Fellow summary authored by James Smithy, MD, MHS

Issues in Design of Translational Research Trials

Dr. Susan Hilsenbeck presented a high-yield discussion of clinical trials involving translational endpoints, with particular attention to the incorporation of biomarkers in trial design. At the outset, Dr. Hilsenbeck explained that the incorporation of biomarkers into trials adds another level of complexity in their interpretation—with a negative result, it may be that either the intervention is ineffective or the biomarker is ineffective at distinguishing responders. All biomarkers need to undergo multiple steps of validation before they are adopted into routine clinical practice. Analytic validation demonstrates that the biomarker can be measured in a reproducible manner. Clinical validation demonstrates that the biomarker is associated with a specific pathophysiologic state. Lastly, clinical utility demonstrates that incorporating the biomarker into practice leads to better outcomes for patients in terms of survival or other metrics. Several types of biomarkers were defined: those associated with an increased risk of developing a condition (e.g., BRCA1 germline mutations), diagnostic biomarkers (e.g., prostate specific antigen), prognostic biomarkers (e.g., OncotypeDx), biomarkers predictive of treatment benefit (e.g., BRAF V600E mutation), and biomarkers used to monitor response (e.g., circulating tumor DNA). Dr. Hilsenbeck then covered the different ways biomarkers can be incorporated into clinical trials. Ancillary or exploratory biomarkers can be measured without pre-specified hypotheses, in an effort to gather preliminary data to better understand biology. Integrated biomarkers are used to test hypotheses with pre-specified statistical plans, but they are not used to inform treatment decisions. Integral biomarkers are used to inform trial eligibility, treatment assignment, or stratification. Unlike biomarkers in the other categories, integral biomarkers must be performed in a CLIA-certified lab. As the impact of a biomarker on the efficacy of a given treatment is not always apparent prior to the time a study is conducted, Dr. Hilsenbeck introduced the idea of an adaptive trial design as a potential way to address this. In an adaptive trial design, both marker-positive and marker-negative populations are randomized to a given intervention. Before accruing either cohort fully, both sub-populations can have a separate early stopping rule for futility. Lastly, Dr. Hilsenbeck cautioned that testing multiple biomarkers in one trial increases the risk of type I error, and that correction for multiple comparisons is necessary in these settings. All post-hoc analyses should be interpreted with caution, and the most weight should be placed on conclusions drawn from pre-specified analysis.