28 – Analysis of Count Data
Poisson Modeling for LOD Assessment for BCRABL Testing
Zhonggai Li
Novartis
Tianlei Chen
Novartis
Brian Mullaney
Novartis
The Gleevec and Tasigna for Philadelphia chromosome-positive chronic myeloid leukemia (Ph+ CML) patients have been clinically proved to be highly effective. After certain months of treatment, the BCRABL/ABL ratio (IS) typically reach to a very lower level, such that it requires a highly sensitive BCRABL assay that can detect the BCRABL/ABL ratio (IS) to 4.5log10 below the IS baseline (MR4.5). Most of BCRABL assays use qPCR technology, and utilize ABL gene as control for relative quantification. In the CML BCRABL testing community, when the BCRABL is not detected, the assay algorithm will substitute it with 1 copy and use (1/ABL copy)*IS Conversion Factor to report the BCRABL/ABL ratio (IS). However, in IVD practice, it is usually substituted with an assay specific LOD, which is higher than 1 copy due to sampling error in these very low copy number samples. This paper will use poisson modeling to construct the connection and separation between the qPCR assay and sampling error, which will offer quantifiable tool to evaluate the difference and appropriateness of substitute 1 BCRABL copy vs LOD of BCRABL copy in the clinical diagnostics.