Prevalence and Associated Factors of Minor Neuromotor Dysfunctions at Age 5 Years in Prematurely Born Children: Title and subTitle BreakThe EPIPAGE Study FREE

Important progress in obstetric and neonatal care in recent decades has improved the survival of increasingly premature infants, but the prevalence of the most severe neurodevelopmental sequelae, such as cerebral palsy,^{1 - 3} cognitive impairment, or behavioral problems,⁴ remains high. Research has focused on children born extremely preterm or with extremely low birth weights and mostly on the severe long-term neurodevelopmental outcomes.^{5 - 8} However, in children without obvious neurological deficits, subtle abnormalities such as difficulties with fine motor function, coordination, and learning may occur.⁹ These minor neuromotor dysfunctions (MNDs), which may have adverse effects on family life and require specific social, educational, and health care resources, can also be found in children born less prematurely.^{10 - 13}

In early childhood (ages 4-6 years), these MNDs can reliably be observed using a detailed and standardized neurological examination such as the one described by Touwen.¹⁴ Some studies have indicated that the quality of general movements and the presence of neurological soft signs in early childhood are related to the type and severity of later neurological and behavioral developmental difficulties^{15 - 17} as well as the presence of learning disorders.¹⁸

We analyzed the results of the Touwen examination at age 5 years in a cohort of children born very preterm and in 2 control groups of children born at 33 to 34 and 39 to 40 weeks of gestation followed up from birth with the same methods. Our aims were to assess the frequency of MND according to gestational age (GA), to describe their association with behavioral problems and learning difficulties, and to determine which sociodemographic or neonatal factors are associated with such dysfunctions in the very preterm group.

The children included are from the EPIPAGE (Etude Epidémiologique sur les Petits Ages Gestationnels) project,¹⁹ which is a population-based cohort of children followed up from birth to age 5 years recruited in 9 French regions in 1997. Three groups of children were included: all of the children born before 33 completed weeks of gestation and 2 control groups, including all of the children born at 33 to 34 weeks during April and October and 1 in 4 children born at 39 to 40 weeks during 1 week. Ethics approval was sought from the appropriate body and parental informed consent was obtained.

All of the children whose parents agreed to participate were enrolled in the study. In 2 regions, half of the infants born at 32 weeks were randomly excluded from the follow-up to reduce the workload. At age 5 years, each child was seen both by a pediatrician who undertook a detailed and standardized physical and neurological examination and by a psychologist who performed a test of cognitive abilities. All of the pediatricians and psychologists were trained to administer these tests. They were not informed of the child's perinatal history at the time of examination. Parents were also asked to fill in a questionnaire on the health, development, and behavior of their child.

The neurological examination was a shorter version of the Touwen examination previously validated in a population of 5-year-old children born very preterm.²⁰ A neurological profile is derived from 16 items grouped into 4 subsystems representing a subset of the neurological repertoire: posture and muscle tone, reflexes, coordination and balance, and motor behavior of the face and eyes. Each of these 4 subsystems is rated as optimal or nonoptimal according to specific criteria based on the presence of a number of dysfunctional signs. The children are then classified as healthy (MND-0) if the 4 subsystems are optimal or as having mild MND (MND-1) or moderate MND (MND-2) if up to 2 subsystems or more than 2 subsystems, respectively, show dysfunctional neuromotor signs. Available information suggests that the reliability of this pediatric neuromotor assessment is fairly good (κ values of interrater agreement for various subsystems, > 0.7).^{17 ,21} As the examination aims to detect minor abnormalities, it is not applicable to children with cerebral palsy, cannot be carried out when severe mental (IQ < 50) or sensory impairment is present, and can only be interpreted if the children perform all of the required tests.

The following social and demographic characteristics were recorded at birth and included in the analysis: mother's nationality (French or other), marital status (mother living alone, yes or no), parity (0, 1-2, or ≥ 3), maternal educational level (university level or not), and mother's age (< 21, 21-34, or > 34 years).

