low educational achievement but do not have a diagnosed developmental disorder, low educational achievement has been demonstrated to be associated with early developmental difficulties.⁴⁴  Including educational achievement also gives the opportunity to examine outcomes at a later stage of childhood, demonstrating that the association between birth before full term and developmental disorder can persist into adolescence.

The tools measuring outcomes were numerous and varied, as described in Supplemental Table 8, which likely accounts for some of the different prevalence (eg, DCD Supplemental Fig 13). Furthermore, although some studies used diagnostic codes in medical records to determine their outcome, many studies used questionnaires, tests, and tools (eg, the Ages and Stages Questionnaire) that were not developed to make formal diagnoses, but rather were intended as screening tools for developmental problems.⁴⁵  Thus, when considering the results of this review, it is crucial to bear in mind that although some children may screen positive for a potential developmental disorder or be highlighted as a “cause for concern,” that does not equate to a formal diagnosis of a specific developmental disorder.

As with other reviews,¹⁰  the children were assessed at different ages and the term comparison groups were variable, for example ≥37 weeks, >37 weeks, ≥39 weeks, 39 to 41 weeks, or 40 weeks. Comparator groups were (where possible) children born at full term (39–40/41 weeks), or if this data were not presented, at term (≥37 weeks). This could account to some extent for why the prevalence of some disorders (eg, not school ready, low educational achievement 6–11 years) is lower in the 37 to 38 week group than in the term group. Although it would have been preferable to have a homogenous term comparator group, only including studies that used a full term (39–40/41 weeks) comparison group would have resulted in a substantial loss of data; of the 76 included studies, only 24 used a full term comparison group. A sensitivity analysis was undertaken for CP, global developmental delay, educational outcomes aged 6 to 11 years, and ADHD, where only the papers with a full term comparison group were included (Supplemental Figs 70–73). In some cases, this resulted in only 1 study in each gestational age subgroup. The overall interpretation was unchanged.

Some studies used atypical cohorts, eg, Drougia 2007, Klassen 2004, and Polic 2017 only included children who admitted to a neonatal unit; in these cases, their term comparator groups would not represent typical term-born children.

The majority of studies were from economically developed countries and all were published in English, possibly limiting generalizability. Outside of the United States, there was limited data on children with non-European heritage; this is important, as race or ethnicity may impact the likelihood of screening positive for, or receiving a diagnosis of, a developmental disorder.^3,18,46 

Understanding the long-term implications of birth before full term when balanced against short term risk to the mother and fetus may influence obstetric decision making.⁸  It is vital that all healthcare professionals, particularly pediatricians, are well informed of the potential consequences of preterm birth in order that they can give evidence based information to families and so opportunities for early intervention are not missed. Children born at 32 to 38 weeks may benefit from increased monitoring of their development, but most neonatal follow-up programs only apply to children born very preterm, in line with American Academy of Pediatrics guidelines.^6,32,47  However, it is likely that, especially among the more mature gestations, many children will not have any developmental disorders.³¹  Depending on the structure and financing of the healthcare systems, routine appointments for all these children may be impractical and undesirable. A more effective approach would be collaborating with the education sector (which currently bears the majority of the cost associated with prematurity⁴⁸ ). Teachers should be informed if they have students who are born preterm and early term and receive training on how to support them.^8,49  It is also likely that early childhood risks for poorer outcomes are additive; determining which groups of children born 32 to 38 weeks are most at risk for developmental disorder, selectively following them up, and providing family support would be a pragmatic approach.⁶  Empowering parents with information on developmental risks is important, not only for the early recognition of problems, but also to give parents agency.⁴⁷  It is currently unclear which children born between 32 and 38 weeks are at the highest risk³²  or to what extent early interventions shown to benefit very preterm children might also benefit children born 32 to 38 weeks.^{12,32,48–52} 

We have demonstrated that for many gestational subgroups the research into developmental disorders is sparse (Table 3) and gaps in the empirical knowledge persist. In future research, consistent outcome measurements, full term control groups (39–40/41 weeks) and standardized gestational age groups should be used, to allow easier comparison.^8,19,32 

This review has found evidence of an inverse relationship between birth before full term and the majority of developmental disorders. Low educational achievement, DCD, global developmental delay, and cognitive impairment were the most prevalent disorders among children born 32 to 38 weeks. Some of the increased risks are small but may have significant consequences both clinically and economically at a population level, as birth between 32 and 38 weeks is common. Future research should focus on determining which subgroups of children born 32 to 38 weeks are at particularly high risk and how these children can be supported to reach their potential.

Thanks to David Brown, Academic Liaison Librarian for Health Sciences, for help with the search strategy design.

ADHD

attention deficit hyperactivity disorder

ASD

autism spectrum disorder

CI

confidence interval

CP

cerebral palsy

DCD

developmental coordination disorder