Table 1.

Methods of carrying out pharmacogenomic study in early-phase oncology trials

SettingMethods
General considerationsUniversally include collection and storage of germline and tumor DNA whenever feasible
Recognize that almost all pharmacogenomic findings in early-phase studies will need independent confirmation in a separate population, meaning endpoints will necessarily be exploratory or hypothesis generating
Consider the expected prevalence of a given marker in the study population prior to testing; if the prevalence is expected to be very low, associations will be unlikely, and testing is therefore unlikely to be fruitful
Recognize that testing of a large number of variants will limit statistical power to positively associate any single variant, once the penalty for multiple comparisons is properly applied
When possible, identified pharmacogenomic markers should be verified as independent predictive factors alongside other clinical factors influencing interindividual drug-response variability, such as organ function, disease stage and/or severity, performance status, or drug–drug interactions
Phase IChoice of variants to test might appropriately be informed by prior case reports or by preclinical information about a drug's metabolism, mechanism, and/or purported target
Consider focusing on pharmacokinetic phenotypes in relation to rational pharmacogenomic variants, because PK data usually will be collected, and because even small sample sizes can show significant pharmacogenomic–pharmacokinetic relationships
Consider focusing on toxicity pharmacogenomics, because disease heterogeneity of patients in phase I trials may limit drawing conclusions about tumor-related and/or response biomarkers
Phase IIMay be the first opportunity, at a fixed drug dose, to extensively evaluate pharmacogenomic variants related to a given drug
Choice of variants to test might appropriately be informed by prior case reports, phase I studies, or by preclinical information about a drug's metabolism, mechanism, and/or purported target
Consider pharmacogenomic evaluation of toxicity, not just response, because probability of finding a positive association may be higher for a toxicity phenotype (rather than a response phenotype) if the incidence of the toxicity is more common than the incidence of response; pharmacogenomic risk stratification for toxicity could mitigate concerns about moving a drug to phase III if the highest toxicity-risk patients could be preidentified and excluded
Randomized designs offer the advantage to specifically evaluate a marker with respect to a drug (predictive), rather than just the disease (prognostic), a step that is ultimately essential for a genetic marker to be categorized as pharmacogenomic
Other innovative designs (such as an enrichment, or tandem, 2-step designs; Fig. 3) may accelerate confirmation of a potential marker