Original Article | Published:

Television viewing and obesity: a prospective study in the 1958 British birth cohort

European Journal of Clinical Nutrition volume 62, pages 13551363 (2008) | Download Citation

Contributors: TJP, OM and CP jointly developed the idea for the study, and planned the analysis. TJP did the analysis and wrote the first draft of the paper. All the authors contributed to the final version. All authors declare no conflict of interest.

Abstract

Objective:

To assess whether frequency of television viewing in adolescence (11 and 16 years) or early adulthood (23 years) affected subsequent changes in body mass index (BMI) through to mid-adulthood life, and waist–hip ratio in mid-adulthood.

Subjects:

The 1958 British birth cohort includes all births in 1 week in March 1958 in England, Scotland and Wales. The main analyses included at least 11 301 participants. Outcome measures included BMI at 16, 23, 33 and 45 years and waist–hip ratio at 45 years.

Results:

Watching television ‘often’ at 16 years (but not 11 years) was associated with a faster gain in BMI between 16 and 45 years in males (0.011 kg m−2 per year, 95% confidence interval (CI) 0.003, 0.019) and females (0.013 kg m−2 per year, 95%CI 0.003, 0.023). More frequent television viewing at 11, 16 and 23 years was associated with a faster gain in BMI between 23 and 45 years in females, but not in males. Television viewing at 23 years was associated with waist–hip ratio at 45 years: participants watching 5 times per week had a waist–hip ratio 0.01 higher than those watching less often. At 45 years, those watching television for 4 h day−1 had a waist–hip ratio 0.03–0.04 higher than those watching for <1 h day−1.

Conclusions:

More frequent television viewing in adolescence and early adulthood is associated with greater BMI gains through to mid-adulthood and with central adiposity in mid-life. Television viewing may be a useful behaviour to target in strategies to prevent obesity.

Introduction

Increasingly sedentary lifestyles and declining physical activity are prime suspects among the lifestyle factors contributing to the recent and rapid increase in obesity. Television viewing is the most popular leisure-time pursuit; in 1999, UK children aged 4–15 years watched an average of about 18 h of television per week and adults about 27 h (Church et al., 2001).

Recent reviews have shown that higher levels of television viewing are related to adiposity in children (Marshall et al., 2004) and adults (Foster et al., 2006). Marshall et al. (2004), concluded from a large meta-analysis of mostly cross-sectional studies that although a significant relationship between television viewing and fatness exists in children, it was likely to be too small to be clinically important. However, small effects may be important at both a population and individual level if they accumulate over time. There are fewer longitudinal studies in children, but reports suggest that television viewing may have cumulative effects on body mass index (BMI) (Hancox and Poulton, 2006) and influence subsequent change in fatness (Horn et al., 2001; Berkey et al., 2003; Kaur et al., 2003; Proctor et al., 2003; Hancox et al., 2004; Jago et al., 2005; Viner and Cole, 2005). The large number of people now overweight or obese necessitates prevention strategies, and many argue for an emphasis on childhood and efforts to instil healthy behaviours early in life (Lobstein et al., 2004). Interventions to reduce television viewing in children have met with some success, and suggest that this is a useful target behaviour in mitigating obesity risk (Robinson, 1999).

We have previously investigated in the 1958 British birth cohort, cross-sectional associations between television viewing and BMI in adolescence and early adulthood (Parsons et al., 2005). The aim here was to investigate longitudinal relationships: whether frequency of television viewing in adolescence and early adulthood influences (i) subsequent changes in BMI through to mid-adult life, and (ii) central adiposity in mid-adulthood.

Participants and methods

Study population

All children born in England, Scotland and Wales in 1 week in March 1958 were included in the 1958 birth cohort. From a target population of 17 733 births, information was obtained on 98%. Follow-up sweeps of surviving children were conducted at ages 7, 11, 16, 23, 33, 42 and 45 years (Power and Elliott, 2006). At age 45 years, of 11 971 individuals invited to participate, 9216 provided BMI data. Sample attrition has resulted in a slight under-representation of those who are most disadvantaged (for example, participants entitled to benefits due to financial hardship in their family), but the remaining sample is generally representative of the original sample (Ferri, 1993). For the purposes of our study, we compared mean BMI at 7 and 11 years for two groups: (1) all participants with data (from original data set) and (2) participants in our analyses of BMI trajectories (that is those with at least one BMI measure between 16 and 45 years plus data on television viewing at 16 years). Differences in mean BMI at 7 or 11 years between the two groups were small, 0.04 kg m−2 in males and females. Ethical committee approval for the 45 years survey was obtained from the South-East Multi-Centre Research Ethics Committee, UK. For the 33 years survey, ethical committee approval was not sought, although participants were asked to give written consent for access to medical records.

