Sensitivity analyses
A number of sensitivity analyses were conducted on the primary outcome.
Treatment effects were also estimated using a linear mixed model with
autoregressive correlation. Post-hoc sensitivity analyses used an
autoregressive or exchangeable correlation structure with GEEs.
A further post-hoc sensitivity analysis was performed at all timepoints
using the same analysis as the primary outcome. Since the intervention
was only administered in a clinical setting, the outcome was whether a
repeat infusion was administered or not; any participants who did not
present for a follow-up appointment were coded as not having received a
repeat infusion. Finally, the difference in proportions of participants
who (1) received an infusion whether it was required or not and (2)
received an infusion per protocol was analysed.
As a secondary post-hoc analysis, the primary outcome was estimated
using all randomised participants to assess the possibility of bias and
the effect of reduced power after ineligible participants were excluded.