- Split View
-
Views
-
Cite
Cite
Tessa J Parsons, Orly Manor, Chris Power, Physical activity and change in body mass index from adolescence to mid-adulthood in the 1958 British cohort, International Journal of Epidemiology, Volume 35, Issue 1, February 2006, Pages 197–204, https://doi.org/10.1093/ije/dyi291
- Share Icon Share
Abstract
Background Prevention of obesity has focused on childhood as a target period. Our aim was to assess whether frequency of adolescent physical activity affected subsequent body mass index (BMI) gain through to mid-adulthood.
Methods The British birth cohort of all births in 1 week in March 1958, includes information on physical activity frequency and BMI for several ages, 11–45 years. We examined relationships between activity in adolescence and trajectories of BMI between 16 years (or 23 years) and 45 years using multi-level models. Effects of change in activity on BMI and on change in BMI were tested using ANOVA.
Results Physical activity at 11 years had no effect on the BMI trajectories, in males or females. More active females at 16 years gained BMI more slowly than others, by 0.007 kg/m2/year per activity category over the period 16–45 years, whereas the most active males gained BMI faster than others, by 0.005 kg/m2/year per activity category. This effect in males was not evident on the BMI trajectory from 23 to 45 years. Consistent with these analyses, change in activity was associated with change in BMI in females, e.g. females active at 16 and 42 years gained less BMI than inactive females (2.1 vs 2.5 kg/m2/10 years). Results for males were inconsistent over the time periods examined.
Conclusions Physical activity may lessen the gains in BMI from adolescence onwards, but relationships vary with age, and in later adolescence show opposite effects for males and females. Decreasing activity between adolescence and mid-adulthood in males, and inactivity in both life stages in females may increase BMI gain.
Introduction
With rates of obesity continuing to rise rapidly in adults and children worldwide, a recent report emphasized ‘the need for prevention as the only feasible solution for developed and undeveloped countries alike’.1 Childhood is a target period for prevention for several reasons; (i) obesity tends to track, so that fatter children are more likely to become fatter adults,2 (ii) evidence is accumulating that children, like adults, suffer from adverse health consequences relating to their obesity,3 (iii) children may be more amenable to intervention strategies than adults,4 and (iv) treatment of obesity is difficult, with limited effectiveness.5 In addition, lifestyle factors with the potential to influence obesity, such as physical activity and diet, generally show moderate tracking over time,6–8 and it has been argued that healthy lifestyle habits should be encouraged in children, rather than allowing unhealthy habits to become established, which may require change later.9
Based on indirect evidence of transport use, and sedentary activities such as television viewing, it is argued that physical activity is a primary factor underlying the current high rates of obesity.10 Time trend data show that motorized transport use and hours spent watching television have been increasing over the last 30 years or so,11,12 providing suggestive evidence for declining activity levels. Walking and cycling have decreased in all age groups between 1985–86 and 1997–99, with the greatest falls being in school-age children and young adult males.12 In children aged 14 years or less, walking and cycling fell by 20 and 26%, respectively, in the UK between 1985 and 199213 and by 40% in the US between 1977 and 1995.14 In contrast, car travel by children in the UK increased concurrently by 40%.13 While arguments on the role of decreasing physical activity are persuasive, conclusive evidence for a causal link between physical inactivity and obesity is lacking, partly because there are few prospective studies with relevant data to address this issue. Even with relevant prospective data, the direction of association is not always clear-cut, as suggested by our study of cross-sectional relationships between physical activity and body mass index (BMI) in the 1958 British birth cohort, in which the relationships changed with age.15 This variation was particularly evident in males, in whom the direction of the relationship changed, such that in adolescence (16 years), BMI increased with increasing activity, whereas in childhood and adulthood BMI increased with decreasing activity.15 Changing selection effects across the lifecourse probably explain this phenomenon; bigger adolescent boys have an advantage in the sports they are likely to participate in, and are therefore preferentially selected into sports activities.16,17 This makes investigating longitudinal relationships between physical activity and BMI particularly complex. The aim of this study was to investigate whether adolescent physical activity influences subsequent changes in BMI through to mid-adulthood life. Specifically, we examine physical activity on two occasions (11 and 16 years) in relation to trajectories of BMI from 16 to 45 years, within the well-characterized 1958 British birth cohort.
Participants and methods
Study population
All children born in England, Scotland, and Wales in 1 week of March 1958 were included in the 1958 British birth cohort. From a target population of 17 733 births, information was obtained on 98%. Follow-ups of surviving children were conducted at ages 7, 11, 16, 23, 33, 42, and 45 years.18 At age 45 years, from a target sample of 12 070 participants, 9377 provided some information. Sample attrition has resulted in a slight under-representation of those who are most disadvantaged, but the remaining sample is generally representative of the original sample.6,19 For the purposes of this study, we compared mean BMI at 7 and 11 years for all cohort members with data (from original dataset), with means for cohort members included in analyses of BMI trajectories between 16 and 45 years (those with at least one BMI measure between 16 and 45 years plus a measure of physical activity at 16 years). Differences in mean BMI at 7 or 11 years between the original dataset and sub-sample included in analyses were very small, in males and females, ≤0.04 kg/m2.
