Table 3.

Proteinuria at study end (ANCOVA model of log-proteinuria)

Group Mean (95% CI)% Reduction (95% CI)P
Model 1 (R2 0.23)
    control0.36 (0.21 to 0.62)−69.80 (−86.04 to −34.68)0.003
    intervention0.11 (0.06 to 0.20)
Model 2 (R2 0.29)
    control (no toxicity)0.55 (0.26 to 1.18)−83.59 (−94.24 to −53.18%)0.001
    intervention (no toxicity)0.09 (0.03 to 0.22)
    control (toxicity)0.22 (0.10 to 0.48)−35.66 (−78.62 to 93.66%)0.425
    intervention (toxicity)0.14 (0.06 to 0.30)
  • Geometric means (95% confidence intervals [CIs]) of daily proteinuria at study end and percentage change due to intervention (exponentiated differences of log-values) without considering toxicity (model 1) and by absence and presence of drug toxicity (model 2). Both models include baseline proteinuria, intervention group, and prednisone dose. The models failed to reject the null hypothesis of inferiority of rituximab (RTX) as compared to standard therapy. In model 2 (which includes also the interaction between toxicity and intervention) the average proteinuria values at 3 months are similar to (children with toxicity [P = 0.425] but upper 95% CI <400%, the noninferiority margin of 3 times) or lower than baseline values (children without toxicity [P = 0.001]). Of note, the use of prednisone and calcineurin inhibitors was reduced only in the RTX arm of both strata (Figure 2). ANCOVA, analysis of covariance.