Building Bridges for Companion Diagnostics
*James Roger Ranger-Moore, Roche Tissue Diagnostics  Crystal Schemp, Roche Tissue Diagnostics 

Keywords: Bridging Studies, Companion Diagnostics

In 2012, Ventana Medical Systems faced the challenge of bridging one of its immunohistochemistry (IHC) assays to an already FDA-approved fluorescent in-situ hybridization (FISH) assay for selecting assay-positive patients for treatment with a targeted therapy in lung cancer.

This bridging strategy was motivated by IHC’s lower cost, ability to use bright-field microscopy (thereby preserving morphologic information), and wider availability. There was an ongoing phase III selection trial that provided the opportunity for comparing the IHC assay to the FISH assay. An IHC scoring algorithm was developed to maximize concordance with FISH while maintaining potential for high reader precision through algorithm simplicity. The algorithm training occurred in independent samples prior to its use in the phase III trial.

The primary endpoints for the bridging study were positive percent agreement (PPA) and negative percent agreement (NPA) of the IHC assay with the FISH assay. The secondary endpoint was progression-free survival (PFS), with the analysis focused on the hazard ratio (HR) between two treatment arms. Statistical analysis had to consider a number of limitations, including that (1) some cases were unavailable for IHC analysis, (2) results for IHC+ cases that were FISH- (or unevaluable) were not available (and also, being screen fails, had limited amounts of other clinical information available), (3) there were multiple plausible ways to define best and worst case scenarios where imputation or simulation occurred, and (4) the trial’s inclusion/exclusion criteria were designed with FISH, not IHC, in mind.

The complex nature of analysis in this setting highlighted many of the challenges of performing a successful bridging study for a companion diagnostic device; the lessons learned during this undertaking are also presented.