Ethical committee approval for the 45 year survey was obtained from the South-East Multi-Centre Research Ethics Committee, UK. For the 33 year survey, ethical committee approval was not sought, although cohort members were asked to give written consent for access to medical records.
Body mass index
BMI (kg/m2) was calculated from heights and weights. At 11 and 16 years, heights (to the nearest inch) and weights (in underclothes, to the nearest pound) were measured by trained medical personnel. At 23 years, self-reports of weight and height were obtained. At 33 and 45 years, weight was measured with indoor clothing, without shoes, to the nearest 0.1 kg, and height was measured to the nearest centimetre at 33 years and to the nearest millimetre at 45 years. Data have been checked to detect measurement and data entry errors.19,20 BMI values for women who were pregnant at 33 years (n = 256), were excluded.
Physical activity
Physical activity frequency was assessed by questionnaire at 11 and 16 years. At 11 years, the mother responded to four questions about how often the child used parks, recreation grounds, swimming pools, and indoor play centres, and the child answered one question on how often (s)he played outdoor sport or took part in sport outside school hours. These five activity variables were combined to give a single variable with four categories ranging from least active to most active.15 At 16 years, participants were asked how often they played outdoor and indoor games and sports, and went swimming and dancing. We combined the four activity variables to give a single four category variable.15
In adulthood, participants responded to a single question, about frequency of sports participation at 23 years, and of regular physical activity at 33 and 42 years.15 As an aid to recall, the question at 33 and 42 years provided several examples of activities, but this was not provided at 23 years. Original categories ranged, at 23 years, from ‘5 times per week or more’, to ‘not at all in the last 4 weeks’ and at 33 and 42 years, ‘most days’ to ‘<2–3 times per month’. We collapsed the original six categories to four, to avoid small proportions of participants in some categories.15
Potential confounding factors
The influence of several previously identified potential confounding factors was considered.
Maternal body mass index: maternal height without shoes was measured, and pre-pregnant weight self-reported in categories of 1 stone, shortly after the birth of the cohort member. To estimate BMI, mothers were assigned a weight equivalent to the mid-point of their weight group.
Social class was defined using father's occupation, according to the UK 1951 General Registrar's classification. We used two categories in our analyses; (i) classes I, II, and III-NM (professional, managerial, and skilled non-manual) and (ii) III-M, IV, and V (skilled, semi-skilled, and unskilled manual, and those recorded as having ‘no male head of household’).
Pubertal stage was assessed by doctors at 11 and 16 years. Measures used in this study (based on previous work21) include: for girls, breast development at 11 years was assessed on a scale of 5 (1 = preadolescent, 5 = mature) and age at menarche reported at 16 years; for boys, axillary hair at 16 years was assessed as absent, sparse, intermediate, or adult, and age of voice breaking reported at 16 years.
Television viewing was assessed by questionnaire, at 11, 16, and 23 years. At 11 and 16 years cohort members reported how often they watched television as often, sometimes, or never/hardly ever. Few children reported watching television never or hardly ever (2–6%) and, therefore, the sometimes and never/hardly ever categories were combined. At 11 years ‘often’ was defined as nearly every day, but at 16 years ‘often’ was undefined. At 23 years, cohort members reported how often they watched television in six categories ranging from ‘5 times per week or more’, to ‘not at all in the last 4 weeks’. In this paper we used two categories; ≥5 times per week and ≤4 times per week.
Statistical analyses
Measures of BMI at different time-points within individuals are correlated: in the 1958 birth cohort correlation coefficients between successive ages ranged from 0.64 to 0.79, with a lower coefficient of ∼0.5 for the longer time period between 16 and 45 years. We used multilevel models to model the BMI trajectory over time and to allow for the correlation between repeated BMI measures. These models allow for missing outcome data, so that all participants with at least one BMI measure (and data on all independent variables) are included in the analyses. It is assumed that outcome data are missing at random, i.e. that the likelihood of a measurement being missing is unrelated to the actual measurement value. Complete data on BMI (i.e. at all four time-points) was available for about a third of the participants and a further 29% had data at three time-points. On investigation, we found no clear trend between BMI at 45 years and number of measurements available, supporting the model assumption of data missing at random. All analyses were performed for males and females separately, using MLWin version 1.10, Institute of Education, London, UK.22
To make inferences about causality, analyses were designed so that the timing of predictors preceded the outcome, or in some instances coincided with the first BMI outcome measure. Thus, for analyses of physical activity at 11 or 16 years, the BMI trajectory between 16 and 45 years was modelled. To establish effects of activity in adolescence on adult BMI trajectories, we also modelled BMI between 23 and 45 years. This allowed us to include activity at 23 years as an additional predictor.
Stages of analyses were as follows. First, change in BMI with age was modelled, using both linear and quadratic age terms. Because each follow-up of the cohort members took place over a period of time (6–19 months), exact age (to the nearest month) was used wherever possible. Age was centred at 33 years so that the intercept of the model represents mean BMI at 33 years. The slope represents the linear change in BMI per year. Random effects at the individual level were included for the intercept and linear age term, allowing both intercept and linear slope to vary between individuals. Random effects for the quadratic age term were not significant. Second, we added physical activity (as a continuous variable) to the model, allowing the intercept (BMI at age 33) to vary by level of activity. An interaction term for physical activity by age was also added to test whether the slope of the BMI trajectory varied by physical activity. To examine whether any effect of physical activity in childhood operated through continuities in activity habits over time, we adjusted for activity at 33 years and its interaction by age term. Adjustment for potential confounding factors, including childhood BMI, mother's BMI, social class, pubertal stage, and television viewing and their interaction terms by age were also examined. Third, we carried out several sensitivity analyses. We repeated main analyses (i) with physical activity as a categorical variable, (ii) after transforming BMI using a natural logarithm, and (iii) accounting for the fact that the variance of BMI increases with age by adding a random effect at the time-point level to the age coefficient. All sensitivity analyses gave similar results to those presented here.
