6. Preanalytic Procedures and Testing Protocols
646
Many IVD companion diagnostics require a number of preanalytic steps to prepare the
647
analyte(s) for measurement (e.g., tissue fixation, DNA and RNA extraction, melanin
648
removal, whole genome amplification, bisulfite modification). Preanalytic reagents and
649
instrumentation are typically considered to be part of the test system and should be validated
650
with the IVD.
651
652
Variations in preanalytical steps at different testing sites may make it difficult to interpret
653
analytical performance studies. Thus, for all steps of preanalytical specimen handling and
654
preparation, sponsors should have a detailed standard operating procedure (SOP) or protocol
655
that is followed at each site that performs any of the preanalytical steps. The sponsor should
656
ensure that all sites handling the specimens are trained to use the specific method, follow the
657
SOPs, and record any deviations from the SOP.
658
659
FDA bioresearch monitoring (BIMO) personnel may, and in some cases (e.g., when a PMA
660
for an IVD is under review) generally do, examine laboratory records to determine whether
661
protocols have been followed (see also Section III. F.1.iii. of this guidance). In cases where
662
there is significant and/or uncontrolled deviation from the specimen testing protocol , FDA
663
may be unable to approve the regulatory submission because it may deem the data derived
664
from poorly controlled testing to be unreliable and non-representative of the IVD companion
665
diagnostic’s performance under its proposed instructions for use.
666
667
7. Planning Ahead for Analytical Validation Studies
668
The IVD sponsor should consider the types of studies needed for analytical validation to
669
Contains Nonbinding Recommendations
Draft - Not for Implementation
20
support marketing authorization of an IVD companion diagnostic and plan accordingly.
56
670
For example, if the analyte is labile, a plan to collect several specimens from a small
671
number of subjects to assess lability to inform appropriate limitations on storage and
672
transport durations may be appropriate. Note that some analytical validation studies may
673
not require use of samples from therapeutic product clinical trial subjects, although the
674
studies should be conducted with samples from the same target population to ensure that
675
the variability parameters defined are relevant to the population to be tested.
676
677
It is important to ensure that appropriate specimens are collected and banked (where
678
analyte stability allows) in sufficient quantities and maintained adequately to support the
679
full range of analytical studies. Collecting the appropriate pathologic-based annotation
680
(e.g., tumor content, necrosis, adiposity, presence of large amounts of stroma, and other
681
characteristics) for the samples may help to support conclusions about the performance of
682
the assay. Appendix 2 provides additional detail on specimen handling considerations.
683
684
In cases where multiple markers will be detected/measured by the test, analytical validation
685
of each reported marker may be required regardless of each marker’s prevalence. When it is
686
not possible for sponsors to obtain specimens containing a particular marker, validation
687
studies with contrived samples may be permitted.
57
Analytical validation studies may also be
688
complicated for IVDs that have the potential to detect a very large number of markers, in
689
which case it may be necessary for the study to use a representative sampling of markers.
690
For example, for next generation sequencing panels, the ability of the IVD to detect single-
691
nucleotide polymorphisms, copy-number variations, inversions or deletions, and other
692
relevant variant classes should be studied. Sponsors who are concerned about the feasibility
693
of conducting analytical validation studies for all markers detected by an investigational IVD
694
should consult with FDA before beginning sample collection and analytical validation
695
studies.
696
D. Therapeutic Product Clinical Trial Design Considerations
697
When planning therapeutic product clinical trials designed to rely on information
698
provided by an IVD, whether for enrollment, stratification, dose, or other uses, sponsors
699
should consider clinical trial designs that can be used to support the claims for both the
700
therapeutic product and IVD companion diagnostic, and consider whether the IVD
701
companion diagnostic development strategy is aligned with the approval goals for the
702
therapeutic product.
703
704
Understanding the population of subjects enrolled in a clinical trial is critical. It is
705
conceivable, for example, that assessment of preclinical or early clinical studies indicates a
706
56
Sponsors may find it helpful to consider resources on analytical validation studies, e.g., Mansfield, E., et al.
“Biomarkers for pharmacogenetic and pharmacogenomic studies: Locking down analytical performance.”
Drug Discovery Today: Technologies. 2007, Vol. 4, No. I, pp. 17-01.
