Motor Performance After Neonatal Extracorporeal Membrane Oxygenation: A Longitudinal Evaluation
OBJECTIVE: To assess longitudinally children’s motor performance 5 to 12 years after neonatal extracorporeal membrane oxygenation (ECMO) and to evaluate associations between clinical characteristics and motor performance.
METHODS: Two hundred fifty-four neonatal ECMO survivors in the Netherlands were tested with the Movement Assessment Battery for Children at 5, 8, and/or 12 years. Percentile scores were transformed to z scores for longitudinal evaluation (norm population mean = 0 and SD = 1). Primary diagnoses: meconium aspiration syndrome (n = 137), congenital diaphragmatic hernia (n = 49), persistent pulmonary hypertension of the newborn (n = 36), other diagnoses (n = 32).
RESULTS: Four hundred fifty-six tests were analyzed. At 5, 8, and 12 years motor performance was normal in 73.7, 74.8, and 40.5%, respectively (vs 85% expected based on reference values; P < .001 at all ages). In longitudinal analyses mean (95% confidence interval [CI]) z scores were –0.42 (–0.55 to –0.28), –0.25 (–0.40 to –0.10) and –1.00 (–1.26 to –0.75) at 5, 8, and 12 years, respectively. Mean score at 8 years was significantly higher than at 5 years (difference 0.16, 95% CI 0.02 to 0.30), and mean score at 12 years was significantly lower than at both other ages (differences –0.59 and –0.75; 95% CI –0.33 to –0.84 and –0.49 to –1.00, respectively). Children with congenital diaphragmatic hernia encountered problems at all ages. The presence of chronic lung disease was negatively related with outcome.
CONCLUSIONS: Motor problems in neonatal ECMO survivors persist throughout childhood and become more obvious with time.
- CDH —
- congenital diaphragmatic hernia
- CI —
- confidence interval
- CLD —
- chronic lung disease
- CUS —
- cranial ultrasound
- ECMO —
- extracorporeal membrane oxygenation
- M-ABC —
- Movement Assessment Battery for Children
- MAS —
- meconium aspiration syndrome
- PPHN —
- persistent pulmonary hypertension of the newborn
- VA —
What’s Known on This Subject:
After neonatal extracorporeal membrane oxygenation treatment, children are at risk for neurodevelopmental problems including delayed motor function. So far this has only been studied cross-sectionally until age 7 years.
What This Study Adds:
We describe, in a nationwide evaluation, the longitudinal course of motor function development after neonatal extracorporeal membrane oxygenation with persisting problems up to 12 years. At risk are children with congenital diaphragmatic hernia and those with chronic lung disease.
Extracorporeal membrane oxygenation (ECMO) is a cardiopulmonary bypass technique providing life support when conventional treatment of severe respiratory (and/or cardiac) insufficiency fails. Despite changes in indications for neonatal ECMO treatment, the most frequent diagnoses still are meconium aspiration syndrome (MAS), congenital diaphragmatic hernia (CDH), persistent pulmonary hypertension of the newborn (PPHN), sepsis, and pneumonia.1 The collaborative UK ECMO trial showed improved survival of term infants with severe respiratory failure who were treated with ECMO.2–5 Previously our research group published nationwide data on overall morbidity6 and motor performance7 in Dutch neonatal ECMO survivors at 5 years of age. Notably ball and balance skills were affected. Motor performance beyond age 5 years has been little studied. The UK trial found normal neuromotor development in 24 of 56 (43%) ECMO treated survivors at age 7.3 On the basis of a longitudinal evaluation of exercise capacity after neonatal ECMO, we reported a significant decline of exercise capacity in children of 5, 8, and 12 years.8
We hypothesized, in line with the results of exercise capacity, that problems with motor development would persist throughout childhood.
Therefore, we aimed to evaluate long-term motor performance in neonatal ECMO survivors, the changes in level of motor performance over time in these patients, and associations between clinical characteristics (including physical growth) and motor performance.