Finally we examined changes in activity between adolescence and adulthood, in relation to BMI at 16 years and to subsequent BMI gain. We obtained categories of change in activity from a cross-classification of the top two vs the bottom two physical activity categories at two time-points, (i) 16 and 23 years and (ii) 16 and 45 years. ANOVA estimates were used to test whether those changing their activity status between 16 and 23 years (or 45 years) differed from those remaining in the same activity group at both ages. For example, whether those who changed from inactive to active differed from those who remained inactive. We also compared the active at both time-points with the inactive. Bonferroni methods were used to allow for multiple comparisons. To check whether results were sensitive to the classification of activity we repeated analyses with a different classification; findings were similar. All non-multi-level analyses (means, correlations, distributions, ANOVA) were carried out using SPSS for Windows, version 12.0.1, SPSS Inc., Chicago, Illinois.
Results
BMI increased with age, and was higher in females than males at 16 years, but higher in males from 23 years onwards (Table 1). Males were more active than females at 11, 16, and 23 years, whereas activity was similar at 33 and 42 years (Table 2). Physical activity had a low to moderate correlation between successive time-points, with coefficients between 0.1 and 0.2.
. | Males . | . | . | Females . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
. | Mean . | SD . | n . | Mean . | SD . | n . | ||||
BMI at 16 years | 20.24 | 2.72 | 5698 | 21.00 | 2.96 | 5342 | ||||
BMI at 23 years | 23.10 | 2.90 | 6134 | 22.12 | 3.25 | 6145 | ||||
BMI at 33 years | 25.62 | 3.99 | 5496 | 24.60 | 4.89 | 5424 | ||||
BMI at 45 years | 27.80 | 4.24 | 4590 | 26.96 | 5.53 | 4626 |
. | Males . | . | . | Females . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
. | Mean . | SD . | n . | Mean . | SD . | n . | ||||
BMI at 16 years | 20.24 | 2.72 | 5698 | 21.00 | 2.96 | 5342 | ||||
BMI at 23 years | 23.10 | 2.90 | 6134 | 22.12 | 3.25 | 6145 | ||||
BMI at 33 years | 25.62 | 3.99 | 5496 | 24.60 | 4.89 | 5424 | ||||
BMI at 45 years | 27.80 | 4.24 | 4590 | 26.96 | 5.53 | 4626 |
. | Males . | . | . | Females . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
. | Mean . | SD . | n . | Mean . | SD . | n . | ||||
BMI at 16 years | 20.24 | 2.72 | 5698 | 21.00 | 2.96 | 5342 | ||||
BMI at 23 years | 23.10 | 2.90 | 6134 | 22.12 | 3.25 | 6145 | ||||
BMI at 33 years | 25.62 | 3.99 | 5496 | 24.60 | 4.89 | 5424 | ||||
BMI at 45 years | 27.80 | 4.24 | 4590 | 26.96 | 5.53 | 4626 |
. | Males . | . | . | Females . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
. | Mean . | SD . | n . | Mean . | SD . | n . | ||||
BMI at 16 years | 20.24 | 2.72 | 5698 | 21.00 | 2.96 | 5342 | ||||
BMI at 23 years | 23.10 | 2.90 | 6134 | 22.12 | 3.25 | 6145 | ||||
BMI at 33 years | 25.62 | 3.99 | 5496 | 24.60 | 4.89 | 5424 | ||||
BMI at 45 years | 27.80 | 4.24 | 4590 | 26.96 | 5.53 | 4626 |
Age . | Activity categorya . | Males (%) . | Females (%) . |
---|---|---|---|
11 years | Activity level 4 | 19 | 13 |
Activity level 3 | 27 | 26 | |
Activity level 2 | 27 | 29 | |
Activity level 1 | 27 | 32 | |
Total n | 7708 | 7298 | |
16 years | Activity level 4 | 23 | 14 |
Activity level 3 | 16 | 13 | |
Activity level 2 | 19 | 19 | |
Activity level 1 | 41 | 54 | |
Total n | 6114 | 5863 | |
23 years | ≥3 times a week | 21 | 8 |
Once or twice a week | 21 | 14 | |
1–3 times in last 4 weeks | 17 | 14 | |
Not at all in last 4 weeks | 41 | 64 | |
Total n | 6253 | 6266 | |
33 years | 4–7 days a week | 24 | 27 |
2–3 days a week | 23 | 19 | |
Once a week | 21 | 23 | |
2–3 times per month or less | 32 | 31 | |
Total n | 5572 | 5770 | |
42 years | 4–7 days a week | 25 | 28 |
2–3 days a week | 22 | 21 | |
Once a week | 20 | 17 | |
2–3 times per month or less | 34 | 35 | |
Total n | 5599 | 5772 |
Age . | Activity categorya . | Males (%) . | Females (%) . |
---|---|---|---|
11 years | Activity level 4 | 19 | 13 |
Activity level 3 | 27 | 26 | |
Activity level 2 | 27 | 29 | |
Activity level 1 | 27 | 32 | |
Total n | 7708 | 7298 | |
16 years | Activity level 4 | 23 | 14 |
Activity level 3 | 16 | 13 | |
Activity level 2 | 19 | 19 | |
Activity level 1 | 41 | 54 | |
Total n | 6114 | 5863 | |
23 years | ≥3 times a week | 21 | 8 |
Once or twice a week | 21 | 14 | |
1–3 times in last 4 weeks | 17 | 14 | |
Not at all in last 4 weeks | 41 | 64 | |
Total n | 6253 | 6266 | |
33 years | 4–7 days a week | 24 | 27 |
2–3 days a week | 23 | 19 | |
Once a week | 21 | 23 | |
2–3 times per month or less | 32 | 31 | |
Total n | 5572 | 5770 | |
42 years | 4–7 days a week | 25 | 28 |
2–3 days a week | 22 | 21 | |
Once a week | 20 | 17 | |
2–3 times per month or less | 34 | 35 | |
Total n | 5599 | 5772 |
At 11 and 16 years activity level 1 is least active, level 4 is the most active.
