Differences in categorical measures across ordinal groups (i.e., PI3K inhibitor drugs educational attainment and iron scores) were assessed using a Jonckeere-Terpstra test, or the exact equivalent. Continuous measures were summarized as medians and interquartile ranges with differences in group distributions assessed using a Wilcoxon rank sum test (for comparison of two groups) or a Kruskal-Wallis test (for comparison

of more than two groups). To assess whether changes in lipid profile measures significantly differed from baseline, a Wilcoxon sign rank test was used. Associations between the proportion of PEG-IFN and ribavirin taken with changes in serum selleck chemical lipids were assessed using Spearman’s correlation analyses. For all statistical tests, P < 0.05 was considered statistically significant. To evaluate factors associated with SVR, a relative risk model was employed with a robust variance

estimator.38 In regression models, TG, HDLc, and TC were transformed to the natural logarithm scale to achieve normality. All continuous predictors were centered. The relationships between baseline and 24-week changes during treatment in lipid profile measures and the probability of SVR were graphically assessed using smoothing spline plots. Due to different patterns observed by gender, smoothing spline plots for HDLc were examined separately for males and females. Two types of multivariable models of SVR were constructed using a stepwise approach. One type of multivariable model (models 1 and 2) allowed Farnesyltransferase pretreatment characteristics and the amount of PEG-IFN taken during the first 24 weeks as eligible predictors. Model 2 allowed as additional eligible predictors the baseline lipid profile measures. A second type of multivariable model (model 3) also

adjusted for body weight changes and allowed for the inclusion of variables representing baseline and changes in lipid profile measures during the first 24 weeks of therapy as eligible predictors. To compare the prediction of multivariable models, differences in area under the receiver operating curves (AUROCs) were assessed using a nonparametric method.39 Baseline characteristics of the 330 participants are shown in Table 1. AAs did not significantly differ from CAs by age, gender, employment status, health risk behaviors (smoking status and weekly alcohol consumption), viral level, aspartate aminotransferase, international normalized ratio, white blood cell count, platelet count, percent iron/total iron-binding capacity, Ishak fibrosis, total histological activity index score, steatosis, TG, HDLc, or TC. Compared with CAs, a larger percentage of AAs had health insurance coverage (87% versus 78%, P = 0.