A longitudinal follow-up study was conducted in children who all received ECMO support as neonates between January 1996 and December 2006 in either of the 2 ECMO centers in the Netherlands. Both centers used similar inclusion criteria9 and treatment protocols during first admission and at follow-up. Ligation of the carotid artery had been performed in all patients who underwent venoarterial (VA) ECMO. Cranial ultrasound (CUS) examinations had been performed before and serially during ECMO (started on daily basis, with lower frequency after stabilization).10 On the basis of the national consensus on neonatal follow-up and the Dutch Ministry of Health’s requirement to provide relevant data, the assessment protocol is the standard of care in the Netherlands after ECMO. The Medical Ethical Review Board Erasmus Medical Center waived Institutional Review Board approval because “Medical Research in Human Subjects Act does not apply to this study, since subjects are not being submitted to any handling, nor are there rules of human behaviour being imposed.” All parents provided permission to use the data for research purposes.
Procedures and Study Design
Demographic and clinical data were retrieved from medical records and a computerized patient data management system.9 Chronic lung disease (CLD) was defined as oxygen dependency at day 28.11 The assessment protocol encompassed hospital visits at ages 5 and 8 years (Nijmegen) and at ages 5, 8, and 12 years (Rotterdam) with medical assessments including standardized physical examination, evaluation of growth, and questions regarding sports activities.
Motor Performance Assessment
Children’s motor skills were assessed with the Movement Assessment Battery for Children (M-ABC).12 In this study, we have used the first edition of the M-ABC. The test contains 32 items organized into 4 levels of 8 items, each level designed for the use with children of a specific age band. The requirements of the 8 items in each level of the test are identical. The M-ABC evaluates motor function in daily life and is suitable for children without neurologic impairments who can understand and act on instructions. A Dutch standardization study has shown that the original norm scores and cutoff points can also be applied to Dutch children.13 The M-ABC consists of 3 manual dexterity items, 2 ball-skill items, and 3 balance items. Scores for each item are provided ranging from good (0) to very poor.5 A profile of the child’s motor performance for each domain is obtained by summing the relevant item scores. Summation of all item scores produces the total impairment score. The 3 subtest scores and the total impairment score should be interpreted using age-related normative data tables. The range between the 100th and 16th percentile is regarded as “normal”; between the 15th and 6th percentile as “borderline.” The fifth percentile and below is regarded as “definite motor problem.” All tests were administered by experienced pediatric physical therapists. The M-ABC has good interrater reliability (as reflected by κ coefficients ranging from 0.95 to 1.00).14
We calculated z scores for height and BMI using the Dutch Growth Analyzer, version 3.0.15–17 Mean z score of height and BMI were compared with 0 using 1-sample t tests. Differences in medical background variables between the groups “participants,” “children lost to follow-up,” and “children unable to perform the M-ABC” were evaluated by using Kruskal-Wallis tests; differences between the 2 centers using Mann-Whitney tests. Percentages of M-ABC outcome were compared with normative proportions (normal = 85%, borderline = 10%, and motor problem = 5%) using χ2 tests. Interpretation of raw scores of the M-ABC depends on age (different tables for children aged 4 and 5 years and children aged ≥6 years). In the longitudinal analysis, the results for children aged 5 years are combined with the results of older children. Therefore, the raw scores cannot be used in the longitudinal analysis, and the main outcome of the M-ABC is the percentile score. Because the percentile scores of the M-ABC had a skewed distribution, they were transformed into z scores using a probit transformation (ie, an inverse normal transformation).18 These z scores (with mean = 0 and SD = 1 in reference population) served as dependent variable in a linear mixed model, which allows subjects who have some missing outcome data to be included in the analysis, provided that the data are missing at random.19
We included a random intercept for each patient to take into account the repeated measures design of the study. The linear mixed model served to investigate whether age group, underlying diagnosis, and other determinants had a significant influence on motor function outcome. The following determinants were used as covariates in this linear mixed model: time on ECMO, highest oxygenation index before ECMO, highest mean airway pressure before ECMO, presence of abnormal CUS, presence of CLD, z score height and BMI at follow-up, and sports participation. The covariates were first individually entered into the model and afterward all together in a multivariate model. Age group, underlying diagnosis, and the interaction between these 2 variables were also included as independent variables in all linear mixed models. A linear mixed model with only age group as independent variable served to compare mean z scores between different age groups.
In addition, a subgroup of 36 children who underwent all 3 consecutive assessments was analyzed similarly. Statistical significance was accepted at a 5% level. Analyses were performed by using SPSS 20.0 (IBM, Chicago, IL).
