A Systematic Review and Meta-Analysis of Long-Term Development of Early Term Infants

Historically, gestation has been treated as a binary factor, with 37 gestational weeks defined as the threshold of fetal maturity [1]. Compared with term infants (37 0/7 to 41 6/7 gestational weeks), neonates delivered before 37 weeks are associated with a significantly higher risk of morbidity and mortality, and the risk increases with increasing degree of preterm birth [2]. Infants born at term, on the other hand, are considered healthy and treated as a homogeneous group with respect to neonatal outcomes. However, emerging evidence shows that term births prior to 39 gestational weeks are more likely to have adverse outcomes than those after 39 weeks, indicating gestation is a biologic continuum and the dose-effect exists even within the term spectrum [3]. In the last decade, births occurring between 37 0/7 and 38 6/7 weeks, newly defined as early term infants (ETIs), increased by nearly 20%, while the delivery of late preterm infants (34 0/7 to 36 6/7 weeks) increased by 12% and births after 40 gestational weeks have declined [1,4]. ETIs are the fastest growing subgroup of neonates, and account for 17.5% of the annual live births in the United States, whereas preterm births constitute approximately 12.5% in the same cohort [5,6]. Data from other regions in the world are largely unknown due to the general unawareness of infants born early term. From a public health perspective, ETIs are the main target of health services because the majority of adverse outcomes are attributable to this special subgroup of neonates.

The negative effect of being born early term on short-term morbidity and mortality has been well documented [3,6]. Although ETIs have a significantly higher risk of morbidity and are more likely to die in infancy when compared with term neonates born after 39 gestational weeks, an increasing number of infants born early term can survive into adolescence or adulthood on account of the progress in neonatal care [3]. In contrast to the abundance of information on short-term outcomes, there is a dearth of research relating to the long-term development of ETIs, and this problem has not been systematically evaluated as far as we are aware. Addressing this issue will help optimize management of a majority of neonates and improve general public welfare. In light of the fact, we undertook a systematic review to explore the long-term effect of being born early term on development.

A list of literature was identified by searching electronic databases including MEDLINE (1980–2012), EMBASE (1980–2012), Cochrane Database of Systematic Reviews (the Cochrane Library, issue 1, 2012) and Cochrane Central Register of Controlled Trials (CENTRAL, the Cochrane Library, Issue 1, 2012) in January, 2012. The following terms were used for searching: early term, or 37 to 38 weeks, and long-term outcome, development, neurodevelopment, child development, health, cognitive, motor, behavior, attention deficit, intelligence, and school. No language restriction was applied. Additionally, the abstract books from the Society of Pediatric Research from 1998 onwards were manually searched during the same time. The reference list of identified studies and reviews were sought for additional citations.

Retrieved articles were assessed for eligibility. To be included in the study, the participants had to be ETIs (37 0/7 to 38 6/7 gestational weeks). The control group (term infants born after 38 completed gestational weeks) was defined; long-term developmental outcomes of both ETIs and infants in the control group were reported; data and research methodology were adequately described. Studies reporting outcomes by week of gestation were also deemed eligible as long as term births were included. Reviews and follow-up studies on a certain preexisting disorder (e.g. small for gestational age, SGA) that would possibly increase the likelihood of a compromised developmental outcome were excluded. Two reviewers searched literature independently, and discrepancies were resolved by discussions until consensus was achieved. Data on authors, study location, study design, participants, age at assessment and outcomes were tabulated. Where data were incomplete, the corresponding author was contacted for further information.

Included were non-randomized prognostic studies aimed at investigating long-term outcomes of infants. In light of limited information on the appraisal of such studies, we performed the quality assessment by referring to previous studies [7]. Hayden’s framework and the International Society for Pharmacoeconomics and Outcomes Research retrospective database checklist were used to assess the following 7 areas of potential bias: data source, study participation, study attrition, prognostic factor measurement, confounding measurement and account, outcome measurement and analysis [8,9]. An ultimate appraisal of yes, partly, no or unsure was given to each area after generally assessing a list of quality items relevant to it [8]. This type of quality assessment tool is recommended against a score scale because scales are considered to reduce scientific judgment and are more likely to include criteria that do not directly relate to study validity [8,10]. Two reviewers assessed the quality of studies independently. Kappa value was used to evaluate the inter-rater agreement.

Outcomes of included studies were divided into the following 4 categories: neurodevelopment, schooling problems, behavior and emotional status, and long-term social outcomes. Relative risk (RR) and 95% confidence interval (CI) were calculated for dichotomous outcome variables. Continuous outcome variables were expressed as mean difference and 95% CI. If two or more studies regarding the same outcome were retrieved, a meta-analysis was conducted by using RevMan 5. If appropriate, a fixed effect model was used for combining results in the meta-analysis. Heterogeneity was defined as a significant test of p < 0.1, and the I² statistic was used to estimate the heterogeneity. In addition, a sensitivity analysis was carried out to assess the heterogeneity. On the other hand, a descriptive methodology was chosen and a narrative synthesis was performed for studies with irrelevant outcomes.

