|
| |
Case Study: Statistical Assay Analysis
StatLIA's statistical analysis and the Shewart 22S control rule were used to analyze the results of more than 1600 assays performed in a high-volume reference laboratory in the Midwest. StatLIA's statistical analysis flagged 43 assays, compared to 15 assays flagged by the multi-rule method. In 17 of the 43 StatLIA-flagged assays, the amount of error shifted the results of the unknowns enough to change the diagnostic interpretation. Only 4 of these failed assays were detected by the multi-rule method. The other 11 assays of the total 15 flagged by the multi-rule method were due to bad control specimens, not bad assays, a distinction easily determined using StatLIA’s statistical analysis.
Applications
|
|
Characterize performance of standard curve by monitoring more than 50 standard curve parameters, including standard responses, min/max detectable concentrations, curve fit statistics, control responses and concentrations |
|
Troubleshoot assays immediately and document the failed component (antibody, tracer, etc.) |
|
Develop, characterize, and validate new tests |
|
Establish test verification standards in StatLIA, and then monitor them automatically |
|
Distinguish between a bad assay or just a bad control |
|
Document and justify why a standard curve was edited |
|
Generate Quality Assurance reports to verify the performance consistency of a test over time |
|
Monitor Levey-Jennings Charts that plot response, normalized response, and concentration for each control |
|
Analyze assay performance by verifying accuracy, precision, reportable range, specificity, reliability |
|
Optimize assay performance with a complete statistical comparison of one or more assays to a stable pool of historical reference assays |
|
Compare each sample’s observed variance to its predicted variance based on the variances from the test’s historical reference assays, and compute a Precision Probability based on that comparison |
|
Calculate matrix effects by computing the analysis of variance for both the standards and the unknowns |
