Evaluation of measurement precision and bias are two key components of establishing analytical performance characteristics of diagnostic tests. Recent updates of the relevant CLSI guidelines (EP05-A3 and EP09-A3) have increased scientific rigor but also complexity of statistical methodology. At time of guideline finalization publicly available statistical software was missing key analysis features. In this presentation we will give an overview how we have implemented the statistical methods recommended by CLSI for precision and bias evaluation in two open source R packages (“VCA” and “mcr”).

The majority of package functionality is implemented in R using the S4 object model, some demanding numerical functions are implemented in C for computational efficiency. Packages are validated by unit and system test cases, including comparison to external reference results. For statistical expert users package functionality and documentation is provided via the standard R interface. In order to directly support users in a laboratory setting a generic wrapper to the .Net framework has been developed and the R packages are fully integrated into our in house laboratory software solutions.

We routinely use the developed R packages to establish and validate the performance characteristics of our diagnostic tests, including the support of regulatory submissions. Sharing our statistical software as open source with the R community has provided many benefits, including transparency to customers and regulators, as well as improved usability and software quality due to testing and code review by the user community.