Introduction

Several childhood factors are suspected to influence the development of obesity,¹ although the factors generally considered to be important, namely dietary intake and physical (in)activity, are inherently difficult to measure. Thus, many studies rely on indicator variables, such as birthweight, height and socioeconomic position to investigate the role of childhood factors on adult health.^2,3,4 The relation between obesity and socioeconomic position measured at different life stages is therefore of interest in this context. In the Health Survey for England, 14% of women in social class I (professional occupations) are obese, increasing to 25% in class V (unskilled manual).⁵ As in other developed countries, no consistent trends are seen for men or children.⁶ But in several longitudinal studies, social position is negatively related to adult obesity in both sexes, and the effect is generally more pronounced for class of origin than for adult social position.⁷ While this implicates socioeconomic conditions in childhood, it is unclear whether particular childhood stages are more influential than others.

We therefore investigate relation between obesity and socioeconomic position in both childhood and adult life. Given that obesity is related to parental obesity and to education,^1,8,9,10,11 our second aim is to establish whether socioeconomic conditions in childhood have an independent association with adult obesity. We use the nationally representative 1958 birth cohort study to examine cross-sectional and longitudinal association.

Methods

Sample

The 1958 birth cohort includes all persons born in England, Wales and Scotland, during 3–9 March 1958. Information was obtained on 98% of births totalling 17 414, with follow-up of survivors at ages 7, 11, 16, 23 and 33 y (11 405 subjects were included at age 33 y). Full details of the study have been presented elsewhere.¹² Biases associated with sample attrition are small: within the sample used for multivariate analyses (3267 men and 3808 women), 18.9% of men were born into classes IV and V compared to 21.8% in the original sample; for women the figures are 20.5 and 22.1%, respectively.

Measures

Height and weight were: (i) measured by trained medical personnel at 7, 11 and 16 y (subjects were weighed in underclothes to the nearest pound, height was measured to the nearest inch); (ii) self-reported in 1981 at age 23 y, (iii) measured at age 33 y (height was measured without shoes using a stadiometer reading to the nearest centimeter and weight was measured in indoor clothing using Salter portable scales). Reported heights at age 23 y were checked to detect errors;¹³ these reports agree with a national study in which measurements were obtained.¹⁴

Body mass index (BMI) was calculated as weight/height² (kg/m²). The 90th percentile for BMI of the cohort was used as an arbitrary cutoff to define the fattest group to achieve adequate statistical power, with values as follows:(SeeTable 4)

Table 4.

Full table

Obesity at age 33 y was defined as BMI (kg/m²) 30 according to WHO recommendations.¹⁵ Pregnant women at age 33 y (n=256) were excluded from analyses. Heights and weights of parents were self-reported when cohort members were aged 11 y in 1969.

Social class (Registrar General's classification) was used to represent socioeconomic conditions at different ages, based on the father's occupation at the time of the cohort member's birth, and when s/he was 7, 11 and 16 y; and on the cohort member's own occupation at ages 23 and 33 y. To avoid small numbers in extreme groups, the six Registrar General's categories were collapsed to four: (1) I and II (professional and managerial), (2) III nonmanual (unskilled), (3) III manual (skilled) and (4) IV and V (semi- and unskilled manual). Most subjects (91% men, 72% women) were classified using their current occupation and the remainder according to their most recent occupation. Educational qualifications at age 33 y, were: above 'Advanced' level (29% men and 26% women), 'Advanced' level or equivalent (24 and 10%), 'Ordinary' level or equivalent (24 and 36%), less than 'Ordinary' level (14 and 17%), and no qualifications (9 and 11%). These classifications are broadly comparable to US classifications: above high school diploma, high school diploma/grade 12, grade 10, less than grade 10 and no qualifications.

Data analysis

Initially, cross-sectional associations between fatness and social class were assessed in logistic regression analyses for each age separately, that is, at 7, 11, 16, 23 and 33 y. Social class was used as a continuous variable with values 1 (classes I and II), 2 (class IIInm), 3 (class IIIm) and 4 (classes IV and V). A linear model was appropriate for analysis of obesity and social class in this sample and the results were not sensitive to the values assigned to social class.¹¹

All longitudinal analyses thereafter use obesity at age 33 y as the outcome. Associations were estimated between obesity and social class at each age. Before constructing models of social class effects on obesity at 33 y, it was important first to examine the correlation between class measures at different life stages. Correlations between child and adult social position were weak (r<0.3), and thus models with childhood and adult measures showed no collinearity problems. However, correlations between childhood measures were stronger, for example, r=0.73 for class at ages 11 and 16 y, and in order to avoid potential problems with multicollinearity it was necessary to test several models of childhood social class, omitting one age at a time. For example, in a model for women including only ages 7 and 11 y, the 7-y measure had a significant effect while the 11-y measure did not. Similarly, when modelling 7- and 16-y measures simultaneously, the 7-y measure again remained significant and the 16-y measure was not. After assessing a series of such models, we selected social class measures at birth, 7 and 16 y for presentation here. We confirmed that there was no effect of collinearity in the final models using an adaptation of the variance inflation factor for logistic regression.¹⁶