The initial search yielded 1,671 potential related citations. After excluding reviews, irrelevant studies and studies not reporting on ETIs, a total of 12 studies were identified. Attempts were made to contact authors of 2 studies for further information [11,12]. One of them (Dr. Mathiasen) gave a response and the other failed to be contacted. Finally 11 eligible studies were included and the selection process is detailed in figure 1. Characteristics of the 11 studies are summarized in table 1 and quality assessment is presented in table 2[11,13,14,15,16,17,18,19,20,21,22]. None of the 11 studies focused solely on ETIs, and all of them included a subgroup of infants born at 37–38 gestational weeks within the study population. Most of the included studies showed an overall low level of risk for bias. The quality assessment demonstrated high concordance between the 2 raters as measured by kappa value (ĸ = 0.741; p < 0.0001), indicating there were no major systematic disagreements. A summary of the results of included studies were shown in table 3.

Neurodevelopmental disabilities have been defined as ‘a group of heterogeneous conditions that share a disturbance in the acquisition of basic developmental skills in a chronologically appropriate manner [23].’ Cerebral palsy (CP) and intellectual disability were included in this broad definition. Data for CP were reported in 3 studies [11,17,22], and a random-effects meta-analysis showed that the pooled risk of CP was 1.75 (95% CI 1.32, 2.31) times higher for ETIs than infants born at 39–41 weeks’ gestation (fig. 1). However, the χ² test for heterogeneity was significant (χ² = 20.38, p < 0.0001). Among the 3 studies, Moster et al. [22] showed the lowest risk of CP for ETIs, which could be explained by the fact that their data are retrieved by using a disability register, and only children surviving to 4 years old are included, thus fewer patients may be included. Heterogeneity was no longer significant after excluding the data reported by Moster et al. [22] from this meta-analysis, and the pooled risk of CP was 1.99 (95% CI 1.79, 2.22), suggesting that the follow-up period is a factor explaining the heterogeneity. There were 3 studies that focused on the impact of being born early term on intelligence scores [13,18,21]. However, the data were not eligible for quantitative syntheses due to the widely varying scales used for measurement. It is noteworthy that the score gap between early term group and control group narrowed down when confounding factors like maternal education and family socioeconomic status (SES) were adjusted [13,21]. Ekeus et al. [18] showed that accounting for perinatal factors such as SGA would further dilute the association between gestation and intelligence scores.

Schooling problems have been broadly defined as the need for special education, education below age level, special support in regular school, or poor performance in comparison with pupils in the same class [24]. The 5 studies relating to schooling problems were highly variable with respect to the population size, demographic factors, outcomes and follow-up periods [11,14,15,16,19]. Two of the 5 studies reported data on difficulties in requiring mathematical skills and were subjected to meta-analysis [14,15]. We noted that the risk of having mathematical difficulties for ETIs was not significant in Kirkegaard’s study [RR 0.82 (95% CI 0.42, 1.6)], which could be explained by the small population size and a low reporting rate [15]. However, the pooled result showed a significantly higher risk of having mathematical difficulties among early term group than their controls [RR 1.13 (95% CI 1.04, 1.21)] with no significant heterogeneity between the studies (I² = 0%; p = 0.35), indicating the power of detecting an effect was increased after data combination (fig. 2). It can be seen from table 3 that after taking a series of confounders into account, the risks of having several schooling problems for ETIs decreased [14,15,16]. However, the significance remained even after an extensive adjustment was performed.

Lindström et al. [17] carried out a population-based cohort study to analyze the effect of preterm birth on the risk of attention-deficit/hyperactivity disorder (ADHD) among school age children. The adjusted risk of receiving ADHD medications was significantly higher among ETIs than that among infants born at 39–41 weeks of gestational age (GA). However, the actual risk for ETIs might be underestimated due to the fact that receiving ADHD medications indicates severe cases of ADHD, and it is highly probable that much milder cases may be left unaccounted for [17]. In another cohort from the same country, Lindström et al. [20] explored the hazard ratio of having a psychiatric disorder for ETIs. The risk was reduced from 1.11 (95% CI 1.05, 1.16) to 1.06 (95% CI 1.01, 1.12) after taking perinatal factors, SES and the home environment into account [20].

A large-scale cohort study analyzed the effect of GA on disability and vocational success among young adults [19]. Potential confounding factors included maternal factors, SES and single-parent household. In comparison to adults born at 39–41 gestational weeks, those born early term carried significantly increased risk for disability [adjusted RR 1.26 (95% CI 1.17, 1.35)]. However, the chance of getting employed [adjusted RR 0.99 (95% CI 0.98, 1.02)] and the mean net salary (mean difference: –6 EUR; p = 0.907) were identical between the two groups, suggesting the adverse effect of being born early term on certain social outcomes might be attenuated by long-term postnatal environments [19].

