Subjects and methods

SAMPLE

The 1958 British birth cohort study includes all children born in Scotland, Wales, and England between 3 and 9 March 1958.30From a target population of 17 733 births, information was obtained on 98%. Major follow ups of surviving children were conducted at the ages of 7, 11, 16, 23, and 33 years.31 Immigrants to Britain born during the same week in 1958 were incorporated into the survey at ages 7, 11, and 16. Sample attrition has occurred, but despite this the representativeness of the study has been maintained, in general, through to the age of 33 years, when 11 407 subjects (69% of the target population) provided information. There is some evidence to suggest a slight under representation of ‘disadvantaged’ subjects.31 Information at ages 23 and 33 years was obtained through an interview with the study subjects. Previous sweeps collected data from several sources, including parents and schools (teachers and doctors), as well as from the individual subjects.

MEASURES

The heights and weights of the cohort members (the offspring) were measured by trained medical staff at 7, 11, and 16 years. The subjects were weighed in their underclothes to the nearest pound, and their height was measured to the nearest inch. Self reports of height and weight were obtained at the age of 23 years and were checked to detect coding errors.32 These self reports are in agreement with the measured heights in a national sample of British adults surveyed in 1980.33 At the age of 33 years, heights were measured without shoes using a stadiometer reading to the nearest centimetre, and weights were measured in indoor clothing using Salter portable scales. The dates of measurement were recorded at the ages of 7, 11, and 16 years.

The heights and weights of the parents were self reported when their children were aged 11 years in 1969. Heights were reported to the nearest inch; weights were classified into one of 27 groups ranging from 6 stone 4 pounds (39.9 kg) to 19 stone 10 pounds (125.2 kg) in increments of 6 pounds (2.7 kg). For the purpose of estimating Quetelet’s BMI (defined as weight/height²), the parents were assigned a weight equivalent to the midpoint of their weight group. The BMI was calculated at a single time for the parents (that is, when their children were aged 11 years) and at each of the five ages of follow up for the children. Parental BMI was first classified as underweight, normal, overweight, or obese, as defined by the 85th centile (see table 1 for cut offs). This particular cut off corresponds to the centile used to define obesity in the USA,34although the definition is arbitrary and other centiles have been suggested.28 The 85th centile ensured adequate sample sizes in the obese groups. Six parental obesity groups were then constructed (the abbreviated names of these groups are given in quotes and these are used throughout this paper). These are ordered in increasing parental fatness based on the combination of the mother’s and father’s mean BMI (see table 2 for means): (a) ‘one underweight’—one parent underweight, one not obese (that is, underweight, normal, or overweight); (b) ‘both normal’—both parents of normal BMI; (c) ‘one overweight’—one parent overweight, one of normal BMI; (d) ‘both overweight’—both parents overweight; (e) ‘one obese’—one parent obese, one not obese (that is, underweight, normal, or overweight); and (f) ‘both obese’—both parents obese.

STATISTICAL METHODS

The relation between the BMI of the parents and their offspring was assessed using partial correlation coefficients, adjusted for measurement date in childhood, at ages 7, 11, 16, 23, and 33 years. To investigate the tracking of adiposity from child to adult life, in relation to parental obesity status, partial correlation coefficients were then calculated between the child’s BMI at ages 7 and 33 years (adjusted for measurement date at age 7 years) separately for each of the six parental obesity groups. To reduce the skewness of the BMI distributions, correlations were also calculated using log_e(BMI) and −1/BMI. Trivial differences (−0.06 to 0.03) were found between correlations using transformed and not transformed BMIs, so only the latter are presented. Differences in the correlation coefficients (ages 7 and 33 years) between parental obesity groups were tested using Fisher’s transformation.35

Results

Table 1 shows that approximately 50% of parents had BMIs in the normal range and, as predetermined by the definition of parental obesity, approximately 15% were obese. About 3% of the children had two obese parents, 9% had two who were overweight, and over a quarter had two parents of normal BMI (table 2). In general, the mean BMIs were similar for children with either an underweight, overweight, or obese mother or father, so these are combined and reported as the one underweight, one overweight, or one obese parent groups, respectively.