57
For example, see FDA guidance “Guidance on Pharmacogenetic Tests and Genetic Tests for Heritable
Markers”
(
http://www.fda.gov/MedicalDevices/DeviceRegulationandGuidance/GuidanceDocuments/ucm077862.htm
).
Contains Nonbinding Recommendations
Draft - Not for Implementation
21
therapeutic product may be beneficial in the test-positive subgroup
58
and harmful in a test-
707
negative subgroup. In such cases, subjects with false-positive results may be harmed by the
708
therapy, and subjects with false-negative results may be deprived of beneficial therapy.
709
Additionally, false-positive results could lead to underestimation of effect size, whereas
710
false-negative results could lead to underestimation of the proportion of subjects who are
711
more likely to respond. Therefore, the therapeutic product and IVD sponsors should work
712
closely to understand how the IVD’s analytical performance affects the selection of subjects
713
in the trial. To minimize the proportion of incorrect test results (i.e., false positives and false
714
negatives that would result in misclassification),
59
sponsors should ensure that the
715
appropriate analytical validation studies are carried out and that the level of analytical
716
validation of the proposed IVD(s), in relation to its specific role in the clinical trial, has been
717
adequately assessed. This is especially important when progressing from the versions of the
718
test used in a trial to the candidate IVD companion diagnostic (see Section III.E.3. of this
719
guidance).
720
721
Sponsors should also be aware of, and plan to address, potential sources of bias or error
722
associated with IVD development such as prescreening, preanalytical processing
723
(discussed in Section III.C of this guidance), and bridging studies when necessary (see
724
Section III.E of this guidance).
725
726
The following sections discuss considerations for the design of clinical trials for a
727
therapeutic product for use with a developmental IVD companion diagnostic.
728
729
1. General Considerations for Early Therapeutic Product
730
Development
731
Performing tests for exploratory purposes (referred to as exploratory testing) to identify
732
potential biomarkers in early therapeutic product development may lead to a codevelopment
733
program. Sponsors should be aware that using exploratory testing that is not sufficiently
734
analytically validated or is validated with inappropriate analysis methods may produce
735
spurious associations.
60
This could result in the failure of a codevelopment program if, for
736
example, a late-phase clinical trial enrolls only “marker-positive” subjects, when positivity is
737
based on flawed exploratory programs. When using exploratory testing, it is advisable for
738
sponsors to establish procedures that specify the process for sample acquisition and handling
739
58
Note that the terms “test-positive” and “test-negative” are often used interchangeably with the term “marker-
positive” and “marker-negative;” however, it is important to be aware that tests for the same marker that have
different performance characteristics may identify different subpopulations of “marker-positive” patients.
59
For example, molecular tests that are intended to select for one target but have undetected cross-reactivity
with other targets may result in selection of a substantial number of patients with the cross-reactive target but
not the target of interest.
60
Sponsors should consider principles laid out in the National Cancer Institute publication, “Criteria for the use
of omics-based predictors in clinical trials,” McShane, et al., Nature. 2013, Vol 502, pp. 317-320; and FDA
guidance for industry “Clinical Pharmacogenomics: Premarket Evaluation in Early-Phase Clinical Studies and
Recommendations for Labeling”
(
http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM337169.pd
f
).
Contains Nonbinding Recommendations
Draft - Not for Implementation
22
and the testing and analysis plans so that the preliminary evidence that is generated is most
740
likely to be informative.
741
742
Some early therapeutic product clinical trial designs employ testing for multiple markers to
743
assign subjects to one of multiple different therapeutic arms with the goal of testing multiple
744
hypotheses under one study protocol. Sponsors of these clinical trials should consider the
745
pathway for continued development of selected therapeutic products with accompanying
746
IVDs in the event that such trials support further development of a candidate IVD companion
747
diagnostic.
748
749
2. General Considerations for Late Therapeutic Product
750
Development
751
When a clinical trial is properly designed to establish the safety and effectiveness of a
752
therapeutic product in a population based on measurement or detection of a marker, the
753
results of the clinical trial can also be used to establish the clinical validity of the IVD
754
companion diagnostic.
