Biomedical Engineering Reference
In-Depth Information
assessed by using a statistical test to determine if the treatment-
by-marker interaction is statistically significantly different from
zero. If the interaction test is significant, it can be assessed whether
the treatment effect in the two biomarker subgroups point in the
same direction but differ in magnitude (quantitative interaction),
or if the treatment effects point in opposite directions (qualitative
interaction).
R
a
n
d
o
m
i
z
e
Treatment A
Biomarker
positive
Treatment B
Stratify on
biomarker
Assess
biomarker
R
a
n
d
o
m
i
z
e
Treatment A
Biomarker
negative
Treatment B
Figure 2.2
Biomarker-stratified design. Treatment A is directed toward
the biomarker and Treatment B is standard therapy. The
biomarker is used to stratify treatment randomization but
does not influence what treatments patients receive.
Qualitative interactions are far more useful than quantitative
interactions because they can provide a clearer indication of optimal
choice of therapy for the biomarker-positive and biomarker-negative
subgroups. Because the treatment differences point the same
direction in both biomarker subgroups in the case of a quantitative
interaction, one might conclude that all patients should receive the
same (superior) treatment, regardless of biomarker status. However,
it is important to consider whether a more modest treatment benefit
in the biomarker-negative subgroup is still worthwhile if there are
added toxicities or other costs of the new treatment. This must
be balanced against the realization that if it is decided to forego
the modest benefit of the new therapy in the biomarker-negative
subgroup, a cost will be incurred to perform the biomarker testing
to identify the biomarker-negative patients. Further complicating
