Design
#3 addresses the concerns inherent in Design #1, corrects
some of the potential sources of bias in Design #2,
and favorably increases the period of monitoring. The
only patients who will be studied will be those with
breast cancer, and treatment diffusion from study patients
to comparison patients is eliminated by the collection
of comparison data from another site. The trade-offs
in these approaches, namely concern about generalizability
of the results and concern about the use of different
data collectors and pre-existing institutional distinctions,
seem worth making given the likelihood of bias otherwise.
The sample size is not very large, but there are no
data available to inform a power calculation, and the
number seems reasonable. Follow-up data are collected
at specified time points and the follow-up is 6 months.
Like design #2, the recruitment will be done to ensure
that there are 50 patients per group with at least 4
months of follow-up. This will again allow both an "intent-to-treat"
analysis and a "completer" analysis. The improvements
in Design #3, which eliminate the possibility that a
different number and timing of outcome assessments will
bias the data, suggest that it is most likely to yield
interpretable data.