In multivariate analyses, we examined the effect of childhood and adult social class simultaneously on 33-y obesity. Models were constructed taking into account the timing of the social class measures. Starting with childhood (birth, 7 and 16 y) we selected measures for ages that remained significant in mutually adjusted models, thereafter adding adult (23 and 33 y) measures of social class to the models. Next, we assessed associations with obesity of childhood and adulthood social class after adjustment for parental BMI (kg/m²) and own education. For variables significant in the fully adjusted model, we found no significant interactions. We confirmed our results in analyses using social class as a categorical variable (data not presented).

Results

Cross-sectional association between fatness and social class are presented for several ages in Table 1. Fatness and social class were unrelated during childhood, for both sexes at ages 7 and 11 y, and among men, at age 16 y. A significant social class gradient was evident for women by age 16 y, with increasing prevalence of fatness from classes I and II to classes IV and V. Thereafter, during early adulthood, the social gradient was seen for both sexes. At age 23 y, for example, the risk of fatness increased by 38% (men) and 39% (women) for each increment in social position from classes I and II to classes IV and V. These associations suggest a predominant role for socioeconomic conditions in adulthood in the development of obesity and, conversely, that conditions in childhood are relatively unimportant for childhood fatness.

Table 1 - Prevalence (%) of fatness by social class at different ages: cross-sectional associations.

Full table

However, examining the data longitudinally in relation to adult obesity (affecting 10.8% of men and 12% of women at age 33 y), social class shows significant associations with each age during childhood (Table 2). Associations for class at birth for men (unadjusted odds ratio (OR)=1.27) and at age 7 y for women (OR=1.38) were most marked, although effects weakened slightly after allowing simultaneously for other ages in childhood. No independent associations persisted for other stages of childhood. The effect of class at birth (men) and age 7 y (women) also persisted in models with adult social class, which for men was also associated with obesity after allowing for childhood class (adjusted OR for age 23 y=1.16). Thus for men, class at birth and at age 23 y both predict adult obesity, with comparable effects of each age and no significant interaction. A slightly different pattern emerged for women: the only substantial effect was for class at age 7 y.

Table 2 - Relation between obesity at age 33 y and social class at birth and ages 7, 16, 23 and 33 y (ORs and 95% CI).

Full table

We next examined whether parental obesity or educational level of the cohort member could account for the social class/adult obesity association. ORs for obesity by social class were modestly attenuated after adjustment for parental BMI (Table 3). Education level was strongly related to obesity at age 33 y, with the highest rates for those with no qualifications (Figure 1). Class at age 23 y was of borderline significance after adjustment for education, but effects of class at birth for men and at age 7 y for women persisted (OR=1.22 and 1.28, respectively, Table 3). Finally, after adjustment for both educational level and parental BMI, a significant association remained for class at birth among men, the OR reducing from 1.35 to 1.18. A significant association also persisted for class at age 7 y among women (from 1.41 to 1.23). It should be noted that the association for education was nonlinear for women (Figure 1), but incorporating a quadratic term in the model did not affect results for social class.

Figure 1.

Prevalence of obesity (BMI (kg/m²)30) (%) and educational qualifications (see the section methods for equivalent U.S. education grades) obtained by age 33 y.

Full figure and legend (52K)

Table 3 - Relation between obesity at age 33 y, social class, parental BMI (kg/m²) and education level (ORs and 95% CI).

Full table

Discussion

In this British cohort, social class in early life, at birth for men and age 7 y for women, showed a persisting effect on obesity at age 33 y: those from unskilled manual origins had the highest risk of obesity in adulthood. The estimate from this study is equivalent to an almost twofold difference in adult obesity between those with the lowest and highest social origins. This influence of childhood circumstances was not explained by parental BMI or by the individual's own social position or educational level in adult life. Our study therefore adds to previous studies suggesting that factors in early life affect the risk of adult obesity.^{1,17,18,19,20,21}

Strengths and weaknesses

As in other epidemiological studies, we do not have a direct measure of body fat, but rely on BMI, with a cutoff of 30 kg/m²,¹⁵ identifying 10.8% of men and 12% of women as obese at age 33 y. BMI was also used to indicate childhood adiposity, which in cross-sectional comparison was categorised using the 90th percentile. The availability of data collected prospectively allows simultaneous investigation of socioeconomic position at multiple ages over three decades in a large nationally representative British sample. The findings therefore apply to contemporary young adults.

