Movement Difficulties at Age Five Among Extremely Preterm Infants

We used data from the Screening to Improve Health in Very Preterm Infants in Europe (SHIPS) study²⁵  that followed-up the Effective Perinatal Intensive Care in Europe population-based, prospective cohort of children born before 32 weeks' GA, between 2011 and 2012, in 19 regions in 11 European countries. Data were collected from obstetrical and neonatal records during the neonatal hospitalization using a standardized, pretested instrument by study investigators or medical personnel. When the child was 2 and 5 years of age, questionnaires on health, general development, and socioeconomic circumstances were sent to parents. Clinical assessments evaluating neurocognitive functioning and motor skills (using the Movement ABC-2) were also conducted at 5 years of age for children born EPT.

Our study population was limited to children born EPT who were eligible for the Movement ABC-2 assessment. Out of 1671 EPT infants alive at discharge (Fig 1), 1654 were alive at 5 years, and 1021 were followed-up. As in most studies using the Movement ABC-2, we excluded children with a CP diagnosis (n = 98),^26–28  as this is a well-defined neurodevelopmental disorder,²⁹  as well as children with a severe neurodevelopmental impairment defined as an intelligence quotient (IQ) ≤54 (<−3 standard deviation [SD]) or severe hearing or visual impairment (n = 32) because the Movement ABC-2 was not designed to assess movement abilities in these children.²⁴  Lastly, we excluded children with missing Movement ABC-2 data (n = 119): 88 children were assessed by parental questionnaire only, whereas in 31 the test was incomplete or missing and imputation was not possible. The final sample was comprised of 772 children.

The Movement ABC-2,²⁴  is a validated test to evaluate movement difficulties by age category,^10,30  even in high-risk populations of EPT children.²⁷  It assesses performance on 8 motor tasks in 3 motor components: manual dexterity, aiming and catching, and balance. Scores from all tests were summed and converted to an age adjusted percentile score based on United Kingdom norms that classify movement difficulties into 3 categories: none (>15th percentile), at risk (6th–15th percentile), and significant (≤5th percentile), within each component and globally.²⁴  As national norms exist only in Belgium, France, Italy, the Netherlands, and the United Kingdom,^24,31–33  we uniformly applied the United Kingdom norms that are the most commonly used in the literature.^24,34  The Movement ABC-2 was administered by trained psychologists or physiotherapists. Assessments were conducted by staff in local routine follow-up programs where available (Belgium, the Netherlands, and Sweden) or by the SHIPS research team (Denmark, Estonia, France, Germany, Italy, Poland, Portugal, and the United Kingdom). Although it was not possible to carry out interrater reliability across countries, common data collection guidelines and a core data collection form were developed to standardize procedures and to ensure consistent reporting of the assessment results. Training sessions were held locally, and an online discussion forum was set up to discuss possible problems emerging during the data collection.

Children who had missing Movement ABC-2 data were reviewed on a case-by-case basis by neurodevelopmental specialists and an epidemiologist (R.C., U.A., S.J., and J.Z.). If a child was unable to complete a test-item or component because of severe motor impairment, the lowest score was assigned for that test-item and/or component (n = 7). If data were missing for a test-item or component score in the absence of other developmental problems, the average of the other test-items within the component was used for imputation (n = 11). For 11 Belgian children, percentile scores from the Movement ABC (First Edition) were used.³⁵  In all other cases, scores were left as missing.

Variables selected for this analysis were those hypothesized to affect risks of movement difficulties based on biological plausibility and the scientific literature. Sociodemographic factors included maternal age at childbirth, parity, maternal country of birth (from neonatal records), parental cohabiting status, maternal educational level,³⁶  and household unemployment status (from parental questionnaires). Perinatal and neonatal factors were GA, small for gestational age (SGA),³⁷  sex, multiple birth, premature rupture of membranes >12 hours, any antenatal corticosteroids, congenital anomaly,³⁸  severe brain lesions (intraventricular hemorrhage [IVH] grade III or IV or cystic periventricular leukomalacia [cPVL]), retinopathy of prematurity stage III or more, necrotizing enterocolitis (requiring surgery or peritoneal drainage), bronchopulmonary dysplasia ([BPD], defined as supplemental oxygen at 36 weeks’ postmenstrual age), postnatal corticosteroids,³⁹  and breastfeeding at discharge.

