Predictors of cerebral palsy in very preterm infants: the EPIPAGE prospective population-based cohort study
Abstract
Aim The aim of this study was to assess the independent role of cerebral lesions on ultrasound scan, and several other neonatal and obstetric factors, as potential predictors of cerebral palsy (CP) in a large population-based cohort of very preterm infants.
Method As part of EPIPAGE, a population-based prospective cohort study, perinatal data and outcome at 5 years of age were recorded for 1812 infants born before 33 weeks of gestation in nine regions of France in 1997.
Results The study group comprised 942 males (52%) and 870 females with a mean gestational age of 30 weeks (SD 2wks; range 24–32wks) and a mean birthweight of 1367g (SD 393g; range 450–2645g). CP was diagnosed at 5 years of age in 159 infants (prevalence 9%; 95% confidence interval [CI] 7–10%), 97 males and 62 females, with a mean gestational age of 29 weeks (SD 2wks; range 24–32wks) and a mean birthweight of 1305g (SD 386g; range 500–2480g). Among this group, 67% walked without aid, 14% walked with aid, and 19% were unable to walk. Spastic, ataxic, and dyskinetic CP accounted for 89%, 7%, and 4% of cases respectively. The prevalence of CP was 61% among infants with cystic periventricular leukomalacia, 50% in infants with intraparenchymal haemorrhage, 8% in infants with grade I intraventricular haemorrhage, and 4% in infants without a detectable cerebral lesion. After controlling for cerebral lesions and obstetric and neonatal factors, only male sex (odds ratio [OR] 1.52; 95% CI 1.03–2.25) and preterm premature rupture of membranes or preterm labour (OR 1.72; 95% CI 0.95–3.14) were predictors of the development of CP in very preterm infants.
Interpretation Cerebral lesions were the most important predictor of CP in very preterm infants. In addition, infant sex and preterm premature rupture of membranes or preterm labour were also independent predictors of CP.
List of Abbreviations
-
- IPH
-
- Intraparenchymal haemorrhage
-
- IVH
-
- Intraventricular haemorrhage
-
- PPROM
-
- Preterm premature rupture of membranes
-
- PTL
-
- Preterm labour
-
- PVL
-
- Periventricular leucomalacia
Very preterm infants (born before 33wks of gestation) are at high risk of developing motor deficiencies during childhood, particularly cerebral palsy (CP).1 Cerebral lesions and several other neonatal and obstetric risk factors for CP have been described in the literature.2–4 These factors include, in particular, gestational age, low birthweight, multiple gestation, intrauterine infection, placental abruption, preterm labour or pre-labour rupture of membranes, respiratory distress syndrome, and neonatal sepsis.
Since the early 1990s, advances in perinatal and neonatal care have improved the survival of very preterm infants.5 Although the incidence of severe cerebral lesions (i.e. white matter disease) has decreased significantly,6,7 that of CP has not.1,8 Indeed, in half of very preterm children who develop CP, no severe cerebral lesion is apparent on neonatal cranial ultrasound scans.9 Nevertheless, it is still strongly recommended that ultrasonography be performed in all infants born at a gestational age of less than 30 weeks and that it be used to predict long-term neurodevelopmental outcome.10
Recent studies have evaluated the relationship between risk factors at birth and motor outcome in very preterm infants, but none has assessed the ability of a wide range of perinatal factors to predict the development of CP taking into account the results of ultrasound scanning. This information could help clinicians to make a prognosis regarding individuals’ motor function as well as to define diagnostic entry criteria and evaluate the effects of interventions in clinical trials studying the impact of neuroprotective strategies on long-term outcomes.10 Our aim was to assess the independent role of cerebral lesions detected on ultrasound scans and several other neonatal and obstetric factors as potential predictors of the development of CP in a large population-based cohort of very preterm infants (EPIPAGE study [Epidémiologique des Petits Ages Gestationnels]; see Appendix).
Method
This study was approved by the French data protection agency (Commission Nationale de l’Informatique et des Libertés – CNIL).
