The impact of genomic variation on drug response may require investigation in clinical studies.
For genomic variation that may affect the PK of a medicinal product, clinical studies should be designed in accordance with the Guideline on the use of pharmacogenetic methodologies in the pharmacokinetic evaluation of medicinal products.
Genetic variants may be evaluated in clinical studies based on biological plausibility or non-clinical (in vitro or in vivo) findings. Alternatively, a predictive genetic variant may be identified for the first time during clinical studies.
A predictive genetic variant may be a single marker or be comprised of a multi-marker signature or an algorithm. Predictive genetic variants may be binary, categorical or continuous. For continuous genetic variants (e.g. gene expression levels), the test result is usually dichotomised using a pre-defined cut- off; in which case the predictive value is a function of the cut-off.
Selection bias may affect the apparent predictive value of the genetic variant at any stage of clinical development. Bias may arise due to non-availability of biological samples from all patients, the use of convenience samples (e.g. from certain centres or larger tumours), assay failure or the practice of pre-screening. Sources of selection bias should be minimized.
The chosen biomaterial should be appropriate for the study objective. All sources of variability (technical or biological) should be identified and controlled for. The analytical validity of the assay should be demonstrated as early as possible during clinical development. This is particularly important when the genetic variant is used to include/exclude patients or allocate treatment.
Analytical validity at an early stage also optimizes the relevance of early clinical findings to later clinical development.
It is recommended to seek scientific advice on the proposed development strategy.
Exploratory stage
Exploration of candidate predictive genetic variants can be part of both early and late phases of clinical development. Early phase clinical studies are often small single arm cohort studies. Such study designs may provide evidence of an association between genetic variability and drug response. However, a control arm may be required to confirm that a genetic variant is predictive of drug response rather than prognostic. Selection bias may also be a particular problem for early phase studies.
The design of clinical studies to explore candidate predictive genetic variants may depend on the type of genetic variant being investigated: single binary, continuous (requiring exploration of different cut- offs) or candidates for multi-marker signatures. In addition, exploration may be conducted in the context of an umbrella or basket design. Furthermore, flexibility may be incorporated by use of an adaptive design which allows for pre-planned interim analysis and design modification (see EMA Reflection Paper on Methodological Issues in Confirmatory Clinical Trials Planned with an Adaptive Design – section 3).
Genetic variants predictive of drug response may also be identified by retrospective analysis of large clinical datasets including randomised controlled trials (RCTs), for example using a genome wide association study (GWAS) approach. In the case of GWAS, due to multiplicity issues, only very strong associations are likely to be reproducible during subsequent development. Furthermore, only relatively common genetic variants >1% can be identified using GWAS. This approach has been used to identify genetic variants predictive of toxicity.
Confirmatory stage
Confirmation of the clinical validity of a predictive genetic variant involves replication in an independent cohort of the observed genotype-dependent outcome. The frequency of the genetic variant in the target population and the outcomes of interest will drive the design of the study, including analyses to confirm the clinical relevance and utility of the genetic variant. In some cases, the pivotal trial for clinical efficacy and safety can provide sufficient evidence of clinical validity.
In some cases, confirmation of the clinical validity of the genetic variant could be based on retrospective data rather than prospective studies, provided the confirmatory dataset is distinct from that used in the exploratory stage.
To investigate the clinical validity of a genetic variant, a stratified RCT should be considered. Patients with or without the genetic variant are included, and randomization is stratified by variant status. This design allows measurement of the capacity of the genetic variant to separate patients into those more likely or less likely to demonstrate a drug response, within the same study. The analysis of treatment effect can be conducted in the overall pooled population as well as in the populations defined by the genetic variant; the primary analysis population should be pre-specified in the statistical analysis plan.
For continuous genetic variants, the cut-off point for determination of variant status should be pre- specified and adequately justified.
In a targeted (enrichment) design, study inclusion is dependent on genetic variant status, and therefore it is not possible to confirm the clinical validity of the genetic variant. However, it may be possible to demonstrate a drug response in the presence of the genetic variant. This design is only appropriate if there is strong biological plausibility, and compelling non-clinical and/or clinical evidence, that drug response is likely to be variant-dependent. An enriched design may also refer to a variation on the stratified design in which a small number of variant negative patients are included; in this case, it may be possible to confirm the clinical validity of the genetic variant.
For a genetic variant that might predict a rare severe adverse drug reaction, a case control design may be more applicable than a prospective RCT.
Read also: DNA Analytics for Genomic Study