Neonatal characteristics were determined from medical records: GA (defined as completed weeks of amenorrhea based on the date of the last menstrual period and results of early prenatal ultrasonography), sex, multiple births, antenatal steroids, birth weight, small for GA (defined as a birth weight below the 10th percentile of the live births in the EPIPAGE study for a given GA and sex), acute fetal distress, Apgar score less than 7 (5 minutes after birth), Clinical Risk Index for Babies score (validated tool for assessing the initial neonatal risk and severity of illness of the preterm infant, based on birth weight, GA, oxygen requirement, base excess in the first 12 hours of life, and presence of congenital abnormalities, and categorized as < 5, 5-10, or > 10 [with higher scores indicating high mortality and morbidity]),²² confirmed maternal-fetal infection, persistent patent ductus arteriosus, postnatal corticosteroid therapy, bronchodysplasia (defined as oxygen dependence at 36 weeks²³ ), and breastfeeding at discharge. Furthermore, the results of cranial ultrasonographic examinations (1-3 per child) were grouped into 3 levels: infants with consistently normal examination results, those with subependymal hemorrhage or intraventricular hemorrhage (IVH) without intraparenchymal involvement (ie, grade I-II IVH), and those with grade III IVH, white matter abnormalities (persisting echodensities, ventricular dilatation, or unilateral or bilateral cystic periventricular leukomalacia), or intraparenchymal hemorrhage.^{24 - 26}

Cognitive development was assessed using the French version of the Kaufman Assessment Battery for Children,²⁷ which gives a global intelligence scale (equivalent to IQ).

The following parent-reported items were used to assess the children's learning difficulties at age 5 years: learning difficulties at school (numerous or none/few), language impairment (yes or no), and speech therapy (yes or no). Child behavior was assessed using the parent form of the Strengths and Difficulties Questionnaire^{28 - 29} (designed for children aged 4-16 years), which contains 25 items yielding 4 symptom scales (conduct, hyperactivity, emotion, and peer problems) combined into a total difficulties score. An additional domain, the prosocial scale, assesses social competence. The 90th percentiles in a large sample of British children from the general population have been defined as thresholds identifying high risk of psychological problems²⁹ but are not validated for the French version. Therefore, we defined thresholds such that 10% of the children in the EPIPAGE control group (born at 39-40 weeks) were considered at high risk for having a psychological problem: conduct problem, score of 5 or greater; hyperactivity, score of 7 or greater; emotional symptoms, score of 5 or greater; peer problems, score of 4 or greater; total difficulties, score of 16 or greater; and prosocial behavior, score of 6 or less.

For the children who completed all of the 16 items, the results of the Touwen examination were described and compared for each GA group using a trend analysis for proportions. The statistical analyses were weighted to take into account the differences in the proportions of children born at 32 weeks' GA who underwent follow-up in some regions. The Pearson χ² statistic was corrected for survey design.³⁰

To determine the factors associated with MNDs in the group of very preterm children, 2 separate multivariate logistic regression analyses with the same reference group (equivalent to multinomial regression) were carried out: the first compared children with MND-1 with children without MND, and the second compared children with MND-2 with the same children without MND. We tested for interactions between the significant variables in the final model.

We also compared the characteristics of the children with missing items in the neurological examination with those of the children who completed all of the items for the variables that we found to be associated with MND.

The statistical analyses were performed using STATA statistical software version 9.³¹

In total, 2239 children underwent a pediatric examination. Their mean (SD) chronological age was 5.1 (0.2) years, with no difference according to GA at birth. Of these children, the Touwen examination was not applicable to 154 children with a severe disability (Table 1).

Of the 367 children who did not complete the Touwen examination, 275 were born very preterm (18.2% of the children of similar GA to whom the examination was applicable), 47 were born at 33 to 34 weeks (19.4% of the children of similar GA to whom the examination was applicable), and 45 were born at 39 to 40 weeks (13.6% of the children of similar GA to whom the examination was applicable) (Table 1). There was no statistically significant difference in the proportions of unclassified children according to GA (P = .09).

