OBJECTIVE: To analyse the association of time watching television (TV) and physical activity with obesity in the Mediterranean area of Spain with the highest prevalence of obesity.
DESIGN: Cross-sectional study.
SETTING: Valencia Region in Spain.
PARTICIPANTS: A representative sample of 814 men and 958 women, aged 15 y and older, participating in a Health and Nutrition Survey conducted in 1994.
MEASUREMENTS: Height and weight were directly measured during home interviews. The outcome measure was obesity, defined as a body mass index ≥30 kg/m2. Covariates were self-reported hours of TV viewing, physical activity habits, sleeping duration, age, gender, educational level, smoking and marital status. Prevalence odds ratios (POR) estimated by logistic regression were used as effect measures.
RESULTS: Obese people reported to spend more time watching TV (mean±s.d.: 3.6±1.5 h/day) than non-obese ones (3.0±1.4 h/day), and less sleeping time. In multivariate analysis, obesity was associated with TV viewing, sleeping time and physical activity at work. People watching TV ≥4 h/day showed a higher adjusted prevalence odds ratio of obesity, POR=2.38 (95% confidence interval, 1.54–3. 69), compared with those watching TV ≤1 h/day. People who reported to sleep ≥9 h/day presented a lower POR of obesity than those sleeping ≤6 h/day, POR=0.43 (0.27–0.67). Statistically significant dose–responses were observed for both associations, so that the prevalence odds ratio of obesity was 30% higher for each hour of increased TV viewing and 24% lower for each additional hour of sleeping time. In addition, the prevalence of obesity was lowest among single people, those more physically active at work, and those with a high educational level.
CONCLUSION: Time spent watching television and a low physical activity at work were related to obesity in adults. The inverse association between obesity and sleep duration deserves further research.
Obesity has been associated with increased mortality and mortality from cardiovascular diseases, cancer, diabetes mellitus and overall mortality,1,2 and its prevalence seems to be rising in the US and most European countries.3,4,5 It has been pointed out that this increasing prevalence may be mostly attributed to reduction of physical activity,6 and that major public health efforts are needed to identify those factors related to weight gain with age in order to implement effective public health interventions.7,8
Television (TV) viewing represents a major sedentary behaviour which is very extensive among people of industrialized countries, and that has been associated with obesity in cross-sectional and prospective cohort studies,9,10,11,12,13 mostly in young US populations. In Spain, the prevalence of obesity may be considered high among adults with some evidence that it is increasing, at least among the lower socioeconomic status groups.14,15 The Region of Valencia, with close to 4 million inhabitants, has the highest prevalence of obesity in Spain: 17.1% for people aged 15 y and older.15,16
The aim of the study was to identify whether time spent watching television and physical activity were independently associated with adult obesity in this Mediterranean population, exploring the influence of other modifiable risk factors as well.
Data for this study came from the Health and Nutritional Survey of Valencia (HNSV) conducted in 1994. Survey methods have been described in detail elsewhere.16 A representative sample of 2439 residents aged 15 y and upwards (1159 men and 1280 women) was selected using a two-stage stratified sampling procedure that included in the first stage 28 municipalities randomly selected with probability proportional to the population size. In the second stage, individuals were randomly selected by age (15–24, 25–34, 35–49, 50–64, and 65 y old and over) and sex quota. The overall participation rate was 74.4%. Non-response rates for the variables of interest were small, and below 2.3% in all cases. Final analyses were performed on the basis of 814 men and 958 women with complete information. When necessary, non-response (missing data) for physical activity variables were included in multivariate analyses by adding an extra category for missing values.
Information was collected in household visits using a questionnaire that included information on anthropometry, physical activity, TV viewing, educational level, marital status and smoking status. Weight was measured in kilograms using electronic weight scales with direct digital reading (Tefal®, Topline model, precision±100 g). Height was measured in centimetres using a flexible unextendible measurement tape with the person standing and without shoes. Two weight and height measurements were taken and recorded in all cases, and a third one performed if discrepancies existed between the two prior ones. Obesity was defined as a body mass index (BMI) equal to or greater than 30 kg/m2.
