Fig. 3: Key secondary endpoints.

a, The proportion of patients with HES flares was analyzed using a logistic regression model with treatment group and region as covariates to estimate the OR and 95% CIs. This analysis included patients who withdrew from the study without having flares as an event; these patients were censored at the time of study withdrawal in the analysis of time to first flare. b, The number of HES flares (annualized rate) was analyzed using a negative binomial model with treatment group and region as covariates. The logarithm of follow-up time was used as an offset variable. Flare rates were estimated for each treatment group (error bars represent 95% CIs) and the RR and 95% CIs were calculated for benralizumab versus placebo. c, The time to first hematologic relapse (AEC ≥1,000 cells µl⁻¹) was analyzed using a stratified log-rank test, adjusted for region. HR and 95% CIs were estimated using a Cox proportional hazards model with treatment group and region as covariates. Patients who did not experience hematologic relapse were right censored at the end of the double-blind period corresponding to the earliest date of: the first benralizumab open-label dose, study day 183, the date of last contact and the data cutoff date. d, The LS mean change from baseline to week 24 in PROMIS Fatigue scores was analyzed using a mixed model for repeated measures, with treatment group, baseline score, visit, region and treatment visit interaction as covariates. All available observations were included with no imputation for missing data, assuming missing data were missing at random. The standardized T-score range for Short Form 7a is from 29.4 to 83.2; higher scores indicate more severe fatigue. P values from week 4 to 20 are nominal. Hierarchical fixed sequence testing of key secondary endpoints was conducted in the order presented (a–d) at a two-sided significance level of 0.05. Error bars represent 95% CIs. Values of HR <1, OR <1, RR <1 and LS mean difference <0 favor benralizumab.