CREATING CONFIDENCE

RESEARCH

Statistical evaluation of test results on run controls

To enable statistically valid analysis of response values on our run controls the first requisite is to ensure a linear dose response relation on our standard dilutions. This requirement is fulfilled for Ct values in quantitative NAT assays but not for viral load expressed in IU or copies/mL or S/CO values reported by viral serology or qualitative NAT methods. Often there is a log linear relationship between concentration (or dilution factor) and assay response value. Hence linearization is achieved when the log concentration is plotted against the log response value (e.g. log IU/mL or log S/CO). If the requirement of linearization of the dose response relation is fulfilled by the log transformation there is a normal distribution of test values on the run control and then one can calculate the 95% or 99% confidence intervals and apply the Westgard rules in a statistically valid manner. The method for calculation of the 95% confidence limits are given in each package insert of the assay specific run controls. If one uses a run control with a too high concentration all S/CO ratios in the Ultrio versions reach values in the saturated response range of the assay. In that case one can assume a normal distribution but the statistical analysis becomes meaningless (because if the analytical sensitivity of the assay would significantly decrease the S/CO values would not change). The positioning of the ViraQ Check and Trend Controls is close enough to the 95% LOD of the assay that a significant shift to lower analytical sensitivity would immediately become apparent. In that case there will be a higher proportion of S/CO values in the dynamic range of the assay with a higher probability of nonreactive results. The S/CO values of the Ultrio versions on the ViraQ Check and Trend Controls can be best described by a Gumbel (extreme value) distribution. This statistical approach allows trend analysis and comparison of the analytical performance of Ultrio (Plus and Elite) reagent batches and Tigris or Panther instruments.

Read more on the statistical methods for evaluation of the NAT test results on our run controls (PDF). Run control data sheets for the statistical evaluation of assay response values are in development.