Statistical analysis

PASW statistics 18 (formerly SPSS Statistics; http://www.spss.com.hk/statistics) was used to analyze the data. The Hardy-Weinberg equilibrium (HWE) testing was calculated using the Finetti method (https://ihg.gsf.de/cgi-bin/hw/hwa2.pl), and the gender differences for alleles and genotypes’ distributions were also evaluated. All variables were categorized by the median value in study sample and t-tests were performed to estimate the effects of non-genetic variables on individuals’ serum lipid levels. A one-way analysis of variance (ANOVA) and LSD post hoc test were performed to examine the influence of each tagSNP on individual serum lipid status (including TCHO, TG, HDL-c and LDL-c). The non-parametric test, Kruskal-Wallis H test (K-W test) was also used. Linkage condition and haplotypes construction of positive SNPs were estimated by UNPHASED software (version 3.0.13) [25]. More stringent criteria for positive tag SNP(s) was adopted: tag SNP(s) which demonstrated a significant association using both ANOVA and the K-W test were considered.

To partial out other non-genetic variables (i.e., age, gender, BMI, and WHR) that might have influences on the relationship between SNPs and individuals’ serum lipid status, a multiple regression analysis was performed in which the genotype group served as one of the class variables. For the multiple linear regression analysis, we performed the following steps: (step 1) entering control variable(s) and (step 2) entering both variables and the genetic information. The statistical power analysis was referred to as p < 0.05 for two-tailed tests.