BMI and central adiposity

BMI (kg m−2) was calculated from heights and weights. At 11 and 16 years, heights (to the nearest inch) and weights (in underclothes, to the nearest pound (0.454 kg)) were measured. 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 nearest millimetre at 45 years. Central adiposity, indexed by waist and hip circumferences, was measured at 45 years. Both circumferences were measured to the nearest millimetre, waist circumference midway between the lower ribs and iliac crest in the mid-axillary line, and hip circumference at the widest part of the body below the waist.

Television viewing

Participants reported television-viewing frequency at 11, 16 and 23 years, and daily duration at 45 years. At 11 and 16 years, categories were ‘often’, ‘sometimes’ or ‘never or hardly ever’. Few children watched television never/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 was undefined at 16 years. At 23 years, frequency categories were as follows: 5, 3–4, 1–2 times per week, 2–3 times in the last 4 weeks, once in the last 4 weeks or not at all in the last 4 weeks. For analyses presented here, we used two categories: 5 and 4 times per week. At 45 years, participants reported how much television they had watched during the last year: >4, 3–4, 2–3, 1–2, <1 h day−1 or none.

Potential confounding factors

The following potential confounding factors were considered.

Maternal BMI: in 1958, maternal height without shoes was measured. Pre-pregnant weight was self-reported by mothers in categories of 1 stone (14 pounds or 6.35 kg), and the midpoint of their weight group was used for calculations of BMI (Lake et al., 1997).

Social class was defined using father's occupation, according to the UK 1951 General Registrar's classification. We used two categories: (i) non-manual (professional, managerial and skilled non-manual), and (ii) manual (skilled, semi-skilled and unskilled manual, and those recorded as ‘no male head of household’) (Power et al., 2003).

Puberty: we used reports of age at menarche (girls) and age of voice breaking (boys) at 16 years.

Physical activity: at 11 years, mothers reported how often the child used parks, recreation grounds, swimming pools and indoor play centres and the child was asked how often (s)he played outdoor sport or took part in sport outside school hours. At 16 years, participants were asked how often they played outdoor and indoor games and sports, and went swimming and dancing. As previously described, for each age, variables were combined into single variable (Parsons et al., 2005) and dichotomized; the most active and others. At 23 years, participants responded to a single question, about frequency of regular sports and exercise (Parsons et al., 2005), categorized here into two groups: most active (active 3 times per week) and others (active 2 times per week).

Alcohol consumption: at 16 years, participants were asked (without parents present) how long since they had consumed an alcoholic drink and if within the past week how much they had consumed. We use three categories: drank >5 weeks ago, 1 U in the last month, 2 U in the last week. At 23 years, participants were asked how frequently they consumed alcoholic drinks, and we use the following categories: most days, 1–2 times per week and less often.

Cigarette smoking: at 16 years, participants were asked (without parents present) how many cigarettes they smoked per week. At 23 years, participants reported whether they had ever, or currently, smoked regularly or not, and were defined as never smokers, ex-smokers, smoker 1–19 cigarettes per day or smoker 20 cigarettes per day (Jefferis et al., 2004b).

Diet: at 33 years, participants reported how often they ate fresh fruit in summer, salads or raw vegetables in winter, chips, fried food (excluding chips), sweets or chocolates, and biscuits, in one of six frequency categories: >once per day, once per day, 3–6 days per week, 1–2 days per week, <1 day per week or never. We constructed a healthy eating score using frequency data for the six foods as described previously (Parsons et al., 2006b), treating fruit and salad as ‘healthy’ foods, and chips, sweets, biscuits and fried food as ‘unhealthy’ foods.

In addition, we conducted several analyses to explore relationships between disability (at 16 years) or longstanding illness (at 23 years), BMI and television viewing: relationships were few and inconsistent and therefore we did not consider disability as a confounding factor.

