Improved understanding of cancer biology and advances in biotechnology bring us closer to the concept of personalized treatment of cancer. A key component of this new paradigm is development of biomarkers that can guide application of new and existing treatments. This requires a thorough understanding of the relationship between the biomarker and the treatment effect.
Traditionally, most randomized clinical trials (RCTs) focus on obtaining a reliable estimate of the average treatment effect in a broad patient population. Evaluation of biomarkers (and targeted therapies) often requires larger trials with more complex designs to provide a comprehensive assessment of the relationship between the biomarker and the treatment effect. However, in practice, clinical studies involve a delicate balance between the need for reliable evidence, the need to provide this evidence quickly, and feasibility. As we will discuss, achieving this balance in biomarker RCTs often requires a compromise between these competing considerations in both designing and monitoring these trials.
Biomarkers that are informative for clinical outcome can be broadly categorized as prognostic or predictive biomarkers. Prognostic biomarkers classify patients treated with standard therapies (including no treatment if that is standard) into subgroups with distinct expected clinical outcomes. The types of prognostic markers considered here are those for which the prognostic information has some implications for therapy decisions. For example, if the prognostic biomarker can identify a group of patients with very low risk of recurrence, additional treatment might not be considered, whereas higher-risk patients would be treated. Predictive biomarkers identify patients whose tumors are likely to be sensitive and/or resistant to a specific agent. For example, in advanced colorectal cancer, the benefit of cetuximab appears to be limited to patients with tumors that have the wild-type KRAS
). Note that biomarkers that predict toxicity to a certain agent are often treated as a separate type of biomarker. However, for the purpose of evaluating biomarker designs, we will consider toxicity biomarkers as a type of predictive biomarker (2
The purpose of this commentary was to provide a comprehensive comparison of the commonly used biomarker RCT designs. Ongoing or recently completed trials are used throughout the discussion as illustrative examples. Issues related to interim monitoring of biomarker trials are also discussed because standard futility and superiority monitoring may be inadequate due to the multiple subgroups and hypotheses being considered.