Article Figures & Data
Figures
- Figure 1.
A, published studies of antineoplastics incorporating the term “polymorphism,” 2001 to 2010 (includes phase I, II, and III trials, plus observational studies or meta-analyses). It is acknowledged that some additional retrospective and case–control studies of oncology pharmacogenomics were likely not captured by our specific MEDLINE search criteria. B, the probability of “success” for showing pharmacogenomic endpoints differs substantially between published phase I and phase II trials. Interestingly, the positive pharmacogenomic results reported by 2 of the 3 “pharmacogenomically positive” phase I trials (34, 35) were, indeed, also reported as positive in separate phase II trials (36, 37). However, it is not clear that either of the findings were refined in phase II. In 1 case, the positive phase I finding (because it was reported as part of a phase I drug combination study; ref. 34) was published after the phase II positive finding (36). In the other case, the phase I positive finding was published first (35), but the separate phase II positive finding (published only 12 months later; ref. 37) showed the allele having the exact opposite direction of effect on the phenotype (progression-free survival). Success of contemporary pharmacogenomic analyses from phase III studies is also shown for comparison. Note that in our analyses, for the purposes of this illustrative examination, we allowed the investigators to define “success,” meaning we simply reported here whether the variant being evaluated was associated in a “positive” or “negative” manner with the phenotype being studied by the authors (typically, tumor response, survival, or drug toxicity). It is important to state that in some of the reportedly “positive” trials, this does not necessarily mean that the implicated pharmacogenomic marker–phenotype relationship meets the statistical rigors of multiple testing; we simply reported whether the authors designated an association as positive or negative.
- Figure 2.
Common-arm approaches analyzing 2 or more common drug-treatment arms from different clinical trials can provide a rich, large data set for pharmacogenomic discovery. Common-arm approaches take advantage of the fact that several different independent clinical trials (all of which could be phase II trials) may be testing a similar common (or “standard”) arm against various other arms. The investigator may be interested in a pharmacogenomic variant related to the agent used in the common arms, rather than the comparator arms. As shown, a given cancer is being tested for the development of 2 different novel therapies: drug A and drug B. In 1 randomized trial, novel drug A is being compared with the existing standard therapy. In the separate trial, novel drug B is being compared against the existing standard therapy. The existing standard therapy is, however, the same for both studies. Therefore, by combining the patients from the 2 standard therapy “common” arms from these 2 independent trials, one doubles the number of patients in which to examine a potential pharmacogenomic marker(s) for the standard therapy. In addition, the randomization feature allows one to draw conclusions that the studied pharmacogenomic marker is, indeed, potentially predictive for the therapy in the standard arm, because the novel therapy arms can serve as “control” populations having the same disease.
- Figure 3.
Tandem, 2-step phase II study design for evaluation of pharmacogenomic biomarkers. The hypothesis is that a drug will show a higher response rate in selected patients carrying the susceptibility marker. In the first stage of the tandem design, all patients meeting eligibility criteria are treated, and an early-stopping role is instituted. If adequate responses were seen at the early-stopping analysis point, then the second stage of the study would be completed with no selection of patients by the presence of a pharmacogenomic marker. If too few responses are seen in the first stage, patients will proceed on to a “limited eligibility” stage of the trial, in which only “marker-positive” patients would be enrolled and the question would be asked whether a prespecified response rate (or other outcome) is or is not met in this selected population. If too few responses are again seen, that pharmacogenomic marker would be considered invalid. If an adequate number of responses are seen in the marker-positive patients of the 2-step design, then that second stage of the study would be completed, possibly allowing a drug to be defined for a select patient population having that pharmacogenomic marker.
Tables
- Table 1.
Methods of carrying out pharmacogenomic study in early-phase oncology trials
Setting Methods General considerations Universally 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 I Choice 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 II May 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
Volume 18, Issue 10
Jump to section
- Article
- Abstract
- Introduction
- Oncology Clinical Trial Incorporation of Pharmacogenomics Is Rapidly Growing
- Pharmacogenomics in Early-Phase Trials
- Phase II Example
- Randomization in Phase II Trials
- Pharmacogenomic Phenotypes in Phase II Trials: Response Compared with Toxicity
- Study Population Challenges When Incorporating Pharmacogenomic Endpoints in Phase II Trials
- What's Next after a Positive Pharmacogenomic Finding in Phase II?
- Conclusions
- Disclosure of Potential Conflicts of Interest
- References
- Figures & Data
- Info & Metrics