Table 3 shows that, in general, the mean BMI of the children increases as the degree of parental fatness increases. This is consistent for sons and daughters at each age from 7 to 33 years. The largest difference between adjacent groups in mean BMI was for those with both parents obese compared with those with one obese (range 0.8 to 2.5 kg/m²). The smallest difference was found between children with one underweight parent compared with both with normal BMI (range 0.2 to 0.4 kg/m²). Differences in mean BMI increased with increasing age of the child—for example, 7 year old daughters with two obese parents had a mean BMI 9.7% greater than those with two normal BMI parents and this difference increased to 20.4% at the age of 33 years. To some extent the mean BMIs underplay large differences in the risk of adult obesity at the age of 33 years for those with obese or overweight parents. The odds ratios for men and women with two obese parents (relative to those with normal BMI parents) were 8.4 (95% confidence interval 5.5 to 13.0) and 6.8 (4.6 to 10.0), respectively (table 4).

Table 5 shows further the association between parental BMI and that of their children. This shows weak correlations, ranging from 0.15 to 0.25 at the different ages of the children. From the age of 11 years onwards, the daughters’ BMI was more strongly correlated with the BMI of their mother than with their father. A similar, but weaker, pattern was evident for the sons. Correlations with mothers’ BMI was similar for sons and daughters at all ages, but the father-son correlations in adulthood were significantly stronger than those for father-daughter (p <0.05).

Table 6 shows the extent of tracking between childhood and adulthood for the children according to parental BMI. It presents correlations of offspring BMI between the ages of 7 and 33 years for each of the six parental BMI groups. In general, daughters have slightly stronger BMI correlations than sons. Long term associations, denoted by the 7 to 33 year correlation, tend to be moderate and similar for those with both parents with normal BMI (0.25 for sons, 0.32 for daughters). Child to adult correlations are strongest for both sons and daughters with two obese parents (0.46 for sons, 0.54 for daughters) and differ significantly from the group with two normal BMI parents.

Discussion

Most investigations of the tracking of BMI from childhood to adulthood report only weak to moderate associations.27 28This is apparent from the literature review of Serdula et al.27 In agreement with this, we reported elsewhere for this large nationally representative birth cohort sample that associations between child and adult BMI were only moderate (r = 0.33 for men and r = 0.37 for women).29 Thus it is no surprise that in the present analyses child to adult BMI correlations are low to moderate for the 28% of subjects whose parents are both within the normal BMI category (correlation coefficients between ages 7 and 33 years, r = 0.25, 0.32, sons and daughters, respectively). For the smaller percentage (3%) with both parents obese, however, the associations between child and adult BMI were considerably stronger (r= 0.46, 0.54, sons and daughters, respectively). This suggests that, compared with those with parents in the other parental obesity groups, the prediction of adult BMI from that in childhood is much better for children with two obese parents and that the stronger tracking of BMI over the life course indicates less variable BMIs among these children. Thus we would expect the fattest children to be more likely to stay in the upper BMI centiles throughout life (indicated by the stronger correlation between BMI at 7 and 33 years). Subjects with one obese parent resembled those with two parents with normal BMI showing a weaker child to adult BMI relation, although at each age their mean BMI was intermediate between those with two normal BMI parents and those with two obese parents. It is, of course, well known that parental adiposity is related to that of their offspring,3-13 but the extent to which parental obesity influences the BMI trajectory of their children does not appear to have been documented previously.

