The transition between the phenotypes of Prader-Willi syndrome during infancy and early childhood

Prader–Willi syndrome (PWS) is a genetic disorder characterized by obesity, hyperphagia, short stature, cryptorchidism, and intellectual disability,¹ with hypotonia and failure to thrive in the neonatal period. PWS results from the loss of paternal contribution of genes that are maternally imprinted or paternally expressed in the region 15q11–13. This absence of expression may occur as a result of (1) a deletion at q11–13 on the chromosome 15 of paternal origin, (2) inheritance of two copies of the maternal chromosome 15 (uniparental maternal disomy), an imprinting centre defect, or (3) an unbalanced chromosomal translocation. PWS has historically been felt to have two distinct phenotypic stages. In infancy, it is characterized by poor suck and feeding problems, followed in later childhood by hyperphagia that may lead to excessive weight gain occurring from 12 months of age and invariably before 6 years (in the absence of intervention).² This obesity can have serious repercussions for future health and well-being.³

Several studies have investigated the height and weight of children with PWS, although few have related these directly to development of overeating. Ehara et al.⁴ reported poor weight gain in the first 6 months and a slow rate of both height and weight increase up to 18 months in eight of 11 cases of presumed PWS in children younger than 13 years. The authors attributed the subsequent rapid weight gain to the development of hyperphagia. Growth is nearly normal during the first year in PWS, with low birthweight and normal birth length. However, short stature is noted thereafter in half of all cases, and after 3 years the 50th height centile in PWS is equivalent to the third height centile in the healthy population.⁵ Body weight has been reported as normal during the first 2 years with rapid gain thereafter.⁵ In adults with PWS, hyperphagia may be due to an abnormality of the satiety response to food intake as suggested by functional magnetic resonance imaging studies,^6,7 with hormonal and neuroendocrine defects playing a role.⁸ It is not known when this abnormality in satiety begins. The age at which increased eating behaviour begins may indicate the failure to develop a normal satiety response to food intake at that time. Anecdotal findings suggest that obesity may in fact develop before the onset of overeating behaviour, indicating that there may be other metabolic pathophysiological mechanisms involved.^9,10 Whether such changes in satiety are present before any disproportionate increase in food intake is not known. So far, there has been little published work investigating the shift from the early to the later phenotype, and changes in height and weight in relation to eating behaviour. Investigating this may provide information about the development of the syndrome, and hold clues as to how best it could be managed in these early years. Determining how changes in eating behaviour coincide with changes in body mass index (BMI) may help elucidate the nature and mechanism of the transition between the phenotypic stages.

This study aimed to establish (1) the relation between weight and height in PWS with age during infancy, (2) at what age interest in food appears to increase, (3) the age relation between increasing weight and increasing food interest, (4) how growth hormone therapy, a commonly used treatment in PWS, affects these trajectories, and (5) whether interventions by parents are successful in influencing the consequences of the emergence of abnormal eating behaviour.

Method

Results

Figure 1a,b show retrospective weight and height data relative to age. Height, weight, and BMI at the time of interview are shown in Table II. Measurements taken at the time of interview indicated a positive correlation between weight SDS and age (Table II; r=0.5, p<0.01, 95% confidence interval [CI] 0.22–0.69). Height SDS showed a slight decrease with age, with levels well below the 50th centile, in all except two instances. At the time of interview, 33 of the 46 children had a weight SDS exceeding height SDS (Table II). The mean age of this group was 38 months (SD 18.2), which was older than the 13 whose height SDS was equal to or exceeded weight SDS (mean age 21mo, SD 15.6, t=2.8, p<0.01). BMI SDS increased steadily with age (Table II, r=0.54, p<0.01, 95% CI 0.31–0.71).

Cross-sectional results were supplemented by using retrospective data. These were used to calculate mean SDS at different ages. After an initial dip in weight during the first 3 months (Fig. 1a), mean weight SDS recovered and then continued to rise steadily approaching the norm for the general population (SDS=0) in the 4th year. Meanwhile, the height SDS (Fig. 1b) appears to fall from near normal at birth until it stabilizes from around 2 years at a level (approximately 1.5SDS) below the norm.

