Clinical Laboratory Standards Institute (CLSI) provides important guidance for interference testing.1 Many laboratorians find the CLSI document difficult to follow, which is why we have developed simple instructions and Excel Interference Worksheets to make the interference testing experiment simpler and easier to interpret.
CLSI recommends a screening experiment (paired difference testing) to identify assay interference, followed by a dose-response experiment to determine the exact concentration of interferent that is clinically significant. Here, we’ll focus on the screening experiment.
The first step in the interference testing process is to define the acceptance criteria; e.g. what difference caused by an interfering substance is considered clinically significant. The maximum acceptable difference (Dmax) is defined as the difference between the mean concentration of the “Test” Pool and the mean concentration of the “Control” Pool that is considered medically significant. Fortunately, there are resources available to help with this.
Total allowable error (TAE) is a term that is familiar to most laboratorians. TAE encompasses all sources of error associated with an assay and represents the total error “budget” for our assay. Setting the Dmax for our interference experiment means that we must choose a fraction of our TAE budget that we can devote to bias from interference. A reasonable choice is one-fourth to one-half of the TAE, but this is a decision for the Laboratory Director. For example, CLIA defines 10% as a TAE for glucose, but we don’t want to apply our entire error budget to potential interference. We need to leave some room in the error budget for random imprecision, calibration bias, etc. A reasonable Dmax for glucose would be 2.5% to 5%.
Now let’s prepare our “Test Pool” and “Control Pool.” The Test Pool contains a high concentration of the interferent in the appropriate sample matrix (serum, plasma, urine, etc.). The Control Pool contains the same sample matrix with little or no interferent. The Control Pool should contain the same volume of solvent as used to add interferent to the Test Pool. The solvent may be normal saline or water if the interferent was solubilized in water or saline, or it may be 0.1M NaOH (used to solubilize unconjugated bilirubin) or ethanol or some other solvent. As an example, if the Test Pool was created with 0.95 mL of a serum pool plus 0.05 mL unconjugated bilirubin (in 0.1M NaOH), the Control Pool should be 0.95 mL of the same serum pool plus 0.05 mL 0.1M NaOH.
The next obvious question is how much interferent should be added to create the Test Pool. For most potential interferents CLSI provides recommended interferent concentrations in document EP-37.2 For interferents not included in this document a general recommendation is to test 3-times the highest concentration ever observed or documented, although many of us think this is excessive. The Laboratory Director should make this call. As a result, the Test and Control Pools will have the same matrix and the same concentration of measurand (analyte of interest). Any difference between Test and Control Pools is therefore due to the presence of the interfering substance.
We have considered the concentration of the interferent, but the concentration of the measurand is equally important, as both contribute to the magnitude of the interference. Again, CLSI EP7 provides suggestions in Appendix A for appropriate measurand concentrations. It is a good idea to test for interference using two different measurand concentrations.
Lastly, how many replicates of the Test and Control Pools should I measure? CLSI EP7 provides sample size equations and Table 2 to help answer this question. You need to define (1) the probability of a type 1 error (the chance of falsely rejecting the true (null) hypothesis), (2) the probability of a type II error (the chance of accepting a false null hypothesis), (3) ratio of Dmax/repeatability (within-run standard deviation), both expressed in concentration terms or as a percent. Generally, it is advisable to use 5% and 10% for type 1 and type 2 error probabilities, respectively, also known as 95% confidence and 90% power. The other component of the sample size calculation to determine the number of replicates is to decide on a one-tailed or a two-tailed test. The null hypothesis for a one-tailed test is that the Test Pool mean is higher than the Control Pool or that the Test Pool mean is lower than the Control Pool. For a two-tailed test, we hypothesize that the Test Pool is different from the Control Pool, but we cannot predict the direction of this difference. A two-tailed test is usually appropriate. The Sun Diagnostics Interference Worksheets helps with this.
After the analysis, we can interpret the interference testing results. CLSI keeps this simple: we do not need statistical analysis of significance. We are only going to interpret the difference with respect to medical significance—if the difference between the Test Pool mean and the Control Pool mean is greater than Dmax, then the difference is medically significant. The screening experiment does not tell us what level of interference is acceptable, only that there is or is not medically important interference at the concentration tested. If there is medically important interference it is necessary to conduct a dose-response experiment to characterize the degree of interference associated with different concentrations of interferent. More on dose-response experiments in a later blog.
Sun Diagnostics has created an Interference Worksheet to help with all aspects of interference testing from creating Test Pools, calculating the required number of replicates and differences in Test and Control Pools, to interpreting the results. We’ll introduce our Interference Worksheets and instruction in a follow-up blog.
John H. Contois, PhD, DABCC, FAACC
1 CLSI. Interference Testing in Clinical Chemistry, 3rd ed. CLSI guideline EP07, Wayne, PA: Clinical and Laboratory Standards Institute; 2018.
2 CLSI. Supplemental Tables for Interference Testing in Clinical Chemistry, 1st ed. CLSI guideline EP37, Wayne, PA: Clinical and Laboratory Standards Institute; 2018.