Participants
All infants born between 22 and 32 weeks of gestation in nine regions of France in 1997 (1 January to 31 December) were included in the EPIPAGE study.11 Among 2901 live births (Fig. 1), 127 infants died in the delivery room, including all infants born at 22 weeks of gestation. Among the 2774 infants admitted to neonatal intensive care units, 315 died, including all those born at 23 weeks of gestation; 2459 infants were discharged alive. The protocol allowed regions the option of following at random only one out of every two infants born at 32 weeks of gestation to reduce their workload. Two regions exercised this option and, as a result, 77 infants were not included in the follow-up. Twenty-five infants died after hospital discharge but before the age of 5 years. Overall, 2357 infants were eligible for follow-up. Parents were given written information and provided oral consent to the medical team in charge of the study.
Study population. aThe protocol gave the regions the option of following randomly one out of every two infants born at 32 weeks of gestation in order to reduce their workload. Two regions exercised this option and, as a result, 77 infants were not included in the follow-up. NICV, neonatal intensive care unit.
Cerebral lesions
Neonatal cranial ultrasound scanning is standard practice in very preterm infants and is repeated during hospitalization to diagnose and/or follow the progression of cerebral lesions. In France, scanning is usually performed on one to three occasions during the first 2 weeks of life, then once a week in the case of infants with lesions or once every 2 weeks in infants without lesions. In the EPIPAGE study, cranial ultrasound scanning was performed using high-frequency 7.5MHz transducers by qualified neonatologists or radiologists who routinely performed cranial ultrasonography. In total, 97% of infants in the EPIPAGE study underwent cranial ultrasonography at least once during the neonatal period; among this group, scanning was performed once in 11%, twice in 23%, and three or more times in 66%.12
Two major types of cerebral lesion were assessed: intraventricular haemorrhage (IVH) and white matter disease (intraparenchymal haemorrhage [IPH], periventricular leucomalacia [PVL], and ventricular dilatation).13 Subependymal IVHs were classified as grade I, intraventricular IVHs were classified as grade II, and IVHs associated with ventricular dilatation were classified as grade III. IPH included large unilateral parenchymal hyperdensity or a large unilateral porencephalic cyst. PVL was defined as the presence of periventricular white matter echolucencies (cystic PVL) or echodensities persisting for more than 14 days without cyst formation. Cystic PVLs were characterized by their laterality and their localization (parieto-occipital, frontal, or other). Ventricular dilatation was defined as isolated dilatation of ventricles with no associated IVH.
When several cerebral lesions were observed, the most severe cerebral lesion was considered. In order of decreasing severity, cerebral lesions were cystic PVL (class 1), IPH (class 2), persistent echodensities or ventricular dilatation (class 3), and grade III (class 4), grade II (class 5), and grade I IVH (class 6). Infants without an identified cerebral abnormality constituted the reference group.
Perinatal risk factors
Perinatal data, including data on known and potential risk factors for CP, were collected using questionnaires administered in the maternity ward and the neonatal transfer unit. Obstetric information included gestational age, in completed weeks of amenorrhoea (based on the date of the last menstrual period and an early prenatal ultrasound, which is standard practice for almost all pregnant females in France); infant sex; whether the infant was small for gestational age (birthweight below the 10th centile for gestational age and sex among live-born infants in our study sample); multiple pregnancy; and complications of pregnancy, including maternal hypertension (systolic blood pressure ≥140mmHg or diastolic blood pressure ≥90mmHg during pregnancy), preterm premature rupture of membranes (PPROM) at least 12 hours before the beginning of labour, and idiopathic preterm labour (PTL; spontaneous onset of labour before rupture of membranes). We combined PPROM and PTL, as we hypothesized that they have the same underlying mechanism, namely intra-amniotic infectious conditions. Caesarean delivery was not included in the analysis owing to its strong correlation with complications of pregnancy. Postnatal data included Apgar scores at 1 and 5 minutes, postnatal corticosteroid administration, respiratory distress syndrome, necrotizing enterocolitis, maternal–fetal infection (maternal-acquired culture-proven neonatal sepsis), bronchopulmonary dysplasia at 36 weeks (need for oxygen and/or breathing assistance at corrected gestational age of 36wks), and anaemia (Hb<13g/dL following perinatal haemorrhage). We lacked reliable information on neonatal sepsis.
