To our knowledge this is the first study to show two independent effects of both beverage sugar and animal protein intake during infancy and overweight at age 8 years. Beverage sugar intake during infancy increases the risk of overweight at 8 years by more than 10% extra risk per energy percent beverage sugar increase. For animal protein infants with the highest tertile of intake had a more than 9 times higher risk of becoming overweight at 8 years.
In 1995 Rolland-Cachera suggested the early protein hypothesis; a high protein intake in excess of metabolic requirements enhances weight gain in infancy and increases the risk of obesity later in life . This hypothesis has been confirmed in similar small studies [6–8], although not in hispanic youth . The period of dietary exposure, the timing of overweight assessment and the definition of overweight might in part explain different outcomes. Gunther et al. suggested that particularly animal protein might be responsible for the association . When the present analysis was performed with two dummies for second and third tertile of animal protein intake (data not shown), the second tertile did not show a significant relationship. In line with this observation is the hypothesis that animal protein above a certain level of intake might deregulate insulin and IGF-1 with an impact on preadipocyte differentiation and multiplication [2, 27]. High protein intakes may also decrease human growth hormone secretion and reduce lipolysis, which might aid fat accumulation . Lower protein levels in infant formula seems to reduce weight up to age 2 years , and might therefore contribute to reduction of later overweight.
A recent review by Gibson  showed an overwhelming amount of studies on sugar sweetened soft drinks and obesity. Gibson included all cold beverages containing added sugars, carbonated or not. The evidence seems to be inconclusive, although a small effect of sugar sweetened soft drinks on BMI is suggested. A recent study by Herbst et al.  is the single study that prospectively investigatedinfant beverage sugar intake during the first year of life. Herbst et al.  show no or possibly a protective effect of sugar from beverages and sweets, however the level of beverage sugar intake is very low. At 1 year of age the beverage sugar intake is 0 percent of total energy intake in the Donald study  compared to more than 5 percent in the present study. A reasonable explanation for this difference is that the Donald study has selected infants from a higher socio-economic level: > 80% of infants was breastfed for more than 4 months versus 41% in our study, > 60% of parents had a high socio-economic status versus one third in our study, and maternal overweight was only 25% versus > 40% in our study). In line with our study, the authors conclude that when added sugar intake increases to higher levels this might be detrimental to BMI development.
Beverage sugar was consumed through apple and orange juice, most often from apple concentrate, and lemonade also from sugar concentrate. Apple juice has a high fructose content. Fructose consumption results in decreased circulating levels of insulin and leptin when compared with glucose . Because insulin and leptin function as key signals to the central nervous system in the regulation of energy balance, it has been assumed that over a longer period of time high fructose intake could lead to increased caloric intake or decreased caloric expenditure and weight gain . However, at this moment and from this study it is not possible to draw any firm conclusions about the mechanism involved.
Limitations of the study are the total number of infants, although similar studies were of similar size [6–8]. The low response rate to the initial study is also a limitation, but each child serves as its own control. Confounders were not all available for the whole sample of 120 children, but for the subsample of 63. However, the primary effect of beverage sugar and animal protein did not seem to change when tested in the subsample (from model 3 to model 4). Therefore we assume that the effects we found are very robust and would not be different when the whole sample would have provided all information on confounders. Although a large random sample was the basis of the study, socio-economic status (maternal education) was slightly higher compared to Dutch average http://statline.cbs.nl/statweb/. The educational level of parents was very similar to the only other Dutch investigation into dietary intake of infants, the Nutrient Intake Research (VIO) study . This study used a representative sample of 300 infants for both the 9 and the 12 month old infants. We have evaluated the quality of the self reported dietary intake data in an earlier study as good , also based on the similarity with the VIO study. Two day food records at one time point is a limited assessment of longer-term exposure to food constituents. However, providing high levels of beverage sugar and/or food products high in animal protein are most likely related to long established food habits of (primarily) the mother. It should also be stressed that both beverage sugar and animal protein were not related to poor food habits. Beverage sugar is largely from fruit juices and from apple concentrate, which the parents judge as 'fruit' and therefore as 'good'. Animal protein is derived from normal healthy food groups and infant formula. The level of intake of percentage energy beverage sugar or animal protein is not significantly different between high and low socio-economic status groups. The relationship between high animal protein intake (highest tertile) with overweight at 8 years is even stronger in high compared to low socio-economic status group (p = 0.006 vs p = 0.102 respectively). This may have resulted in low underreporting as well as high consistency in food 'habits'. Comparing responders with non-responders, no differences appeared in socio-economic status, but parents with breastfed infants at the time of investigation in 2001 had responded to the follow-up call in 2009 significantly more often (26/38 vs 93/188, p = 0.009). Information on maternal smoking is missing, however it is not likely that smoking is associated with beverage sugar or animal protein intake at infancy.
Also, the intake of mother's milk may be more inaccurately estimated than the intake of formula milk.
However, considering the low contribution of mother's milk to the total consumption, it is unlikely to play a key role in effect size and/or significance.
The use of BMIsds as definition of overweight is not ideal but acceptable as an outcome measure; no definition of obesity is ideal at present . Both weight and height are self reported which might introduce some bias. However, recently it has been shown that BMI misclassification was almost absent in an adult population when self reported weight and height were evaluated . Children are more often subject to measurement of weight and height, therefore BMI misclassification is assumed to be limited. Although some BMI misclassification may have occurred it is not likely that the amount of BMI misclassification is related to protein and beverage sugar intake in infancy. It is unlikely this would have caused the significant effect found in this study since the effect of non-differential misclassification on estimates of relative risk will be to always reduce the true relative risk . Physical activity is also self reported, however self reporting is probably less important than selection of a method with a high internal and external validity . Another limitation is the observational nature of the design, which makes it diffecult to conclude a causal nature .
Provision of sugar containing beverages and a high animal protein diet in the first year of life appears to increase the risk of overweight at 8 years. Although childhood overweight is clearly multi-factorial, each significant and clinically meaningfull contributing factor that can be influenced by the parents should be addressed accordingly. Sugar containing beverages provided during infancy are much less difficult to address than lifestyle changes during later stages of life [32, 33].
Future studies with a larger sample size should confirm these observations.