Age . | Activity categorya . | Males (%) . | Females (%) . |
---|---|---|---|
11 years | Activity level 4 | 19 | 13 |
Activity level 3 | 27 | 26 | |
Activity level 2 | 27 | 29 | |
Activity level 1 | 27 | 32 | |
Total n | 7708 | 7298 | |
16 years | Activity level 4 | 23 | 14 |
Activity level 3 | 16 | 13 | |
Activity level 2 | 19 | 19 | |
Activity level 1 | 41 | 54 | |
Total n | 6114 | 5863 | |
23 years | ≥3 times a week | 21 | 8 |
Once or twice a week | 21 | 14 | |
1–3 times in last 4 weeks | 17 | 14 | |
Not at all in last 4 weeks | 41 | 64 | |
Total n | 6253 | 6266 | |
33 years | 4–7 days a week | 24 | 27 |
2–3 days a week | 23 | 19 | |
Once a week | 21 | 23 | |
2–3 times per month or less | 32 | 31 | |
Total n | 5572 | 5770 | |
42 years | 4–7 days a week | 25 | 28 |
2–3 days a week | 22 | 21 | |
Once a week | 20 | 17 | |
2–3 times per month or less | 34 | 35 | |
Total n | 5599 | 5772 |
Age . | Activity categorya . | Males (%) . | Females (%) . |
---|---|---|---|
11 years | Activity level 4 | 19 | 13 |
Activity level 3 | 27 | 26 | |
Activity level 2 | 27 | 29 | |
Activity level 1 | 27 | 32 | |
Total n | 7708 | 7298 | |
16 years | Activity level 4 | 23 | 14 |
Activity level 3 | 16 | 13 | |
Activity level 2 | 19 | 19 | |
Activity level 1 | 41 | 54 | |
Total n | 6114 | 5863 | |
23 years | ≥3 times a week | 21 | 8 |
Once or twice a week | 21 | 14 | |
1–3 times in last 4 weeks | 17 | 14 | |
Not at all in last 4 weeks | 41 | 64 | |
Total n | 6253 | 6266 | |
33 years | 4–7 days a week | 24 | 27 |
2–3 days a week | 23 | 19 | |
Once a week | 21 | 23 | |
2–3 times per month or less | 32 | 31 | |
Total n | 5572 | 5770 | |
42 years | 4–7 days a week | 25 | 28 |
2–3 days a week | 22 | 21 | |
Once a week | 20 | 17 | |
2–3 times per month or less | 34 | 35 | |
Total n | 5599 | 5772 |
At 11 and 16 years activity level 1 is least active, level 4 is the most active.
Physical activity at 11 years had no effect on BMI at age 33 years (intercept), and no effect on the slope of the BMI trajectory, either from 16 to 45 years or from 23 to 45 years in males or females (Table 3). Physical activity at 16 years showed significant effects on both BMI at 33 years and on the slope of the trajectory from 16 to 45 years, but in opposite directions for males and females. More active females at 16 years had a lower mean BMI at 33 years than their less active peers (by 0.15 kg/m2 per activity category), but the more active males at 16 years had a higher mean BMI (by 0.23 kg/m2 per activity category). Between 16 and 45 years, more active females gained BMI more slowly than less active females, by 0.007 kg/m2/year per activity category, while more active males gained BMI faster than less active males, by 0.005 kg/m2/year per activity category (Table 3). This effect of activity at 16 years in females was similar in a model where the outcome was a BMI trajectory between 23 and 45 years, but in males, activity at 16 years had no effect on the BMI trajectory between 23 and 45 years (Table 3). Also in males, activity at 23 years had no effect on BMI at 33 years (intercept), or on the BMI trajectory between 23 and 45 years. More active females at 23 years had a lower BMI at 33 years (by 0.38 kg/m2 per activity category) and gained BMI more slowly between 23 and 45 years, by 0.010 kg/m2/year per activity category (Table 3).