Between 1996 and 2006, 434 neonates received neonatal ECMO support. Of those, 116 died before age 5 years (27%); 70 (60%) of the nonsurvivors had CDH, 12 (11%) had MAS, 5 had PPHN (4%), and 29 (25%) had other diagnoses. Of the 318 survivors, 278 children (87%) participated in our follow-up program because 40 children were lost to follow-up (refusal n = 26, not traceable n = 12, organizational reasons n = 2). In 24 follow-up participants (9%), the M-ABC could not be assessed: 15 of them (5% of the survivors) had cerebral palsy with severely impaired motor function. In 8 of these 15 children (53%), cerebral palsy was expected based on abnormal CUS during the ECMO period (see Supplemental Table 6). The others had normal CUS before or during ECMO.
The final analyses concerned 456 assessments performed in 254 children (Fig 1). Time on ECMO and duration of ventilation significantly differed between the Rotterdam and Nijmegen centers: median (range) time on ECMO was 125 (24–510) and 158 (81–367) hours, respectively (P < .001); the median (range) duration of ventilation was 11 (1–130) and 17 (7–54) days, respectively (P < .001). Children who were unable to perform the M-ABC more frequently had abnormal CUS than children in the participants group and children lost to follow-up (P = .008; Table 1). Three participants had abnormal CUS before ECMO: 2 had normal motor function performance, and the third, a child with post asphyxia lesions and deep gray matter injury, had a definite motor function problem without cerebral palsy (Supplemental Table 6).
Patient characteristics at time of follow-up are presented in Table 2.
At 5, 8, and 12 years, motor development was normal in 73.7%, 74.8%, and 40.5%, respectively (vs 85% expected based on reference values; P < .001 at all ages; Table 2). At 5 years, ball and balance skills were most frequently affected, whereas manual dexterity did not differ from reference values. At 8 years, manual dexterity and ball skills were most frequently impaired. At 12 years, scores at all domains were beneath the norm. At 5 and 8 years, the M-ABC outcomes did not differ significantly between both centers (data not shown).
Longitudinal analyses of the z scores of the M-ABC showed mean (95% CI) scores of –0.42 (–0.55 to –0.28), –0.25 (-0.40 to –0.10) and –1.00 (–1.26 to –0.75) at 5, 8, and 12 years, respectively. Mean z score at 8 years was significantly higher than at 5 years (difference = 0.16, 95% CI 0.02 to 0.30), and mean z score at 12 years was significantly lower than at both other ages (difference = –0.59 and –0.75, 95% CI –0.33 to –0.84 and –0.49 to –1.00 respectively; Fig 2).
We evaluated a subgroup of 36 children who underwent 3 consecutive measurements separately (for characteristics and results see Tables 3 and 4). All underwent VA-ECMO. Nine of these children had abnormal CUS (P = .006 presence of abnormal CUS compared between this subgroup and the total group). Other perinatal characteristics did not differ between this subgroup and the other children. Of these 36 children, fewer than expected had normal motor performance (P < .001 at all ages). Longitudinal analyses of the z scores of the M-ABC in this subgroup showed mean (95% CI) scores of –0.54 (–0.88 to –0.21), –0.59 (–0.92 to –0.26), and –1.07 (–1.41 to –0.74) at 5, 8, and 12 years, respectively. Mean z scores at 5 and 8 years were similar (P = .75) and significantly lower at 12 years than at both other ages (P = .001 and P = .002).
At age 5, z scores of height and BMI were both significantly below the norm with a catch-up effect throughout the years (Table 2).
Association of Clinical Characteristics and Motor Performance
At age 5, problems with motor performance (comparison with normative proportions, normal = 85%, borderline = 10%, and motor problem = 5%, using χ2 tests) were mostly encountered in children with CDH and in children from the group with other diagnoses (both groups P < .001), whereas children with MAS and PPHN scored within normal range. At age 8, motor problems were mostly found in children with CDH (P < .001) and PPHN (P = .041), whereas at age 12, children from all subgroups encountered motor problems (MAS, CDH, and other diagnoses P < .001, PPHN P = .005).
