Tips on Managing the Quality of Immunoassays
EXAMPLE ANALYTES

R. Neill Carey, Ph.D.
Clinical Chemist
Peninsula Regional Medical Center
Salisbury, MD

• Theophylline
• Cortisol
• Thyroxine
• Folate
• Back to the main text

Example Analytes

The following examples demonstrate how to apply these tools and thought processes. In each case, the assay is stable, and has a very low frequency of large errors (< 2%). These are real assays from the laboratory, and the QC data represent precision over a period of about 4 months. PT specifications were used for allowable error.

Theophylline on Abbott AxSYM, PT Limit is Target Value +/- 25%

Since there is a single CV criterion for all concentrations, look for the control with the highest CV. Our worst CV is 3.34%. Since allowable error (25%) is large relative to the CV (remember the Westgard-Burnett rule?), it should be pretty easy to control this method even with 2 control measurements per run. However other tests on this system are not so easily controlled and require us to analyze 3 different control materials. Because people may get confused if you have different numbers of controls for different tests on the same analyzer, I used the automatic rule selection feature for 3 materials. As shown by the OPSpecs chart, even the 1_3.5s rule has P_ed over 90% for errors which could cause PT failure. I actually chose the 1_3s rule because our laboratory computer software does not allow us to choose the 1_3.5s. P_fr is still very low, only 1% or so.

Cortisol on Abbott TDx, PT Limit is Target Value +/- 25%

Cortisol also has a fixed PT limit of target value +/- 25%. Our worst CV (low control) is 9.97%. The CV is greater than one third of allowable error.

The OPSpecs® chart shows that there is very little ability to detect errors that would cause PT failure. The 13s/(2of3)2s/R4s/31s rule with 6 controls barely achieves Ped = 25%. It is critically important to minimize bias in the low concentration range. [Note that this OPSpecs chart is for 25% AQA(SE).]

Fortunately, the lowest cortisol concentration in the CAP K series in 1996 was 13 mcg/dL. Looking at the situation from the perspective of our average performance makes a brighter picture. The average CV over the range of our three controls is 6.74%, which just over one fourth of allowable error. With 6 controls, the 1_3s/(2of3)_2s/R_4s/3_1s/6_x and 1_3s/(2of3)_2s/R_4s/3_1s rules have P_ed over 90%. Since our laboratory computer does not support the 3_1s or 6_x rule, I chose the 1_3s/(2of3)_2s/R_4s rule with P_ed around 80% for errors which could cause PT failure. Its P_fr is reasonable at 0.02.

For three controls in daily runs, Ped for the 13s/(2 of3)2s/R4s rule is 58%, which is acceptable for a method with a low frequency of large errors.

This is an example of multistage QC. If problems occur, it will usually be when new reagent lots are calibrated or after recalibrations. Maximum sensitivity to errors is important, even if it costs increased false rejections. Six controls are run at calibration. Less sensitivity to errors is acceptable for daily operation because of the low frequency of errors, and QC procedures with 3 controls are satisfactory.

In summary, we are OK for cortisol on the average with 3 controls, but we could have problems with PT samples at low concentrations if bias exceeds 0.86 SD, and we have little ability to tell if bias has become too large.

Thryoxine on Abbott AxSYM, PT Limit is Target Value +/- 20% or 1.0 mcg/dL, Whichever is Greater

The limit for thyroxine is target value +/- 1.0 mcg/dL for concentrations up to 5.0 mcg/dL, and +/- 20% for concentrations above 5.0 mcg/dL. Cumulative statistics from four months’ data are shown below for our current lot of tri-level Immunoassay control:

Consider performance at low concentrations first. I converted the allowable error of 1.0 mcg/dL to a percentage by dividing it by the concentration of the low control (3.96 mcg/dL) to give 25.2% allowable error. Since the CV is about one-fourth of allowable error, we should be OK here with multirule. I use multirule with N=3, with P_ed of 0.70. For an assay with a low frequency of errors, P_ed of 0.70 is satisfactory. Even the 1_3s with P_ed of 0.53 would be acceptable. The fixed limit of 1.0 mcg/dL for thyroxine lessens the odds for trouble at low concentrations.

At higher concentrations, the CV is less than 25% of allowable error, and performance of these QC rules would be even better.

Folate by RIA, PT Limit is Group Mean +/- 3 SD

When the limit is target value +/- 3 SD, the PT limit is calculated from recent PT data for the method. Group the PT results for a recent period by concentration ranges that correspond roughly to the concentrations of the control materials you are using and/or where you see patterns of similar CVs in the PT data. We used PT data from the 1996 College of American Pathologists K series for Bio-Rad Quantaphase II. The CVs were relatively constant, ranging from 9.3% to 11.1%. The average CV was 10.2%. The PT allowable error is calculated by multiplying the average CV by 3 to give 30.6%. Cumulative control data for folate are as follows for our tri-level immunoassay controls:

Our worst CV is only 14% of allowable error. With 3 controls, the 1_3.5s rule has P_ed exceeding 90% for the critical error.

Return to the main text