Blood samples were obtained from 2180 adult outpatients, ages >18 years, at the department of clinical laboratory of Zhongshan Hospital. Blood was collected in tubes without anticoagulant from subjects after an overnight fast. The samples were allowed to clot at room temperature, and serum was obtained by centrifugation at 3000 rpm for 15 minutes. All blood lipid analyses were performed within 1 day. All subjects were classified into three groups according to the TG concentrations (A: < 200 mg/dl, *n* = 1220; B: 200-400 mg/dl, *n* = 480; C: 400-1000 mg/dl, *n* = 480). The Non-HDL-C concentrations in all samples were less than 300 mg/dl. To convert values for TG and cholesterol to millimoles per liter, we multiply the values with 0.0113 and 0.0259, respectively.

The Non-HDL-C value was estimated by the formula as follows [

16]:

Lipid measurements were performed on a Hitachi 911 automatic analyzer. The LDL-C assay was performed according to Roche manufacture's specifications. At the same time, the LDL-C values were also calculated by the FF and MFF. TC and TG concentrations were determined enzymatically using CHOD-PAP and lipase/GPO/PAP methods, respectively. The HDL-C concentration was measured by phosphotungstic acid and MgCl_{2} precipitation approach. The reagents were obtained from Roche Diagnostics. The procedures and efficiency of lipid assays had been demonstrated previously [17]. The total error used in precision assessment was 3.95%-7.85% for the Roche method, as recommended by the National Cholesterol Education Program.

The FF was transformed as follows:

Multivariate linear regression analysis was used to investigate the relationship between LDL-C (expected value), TG and Non-HDL-C (explanatory variables) concentrations. Repeatability of the new formula was evaluated by Bland-Altman analysis [

18]:. We compared the agreement between FF and our new formula, and calculated the mean and standard deviation of the differences (formula and lab value). The mean difference of both FF and new formula were close to zero. We concluded the MFF as follows:

Statistical analysis was performed using SPSS 11.5 for Windows (SPSS Inc., USA). Linear regression analyses were used to assess the correlations between the methods of formula calculation and direct measurement. To examine the degree of consistency between values obtained by the two methods, we used the graphical procedure outlined by Bland and Altman. Comparisons between groups were performed using the method of ANOVA. The test of Pearson chi-square was used to compare discrete variables. *P* values less than 0.05 were considered significant.