Overweight and obesity are important risk factors for several diseases including CVD and cancer(Reference Kopelman1, Reference Adami and Trichopoulos2). Since the latency of many chronic diseases is long and several lifestyle patterns are established in childhood and adolescence, the prevalence of overweight and obesity and their determinants during these stages of life are important public health issues. Several studies have documented that the prevalence of obesity in childhood and adolescence in several European countries is high and increasing(Reference Flodmark, Lissau, Moreno and Pietrobelli3, Reference Kosti and Panagiotakos4) and the problem is particularly evident in Greece(Reference Mamalakis, Kafatos, Manios, Anagnostopoulou and Apostolaki5–Reference Papadimitriou, Kounadi, Konstantinidou, Xepapadaki and Nicolaidou8). We report here the results of a cross-sectional study on correlates of overweight undertaken among pre-adolescent primary-school children in the capital city of Greece, Athens.
Methods
In January 2003, an interviewer-administered questionnaire was developed to ascertain sociodemographic profile, dietary patterns and physical activity of 5th and 6th grade primary-school pupils in Athens, Greece. Under the guidance of the lead author of this study, the questionnaire was discussed among the five students of the Health Visiting Unit of Athens Technological Educational Institute who would act as the interviewers. Then, each of the interviewers, in turn, submitted the questionnaire to his/her remaining four colleagues, in order to achieve a high degree of clarity and repeatability. Between March 2003 and June 2003, the questionnaire was distributed by the five trained interviewers to eleven primary schools dispersed in the Greater Athens area. The schools were chosen to represent the spectrum of schools in the Athens area. Although a small but not recorded number of pupils were not present at the visit of the interviewers, virtually all present volunteered to fill in the questionnaire and only six questionnaires were judged incomplete. Height and weight were measured at the same time as the interviews, with sets of five SECA 220 stadiometers and five SECA Bellissima 841 scales, respectively. All instruments were calibrated, regularly checked and adjusted when needed. Measurements were carried out by the trained investigators while the pupils were lightly dressed and without shoes and their BMI was calculated in kg/m2. Subsequently, growth charts developed for Greek children by the First Department of Pediatrics of Athens University Medical School were used in order to identify who among the examined children had a BMI equal to or above the 85th centile in the corresponding age and gender BMI curve(Reference Hiotis, Tsiftis, Hatzisimeon, Krikos, Maniati and Dakou-Voutetaki9).
The questionnaire included entries concerning year and place of birth, birth order, paternal education (an indicator of socio-economic status) and a detailed FFQ that included 25 food item categories generally consumed by pre-adolescent boys and girls in Greece. The questionnaire was comprehensive enough to allow calculation of energy intake as well as protein, carbohydrate and lipid intake on the basis of food composition tables for Greece(Reference Trichopoulou10). The FFQ was based on the one used in a series of case–control studies conducted in Greece by the Department of Hygiene and Epidemiology of the University of Athens’ Medical School. This questionnaire has not been formally validated but the results of these studies have been in line with those of major nutritional epidemiology investigations previously or subsequently undertaken(Reference Trichopoulos, Tzonou, Katsouyianni and Trichopoulou11, Reference Trichopoulou, Toupadaki, Tzonou, Katsouyianni, Manousos, Kada and Trichopoulos12). The questionnaire also had entries allowing the calculation of hours per day of vigorous physical activity (sports, bicycling, swimming, gymnastics) as well as hours of essential inactivity excluding sleep (watching television or working/playing with the computer)(Reference Lagerros and Lagiou13). No formal validation of this questionnaire was undertaken.
For the statistical analysis, the study subjects were cross-classified by gender, overweight (≥85th or <85th centile) and the indicated principal study variables. Subsequently, the data were modelled through multiple logistic regression using as outcome overweight (≥85th centile) v. else (<85th centile). All covariates indicated in the tables were mutually adjusted for, except for carbohydrates, in order to avoid collinearity, since total energy intake as well as energy from protein and from total fat were included in the model. The SPSS 12.0 Statistical Software Package (SPSS Inc., Chicago, IL, USA) was used.