People were asked to classify their physical activity at work according to four categories: (a) sitting most of the day (sedentary); (b) standing most of the day but little motion (moderately active): (c) walking or carrying light weights (active); (d) work requiring intense physical activity (very active). Similarly, four categories of physical activity at leisure time were used: (a) sitting most of the time (sedentary); (b) walking, cycling, gardening (moderately active); (c) heavier activities—jogging, cycling, tennis (active); (d) athletics or regular sports training (very active). Due to the small number of individuals, categories (c) and (d) were collapsed into one ‘active’ category. An additional question that has been shown to discriminate physically active people was also included.17
Time spent watching television (TV) was estimated by asking the person: ‘How many hours per week do you usually spend watching TV?’ The number of hours per day (h/day) estimated from that response was categorized as: ≤1 h/day; 2 h/day; 3 h/day and ≥4 h/day. Hours slept per day, including afternoon naps, were also grouped into four categories (≤6 h/day, 7 h/day, 8 h/day, and ≥9 h/day). People were also classified into three categories by smoking status: never, former and current smoker.
To permit generalizability to the adult Valencian population, prevalence estimates of obesity and average body mass index (BMI) presented in Table 1 were calculated using the weights corresponding to the survey design. To explore the association between obesity and the independent variables (eg TV watching), prevalence odds ratios (POR) were calculated by unconditional logistic regression analysis. Thus, PORs were interpreted as relative risks. Indicator (factored) variables for each category of the independent variables were automatically created and the lowest category was used as the reference category. PORs with 95% confidence intervals (95% CI) were estimated for the variables included in Tables 1 and 2 using the statistical program STATA.18 Tests for trend were performed for each ordinal variable after unfactorizing and adding it to a previous model, including potential confounders. The likelihood ratio statistic was used to evaluate the significance of linear trends.
The overall estimated prevalence of obesity was 16.0% for men and 18.1% for women. The prevalence of obesity (BMI≥30), the distribution of average BMI and the prevalence rate ratios (PORs), adjusted by study design variables are shown in Table 1. Obese people reported a higher average time of TV watching (mean±standard deviation: 3.6±1.5 h/day) than non-obese ones (3.0±1.4 h/day). People in the highest category of TV watching (≥4 h/day) showed higher POR of obesity than those in the lowest category (≤1 h/day), POR=2.36 (95% confidence interval, CI: 1.55–3.60). People who slept nine or more hours per day presented a lower POR of obesity than people sleeping six hours of less per day, POR=0.47 (95% CI=0.30–0.73). Obesity was also associated with marital status, educational level but not with smoking status or other physical activity variables, although some evidence for a protective effect of physical activity at work was present (Table 1). Married and widowed persons presented higher PORs than single ones, and those who had not completed primary school had higher risk of obesity than those with higher educational attainment.
To determine if the association between watching TV and obesity was independent of other variables used in the univariate analyses, we fitted a multivariate model including all variables presented in Table 1. After controlling for the effect of all variables, people viewing TV at least 4 h per day had higher risk of obesity than those viewing 1 h or less, POR=2.38 (95% CI=1.54–3.69). An increasing and statistically significant trend across prevalence odds ratios was observed, so that adjusted PORs of obesity were 30% higher for each additional hour spent viewing TV per day, POR=1.30 (95% CI: 1.14–1.48), (Table 2).
Sleep duration showed an inverse and statistically significant association with obesity. People who slept nine or more hours per day had less than half the risk of obesity than those sleeping only 6 h or less per day, POR=0.43 (95% CI: 0.27–0.67). A significant monotonic descendent trend was also observed across PORs for this association. The prevalence odds ratio of obesity was 24% lower for each additional hour spent sleeping per day, POR=0.76 (85% CI: 0.66–0.87). The magnitude of associations of marital status and educational level with obesity persisted in the multivariate analysis. The protective effect of physical activity at work became more obvious since the POR of obesity was significantly lower among active people than among sedentary ones, POR=0.54 (95% CI: 0.33–0.88), with evidence of a significant decreasing trend across prevalence ratios. As in the univariate analyses, no significant associations were observed for the remaining variables.
Formal tests for interaction were performed between TV viewing and other variables associated with obesity, but no statistically significant results were observed. Finally, the analyses were repeated including overweight (BMI≥25 kg/m2) as the dependent variable with similar results (data not shown).