In 30 participants with abnormal CUS without cerebral palsy, 8 (27%) had definite motor problems, 4 (13%) had borderline motor performance, and 18 (60%) had normal motor performance. In 2 of those 8 patients with definite motor problems, the lesions did not correlate with outcome; in 3, correlation was not evident; and in another 3 patients, we suspected a correlation between abnormal CUS and outcome (see Supplemental Table 6).
Further analysis using linear mixed models that included age group, diagnosis, and the interaction between age group and diagnosis as independent variables did not show significant relationships between time on ECMO, highest highest oxygenation index before ECMO, highest highest mean airway pressure before ECMO, presence of CUS, z score height at follow-up, z score BMI at follow-up, and sports participation with z scores of the M-ABC. The presence of CLD was negatively related with motor problems (P = .007). Multivariate analyses confirmed that the presence of CLD is predictive for motor problems (P = .02; Table 5).
To our knowledge this is the first longitudinal analysis of motor performance after neonatal ECMO in primary school–age children. We hypothesized that these children would have problems with motor development throughout childhood. At all 3 test ages, the participants scored significantly lower than reference values appropriate for Dutch children, and level of motor performance declined over time. Ball and balance skills seemed to be the most frequently affected. Also in a subgroup of 36 children who underwent 3 consecutive assessments, motor performance was worse at 12 years. This suggests that motor problems not only persist but even increase over time. Children with CDH had motor problems at all ages. The presence of CLD was negatively related with outcome. Of the 24 children (9%) who did not perform the M-ABC because of neurologic or behavioral problems, 15 had cerebral palsy with severely impaired motor function. We therefore excluded these 15 patients with a known disorder of movement or posture who would have had motor function problems at assessment. Should we have included them, the overall results on motor performance would have been worse. In 8 of these 15 children (53%), cerebral palsy was expected based on abnormal CUS during the ECMO period.
Children in the UK Collaborative ECMO trial were assessed at 7 years of age.3 Only 24 of the 55 ECMO-treated children (43.6%) had normal motor development versus 74.8% 8-year-old participants in our cohort. Although the M-ABC was used in both studies, results are difficult to compare because of differences in study group composition (5% of children in the UK trial were CDH survivors vs 22% in our cohort), and differences in assessment (in the UK trial only manual dexterity and balance items of the M-ABC were assessed). Problems with gross motor function and normal manual dexterity were also reported by Glass and coworkers in 5-year-old ECMO-treated children.20
In the current study, a neonatal diagnosis of CDH was a risk factor for delayed motor function. This confirms our earlier observations in 35 eight-year-old CDH survivors who had long-term problems with motor function and concentration, even when they had not been treated with ECMO.21 A high presence of CLD and increased airflow obstruction that deteriorates over time22 may contribute to poor motor function performance in these CDH patients.
We considered several options to explain why motor function performance was worst at 12 years. First, one could argue that the 12-year-old children were treated in an era when ECMO treatment was relatively new and that these children had been sicker before starting ECMO. However, the 12-year-olds were born between May 1996 and April 2000, and ECMO had already been initiated in Rotterdam in the early 1990s. Second, we assume that outcome might be related to the high proportion of abnormal CUS findings during ECMO treatment in these 12-year-olds. Third, despite the higher proportion of abnormal CUS findings in the 36 children with 3 assessments, the mean z scores of the M-ABC at 5 and 8 years of age were not significantly worse in this subgroup compared with the other participants. It is likely, therefore, that at older age when more complicated tasks are being demanded, problems in motor performance become more obvious.
This assumption parallels the observation that the rate of acquisition of new complex skills is reduced in children after severe traumatic brain injury23 and is also in line with findings from our earlier study on executive functions and school performance after neonatal ECMO.24
Almost all participants underwent VA-ECMO. The question is whether the presence of motor function problems would be lower in a population treated with venovenous-ECMO. However, it can be assumed that preexisting conditions might have contributed to the poor motor performance. The finding that having chronic lung disease was the best predictor for poor motor function performance supports the assumption that critical illness in the neonatal period (with the need for ECMO because of illness severity) and not the ECMO treatment itself might contribute to long-term morbidity.
A limitation of the study is that 1 center had no follow-up data beyond 8 years. However, because motor performance at 5 and 8 years did not differ between centers, we assume that a similar pattern could be expected at age 12 as well. Therefore, we applied linear mixed model analysis, which accounts for missing data provided that data are missing at random.