Social conditions in childhood

No social class trend in childhood obesity was evident in our cross-sectional analyses. This is consistent with many previous studies^6,22 and with recent studies of British children.^23,24 By early adulthood a social gradient had emerged, with increasing risk of obesity from professional/managerial to unskilled manual classes. This corresponds to trends in other cross-sectional studies of adults, although usually only for women.⁵ While these social trends from birth to age 33 y in the 1958 cohort suggest that class-related factors in adulthood have a predominant influence on adult obesity, the longitudinal relation suggest otherwise. Indeed a major finding of our study is that cross-sectional trends can be misleading, by obscuring important effects of childhood socioeconomic conditions on adult obesity.

Critical periods in the development of obesity?

A consistent pattern is now emerging in which the salience of childhood social position is recognised.¹ However, most if not all studies on the influence of child and adult social position, including our own previous studies of this cohort,^13,25 represent the entire childhood period with one socioeconomic measure. Consequently, these studies cannot differentiate between stages in early life; yet, more than one developmental stage may have long-lasting effects. At least three periods in childhood have been proposed as critical for obesity, namely (1) gestation and early infancy, (2) the period of adiposity rebound (5–7 y), and (3) adolescence,¹⁸ but some uncertainty exists on whether each or all of these are critical.²⁶ Our findings on the long-lasting impact of early life conditions shed some light on this issue. The associations for social class at birth (men) and age 7 y (women) are consistent with gestation/early infancy or the time of adiposity rebound being important periods during which factors might influence adult obesity. This interpretation is supported by previous work on this cohort, which showed that individuals from lower class origins gained more weight from childhood through to early adulthood than those from higher social classes.^13,25 Evidence is also now growing for an influence of maternal smoking in pregnancy increasing offspring risk of obesity, which argues for prenatal effects.^27,28 Less support is available here for adolescence being an important life stage, although we are unable to discount this possibility. It should be remembered that the cohort has a shared generational environment of increasing high-fat diets and reducing physical demands of work, home and travel.²⁹ Thus, for this generation, the effects of early life factors may well be conditional upon the environment in adolescence and adulthood.

Implications for social inequalities in obesity

Social inequalities in obesity are observed at least among adult women within Britain,⁵ and across developed countries,⁶ and as seen here in the 1958 cohort. Our study suggests that this trend is due in large part to class-related factors in early life and that inequalities in obesity will not be explained by adult life factors alone. Consistent with this interpretation is one recent study in which inequalities in obesity among women were not explained by differences in adult lifestyles, namely physical activity, smoking and alcohol consumption.³⁰

Specific factors

Although our study identifies early life as an important stage, it does not identify the specific factors involved in the development of adult obesity. The parent–offspring obesity relation is well established, yet it is only rarely that social position is examined taking this factor into account.¹ Our study suggests that the effect of social position in early life is not primarily a reflection of parental size. Education is clearly another important factor,^8,9,21 particularly since recent increases in the prevalence of obesity have been greater among those with lower educational levels than those with higher education.^31,32 Not surprisingly, we found a strong trend in obesity by educational level, and this association appeared to underlie the relationship among men with adult social class. However, confounding by education failed to explain the relation for social class in early life. This directs our attention to factors in early life, such as prenatal growth, infant feeding, childhood diet and activity levels. Future research should therefore investigate early life environmental factors that might account for our findings.

