Biomedical Engineering Reference
In-Depth Information
is relaxed to allow enrichment on other patient characteristics
thought to be positively associated with a biomarker of interest.) We
consider these goals in the settings of a single biomarker of interest
(Section 2.5.1) and when there are multiple biomarkers of interest
(Section 2.5.2).
2.5.1
Designs Involving Single Biomarkers
To assess whether a new therapy works in a biomarker-enriched
population (goal 1) using a well-defined biomarker, one can perform
a standard phase II trial restricted to this population. Some care is
required here: Whereas a single-arm phase II trial might normally be
appropriate in a certain setting, when the population is enriched the
historical targets may be inappropriate because the biomarker may
be prognostic, falsely leading one to conclude that the new therapy
is beneficial when any therapy might perform well on the enriched
population [27]. In this situation, or whenever a single-arm trial
would not be appropriate, a randomized phase II trial design can be
used with restriction to the enriched subpopulation. With either the
enriched single-arm or randomized phase II designs, a large number
of patients may need to be screened to obtain the required number
of biomarker-positive patients to be treated if the prevalence of the
enrichment subpopulation in the general patient population is low.
To determine if the new therapy should be developed for all
patients or an enriched population only (goal 2), both biomarker-
negative and biomarker-positive patients will need to be treated
with the new agent. This can sometimes be done retrospectively by
evaluating the biomarker on stored specimens of patients treated
with the new agent. For example, the first demonstrations that EGFR
mutation status was predictive of response to gefitinib were done
retrospectively and revealed that the responses were confined to
EGFR-mutation positive tumors [28, 29].
With a response rate endpoint and a single biomarker, a
prospective trial of a new agent could again compare response rates
in the biomarker-positive and biomarker-negative subgroups. One
might want to consider enrolling a sufficient number of biomarker-
positive patients for evaluation in that subgroup. This might be done
in the context of a two-stage design that enrolls differing numbers
of patients in the subgroups in the second stage of accrual based on
the number of responders in the two subgroups in the first stage [30,