61
There are a variety of clinical trial designs that may be used to
755
study a developmental IVD companion diagnostic in combination with a therapeutic
756
product in premarket codevelopment programs. The appropriate clinical trial design to
757
support the diagnostic strategy depends on the proposed claim(s) for the IVD and what
758
has already been established about the predictive, prognostic, or other critical properties
759
of the marker.
62
The success of a clinical trial design strategy depends on many factors,
760
including but not limited to the following: a) the characteristics of the marker as applied
761
to the target population for whom the therapeutic product will be indicated, specifically
762
the mechanistic rationale for selecting the marker, its predictive/prognostic/other utility
763
and its intrinsic properties (e.g., variability and specificity with respect to the disease); b)
764
the nature of the disease; and c) the need to fully characterize the therapeutic product’s
765
benefits and risks, such as the safety profile (e.g., taking into account a possible lack of
766
benefit in the test-negative population), and the degree of observed benefit, if any, in the
767
population for whom the therapeutic product may not be indicated (e.g., test-negative
768
subjects).
769
770
Two marker-based clinical trial designs that are commonly used are illustrated in Figure
771
1; however, other designs could be appropriate and should be discussed with the
772
appropriate therapeutic product review center.
63
773
774
61
For IVDs, clinical validity typically refers to the accuracy with which the test identifies, measures, or predicts
the presence or absence of a clinical condition or predisposition in a patient. In the case of an IVD companion
diagnostic, clinical validity typically refers to the accuracy with which the test identifies the patients for whom
use of the therapeutic product is safe, effective, or both.
62
See Section III.D.3. and Section III.G.1 for additional discussion of predictive and prognostic markers.
63
For additional trial designs and further discussion, please also refer to FDA draft guidance “Enrichment
Strategies for Clinical Trials to Support Approval of Human Drugs and Biological Products”
(
www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM332181.pdf
).
FDA draft guidance represents FDA’s proposed approach on this topic. When final, this guidance will
represent the FDA’s current thinking on this topic.
Contains Nonbinding Recommendations
Draft - Not for Implementation
23
Figure 1. Clinical Trials Involving Markers. Trial design A, called an interaction or
775
biomarker-stratified design, is designed to evaluate treatment and marker effects, and
776
their interaction, by stratifying randomization based on marker status, as determined by
777
an IVD. Trial design B, called a targeted or selection design, is designed to evaluate
778
treatment effects in a targeted population by selecting only those who are test-positive.
779
Key: test-positive, +; test-negative, -; randomize, R. Treatment A is typically the
780
experimental arm and Treatment B is typically standard-of-care or placebo.
781
782
783
In many efficacy trials, it is generally desirable to obtain information about the safety and
784
effectiveness of the therapeutic product for all subjects (rather than for only those
785
subjects with a particular marker status), to ascertain the appropriateness of restricting the
786
therapy to a patient population on the basis of a marker. However, this does not mean all
787
subjects, regardless of marker status, should be randomized. The study could enroll
788
marker-positive subjects and include only a sample of marker-negative subjects, e.g.,
789
when marker-positive subjects are only a small percentage. Testing for the presence of
790
particular markers may provide information on prognosis, prediction of response (i.e.,
791
response, non-response, or toxicity), or both.
64
The clinical trial design depicted in
792
Figure 1A, in which both test-positive and at least some test-negative subjects are
793
enrolled and randomized, is the most informative design because treatment by marker
794
interaction, as well as the prognostic versus predictive value of the marker, can be
795
assessed. This approach may be particularly valuable when the biological plausibility or
796
medical relevance of the biomarker is not well understood (e.g., based on findings from
797
exploratory studies or post-hoc analyses in other trials). Other variations on this design
798
exist, such as those including interim futility analysis where, for example, further
799
enrollment could be limited to test-positive subjects if harm or lack of efficacy is
800
64
A purely predictive marker will predict that patients, given a particular marker status, will have better or
worse outcomes than patients without the marker, solely as a result of having received the investigational
therapeutic product; that is, there is a clear therapy-marker interaction. A prognostic marker would suggest that
patients with the marker would, as a consequence of the natural history of the disease, have better or worse
outcomes even absent treatment with the investigational therapeutic product; that is, the marker has little or no
interaction with the therapy. Some markers may have both predictive and prognostic properties in a given
disease/therapy setting. For example, the presence of HER-2 protein overexpression indicates a poorer
prognosis in patients with breast cancer than in patients who do not overexpress HER-2, but the same marker
also predicts greater likelihood of response to the drug trastuzumab (Herceptin). Thus, it is important to
understand the role the marker is expected to play in the therapeutic product trial. The prognostic value of the
marker, if unknown at the time of the therapeutic product trial, should be assessed in clinical trials that are
stratified by marker status.