. | Males . | . | . | Females . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
Activity predictor . | Intercept at age 33 years (kg/m2) coefficient (95% CI) . | Linear change from age 16 years (kg/m2/yr) coefficient (95% CI) . | n . | Intercept at age 33 years (kg/m2) coefficient (95% CI) . | Linear change from age 16 years (kg/m2/yr) coefficient (95% CI) . | n . | ||||
11 years | −0.010 (−0.086 to 0.066) | 0.000 (−0.004 to 0.004) | 7124 | −0.087 (−0.191 to 0.017) | −0.001 (−0.005 to 0.003) | 6797 | ||||
16 years | 0.233 (0.159 to 0.307) | 0.005 (0.001 to 0.009) | 5897 | −0.148 (−0.250 to −0.048) | −0.007 (−0.011 to −0.003) | 5674 |
. | Males . | . | . | Females . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
Activity predictor . | Intercept at age 33 years (kg/m2) coefficient (95% CI) . | Linear change from age 16 years (kg/m2/yr) coefficient (95% CI) . | n . | Intercept at age 33 years (kg/m2) coefficient (95% CI) . | Linear change from age 16 years (kg/m2/yr) coefficient (95% CI) . | n . | ||||
11 years | −0.010 (−0.086 to 0.066) | 0.000 (−0.004 to 0.004) | 7124 | −0.087 (−0.191 to 0.017) | −0.001 (−0.005 to 0.003) | 6797 | ||||
16 years | 0.233 (0.159 to 0.307) | 0.005 (0.001 to 0.009) | 5897 | −0.148 (−0.250 to −0.048) | −0.007 (−0.011 to −0.003) | 5674 |
. | . | Linear change from age 23 years (kg/m2/yr) . | . | . | Linear change from age 23 yr (kg/m2/yr) . | . |
---|---|---|---|---|---|---|
11 years | 0.002 (−0.076 to 0.080) | 0.000 (−0.004 to 0.004) | 6576 | −0.087 (−0.191 to 0.017) | −0.002 (−0.008 to 0.004) | 6412 |
16 years | 0.249 (0.173 to 0.325) | 0.000 (−0.004 to 0.004) | 5397 | −0.153 (−0.257 to −0.049) | −0.007 (−0.013 to −0.001) | 5311 |
23 years | −0.013 (−0.086 to 0.060) | −0.004 (−0.008 to 0.000) | 6217 | −0.376 (−0.484 to −0.268 | −0.010 (−0.016 to −0.004) | 6233 |
. | . | Linear change from age 23 years (kg/m2/yr) . | . | . | Linear change from age 23 yr (kg/m2/yr) . | . |
---|---|---|---|---|---|---|
11 years | 0.002 (−0.076 to 0.080) | 0.000 (−0.004 to 0.004) | 6576 | −0.087 (−0.191 to 0.017) | −0.002 (−0.008 to 0.004) | 6412 |
16 years | 0.249 (0.173 to 0.325) | 0.000 (−0.004 to 0.004) | 5397 | −0.153 (−0.257 to −0.049) | −0.007 (−0.013 to −0.001) | 5311 |
23 years | −0.013 (−0.086 to 0.060) | −0.004 (−0.008 to 0.000) | 6217 | −0.376 (−0.484 to −0.268 | −0.010 (−0.016 to −0.004) | 6233 |
. | Males . | . | . | Females . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
Activity predictor . | Intercept at age 33 years (kg/m2) coefficient (95% CI) . | Linear change from age 16 years (kg/m2/yr) coefficient (95% CI) . | n . | Intercept at age 33 years (kg/m2) coefficient (95% CI) . | Linear change from age 16 years (kg/m2/yr) coefficient (95% CI) . | n . | ||||
11 years | −0.010 (−0.086 to 0.066) | 0.000 (−0.004 to 0.004) | 7124 | −0.087 (−0.191 to 0.017) | −0.001 (−0.005 to 0.003) | 6797 | ||||
16 years | 0.233 (0.159 to 0.307) | 0.005 (0.001 to 0.009) | 5897 | −0.148 (−0.250 to −0.048) | −0.007 (−0.011 to −0.003) | 5674 |
. | Males . | . | . | Females . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|---|
Activity predictor . | Intercept at age 33 years (kg/m2) coefficient (95% CI) . | Linear change from age 16 years (kg/m2/yr) coefficient (95% CI) . | n . | Intercept at age 33 years (kg/m2) coefficient (95% CI) . | Linear change from age 16 years (kg/m2/yr) coefficient (95% CI) . | n . | ||||
11 years | −0.010 (−0.086 to 0.066) | 0.000 (−0.004 to 0.004) | 7124 | −0.087 (−0.191 to 0.017) | −0.001 (−0.005 to 0.003) | 6797 | ||||
16 years | 0.233 (0.159 to 0.307) | 0.005 (0.001 to 0.009) | 5897 | −0.148 (−0.250 to −0.048) | −0.007 (−0.011 to −0.003) | 5674 |
. | . | Linear change from age 23 years (kg/m2/yr) . | . | . | Linear change from age 23 yr (kg/m2/yr) . | . |
---|---|---|---|---|---|---|
11 years | 0.002 (−0.076 to 0.080) | 0.000 (−0.004 to 0.004) | 6576 | −0.087 (−0.191 to 0.017) | −0.002 (−0.008 to 0.004) | 6412 |
16 years | 0.249 (0.173 to 0.325) | 0.000 (−0.004 to 0.004) | 5397 | −0.153 (−0.257 to −0.049) | −0.007 (−0.013 to −0.001) | 5311 |
23 years | −0.013 (−0.086 to 0.060) | −0.004 (−0.008 to 0.000) | 6217 | −0.376 (−0.484 to −0.268 | −0.010 (−0.016 to −0.004) | 6233 |
. | . | Linear change from age 23 years (kg/m2/yr) . | . | . | Linear change from age 23 yr (kg/m2/yr) . | . |
---|---|---|---|---|---|---|
11 years | 0.002 (−0.076 to 0.080) | 0.000 (−0.004 to 0.004) | 6576 | −0.087 (−0.191 to 0.017) | −0.002 (−0.008 to 0.004) | 6412 |
16 years | 0.249 (0.173 to 0.325) | 0.000 (−0.004 to 0.004) | 5397 | −0.153 (−0.257 to −0.049) | −0.007 (−0.013 to −0.001) | 5311 |
23 years | −0.013 (−0.086 to 0.060) | −0.004 (−0.008 to 0.000) | 6217 | −0.376 (−0.484 to −0.268 | −0.010 (−0.016 to −0.004) | 6233 |
Adjustment for adult activity (at 33 years) had little effect on our findings. Adjustment for potential confounding factors (mother's BMI, socioeconomic status, pubertal stage, television viewing, activity at 33 years) had little or modest effects on our findings. To illustrate the latter, in females, adjusting for television viewing slightly reduced the effect of activity at 23 years on the 23–45 years BMI trajectory (−0.010 to −0.008 kg/m2/year).
