Statistics
To test the association between LPL SNPs and haplotypes (HTs) and lipid concentration, linear regression analyses, adjusted for age, gender, and body mass index (BMI), were performed. For stage 1 and imputed data, the residence area of each participant was additionally adjusted. The influence of the interaction between LPL SNPs and lifestyle variables on lipid levels was analyzed by using linear regression, including main and interaction effects, adjusted for residence area, age, gender, and BMI. The effect of energy and fat intake, physical activity, smoking, and drinking status on lipid profiles was examined. Energy intake, fat intake, and physical activity were measured as the total intake of kilocalories per day, daily fat intake, and metabolic equivalent scores per hour, respectively. Smoking status had two categories, non-smokers and smokers, and alcohol consumption was divided into non-drinkers and drinkers. These lifestyle factors were available for subjects from stage 1 and imputed data but not from stage 2.
All analyses were performed by log-transforming HDLC, TG, and TCHL values to normalize their distribution. Untransformed LDLC values were used for all analyses. To determine the effect of SNPs, untransformed lipid concentrations were used.
Linkage disequilibrium (LD) block partition was performed by Gabriel's rule, applied in Haploview version 4.1 (Broad Institute of MIT and Harvard University, Boston, MA, USA), and HT inference was performed using the expectation maximization algorithm with PLINK version 1.07 (http://pngu.mgh.harvard.edu/~purcell/plink). All statistical analyses were performed in an additive model (AA vs. AB vs. BB or +/+ vs. +/- vs. -/-) using PLINK.