The stronger BMI child to adult correlation observed here for the group with obese parents has important implications. When the prevalence of parental obesity is relatively low, as in the present study, the overall trend is dominated by the moderate correlations of the normal BMI groups. Under these circumstances the identification and prevention of childhood fatness may not be seen as an important strategy for the prevention of adult obesity. When the prevalence of obesity among parents increases, however, as is now happening in several developed countries,36-40 then the pattern of tracking from childhood to adulthood will be more influenced by the stronger child-adult relations shown here for those with obese parents. Using a standard definition for parents and their offspring at the age of 33 years, it is estimated that the prevalence of two obese parents in the 1958 cohort has doubled from 1969 to 1991 (3.2 to 6.0%). Thus the child to adult tracking of BMI is likely to strengthen in the next generation—that is, among the offspring of 1958 birth cohort members.

Only one other study was found that provided information on the tracking of the child’s BMI in relation to parental fatness with which the present study can be compared. Charney et al showed that of the fattest 25% of infants (based on their weight status at 6 months), 51% were still overweight at the age of 20 years for those with at least one overweight parent, compared with 20% for those with neither parent overweight.6 If we compared a group of children consisting of those with at least one obese parent with those with no obese parent (that is, both with normal BMI) in the current study, we would still show stronger tracking in the former group, which is consistent with the results of Charney et al.6

There is extensive evidence showing a parent-child relation for BMI at a particular age in childhood.3 4 9 10 Stunkardet al found approximately linear relations with the BMI of adult adoptees and the BMI of biological parents, but not the adoptive parents, which they attribute to a strong genetic influence.9 In the present investigation the mean BMI of the children increased linearly with the number of obese parents: 7 year old boys with parents within the normal BMI range had a mean BMI of 15.7, increasing to 16.2 for those with one obese parent and to 17.0 for those with two obese parents. This trend strengthened slightly with increasing age. Focusing on adult obesity at the age of 33 years, the population attributable risks associated with having one obese parent were 37% and 32% for men and women, respectively, whereas the attributable risks associated with the small percentage (3%) with both parents obese were 20% and 15% (sons and daughters, respectively). For obesity among the offspring therefore, there appears to be a substantial risk associated with parental BMI.

These risks are obscured to some extent when the overall relation between parent and child BMI is examined: in the 1958 birth cohort 2–6% of the variability in the children’s BMI is explained by the parental BMI. Correlations reported here between child and parental BMI were weak (r = 0.15–0.25), as has been shown previously.3 10 12 26 41 Mother-child correlations were slightly stronger than father-child correlations. This greater influence of maternal adiposity is consistent with some studies,10 26 but not all.7 10 12 In the present study the stronger maternal correlations may be due to the poorer quality of the father’s height and weight data. The adequacy of the self reported parental height and weight data is a limitation in this investigation. Self reported data tends to underestimate BMI and hence obesity as height tends to be overestimated and weight underestimated.42 Thus a data validation check was undertaken for the mothers for whom measured and self reported heights were available in 1958 and 1969: the correlation was weaker than expected (r = 0.79) with an associated measurement error of 3 cm. Reported parental weights are likely to be less precise, partly because they were recorded in categories with the exact weight unknown. It is unlikely, however, that the relation described here will be affected by an underestimation of the absolute value of the BMI because the obese group was defined using a population specific relative value, namely the 85th BMI centile. The limitations of the self reported data do not appear to have had a major effect on the relations that are of interest because, as mentioned earlier, the parent-child correlations are consistent with other studies based on measured heights and weights.41

In conclusion, the children of obese parents are at increased risk of obesity throughout childhood and early adult life, particularly if both parents are obese. Tracking of BMI from the age of 7 to 33 years among the children with two obese parents was much stronger than for those with two parents of normal BMI, suggesting that the BMIs of this group are less variable with increasing age. At present, this stronger tracking pattern is not predominant, at least within this British population, therefore the prevention of adult obesity is not strongly influenced by the prevention of childhood fatness.29 As the prevalence of obesity increases among adults, their offspring are likely to be affected by the stronger child-adult relations shown here. In this instance, in which the prevalence of obesity is relatively high, the prevention of adult obesity may be more dependent on the identification and prevention of fatness in childhood.