Contains Nonbinding Recommendations
Draft - Not for Implementation
24
identified in the test-negative population.
65
801
802
In the approach depicted in Figure 1B, only a subgroup identified by the marker status is
803
enrolled (e.g., only subjects deemed positive by the test are enrolled into the clinical
804
trial). With this design, the predictive value of the test cannot be determined because
805
there is no information on the treatment effect in the test-negative population. Likewise,
806
there is no information about whether the assigned assay cutoff adequately distinguishes
807
those who will respond from those who will not. FDA does not object to this approach
808
categorically because it may be appropriate in some situations (see also Section III.D.3 of
809
this guidance). A modification of the design, however, could stratify by assay cutoff.
810
811
Sponsors planning to evaluate the safety and effectiveness of a therapeutic product only
812
in a subset of subjects identified by an IVD should consider whether there is persuasive
813
evidence (e.g., evidence from strong preclinical data, preliminary clinical data, or from
814
clinical trials with similar therapeutics) for the marker as a predictive measure of
815
response or non-response. Although the sponsor may select any cutoff , FDA
816
recommends that sponsors choosing a marker-positive only approach assure that the
817
chosen marker and assigned assay cutoff are relevant to the disease under study (i.e.,
818
known prevalence of marker positivity in the general patient population) within the
819
context of likelihood of a subpopulation’s response (e.g., biologic plausibility,
820
mechanism of action), and that sponsors make a persuasive case for use of the IVD to
821
identify patients who are to be treated.
822
823
3. Prognostic and Predictive Markers
824
In clinical trial designs, prognostic markers can be used either to identify the population
825
to be enrolled or to stratify treatment randomization. For putative prognostic markers, no
826
difference in the effect size is expected in marker-negative versus marker-positive
827
subjects. Effect size may be measured in different ways, depending on the clinical trial.
828
In oncology trials with time to death as an endpoint, a hazard ratio may be used.
829
Potential study designs for markers expected to be predictive of therapeutic response are
830
discussed elsewhere.
66
831
832
With respect to a predictive marker, the clinical trial can stratify by the marker test result
833
and randomly assign subjects with the same marker status to the experimental treatment
834
and control (Figure 1A). If there is little possibility of any effect in marker-negative
835
subjects, however, only marker-positive subjects might be randomly assigned to
836
treatment (Figure 1B), but this provides no formal test of whether the marker predicts
837
65
See note 63. Sponsors may also find it helpful to consider resources on this topic, e.g., Wang SJ, O’Neill RT,
Hung HMJ. “Approaches to evaluation of treatment effect in randomized clinical trials with genomic subset.”
Pharmaceutical Statistics Vol. 6, pp.227-244.
66
See note 63. Additionally, sponsors may find it helpful to consider resources on clinical trial designs, e.g.,
Fridlyand, J. et al. “Considerations for the successful co-development of targeted cancer therapies and
companion diagnostics.” Nat Rev Drug Discov. 2013. Vol. 10, pp. 743-55; Temple, R. “Enrichment of clinical
study populations.” Clin Pharmacol Ther. 2010. 88(6), pp. 774-8.
Contains Nonbinding Recommendations
Draft - Not for Implementation
25
treatment benefits only in such marker-positive subjects. In clinical trial designs depicted
838
in Figure 1 above, for a continuous marker for which a firm cutoff has not been
839
determined, there could be randomization at varying degrees of marker positivity, or less
840
formally, there could be a post-hoc analysis of the treatment effect at a range of cutoff
841
values. As noted, if the marker is both prognostic and predictive, then post-hoc analyses
842
of response by marker positivity in the clinical trial designs depicted in Figure 1A or 1B
843
are likely to be confounded, and stratification by degree of marker positivity is strongly
844
recommended.
845
846
Dostları ilə paylaş: |