Are short normal children at a disadvantage? The Wessex growth studyBMJ 1997; 314 doi: http://dx.doi.org/10.1136/bmj.314.7074.97 (Published 11 January 1997) Cite this as: BMJ 1997;314:97
- A Bruce Downie, research assistant (psychology)a,
- Jean Mulligan, data managera,
- Robert J Stratford, senior lecturer in psychologyb,
- Peter R Betts, consultant paediatricianC,
- Linda D Voss, senior research fellowa
- a Department of Child Health, Southampton University Hospitals Trust, Southampton General Hospital, SO16 6YD
- b Department of Psychology, University of Southampton, Southampton, SO17 1BJ
- c Department of Paediatrics, Southampton University Hospitals Trust, SO16 6YD
- Correspondence to: Dr Voss
- Accepted 28 October 1996
Objective: To examine whether short stature through childhood represents a disadvantage at around 12 years.
Design: Longitudinal non-intervention study of the physical and psychological development of children recruited from the community in 1986-7 after entry into primary school at age 5-6 years; this is the second psychometric assessment made in 1994-5 after entry into secondary school at age 11-13 years.
Setting: Southampton and Winchester health districts.
Subjects: 106 short normal children (<3rd centile for height when recruited) and 119 controls of average stature (10th-90th centile).
Main outcome measures: Psychometric measures of cognitive development, self concept development, behaviour, and locus of control.
Results: The short children did not differ significantly from the control children on measures of self esteem (19.4v 20.2), self perception (104.2 v 102.4), parents' perception (46.9 v 47.0), or behaviour (6.8 v 5.3). The short children achieved significantly lower scores on measures of intelligence quotient (IQ) (102.6 v 108.6; P<0.005), reading attainment (44.3 v 47.9; P<0.002), and basic number skills (40.2 v 43.5; P<0.003) and displayed less internalisation of control (16.6 v 14.3; P<0.001) and less satisfaction with their height (P<0.0001). More short than control children, however, came from working class homes (P<0.05). Social class was a better predictor than height of all measures except that of body satisfaction. Attainment scores were predicted by class and IQ together rather than by height. Height accounted for some of the variance in IQ and locus of control scores.
Conclusions: These results provide only limited support for the hypothesis that short children are disadvantaged, at least up until 11-13 years old. Social class seems to have more influence than height on children's psychological development.
Key messages Most studies into the psychology of short stature in childhood have been of children referred to clinics, who do not necessarily come from a range of social classes
In this study, in which referral bias was avoided, short children displayed normal psychosocial adjustment up to age 11-13 years
Social class was shown to be a better predictor than stature of psychometric performance.
Early research generally portrayed short children as psychosocially disadvantaged.123 More recent studies, however, have found their self concept to be normal,4 and several studies indicate that short children's self esteem is equivalent to or higher than norms.567 None the less, short children have been reported to display less social competence and more behavioural difficulties compared with normative samples46 or with controls.8 Lower levels of attainment are still reported despite average intelligence.8910 Most studies, however, have been conducted with referred samples and without controls4 9 and often with mixed diagnostic groupings.5 It is therefore difficult to generalise their findings to all short children.
The Wessex growth study is an ongoing longitudinal study, begun in 1986, comparing the physical growth and psychological development of a cohort of short “normal” children with that of a cohort of children of average stature.11 All the children were recruited from the community at age 5-6 years, and the sample avoids the imbalances in social class and sex found in referred samples. The children have been measured for height and weight every six months, and three major psychometric assessments were planned: after entry into primary school (age 7-9 years), after entry into secondary school (age 11-13 years), and at the end of compulsory education (age 16 or over). The first was made in 1989-91, when no differences were found between short and control children in intelligence quotient (IQ), attainment, self esteem, and behaviour once allowance had been made for social class.12 This article reports the second assessment, made in 1994-5.
Subjects and methods
All children beginning primary school in two adjacent health districts in Wessex during 1985-6 and 1986-7 were screened for short stature.11 13 There were 180 children below the third centile for height by Tanner-Whitehouse standards.14 Forty children, including 32 diagnosed as having an organic cause for short stature, did not enter the study. The 140 remaining short normal children were each matched for sex, age, and school class with a control of average height (10th-90th centile14).
Sample attrition-Of the 140 short normal children originally recruited, 21 have subsequently received treatment as part of a growth hormone trial and therefore were not available for the present assessment. The results to date of the trial have been reported elsewhere.15 The treated children were compared on all measures with the short children remaining in the larger study and no significant differences were found.15 There were 106 short children and 119 controls remaining at the time of the 1994-5 assessment. Of the 13 short children who left the study, six left the area and seven refused to continue. Of the 21 control children who left the study, 19 left the area, one refused, and one developed diabetes. Table 1 shows the characteristics of the remaining sample.