We first described the characteristics of children included in the sample and those excluded because of missing Movement ABC-2 scores. We then described the Movement ABC-2 classifications overall and by component. We produced 3 models to measure the association of sociodemographic, perinatal, and neonatal variables with the probability of being at risk or having significant movement difficulties using multinomial logistic regression taking into consideration clustering within multiple pairs: (1) with no adjustment except country modeled as a fixed effect (termed “unadjusted”), (2) adjusted on sociodemographic and perinatal factors, and (3) with additional adjustments on neonatal factors.

To take into account loss to follow-up, we used inverse probability weighting (IPW).^40,41  As described previously for this cohort,^42–44  we compared the characteristics of responders and nonresponders at the 5-year follow-up (Supplemental Table 5). Variables potentially affecting loss to follow-up were used to estimate the probability of response using multivariate logistic regression and to define a weight inversely proportional to this probability (Supplemental Table 6). Multiple imputation by chained equations was used (m = 20) to impute missing data for covariates used to create the weights.^45,46  We also imputed data for covariates in the final models (m = 20). Three had a percentage of missing >3.0%: household unemployment status (6.7%), parental cohabiting status (5.8%), and maternal education (4.0%). Data were assumed to be missing at random after taking into consideration observed covariates. We did not impute data for children with missing Movement ABC-2 scores as the missing at random assumptions likely did not hold. However, we compared the characteristics of children with and without Movement ABC-2 scores. Final models used IPW and multiple imputation.⁴⁷ 

We conducted sensitivity analyses by rerunning models (1) using the unweighted and complete case samples and (2) using national norms in the countries where these were available, as these may affect movement difficulties classification.³⁴ 

All analyses used the statistical software Stata version 15.0 (StataCorp, College Station, TX, USA).

Children in the study sample were assessed at 5.7 (0.4) years (mean SD) (Table 1). There were 28.6% born <26 weeks' GA, 49.8% were males, and 71.9% were singletons. Their mothers were aged 35 years or older for 26.1%, 58.5% were primiparous, 37.0% had at least a bachelor’s degree, and 18.6% were born outside of Europe. Compared with children in the sample, those without Movement ABC-2 scores were less likely to have young (<25 years) or old (≥35 years) mothers and were more likely to have lower GA and to not be breastfed at discharge from the neonatal unit. For children followed at 2 years of age, missing Movement ABC-2 scores were more common among those with impairment. Loss to follow-up was mainly related to social disadvantage and GA (Supplemental Table 5).

Children classified at risk and with significant movement difficulties comprised 23.2% and 31.7% of the sample, respectively (Table 2). The prevalence of significant movement difficulties was higher for the manual dexterity component (37.4%) than the aiming and catching and balance components (19.2% and 19.7%, respectively).

The prevalence of movement difficulties by risk factor group as well as unadjusted relative risk ratios (RRR) and their 95% confidence intervals (95% CI) are presented in Table 3. Some factors had strong associations with both being at risk for movement difficulties and having significant movement difficulties (ie, household unemployment status, SGA, male sex, retinopathy of prematurity, and BPD), whereas some others were associated with significant movement difficulties only (ie, congenital anomalies, severe brain lesions, and postnatal corticosteroids). Prevalence of significant movement difficulties ranged from 12.4% (the Netherlands) to 72.3% (Poland), whereas the at-risk group ranged from 11.3% (Sweden) to 39.0% (Belgium).

Adjustment on sociodemographic and perinatal factors (model I) slightly attenuated these associations (Table 4). Adjusting for neonatal factors (model II) further reduced the magnitude of associations of low GA, SGA, male sex, and congenital anomalies with significant movement difficulties. Severe brain lesions and postnatal corticosteroids were risk factors for significant movement difficulties, whereas BPD were risk factors for both at risk and significant movement difficulty groups. The other morbidities were no longer significant. These adjustments did not strongly impact the estimated associations with sociodemographic factors. The large differences between countries persisted in adjusted models.

Sensitivity analyses without IPW did not show differences in sample characteristics or prevalence of movement difficulties (Supplemental Tables 7 and 8), and final models using the unweighted and complete case samples yielded similar conclusions (Supplemental Table 9). Redoing models using Movement ABC-2 national norms (Supplemental Table 10) gave broadly similar results and did not affect study conclusions about the key risk factors or the wide variation between countries, although country rankings changed.

