Schreiber, William E.

Abstract Context. —As part of a College of American Pathologists (CAP) proficiency testing survey, a comparison is made between human chorionic gonadotropin (hCG) results from an actual patient pool and a similarly targeted artificial sample. The goal is to gain insight into the possible source(s) of bias attributable to the proficiency testing material (PTM) with a view toward creating more appropriate survey materials. Objective. —To compare hCG and related variants in a pool of fresh frozen sera (FFS) with that found in PTM. Design. —The 2003 CAP K/KN-A Survey included a FFS specimen along with admixtures of PTM. The FFS (K-02) and 1 PTM admixture (K-01) had similar mean hCG values. Five hCG-related analytes were measured on these 2 samples by a reference laboratory. Participants. —Approximately 1800 clinical laboratories and diagnostic test kit manufacturers participated in the K/ KN-A Survey. Main Outcome Measures. —Method imprecision (coefficient of variation) and method bias (relative difference between peer group mean and all-method median) were computed for the 2 samples. Differences were evaluated with respect to hCG-related analytes levels. Results. —All-method hCG results were 12.9 mIU/mL (12.9 IU/L) for the PTM material and 21.6 mIU/mL (21.6 IU/L) for the FFS material. Method biases for 14 manufacturers were greater for PTM than for FFS (−40% to +35% and −16% to +23%). Twelve of 14 methods had higher coefficients of variation on PTM. Total hCG and free β hCG measurements by the reference laboratory were 14.1 mIU/ mL (14.1 IU/L) for the PTM material and 18.5 mIU/mL (18.5 U/L) for the FFS material (FFS), and 2.4 (PTM) and 0.7 (FFS) mIU/mL (2.4 and 0.7 IU/L), respectively. On a molar basis, free β represented 17% and 4% of the total hCG, respectively. Levels of hyperglycosylated hCG, nicked hCG, and β core fragment were not measurable in either sample. Conclusions. —It is unlikely that the hCG added to the PTM is the source of the increased bias and variability. The main difference is a 3-fold increase in free β found in the PTM, but methods previously found to strongly react with free β were not systematically elevated. The biases between manufacturers found for the FFS specimen are likely attributable to calibration differences.

Abstract Context. —There are 2 general types of assays measuring MUC1 gene–derived glycoprotein: assays for cancer antigen (CA) 15-3, which are sandwich assays, and assays for CA 27.29, which are competitive assays. These 2 types of assays measure slightly different parts of this tandem-repeat molecule. Across-method assay differences hinder the exchange of patient test values among integrated health care networks and among countries. Objective. —This report evaluates the method differences among these assays to determine if the differences between these assays are mainly related to variations in calibration or differences in analyte specificity. Design. —Data from 22 College of American Pathologists survey challenges were analyzed to compare 10 commercial assay methods for these 2 related analytes. In addition, data from 58 patient samples were analyzed to compare 3 of these assays. Results. —The linear correlation coefficients comparing the within-method medians of these proficiency test distributions were very high (>0.99) for all of the methods; however, the regression slopes varied from 0.836 to 1.095. The regression slopes for the patient specimens varied similarly, but the correlation coefficients were lower. Conclusions. —This study indicates that many of the test value differences for these measurements are due to differences in assay calibration rather than differences in the specificity of the assay measurement systems. Survey test data potentially could be used to help harmonize these assay differences.