METHODS

Participants

All infants born before 32 completed weeks in 1991–92 in the eight hospitals within the Liverpool postal districts were ascertained. Those who died before discharge from hospital, or whose mothers were not resident within a Liverpool postal district at the time of birth were excluded.

Initial contact was made with the family doctor via the child’s hospital paediatrician to ascertain current health status and school placement. The parents of those children who were alive and attending mainstream schools were then approached to seek consent to take part in the study. The individual children’s schools were then contacted to arrange assessment visits, and to request that the class teacher choose the child of same sex and first language in the class whose birthday was closest to that of the index child. The parents of that control child were then approached with information about the study, and consent for their child to participate sought. Most children were tested at their schools, although a few were tested at the Institute of Child Health, Royal Liverpool Children’s Hospital, at their parents’ request. The study protocol was approved by the local research ethics committee.

Growth measurements

Height, weight, and head circumference (occipitofrontal circumference, OFC) were measured by a single observer (LF-H) using standardised procedures recommended by the Child Growth Foundation UK.¹² A portable Leicester height stick was used to measure the child’s height in centimetres (to nearest millimetre). The children were measured in bare feet standing with their back to the height stick, feet together, and heels, buttocks, and shoulder blades touching the vertical measure. Upward pressure was applied to the mastoid processes. Children were measured at the same time of day. Weight was assessed using solar scales provided by the Child Growth Foundation and recorded in kilograms (to nearest 100 grams). They were weighed wearing their school uniforms but without shoes or socks. OFC was measured in centimetres (to the nearest millimetre) using a lasso tape measure. Measurements were taken from midway between the eyebrows and the hair line at the front of the head and from around the occipital prominence at the back of the head. All measurements were recorded on “4 in 1” growth charts for boys and girls and plotted to obtain age appropriate centiles. Body mass index (BMI) was computed and centiles also derived. Perinatal data for the children born preterm were obtained from hospital records by a clerical research assisstant.

Test instruments

Fine and gross motor skills were assessed using age band 2 of the Movement Assessment Battery for Children (Movt-ABC).¹³ The test comprises eight items, divided into four subsections: manual dexterity, ball skills, and static and dynamic balance. The scoring system for each item is from 0 to 5, ranging from no impairment to severe impairment. The scores for each item are added and converted to percentiles. A score between the 5th and 15th centile for age is considered “borderline” impairment; at or below the 5th centile is “definitely impaired”, and is sometimes used as a diagnostic criterion for developmental coordination disorder (DCD).

Integration of visual and motor abilities was assessed using the Developmental Test of Visual-Motor Integration (VMI).¹⁴ It consists of 27 geometric forms which increase in complexity and are in a developmental sequence. There is a one point scoring system for each shape. Scores are then standardised for age. Standard scores have a mean of 100 and a standard deviation of 15. The Movt-ABC and VMI were measured by two research assistants (an occupational therapist and a graduate psychologist) trained in their use.

General intelligence was measured using the Wechsler Intelligence Scale for Children UK (WISC III UK),¹⁵ by three research assistants with prior experience in its use (two graduate psychologists and a teacher). Total (TIQ), verbal (VIQ), and performance (PIQ) scores were calculated. Research assistants could not practically be blinded to the birth status of the children being examined.

Statistical analysis of data was carried out using SPSS-10, using parametric or non-parametric tests depending on the distribution of the data.

RESULTS

Of a potential cohort of 382 preterm children identified, 33 had moved out of the area or could not be traced, 18 had died, 29 were attending special schools, 16 parents refused permission for their children to be tested, and six were not tested as suitable appointments could not be made. In addition to these 280 children, 210 term controls were also recruited. In 70 cases a control was not obtained, or the parents of the selected control would not agree to testing. Time constraints at school meant that a few children did not complete all the tests. The index children comprised 151 (53.9%) males and 129 (46.1%) females. The term controls comprised 112 (53.3%) males and 98 (46.7%) females. There were 215 singleton births, 56 twins, and nine triplets among the preterm group. The mean age at testing was 89.8 (range 82–101) months for the index children and 89.9 (range 72–107) months for the controls.

An idea of the illness severity of the preterm group is given by the frequency of perinatal variables listed in table 1. The mean gestational age for the index group was 29.8 (range 23–32) weeks and the mean birth weight 1467 (range 512–2860) grams. A total of 21.4% were of ⩽28 weeks gestation, and 3.6% ⩽24 weeks; 50.4% were <1500 g birth weight and 14.6% ⩽1000 g. The age specific birth weights followed a normal distribution about the 50th centile; 8.9% preterm infants had a birth weight below the 9th centile for their gestational age.

The preterm index group were significantly lighter, shorter, and had smaller OFC and lower BMI than the term controls (table 2). Median centiles for weight, height, OFC, and BMI for the index children were 25th, 25th, 9th, and 50th for boys, and 50th, 25th, 9th, and 50th for girls, respectively. The median centiles for the term controls of both sexes were 50th, 50th, 50th, and 75th, respectively.

The term control children as a group scored significantly better than the preterm index children on all measures of motor function and intelligence (table 3). Eighty six (30.7%) of the index and 14 (6.7%) of the control group were on or below the 5th centile for the Movt-ABC and could be categorised as having DCD. No significant differences in the frequency of DCD were seen between the sexes in index or control children.

