Impact of Biomarker-based Design Strategies on the Risk of False-Positive Findings in Targeted Therapy Evaluation
- 13 December 2018
- journal article
- research article
- Published by American Association for Cancer Research (AACR) in Clinical Cancer Research
- Vol. 24 (24) , 6257-6264
Abstract
Purpose: When there is more than one potentially predictive biomarker for a new drug, the drug is often evaluated in different subpopulations defined by different biomarkers. We aim to (i) estimate the risk of false-positive findings with this approach and (ii) evaluate the cross-validated adaptive signature design (CVASD) as a potential alternative. Experimental Design: By using numerically simulated data, we compare the current approach and the CVASD across different settings and scenarios. We consider three strategies for CVASD. The first two CVASD strategies are different in terms of the partitioning of the overall significance level (between the population test and the subgroup test). In the third CVASD strategy, the order of the two tests is reversed, that is, the population test is realized when the prioritized subgroup test is not statistically significant. Results: The current approach results in a high risk of false-positive findings, whereas this risk is close to the nominal level of 5% once applying the CVASD, regardless of the strategy. When the treatment is equally effective to all patients, only the CVASD strategies could specify correctly the absence of a sensitive subgroup. When the treatment is only effective for some sensitive responders, the third CVASD strategy stands out by its ability to correctly identify the predictive biomarker(s). Conclusions: The drug–biomarker coevaluation based on a series of independent enrichment trials can result in a high risk of false-positive findings. CVASD with some appropriate adjustments can be a good alternative to overcome this multiplicity issue.Keywords
Funding Information
- Conseil Régional, Île-de-France (Île-de-France Regional Council) (Bourse Master « Île-de-France »)
This publication has 30 references indexed in Scilit:
- Trial Designs for Personalizing Cancer Care: A Systematic Review and ClassificationClinical Cancer Research, 2013
- A Robust Method for Estimating Optimal Treatment RegimesBiometrics, 2012
- Subgroup identification from randomized clinical trial dataStatistics in Medicine, 2011
- Subgroup identification based on differential effect search—A recursive partitioning method for establishing response to treatment in patient subpopulationsStatistics in Medicine, 2011
- An adaptive seamless phase II/III design for oncology trials with subpopulation selection using correlated survival endpoints†Pharmaceutical Statistics, 2010
- The Cross-Validated Adaptive Signature DesignClinical Cancer Research, 2010
- Randomized Clinical Trials With Biomarkers: Design IssuesJNCI: Journal of the National Cancer Institute, 2010
- Clinical Trial Designs for Evaluating the Medical Utility of Prognostic and Predictive Biomarkers in OncologyPersonalized Medicine, 2009
- The Use of Genomics in Clinical Trial DesignClinical Cancer Research, 2008
- Adaptive Signature Design: An Adaptive Clinical Trial Design for Generating and Prospectively Testing A Gene Expression Signature for Sensitive PatientsClinical Cancer Research, 2005