- Top of page
- What this paper adds
Aim The aim of the study was to assess weight changes after traumatic brain injury (TBI) in children and the factors influencing them.
Method We conducted a longitudinal observational study of children with TBI of mixed severity who were consecutively admitted to one rehabilitation department (39 children; 23 males, 16 females; median age 8y 7mo; 25th to 75th centiles 3y 7mo–11y 6mo). Weight and height before TBI were obtained from the children’s records and were measured monthly for 1 year after TBI. Body mass index (BMI) and BMI z-scores were calculated, and pre-TBI values were compared with the final values using paired tests. Linear mixed-effect interaction models were used to assess the effect of various factors on z-score evolution.
Results Z-score curves revealed early weight loss followed by a rapid increase in weight. The mean BMI gain over the period under study was 0.9kg/m² (p < 0.001) and the mean z-score gain was 0.4 (p = 0.006). Six children had become overweight by the time of final assessment. Factors associated with a greater rate of increase in the post-TBI z-score were mobility restriction, male sex, and older age. Global pre- to post-TBI weight gain was significantly higher in males (z-score 0.7). Pituitary hormonal testing was available for 17 children at 3 months and for 27 at 1 year. Growth hormone deficiency was detected in one child.
Interpretation Weight gain of children during the first year after TBI was rapid and excessive. Male sex was a risk factor for excessive weight gain.
Childhood traumatic brain injury (TBI) is a major public health concern as it is the leading cause of death and acquired disability in children.1 Survivors often suffer from impairments that alter their social functioning, academic achievements, and ultimately, their participation in society and quality of life.2
Hormonal deficits have become a matter of concern, as several studies found hypophyseal dysfunctions after childhood TBI.3,4 These deficiencies could interfere with progress during rehabilitation, and later with behaviour, fatigue, and well-being.4
Few authors have recorded anthropometric measures after childhood brain injury. In a cross-sectional study, Patradoon-Ho et al.5 found a higher prevalence of overweight and obesity among children who were chronically brain injured (19% and 15% respectively) than in the general population (15% and 5% respectively). This study was retrospective, and did not report early weight evolution post TBI or compare anthropometric measures before and after TBI.
In the clinical experience of our paediatric rehabilitation department, excessive weight gain is often observed after TBI, despite the team’s efforts to control it. Our hypothesis was that TBI is a risk factor for becoming overweight or obese, and that weight gain begins soon after the injury. The objectives of this study were (1) to describe changes in weight and body mass index (BMI) in children hospitalized in a rehabilitation setting for TBI during the first year after injury; (2) to compare pre-TBI values with values 1 year after injury; and (3) to identify factors influencing post-TBI weight gain.
- Top of page
- What this paper adds
This observational study took place in a tertiary care neurorehabilitation department (Hôpitaux de Saint-Maurice in Saint-Maurice, France) dedicated to children with acquired brain injury aged 0 to 15 years. Children with TBI necessitating intensive care in the Paris area are admitted to a unique tertiary care centre (Hôpital Necker, Paris, France). Children are subsequently referred to Saint-Maurice’s neurorehabilitation department, either for in-hospital rehabilitation or for outpatient follow-up according to clinical presentation.
This study started in January 2007, and included all children who were hospitalized at this time and all consecutive children admitted over a 1-year period (January–December 2007). Inclusion criteria were accidental TBI sustained less than a year before admission in children who were on a diverse diet at the time of injury.
In France, children’s health and anthropometric follow-up data are recorded by general practitioners and paediatricians in a health booklet, which is given to every child at birth. Data from health booklets and family interviews were used to assess demographic data and weight and height before the accident.
Weight and height were recorded monthly during the first year after TBI. Final weight and height were recorded at 12 months after TBI or at the time of the next follow-up visit after this period. Weight (kg) and height (m) were measured by trained nurses during in-patient hospitalization in the initial phase, and then during follow-up visits. The weight of external devices was deducted from total weight. BMI was calculated as body weight/height². Z-scores adjusted for sex and age were calculated for all BMI measures. The assessments of the World Health Organization (WHO)6,7 were used as reference measures. Z-scores were calculated with software available from the WHO.8
Cut-off values for categories of BMI z-scores were defined according to the WHO classification, as recommended recently.9‘Normal’z-scores ranged from a standard deviation (SD) of −2 to +1. For children aged under 5 years, a BMI z-score above 1SD was defined as ‘at risk of becoming overweight’, above 2SD was defined as ‘overweight’, and above 3SD was defined as ‘obese’. Children aged over 5 years were defined as ‘overweight’ at 1SD and as ‘obese’ at 2SD. ‘Thinness’ was defined for the whole sample as a z-score value below −2SD. Individuals were classified according to these weight categories before TBI and at final assessment.
