Numerous studies in adults have found that diet composition may play an important role in the development of adiposity [2, 4]. Researchers are now investigating the relationship between diet composition and adiposity in children [3, 33]. Several studies investigating the relationship between BMI, waist circumference (WC) and/or skin fold measurements with energy intake have suggested that the macronutrient composition of the diet (protein, carbohydrate, fat) may play an important contributing role to obesity in childhood [33–35] as it does in adults [1–4]. However conflicting results have been found . This study aimed to explore the possible relationship between BMI and WC with energy intake and percentage energy intake from macronutrients in children and adolescents, using a more robust technique by eliminating potential food reporting errors.
Studies using BMI as a parameter for "overweight" and "obesity" such as Rocandio et al., (2001) and Hassapidou et al., (2006) found that overweight adolescences of both sexes reported lower energy intake than non- overweight subjects [36, 37]. However, our findings, where we have accounted for mis-reporters, demonstrate a weak but significant positive correlation between BMI z-score and energy intake. When examining the relationship between BMI z-score and energy intake based on quartiles, there was a significant difference amongst the groups, suggesting that our population of overweight children and adolescents consume more energy than their leaner counterparts. Similar findings were seen when using WC to assess the degree of overweight and obesity. Once again a weak but significant positive correlation between WC and total energy intake across all age bands was demonstrated. When comparing groups based on quartiles of WC there was significant difference between the groups in year five and ten. Yet, no significant differences between the groups were found in younger children (year 1).
Energy intake from carbohydrate has been shown in adults to be inversely associated with body fat [22, 36, 38]. Studies by Hassapidou et al. (2006) and Ortega et al. (1995) found that overweight and obese adolescents consumed fewer carbohydrates than lean subjects [23, 37]. Again, once under-reporters are omitted, the results from this study differ in that there was no association between carbohydrate intake and either BMI z-score or WC. Again, there was no significant difference between the groups when comparing percentage carbohydrate intake based on quartiles of BMI z-score and WC. Similar to studies by Rocandio et al. (2001) and Maffeis et al. (1996) in which no difference in protein intake was found between overweight and non overweight children and adolescents, our findings show that percentage protein intake did not differ between quartiles of BMI z-scores or WC [18, 36].
While the majority of studies have demonstrated a positive association between adiposity and dietary fat our results conflict with these findings [3, 22, 33, 39]. An inverse relationship was observed between fat intake and BMI z-score and similar results were also found when using WC as a reference for adiposity, however these correlations were not significant. When subjects were split into quartiles based on BMI z-score and WC, there was no significant difference between the groups, which support the results found by Hassapiduo et al. in 2006, in which no significant difference in percentage fat intake was found in overweight and non overweight adolescents .
Using multiple regressions we found that the macronutrient composition of the diet had no significant impact on the ability to predict BMI z-score or WC.
Researchers have suggested that the reason why a clear relationship between total energy intake and percentage macronutrient intake and adiposity has not been seen is due to the under reporting of foods. This study applied the McCrory cut off  and those with an "implausible" energy intake were excluded from the analysis, thus applying a more robust method of self reported energy intake. The results suggest that the lack of association between BMI z-score and WC with macronutrient intake was not a direct result of underreporting and that total energy intake is more influential than the macronutrient composition of the diet in the development of childhood obesity.
However, perhaps one of the reasons of conflicting findings from various studies is the use of BMI itself as a measure of adiposity. While a simple, convenient assessment, the accuracy and use of the BMI as an indicator of body fatness in children is questionable . Riley et al. 2000, suggested that BMI cut -offs are non specific, thus tending to identify non-obese "stocky" children as obese . On the other hand, there is also a concern that BMI cut-offs fail to identify the obese child . We attempted to partly overcome this by also exploring WC as a measure of obesity; however our correlations with WC and dietary components were very similar to those with BMI, and added nothing further to the study. It may be that WC should have been adjusted for body size. Studies investigating the relationship between dietary intake and the development of childhood obesity have used various screening tools to group children based on their level of adiposity, and as a result, the difference in methodologies between studies may account for conflicting findings.
Similarly, the use of different assessment tools to assess dietary intake in children and adolescents may explain some of the variability in previous studies assessing dietary intake and adiposity. In the current study, dietary intakes and food habits were assess by a 24 h food and drink record, similar to the 2003 Physical Activity and Nutrition Levels in Western Australian Children and Adolescents Report, which was adapted from the 1995 National Nutrition Survey. In this survey a 24 hour dietary record was used, a common approach in large population based studies. Whilst the 24 hour dietary record may not always represent habitual intake we have attempted to limit the error by screening the intake data via the method described by McCrory et al. .