|
Previous Top Next |
|
|
| 1 | Optimize for precision at the lowest nonzero calibrator and the highest calibrator. If you cannot get acceptable precision over the entire dynamic range of the calibration curve, optimize for precision for calibrators spanning the critical region for clinical decision making. Optimizing for the precision at the lowest calibrator will usually give the same result as optimizing for the Lower Limit of Quantification.
|
|
|
| 2. | Optimizing for the Upper Limit of Quantification can be problematic. The ULOQ will be set to the value of the highest calibrator in a run where the precision there is still better than the LOQ threshold.
|
|
|
| 3. | It is possible to optimize for accuracy, but there is frequently little effect on accuracy due to systematic changes in assay conditions. During optimization, put your primary focus on precision. Otherwise, you may be fitting an optimization to small amounts of random variation around the accuracy.
|
|
|
| 4. | Although some of the example data sets supplied with ProQuant show varying numbers of replicates among the calibrators and controls, the most efficient way to estimate the imprecision is to spread your replicates evenly across all calibrators and controls.
|
|
|
| 5. | Take advantage of the confidence intervals provided by ProQuant. If the confidence intervals for a performance measure (for example, LLOQ) overlap between condition A and condition B, there may not be important differences in performance between the two conditions.
|
|
|
| 6. | Collect as much data as you can afford. Even though reasonable estimates can be made from a calibration curve in duplicate, you may find a robust optimum more quickly with just a little more data. Experiment with removing replicates from some of the example data sets to see the effect on confidence interval width.
|
|
|
| 7. | Understand your DOE tools and software.
|
|
|