In total, 1237 children born before 33 weeks, 195 children born at 33 to 34 weeks, and 287 children born at 39 to 40 weeks were included in the analysis. The characteristics of this sample according to GA are given in Table 2. Table 3 gives the results of the Touwen examination in terms of the frequencies of nonoptimal items, nonoptimal subsystems, and the neuromotor profile (MND status). Across all of the GA groups, the tests that the children failed the most were those assessing coordination and balance. In total, 41.4% (95% confidence interval [CI], 38.6%-44.2%) of children born very preterm (before 33 weeks) had MND-1 and 3.0% (95% CI, 2.1%-4.1%) had MND-2. These proportions were 30.8% (95% CI, 24.4%-37.8%) and 0.5% (95% CI, 0.01%-2.8%), respectively, for children born at 33 to 34 weeks and 22.0% (95% CI, 17.3%-27.2%) and 0.7% (95% CI, 0.1%-2.5%), respectively, for children born at 39 to 40 weeks. Among the children with MND-1, the proportion of children with 2 suboptimal subsystems decreased with increasing GA (from 37.6% [35 of 93 children] for the children born before 28 weeks to 14.3% [9 of 63 children] for the children born at 39-40 weeks).

Whatever their GA at birth, all of the children went to school at age 5 years. Table 4 shows that MNDs were associated with cognitive development (psychologist's assessment), learning difficulties at school, and behavioral problems (parent-reported) in the very preterm group. After adjusting for maternal sociodemographic characteristics, the presence of MND was found to be independently associated with cognitive impairment (odds ratio [OR], 2.1; 95% CI, 1.4-3.3), learning difficulties (OR, 1.6; 95% CI, 1.1-2.2), and behavioral problems (total difficulties score: OR, 1.4; 95% CI, 1.02-1.9; and emotional problems: OR, 1.4; 95% CI, 1.1-1.9).

We determined which of the maternal sociodemographic or neonatal factors were associated with MND in the very preterm group. None of the mothers' social and demographic characteristics (as recorded at birth) were found to be significantly associated with MND-1 or MND-2. Table 5 and Table 6 show the univariate analyses and the final multivariate models that explain the relationships between neonatal factors and the presence of MND-1 and MND-2. Postnatal corticotherapy was independently associated with an increased risk of MND-1 at age 5 years. On the contrary, children from multiple births and breastfed children had a decreased risk of MND-1. In this model, a significant interaction between breastfeeding and multiple births was found. In singletons, breastfeeding was significantly associated with a decreased risk of MND-1 (OR, 0.8; 95% CI, 0.6-0.99), whereas this effect was not significant in children from multiple births (P = .59). However, in all of the children, postnatal corticotherapy doubled the risk of MND-1 (even after adjusting for bronchodysplasia).

Concerning the risk of MND-2 in the very preterm population, the adjusted analysis of neonatal characteristics showed that being a boy, having had acute fetal distress, having had grade III IVH, white matter abnormalities, or intraparenchymal hemorrhage (on early cranial ultrasonography), and having had postnatal corticotherapy were independent risk factors.

The comparison of the group of unclassified very preterm children who did not complete the Touwen examination with the group of classified children who did complete it showed that 4 of the tested characteristics were significantly more frequent in the unclassified group: being from multiple births (36.7% vs 29.9%, respectively; P = .03), associated intellectual impairment (14.9% vs 9.1%, respectively; P = .008), learning difficulties (19.7% vs 14.7%, respectively; P = .05), and behavioral problems (as measured by the Strengths and Difficulties Questionnaire) (26.0% vs 19.3%, respectively; P = .02).

