4.3. Statistical methods

All statistical analyses were performed using the SAS program (SAS Institute Inc., Cary, NC, USA).

In study I, all values were expressed as mean ± standard deviation (SD). The statistical significance of the results was calculated using Student"s t-test.

When comparing the effects of the polymorphisms in study II, the values were expressed as mean ± standard deviation, except for Lp(a), for which the median and range were given. The quantitative variables were compared using the t-test or the analysis of variance with Tukey’s Studentized Range test, while Lp(a) and triglycerides were analysed with the nonparametric tests of Wilcoxon or Kruskal-Wallis. Proportional data were compared using the c2 test.

In study III, the values for total, LDL and HDL cholesterol were expressed as mean ± standard deviation and those for Lp(a), total triglycerides, VLDL cholesterol and VLDL triglycerides as median and range. The quantitative variables were compared using the t-test. For multiple pairwise comparisons, analysis of variance was used, followed by pairwise comparisons with Tukey"s Studentized Range test. Proportions were compared using the c2-test or the Cochran-Mantel-Haenszel test. Logarithmic transformation was applied to plasma total triglycerides, VLDL triglycerides and VLDL cholesterol because of their skewed distributions. Due to the different age distributions of the patient groups, the values of total cholesterol, LDL cholesterol, HDL cholesterol and plasma total triglycerides were age-adjusted for the statistical analyses by linear regression. Stepwise logistic procedure and ranked stepwise logistic procedure were used to compare the lipid risk profiles of the male and female controls and the patients with CAD and different extensions of CAD, respectively. Age, smoking status, hypertension and apolipoprotein E phenotype were forced as parameters into each model, and the lipids included were lipoprotein(a), HDL cholesterol, LDL cholesterol, VLDL cholesterol and VLDL triglycerides.

In study IV, the lipid values were expressed as mean (SD). The quantitative variables were compared using the t-test or the analysis of variance with Tukey"s Studentized Range test, and the analysis of covariance was performed using the general linear models procedure. Logarithmic transformation was applied to the triglycerides due to their skewed distributions. The relationships between variables were measured by the Pearson correlation. Due to the small number of patients, the apo E 2/3 and 3/3 phenotypes were grouped as one group and 3/4 and 4/4 as one group for the statistical comparisons involving CETP.

In study V, the sample size for evaluating the cholesterol-lowering effect of RS-86505-007, 24 for each group, was based on the following power calculations: a 11 % detectable difference between the groups in mean total cholesterol and a 12 % difference in the LDL cholesterol level, and 4 % within groups for total cholesterol and 5 % for LDL cholesterol. The means of the three measurements of total cholesterol, HDL cholesterol and triglycerides made during the diet period were used as the baseline values in the analysis. Each of the sections of the trial (3 mg and 6 mg dose groups) was analysed separately, the main analysis being the comparison of RS-86505-007 and placebo with respect to the change from the baseline to the end of treatment. Patients with triglyceride values higher than 3.95 mmol/l were excluded from the analyses of HDL and LDL cholesterol. The percentage changes were tested using a two-way analysis of variance with a model including the centre, the treatment and their interactions. All statistical tests were two-tailed and considered to be significant with a 95 % confidence interval.

Paired t-test and the analysis of variance were used in the comparisons of lipid changes between the polymorphism groups. The Lp(a) concentrations were compared with the t-test.