The present study showed a positive and independent association between the number of hours spent watching television and the risk of being obese among adults living in the Mediterranean Region of Valencia. In addition, physical activity at work and sleep duration showed an inverse association with the risk of obesity in our study.
The magnitude of the association found between TV watching and obesity is consistent with that found in cross-sectional and prospective studies carried out in different social and cultural settings,9,10,11,19 although lower effects were found in Mexican children 9–16 y old,20 12% increase per hour of TV watching, and larger effects in US youths 10–15 y of age (odds ratio, 5.3), comparing people watching TV more than 5 h per day with those watching less than 2 h.21 It has been suggested that TV viewing may be a good predictor of sedentarism or inactivity. Moreover, that it may actually play a direct role in the obesity epidemic either by the low physical demands of this recreational activity or by inducing the adoption of specific dietary habits.22,23 TV watching may replace other physical activities with higher energy demands, and significantly reduce total energy expenditure increasing net energy balance.24,25,26
One potential caveat of our study is that its cross-sectional design does not allow firm causal judgements. However, the interpretation of the associations of TV with obesity is in concordance with most cross-sectional and prospective studies that have explored the association between TV viewing and obesity.21 An additional point to consider is that the questions used to collect information on physical activity may not have been able to fully characterise total energy expenditure, a likely predictor on the risk of obesity.
We found a moderate inverse association between obesity and physical activity at work, but not with other physical activity variables that have been reported in other studies.27,28 Two explanations could be considered. Firstly, television viewing may truly be a better independent indicator of sedentarism than other variables used to describe physical activity, particularly during leisure time. Secondly, physical activity at work and leisure time may counteract each other. In fact, most people classified as active at work reported being sedentary in their leisure time (61%), and not to practice any sports (68%). Therefore, we feel that our study would support an inverse association between physical activity at work and obesity, but it should not allow to draw definitive conclusions on the association between overall physical activity and obesity in this population.
The unexpected inverse association between sleep duration and obesity was independent of other effects and deserves some comment. As far as we know, only one study has previously shown a similar association although in French children.29 In this study, 5 y old children who slept less than 11 h per day had higher odds ratio of obesity than those sleeping more, OR=1.4 (95% CI: 1.1–1.9), after controlling for the effect of TV watching and other variables, and this association showed a statistically significant trend across categories of sleeping hours per day (P<0.001).
An initial explanation of the association between obesity and sleep might be possible confounding. Physical activity or psychological factors are two potential confounders. It might be argued that obese people, usually less active, sleep less because they might need less time to recover, or alternatively, that obese people tend to stay up late watching TV which, in turn, might leave them too tired to exercise the next day. These two alternatives would not be supported by our data since the association between sleeping time and obesity was adjusted for physical activity and time spent watching television. Obesity has been associated with nervousness, anxiety, and other psychological problems also related to sleeping difficulties,30,31 which deserves further consideration.
An additional possibility to consider is that obese people may sleep less because they sleep worse. Survey data from the US and UK have shown a high prevalence of sleeping disorders in adults, and a clear association between obesity and obstructive sleep apnoea (OSA) has also been found.32,33,34 Our findings would be compatible with an increasing prevalence and severity of OSA episodes or other sleeping disorders with increasing BMI. Finally, a further alternative is that shorter duration of sleep results from hormonal and neuroendocrine alterations related to obesity.31
In our study duration of sleep was directly reported during the interview, not measured, and information bias should always be considered. Differential reporting of sleeping time might have occurred by the obese if they tended to systematically underestimate the number of hours of sleep due to fatigue or sleepiness when awake. Although unlikely, that is certainly an alternative explanation that we cannot rule out in this study.
In conclusion, this study has shown that television viewing is a very prevalent practice in this Mediterranean Region of Spain, and may be an independent and more important predictor of obesity than other physical activities in adults. Since combined interventions may be efficacious in reducing TV watching and decreasing obesity among the young and adults,28,35 feasibility studies exploring the implementation of public health interventions might be helpful in finding ways to stop the increasing epidemic of obesity and its consequences in the new millennium. The association between obesity and sleeping time, for the second time reported in the literature, deserves further research.
This study was supported by a grant from the Dirección General de Salud Pública, Generalitat Valenciana, 1994 and the Fondo Investigacion Sanitaria (FIS 00/0985). The authors are grateful to Dr Matthew P Longnecker for his helpful comments.