Five-year assessment
At 5 years’ follow-up, information on CP was available for 1812 children (i.e. 77% of the study population). Medical information was collected from standardized questionnaires completed by physicians (who had undergone a 1-day training workshop and were unblinded to very preterm or term birth) in centres specifically set up for the study (1635 children) or, when responses to these questionnaires were not available, from simplified questionnaires completed by regular treating physicians (82 children) and parents or health services personnel (95 children). Experts reviewed questionnaires for children with abnormal neurological examination in order to validate the diagnosis. We used the definition of CP proposed by the Surveillance of Cerebral Palsy in Europe (SCPE) collaborative group.14 Children were classified as having CP if they had involuntary movements (dyskinetic CP), loss of coordination (ataxic CP), or at least two of the following criteria: abnormal posture or movement, increased tone, or hyperreflexia (spastic CP). Spastic CP was classified as bilateral or unilateral depending on its localization. The functional severity of CP was classified according to the child’s walking ability as follows: able to walk without an aid (i.e. with no help at all or with only a technical aid regarded as minor by the physician, such as orthopaedic soles or shoes, or splints at night only), able to walk with aid (with the help of another person or a technical aid), or unable to walk.
Statistical analysis
We compared data on cerebral lesions, gestational age at birth, birthweight, infant sex, multiple pregnancy, and parents’ socioeconomic status in infants in whom information on CP at 5 years of age was available (responders) and those in whom this information was not available (non-responders). Differences between responders and non-responders were evaluated using χ2 tests for categorical variables and Student’s t-test for continuous variables.
We determined the prevalence of CP at 5 years of age as a function of neonatal cranial ultrasound abnormalities and other neonatal and obstetric risk factors. The results are presented as proportions of survivors assessed at 5 years of age.
Associations of obstetric and neonatal risk factors with CP were first analysed using univariable logistic regression. Multivariable analyses were then conducted. The strategy of analysis for the multivariable models was based on the main aim of our study, which was to estimate the independent effects that may be associated with cerebral lesions on ultrasound scans and several other neonatal and obstetric risk factors as potential predictors of the development of CP. Our analysis, however, was not designed to provide estimates of causal effects that may be associated with the risk factors included in our study. Thus, logistic regression models identified significant predictors of the risk of CP after taking into account cerebral lesions as follows: model A included obstetric factors together with cerebral lesions; model B included obstetric and neonatal factors (except Apgar scores, because of the large numbers of incomplete observations); model C was the same as model B but with data restricted to children without cystic PVL or IPH, as infants with these cerebral lesions had, in general, a very high risk of CP. Two-way interaction effects tests were conducted between cerebral lesions (divided into cystic PVL or IPH, the other cerebral lesions, and no cerebral lesion) and the other risk factors in model B. Results were expressed as crude and adjusted ORs and their 95% CIs. Statistical analyses were performed using STATA/SE, version 10 (StataCorp, College Station, TX, USA).
Results
The frequency of cerebral lesions, especially grade I IVH, was greater in responders than in non-responders. In addition, gestational age and birthweight were significantly lower, and socioeconomic background significantly higher, in responders than in non-responders (Table I).
| Characteristic | Number of responders (%), n = 1812 | Number of non-responders (%), n = 545 |
|---|---|---|
| Cerebral lesion | 1788 | 517 a |
| Cystic PVL | 4 | 4 |
| IPH | 0.3 | 0.4 |
| Persistent echodensities/ ventricular dilatation | 13 | 14 |
| Grade III IVH | 2 | 0.4 |
| Grade II IVH | 7 | 8 |
| Grade I IVH | 10 | 6 |
| None | 64 | 67 |
| Gestational age at birth (wks) | 1812 | 545 a |
| 24–28 | 25 | 18 |
| 29–30 | 26 | 26 |
| 31–32 | 49 | 56 |
| Birthweight (g) | 1802 | 541 a |
| Mean (SD) | 1367 (393) | 1422 (388) |
| Male/females, n (% male) | 935/877 (52) | 303/242 (56) |
| Multiple pregnancy | 1812 (31) | 545 (28) |
| Parents’ socioeconomic status | 1616 | 424 a |
| High | 19 | 10 |
| Intermediate | 44 | 25 |
| Low | 37 | 64 |
- Numbers in italic are denominators for each characteristic. ap value comparing non-responders with responders p≤0.05. IPH, intraparenchymal haemorrhage; IVH, intraventricular haemorrhage; PVL, periventricular leucomalacia.
