Dennis Freuer1, Christa Meisinger2, Jakob Linseisen2. 1. Chair of Epidemiology at UNIKA-T Augsburg, Ludwig-Maximilians-Universität München, 86156 Augsburg, Germany; Independent Research Group Clinical Epidemiology, Helmholtz Zentrum München, German Research Centre for Environmental Health, 85764, Neuherberg, Germany. Electronic address: d.freuer@unika-t.de. 2. Chair of Epidemiology at UNIKA-T Augsburg, Ludwig-Maximilians-Universität München, 86156 Augsburg, Germany; Independent Research Group Clinical Epidemiology, Helmholtz Zentrum München, German Research Centre for Environmental Health, 85764, Neuherberg, Germany.
Abstract
BACKGROUND: The question whether the proportion of energy provided by fat and carbohydrates in the diet is associated with body mass index (BMI) and waist circumference (WC) is an important public health issue, but determining causality is difficult in epidemiological studies. OBJECTIVES: Using a two-sample bidirectional Mendelian randomization (MR) in both a univariable and multivariable setting, we aimed to determine whether the relative proportion of different macronutrients in the diet (in % of total energy intake (E%)) is causally related to BMI and WC and vice versa. METHODS: All analyses were based on genome-wide association studies including 268,922 Europeans with dietary data (SSGAC Consortium) and at least 232,101 with anthropometric measures (GIANT Consortium). An inverse-variance weighted approach using modified second-order weights within the radial regression framework was performed. Radial MR-Egger, weighted median and mode, Robust Adjusted Profile Score (RAPS), and Pleiotropy RESidual Sum and Outlier (PRESSO) methods were used in sensitivity analyses to verify MR assumptions. Additionally, multivariable MR was conducted to account for inter correlation between macronutrient intakes. All estimates represent the standard deviation (SD) change in each outcome per one SD change in the respective exposure. RESULTS: We found that genetically predicted relative carbohydrate intake (E%) reduced BMI (β = -0.529; 95% CI: -0.745, -0.312; P-value = 2⋅10-6) and WC (β = -0.459; 95% CI: -0.656, -0.262; P-value = 5⋅10-6). Both effects were also supported by the multivariable approach: β = -0.441 (95% CI: -0.772, -0.109; P-value = 0.009) for BMI and β = -0.410 (95% CI: -0.667, -0.154; P-value = 0.002) for WC. Genetically predicted dietary intake of fat (E%) was weaker and positively related to both anthropometric measures. We obtained evidence that a higher BMI and WC increased the relative dietary intake of fat and protein (E%). For example, each SD higher BMI increased protein intake (E%) by 0.114 SD (95% CI: 0.081, 0.147; P-value = 9⋅10-12) and each SD higher WC increased protein intake (E%) by 0.078 SD (95% CI: 0.035, 0.121; P-value = 4⋅10-4). Sensitivity analyses confirmed these findings revealing consistent effect estimates. CONCLUSIONS: Using genetic information to improve causal inference we found evidence, that a low relative carbohydrate proportion (E%) and a high proportion of fat (E%) in the diet is causally related to a higher BMI and a higher WC. Further research considering carbohydrate, fat, and protein quality and possible consequences on micronutrient intake is needed to define the implications for dietary intake recommendations.
BACKGROUND: The question whether the proportion of energy provided by fat and carbohydrates in the diet is associated with body mass index (BMI) and waist circumference (WC) is an important public health issue, but determining causality is difficult in epidemiological studies. OBJECTIVES: Using a two-sample bidirectional Mendelian randomization (MR) in both a univariable and multivariable setting, we aimed to determine whether the relative proportion of different macronutrients in the diet (in % of total energy intake (E%)) is causally related to BMI and WC and vice versa. METHODS: All analyses were based on genome-wide association studies including 268,922 Europeans with dietary data (SSGAC Consortium) and at least 232,101 with anthropometric measures (GIANT Consortium). An inverse-variance weighted approach using modified second-order weights within the radial regression framework was performed. Radial MR-Egger, weighted median and mode, Robust Adjusted Profile Score (RAPS), and Pleiotropy RESidual Sum and Outlier (PRESSO) methods were used in sensitivity analyses to verify MR assumptions. Additionally, multivariable MR was conducted to account for inter correlation between macronutrient intakes. All estimates represent the standard deviation (SD) change in each outcome per one SD change in the respective exposure. RESULTS: We found that genetically predicted relative carbohydrate intake (E%) reduced BMI (β = -0.529; 95% CI: -0.745, -0.312; P-value = 2⋅10-6) and WC (β = -0.459; 95% CI: -0.656, -0.262; P-value = 5⋅10-6). Both effects were also supported by the multivariable approach: β = -0.441 (95% CI: -0.772, -0.109; P-value = 0.009) for BMI and β = -0.410 (95% CI: -0.667, -0.154; P-value = 0.002) for WC. Genetically predicted dietary intake of fat (E%) was weaker and positively related to both anthropometric measures. We obtained evidence that a higher BMI and WC increased the relative dietary intake of fat and protein (E%). For example, each SD higher BMI increased protein intake (E%) by 0.114 SD (95% CI: 0.081, 0.147; P-value = 9⋅10-12) and each SD higher WC increased protein intake (E%) by 0.078 SD (95% CI: 0.035, 0.121; P-value = 4⋅10-4). Sensitivity analyses confirmed these findings revealing consistent effect estimates. CONCLUSIONS: Using genetic information to improve causal inference we found evidence, that a low relative carbohydrate proportion (E%) and a high proportion of fat (E%) in the diet is causally related to a higher BMI and a higher WC. Further research considering carbohydrate, fat, and protein quality and possible consequences on micronutrient intake is needed to define the implications for dietary intake recommendations.