Estimation of Biomarker Responses in Presence of Missing Biomarker Status

• Activity Number: 550
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Shengyan Hong*
• Keywords: missing data ; biomarker response ; estimation ; EM algorithm

Abstract

One of the objectives in early phase clinical trials studying a targeted therapy is to accurately estimate treatment response by the status of the biomarker of interest. This is critical to inform further development strategy, such as whether the therapy should be developed in an enriched population. However, often the collection of baseline biomarker tissues for determining biomarker status is not mandatory in early phase clinical trials and/or some collected baseline biomarker tissue is not useable to determine biomarker status due to operational handling issues or assay sensitivity. A naïve method in presense of missing biomarker status is to estimate the response rate based only on subjects whose biomarker statuses are known, which is a consistent estimate only if the missing biomarker status is missing completely at random (MCAR). In this paper, we improve this approach by utilizing observed response data for subjects with unknown biomarker status by applying the EM algorithm under the assumption of missing at random (MAR). Simulations are conducted to compare these two methods and a case study using the methodology is provided.