ETIs are a newly defined subgroup of neonates who are beginning to arouse attention due to their suboptimal long-term outcomes shown by emerging evidence. In this systematic review of long-term developmental outcomes of ETIs, none of the 11 retrieved articles focused solely on ETIs, highlighting a paucity of relative literature. Although wide heterogeneity existed among included studies, all authors reported a consistent trend of poorer outcomes among ETIs than that among infants born at 39–41 weeks of GA. It should be noted some cohorts were born before ultrasound became a routine procedure to measure GA, and a considerable number of individuals born term were miscoded as preterm, which may lead to underestimation of the effect of being born early term on long-term risks [11,13,17,18,19,20,22]. Compared with infants of 39–41 gestational weeks, a significant increase in the risk of CP and mathematical difficulties among ETIs was shown by our meta-analyses.

The traditional assumption of gestation to be a dichotomous developmental course has been challenged as mounting evidence demonstrates that the dose-dependent effect of gestation on neonatal outcomes exists not only across the spectrum of preterm but also within the boundary of term [1,2]. Over the last several years, risks posed to late preterm infants, a marginally term group, have been sufficiently realized, and much effort has been dedicated to prevent late-term deliveries [25,26]. Recently, ETIs were defined to raise concerns for infants born at the more optimistic stage of term, stressing the fact that gestation is a biological continuum, and any classification tends to be arbitrary [3]. A reverse J-shaped relationship between the risk of suboptimal outcomes and GA across the whole gestational spectrum indicates that the optimal time of gestation occurs somewhere between 39 and 41 completed weeks, a much narrower interval than the historically held 5-week term period [16,22]. A delivery occurring earlier or later may increase the risk of adverse outcomes.

Our results demonstrated that each additional week of gestation prior to 39 gestational weeks was accompanied by increased intelligence scores and a reduced risk of CP. The gradient suggests that the relation between GA and brain development is present across the whole gestation and not confined solely to preterm infants. Active myelination and significant growth in various kinds of brain cells are observed between 34 and 40 weeks of GA, which contributes to the 50% increase in cortical volume and the 25% increase in cerebellar development [27]. Inge et al. [28] assessed the brain volumes of normally developing children when they were 9 years old, finding that ETIs had a smaller cerebellar volume than their counterparts who were born at 39–41 weeks’ gestation even after adjusting for birthweight, age, sex and head size. Davis et al. [29] found that among healthy term infants without any perinatal risk factors such as intraventricular hemorrhage and periventricular leukomalacia, a longer gestation was associated with an increase in gray matter density of certain brain regions. These data suggest that shortened gestations even within the spectrum of term can interrupt the process of brain maturity. Additionally, perinatal risk factors such as neonatal morbidities and medical treatments can compound existing biological vulnerabilities and impede postnatal brain development [29]. Apart from obvious disabilities like CP, early term births are at risk of subtle developmental impairments such as schooling problems, behavioral and emotional problems, and social disadvantages, the underlying mechanisms being largely unknown [11,13,14,15,16,19]. It is noted that associations between GA and adverse long-term outcomes are attenuated with adjustment for potential confounders, most of which are associated with parental characteristics and housing conditions, suggesting that postnatal environments may compensate for neonatal disadvantages. Therefore, a long-term follow-up combined with appropriate interventions may help identify and alleviate potential problems encountered by ETIs during their process of development.

The risk of bias is considered to be higher in a systematic review of non-randomized studies due to the lack of randomization and loss to follow-up. The strength of this review is that most included studies are high-quality population-based cohort studies, which is considered helpful in reducing selection bias. Even though current evidence generally supports that ETIs have an increased risk of suboptimal long-term development, a conclusive statement cannot be reached given the limitations of included studies and the wide heterogeneity among them. Although extensive adjustments are conducted in included studies, most of the participants retrospectively enrolled were born prior the last decade or even as much as 4 decade ago, during which time the medical technology evolved rapidly, and the care provided to ETIs may not be uniform; therefore, the chance of confounding bias remains high. Additionally, the prevalence rate of an adverse outcome among ETIs is considerably low when compared with that among infants born more preterm; a large sample of population might be needed to detect an effect. Therefore, reporting biases will arise as small-size studies with negative results are not likely to be reported or published. As demonstrated in our review, wide variations exist in population size, demographic characteristics, follow-up period and outcomes of interest. The results of our meta-analysis should be interpreted with caution due to the limited number of included studies. Standardization of future study protocols is needed, so that more studies can be quantitatively assessed and the risk of suboptimal outcomes under different categories can be adequately evaluated for ETIs.

ETIs are a subgroup of neonates assuming great importance in public health. However, the long-term risks associated with early term births have not been well understood due to the heterogeneity and paucity of current data. A complex series of factors can impact the trajectory of development in ETIs, which needs to be clarified by further studies with an adequate population size and a standardized study design. As evidence accumulates, the management of ETIs can be guided and optimized.

We would like to thank Dr. Mathiasen for providing additional data of their published study.