Among 772 European children born EPT without CP or severe neurodevelopmental impairment, 23.2% and 31.7% were classified as being at risk and having significant movement difficulties, respectively. Children with lower GA, severe brain lesions, and who received postnatal corticosteroids were more likely to have significant movement difficulties, whereas SGA (ie, <3rd percentile), male sex, and BPD were associated with both being at risk and having significant movement difficulties. Sociodemographic factors, including having younger, primiparous, less educated, and non-European born mothers, were associated with movement difficulties risks. Wide variations in prevalence existed by country and these persisted after adjustment for individual characteristics.

Although children born EPT are more likely to have movement difficulties compared with their term-born peers,^6,10,20,48  prevalence estimates reported in the literature vary by a factor of 5.^{13,18,20–22}  Some of this variability results from the use of different motor function measures, including parental report,⁴⁹  Movement ABC or other tests,^6,10,12  ages of assessment, and population inclusion criteria.²⁰  When comparisons are restricted to studies with samples and methods similar to ours, prevalence is higher with lower variability. For instance, 37.1% of children had Movement ABC-2 scores ≤5th percentile in a Swedish sample born <27 weeks’ GA, after exclusion of children with CP or severe neurodevelopmental impairment.¹³  In an Australian sample of 165 children born <30 weeks’ GA or <1250 g birth weight without CP or IQ <80 points, 47.9% had a Movement ABC (First Edition) score ≤16th percentile.¹⁸  Nonetheless, contextual variations in prevalence persist even between these more standardized samples, as we observed within our European cohort by the differences between countries. Variation is also seen in proportions classified as being at risk or with significant movement difficulties, with some studies finding more children in the most severe category,⁵⁰  as we did. This heavily skewed distribution of scores reflects the high levels of impairment in this population. Finally, when movement difficulties were classified by component, we noted a lower prevalence of impairment for aiming and catching and balance (19% to 20% with significant movement difficulties), compared with manual dexterity (37%). Two previous reviews also noted differences by domain, with lower performance in manual dexterity in some, but not all studies.^10,50 

Some perinatal and neonatal characteristics associated with movement difficulties were reported previously,^6,12,50  including lower GA, SGA, male sex, severe brain lesions, BPD, and postnatal corticosteroids.^27,51  In a recent review, Van Hoorn et al,⁶  reported on several studies showing associations between neonatal brain lesions and motor development. In addition, use of postnatal corticosteroids was associated with motor impairment in 2 out of 3 studies. The risk of developmental coordination disorder (measured by Movement ABC in 22 of 36 studies), was generally inversely related to the GA group, an association found in our sample. Contrary to some reports, premature rupture of membranes, necrotizing enterocolitis, and retinopathy of prematurity were not associated with movement difficulties.^51–53  However, these associations have been inconsistent in the literature^6,12  and may be influenced by population differences (eg, inclusion of children with other major disabilities),⁵²  or methods for measuring comorbidities.⁵¹  Finally, we found that children who did not receive antenatal corticosteroids were less likely to be at risk for movement difficulties, an association not previously reported⁶ ; however, as few children were in this group, this may be a spurious finding. Being able to identify children at risk using perinatal and neonatal risk profiles has the potential to improve long-term motor outcomes as studies find beneficial effects of early intervention.^54,55  Further, the identification of risk factors can generate etiological hypotheses; our finding that risk profiles differed for children classified as at risk versus with significant movement difficulties, requiring confirmation in future work, may suggest varying causal pathways.

Several sociodemographic characteristics increased risks of movement difficulties: low maternal age, lower educational level, parental unemployment, primiparity, and non-European maternal country of birth. Previous studies have reported associations with some sociodemographic characteristics, but often only 1 or 2 characteristics and among younger children below 5 years of age.^12,56–59  Null or contradictory findings may result from sociodemographic factors not being consistently or fully explored.^6,12,50  Understanding how social factors interact with medical and biological risk is an important area for study since it makes it possible to target at-risk groups and understand risk and resilience mechanisms.

After adjustment on sociodemographic and clinical factors, differences in movement difficulties persisted by country of birth. We also found that prevalence estimates at the country-level were sensitive to the choice of norm, as previously shown.³⁴  Taken together, these results raise the possibility that cultural, or policy factors could affect motor outcomes for children born EPT. For instance, preschool, school, or public health programs which differ between European countries^60–62  may influence general motor development and subsequently affect performance on Movement ABC tests, even for children born EPT.^63–65  Further research to confirm and investigate cross-country differences in movement difficulties could inform prevention efforts.

