We also did a per-protocol analysis. Specifi cally, we
analysed the fi rst intravenous catheter per patient
remaining in situ on day 3 that was treated as per the
random allocation. Rates per 1000 days for this analysis
consisted of aggregated rate comparisons (incident rate
ratios) instead of hazard ratios (HR), we did not compare
groups for survival from phlebitis beyond day 3, since in
this analysis all intravenous catheters in the routine
resite group were removed on day 3. We calculated
intravenous catheter dwell times from this analysis since
only fi rst catheters were assessed prospectively as
required for 4 days or more, and to estimate the diff erence
in catheterisation that clinically indicated removal would
achieve. We retrospectively assessed power to detect
equivalence with the per-protocol analysis.