Biology Reference
In-Depth Information
to perform with the same accuracy on an independent set of patients. Further, among those
genes in the function, there would exist a large amount of mutual or correlated informa-
tion, thereby providing redundancy in explaining the separation between the two patient
populations. To address this issue, statistical methods have been developed to remove those
variables that do not significantly contribute to the decision rule or classifier and are highly
correlated with other variables
[21]
. Often, these types of methods can greatly reduce the
analytes in the model to the most predictive set and still maintain a very similar prediction
performance. Various error functions have been used to determine this cost
/
beneit ratio in
the optimization of the best predicting variable set.
7.4.2.11 Instrument
In order to ease transition and prevent additional validation work, it is recommended to
utilize instrumentation and associated software that has already been through the clearance
process and has obtained 510k clearance. This will reduce the regulatory hurdles encoun-
tered in a first-time submission of a platform and associated software. This can be either a
trivial decision, or much more complex, as most commercial diagnostic development part-
ners are closely tied to development on equipment for which they own development rights.
For example, Qiagen develops assays on their proprietary platform, the Rotor-Gene Q
3
. Early
development on the ultimate clinically applicable platform can offer significant advantages
such as decreased bridging, early understanding of performance characteristics, and early
optimization of assay format. If it is possible to understand the ultimate commercial partner
for the assay early in assay development, the appropriate platform can be utilized to avoid
assay transfer and migration to a different platform late in the development process. In addi-
tion to the regulatory and sample requirement considerations of the assay platform, the
investigator must also consider the footprint of the instrument utilized in the device. For
example, if a diagnostic assay is built on an instrument which is not present in the geographi-
cally necessary clinical testing laboratories, there will be additional delays to test distribution
as these laboratories evaluate the risks and benefits of procuring the new platform. With this
in mind, it is wise to understand the footprint necessary for rapid adoption of the particular
companion diagnostic assay of interest and ensure that this is either already present or
appropriately planned to meet that footprint in advance of device approval.
7.4.2.12 Reagent Status
In the pre-IDE setting, the FDA closely evaluates each of the specific reagents included in
the assay. For each reagent, the manufacturing status must meet that of a commercial assay
prior to approval. This means that each assay component must be manufactured to Good
Manufacturing Practices standards. In the past, the FDA had allowed use of what was termed
analyte-specific reagents, but only in the context of laboratory developed tests (LDTs), which
we now know from guidance not to be an appropriate regulatory pathway for companion
diagnostics. In addition to ensuring the appropriate level of manufacturing compliance, the
investigator must identify critical reagents in the assay. For these critical reagents and compo-
nents, the FDA looks for alternative manufacturers or risk mitigation strategies in the scenario
where a manufacturer may discontinue the production of a certain component. Therefore,
Search WWH ::
Custom Search