Neurological development in children born moderately or late preterm: national cohort study

Abstract Objective To assess long term neurodevelopmental outcomes of children born at different gestational ages, particularly 32-33 weeks (moderately preterm) and 34-36 weeks (late preterm), compared with 39-40 weeks (full term). Design Nationwide cohort study. Setting Sweden. Participants 1 281 690 liveborn singleton children without congenital malformations born at 32+0 to 41+6 weeks between 1998 and 2012. Main outcome measures The primary outcomes of interest were motor, cognitive, epileptic, hearing, and visual impairments and a composite of any neurodevelopmental impairment, diagnosed up to age 16 years. Hazard ratios and 95% confidence intervals were estimated using Cox regression adjusted for parental and infant characteristics in the study population and in the subset of full siblings. Risk differences were also estimated to assess the absolute risk of neurodevelopmental impairment. Results During a median follow-up of 13.1 years (interquartile range 9.5-15.9 years), 75 311 (47.8 per 10 000 person years) liveborn singleton infants without congenital malformations had at least one diagnosis of any neurodevelopmental impairment: 5899 (3.6 per 10 000 person years) had motor impairment, 27 371 (17.0 per 10 000 person years) cognitive impairment, 11 870 (7.3 per 10 000 person years) epileptic impairment, 19 700 (12.2 per 10 000 person years) visual impairment, and 20 393 (12.6 per 10 000 person years) hearing impairment. Children born moderately or late preterm, compared with those born full term, showed higher risks for any impairment (hazard ratio 1.73 (95% confidence interval 1.60 to 1.87) and 1.30 (1.26 to 1.35); risk difference 4.75% (95% confidence interval 3.88% to 5.60%) and 2.03% (1.75% to 2.35%), respectively) as well as motor, cognitive, epileptic, visual, and hearing impairments. Risks for neurodevelopmental impairments appeared highest from 32 weeks (the earliest gestational age), gradually declined until 41 weeks, and were also higher at 37-38 weeks (early term) compared with 39-40 weeks. In the sibling comparison analysis (n=349 108), most associations remained stable except for gestational age and epileptic and hearing impairments, where no association was observed; for children born early term the risk was only higher for cognitive impairment compared with those born full term. Conclusions The findings of this study suggest that children born moderately or late preterm have higher risks of adverse neurodevelopmental outcomes. The risks should not be underestimated as these children comprise the largest proportion of children born preterm. The findings may help professionals and families achieve a better risk assessment and follow-up.


Introduction
][4][5] Children born moderately or late preterm represent a major healthcare burden in neonatal medicine, 6 7 and even small increases in adverse outcomes may have important consequences from a public health perspective, including the day-to-day functioning of children and their families. In contrast with studies of children born extremely preterm, [26][27][28][29][30][31] most studies of children born moderately or late preterm are not population based. 8-13 15 17 20-25Population based studies are needed for more accurate risk estimates for children born moderately or late preterm, using standardised outcome measures and thus allowing follow-up of neurodevelopmental outcomes over time. 4n this nationwide cohort of more than one million liveborn singleton children of gestational age 32 +0 weeks to 41 +6 weeks, we assessed long term neurodevelopmental outcomes of children born at different gestational ages, particularly those born moderately or late preterm, compared with children born full term.

Data sources
Using the unique personal identity numbers of mothers and children, 32 we linked data from the Swedish Medical Birth Register 33 to several Swedish national registries: the National Patient Register, 34 Total Population Register, 35 Education Register, 36 and Cause of Death Register. 37Extensive validation of the Medical Birth Register has shown high validity for most variables and coverage of prospectively collected information on almost all births in Sweden since 1973. 33The Swedish National Patient Register provides information on primary and secondary diagnoses at discharge for all patients admitted to hospital care since 1987 and from specialised outpatient care units since 2001.

study population
This population based cohort study included 1 496 950 births recorded in the Swedish Medical Birth Register from 1 January 1998 to 31 December 2012.We excluded stillbirths (n=5255), multiple births (n=43 602), children with major congenital malformations (n=51 858), births with missing information on personal identity number of children or mothers (n=1843), children with missing data on infant's sex (n=7), children who emigrated (n=113) or died (n=2025) before age 28 days, children with missing data on gestational age (n=871), and children with gestational age <32 weeks (n=7616) and ≥42 weeks (n=102 070).After exclusions, the study population comprised 1 281 690 liveborn singleton children without congenital malformations born from 32 +0 to 41 +6 weeks (see supplementary figure A).Supplementary table A provides information on the ICD-10 (international classification of diseases and related health problems, 10th revision) codes for major congenital malformations.