Results
Complete questionnaires were available for a total of 633 pupils (316 boys, 317 girls). Table 1 compares the distributions of normal weight (<85th centile) and overweight (≥85th centile) boys and girls by age, place of birth, birth order, marital status of parents, paternal education (the most important indicator of socio-economic status in Greece), number of eating occasions, frequency of consumption of refreshments (including juices and sodas), conditions surrounding eating (eating under parental pressure), hours of physical activity per day (gymnastics, group sports, bicycling, swimming, etc.) and hours of inactivity per day except sleep (time spent on television and/or computer). Daily intake (median, 25 and 75 centiles) of energy overall and of energy-generating nutrients are also shown.
Table 1 Sociodemographic, lifestyle and diet characteristics of 633 primary-school students in Athens, GreeceFootnote *
* Numbers and percentages; for one boy, height was not available and BMI could not be calculated.
The data in Table 1 serve only descriptive purposes because they do not accommodate possible mutual confounding. Nevertheless, there is evidence that overweight is more common among girls than among boys, among children of parents who do not live together and among children with patent inactivity, whereas overweight is less common among children born outside Greece and among children who distribute their energy intake among many, rather than few, eating occasions. The energy intake data in this table suggest that there are no strong associations between energy intake, overall or from specific macronutrients, on the one hand, and overweight on the other, among either boys or girls. About 15 % of energy intake was derived from proteins, about 35 % from carbohydrates and about 50 % from lipids.
Table 2 shows mutually adjusted odds ratios and 95 % confidence intervals for overweight v. normal weight by specified contrasts or increments of the indicated predictor variables. The probability of overweight is significantly higher among girls than among boys and significantly lower among foreign-born children. Children of separated parents are significantly and considerably more likely to be overweight, although the numbers are small. Controlling for total energy intake, children spreading their intake over several occasions (larger number of daily eating episodes) have significantly lower prevalence of obesity. Neither reported total energy intake nor qualitative aspects of diet, including consumption of beverages, appear to be related to overweight among children in this data set. Reported physical activity is also unrelated to overweight. In contrast, physical inactivity is strongly and significantly related to overweight, so that an additional 1·5 h of inactivity is associated with a 20 % increase in the prevalence of overweight. There was a borderline significant (P = 0·06) inverse association between parental pressure and overweight.
Table 2 Mutually adjusted odds ratios and 95 % confidence intervals for overweight v. normal weight by specified contrasts or increments of the indicated predictor variables
*Energy from carbohydrates was not included in the model in order to avoid collinearity.
Discussion
We have undertaken a cross-sectional investigation of 633 school children in the last two years of their studies in eleven primary schools in variable socio-economic areas of Athens and have defined as overweight children those whose BMI exceeded the 85th centile of the corresponding age-, gender- and height-specific distribution. We have studied the independent predictors of overweight through multiple logistic regression. Overweight was more common among girls than among boys and substantially less common among children born outside Greece. Reported physical activity was unrelated to overweight but patent physical inactivity, operationalised as time spent watching television or working/playing with the computer, was a highly significant predictor of overweight. Composition of diet (energy from protein, carbohydrates and lipids) was unrelated to overweight but the daily number of eating occasions, controlling for total energy intake, was significantly inversely associated with overweight. There was evidence that separation of parents was an underlying cause of overweight.
Excess BMI among girls is a common finding among peri-adolescent girls in comparison to boys of the same age(Reference Wardle, Brodersen, Cole, Jarvis and Boniface14–18) and our finding in this context is in line with the literature. Children born abroad are mostly children of economic migrants of low socio-economic status. These children may be less exposed to the pressure of eating that permeates the Greek society. Children of separated parents, virtually all of them of native Greek parents, tend to be substantially more frequently overweight. This could be due to the fact that parental separation can generate an adverse psychological environment. Overeating and overweight are common manifestations of psychological stress at various stages of life(Reference Pine, Goldstein, Wolk and Weissman19–Reference Stunkard, Faith and Allison21).
