In medical research, biomarkers can be used as indicators of many different disease states. Treatment-selection markers are biomarkers that indicate how well a given disease treatment is working; for some diseases, treatments are not equally effective in all individuals. Understanding variability in treatment effect is essential for clinicians to effectively select individuals who will benefit from a given treatment and avoid unnecessary or harmful procedures on others. To date, treatment-selection markers have been chosen solely by their interaction with treatment, an important criterion, but not directly related to marker performance.
VIDD assistant member Dr. Ying Huang, associate member Dr. Holly Janes and member Dr. Peter Gilbert developed new statistical measures to compare and contrast treatment-selection biomarkers based on a potential outcomes framework, a statistical model for determining cause and effect, and applied their technique to the example of the Step trial, an HIV vaccine trial that was halted in 2007. In the Step trial, vaccine recipients who had previously been infected with the virus Ad5 were more likely to be infected with HIV than placebo recipients infected with Ad5. The statisticians evaluated whether baseline measures of Ad5 were predictive of vaccine-induced risk in HIV infection, and found that baseline Ad5 may have a weak capacity to discriminate those with increased risk of HIV infection from the trial vaccine. – RT
Huang Y, Gilbert PB, Janes H. Assessing Treatment-Selection Markers using a Potential Outcomes Framework. Biometrics. 2012 Feb 2.