A total of 159 infants were diagnosed with CP at 5 years of age (prevalence 9%, 95% CI 7–10%). Among this group, 67% walked without aid, 14% walked with aid, and 19% were unable to walk. Spastic, ataxic, and dyskinetic CP accounted for 89%, 7%, and 4% respectively, of all cases. Among children with spastic CP, bilateral and unilateral localization of CP accounted for 82% and 18% of cases respectively. The prevalence of CP at age 5 years was very high in infants with cystic PVL (61%), whether isolated or combined with IVH, unilateral or bilateral, parieto-occipital or other localization, as well as in those with IPH (50%, Table II). The prevalence of CP was higher in infants with persistent echodensities or ventricular dilatation (14%) and in infants with isolated grade I or II IVH (8–11%) than in infants without a detectable cerebral lesion (4%), although infants without lesions accounted for almost one-third of all children with CP.
| n (%) | Crude odds ratio (95% CI) | |
|---|---|---|
| Cerebral lesions | ||
| Cystic PVL | 66 (61) | 33.41 (19.25–57.96)a |
| Isolated | 33 (54) | |
| Combined with IVH | 33 (67) | |
| Unilateral | 24 (42) | |
| Bilateral | 35 (80) | |
| Parieto-occipital | 52 (69) | |
| Other localization | 11 (36) | |
| IPH | 6 (50) | |
| Persistent echodensities or ventricular dilatation | 241 (14) | |
| Grade III IVH | 32 (19) | 3.75 (2.41–5.85)b |
| Grade II IVH | 117 (11) | 2.82 (1.48–5.36) |
| Grade I IVH | 173 (8) | 1.98 (1.07–3.67) |
| None | 1153 (4) | 1.00 |
| Obstetric factors | ||
| Gestational age (wks) | ||
| 31–32 | 889 (6) | 1.00 |
| 29–30 | 477 (9) | 1.61 (1.33–1.95) |
| 24–28 | 446 (14) | 2.59 (1.77–3.80) |
| Infant sex | ||
| Female | 877 (7) | 1.00 |
| Male | 935 (10) | 1.52 (1.09–2.12) |
| Small for gestational age | ||
| No | 1639 (9) | 1.00 |
| Yes | 163 (5) | 0.58 (0.29–1.16) |
| Multiple pregnancy | ||
| No | 1248 (9) | 1.00 |
| Yes | 564 (7) | 0.78 (0.54–1.12) |
| Preterm premature rupture of membranes or preterm labour | ||
| No | 600 (5) | 1.00 |
| Yes | 1181 (10) | 2.17 (1.45–3.26) |
| Maternal hypertension | ||
| No | 1373 (10) | 1.00 |
| Yes | 439 (5) | 0.53 (0.34–0.83) |
| Neonatal factors | ||
| Apgar score at 1min <7 | ||
| No | 1016 (8) | 1.00 |
| Yes | 693 (10) | 1.35 (0.96–1.89) |
| Apgar score at 5min <7 | ||
| No | 1448 (8) | 1.00 |
| Yes | 184 (10) | 1.31 (0.78–2.18) |
| Respiratory distress syndrome | ||
| No | 1018 (6) | 1.00 |
| Yes | 780 (12) | 1.98 (1.42–2.76) |
| Necrotizing enterocolitis | ||
| No | 1731 (9) | 1.00 |
| Yes | 61 (16) | 2.10 (1.04–4.22) |
| Maternal–fetal infection | ||
| No | 1655 (8) | 1.00 |
| Yes | 123 (16) | 2.13 (1.28–3.55) |
| Bronchopulmonary dysplasia at 36wks | ||
| No | 1545 (8) | 1.00 |
| Yes | 218 (12) | 1.60 (1.03–2.50) |
| Acute anaemia | ||
| No | 1673 (8) | 1.00 |
| Yes | 113 (15) | 1.95 (1.13–3.37) |
| Postnatal corticosteroid use | ||
| No | 1474 (7) | 1.00 |
| Yes | 323 (17) | 2.76 (1.94–3.93) |
- aCrude odds ratio (and its 95% Confidence interval [CI]) for the association between cerebral palsy (CP) and cystic periventricular leukomalacia (PVL) or intraparenchymal haemorrhage (IPH; 72 infants; prevalence of CP 60%). bCrude odds ratio (and its 95% CI) for the association between CP and persistent echodensities or ventricular dilatation or grade III intraventricular haemorrhage (IVH; 273 infants; prevalence of CP 14%).
