Know the Problems in the Assessment of Risk Factors for Obesity in Children

Problems in the Assessment of Risk Factors for Obesity in Children Johnson-Down et al. [91] described that the reported food intakes of overweight children did not meet calculated energy needs. There are different possible explanations for this unexpected finding: One explanation could be that the study participants were on a therapeutic diet. This would be an error due to confusion of chicken and egg. Another explanation could be that study participants did not report their actual eating habits correctly, but told the interviewers what they were supposed to do: eat less. This would be an example of intentional and selective misclassification of the diet resulting in underreporting. With few exceptions it is impossible to investigate all members of a population,
but information on the entire population is desired. So a representative selection of individuals has to be made under the option of generalizability. Thus, during the recruitment of study participants selection biases can occur. This may be due to refusal to participate, or selection of particular study populations like health professionals, or by self-selection, i.e. patients or parents who are more aware of certain health problems for whatever reasons refer themselves to voluntary studies. In a questionnaire study on breastfeeding and childhood obesity breastfeeding mothers of non-obese children might preferentially participate, e.g. to prove the virtue of breastfeeding. This would result in selection bias.
 

When the effect of a certain exposure is distorted by an unapparent exogenous factor, this is called confounding. Confounding can lead to over- or underestimation of an effect or even alter the direction of an effect [90]. Confounding was a major issue in the assessment of the impact of maternal smoking in pregnancy on childhood obesity. Maternal smoking in pregnancy is associated with a number of known and potentially unknown risk factors for childhood obesity. If information on known risk factors is collected, adjustment is possible in the analysis [82]. For unknown confounders such an adjustment is not possible, however, accounting for spurious results caused by residual confounding. Well-designed epidemiological studies, however, may provide clues for the understanding of the causes of obesity in individuals and populations. At the present, the etiology of the adiposity epidemic is not yet understood. Thus, potential causes and the course of the epidemic should be kept under surveillance. This opens the field to intensive epidemiologic research.