We conducted further analyses in order to better understand the effects of adolescent physical activity on the BMI slope, in particular to clarify the change from a positive to null effect for activity at 16 years among males. Figures 1 and 2 present changes in activity from age 16 years in relation to (i) BMI at 16 years and (ii) change in BMI from age 16 years. In both males and females, BMI at 16 years was similar for all activity groups representing stability or change, 16–23 years and 16–42 years. In females, similar patterns were seen for change in activity and change in BMI for both time periods, 16–23 years and 16–45 years. For example, females decreasing their activity between 16 and 42 years had a higher BMI gain (16–45 years) than those remaining active, 2.3 vs 2.1 kg/m2/10 years, (borderline significance) and those remaining inactive gained significantly more than those increasing their activity, 2.5 vs 2.2 kg/m2/10 years (Figure 2b).
In contrast, males who decreased their activity between 16 and 23 years showed a similar mean gain in BMI to those who were active at both time-points, 3.8 vs 3.9 kg/m2/10 years (Figure 1b). Likewise, males inactive at both time-points showed a similar BMI gain, of ∼3.4 kg/m2/10 years, to those who increased their activity by 23 years (Figure 1b). (But as suggested by Table 3, BMI gain between 16 and 23 years was significantly greater in males who were active compared with inactive at both time-points.) Over the longer time period 16–45 years, a shift in the pattern of BMI gain was seen. Males who became inactive showed a significantly larger mean BMI gain than males who were active at 16 and 42 years, 3.2 vs 2.8 kg/m2/10 years, whereas males who became active showed a similar BMI gain to those who remained inactive, 2.8 vs 2.9 kg/m2/10 years (Figure 2b). Thus, those remaining active had a similar BMI gain to those remaining inactive (2.8 vs 2.9 kg/m2/10 years) over this longer time period.
Discussion
The main focus of our study was the influence of physical activity in adolescence on the change in BMI from adolescence through into mid-adulthood. In the 1958 British birth cohort, BMI increased steadily throughout adulthood, and by the age of 45 years mean BMI was nearly 28 kg/m2 in men and 27 kg/m2 in women. Thus over half of all men and women were overweight by age 45 years using the internationally accepted cut-off of 25 kg/m2. As shown previously for this population21 there was considerable tracking, or correlation, of BMI between age 16 and adulthood, indicating a tendency to maintain a relatively high, or low BMI. Results from our study suggest that the gain in BMI over ∼30 years of adult life is influenced by frequency of physical activity in adolescence and the early adulthood period, but there are complexities in the patterns observed.
Our longitudinal analysis therefore adds important information about changes in BMI over time, and in this regard, it is noteworthy that effects of physical activity were seen in relation to rate of change in BMI as well as on BMI at a later age in adulthood (age 33 years, as identified as the intercept in our models). Adolescence is a period during which activity tends to decline,23 and selection effects may be changing.15 A major strength of the 1958 birth cohort is not only that it has a large sample size, but that for this study, both measures of predictor and outcome were available at several ages and over an extended period of time. Therefore, we were able to relate physical activity at more than one time-point to trajectories of BMI, which is relevant given the relatively low tracking of activity and likely changes in selection effects. It is also important to highlight that we have taken account of key confounding factors, in particular parental BMI, own BMI in childhood, childhood socioeconomic position, pubertal stage, and television viewing,24 none of which altered our findings. Consistent with other studies, effects of physical activity on BMI gain were independent of television viewing.15,25–27
We acknowledge that our measures of activity are not direct indicators of physical activity and that misclassification may have occurred. Nevertheless, a recent systematic review found that measures such as access to facilities and programmes, and time spent outdoors, were consistently related to children's physical activity.28 Our activity measures in adulthood provide a measure of frequency, not duration or intensity, but all three aspects of activity have been found to be relevant to body fatness or weight maintenance.29,30 Assuming that misclassification of physical activity in our study is random, the relationships we observed are likely to be attenuated, and under-estimate what is actually occurring.