Significant crude associations were found between CP and cerebral lesions, obstetric factors (gestational age, infant sex, PPROM or PTL, and maternal hypertension) and postnatal factors (respiratory distress syndrome, necrotizing enterocolitis, maternal–fetal infection, bronchopulmonary dysplasia at 36wks, anaemia, and postnatal corticosteroid use; Table II).
Logistic regression model A studied the contribution of obstetric factors to CP risk after taking into account cerebral lesions. It showed that, in addition to cerebral lesions, infant sex was a significant predictor of CP, whereas gestational age and PPROM or PTL were at the limit of significance (p=0.07 and p=0.05 respectively; Table III). Logistic regression model B added postnatal factors to model A and showed that none of the postnatal factors significantly influenced the prognosis of very preterm infants after controlling for cerebral lesions and obstetric factors; cerebral lesions and infant sex were the only significant predictors of CP in these infants, and PPROM or PTL was at the limit of significance (p=0.07; Table III). In infants without cystic PVL or IPH, the same predictor factors were found (model C; Table III). We did not find any statistically significant interaction effects between cerebral lesions and other risk factors studied.
| Risk factor | Adjusted odds ratio (95% CI) | ||
|---|---|---|---|
| Model A | Model B | Model C | |
| Cerebral lesion | a | a | a |
| None | 1.00 | 1.00 | 1.00 |
| Grade I IVH | 1.78 (0.94–3.40) | 1.76 (0.90–3.45) | 1.75 (0.89–3.44) |
| Grade II IVH | 2.53 (1.30–4.93) | 2.56 (1.27–5.18) | 2.52 (1.24–5.12) |
| Grade III IVH or echodensities or ventricular dilatation | 3.25 (2.02–5.22) | 3.40 (2.07–5.60) | 3.31 (2.00–5.48) |
| Cystic PVL or IPH | 29.66 (16.71–52.62) | 28.41 (15.65–51.59) | – |
| Gestational age | 1.08 (0.99–1.18)b | 1.00 (0.89–1.12) | 1.03 (0.92–1.17) |
| Infant sex | a | a | a |
| Female | 1.00 | 1.00 | 1.00 |
| Male | 1.48 (1.02–2.15) | 1.52 (1.03–2.25) | 1.82 (1.19–2.80) |
| Small for gestational age | 0.89 (0.39–2.00) | 0.81 (0.34–1.92) | 0.92 (0.36–2.34) |
| Multiple pregnancy | b | ||
| No | 1.00 | 1.00 | 1.00 |
| Yes | 0.71 (0.47–1.07) | 0.67 (0.43–1.03) | 0.71 (0.45–1.13) |
| Preterm premature rupture of membranes or preterm labour | b | b | b |
| No | 1.00 | 1.00 | 1.00 |
| Yes | 1.78 (0.99–3.18) | 1.72 (0.95–3.14) | 1.91 (0.98–3.73) |
| Maternal hypertension | |||
| No | 1.00 | 1.00 | 1.00 |
| Yes | 0.79 (0.42–1.49) | 0.78 (0.41–1.49) | 0.88 (0.44–1.77) |
| Respiratory distress syndrome | |||
| No | 1.00 | 1.00 | |
| Yes | 1.20 (0.78–1.85) | 1.14 (0.72–1.82) | |
| Necrotizing enterocolitis | |||
| No | 1.00 | 1.00 | |
| Yes | 1.51 (0.64–3.55) | 1.22 (0.45–3.29) | |
| Maternal–fetal infection | |||
| No | 1.00 | 1.00 | |
| Yes | 1.47 (0.79–2.75) | 1.31 (0.66–2.59) | |
| Bronchopulmonary dysplasia at 36wks | |||
| No | 1.00 | 1.00 | |
| Yes | 0.95 (0.53–1.71) | 0.89 (0.47–1.68) | |
| Anaemia | |||
| No | 1.00 | 1.00 | |
| Yes | 0.96 (0.48–1.89) | 1.03 (0.49–2.15) | |
| Postnatal corticosteroid use | |||
| No | 1.00 | 1.00 | |
| Yes | 1.41 (0.82–2.43) | 1.41 (0.78–2.54) | |
- Model A includes obstetric factors and cerebral lesions; model B includes obstetric and neonatal factors; model C is the same as model B in children without cystic periventricular leukomalacia (PVL) or intraparenchymal haemorrhage (IPH). ap<0.05. bp≥0.05 and ≤0.07. CI, confidence interval; IVH, intraventricular haemorrhage.