Our study shows that almost half (44.4%) of the children born very preterm and significant proportions of children born at 33 to 34 weeks (31.3%) and 39 to 40 weeks (22.6%) have mild MND at age 5 years. Furthermore, as GA decreases, the proportion of children classified as having MND-1 with 2 suboptimal subsystems increases. Moderate MNDs are much less frequent (3.0% of children born before 33 weeks).

In the very preterm group, parents of children with MND more frequently report behavioral and learning difficulties than parents of children with no neuromotor problems.

The multivariate model for these children shows that postnatal corticotherapy doubles the risk of MND-1 at age 5 years, whereas children from multiple births and children breastfed at discharge have a decreased risk. The independent risk factors for MND-2 are being a boy, having had acute fetal distress, having had severe abnormalities on cerebral ultrasonography, and having had postnatal corticotherapy.

The population of the EPIPAGE study was defined on a geographical basis, thus avoiding potential recruitment bias. Another strength of this study was that it included 2 control groups of different GAs. Previous studies have shown that GA is a better predictor of mortality and morbidity than birth weight,³² as recruitment based on birth weight leads to overrepresentation of more-mature children with restricted growth.³³ All of the 3 groups and not just the very preterm children were followed up until age 5 years.

At this age, 69.6% of the children who were followed up were examined by a pediatrician. Among the children to whom the Touwen examination was applicable, 17.6% did not perform all of the tests. The circumstances of the examination (distance from home, succession of medical examinations, or unfamiliar environment) may partly explain some children's lack of cooperation. However, refusal to cooperate may also reflect a disability in 1 or several aspects of development. In our study, children who were unclassified following the Touwen examination more often had associated intellectual impairments, learning difficulties at school, and behavioral problems than the classified children. Therefore, it is likely that the true prevalence of MND is underestimated. Wocadlo and Rieger³⁴ found that, compared with cooperative children of a similar GA, a greater proportion of the prematurely born children who refused to cooperate for an examination at age 3 years had minor motor and cognitive difficulties at age 5 years. Another study³⁵ also showed that children monitored with difficulty are more likely to have severe neuromotor disability than those monitored without difficulty.

Our results confirmed that MNDs are more frequent in very premature children than in children born at term and showed an increased risk with decreasing GA as previously described.^{36 - 38} The frequencies found for children born at 33 to 34 and 39 to 40 weeks are close to those published by Hadders-Algra et al³⁶ using the original version of the Touwen examination. The relatively high proportion of MND-1 among children born at 39 to 40 weeks (more than 1 in 5 children) could partly be due to the high sensitivity of the Touwen classification, which is mainly designed for descriptive purposes.¹⁴ However, because this examination has prognostic significance,¹⁶ the children classified as having MND-1 are more likely to have later difficulties than the others. The examination results must obviously be interpreted according to the type and extent of nonoptimal signs and with the child's full history. It should be noted that only 3.2% of the children born at 39 to 40 weeks had at least 2 dysfunctional subsystems. In a study of children born in 1992 and 1993,²⁰ the proportions of MNDs at age 5 years in premature children were lower than the frequencies found in our study. A possible explanation might be that the former study was hospital-based; therefore, the children included may have benefited from closer health surveillance.

The neonatal parameters associated with the existence of MND-2 at age 5 years were mostly the same as the known risk factors for cerebral palsy and could justify the term cerebral palsy a minima sometimes used to describe moderate MND. As has been reported for cerebral palsy,^{1 ,39 - 42} we found that being a boy and having had postnatal corticotherapy were significantly associated with MND-2. The factors most significantly associated with MND-2 were acute fetal distress and ultrasonographic examination results showing grade III or IV IVH or white matter abnormalities. A study by Jongmans et al¹³ confirmed that there exists an association between neonatal cerebral lesions visible using cranial ultrasonography and having MND at age 6 years. Kutschera et al⁴³ have shown that transient periventricular echodensities in very low-birth-weight infants seem to lead to MND in children without major neurological impairment.

