For the obese children our results suggest a consistent picture of poorer general health as compared to the normal weight children. The results indicate that this may be partly explained by a higher incidence of common childhood respiratory infections in obese children. The association between obesity and the use of antibiotics is also indicative for an association between obesity and infections. For the moderately overweight children, the results did not suggest poorer health. The analyses of the influence of parental perception of the child's body shape showed that parents who saw their child as slim/slender reported a lower RAND score, more GP visits and more health-related limitations than parents who did not see their child as slim/slender. Actually being thin was not associated with these outcomes in analyses adjusted for parental perception. When perceived heaviness/chubbiness was included in the regression models, the associations between obesity and the health outcomes became stronger rather than weaker (except for the RAND score), indicating that perception did not explain the associations that were observed between obesity and the health outcomes.
Strengths and limitations of the study
Important strengths of the study were the large study population, the availability of measured, rather than reported, data on weight and height, the range of different health outcomes that could be studied, the possibility to adjust for prospectively collected indicators of parental health related behaviour and the possibility to assess the role of parental perception of the child's body shape.
A number of limitations have also to be considered.
Firstly, the health outcomes that were studied were based on questionnaire data and not on objective measurements. We therefore need to consider the potential influence of recall problems. We expect that parents will mostly have been able to recall the sort of events we asked about (such as "Did your child visit a GP during the last 2 months?"), but the possibility of recall errors can never be excluded. If parents' ability to recall health problems of the child was unrelated to the child's BMI status, incorrect recall will have caused 'noise' but not bias in our results. It could be hypothesized that parents with a relatively low 'health consciousness' or parents with parenting problems might have more obese children and might at the same time more frequently fail to recall disease episodes of their child. If this would be the case, this would have made the associations observed in our study weaker than the 'true' associations.
In addition to recall problems, we need to address the possibility that obese children are as healthy as normal weight children and that the only difference between them is the way their parents report on their children's health. The outcomes studied ranged from parameters that depend largely on subjective parental evaluation, such as the RAND score and health related limitations, to more objective measures like doctor diagnosed respiratory infections and use of antibiotics. Associations with obesity were observed for the more subjective as well as the more objective outcomes. However, although we asked about illnesses that were 'diagnosed by a doctor', a doctor's diagnosis depends not only on the (severity of the) actual illness, but also on parents' readiness to seek medical help. Indicators of parental health-related attitudes and behaviour, i.e. parental education, smoking and breast feeding were included in the analyses and did not substantially change the associations between overweight and the health outcomes. We therefore assume that the associations between obesity and parental reported health outcomes are not the result of confounding by parental health-related attitudes or behaviour. Also parental perception of the child's body shape, which was clearly involved in the association between thinness and general health, did not explain the associations between obesity and the health outcomes. We think therefore that it is unlikely that characteristics of the parents of obese children, rather than the actual health of obese children explain the associations we observed between obesity and the health outcomes studied.
Secondly, participation in the medical examination at the age of 8 years was lower than the response on the questionnaires, resulting in a relatively high proportion (44%) of missing data on measured weight and height. We followed common practice by conducting the analyses in the sub group of children with complete data. Because complete case analysis is liable to bias due to selective missing of data however, we repeated the analyses after missing data had been multiple imputed. The probability of data being missing was strongly associated with a number of variables measured in the study and we had parental reported weight and height data for most of the children who did not have their weight and height measured by the investigators. These two factors made us decide that multiple imputation was the best available method to deal with missing data in our study. Nevertheless, the relatively high percentage of missing values on BMI status remains a reason for cautious interpretation of the results both of the complete case analysis and the analysis in the imputed datasets. Comparison of the results of the analyses in the complete cases and in the imputed data showed some differences in the strengths of the associations. There was no consistent pattern of associations always being weaker or stronger in one analysis than in the other however and all associations were in the same direction. Both analyses suggested a consistent pattern of poorer health in the obese children.
Thirdly, the number of obese children in our study population was relatively low, especially in the complete case analysis, which may have limited the possibility to detect significant associations.