Statistical analyses

Measures of BMI at different time points within individuals are correlated, and therefore to investigate trajectories of BMI over time, we used multilevel models, which allow for these correlations, fitting BMI as a repeated measure. These models allow for missing outcomes; thus all participants with at least one BMI measure (and data on all independent variables) were included. It is assumed that outcome data are missing at random, that is, the likelihood of a measurement being missing is unrelated to its value. About a third of participants had BMI data at all four time points from 16 years, and a further 29% had data at three time points. There was 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 using MLwiN version 1.10 (Institute of Education, London, UK) (Rasbash et al., 2000).

Analyses were designed so that the timing of the predictor preceded, or was contemporary with, the first BMI outcome measure. We examined the influence of television viewing at 11 or 16 years on the BMI trajectory from childhood to adulthood (16–45 years) and also on adulthood trajectories (23–45 years). Using an adulthood trajectory allowed us to examine television viewing at 23 years as a predictor.

First, change in BMI with age was modelled, using both linear and quadratic age terms. Age was centred at 33 years. Random effects at the individual level were included for the intercept and linear age term, allowing both intercept (BMI at 33 years) and linear slope (change in BMI per year) to vary between individuals. Random effects for the quadratic age term were not significant. Second, to investigate the influence of television viewing on change in BMI, television viewing was added to the model, allowing the intercept (BMI at age 33) to vary by frequency of television viewing. A television viewing by age interaction term was added to test whether the slope of the BMI trajectory varied by television viewing. Adjustment for potential confounding factors (childhood BMI, mother's BMI, social class, puberty, concurrent physical activity, alcohol intake, smoking, adult diet) and their interaction terms by age was examined. We adjusted for changes in lifestyle factors over time by including each factor at two ages (16 and 23 years) in models of the BMI trajectory from 23 to 45 years. We also examined whether any of the lifestyle characteristics modified the influence of television viewing on the BMI trajectory, by adding interaction terms of each characteristic and (i) television viewing and (ii) age by television viewing. Lastly, we investigated whether effects of television viewing in childhood remained after adjusting for levels of adult television viewing. Analyses were repeated after transforming BMI using a natural logarithm; results were similar to those presented here, with one exception, described below.

Relationships between television viewing (at 11, 16, 23 and 45 years) and waist–hip ratio at 45 years were investigated using linear regression. Unadjusted models were repeated using the (smaller) sample available for adjusted models and similar results were obtained. Regression models and other non-multilevel analyses were conducted using SPSS for Windows, version 13.0.1 (SPSS Inc., Chicago, IL, USA).

Results

Continuities in BMI and television viewing

There was considerable tracking of BMI over time points, 16–45 years with Pearson's correlation coefficients ranging from 0.51 to 0.79. BMI increased with age, and was higher in females than males at 16 years, but higher in males from 23 years onwards (Table 1). The distribution for television viewing was similar for males and females at all three ages (Table 2). A greater proportion of 16-year-olds from manual backgrounds watched television ‘often’ than adolescents from non-manual backgrounds, and several other potential confounding factors showed weak associations with television viewing at 16 years (Table 3). Some continuity of television viewing frequency between successive ages was evident (Table 4); for example, of those who watched television ‘often’ at 11 years, approximately two-thirds of boys and girls also watched ‘often’ at 16 years, while of those who watched television ‘sometimes’ at 11 years, just under half watched ‘sometimes’ at 16 years. Similar associations were seen between 16 and 23 years and between 23 and 45 years.

Table 1: Mean BMI and waist–hip ratio by gender
Table 2: Distribution of television viewing by age
Table 3: Population characteristics
Table 4: Frequency of television viewing: continuities between successive ages from 11 to 45 years