gestational age
Gestational age (recorded in days) was determined using a hierarchy: early second trimester ultrasonography (88.4%), date of last menstrual period (6.6%), or postnatal assessment (4.9%). 33To analyse gestational age in weeks as a continuous variable, we divided the days by seven and rounded up to one decimal place.To analyse gestational age as a categorical variable, we rounded gestational age down to completed week and categorised children as born moderately preterm (32-33 weeks), late preterm (34-36 weeks), early term (37-38 weeks), full term (39-40 weeks), and late term (41 weeks). 7

Outcomes
We obtained information on neurodevelopmental outcomes, including motor, cognitive, epileptic, visual, and hearing impairments, from the Swedish National Patient Register.Each outcome was defined as at least one diagnosis of any of the outcomes in the register.A composite outcome of any neurodevelopmental impairment was defined as a diagnosis of one or more of motor, cognitive, epileptic, visual, or hearing impairment.A severe or major impairment was defined as a diagnosis of one or more of cerebral palsy, severe mental retardation, generalised epilepsy, and severe hearing or visual impairment.Supplementary table A provides information on ICD-10 codes for these outcomes.All children born from 1998 to 2012 were followed for each outcome from 28 days after birth until the date of first diagnosis of the neurodevelopmental outcome, death, emigration, 16th birthday, or 31 December 2019, whichever came first.Therefore, each child had a minimum of follow-up of seven years.][40]

Risk diference, % (% CI)
covariates Characteristics reported to be associated with both gestational age and neurodevelopmental impairments were considered as potential confounders based on a directed acyclic graph (see supplementary figure B).Maternal characteristics included age at delivery, 41 42 parity, 41 43 44 country of birth, 41 44 cohabiting status, 41 45 body mass index (BMI) during early pregnancy, 46 47 and smoking during pregnancy. 37 43 48Maternal diseases included diabetic and hypertensive diseases. 2 42 44arents' characteristics included parental highest educational level and parental history of neurological or psychiatric disorder. 41 44We also included information on calendar year of delivery to control for temporal changes in obstetric and neonatal practice and in diagnosis of neurodevelopmental outcomes. 49haracteristics of the infants included infant's sex 44 45 and birth weight for gestational age, the latter being calculated based on the Swedish national sex specific reference curve for fetal growth. 41 44 50Supplementary table A provides the ICD-10 codes for parental diseases.

statistical analysis
Parental and infant characteristics were described among children born moderately preterm (32-33 weeks), late preterm (34-36 weeks), early term (37-38 weeks), full term (39-40 weeks), and late term (41 weeks).We calculated the incidence rates of each outcome studied during follow-up by gestational age group.The number of impaired neurodevelopmental outcomes among the affected children was also described.
To assess the association between gestational age and each outcome of interest, we used Cox proportional hazards regression to estimate hazard ratios along with 95% confidence intervals across the five gestational age groups, with 39-40 weeks as the reference, and between each completed week using 40 weeks as the reference.Age of the child was used as the underlying time scale.Schoenfeld residuals were used to test the proportional hazards assumption.We also estimated risk differences as P(X)−P (40), where P(X) is the risk of developing a neurodevelopmental outcome by age 16 years at a certain gestational age X, and P(40) is the corresponding risk at 40 weeks of gestation (reference).To consider the impact of preterm birth on the neurodevelopmental health of the population, we further estimated the population attributable fraction, defined as the proportion of the cases of neurodevelopmental impairment in the entire population attributable to a specific gestational age group, instead of 39-40 gestational weeks.Hazard ratios, risk differences, and population attributable fractions along with the corresponding 95% confidence intervals were adjusted for maternal characteristics (age at delivery, parity, country of birth, cohabiting status, BMI during early pregnancy, smoking during pregnancy, calendar period of delivery), maternal diseases (diabetic and hypertensive diseases), parental characteristics (highest educational level and history of neurological or psychiatric disorder), and birth characteristics of the infants (sex and birth weight for gestational age).In addition, to assess the potential non-linear relationship of each outcome with gestational age on a continuous scale, we used restricted cubic splines with three knots positioned at the 10th, 50th, and 90th centiles of the distribution of the gestational age variable; the hazard ratios and risk differences were estimated using 40 +0 completed gestational weeks as the reference.To assess the impact of birth weight for gestational age on long term outcomes among children born moderately or late preterm, we estimated hazard ratios stratified by birth weight for gestational age categories among children born preterm.Finally, to account for the correlation among full siblings, we used a robust sandwich estimator to correct standard errors in the analyses.
We performed several sensitivity analyses, estimating hazard ratios for the studied associations.Firstly, because we used complete case analysis in the primary analysis, results might have been biased owing to missing values of confounders (missing proportions in the variables ranging from <0.1% to 10.9%).We therefore conducted the Cox regression analysis using multiple imputation of missing values with chained equations. 51Ten imputations with 50 iterations each were implemented, and the imputation was informed using maternal characteristics, maternal diseases, parental characteristics, birth characteristics of infants, gestational age, and each outcome of interest.Secondly, we performed a sibling comparison analysis to control for unmeasured shared genetic and environmental factors.In this analysis, only full siblings discordant for both gestational age (ie, siblings in different gestational age groups) and outcome (ie, siblings with different time to event) were informative and thus were included.Stratified Cox regression was conducted and adjusted for confounding factors except maternal country of birth and parental educational level.Thirdly, we investigated if the level of risk differed by type of onset of labour (spontaneous versus induced) using formal tests for interaction.Fourthly, because of the difference in coverage of calendar years between inpatient and outpatient data in the National Patient Register, we performed an analysis in which we restricted the population to children born from 2001 to 2012, when data on both hospital admission and outpatient care were available.
Data management and preparation were performed using SAS version 9.