Our results show a significantly lower risk of mild MND for children from multiple births than for singletons. Twins are generally considered to be at higher risk than singletons in terms of mortality and morbidity. However, this is mostly due to differences in weight and GA, and the excess risk usually disappears after adjusting for these 2 factors. Few long-term studies on multiple births separately consider children born prematurely. A recent study⁴⁴ compared a cohort of 6- to 12-year-old twins born between 27 and 36 weeks with a control group of singletons of the same age and born at the same GA and found that the twins had a higher risk of minimal brain dysfunction. In another study⁴⁵ of 2-year-old children born before 31 weeks' GA, no significant difference in the incidence of death or severe disability was observed between twins and singletons. However, 2 studies observed as we did that twins were less severely affected than singletons. Bonellie et al⁴⁶ found that the prevalence of cerebral palsy was lower for twins born at 28 to 31 weeks than for singletons of the same GA. Draper et al⁴⁷ studied survival in children born prematurely according to different age groups and weight at birth and found that infants from multiple births had a greater chance of survival. These differences may be explained by the higher rate of pathological maternal complications during pregnancy leading to premature birth in single pregnancies compared with multiple pregnancies.⁴⁸ Also, the frequency of incomplete Touwen examinations in our study was higher in multiple births than in singletons, and these children had similar 5-year characteristics as the children with MND. This potential selection bias could be partly responsible for the relationship between multiple births and lack of MND.

Beneficial effects of breast milk on cognitive skills, behavior, and neuromotor development in very preterm children have been widely demonstrated,^{49 - 51} even though some studies found no effect on neurodevelopmental outcomes.⁵² It has been postulated that the act of feeding at the breast as well as the interaction between mother and child account for more optimal outcomes.⁵³ In our study, breastfeeding was significantly associated with a decreased risk of MND only for singletons. However, a weakness of our study was that the only information we had was breastfeeding at discharge from the neonatal unit without any information on the duration of breastfeeding. Consistent with previous studies,⁵⁴ our findings showed that breastfeeding was significantly less frequent in twins (17.3%) than in singletons (24.7%) (P = .002). It is also likely that the duration of breastfeeding was shorter for multiple births.

The association between MND at age 5 years and parent-reported learning and behavioral difficulties at the same age was particularly strong in this population and even more so for moderate MND. These results are in agreement with previous studies. Children with mild motor delay have lower academic achievement scores at age 8 years.⁵⁵ Batstra et al⁵⁶ reported that 5- to 11-year-old children with MND had poorer school results than others. Behavioral problems at school, especially signs of attention deficit, were also reported in children born prematurely. These children usually are more socially isolated, have less concentration, and are more hyperactive compared with children born at term, and this effect seemed to be mediated by intellectual and neuromotor delays in this population.⁵⁷

Our results show the importance of a close assessment of all children born preterm, whatever their GA. The neuromotor dysfunctions should be tracked through a detailed and systematic examination that requires training⁵⁸ but can be performed in general practice. Owing to their consequences on school learning, it is important to detect these MNDs through screening before school age to offer adapted early intervention. Research should be carried out to identify the type of interventions needed and to assess their efficiency. Furthermore, the relatively high frequency of MND-1 in children born at term requires further investigation.

Birth characteristics, minor abnormalities on neonatal ultrasonography of the brain, and motor milestones have only limited value in the early detection of neuromotor dysfunctions.⁵⁹ It is necessary to investigate whether other early factors could be predictive of dysfunctional neuromotor and behavioral development at school age. Furthermore, it is essential to regularly set up new cohorts as screening methods and health practices are constantly changing.

Correspondence: Catherine Arnaud, MD, Institut National de la Santé et de la Recherche Médicale Unit 558, Faculté de Médecine, 37 allées Jules Guesde, 31073 Toulouse CEDEX, France (carnaud@cict.fr).

Accepted for Publication: May 2, 2007.