Television viewing at 11, 16 and 23 years and the BMI trajectory to 45 years

With respect to our objective, to investigate whether frequency of television viewing in adolescence and early adulthood influences subsequent BMI, two components from the multilevel models are relevant. The association between television viewing frequency and BMI is represented first by the level of BMI (at 33 years) and, second, by change in BMI (expressed as kg m−2 per year) over the period to 45 years. These models indicate that among males, television viewing at 11 years had no effect on BMI at 33 years (model intercept) or on BMI gain either between 16 and 45 years or between 23 and 45 years (Table 5). Females watching television ‘often’ at 11 years had a higher BMI at 33 years, and also experienced a faster gain in BMI between 23 and 45 years, by 0.017 kg m−2 per year. No effect was seen on the BMI trajectory (16–45 years). Television viewing at 16 years also showed a positive effect on BMI at 33 years in females but not males. Both males and females who watched television ‘often’ at 16 years had faster gains in BMI at 16–45 years (by 0.011 and 0.013 kg m−2 per year, respectively), as did females between 23 and 45 years (by 0.012 kg m−2 per year) (Table 5). Television viewing at 23 years had a positive effect on BMI at 33 years in both sexes, and on BMI gain at 23–45 years in females; those watching 5 times per week gained BMI faster, by 0.029 kg m−2 per year than those watching less often (Table 5). In males, the effect of television viewing at 23 years on BMI gain disappeared after transforming BMI using a natural logarithm.

Table 5: Effect of television viewing on BMI trajectory to age 45 years

Estimated effects of television viewing at 16 years on BMI at 33 years or the slope of the BMI trajectory were little or modestly affected by adjusting for earlier BMI, maternal BMI and social class, concurrent physical activity, smoking or alcohol intake, or for adult healthy eating score. For example, adjustment for physical activity at 16 years slightly reduced the effect of television viewing (16 years) on the BMI trajectory (16–45 years), from 0.013 to 0.011 kg m−2 per year in females (no change in effect in males). Frequency of alcohol consumption at 23 years slightly reduced the effect of television viewing (23 years) on the BMI trajectory (23–45 years), from 0.010 to 0.008 kg m−2 per year in males and 0.029–0.025 kg m−2 per year in females. Similarly, adjusting for lifestyle factors at two ages to allow for change over time had modest or negligible effect on the relationship between television viewing and the BMI trajectory. In females, adjusting for television viewing at 23 years reduced the effect of television viewing at 16 years on the 23–45 years trajectory (from 0.012 to 0.006 kg m−2 per year), but not the effect of television viewing at 11 years (Table 5). Interaction terms between television viewing and other lifestyle characteristics, concurrent physical activity, smoking, alcohol intake or adult healthy eating score, were not significant, with one exception. In men, smoking at 16 years modified the effect of television viewing on the BMI trajectory between 16 and 45 years, such that in non-smokers those who watched television ‘often’ experienced faster gains in BMI (by 0.017 kg m−2 per year), whereas in smokers, the trajectory was similar irrespective of the level of television viewing.

BMI and waist–hip ratio at 45 years and television viewing

There was no effect of adolescent television viewing (11 or 16 years) on adult waist–hip ratio (at 45 years), but television viewing at 23 years was associated with waist -hip ratio at 45 years; participants watching television 5 times per week had a waist–hip ratio 0.01 higher than those watching less often (Table 6). This association persisted, although attenuated, after allowing for concurrent BMI at 45 years. Subsequently, at 45 years, men watching television >4 h day−1 had a higher mean BMI by 2.0 kg m−2 and higher waist–hip ratio by 0.04 than men who watched television <1 h day−1; corresponding values for women were 3.4 kg m−2 and 0.035. The relationship between BMI and television was stronger in women, as indicated by a significant BMI by television by gender interaction term (P=0.001), whereas the relationship between waist–hip ratio and television was stronger in men (P=0.05).

Table 6: Relationships between television viewing (at 11, 16, 23 and 45 years) and waist–hip ratio at 45 years

Discussion

Our study suggests that gain in BMI over nearly 30 years of adult life is influenced by frequency of television viewing in adolescence and early adulthood, more consistently in females than males. A generally consistent effect of television viewing at age 16 years was found on change in BMI over time, with excess gains of between 0.011 and 0.013 kg m−2 per year for those watching the TV most frequently. Thus, over the 30-year period those watching television more frequently gained about 0.35 kg m−2 more than those who watched less frequently. Importantly, these effects were little affected by allowing for BMI in childhood and other relevant confounding factors. At 23 years, women watching television more than five times per week showed a faster BMI gain of 0.029 kg m−2 per year, equating to an increase of 0.64 kg m−2 between 23 and 45 years, but no effect was evident in men. Television viewing at 11 years showed no effect on BMI gain in males, while females who watched television often at 11 years experienced faster BMI gains from 16 or 23 years to 45 years. Central adiposity at 45 years was increased (waist–hip ratio 0.01 higher), among those watching television 5 times per week more than 20 years previously at age 23 compared with those watching less often. This association was not entirely due to continuities in television viewing through to mid-life.

