Biomedical Engineering Reference
In-Depth Information
In situations where there is not very much preliminary evidence
that the biomarker will be predictive for the experimental treatment,
an alternative to analyzing treatment effects in the biomarker
positive and negative subgroups is to analyze the treatment effect
in the biomarker-positive subgroup and in the whole randomized
population. This strategy is focused on broadening the population for
which the benefit of the new therapy can be formally demonstrated.
An example of a biomarker-stratified design using this alternative
analysis approach is the S0819 trial [44], which is testing whether
the addition of cetuximab to chemotherapy improves outcomes for
patients with non-small cell lung cancer. The biomarker is EGFR
status, and the co-primary analyses involve the whole randomized
population and the EGFR-positive subgroup.
2.6.2
Enrichment Designs
If there is strong scientific evidence that the benefit of the new
treatment is likely to be limited to only patients whose tumors
harbor a certain biomarker, it is appropriate to restrict eligibility to
that biomarker-defined subset of patients. This design, displayed in
(Fig. 2.4), is called an enrichment design. It is motivated by ethical
considerations to avoid treating patients who are thought unlikely
to benefit from the new therapy. It can be a far more efficient design
than one with broad eligibility that analyzes all patients together
regardless of marker status. If a large number of patients who will
not benefit from the new therapy are randomized, this will mask the
treatment benefit in the biomarker-positive subset when an analysis
is performed in the entire patient cohort. A larger number of patients
would have to be randomized to detect this diluted treatment effect
in the overall population.
In practice, it can be difficult to decide when the background
information is sufficiently compelling to exclude biomarker-
negative patients from a trial. Prematurely excluding some patients
could unnecessarily limit the indication for the new therapy and
deny some patients access to a therapy that could benefit them. For
example, there could be off-target effects of a new targeted agent
that would be missed using an enrichment design. In addition, a
positive trial cannot directly prove the utility of the biomarker, and
prognostic effects of the marker cannot be assessed using this design.
Nonetheless, the enrichment design does allow one to definitively
