Appetite is tightly controlled by internal physiological mechanisms in younger children but a decline in the degree of innate appetite control is observed with age 1-3. It has been postulated that external factors, such as palatability of food, begin to influence internal controls and ultimately override satiety signals 4. The energy density of food has been shown to increase energy intake and short term weight gain in experimental studies 5. Energy density is closely related to the palatability of a food and may disrupt appetite control by over stimulating sensory centres in the brain, which process information on texture and taste of food. An association between fat mass and the FTO gene has recently been identified 6. No functional information about the FTO gene is currently available. Human tissue panel expression studies showed that the FTO gene was most highly expressed in the brain, specifically the cerebral cortex, which may implicate a role in the sensory centres, located in the prefrontal cortex.

I hypothesise that the relationship between dietary energy density and obesity, which we have shown to vary by age in the ALSPAC sample 7, may also vary by the FTO genotype, such that children carrying the high-risk allele A will have a stronger relationship between energy density and obesity than those without. Since the relationship between the FTO gene and BMI z score also appeared to get stronger with age this may be explained by age related changes in appetite control in response to energy dense foods. So I would hypothesise further that the interaction between FTO and energy density on obesity may also get stronger with age.

Data requirements:

This analysis would require access to existing data at ALSPAC from direct assessment and biological samples including:

• Dietary energy density from diet diaries at age 10 years
• Genotype for FTO gene
• Fat mass from DXA at age 13 years
• Height measured at age 13 years (In order to adjust fat mass for height)

The proposed work would be split between MRC HNR and ALSPAC. Laura Johnson, at MRC HNR, would generate the dietary energy density variable from the dietary data at age 10 years. This variable would be checked and cleaned by Laura and sent to ALSPAC. Nick Timpson, at ALSPAC, would perform the analysis of association between the FTO genotype, dietary energy density and fat mass. Laura Johnson would then write up the results of the analysis in a paper.

The plan of analysis to test the hypothesis would be to initially inspect the data for potential associations using cross tabs. If there is evidence for an association then a linear regression should be performed with fat mass (adjusted for height) as the dependent variable and the FTO gene, dietary energy density (DED) and FTO*DED interaction term as independent variables.

If there is evidence for an interaction then a linear regression analysis of fat mass and DED stratified by genotype would be done in order to test the hypothesis that the relationship between DED and obesity is stronger in children with the A allele. Further analysis could include a logistic regression with obesity (the top quintile of fat mass adjusted for height) as the dependent variable and the FTO gene, dietary energy density (DED) and FTO*DED interaction term as independent variables. Followed by stratification by genotype if evidence of an interaction is found.

Calculations in Quanto (http://hydra.usc.edu/gxe/) indicate that using a sample of 5000 children there is a power of 0.99 to detect a true positive effect, using the following parameters to estimate power:

For the continuous outcome fat mass index (kg/m5.8);

Risk allele (A) frequency = 0.39

Mode of inheritance = additive

Dietary energy density standard deviation = 1.5 kJ/g

Fat Mass Index mean = 1.21 kg/m5.8, standard deviation = 0.63

Main effect sizes

Gene B = 0.1 kg/m5.8 per A allele

Diet B = 0.05 kg/m5.8 per 1kJ/g

Gene*Diet B = 0.05 to 0.15

N = 5000

Power for all G*E effect sizes between 0.05 and 1.5 was 0.99, meaning that it should be possible to detect an interaction effect as small as 0.05 kg/m5.8