Prospective studies investigating the effects of television viewing on subsequent weight or BMI, particularly change in weight or BMI, are relatively limited in number (Berkey et al., 2000, 2003; Horn et al., 2001; Kaur et al., 2003; Proctor et al., 2003; Hancox et al., 2004; Jago et al., 2005; Viner and Cole, 2005; Hancox and Poulton, 2006), and few have more than 5 years of follow-up (Proctor et al., 2003; Hancox et al., 2004; Viner and Cole, 2005; Hancox and Poulton, 2006). Overall, these studies suggest that higher levels or increases over time in television viewing result in larger increases in body fatness over time. These effects may not be evident in very young children (3–6 years), and possibly emerge around 6–7 years (Jago et al., 2005), and interestingly, several studies report stronger relationships in females (Horn et al., 2001; Berkey et al., 2003; Hancox and Poulton, 2006). Our study has several repeat measures of BMI over an extensive period, during a life stage in which BMI is typically increasing, and it therefore adds important information about changes in BMI through to mid-adulthood. Information on central adiposity at 45 years is a further strength to the study. We acknowledge that our measure of television viewing in adolescence provides information about frequency, not duration, and it is therefore likely that our results underestimate the size of effect. In addition, the measure of television viewing changed over time. Although some suggest that relationships between body fatness and television viewing reported in the literature are generally weak and perhaps unimportant (Marshall et al., 2004), others counter-argue that television viewing is at least comparable to, and may be a better predictor of body fatness, than other frequently considered causal predictors, such as dietary intake or physical activity (Berkey et al., 2000; Hancox and Poulton, 2006). Our analyses of physical activity and television viewing at 23 years reported here and elsewhere (Parsons et al., 2006a) suggest effects of similar magnitude on BMI through to 45 years.

We have also been able to take into account important confounding factors, notably parental BMI, own BMI in childhood, childhood socioeconomic position, physical activity (Parsons et al., 1999), smoking, alcohol intake and adulthood healthy eating. Effects of television viewing were largely unaffected by adjusting for these factors, which was unsurprising given that they were generally only weakly related to television viewing. Although there are changes in physical activity, smoking, alcohol intake and diet over time, described in detail elsewhere (Jefferis et al., 2003, 2004a, 2005; Parsons et al., 2006b), effects of television viewing on the BMI trajectory were little affected by adjusting for these changes, which may also be due to the weak associations between these factors and television viewing. Also there were no interactions between lifestyle factors (activity, smoking, alcohol and adult healthy eating) present, with one exception. In non-smoking men at 16 years, those who watched television ‘often’ experienced faster gains in BMI (16–45 years; by 0.017 kg m−2 per year) than those who watched less often, whereas in smokers, the trajectory was similar irrespective of the level of television viewing. The reason for this difference is unclear; the proportion watching television often and less often in smokers and non-smokers was almost identical, and BMI at 16 years was similar.

We were able to take into account television viewing in adulthood and found that in females, adjusting adolescent television viewing at 16 years for television viewing at 23 years abolished the effect of television viewing at 16 years on BMI gain, suggesting that continuities in television viewing from adolescence to adulthood may be responsible for this effect. However, effects of television viewing at 11 years appeared to be independent of television viewing in adulthood.

Waist–hip ratio has been shown to be positively associated with BMI, but is also a stronger and often independent predictor for cardiovascular disease mortality (Welborn et al., 2003) and associated risk factors, diabetes and certain cancers (World Health Organization, 1998). We found that television viewing in adolescence had little influence on waist–hip ratio at 45 years, but more frequent viewing behaviour in early adulthood (5 times per week at 23 years) was associated with an increase in waist–hip ratio of 0.01, an effect that persisted after allowing for television viewing over 20 years later. By 45 years, those watching television for >4 h day−1 had a mean waist–hip ratio of about 0.04 higher than those watching television for less than 1 h day−1. This difference has been found to correspond to an increase in blood pressure (2–4 mm Hg systolic or 1–3 mm Hg diastolic) (Canoy et al., 2004). An increase of 1 s.d. in waist–hip ratio (0.06) has been associated with a 1.5–2.0 increase in odds of developing type II diabetes (Snijder et al., 2003).