Children’s characteristics included sex and age at time of injury, severity of TBI (‘severe TBI’ defined as an initial Glasgow Coma Scale score of ≤8), and duration of coma. IQs were measured with the Wechsler Intelligence Scale for Children, and presence of inhibition difficulties were assessed using standardized questionnaires and observation of behaviour. Individuals’ ability to walk independently was recorded upon admission to the rehabilitation department. ‘Weight-interfering medication’ was defined as an antipsychotic, antiepileptic, or antidepressant agent administered for 3 months or more.
Hormonal status was available for some individuals as a routine measure at 3 months and at 1 year after injury, as recommended for adults.10 This included morning measures of serum cortisol, free thyroxine and free triiodothyronine, thyroid-stimulating hormone, insulin-like growth factor I, and prolactin. For males and females aged over 10 years, measures of testosterone (males), 17β-oestradiol (females), follicle-stimulating hormone (males and females), and luteinizing hormone (males and females) were added.
The sample description used the mean (SD) for continuous variables that had normal distributions and the median (25th to 75th centiles) otherwise. As age was found to have a bimodal distribution, with a group of children aged under 6 years and another aged older than 7 years, the sample was dichotomized into preschool and school-aged children for further analysis. Correlations between individuals’ parameters were tested with the Student’s t-test or χ.
A plot of BMI z-score global evolution over the study period (including pre-TBI assessments as reference values) was obtained through spline regression in time with adjustment for participant effects.
Differences in BMI and BMI z-scores from before TBI to final assessment were compared with paired Student’s t-tests. The change in proportion of weight categories was assessed using a paired Wilcoxon rank test.
Linear mixed effect models were used to assess rates of change in BMI z-scores after TBI (excluding pre-TBI measures in order to fit a linear model) and factors that influenced z-score evolution. These factors included sex, age at injury, TBI severity, duration of coma, walking ability, and weight-interfering medication. Mixed effect models allow each child to have his or her own regression line and therefore account for differences among children, as well as for correlations among serial measurements within each child. ‘Fixed effects’ coefficients reflect the global effect of the factor on the sample, whereas random effects reflect intersubject variability.11 Time × factor interaction models were used to identify factors associated with the rate of weight gain. p values and 95% confidence intervals of fixed effects coefficients were obtained by non-parametric bootstrapping methods.
The association between individuals’ parameters and either global weight gain (the difference between final and pre-TBI z-score) or initial weight loss (the difference between nadir and pre-TBI z-score) was also tested.
Statistical analyses were performed using R software, version 2.14.0 (R Foundation for Statistical Computing, Vienna, Austria).
This study was approved by the local research ethics committee (Comité de Protection des Personnes Ile de France VI, Groupe Hospitalier Pitié Salpêtrière, Paris, France) and all children and families gave informed consent to participation in the study and to the publication of the results.
- Top of page
- What this paper adds
During the study period, 39 consecutive children with TBI (23 males; 16 females) were recruited to the study. Median age was 8 years 7 months (25th to 75th centiles: 3y 7mo–11y 6mo). The TBI was severe in 25 cases. The median length of stay in hospital was 20 days (14 to 30d) in acute care and 310 days (90 to 464d) in rehabilitation (including in- and outpatient rehabilitation). Upon admission, 13 children were able to walk. Mean IQ at 1 year post TBI was 81 (SD 16), and 19 children (50%) were considered to have inhibition difficulties.
Children received standard oral feeding upon admission to rehabilitation, except for four individuals who, because of dysphagia, required gastric tube feeding for a period of 1 to 6 months after TBI. The mean number of available anthropometric measures during the study period was 8.6 (SD 2.7) per child. Anthropometric data from acute care were available for 15 children, and pre-TBI data were available for 37 out of 39 children. Final values were available for all children.
Coma duration was associated with a low initial Glasgow Coma Scale score reflecting severe injury (p=0.012) and with inability to walk (p=0.029). School-aged children tended more often to have a low initial Glasgow Coma Scale score (p=0.067). Sex yielded no significant association with severity of injury, ambulatory status, or age.
Global z-score evolution (Fig. 1) showed early weight loss followed by rapid and significant weight gain, which was responsible for a higher final z-score than at the pre-TBI measurement.
Figure 1. Mean evolution curve and 95% confidence region of body mass index (BMI) z-score. TBI, traumatic brain injury.
