Summary statistics for normally distributed quantitative variables were expressed as means and standard deviations. For non-normally distributed variables, we used median and IQR; categorical data were summarized by ratiosand percentages.
Differences in means for continuous variables were comparedusing Student’s t-test (two groups) or analysis of variance (multiplegroups), and differences in proportions were tested by x2 test.
Cox proportional hazards models were used toanalyze the association of serum Gd-IgA1 levels and the primary outcome.Gd-IgA1 was highly skewed to the right in this group patients and natural log transformation was used. Serum Gd-IgA1 was first analyzed as a continuous variable with HRs calculated per s.d. increment of natural log–transformedGd-IgA1, and the Gd-IgA1 quartile as a categorical variable, with the lowest quartile defined as the reference group.
The relationship between Gd-IgA1 andrisk of end point was examined in unadjusted and multivariable-adjusted Coxmodels. Proportional hazards assumptions were verified by testing the interaction of survival time and lnGd-IgA1 and quartiles of Gd-IgA1 (P¼0.23 and P¼0.17,respectively), and by inspecting parallelism of estimated hazard functions.
A two-sided P-value 0.05 was considered statistically significant. All statistical tests were performed using SPSS version 16.0 (SPSS, Chicago, IL).