How reduced television viewing lowers the risk of obesity or central adiposity, and whether it might be acting via increased activity, or changes in dietary intake, is uncertain. Cross-sectional studies generally show weak or no association between television viewing and physical activity (Kronenberg et al., 2000), particularly in children (Parsons et al., 2005), although associations are possibly more consistent and stronger in adults (Salmon et al., 2000). Irrespective of an association, both television viewing and physical activity are generally found to have independent effects on obesity (Berkey et al., 2000; Salmon et al., 2000; Jakes et al., 2003; Parsons et al., 2005). Our results are consistent with this general pattern; effects of television viewing on the rate of BMI gain were independent of physical activity.

A second possible explanation is that effects of television are mediated through dietary habits. In children, television viewing has been found to be positively related to snacking between meals (Marshall et al., 2004) and consumption of energy-dense snacks (Phillips et al., 2004), although whether snacking habits or energy density is related to adiposity is less clear (Guillaume et al., 1998; Phillips et al., 2004; Bell et al., 2005). While we lack dietary data concurrent with our measures of television viewing, the relationship between television viewing and BMI gain was not influenced by a healthy eating score in adulthood. We cannot however exclude the possibility that television viewing affects BMI gain or waist–hip ratio through dietary habits. Television viewing is a complex exposure, and further work is needed to determine how it affects obesity risk.

Our findings suggest that watching television more frequently in adolescence or early adulthood is related to a faster BMI gain through to mid-adult life, particularly in females, and that more frequent television viewing in early adult life increases waist–hip ratio some years later. At 45 years, the cross-sectional relationship between television viewing and BMI mirrored the longitudinal relationship, being stronger in women than in men. In contrast, the relationship between waist–hip ratio and television viewing at 45 years was stronger in men. The reasons for gender differences, in our study and elsewhere in the literature, are unclear; we found that continuities in television viewing between one age and the next were similar for males and females. Possible gender differences in the underlying mechanisms by which television influences adiposity, for example in dietary or compensatory behaviours, might explain the differences in the relationship and require further exploration. Television viewing is one factor among several that impact on subsequent body fatness and weight gain (Parsons et al., 1999), and as such, reducing television viewing may be a useful component in strategies for limiting BMI gain and increases in central adiposity.

References

  1. , , , (2005). Contribution of ‘noncore’ foods and beverages to the energy intake and weight status of Australian children. Eur J Clin Nutr 59, 639–645.

  2. , , , , , et al. (2000). Activity, dietary intake, and weight changes in a longitudinal study of preadolescent and adolescent boys and girls. Pediatrics 105, E56.

  3. , , , (2003). One-year changes in activity and in inactivity among 10- to 15-year-old boys and girls: relationship to change in body mass index. Pediatrics 111, 836–843.

  4. , , , , , et al. (2004). Fat distribution, body mass index and blood pressure in 22 090 men and women in the Norfolk cohort of the European Prospective Investigation into Cancer and Nutrition (EPIC-Norfolk) study. J Hypertens 22, 2067–2074.

  5. , , , , , et al. (2001). Social Trends 31. The Stationary Office: London .

  6. (1993). Life at 33: the Fifth Follow-Up of the National Child Development Study. National Children's Bureau: London.

  7. , , (2006). Altering TV viewing habits: an unexplored strategy for adult obesity intervention? Am J Health Behav 30, 3–14.

  8. , , (1998). Obesity and nutrition in children. The Belgian Luxembourg Child Study IV. Eur J Clin Nutr 52, 323–328.

  9. , , (2004). Association between child and adolescent television viewing and adult health: a longitudinal birth cohort study. Lancet 364, 257–262.

  10. , (2006). Watching television is associated with childhood obesity: but is it clinically important? Int J Obes 30, 171–175.

  11. , , , , (2001). Correlates and predictors of adiposity among Mohawk children. Prev Med 33, 274–281.

  12. , , , , (2005). BMI from 3–6 years of age is predicted by TV viewing and physical activity, not diet. Int J Obes 29, 557–564.

  13. , , , , , et al. (2003). Television viewing and low participation in vigorous recreation are independently associated with obesity and markers of cardiovascular disease risk: EPIC-Norfolk population-based study. Eur J Clin Nutr 57, 1089–1096.