Patient and public involvement
Although we support the importance of patient and public involvement, this study was based on analysis of information available from linkage of anonymised data in national registries.No patients were directly involved in designing the research question or the outcome measures, nor were they involved in doi: 10.1136/bmj-2023-075630 | BMJ 2024;384:e075630 | the bmj developing plans for implementation of the study.No patients were asked to advise on interpretation or writing up of results.The collection of patient data in national healthcare registries in Sweden dates back to the 1970s, when patient and public engagement in healthcare and research was less common.As yet, there are no structured processes in Sweden around those data sources, and how national authorities, professional organisations, and research departments are to manage patient and public involvement.This study also lacked funding for patient and public involvement.However, the impetus for this study was parental concerns about follow-up care of moderately and late preterm infants often expressed by families during their stay in the neonatal intensive care unit.
The total and median follow-up time was 15 772 478.4 person years and 13.1 (interquartile range 9.5-15.9)years, respectively.Overall, 75 311 (47.8Overall, compared with children born full term, children born moderately or late preterm showed higher risks for any impairment; motor, cognitive, epileptic, visual, and hearing impairments; and severe or major neurodevelopmental impairment (table 2).For example, the highest relative risk of neurodevelopmental impairment for children born moderately preterm compared with infants born full term was for motor impairment, with a hazard ratio of 4.70 (95% confidence interval 3.95 to 5.59).
The risk difference for any impairment was 4.75% (95% confidence interval 3.88% to 5.60%)-that is, 475 (95% confidence interval 388 to 560) cases per 10 000 population by age 16 years, when comparing children born moderately preterm with those born full term, showing the highest absolute risk of neurodevelopmental impairment.Children born early term also showed higher risks of neurodevelopmental impairments than children born full term (table 2).When neurodevelopmental outcomes were assessed by gestational age as a continuum, the risks (both relative (hazard ratio) and absolute (risk difference)) for neurodevelopmental impairments were highest at 32 +0 gestational weeks, then gradually declined until 41 +6