Current height-It should be emphasised that the two groups were recruited on the basis of their height at age 5-6 years, and there is now a degree of overlap between the heights of the short and control children. Their present height centile, however, is probably a transient feature of their entry into puberty. Those short children who develop early will temporarily increase their centile whereas late developing controls will decrease theirs. Therefore, in almost all of our analysis, the children are kept within the height group that characterised the major part of their experience and not redistributed on the basis of a passing phase. There is evidence that short normal children at age 5 years are likely to become short adults.16 It is likely therefore that any overlap in height apparent at age 11-13 years will be considerably lessened at final height.
Current pubertal status-There were no significant differences between our short and control girls in age at menarche or in age at peak height velocity. Nearly all the boys in both groups were prepubertal in 1994-5 and had not yet begun their growth spurt. We therefore had no reason to suppose that more of the short normal children would be delayed in puberty relative to the controls.
Parents supplied information needed to update social class.17 Cognitive development was assessed by measures of intelligence (British ability scales (BAS) short form IQ test18) and attainment (BAS word reading and basic number skills18). Locus of control was measured by Nowicki and Strickland's scale19 and behaviour was assessed by teachers on Rutter's children's behaviour questionnaire.20 Self esteem was measured by the culture free self esteem inventory21 and self perception by Harter's self perception profile for children.22 Parents' perception was measured by using an adapted Harter scale.22 Body image was measured by a body satisfaction index, derived from Dowdney et al.23 Statistical analysis was performed with spss. Means were compared by t tests or one way analyses of variance. Analysis of covariance was made on all measures, and when appropriate, multiple analysis of variance was also performed. Ordinal data were compared with the Mann-Whitney U test. Correlations were measured by Pearson's coefficient.
Tables of means
Table 2 shows means (SD) for the tests comparing short normal and control children. The short children achieved significantly lower scores on IQ and both measures of attainment. They showed significantly less internalisation of control on the measure of locus of control (as indicated by higher scores).
Table 3 shows that the social class distributions of short and control children differed significantly (P=0.042). Table 4 shows the means for all measures by social class. Children in classes D and E did “worst” on all measures-that is, recorded the highest scores on the measures of locus of control and behaviour and the lowest on the other measures. Children from non-manual homes, typically those in class B, did “best.”
Analyses of covariance
As class is associated with some of the measures used,19 24 it was necessary to determine and correct for its effect before we investigated the effects of height group and sex. Accordingly, analyses of covariance were performed on all measures with class as the covariate and height and sex as explanatory variables.
Social class had a significant effect on all measures (P values 0.005 for self perception; 0.001 for behaviour; and <0.001 for the remaining measures). After we controlled for class in this way, height group still had an effect on IQ (P=0.014), reading attainment (P=0.006), basic number skills (P=0.023), and locus of control (P=0.008). Table 5 shows the differences between means for these measures before and after adjustment for class. All the differences between short and control children were only slightly reduced and remained significant.
After we controlled for class, sex still had a significant effect on behaviour. As described elsewhere,6 boys had more reported difficulties than girls (7.6 v 5.0; P=0.018). There were no significant height-sex interactions, indicating that sex differences in the short children were parallelled in the controls.
Clearly, IQ is associated with attainment. An analysis of covariance that corrected for individual IQs revealed that class continued to have a significant effect on reading (P=0.007) and number skills (P=0.044) whereas height did not. Thus, attainment is explained by class and IQ together rather than by height.
Multiple analyses of variance
When a measure contains subscales, an analysis of covariance cannot indicate which of these show significant differences. Multiple analyses of variance were therefore performed on these measures, again with class as a covariate. No significant height-sex interactions were found, but height had a significant effect within the subscales for IQ and self perception and sex within subscales for self perception. Univariate F tests revealed that the short children's lower IQ score was explained by the difference in performance on verbal reasoning (short 51.0 v control 54.7; P=0.042) and speed of information processing (54.3 v 59.0; P<0.001). On the self perception profile, short children scored significantly higher on appearance (16.9 v 15.7; P=0.020). Boys and girls differed on self perceived athletic ability (boys 17.0 v girls 15.0; P=0.001), appearance (16.9 v 15.5; P=0.016), and behaviour (16.6 v 17.6; P=0.013).