Furthermore, we did not include a control group in our study. With 2 ECMO centers in the Netherlands, covering a relatively small geographic area, the large majority of neonates with similar severity of illness who are not born prematurely (ie, born after 34 weeks’ gestation with birth weight >2000 g) are treated with ECMO. Therefore, it would be difficult to obtain controls with similar severity of illness who survived without ECMO. As an alternative, infants from countries with less access to ECMO treatment could serve as controls. However, variety in treatment protocols and assessments of motor function may bias the results of such a study. For similar reasons, the use of a control group of infants treated before the ECMO era does not seem appropriate. The study design does not permit providing hard evidence that the long-term motor function problems are directly related to the ECMO treatment, but we showed that this group of patients is at risk for such problems and that complications of ECMO treatment (eg, intracranial hemorrhage or cerebral vascular infarction) are associated with motor problems at later age.
We used the first edition of the M-ABC for all assessments that contain a complex manual dexterity task for 11- and 12-year-old children. The manual dexterity score at 12 years might have been influenced negatively by this task. However, deterioration was observed in the ball and balance skills as well, so it is unlikely that this complex manual dexterity task explains the observed deterioration of motor function at 12 years.
Motor function problems in children treated with neonatal ECMO persist throughout childhood and seem to increase as they get older. We found a negative association between the presence of CLD and motor performance. Hypoxia, increased work of breathing, recurrent airway infections, prolonged hospitalization, especially in CDH patients and mainly within the first years of life, may have contributed to our findings. Additional studies with more children assessed at ≥12 years are needed to place deteriorating results into perspective.
This article was written on behalf of the Dutch ECMO follow-up team, consisting of pediatricians, developmental psychologists, speech-language pathologists, and physiotherapists at both centers. The authors thank the entire team and especially the following: Leontien Toussaint (pediatric physiotherapist) and Marjolein Spoel (physician) from Erasmus Medical Center–Sophia Children’s Hospital. Professor M. Nijhuis-van der Sanden was involved in the setup of the pediatric physiotherapy follow-up program in Nijmegen. Ko Hagoort provided editorial advice.
- Accepted May 15, 2014.
- Address correspondence to Hanneke IJsselstijn, MD, PhD, AAP ID: 1244499, Intensive Care and Department of Pediatric Surgery, Erasmus MC–Sophia Children’s Hospital, Room Sk 1280, Dr. Molewaterplein Wytemaweg 80, 3015 CN Rotterdam, The Netherlands. E-mail:
Drs van der Cammen-van Zijp and IJsselstijn contributed to the study conception and design, acquisition of data, analysis and interpretation of data, and writing the first draft; Drs Janssen, Gischler, and van Heijst contributed to study conception and design, acquisition of data, and critical revision of the manuscript; Dr Raets contributed to acquisition of data, analysis and interpretation of data, and critical revision of the manuscript; Dr van Rosmalen contributed to statistical analysis, interpretation of data, and cowriting and critical revision of the manuscript; Dr Govaert contributed to analysis and interpretation of data and critical revision of the manuscript; Dr Steiner contributed to acquisition of data and critical revision of the manuscript; Dr Tibboel contributed to study conception and design and critical revision of the manuscript; and all authors approved the final manuscript as submitted.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: This work was supported by Fonds Nuts-Ohra and Stichting Swart-van Essen Fonds, Rotterdam.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- ↵ECLS Registry Report: International Summary. Ann Harbor, MI: Extracorporeal Life Support Organization; July 2013
- McNally H,
- Bennett CC,
- Elbourne D,
- Field DJ,
- UK Collaborative ECMO Trial Group
- van der Cammen-van Zijp MH,
- Gischler SJ,
- Hop WC,
- de Jongste JC,
- Tibboel D,
- Ijsselstijn H
- ↵Henderson SE, Sugden DA. The Movement Assessment Battery for Children: Manual. San Antonio, TX: The Psychological Corporation; 1992
- Smits-Engelsman BCM
- Smits-Engelsman BCM,
- Fiers MJ,
- Henderson SE,
- Henderson L
- Madderom MJ,
- Toussaint L,
- van der Cammen-van Zijp MH,
- et al
- Spoel M,
- Laas R,
- Gischler SJ,
- et al
- Copyright © 2014 by the American Academy of Pediatrics