Non-linear causal association of body mass index with serum insulin-like growth factor 1: A Mendelian randomization study

Abstract Objectives: Insulin-like growth factor1 (IGF-1) is a polypeptide hormone mainly secreted from the liver. The synthesis and secretion of IGF-1 are affected by various factors including unusual body weight. Nonetheless, the causal relationship between body mass index (BMI) and IGF-1 is still under debate from existing epidemiological studies, implying their complicated regulation pattern. Aiming to investigate their causal association, we conducted a large-scale linear and non-linear Mendelian randomization (MR) analysis in the UK Biobank (UKB) cohort with BMI as the exposure and IGF-1 as the outcome. Methods: After applying a series of exclusion criteria and covariate adjustment, a total of 244 991 participants from the UKB were eligible for analysis. The polygenic risk score (PRS) of BMI was constructed on 96 instrumental variables (IVs). The non-linear observational association between BMI and IGF-1 was examined by a restricted cubic spline test. Non-linear MR analysis was performed with the piecewise linear method and verified by doubly-ranked stratification and log-transformation methods. In addition to combined analysis, stratified analysis was performed by sex and age groups. A series of sensitivity analyses were performed to examine the robustness of the results. Results: Restricted cubic spline regression demonstrated an inverted U-shaped association between BMI and IGF-1 (P non-linear<0.001), which was also supported by MR analysis (Quadratic P-value: 4.93×10-6, Cochran Q P-value: 2.94×10-5). Specifically, genetically predicted BMI was significantly positively correlated with IGF-1 levels when BMI was less than 25kg/m2, and genetically predicted BMI was significantly negatively correlated with IGF-1 levels when BMI exceeded around 28kg/m2. Stratified analysis showed no difference against sex and different age groups. Sensitivity analyses gave similar results, demonstrating the robustness of the results. Conclusions: This study suggested a non-linear causal relationship between BMI and IGF-1 and this effect may not influenced by sex and age.