Frequency of physical activity at 11 years did not influence subsequent BMI or change in BMI through to age 45 years. In females, the most active at age 16 years had a slower gain in BMI than others, and this effect was consistent in direction and of similar magnitude for the BMI trajectory to 45 years commencing at 16 or at 23 years. Thus over the 30 year period from 16 years onwards, the most active females gained 0.6 kg/m2 less than the least active females. This differential gain in BMI partly explains the cross-sectional relationships for this study population, reported elsewhere, whereby more active females had a lower mean BMI than less active females at age 23 years but not at age 16 years.15 In contrast, physically active males at 16 years had faster gains in BMI between 16 and 45 years than their less physically active peers, so that over this 30 year period, the most active gained 0.4 kg/m2 more than the least active. Active 16-year-old males did not, however, have a faster gain in BMI between 23 and 45 years, suggesting that countervailing influences might be operating around adolescence and early adulthood. Although the direction of effect associated with activity at 16 years might be unexpected, it is evident in cross-sectional findings in this study population.15 As we argue elsewhere, the direction of effect (i.e. the faster gain in the most active adolescent males) is likely to be due to a process whereby bigger adolescent boys are more likely to be selected into sports-related activities,16,17 and these bigger boys gain BMI more rapidly during the 16–23 years period.15 Recognizing that later activity might modify the effect of activity at 16 years on the BMI trajectory, we adjusted for adult activity at 23 or 33 years (separately), but this had negligible effect. (Activity at 33 years had an independent influence on the trajectory with more active males at 33 years having a slower gain in BMI between 16 and 45 years.) It is notable, however, that the positive association between adolescent activity and BMI trajectory is very much weakened when considered from age 23 years rather than from 16 years. Furthermore, activity in males in early adulthood is not positively related to the subsequent BMI trajectory, but shows a negative effect, consistent with the direction of effect seen in women, although in men it is non-significant.
Our results from multi-level analyses were consistent with additional analyses on change in activity over time. Males active at 16 and 23 years had a higher mean BMI gain than those who were inactive at both time-points, an opposite pattern to that seen in females. However, looked at over the longer period (16–42 years), in males there was no longer an excess BMI gain in those active at both time-points compared with the inactive. Furthermore, males who decreased their activity between 16 and 42 years were subject to the greatest gains in BMI. Thus, studying patterns of change in activity status provides a valuable supplement to the multi-level analysis. Effects of changing activity may relate to the types of activities that young males participate in. For example, some activities are largely physically active team sports such as football and rugby, that may require commitment, which leads to an ‘all or nothing’ pattern of participation. A decrease from ‘all’ to ‘nothing’ may, therefore, have a relatively large impact on energy expenditure and BMI.
Prospective studies investigating effects of physical activity at baseline on a subsequent change in weight or BMI are limited in number. A recent review identified six studies in adults, with mixed findings.31 In an earlier review of studies in children, we identified four studies of physical activity and change in fatness and, similarly, findings were inconsistent.24 Most of these studies report data on BMI, or some other measure of fatness at two time-points; baseline and follow-up. Few studies have data on body fatness at several time-points.32,33 Moore et al.33 found that higher average levels of physical activity in childhood between 4 and 11 years predicted slower gains in BMI between 4 and 11 years. Van Lenthe et al. 32 reported that change in physical activity between 13 and 27 years was negatively related to change in subscapular skinfold thickness (13–27 years) in females but not males, which is not inconsistent with our findings. Given the complexity in the relationships seen in our study between frequency of activity and BMI, with changes over time and different gender patterns, it is unsurprising that there have been some inconsistencies in the literature, in which different life stages have been investigated with variable periods of follow-up.
In summary, we found that physical activity at 11 years did not influence BMI gain from 16 years (or 23 years) to 45 years. This might be because the behaviour is too distant in time to influence BMI 4 years later and beyond. Alternatively, physical activity may be poorly differentiated in this cohort at this age, or our measure of activity at 11 years may have resulted in misclassification. Thus we cannot exclude the possibility that activity in earlier adolescence has an influence on later BMI gain. Females physically active at 16 years showed a slower gain in BMI between 16 and 45 years and between 23 and 45 years than less active females, suggesting that activity at 16 years may have a prolonged benefit on weight maintenance. Males physically active at 16 years showed faster gains in BMI between 16 and 45 years than less physically active males, but no effect was seen for the BMI trajectory between 23 and 45 years, suggesting that activity at 16 years primarily influences change in BMI over late adolescence and early adulthood, as indicated with measures at 16 and 23 years in our study population. This interpretation is supported by our analyses in which activity over the longer time-period through to mid-adulthood was no longer associated with the highest gain in BMI, whereas a change from active to less active was seen to have adverse effects on BMI gain over this stage of the lifespan.