Figure S1a (supporting information, published online) shows overall BMI data. Height and weight data were not necessarily recorded at the same time; therefore extrapolation from individuals’ growth charts was required to generate retrospective BMI SDS data. Figure S1a shows mean retrospective BMI SDS. These indicate the steady upward trend that initially reaches the 50th centile (SDS=0) between 15 and 18 months, but then remains relatively stable until a progressive increase in BMI SDS after 30 months, which supports the data from the values obtained at interview. Figure S1a shows that BMI progressively stays above BMI SDS>1 from around 42 months. On average, weight started to exceed height by the end of the first year and BMI crossed the 50th centile (SDS=0) in the latter half of the second year (data not shown). In the 15 individuals in whom weight gain might be considered excessive (BMI SDS>1), BMI started to increase from a mean age of 3 months (SD 4.5) and had crossed the 50th centile by a mean age of 13 months (SD 12.3; data not shown).

There was no statistically significant difference between the two main genotype subgroups: deletion compared with non-deletion. However, the non-deletion group showed greater variability in retrospective mean weight, height, and BMI SDS, and there was a suggestion of a more exaggerated weight gain after 30 months (data not shown).

Eighteen participants with a mean age of 3 years 10 months (SD 15.4mo) were currently on growth hormone treatment, 14 of whom had been prescribed it for more than 3 months, which was deemed an appropriate period owing to the age of the children involved. The mean age at which growth hormone was started was 27 months (SD 13.5, range 6–52). Duration of treatment also varied widely, from 1 to 54 months, (mean 19.4mo, SD 16.4). There was no significant effect of growth hormone on BMI obtained retrospectively (Fig. S1b, growth hormone group; Fig. S1c, non-growth hormone group) or BMI determined from data obtained at interview (data not shown). Mean BMI did not cross the 50th centile (SDS=0) until 36 months in the growth hormone group (who received growth hormone for at least 3mo) but did so shortly after 9 months in the non-growth hormone group (who received growth hormone for <3mo or no growth hormone treatment). This is obviously influenced by a growth hormone effect on height. A comparison of data from those who had been taking growth hormone for more than 3 months with those from other participants in a similar age range from the non-growth hormone group (n=18: 11 with deletion, 7 with non-deletion) indicated that those on growth hormone, despite being a mean of 13 months older, had a lower mean BMI SDS at the time of interview than the non-growth-hormone group (data not shown). This difference is not statistically significant but it does appear to support the reported stabilizing effect of growth hormone on weight gain.¹² Note that at 57 months there are data from only one child; hence there is no SD indicated.

Figure 2 demonstrates changes in eating behaviour. Twelve of the 46 participants had noticeable eating problems, (eating code 4 or 5; Table I), five of whom exhibited hyperphagia. A further 10 had noticeable eating issues (eating code 3). Higher eating codes were found among older children using linear regression (p<0.01; Fig. S1a). A positive relation was identified between eating codes and BMI SDS scores at the time of interview (p=0.01). However, even among the heavier children (BMI SDS>1) there were still children with an eating code less than 3 (Fig. 2b), and only three of the six heaviest children exhibited full hyperphagia (eating code 5). There was no significant effect of genetic subtype, or growth hormone use, on eating code (data not shown).

Figure 2b indicates a relation between BMI SDS and control of the food environment, with high BMI SDS seen in those with less control of the food environment, especially in those with eating code 3 or above. Thirty-five of 46 mothers said that they always controlled the food intake, nine controlled intake ‘to an extent’, and two did not control. Of those who controlled, 44 controlled the type of food eaten, 42 limited snacks, 41 controlled the number of meals, 40 controlled portion size, and seven counted calories. Where in place, controls were imposed from an early age (5mo, SD 10.9), irrespective of actual need.

At interview, 14 of the 26 children with BMI SDS of <0 were reported as having food interests greater than that of their peers (eating code 4.1, SD 0.8). Where appropriate, parents were asked when they first became aware that their child’s interest in food was different from their peers (Fig. 2c,d). Of 13 responses, the mean age at which this was first noted was 32 months (SD 13.9) after the mean BMI SDS first exceeded 0 (mean age 15mo, SD 12.2; Fig 2c). This is after the normalization in BMI stage from around 15 to 30 months and before the BMI SDS consistently stays above 1 from 42 months (Fig. S1a). There were six instances where mothers reported an increased interest in food despite a BMI SDS of less than 0. The mean overall eating code of this group was lower (3.1, SD 0.8). In all cases where parents reported an increased interest in food, this was usually reported as having started later than any consistent rise in BMI, as calculated from the child’s chart.