When measures of growth were correlated with test results in the preterm children for Movt-ABC, VMI, and IQ, statistically significant relations were seen (table 4). For boys, height and head circumference were significantly related to Movt-ABC and IQ respectively, while in girls only head circumference was significantly related to all outcomes. In the control children these relations were less evident, with weight and BMI correlating with TIQ and PIQ, and height with PIQ for boys, and height and weight with the Movt-ABC in girls.

OFC data measured within seven days of birth were found in the hospital records for only 219 (78%) of the 280 preterm infants. Mean OFC was 28.7 (2.2) for males and 28.1 (2.4) for females, with a median gestation and sex specific centile of 50 (IQR 25 to 75) and a median gestation and sex specific standard deviation score (birth OFC-SDS) of 0.25 (IQR −0.72 to 1.0). (Standard deviation score = (actual measurement − expected mean for age and sex)/standard deviation around mean.) An SDS was also calculated for OFC at the time of testing standardised for sex and age (OFC-SDS), median −0.58 (IQR −1.51 to 0.19). The difference between individual birth OFC-SDS and OFC-SDS at time of testing was calculated (dOFC-SDS) as an index of postnatal head growth, median −0.67 (IQR −1.80 to 0.25). Significant correlations were seen between OFC-SDS at time of testing and all outcome measures, between dOFC-SDS and measures of IQ, but there was no significant correlation between birth OFC-SDS and any of the outcomes (table 5).

When birth OFC-SDS was plotted against d-OFC-SDS (fig 1), it is clear that it is the infants with normal head circumferences at birth rather than the most growth retarded which grow the poorest between birth and seven years. When birth OFC-SDS is plotted against current OFC-SDS (fig 2), it can be seen that almost all infants who now have an OFC more than 2 SD below the mean did not have this at birth.

As OFC was the current growth variable most clearly related to outcomes, predictors of OFC at age of testing were sought using multiple regression analysis with OFC as the dependent variable (table 6). Although OFC at age of testing was independently predicted by present weight, height, and OFC at birth, it was most obviously associated with change in OFC since birth.

DISCUSSION

It is clear from the results of this study that very preterm children at school age continue to show significant deficits in weight, height, and particularly OFC when compared to term controls, and that these deficits correlate significantly with motor and cognitive performance. In utero growth restriction has been recognised for a long time as a cause of later reduced growth and cognitive performance, but only if later catch-up growth does not occur. The preterm subjects of this study had a normal birth weight distribution and normal distribution of OFC (where available) and so intrauterine growth retardation is unlikely to explain the present findings. On the other hand, changes in the relative size of OFC indicating postnatal growth failure were associated with poorer outcomes.

The human brain growth velocity is at its peak at term.¹⁶ These infants were born 8–16 weeks preterm, at a time when many are experiencing a variety of adverse influences on growth. Many insults at this time could potentially reduce brain growth at a critical period, leading to a permanent reduction in final brain size, proportionately greater than that to weight or height. Direct cerebral insults such as haemorrhage or periventricular leucomalacia may lead to reduction in both white and grey matter growth, and subsequent brain volume as shown on magnetic resonance imaging scans.¹⁷ Drugs such as steroids given for chronic lung disease have also been shown to contribute to poor subsequent growth.¹⁸ Nevertheless, the majority of infants in this study did not have major perinatal disease. Poor nutrition may also result from present day feeding practices and schedules.¹⁹ An average loss of more than 1 SD in weight over five weeks in otherwise well preterm infants was recently shown to be substantially caused by a cumulative calorie/protein deficiency at this time.²⁰

This study has its limitations. Although attempts were made to obtain as complete as possible a geographically determined cohort of preterm infants for testing, adverse press reports concerning the local children’s hospital reduced cooperation by parents and teachers during the later part of the study period. Fewer than one control per index case was obtained. The tests used however were well standardised on normal populations, and the control children served largely to confirm the accuracy of the examiners’ technique and to control for social and teaching effects on outcomes. Blinding of the observers to the birth status of the children was practically impossible. Many of the preterm children looked smaller and facially different to the controls, and the research assistants performing the tests also made the arrangements for the teachers to select the controls. The preterm children did not have any correction made to their ages for prematurity, which averaged 10 weeks. Such corrections are controversial, especially at 7 years of age. The differences seen between the preterm and term children were far larger than could have been achieved in 10 weeks. The perinatal data were collected retrospectively by a single researcher, but from clinical records of varying quality from eight different centres within the region. Inevitably they are incomplete, and have been recorded by very many different people. The information they contain probably makes their inclusion worthwhile.

The motor, visuoperceptual, and cognitive tests applied all rely on intact cortical development. The poorer performances by the preterm children may reflect their postnatally restricted cerebral, mainly cortical growth.

Recent improvements in perinatal care have resulted in reductions in major neurodevelopmental disabilities such as cerebral palsy,²¹ but the rates of minor impairments seen at school age in this cohort appear very similar to studies of similar children born a decade or two earlier. If postnatal growth failure is related to nutritional factors, it and associated poor developmental performance may be open to improvement by dietary means. A prospective study of factors associated with early postnatal brain growth in this population is indicated, followed by experimental studies of early nutritional management.