References

1. Parsons TJ, Power C, Logan S, Summerbell CD. Childhood predictors of adult obesity: a systematic review. Int J Obes Relat Metab Disord 1999; 23(Suppl 8): S1–S107. | PubMed |
2. Barker DJP. Mothers, babies and disease in later life. BMJ: Tavistock, London, 1994.
3. Gunnell DJ, Davey Smith G, Frankel S, Nanchahal K, Braddon FEM, Pemberton J, Peters T. Childhood leg length and adult mortality: follow up of the Carnegie (Boyd Orr) Survey of Diet and Health in Pre-war Britain. J Epidemiol Commun Health 1998; 52: 142–152.
4. Lynch JW, Kaplan GA, Cohen RD, Kauhanen J, Wilson TW, Smith NL, Salonen JT. Childhood and adult socioeconomic status as predictors of mortality in Finland. Lancet 1994; 343: 524–527.
5. Prescott-Clarke P, Primatesta P. Health survey for England '96. The Stationary Office: London 1998.
6. Sobal J, Stunkard AJ. Socioeconomic status and obesity: a review of the literature. Psychol Bull 1989; 105: 260–275. | Article | PubMed | ISI | ChemPort |
7. Power C, Parsons T. Nutritional and other influences in childhood as predictors of adult obesity. Proc Nutr Soc 2000; 59: 1–6. | PubMed |
8. Braddon FEM, Rodgers B, Wadsworth MEJ, Davies JMC. Onset of obesity in a 36 year birth cohort study. BMJ 1986; 293: 299–303. | PubMed | ChemPort |
9. Flegal KM, Harlan WR, Landis JR. Secular trends in body mass index and skinfold thickness with socioeconomic factors in young adult men. Am J Clin Nutr 1988; 48: 544–551. | PubMed |
10. Kuskowska-Wolk A, Bergstrom R. Trends in body mass index and prevalence of obesity in Swedish men 1980–89. J Epidemiol Commun Health 1993; 47: 103–108.
11. Manor O, Matthews S, Power C. Comparing measures of health inequality. Soc Sci Med 1997; 45: 761–771.
12. Ferri E (ed.). Life at 33: the fifth follow-up of the National Child Development Study. National Children's Bureau: London; 1993.
13. Power C, Moynihan C. Social class and changes in weight-for-height between childhood and early adulthood. Int J Obes Relat Metab Disord 1988;12: 445–453. | ChemPort |
14. Knight I. The heights and weights of young adults in Great Britain. HMSO: London; 1984.
15. WHO Expert Committee. Physical status: the use and interpretation of anthropometry. WHO Technical Report Series 854 Geneva: WHO,1995.
16. Wax Y. Collinearity diagnostics for a relative risk regression analysis. Stat Med 1992; 11: 1273–1287. | Article | PubMed | ISI | ChemPort |
17. Whitaker RC, Dietz WH. Role of the prenatal environment in the development of obesity. J Pediatr 1998; 132: 768–776. | PubMed | ISI | ChemPort |
18. Dietz WH. Critical periods in childhood for the development of obesity. Am J Clin Nutr 1994; 59: 955–959. | PubMed | ISI | ChemPort |
19. Laitinen J, Power C, Jarvelin MR. Family social class, maternal body mass index, childhood body mass index, and age at menarche as predictors of adult obesity. Am J Clin Nutr 2001;74: 287–294. | PubMed | ISI | ChemPort |
20. Whitaker RC, Pepe MS, Wright JA, Seidel KD, Dietz WH. Early adiposity rebound and the risk of adult obesity. Pediatrics 1998; 101: E5. | Article | PubMed | ChemPort |
21. Hardy R, Wadsworth M, Kuh D. The influence of childhood weight and socioeconomic status on change in adult body mass index in a British national birth cohort. Int J Obes Relat Metab Disord 2000; 24: 725–734. | Article | PubMed | ChemPort |
22. Peckham CS, Stark O, Simonite V, Wolff OH. Prevalence of obesity in British children born in 1946 and 1958. BMJ 1983; 286: 1237–1242.
23. Duran-Tauleria E, Rona RJ, Chinn S. Factors associated with weight for height and skinfold thickness in British children. J Epidemiol Commun Health 1995;49:466–473.
24. Prescott-Clarke P, Primatesta P. Health Survey for England: the health of young people '95–'97. The Stationary Office: London; 1998.
25. Power C, Matthews S. Origins of health inequalities in a national population sample. Lancet 1997;350: 1584–1589. | Article | PubMed | ChemPort |
26. British Nutrition Foundation Task Force on Obesity. Obesity: the report of the British Nutrition task force. Blackwell Science: Oxford, 1999.
27. Williams S, Poulton R. Twins and maternal smoking: ordeals for the fetal origins hypothesis? A cohort study. BMJ 2000; 318: 897–900.
28. Ong KK, Ahmed ML, Emmett PM, Preece MA, Dunger DB. Association between postnatal catch-up growth and obesity in childhood: prospective cohort study. BMJ 2000; 320: 961–971.
29. Prentice AM, Jebb SA. Obesity in Britain: gluttony or sloth? BMJ 1995;311: 437–439. | PubMed | ChemPort |
30. Jebb SA, Prentice AM, Cole TJ. Obesity and social class in women: effects of smoking and physical activity. Proc Nutr Soc 1997; 56: 159A.
31. Pekkanen J, Uutela A, Valkonen T, Vartiainen E, Tuomilehto J, Puska P. Coronary risk factor levels: differences between educational groups in 1972–87 in eastern Finland. J Epidemiol Commun Health 1995; 49: 144–149.
32. Seidell JC, Flegal KM. Assessing obesity: classification and epidemiology. Br Med Bull 1997; 53: 238–252. | PubMed | ChemPort |

Acknowledgements

The research was supported by the UK Economic and Social Research Council under the Health Variations Programme (L128251021). The Canadian Institute for Advanced Research supports C Power as a Fellow. Data acknowledgment: Centre for Longitudinal Studies Institute of Education. National Child Development Study Composite File including selected Perinatal Data and sweeps one to five [computer file]. National Birthday Trust Fund, National Children's Bureau, City University Social Statistics Research Unit [original data producers]. The Data Archive. SN: 3148. Colchester. 1994.