Translating Biomarker Discovery into Companion Diagnostics through Validation and Regulatory Consideration - Genomic Biomarkers for Pharmaceutical Development - page 162

co-development that should be addressed. The first of these concepts is the implementation and
utilization of the assay in clinical trial design.

7.4.2.1 Implementation of Companion Diagnostic in Clinical Design
To simplify the possibilities of how a diagnostic can be implemented in the clinical trial
setting, we will explain the three potential designs as noted in the draft Drug Diagnostic
Co-Development Concept Paper [3] , and discuss the benefits and drawbacks of each of the
proposed designs. The three designs are the classic two-arm clinical design, randomiza-
tion with the diagnostic, and the restricted approach. The classic two-arm clinical design is
depicted in Fig. 7.4 .
In this approach all subjects that meet enrollment criteria are randomized into either drug
or placebo arms regardless of their diagnostic status. All subjects are still evaluated at base-
line for the diagnostic status, but the diagnostic status is not utilized in the trial design. This
is the most classic of the design types and allows only for retrospective analysis and does
not allow for adaptive clinical trial design. In addition to only allowing a post-hoc analy-
sis approach, this strategy also opens up the investigator to the potential of enrolling the
majority of the signature positive or negative patients into the placebo arm. Without prior
understanding of the prevalence of the diagnostic positive and negative patient population,
this strategy would not be the preferred approach for this reason. While there are significant
drawbacks in this approach, there are also some benefits. For example, classical trial design
does not slow enrollment by waiting on the results of a diagnostic test. This will likely result
in more rapid patient enrollment, albeit at the risk of biased and unbalanced stratification
by diagnostic status in the treatment and placebo arms. In addition to the likely benefit of
shortened enrollment, this design also allows for collection of samples at baseline to be
tested later with the diagnostic assay, assuming stability of the analytes. This benefit is most
readily applicable when an assay has not reached the appropriate level of validation prior
to trial initiation. The design would allow time for validation of the assay to the appropriate
level prior to the testing of clinical samples. Finally, this trial design allows the investigator
to get a sense of the prevalence of diagnostic negative and positive subjects to be applied
to future clinical development. This trial design is most commonly applied early in clinical
development, and does not result in a significant confirmation of hypothesis testing due to
the potential for positive and negative subjects to be unevenly distributed in the treatment
and placebo arms. It is therefore not a recommended approach when seeking to test a diag-
nostic hypothesis.
The second approach is to randomize the trial by diagnostic status. A diagram of
this approach from the draft Drug Diagnostic Co-Development Concept Paper [3] is shown

All subjects
tested but
result not used
for randomization

Drug

All subjects

Placebo

FIGURE 7.4 A figure taken from the Drug Diagnostic Co-Development Concept Paper depicting a classical
two-arm clinical design in which patients are randomized into drug or placebo without regard for the diagnostic
call.