weeks (fig 1 and supplementary table D).
Population attributable fractions corresponding to changes in gestational age group showed that the greatest reduction in absolute risk for any neurodevelopmental impairment would be seen in children born at 37-38 weeks if they were born later at 39-40 weeks (2.24%, 95% confidence interval 1.71% to 2.76%).For severe or major impairment, the highest population attributable fractions were observed for children born moderately or late preterm (see supplementary table E).Among children born preterm, birth weight for gestational age between the third and 10th centile was associated with higher risks of any impairment, as well as motor, cognitive, and hearing impairment; these risks, plus those of epileptic, visual, and severe or major impairments, were highest in the lowest birth weight for gestational age (<3rd centile) category (table 3).
After multiple imputations of missing data, the association between gestational age and neurodevelopmental impairment was largely unchanged (see supplementary table F).A comparison analysis on a subset of 349 108 full siblings showed similar results except that no evidence was observed for associations between gestational age and epileptic or hearing impairment; children born early term had a higher risk for cognitive impairment only, compared with children born full term (see supplementary table G).After stratifying on onset of labour, we observed overall similar risk patterns between spontaneous and induced labour, with some higher risks for motor and severe or major impairment for children born spontaneously at 32-33 weeks, and for any and cognitive impairment for children born spontaneously at 37-38 weeks, compared with their counterparts born through induced labour (see supplementary the first visit to antenatal care to discharge from delivery hospital, as well as inpatient and outpatient care.Apart from hazard ratios, we also estimated risk differences and population attributable fractions to provide a comprehensive picture of the studied associations and the public health impact of preterm birth.This study has also some limitations.We were unable to provide precise information on  comparison with other studies Our findings confirm and expand on the results of earlier studies describing higher risks of adverse neurodevelopmental outcomes among children born moderately or late preterm. 8-15 17-25 55][24] Nevertheless, the prevalence of motor, visual, and hearing impairment for infants born at 32-34 weeks in our study are in line with those reported from the EPIPAGE-2 (an epidemiological study on small gestational ages) cohort study, 15 even if the exact definitions of outcomes and lengths of follow-up were not similar.Moreover, we described in detail associations between gestational week and risks of different outcomes with long term follow-up.Interestingly, not only children born moderately or late preterm but also those born early term faced higher risks of adverse neurodevelopmental outcomes.When looking at the whole spectrum of term gestation, children born early term have been reported to have higher risks compared with children born full term for neonatal morbidities during the neonatal period, 7 and for motor and cognitive impairments and lower academic performance during early childhood. 14 22 56n the sibling comparison analysis, the associations between early term birth and neurodevelopmental impairments were attenuated to null.This suggests that the associations between early term birth and adverse neurodevelopmental outcomes might be explained by shared genetic and environmental factors.However, null findings may also imply that the impact of early term birth on neurodevelopment mediated only through familial factors is "controlled away" in sibling comparison analysis. 57Moreover, given that the subset of full siblings only accounts for about a quarter of the entire population, this result might be prone to type II error and should be interpreted with caution.Weekly increased risks have already been reported for autism spectrum disorder by decreasing gestational weeks, in children born full term to early term and to preterm in Sweden. 38 39All these increased risks have an adverse impact on early school performance, 13 21-23 income, and possibilities of completing a university education. 58Although absolute risks are low, even small shifts in the gestational age spectrum might have implications for public health, as moderately or late preterm births constitute 84% of preterm births in Sweden and nearly 80% of preterm births in other high income countries. 41 59plications and future work Compared with children born extremely or very preterm, those born moderately or late preterm are considered as low risk, and in many countries are not included in follow-up programmes. 52However, our results support the findings of no clear cut-off limit before 40 gestational weeks when children can be considered as fully mature, 7 60-62 as children born moderately or late preterm and also early term are more vulnerable compared with children born full term.Results on low absolute risks may help professionals when advising parents and families about risk, to avoid unnecessary anxiety and reassure them.Our findings may also help obstetricians and neonatologists balance the advantages and disadvantages of induced labour in cases of non-spontaneous birth.Professionals must be aware that it might be possible to lower risks in children born preterm or early term by delaying birth and restricting induction of labour before 39 weeks, except for medical reasons. 63During followup of this large population of children born preterm, primary care practitioners, general practitioners, and paediatricians need to be aware of the difficulties that families might face, and be alert to parental concerns to avoid delayed referrals to specialised services for these children, particularly for those born preterm and small for gestational age.Our findings support the strategy to prevent births before full term to decrease the risk of neurodevelopmental impairments.Targeting health policies focused on population risk factors for the full spectrum of early delivery (<39 weeks), including pregnancy complications, maternal sociodemographic and lifestyle characteristics, environmental factors, and medical practices (eg, provider initiated delivery) could have a synergistic impact on the avoidance of early delivery. 41Future studies could evaluate causal pathways resulting in adverse outcomes, such as the reason for prematurity and neonatal morbidities, 64 and strategies for prevention or intervention.It might also be considered whether a larger proportion of children born preterm should be subjected to some structural follow-up after discharge from neonatal care, especially those born small for gestational age.Also, improving the knowledge of education professionals about the needs of children born preterm might improve early recognition and referral to specialised services and thus enhance appropriate support for these children. 15