Author Contributions: Dr Arnaud had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Arnaud, Grandjean, and Kaminski. Acquisition of data: Pierrat, Larroque, Alberge, Marret, Burguet, Ancel, and Supernant. Analysis and interpretation of data: Arnaud, Daubisse-Marliac, White-Koning, Pierrat, Larroque, Grandjean, and Kaminski. Drafting of the manuscript: Arnaud, Daubisse-Marliac, and White-Koning. Critical revision of the manuscript for important intellectual content: Pierrat, Larroque, Grandjean, Alberge, Marret, Burguet, Ancel, Supernant, and Kaminski. Statistical analysis: Arnaud, Daubisse-Marliac, and White-Koning. Obtained funding: Larroque, Marret, Ancel, and Kaminski. Administrative, technical, and material support: Larroque and Supernant. Study supervision: Grandjean and Kaminski.

EPIPAGE Study Group: Béatrice Larroque, MD, PhD, Pierre-Yves Ancel, MD, PhD, Béatrice Blondel, PhD, Gérard Bréart, MD, PhD, Michel Dehan, MD, PhD, Micheline Garel, MsC, Monique Kaminski, MsC, Françoise Maillard, DUT, Christiane du Mazaubrun, MsC, Pascale Missy, PhD, Fadila Sehili, MD, Karine Supernant, DUT, Institut National de la Santé et de la Recherche Médicale Unit 149, Paris, France; Myriam Durant, DUT, Jacqueline Matis, MD, Jean Messer, MD, Alain Treisser, MD, Hautepierre Hospital, Strasbourg, France; Antoine Burguet, MD, PhD, Laurence Abraham-Lerat, DUT, Alain Menget, MD, Gérard Thiriez, MD, St. Jacques Hospital, Besançon, France; Catherine Lévêque, DUT, Stéphane Marret, MD, PhD, Loic Marpeau, MD, PhD, Charles Nicolle Hospital, Rouen, France; Pierre Boulot, MD, PhD, Jean-Charles Picaud, MD, Arnaud de Villeneuve Hospital, Montpellier, France; Anne-Marie Donadio, DUT, Bernard Ledésert, MD, PhD, Observatoire Régional de la Santé, Montpellier; Monique André, MD, Jean-Louis Boutroy, MD, Jeanne Fresson, MD, PhD, Jean-Marie Hascoët, MD, Maternity Hospital, Nancy, France; Catherine Arnaud, MD, Sylvie Bourdet-Loubère, PhD, Hélène Grandjean, MD, PhD, Institut National de la Santé et de la Recherche Médicale Unit 558, Paul Sabatier University, Toulouse, France; Michel Rolland, MD, PhD, Alain Fournié, MD, PhD, Children and Maternity Hospital, Toulouse; Catherine Leignel, DUT, Pierre Lequien, MD, PhD, Véronique Pierrat, MD, PhD, Francis Puech, MD, PhD, Damien Subtil, MD, PhD, Patrick Truffert, MD, PhD, Jeanne de Flandre Hospital, Lille, France; Georges Boog, MD, PhD, Valérie Rouger-Bureau, DUT, Jean-Christophe Rozé, MD, PhD, Maternity and Children Hospital, Nantes, France; Michel Dehan, MD, Véronique Zupan, MD, Antoine Béclère Hospital, Clamart, France; Michel Vodovar, MD, Marcel Voyer, MD, Institut de Puériculture, Paris.

Financial Disclosure: None reported.

Funding/Support: This study was supported by grants from Institut National de la Santé et de la Recherche Médicale (French National Institute of Health and Medical Research), Merck-Sharp, Dohme-Chibret, Medical Research Foundation, Directorate General for Health of the French Ministry for Social Affairs, and the French Hospital Program of Clinical Research, Paris, France.

Role of the Sponsors: The sponsors of the study had no role in the design and conduct of the study, in the collection, analysis, and interpretation of the data, or in the preparation, review, or approval of the manuscript.

Previous Presentation: This study was presented in part at the XX European Congress of Perinatal Medicine; May 26, 2006; Prague, Czech Republic.