Of the 75 samples analyzed using the BSIN method, 2 had IgG4 measurements greater than the analytic measuring range of the LC-MS/MS method (20 g/L) and, therefore, were not included in subsequent statistical regression analyses.
After review of the low coefficient of determination of the BSIN vs LC-MS/MS intermethod comparison for the IgG2 subclass, additional samples were sent for analysis by the other major vendor of IgG subclass IN, SIN, to determine whether the apparent IN errors were vendorspecific.
This finding was similar to that observed for BSIN IgG2 vs LC-MS/MS IgG2 (discussed below).
Because we had previously noted an IgG4-dependent discrepancy between SumIgG vs IgG total IN measurements (6), we reviewed the effect of IgG4 concentration on the linear regression analyses for both the BSIN vs LC-MS/MS and the SIN vs LC-MS/MS comparisons.
IgG2: effect ofIgG4concentration on BSIN vs LC-MS/MS.
This hypothesis of BSIN IgG2 reagent crossreactivity with IgG4 was further borne out by the observation that the SumIgG, when measured by BSIN, was frequently (13 of 75) >20% higher than the SIN total IgG.
The BSIN IgG1 multivariate linear regression on the LC-MS/MS IgG subclasses also showed a significant contribution from LC-MS/MS IgG4 with a coefficient (0.40) that was slightly <50% of the corresponding LCMS/MS IgG1 coefficient (0.89).
Statistically significant associations between LC-MS/MS IgG1 measurements and IN IgG4 measurements (both SIN and BSIN) were observed (Table 2).
The discordance was most evident for the IgG2 subclass with the magnitude of errors being substantially larger using BSIN reagents, but also present when using SIN reagents.
In the case of BSIN vs LC-MS/MS, the BSIN IgG2 measurements had a substantial positive bias that, on average, was of the same magnitude of the IgG4 concentration, suggesting 100% cross-reactivity of the reagent antisera.
In the case of both the SIN IgG2 reagents and the BSIN IgG1 reagents, the respective positive interferences, while also proportional to IgG4 concentration, were of smaller magnitude and less predictable than the IgG4-mediated interference on the BSIN IgG2 measurement.
In summary, we observed biases of the IN IgG subclass measurements compared with the corresponding LC-MS/MS measurements, which potentially reflect 2 analytical phenomena: (a) cross-reactivity of sample IgG4 with the BSIN IgG2 reagents and (b) IN measurement of IgG1 and IgG2 immunoglobulins that represent an aggregate of the target immunoglobulin and nonspe cifically bound IgG4 (IgG4 bound to either the target immunoglobulin or the reagent immunoglobulin).