conclusion
In this large population based cohort study, we found long term neurodevelopmental impairments in a broad range of areas among the largest group of children born preterm, reflecting the continuity of risk across the gestational age spectrum.This global perspective is important when advising parents and health professionals, and also when planning healthcare systems for children born preterm.Our findings support that preventing moderately or late preterm delivery may have implications for public health, and that higher risks faced by these groups of children and their families should not be underestimated.
Fig 1 | association between gestational age and neurodevelopmental outcomes among liveborn singleton children without congenital malformations in sweden 1998-2012.risk difference is the difference in risk of neurodevelopmental outcome by age 16 years comparing different gestational ages.Hazard ratios and risk differences are adjusted for maternal age at delivery, parity, country of birth, cohabiting status, body mass index during early pregnancy, smoking during pregnancy, diabetic and hypertensive diseases, calendar period of delivery, parental highest educational level, parental history of neurological or psychiatric disorder, and infant's sex and birth weight for gestational age.children born at 40 +0 weeks are the reference.any impairment was defined by at least one of the following: motor, cognitive, epileptic, visual, or hearing impairment.any severe or major impairment was defined by a diagnosis of cerebral palsy, severe mental retardation, generalised epileptic disorder, or severe hearing or visual impairment

table H
In this Swedish nationwide cohort study of more than one million children born at 32-41 weeks, we found those born moderately preterm(32-33 weeks)or late preterm (34-36 weeks) showed higher risks of any long term neurodevelopmental outcome, such as motor, cognitive, and visual impairment, than children born table 1 | characteristics of parents and of liveborn singleton children of gestational age 32-41 weeks without congenital malformations in sweden 1998-2012.values are number (column percentage) unless stated otherwise At least one of motor, cognitive, epileptic, visual, or hearing impairment.†Diagnosis of cerebral palsy, severe mental retardation, generalised epileptic disorder, or severe hearing or visual impairment.‡Number with outcome per 10 000 person years.§Adjusted for maternal age at delivery, parity, country of birth, cohabiting status, body mass index during early pregnancy, smoking during pregnancy, diabetic and hypertensive diseases, calendar period of delivery, parental highest educational level, parental history of neurological or psychiatric disorder, infant's sex, and birth weight for gestational age.¶Difference in risk of a specific neurodevelopmental outcome by age 16 years comparing different gestational age groups.
discussion BMI=body mass index.*Numbers and row percentages.†Includes Sweden, Denmark, Finland, Iceland, and Norway.doi: 10.1136/bmj-2023-075630 | BMJ 2024;384:e075630 | the bmj strengths and limitations of this study A major strength of the study is the population based design and the large sample size using comprehensive a nationwide cohort.As children born moderately or late preterm receive the same routine care as children born at term in Sweden as in many other countries, 52 53 misclassification of outcomes related to gestational age is unlikely.We were able to adjust for potential confounders known to affect both gestational age and neurodevelopment, based on prospectively collected data on gestational age, covariates, and outcomes from 1.08 (1.06 to 1.11) 1.28 (1.20 to 1.38) 1.14 (1.10 to 1.17) 1.06 (1.01 to 1.11) 1.10 (1.05 to 1.14) 1.04 (1.00 to 1.08) 1.10 (1.03 to 1.17) * association between gestational age and neurodevelopmental outcomes among liveborn singleton children without congenital malformations in sweden 1998-2012.risk difference is the difference in risk of neurodevelopmental outcome by age 16 years comparing different gestational ages.Hazard ratios and risk differences are adjusted for maternal age at delivery, parity, country of birth, cohabiting status, body mass index during early pregnancy, smoking during pregnancy, diabetic and hypertensive diseases, calendar period of delivery, parental highest educational level, parental history of neurological or psychiatric disorder, and infant's sex and birth weight for gestational age.children born at 40 +0 weeks are the reference.any impairment was defined by at least one of the following: motor, cognitive, epileptic, visual, or hearing impairment.any severe or major impairment was defined by a diagnosis of cerebral palsy, severe mental retardation, generalised epileptic disorder, or severe hearing or visual impairment doi: 10.1136/bmj-2023-075630 | BMJ 2024;384:e075630 | the bmj

table 3 |
neurodevelopmental outcomes by birth weight for gestational age among preterm (32-36 weeks) liveborn singleton children without congenital malformations in sweden 1998-2012 (n=55 746) *At least one of motor, cognitive, epileptic, visual, or hearing impairment.†Diagnosis of cerebral palsy, severe mental retardation, generalised epileptic disorder, or severe hearing or visual impairment.‡Number with outcome per 10 000 person years.§Adjusted for maternal age at delivery, parity, country of birth, cohabiting status, body mass index during early pregnancy, smoking during pregnancy, diabetic and hypertensive diseases, calendar period of delivery, parental highest educational level, parental history of neurological or psychiatric disorder, infant's sex, and gestational age.