Discussion
To our knowledge, this is the first study to assess, in a prognostic manner, the associations between CP and a detailed set of obstetric and neonatal risk factors in a large population-based cohort of very preterm infants. We found that cerebral lesions on ultrasound were the most important predictor of CP in very preterm infants. In particular, after adjustment for other factors, the odds of developing CP was increased 30-fold in infants with cystic PVL or IPH. In addition, male sex and the presence of PPROM or PTL were independently associated with a higher risk of developing CP, although the strength of their association was much lower than that of cerebral lesions.
Neurological outcome in the infants in our study was assessed at 5 years of age. This age is considered sufficiently advanced for a reliable diagnosis, and thus is frequently used in CP registers.14,15 At the medical assessment at 5 years, we found that mean gestational age was higher in non-responders than in responders and that non-responders did not have more severe cerebral lesions; thus, the prevalence of CP in our study is unlikely to be substantially underestimated. The prevalence of cystic PVL and IPH is decreasing,5 and less severe cerebral lesions are now taken more into consideration when assessing the future burden of CP. Thus, validating our findings in infants without cystic PVL or IPH is particularly relevant. It would also be interesting to carry out separate assessments of predictors of CP in infants with cerebral lesions of high or intermediate severity, or no cerebral lesions. However, our sample sizes were too small to enable us to carry out multivariable analysis in each stratum.
In our population-based study, information on cerebral lesions was obtained during routine practice. Thus, there was no standardized protocol for cranial scanning follow-up, and staff qualifications varied. Although cranial ultrasound scanning was performed with high-frequency 7.5MHz transducers by qualified neonatologists or radiologists who routinely performed ultrasonography, interobserver reliability in interpreting cranial ultrasound was shown to be poor for low-grade IVHs, resulting in some probable misclassifications.16 In one-third of all children with CP in our study, no cerebral lesions were apparent on the neonatal cranial ultrasound. However, cerebral lesions may have been missed in some of these infants. In a previous study, which involved weekly ultrasound scans until discharge, and again at 40 weeks postmenstrual age, using a 7.5MHz transducer in 429 very preterm infants, only 8% of infants with CP had no cerebral lesions.17 Use of an optimal ultrasound scanning protocol may have improved the accuracy of cerebral lesion diagnosis. Although magnetic resonance imaging is more sensitive than cranial ultrasound scanning, especially for detecting diffuse cranial abnormalities,18–20 it is not yet standard practice in very preterm infants because it is expensive and requires the infants to be sedated and transported within departments.
The results of our large population-based cohort are of particular interest given the limitations of previous studies:4,21–24 small numbers of infants followed up; moderately severe cerebral lesions not always included, despite the low and decreasing incidence of severe cerebral lesions;5 only bivariate analyses performed to evaluate the relationship between several risk factors and outcome; or overall outcome grouped into neurological, developmental, or sensory disabilities, despite the fact that each impairment may have specific risk factors and causal pathways, and involve different clinical management and intervention programmes.
Two studies23,24 have assessed the role of perinatal factors in predicting outcome of very preterm or very-low-birthweight infants using multivariable analyses. In a retrospective multicentre cohort23 of 2103 very-low-birthweight infants assessed at 18 to 22 months corrected age, the addition of head ultrasound findings to models with clinical variables did not improve prediction of neurodevelopmental impairment, even though cerebral lesions were significantly associated with the risk of CP. A more recent multicentre prospective cohort study24 of 4446 extremely preterm infants assessed at a corrected age of 18 to 22 months examined the association between obstetric risk factors and development of CP, although cerebral lesions and neonatal risk factors were not included. The authors found that the outcome of infants with intensive care defined as death or profound or any neurodevelopmental impairment was better predicted by a combination of low gestational age, sex, antenatal corticosteroid administration, multiple birth, and low birthweight. In both of these studies, neurodevelopmental impairment was defined as psychomotor (Psychomotor Developmental Index), mental (Mental Developmental Index of the Bayley Scales of Infant Development), neurological (CP), or sensory deficiencies.