  14. , , , (2003). Cigarette consumption and socio-economic circumstances in adolescence as predictors of adult smoking. Addiction 98, 1765–1772.

  15. , , , (2004a). Changing social gradients in cigarette smoking and cessation over two decades of adult follow-up in a British birth cohort. J Public Health (Oxford) 26, 13–18.

  16. , , , (2004b). Effects of childhood socioeconomic circumstances on persistent smoking. Am J Public Health 94, 279–285.

  17. , , (2005). Adolescent drinking level and adult binge drinking in a national birth cohort. Addiction 100, 543–549.

  18. , , , (2003). Duration of television watching is associated with increased body mass index. J Pediatr 143, 506–511.

  19. , , , , , et al. (2000). Influence of leisure time physical activity and television watching on atherosclerosis risk factors in the NHLBI Family Heart Study. Atherosclerosis 153, 433–443.

  20. , , (1997). Child to adult body mass index in the 1958 British birth cohort: associations with parental obesity. Arch Dis Child 77, 376–381.

  21. , , (2004). Obesity in children and young people: a crisis in public health. Obes Rev 5 (Suppl 1), 4–104.

  22. , , , , (2004). Relationships between media use, body fatness and physical activity in children and youth: a meta-analysis. Int J Obes 28, 1238–1246.

  23. , , (2006a). Physical activity and change in body mass index from adolescence to mid-adulthood in the 1958 British cohort. Int J Epidemiol 35, 197–204.

  24. , , , (1999). Childhood predictors of adult obesity: a systematic review. Int J Obes 23 (Suppl 8), S1–S107.

  25. , , (2005). Physical activity, television viewing and body mass index: a cross sectional analysis from childhood to adulthood in the 1958 British cohort. Int J Obes 29, 1212–1221.

  26. , , (2006b). Longitudinal physical activity and diet patterns in the 1958 British Birth Cohort. Med Sci Sport Exerc 38, 547–554.

  27. , , , , , et al. (2004). Energy-dense snack food intake in adolescence: longitudinal relationship to weight and fatness. Obes Res 12, 461–472.

  28. , (2006). Cohort profile: 1958 British birth cohort (National Child Development Study). Int J Epidemiol 35, 34–41.

  29. , , (2003). Child to adult socioeconomic conditions and obesity in a national cohort. Int J Obes 27, 1081–1086.

  30. , , , , , et al. (2003). Television viewing and change in body fat from preschool to early adolescence: The Framingham Children's Study. Int J Obes 27, 827–833.

  31. , , (2000). A User's Guide to MlwiN.2. Institute of Education: London.

  32. (1999). Reducing children's television viewing to prevent obesity: a randomized controlled trial. JAMA 282, 1561–1567.

  33. , , , , (2000). The association between television viewing and overweight among Australian adults participating in varying levels of leisure-time physical activity. Int J Obes 24, 600–606.

  34. , , , , , et al. (2003). Associations of hip and thigh circumferences independent of waist circumference with the incidence of type 2 diabetes: the Hoorn Study. Am J Clin Nutr 77, 1192–1197.

  35. , (2005). Television viewing in early childhood predicts adult body mass index. J Pediatr 147, 429–435.

  36. , , (2003). Waist–hip ratio is the dominant risk factor predicting cardiovascular death in Australia. Med J Aust 179, 580–585.

  37. World Health Organization (1998). Obesity: Preventing and Managing the Global Epidemic. Report of a WHO consultation, Geneva, 3–5 June 1997, WHO: Geneva.

Download references

Acknowledgements

Data were obtained from the UK Data Archive, University of Essex (files: National Child Development Study, SN 3148, SN 4396). The following are the 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. 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. Medical Research Council, UK (Special training fellowship held by T Parsons).

Author information

Affiliations

  1. Department of Paediatric Epidemiology and Biostatistics, Institute of Child Health, University College London, London, UK

    • T J Parsons
    •  & C Power
  2. School of Public Health and Community Medicine, Hebrew University, Jerusalem, Israel

    • O Manor

Authors

  1. Search for T J Parsons in:

  2. Search for O Manor in:

  3. Search for C Power in:

Corresponding author

Correspondence to T J Parsons.

About this article

Publication history

Received

Revised

Accepted

Published

DOI

https://doi.org/10.1038/sj.ejcn.1602884

Further reading