Biomedical Engineering Reference
In-Depth Information
2.6.3
Biomarker-Strategy Designs
In the biomarker-strategy design, patients are randomly assigned
to a control-arm standard treatment
or a biomarker-directed
treatment arm. In the simplest case of a binary biomarker, patients
in the biomarker-directed arm receive treatment
B
A
if the biomarker
is positive and treatment
if the biomarker is negative (Fig. 2.5).
At first glance, this type of design has appeal because it provides
a direct estimation of benefit from using the biomarker in clinical
practice (the biomarker-directed treatment arm) relative to the
standard treatment (the control arm). However, the design is not
efficient because many patients on both arms will be treated with
the same therapy, treatment
B
. Furthermore, one can use the arm-
specific outcomes of a biomarker-stratified or enrichment design to
estimate the same between-arm difference that would be seen in a
biomarker-strategy design, but with a smaller required sample size.
For example, suppose the biomarker positivity rate was one-half
and one was considering performing a trial with 800 patients using
a biomarker-strategy design. A calculation shows that one could
perform a biomarker-stratified design with 400 patients (or screen
400 patients and randomize 200 patients in an enrichment design)
to be able to detect a given treatment effect with the same power as
the 800 patient biomarker-strategy trial.
B
Biomarker-
directed
arm
R
a
n
d
o
m
i
z
e
Biomarker positive
Treatment A
Biomarker negative
Treatment B
Assess
biomarker
Control
arm
Treatment B
Figure 2.5
Biomarker-strategy design. Treatment A is directed toward
the biomarker and Treatment B is standard therapy. Patients
randomized to the biomarker-directed arm receive Treatment
A if they are positive for the biomarker and Treatment B
otherwise. All patients randomized to the control arm receive
Treatment B.
Although we see no advantage to using the biomarker-strategy
design in the simplest case (Fig. 2.5), does it have a role when the
biomarker is choosing among many possible treatments rather
than just two? With, say, a dozen treatments being chosen by the