In our study, we assessed the risk of CP at 5 years of age rather than 18 to 22 months, as in previous studies, thus allowing a more adequate evaluation of the risk of CP.14,15 We found that cerebral lesions were the most important predictors of CP in very preterm infants. Male sex and PPROM or PTL were also independently associated with a higher risk of developing CP. However, other obstetric and neonatal factors were not associated with the risk of developing CP after taking into account the presence of cerebral lesions on ultrasound scans, infant sex, and PPROM or PTL. In contrast to the two previous studies, we assessed the role of PPROM or PTL, as they have previously been found to be risk factors for CP. Gestational age was not a significant predictor of CP in our multivariable models. A plausible explanation for this is that we controlled for several factors, particularly cerebral lesions, which are likely to be involved in the causal pathway between gestational age and CP.
Conclusions
In conclusion, our findings emphasize the importance of cerebral lesions on ultrasound scans for predicting CP in very preterm infants. After taking into account cerebral lesions on ultrasound scans, as well as infant sex and PPROM or PTL, other obstetric and neonatal risk factors, in particular gestational age, were not significantly associated with the risk of CP. Our study also shows the limits of currently known risk factors for predicting CP for very preterm infants, as about one-third of infants who developed CP had no detectable cerebral lesion. Moreover, risk factors other than cerebral lesions were not sufficiently strongly associated with the risk of CP to be useful for prediction of CP at an individual level.25 Therefore, all very preterm infants need to be closely monitored and their motor prognosis re-evaluated during early childhood. Finally, very preterm infants are also at high risk of adverse cognitive outcomes. Hence, it is important to assess also the precise role of obstetric and neonatal risk factors in relation to long-term cognitive outcomes in very preterm infants.
What this paper adds
- •
Cerebral lesions on ultrasound scan are important predictors of cerebral palsy in very preterm infants.
- •
However, one-third of infants who develop cerebral palsy have no detectable cerebral lesion.
- •
Male sex and preterm premature rupture of membranes/preterm labour are independent predictors of cerebral palsy in very preterm infants.
Acknowledgements
This study was supported by grants from INSERM (French National Institute of Health and Medical Research), Merck-Sharp, Dohme-Chibret, the Medical Research Foundation, the French Ministry of Public Health, the General Directorate for Health of the French Ministry for Social Affairs, and the ‘Hospital Program for Clinical Research 2001 No. AOM01117’ of the French Department of Health. Dr Ghada Beaino was financially supported by La Fondation Motrice and the Ile-de-France Region. None of the funding bodies participated in any of the following: study design; collection, analysis, and interpretation of data; writing of the report; or decision to submit the manuscript for publication.
We thank all the children in the study and their families for helping to bring about research progress.
Appendix
APPENDIX I
The EPIPAGE Study Group
INSERM U953: B Larroque (national coordinator), P-Y Ancel, B Blondel, G Bréart, M Dehan, M Garel, M Kaminski, F Maillard, C du Mazaubrun, P Missy, F Sehili, K Supernant.
Alsace: M Durant, J Matis, J Messer, A Treisser (H⊚pital de Hautepierre, Strasbourg).
Franche-Comté: A Burguet, L Abraham-Lerat, A Menget, P Roth, J-P Schaal, G Thiriez (CHU St Jacques, Besançon).
Haute-Normandie: C Lév≖que, S Marret, L Marpeau (H⊚pital Charles Nicolle, Rouen).
Languedoc-Roussillon: P Boulot, J-C Picaud (H⊚pital Arnaud de Villeneuve, Montpellier), A-M Donadio, B Ledésert (ORS Montpellier).
Lorraine: M André, J-L Boutroy, J Fresson, JM Hascoët (Maternité Régionale, Nancy).
Midi-Pyrénées: C Arnaud, S Bourdet-Loubère, H Grandjean (INSERM U558, Toulouse), M Rolland (H⊚pital des enfants, Toulouse).
Nord-Pas-de-Calais: C Leignel, P Lequien, V Pierrat, F Puech, D Subtil, P Truffert (H⊚pital Jeanne de Flandre, Lille).
Pays de la Loire: G Boog, V Rouger-Bureau, J-C Rozé (H⊚pital Mère-Enfants, Nantes).
Paris-Petite-Couronne: P-Y Ancel, G Bréart, M Kaminski, C du Mazaubrun (INSERM U149, Paris), M Dehan, V Zupan (H⊚pital Antoine Béclère, Clamart), M Vodovar, M Voyer (Institut de Puériculture, Paris).