As in other epidemiological investigations, we found no association of overweight with total energy intake, a null finding which may be due, at least in part, to the fact that overweight persons tend to underestimate their dietary intakes(Reference Willett and Stampfer22). In our study, about 15 % of energy intake was derived from proteins, about 35 % from carbohydrates and about 50 % from lipids. High intake of fat, largely in the form of olive oil, has been reported in many studies in the Greek population, among adults as well as among children(Reference Trichopoulou, Toupadaki, Tzonou, Katsouyianni, Manousos, Kada and Trichopoulos12, Reference Trichopoulou, Psaltopoulou, Orfanos, Hsieh and Trichopoulos23–Reference Hasspidou, Fotiadou, Maglara and Papadopoulou25). The possible relationship of composition of diet with overweight is a controversial topic. Some authors have reported that controlling for energy intake, consumption of lipids is associated with increased BMI(Reference Astrup, Astrup, Buemann and Raben26), whereas other studies have found no important association(Reference Trichopoulou, Gnardellis, Benetou, Lagiou, Bamia and Trichopoulos27, Reference Field, Willett, Lissner and Golditz28). Among children, there have been reports that foods of animal origin may be associated with overweight(Reference Gunther, Buyken and Kroke29–Reference Roland-Cachera, Deheeger, Akrout and Bellisle32) but the evidence remains inconclusive. Our results do not provide strong support for differential associations of distinct macronutrients with overweight, but power limitations do not allow exclusion for such a possibility. An intriguing finding that has also been reported by other investigators in studies among both adults(Reference Ma, Bertone, Stanek, Reed, Hebert, Cohen, Merriam and Ockene33, Reference Randi, Pelucchi and Gallus34) and children(Reference Toschke, Küchenhoff, Koletzko and von Kries35, Reference Franko, Striegel-Moore, Thompson, Affenito, Schreiber, Daniels and Crawford36) is that controlling for total energy intake, spreading dietary intake over several eating occasions may impede weight gain. The underlying physiology of such an association, however, is not clear, which clouds the biomedical plausibility of the finding(Reference Mattson37).
Controlling for energy intake, an inverse association between physical activity and BMI (and overweight) should be expected on the basis of laws of physics. The empirical evidence, however, is less clear-cut, mainly because exposure ascertainment is difficult and non-differential misclassification extensive(Reference Lagerros and Lagiou13). In our study, the low tail of activity, that is patent inactivity (times spent in television or computer), is probably better assessed and this explains the clear and statistically significant positive association between inactivity and overweight. The suggestive inverse association between parental pressure and overweight probably reflects reverse causation, since overweight children are less likely to receive parental pressure for eating.
Strengths of this investigation are ascertainment of overweight children on the basis of standardised curves, the sample selection from schools chosen to represent parts of Athens of different socio-economic status and the high participation rates. We did not attempt international comparisons that would have necessitated the use of the international Cole criteria(Reference Cole38), but the application of standardised BMI curves for Greek children allowed control of growth pattern. The study was of moderate size but there was enough statistical power for the documentation of several findings. Limitations of the study are lack of formal validation of the physical activity questionnaire and restriction of the sample to the urban population of Athens.
In conclusion, we have found evidence that the principal factor associated with overweight among children in Athens, and inferentially in Greece, is the extended inactivity imposed by the modern childhood lifestyle. This lifestyle involves prolonged watching of television and working and playing on computer. An intriguing finding that has been previously reported but still needs further investigation is that spreading energy intake over several eating occasions was inversely related to childhood obesity.
Acknowledgements
Conflict of interest:There is no conflict of interest and no financial support that needs to be acknowledged.
Author contributions:A.L. designed the study, coordinated its implementation, analysed the data and drafted the manuscript. M.P. contributed in the analysis and the literature search. Both authors have approved the submitted manuscript.
Acknowledgements:The authors wish to thank Dimitrios Alexakos, Athina Ninidaki, Andrianna Papadopoulou, Anastasios Papakonstantinou and Marina Vasilakou, for their contribution in data collection.