Finally, we specifically focused on health problems and diseases that have a high prevalence in children in the general population and we did therefore not address all the health problems obese children may face.
Findings of other studies
Very few studies have investigated the association between overweight and the
general health outcomes addressed in this paper. Pinhas-Hamiel et al. studied the health-related quality of life in Israeli overweight children and adolescents and observed that adverse associations were particularly strong in the physical domain as compared to the social, emotional and school domains [
16]. In a study in Australian 4-5-year-old children, Wake et al investigated parent reported child global health, health related quality of life, mental health problems, asthma, sleep problems, injuries and special health care needs in relation to weight status [
17]. They found no significant associations, except for a higher prevalence of special health care needs in obese children, and concluded that overweight and obese pre-school children experience few additional health burdens as compared to normal weight children. The authors suggest that morbidity rates may be higher in older children and the age difference between their study population and ours may, besides the differences in health outcomes studied, indeed explain the difference in the results between this study in pre-school children and our study. The same group recently published their results on comorbidities in overweight adolescents and reported that in this age group current obesity was associated with lower scores on 'physical health' and 'global health' [
18].
With respect to the association between overweight and
respiratory infections, evidence has been reported for a higher prevalence of wheezing, cough, bronchitis and pneumonia in U.S. children with a high BMI [
19], for a higher incidence of respiratory and ear problems among Dutch overweight children [
20] and for an association between obesity and the occurrence of otitis media with effusion in Korean children [
21].
For adults, there is some evidence for associations between overweight and poor respiratory function [
22], nose and throat complaints [
23], common cold and influenza [
24], respiratory diseases, including upper airways infections [
25] and prescription of antibiotics and drugs for respiratory diseases [
26,
27]. Remarkably, the Centers for Disease Control recently published evidence that people who are obese but otherwise healthy may be at increased risk of severe complications and death from Influenza A (H1N1) [
28].
Interpretation of the results
Our data showed associations between obesity and respiratory infections and between obesity and the use of antibiotics. Similar results have been reported in the few studies that have covered these outcomes in children [
19‐
21] and in adults [
22‐
28]. Different explanations for these associations need to be considered. Reverse causation, illness causing overweight, cannot be excluded, but does not seem a plausible explanation. Illness tends to be associated with loss of appetite, and disease processes, especially infections with fever, increase metabolic rate [
29], so that infectious disease is more likely to result in weight loss than in weight gain. Confounding could be an explanation. Although our analyses did not show evidence for confounding either by parental lifestyle or by parental perception of the child's body shape, we can not exclude the possibility that, for example, a low quality diet with a high energy and a low nutrient content could increase the risk of both overweight and respiratory infections. Given the limited effect on our results of breast feeding and smoking - factors that tend to be associated with families' eating habits - it seems unlikely however that an unhealthy diet could entirely explain the associations observed. Our results therefore suggest that obesity is likely to be an independent risk factor for doctor diagnosed respiratory infections in children.
How could obesity increase the risk of respiratory infections? Although we cannot tell from our data, we think that it is more likely that obesity may increase the severity and/or duration of respiratory infections than that it increases the incidence of infection. The observations on obese Influenza A (H1N1) patients seem to point in this direction. With respect to our data, this would imply that the obese children may not have had more infections than the normal weight children, but that they may have been ill longer or more seriously and therefore more often required GP attention and treatment with antibiotics. One possible explanation for an association between obesity and duration or severity of respiratory infections could be that excess body fat reduces lung volume, resulting in suboptimal ventilation and reduced clearance of micro-organisms from the airways. Another explanation could be that obesity influences immune responses to infection. Evidence for obesity related changes in immune responses has been observed in mice. Diet-induced obese mice, infected with influenza virus, were found to have a significantly higher mortality rate than their lean controls as well as elevated lung pathology [
30]. Along with the increase in mortality, an altered immune response was observed in these mice, including delayed proinflammatory cytokine expression. Although these mechanisms seem plausible, the question if and to which extent they explain the associations between obesity and respiratory infections observed in epidemiological studies needs further study.