Discussion

This is the first study to explore systematically the nature and timings of the change in phenotypes in PWS. As expected, weight SDS increased with age, and height SDS decreased with age. Data indicate that the initial increases in BMI SDS start before the eating behaviour changes. There is a stability of BMI that is evident from around 15 to 30 months, after which there was a progressive rise in BMI. An intriguing finding is that the age at which BMI starts to increase beyond normal predates the age at which parents note an increase in eating: just over half of the children whose BMI SDS exceeded 0 at interview were reported to have higher than normal food interest. There were five children who exhibited pronounced hyperphagia, which was far fewer than expected based on previous reports.¹¹

Overall, changes from the early to the later phenotype are gradual and undergo transition through a ‘normalized’ stage. Weight gain appears to start between 4 and 5 months, and by the latter half of the second year BMI appears to reach the norm for the general population (SDS=0). The very earliest increase in BMI within the first year is compounded by the relative decline in height seen during the early months. There then appears to be a period of stability with BMI SDS around 0, until a progressive increase in BMI SDS starts from after 30 months, at SDS less than 1 from 42 months. During the 4th and 5th years, some of the children became heavy, with BMI SDS of more than 2. The initial weight decrease from birth may be attributed to poor feeding due to the sucking difficulties experienced by all participants. The subsequent increase in weight in the latter half of the first year may coincide with improvement in suck, the start of weaning and more varied diet, and the point at which most parents reported a normalization of their child’s attitude to food.

Although quantifiable changes in eating appeared after the increase in BMI, there was anecdotal evidence from parents to say that eating behaviour seemed qualitatively different from a young age. There may be subtle differences in attitudes to food among these children even from birth, which are masked by their physical disabilities and lack of autonomy. Again, as a result of early diagnosis, most parents said they had instituted controls, regardless of whether they were perceived as necessary, from an early age. Without observing the family over time, it is not known how strictly such controls were adhered to. However, evidence indicates that any relaxation of control did have a deleterious effect on BMI once an eating problem had been identified.

Previous research¹² has shown that growth hormone therapy has positive results on children’s growth. Owing to the variation in treatment length and age of prescription, no statistical comparison could be drawn. Nonetheless, retrospective data suggested a stabilizing effect of growth hormone on BMI that seemed more pronounced from 3 years of age. From retrospective data, the mean age at which BMI crossed the 50th centile was apparently being delayed by almost 2 years in the growth hormone group. The nature of growth hormone treatment on changing BMI SDS, and potential differences between genotypes, is an area for further investigation.

There are several limitations with our study. The main problem was reliance on retrospective data. Available hospital data were used to corroborate parental recall, although this was not standard between hospitals and there were often large gaps between appointments. However, anecdotal evidence suggests that parents of children with PWS are highly attuned to their child’s behaviour, especially eating behaviour, which suggests that responses would be accurate. Heterogeneity was also a limiting factor. There was large heterogeneity within the growth hormone subset, with variation in length of treatment and the age at which growth hormone started, both of which affect the efficacy of growth hormone. Separating groups according to growth hormone treatment would have rendered group sizes too small to have any statistical relevance. It is suggested that a larger-scale study may be able to overcome this. Another limitation may be selection bias. Ethical constraints meant that families could not be contacted directly. All participants were members of the UK Prader–Willi Syndrome Association, and it is not clear if this group has different characteristics to non-members with PWS, although this bias is an inevitable part of any study involving voluntary recruitment. Owing to the widespread geographical area from which responses came, there is no reason to think that these results are not generalizable to the wider PWS community.

Although a change in eating behaviour accounts for some of the increase in weight, and subsequent BMI, it may not be the driving force. Those with major eating problems do have increased weight; however, conversely, increased weight was not necessarily related to major eating problems. Our findings are in line with recent suggestions that the natural history of PWS is more complex than the two phenotypic subdivisions and that eating behaviour changes after the onset of obesity.⁹ McCune and Driscoll suggest that there are three distinct nutritional phases in childhood PWS: failure to thrive, obesity, and hyperphagia (which worsens the obesity). During the second half of the first phase, phase 1b,¹⁰ the infant is growing steadily along a normal growth curve rate. In the second phase, body weight starts to increase, between 18 and 36 months of age, without a significant increase in food intake. Our findings support this description of the nutritional development, with a stage of food-related normalization and indication that the obesity begins before a substantial increase in food intake or interest.

This study presents a step towards understanding the transition between the early and later phenotypes in PWS and, importantly, provides justification for a long-term study. Increasing knowledge about this transition will help to develop understanding of the physiology of PWS and may lead to better management of the eating behaviour. A larger study with an increased sample size, tracking children throughout growth development, drug regimes, and parental control would negate the reliance on recall data and would increase the statistical power of our findings. A larger study would also enable valid comparisons to be made between genotypes and the effect of growth hormone.