Download figure to PowerPoint
Comparisons of pre-TBI and final values of BMI, BMI z-scores, and proportion of weight categories are summarized in Table I. All paired tests reached significance (p<0.001 for BMI; p=0.006 for BMI z-score; p=0.023 for weight categories). The mean BMI of the sample increased by 0.9kg/m² and the mean BMI z-score by 0.4. Seven children who were in the ‘normal’ category before the TBI became ‘at risk of becoming overweight’ or ‘overweight’, three children who were ‘at risk of becoming overweight’ became ‘overweight’, and one ‘overweight’ child became ‘obese’.
Table I. Evolution of body mass index (BMI), z-score, and weight category proportion
| ||Pre-TBI (n = 37)||1y post-TBI (n = 39)||p valuea|
|BMI (kg/m²): mean (SD)||17.8 (2.8)||18.7 (2.9)||<0.001|
|BMI z-score: mean (SD)||0.36 (1.0)||0.75 (1.1)||0.006|
|Weight category (frequency)|
| ‘Risk of becoming overweight’||4||4|
In mixed-effect models (Table II), the rate of increase in post-TBI z-score was higher in males than in females (p<0.001), being 0.07 per month for males and 0.03 for females. Accordingly, the difference between final and pre-TBI z-scores was higher in males (0.7) than in females (0.1; p=0.040). The rate of increase in post-TBI z-score was also increased by an inability to walk (p=0.014) and older age (p=0.011), but the difference between final and pre-TBI z-scores was not significantly higher for children who were unable to walk or for older children.
Table II. Interaction between individuals’ factors and body mass index z-score increase post-injury. Results of mixed-effect linear models (n = 39)
| ||Fixed effect (per month)||95% CIa||p valueb|
| Sex (males vs females)||0.04||0.01–0.07||<0.001c|
| Age at injury (school age vs preschool)||0.03||0.009–0.06||0.011c|
| Initial Glasgow Coma Scale score (severe vs light/moderate)||0.007||−0.02 to 0.04||0.674|
| Length of coma (per additional day)||0.003||−0.003 to 0.005||0.161|
| Walking ability upon admission (yes vs no)||−0.03||−0.06 to 0.003)||0.014c|
| Weight-interfering medication (yes vs no)||0.005||−0.05 to 0.08||0.559|
Initial loss of weight showed a significant association with ability to walk (p=0.013), a trend with age (p=0.075), and no association with sex (p=0.652).
Routine pituitary tests were available at 3 months post-TBI for 17 children and at 1 year post-TBI for 27 children. At 3 months, no dysfunction in the thyrotrophic or gonadotropic axes was found, and morning cortisol was normal for all children. Five children had hyperprolactinaemia, which did not exceed 35ng/ml (norm <19ng/ml). Two children had low insulin-like growth factor I values: for one child, this abnormality disappeared on later control. Dynamic testing of the second child confirmed the diagnosis of growth hormone deficit, and substitution therapy was started. At 1 year post TBI, no de novo hormonal abnormality was found, except for hyperprolactinaemia (still lower than 35ng/ml) in three children.
- Top of page
- What this paper adds
This is the first longitudinal study to address weight gain in children post TBI. Between the pre-TBI level and the final measurement, mean BMI increased by 0.9kg/m², and mean BMI z-score, which should have remained stable, increased by 0.4, this change would amount to a gain of roughly 4 kg in an 18-year-old male of stable medium height. Eleven children went up a weight category, and there were six additional overweight children at the end of the study period.
These findings strengthen results from Patradoon-Ho et al.,5 who found a higher frequency of being overweight and obese in children with a chronic TBI. In the adult literature, there are case reports of post-TBI eating disorders, ranging from reduction of food intake or total anorexia to hyperphagia,12 and several post-TBI weight evolution patterns have been described, with weight gain in 40% of individuals and loss of weight in 30% reported in one study.13 This phenomenon could then be more significant in higher-risk groups, explaining why other authors, studying hormonal status in a smaller sample of children with chronic TBI (20 children out of 127 screened),14 did not find any difference in their BMI z-score compared with a healthy sample.
Our results further reveal that this important weight gain occurred rapidly during the first year after the injury, after a transitory weight loss during the acute phase. This initial weight loss has been studied before, and Krakau et al.15 reported a 68% prevalence of malnourished children 2 months after a severe TBI. Early nutritional support and rapid return to normal weight, in this context of neuroplasticity, could improve outcomes in terms of survival and disability.16 In our sample, four individuals had nadir BMI z-scores below the thinness level, but they returned rapidly to the pre-TBI level and no child suffered medical complications related to malnutrition. Our results suggest that, following the initial concern of malnutrition and its potential effects on TBI recovery, the next concern during the rehabilitation phase post-TBI is later excessive weight gain.