The main strengths of this study are its large, population-based sample of >750 EPT children, standardized collection of sociodemographic, perinatal, and neonatal data and the measurement of movement difficulties using a validated and widely used clinical assessment.^10,12  Few studies have been conducted on prognostic factors for motor impairment in EPT children without CP and most studies of Movement ABC-2 have much smaller samples.¹²  To harmonize results between countries, we applied the Movement ABC-2 United Kingdom norms²⁴  and were able to conduct sensitivity analyses to confirm that risk factor results were robust to use of national norms. An additional strength is that we analyzed children at risk for movement difficulties as well as those with significant movement difficulties. Minor difficulties are less often examined, but they may substantially impact quality of life and have been identified as an important priority for future investigation.⁹ 

Limits include possible attrition bias because of loss to follow-up (61.7% follow-up at 5 years) or bias because of children with a missing Movement ABC-2 score. Because we have full data on the cohort at inclusion, these biases could be described and we were able to use IPW to adjust for loss to follow-up, principally related to social factors, as found in other cohorts.^42–44  In contrast, we did not use statistical methods to adjust for bias because of missing Movement ABC-2 scores. These children were a higher risk subgroup and their Movement ABC-2 assessments were likely not missing at random. This may lead to underestimation of movement difficulties in this population. Further, we found large differences in movement difficulties prevalence between countries but could not explore them in more detail because of limited sample sizes per country. We did not have a control population of children born at term, however studies with controls report good calibration of Movement ABC-2 norms in their control samples.⁵⁰  Although test and examiner reliabilities across sites were not assessed, recommendations by the European Academy of Childhood Disability on developmental coordination disorder concluded that the Movement ABC-2 has good to excellent interrater reliability and test-retest reliability.¹⁷  A final limitation is the study’s observational design and the measurement of some sociodemographic characteristics at the same time as our study outcome. However, our objective was to identify risk-factor associations, not determine causality.⁶⁶ 

In this large population-based sample of 5-year-old children born EPT in Europe without CP, we found that over half were at risk for movement difficulties or had significant movement difficulties. We identified multiple clinical risk factors that could be used to prioritize at-risk children for follow-up and early intervention services. We also showed associations with social factors illustrating the importance of integrating social circumstances into public health programs for children born EPT. Finally, our study found unexplained variability in the prevalence of at risk for or significant movement difficulties by country that raise questions about the role of the broader social or health context in mitigating risks in this population.

All study regions obtained ethical approval according to national legislations. The study was also approved by the French Advisory Committee on Use of Health Data in Medical Research and the French National Commission for Data Protection and Liberties. Parents gave their written informed consent to participating in the study before any data collection.

Belgium (J. Lebeer, I. Sarrechia, P. Van Reempts, E. Bruneel, E. Cloet, A. Oostra, and E. Ortibus); Denmark (K. Boerch and P. Pedersen); Estonia (L. Toome, H. Varendi, and M. Männamaa); France (P.Y. Ancel, A. Burguet, P.H. Jarreau, V. Pierrat, and A. Nuytten); Germany (R.F. Maier, M. Zemlin, B. Misselwitz, and L. Wohlers) Italy (M. Cuttini, I. Croci, V. Carnielli, G. Ancora, G. Faldella, and F. Ferrari); the Netherlands (A. van Heijst and C. Koopman-Esseboom); Poland (J. Gadzinowski, J. Mazela, A. Montgomery, and T. Pikuła) Portugal (H. Barros, R. Costa, and C. Rodrigues); Sweden (U. Aden); United Kingdom (E.S. Draper, A. Fenton, and S.J. Johnson); EFCNI (S. Mader, N. Thiele, and J.M. Pfeil); Health Economics team (S. Petrou, S.W. Kim, and L. Andronis); and Inserm Coordination (J. Zeitlin, A.M. Aubert, C. Bonnet, R. El Rafei, and A.V. Seppänen).

BPD

bronchopulmonary dysplasia

CI

confidence interval

CP

cerebral palsy

cPVL

cystic periventricular leukomalacia

EPT

extremely preterm

GA

gestational age

IPW

inverse probability weighting

IQ

intelligence quotient

IVH

intraventricular hemorrhage

Movement ABC-2

Movement Assessment Battery for Children‒ Second Edition

RRR

relative risk ratio

SD

standard deviation

SGA

small for gestational age

SHIPS

Screening to Improve Health in Very Preterm Infants in Europe