Contributorship
Tessa J Parsons, Orly Manor, and Chris Power jointly developed the idea for the study and planned the analysis. Tessa J Parsons did the analysis and wrote the first draft of the paper. All the authors contributed to, and approved, the final version. Tessa J Parsons will act as guarantor.
In the 1958 British birth cohort, comprising up to 14 000 individuals in these analyses, a decrease in physical activity between adolescence and mid-adulthood in males, and low levels of activity at both stages in females predicted greater gains in BMI.
Physical activity may reduce the gains in BMI, which occur between adolescence and mid-adulthood.
Relationships varied with age, particularly in males, in whom activity at 16 years appears to predict faster gains in BMI, which is likely to be due to bigger boys being selected into physical activities at this age. Variation in selection effects at different life-stages may partly explain why the literature on physical activity and BMI has shown small and inconsistent effects.
T.J.P. holds a Medical Research Council Special Training Fellowship in Health Services and Health of the Public Research. Research at the Institute of Child Health and Great Ormond Street Hospital for Children NHS Trust benefits from R&D funding received from the NHS Executive. Data obtained from the UK Data Archive, University of Essex, (files: National Child Development Study, SN 3148, SN 4396). Data providers: Centre for Longitudinal Studies, Institute of Education and National Birthday Trust Fund, National Children's Bureau, City University Social Statistics Research Unit (original data producers). Data collection at 45 years was funded by the Medical Research Council, grant G0000934.
Conflict of Interest Statement: We declare we have no conflict of interest.
References
Lobstein T, Baur L, Uauy R. Obesity in children and young people: a crisis in public health.
Power C, Lake JK, Cole TJ. Measurement and long-term health risks of child and adolescent fatness.
Reilly JJ, Methven E, McDowell ZC et al. Health consequences of obesity.
Epstein LH, Valoski AM, Kalarchian MA, McCurley J. Do children lose and maintain weight easier than adults: a comparison of child and parent weight changes from six months to ten years.
Summerbell CD, Ashton V, Campbell K, Edmunds L, Kelly S, Waters E. Interventions for treating obesity in children.
Parsons TJ, Power C, Manor O. Tracking of activity and diet in a large prospective study (1958 British birth cohort).
Telama R, Leskinen E, Yang X. Stability of habitual physical activity and sport participation: a longitudinal tracking study.
Twisk JW, Kemper HC, Van Mechelen W, Post GB. Tracking of risk factors for coronary heart disease over a 14-year period: a comparison between lifestyle and biologic risk factors with data from the Amsterdam Growth and Health Study.
Young and active? A policy symposium on young people and health-enhancing physical activity. June18–20, 1997. London: Health Education Authority,
Social Trends 31. London: The Stationary Office,
DiGuiseppi C, Roberts I, Li L. Influence of changing travel patterns on child death rates from injury: trend analysis [published erratum appears in BMJ 1997 May 10;314(7091):1385].
US Department of Transportation, Federal Highway Administration, Research and Technical Support Center. Nationwide Personal Transportation Survey. Lantham, MD: Federal Highway Administration,
Parsons TJ, Power C, Manor O. Physical activity, television viewing and body mass index: a cross sectional analysis from childhood to adulthood in the 1958 British cohort.
Malina RM. Physical growth and biological maturation of young athletes.
Brewer J, Balsom PD, Davis JA. Seasonal birth distribution amongst European soccer players.
Power C, Elliott J. Cohort profile: 1958 British birth cohort (National Child Development Study).
Lake JK. Body Size in Child and Adulthood: Implications for Adult Health. London: University of London,
Power C, Moynihan C. Social class and changes in weight-for-height between childhood and early adulthood.
Power C, Lake JK, Cole TJ. Body mass index and height from childhood to adulthood in the 1958 British born cohort.
Rasbash J, Browne W, Goldstein H. A User's Guide to MLwiN. 2nd edn. London: Institute of Education,
Health Survey for England 2002. The Health of Young People. London: The Stationary Office,
Parsons TJ, Power C, Logan S, Summerbell CD. Childhood predictors of adult obesity: a systematic review.
Berkey CS, Rockett HR, Field AE et al. Activity, dietary intake, and weight changes in a longitudinal study of preadolescent and adolescent boys and girls.
Salmon J, Bauman A, Crawford D, Timperio A, Owen N. The association between television viewing and overweight among Australian adults participating in varying levels of leisure-time physical activity.
Jakes RW, Khaw K, Day NE et al. Patterns of physical activity and ultrasound attenuation by heel bone among Norfolk cohort of European Prospective Investigation of Cancer (EPIC Norfolk): population based study.
Sallis JF, Prochaska JJ, Taylor WC. A review of correlates of physical activity of children and adolescents.
Saris WH, Blair SN, van Baak MA et al. How much physical activity is enough to prevent unhealthy weight gain? Outcome of the IASO 1st Stock Conference and consensus statement.
Field AE, Manson JE, Taylor CB, Willett WC, Colditz GA. Association of weight change, weight control practices, and weight cycling among women in the Nurses' Health Study II.
Fogelholm M, Kukkonen-Harjula K. Does physical activity prevent weight gain—a systematic review.
Van Lenthe FJ, Van Mechelen W, Kemper HC, Post GB. Behavioral variables and development of a central pattern of body fat from adolescence into adulthood in normal-weight whites: the Amsterdam Growth and Health Study.