Multiple regression analysis
Multiple regression analysis was performed on those measures where both class and height were seen to have a direct effect. Social class was found to explain 14% of the variance in IQ scores while height explained only a further 2%. Analysis of the locus of control scores showed that social class explained 7% of the variance in scores while height explained 3%.
There was a significant difference in satisfaction with height on the measure of body image. Thirteen (12%) short children were happy with their height compared with 56 (47%) controls (P<0.0001); and 92 (87%) short children wanted to be taller compared with 49 (41%) controls (P<0.0001).
The short children in the Wessex growth study have, in two separate assessments made some four years apart, shown normal psychosocial adaptation equivalent, with some exceptions, to that of the controls. While stature did affect some measures-IQ, locus of control, and body satisfaction-its influence fell far short of that of social class. As this study has avoided the (generally middle) class bias of referred samples5 8 9 and because it includes representation from the full range of social class, it has been possible through analysis of variance to determine the independent effects of class and stature.
Cognitive measures and locus of control
Performance in cognitive tests was clearly related to social class (see table 4). Multivariate analysis supported the view that social class is a better predictor than height of measured intelligence.
None the less, height did seem to have some effect on IQ. The short children were significantly less able on two subscales, but on these the controls had scores well above established norms. There is no indication, therefore, that the short children may have had specific cognitive deficits, as suggested elsewhere.25 It is also unlikely that the scores reflected differences in biological maturity as neither gestational age nor pubertal status differed significantly between our groups.
Alternatively, short children may be treated differently by their parents and teachers.26 If they are perceived to be less “mature” then expectations may be lowered and reflected in the results of these tests. The locus of control measure indicates the degree to which children have internalised control or taken responsibility for their own actions, and, as may be expected, this internalisation typically increases with age. Short children in this sample displayed significantly less internalisation of control than the average stature group even after we corrected for social class. This finding may also relate to the way in which short children are perceived and treated by adults.
Behaviour and self esteem
The behaviour ratings of the children by their teachers did not distinguish between the two groups, which conflicts with reports elsewhere of behavioural disturbance in short children.4 68 Our results accord with other work in finding that short children's self esteem seems to be healthy.4567 On two measures of this construct,21 22 the groups displayed similar levels of self esteem equivalent to or higher than norms.21 27 The differences between sexes in self perception and behaviour displayed by this sample are consistent with those reported in other studies6 7 and with norms,20 27 suggesting that the sample is not unrepresentative. These findings were supported by parental perceptions.
The parity between the short and control groups in this study on the measures of behaviour and attainment lends some support to speculations that samples referred to clinics are more likely than community samples to include short children with existing psychosocial difficulties.6
While the body satisfaction index showed that fewer short than control children were happy with their height, the short children recorded more satisfaction with their appearance on the self perception profile.22 It seems that any dissatisfaction elicited when children were questioned specifically about their height did not affect other areas of their emotional wellbeing and that short stature per se is not a major source of distress. Most of the children, irrespective of their short or control status, reported that they would like to be “quite tall” or “average,” indicating a desire for conformity.
Comparison on the basis of current height
We have argued that it is psychologically appropriate to keep the children in their original height groups. In support of this, we found no evidence that the children's current height is a better predictor than original height of psychological wellbeing or that the shortest children are more likely to experience significant problems. The only significant differences between the original short and control groups were in IQ, attainment, locus of control, and body image. No further differences were found when the children were divided on the basis of their height at the time of testing into very short (<2nd centile), quite short (2nd-10th), and average (>10th) groups. The average children still had significantly “better” scores in IQ, attainment, locus of control, and body image. The quite short and very short groups, however, did not differ significantly on these or any other of the measures. These results are consistent with others5 6 that suggest that the very shortest children are not necessarily the most psychologically vulnerable.
These results provide only limited support for the hypothesis that short children are disadvantaged. Stature seems to have no direct effect on the measures of attainment, behaviour, self esteem, parent's perception, or self perception. Stature does, however, seem to influence IQ, locus of control, and body satisfaction. By recruiting from the full range of social class, we have shown that class has more effect than stature on children's psychological development. These findings apply up to age 11-13 years. What will happen in later adolescence and adulthood remains to be seen.
We thank the children and their parents and teachers for their patient contribution to this long term study. We are very grateful to our colleagues Professor Jim Stevenson, department of psychology, and Dr Ruth Pickering, department of medical statistics, University of Southampton, for their invaluable advice on the analysis of these data.
Funding: Wessex Regional Health Authority and Wessex Medical Trust from Pharmacia Upjohn.
Conflict of interest: None.