Asymmetric two-sided group
sequential design with binding futility bound, 2 analyses, time-to-event
outcome with sample size 2100 and 140 events required, 80 percent power,
2.5 percent (1-sided) Type I error to detect a hazard ratio of 1 with a
null hypothesis hazard ratio of 1.63. Enrollment and total study
durations are assumed to be 5 and 10 years, respectively. Efficacy
bounds derived using a Lan-DeMets O’Brien-Fleming approximation spending
function. Futility bounds derived using a Lan-DeMets O’Brien-Fleming
approximation spending function.
An interim analysis will also be planned once 50 patients are accrued in
each arm and in this analysis, we will be evaluating the acute toxicity
of the two interventions. The sample size of 50 patients in each arm
allows us 80% power to exclude a within-group rate of CTCAE Grade 3 or
more acute skin toxicity exceeding 10% over the target rate of 5%
(one-sided type I error of 5%).
Methods for additional analyses (e.g. subgroup analyses)
{20b}
The following pre-specified subgroup analyses will be conducted on the
primary and the other time to event endpoints stratified by:
- Menopausal Status: Premenopausal vs Postmenopausal
- Nodes Positive: None vs 1 - 3 vs > 3
- Subtype: Luminal vs Her2-enriched vs TNBC
- Regional Nodal Radiation: None vs SCF alone vs IMN + SCF
- Boost Type: SIB vs Sequential Boost
These subgroups are chosen as they have an implication on the primary as
well as the secondary time to event outcomes and are independently
prognostic. Results will be presented as forest plots with interactions
results alongside. The interaction test will test if the treatment
effect is modified by the subgroup.
In order to evaluate the impact of prognostic factors on the treatment
effect, a nomogram will be built using Cox regression that includes
treatment as a factor. The use of this nomogram will enable the user to
judge the benefit arising from the use of a particular fractionation
regimen in a patient with a given combination of prognostic factors.
Methods in analysis to handle protocol non-adherence and any
statistical methods to handle missing data
{20c}
For time to event endpoints, we will censor the patient at the date of
last follow up unless the patient is documented to have an event. We
will attempt to contact the patients telephonically or by other methods
to ensure adherence to follow up. If more than 10% of the patients have
missing data then an additional sensitivity analysis will be undertaken
where all lost to follow up patients would be assumed to have the event
of interest. For quality of life data, we will under complete case
analysis and as a sensitivity analysis missing values will be imputed
using an imputation based approach.
Finally, as this is a non-inferiority trial, if there is high
non-compliance with the test treatment, then an analysis of only those
receiving per-protocol treatment will be conducted separately.
Plans to give access to the full protocol, participant
level-data and statistical code
{31c}
The full protocol as well as the statistical analysis plan is available
as supplementary material with the manuscript. At present there are no
plans to make patient level data available publically but we will
consider the same once primary reporting of the outcomes is completed or
as mandated by regulations.