In addition to its cardiovascular, metabolic, respiratory, osteoarticular, and psychosocial consequences,17 being overweight or obese could be an additional burden on children with TBI, who already suffer from motor and cognitive deficits. Dietary interventions in this population could be a challenge, considering the behavioural difficulties and the intolerance of frustration secondary to frontal lobe injuries.18 Antiepileptic drugs have been tested to reduce eating binges in adults with brain injuries,19 but no evidence is available on the efficacy of such interventions. It is, thus, essential to raise awareness about this issue, as early prevention of weight gain through monitoring of caloric intake during initial hospitalization could be more effective than later strategies in addressing excess weight.
In order to explain these findings, the model of hypothalamic obesity in children20 could provide some hypotheses, as in TBI, shearing forces and haemorrhage can cause injury to the hypothalamus or to the pituitary gland.10 In children with craniopharyngioma treated surgically, the prevalence of being obese or overweight is close to 50%.20,21 Weight gain could result from increased caloric intake and hyperphagia, secondary to damage to satiety centres, hypothalamic leptin resistance, or hyperinsulinaemia.22 A reduction in movement and energy consumption could be another physiopathological mechanism.23
Children with post-TBI growth hormone deficiency have been reported to have higher BMI z-scores than non-deficient children with TBI,24 and Klose et al.25 found that BMIs in adults with TBI were significantly higher in those who also had hypopituitarism. In our sample, hormonal deficiencies could not explain all the excessive weight gain, as more than half of the sample was tested for pituitary deficiencies, but growth hormone deficiency was confirmed in only one child.
Another reason for diet imbalance could be related to cognitive and behavioural difficulties, which were frequent in our sample. Lack of self-control18 can alter the quantity of food intake. Henson et al.26 compared the weekly food intake of adults with brain injuries and an age-matched group and found that the individuals with brain injuries ate significantly more at each meal.
Finally, an accident that results in TBI constitutes a sudden disruption to the child and his or her family’s rhythm, especially when it is followed by a long period of hospitalization. Usual landmarks (e.g. meal habits, school and leisure rhythm) are lost, as the child alternates weeks in the hospital and weekends at home. These profound changes could interfere with meal rhythm and diet balance.
The pattern of post-TBI weight evolution differed according to several factors. The rate of weight increase was higher in males, with the mean BMI z-score increasing by 0.7 points in males compared with 0.1 in females. These results are new in the TBI literature; Patradoon-Ho et al.,5 for example, did not observe an association between sex and risk of becoming overweight or obese. This could be explained by behavioural differences, as dietary management could be more difficult to implement in males with TBI. The male population should be a priority target for dietary intervention after TBI.
Mobility restriction and older age were also associated with greater post-TBI weight increase, but were not associated with differences between pre-TBI and final measurements. These variables enhanced early weight loss, explaining the greater subsequent weight gain. Both mobility restriction and older age were associated in our study with severity-related variables (Glasgow Coma Scale score or duration of coma), and Krakau et al.15 reported that some severity-related variables were associated with initial weight loss. Age and mobility restrictions in our study were thus risk factors for initial weight loss, but further studies are needed to determine whether they could also be independent risk factors for excessive later post-TBI weight gain.
This study has several strengths. The pathway of care for children with TBI in the Paris area of France is centralized in a unique primary care unit and a unique rehabilitation unit. Thus, all the children in our sample, although originating from a vast geographical area, benefited from the same care, with similar dietary management.
This study included all consecutive children admitted for in-patient rehabilitation for a 1-year period, with few exclusion criteria, in order to limit recruitment bias and to maximize generalizability of results. Recruitment included children of mixed age and TBI severity, allowing the description of weight evolution in a large range of clinical situations.
This study has several limitations. We encountered difficulties in obtaining monthly weight and height data in all cases. In hospital, anthropometric follow-up in the infant population is notoriously difficult,27,28 and in a study on children with TBI,27 regular data on weight and height evolution were available in only 17% of children. Here, data from the acute care centre were not always available, and the reliability of the data was not checked. However, as median acute care length of stay was short (20d), most data came from the rehabilitation unit, and the use of mixed-effect models lowered the risk of bias due to missing data.
The reliability of pre-TBI data could not be ensured retrospectively. But children’s health booklets, from which pre-TBI data was collected, are usually filled in by trained health practitioners with reasonably reproducible methodology. Exact dates for measurement of height and weight before TBI measures could not be assessed during family interviews, but it was assumed that z-scores, in the absence of a pre-TBI condition, had been stable before the injury. Final measures did not always take place at exactly 12 months after TBI owing to the organization of outpatient follow-up, but the objective of the study was to compare globally the final BMI with the pre-injury status.
Despite these limitations, this first attempt to study weight gain during the first year post-TBI offers several hypotheses on which further studies could be directed to improve research in this rarely studied field.