Interlaboratory Comparison of Antibody-Free LC-MS/MS Measurements of C-peptide and Insulin

Abstract

Background: The enhanced precision and selectivity of liquid chromatography-tandem mass spectrometry (LC-MS/MS) makes it an attractive alternative to certain clinical immunoassays. Easily transferrable work flows could help facilitate harmonization and ensure high-quality patient care. We aimed to evaluate the interlaboratory comparability of antibody-free multiplexed insulin and C-peptide LC-MS/MS measurements.

Methods: The laboratories that comprise the Targeted Mass Spectrometry Assays for Diabetes and Obesity Research (TaMADOR) consortium verified the performance of a validated peptide-based assay (reproducibility, linearity, and lower limit of the measuring interval [LLMI]). An interlaboratory comparison study was then performed using shared calibrators, de-identified leftover laboratory samples, and reference materials.

Results: During verification, the measurements were precise (2.7% to 3.7%CV), linear (4 to 15 ng/mL for C-peptide and 2 to 14 ng/mL for insulin), and sensitive (LLMI of 0.04 to 0.10 ng/mL for C-peptide and 0.03 ng/mL for insulin). Median imprecision across the 3 laboratories was 13.4% (inter-quartile range [IQR] 11.6%) for C-peptide and 22.2% (IQR 20.9%) for insulin using individual measurements, and 10.8% (IQR 8.7%) and 15.3% (IQR 14.9%) for C-peptide and insulin, respectively, when replicate measurements were averaged. Method comparison with the University of Missouri reference method for C-peptide demonstrated a robust linear correlation with a slope of 1.044 and r2 = 0.99.

Conclusions: Our results suggest that combined LC-MS/MS measurements of C-peptide and insulin are robust and adaptable and that standardization with a reference measurement procedure could allow accurate and precise measurements across sites, which could be important to diabetes research and help patient care in the future.

Figure 1.. Linearity verification study.

The observed peak area ratio is plotted vs. the expected concentration for two dilutional series of samples analyzed for C-peptide at either CSMC (A) or PNNL (B). Similar data for insulin are also shown for CSMC (C) and PNNL (D). Samples were analyzed in triplicate. Error bars are SD.

Figure 2.. Lower limit of the measuring interval verification study.

Imprecision is plotted vs. the observed concentration of C-peptide (A, B) and insulin (C, D) for a diultional series of five samples run at CSMC (A, C) and a series of seven samples run at PNNL (B, D). A power function was used in the regression of imprecision (%CV) vs. observed concentration and solved for the estimated LLMI (i.e., at %CV = 20%).

Figure 3.. Between-lab comparison of the measurement of C-peptide and insulin calibrated with SPC or in-house calibrators.

The concentration observed for each individual sample in each individual laboratory is plotted vs. the mean concentration observed across all three labs. Results for C-peptide (A, B) and insulin (E, F) are illustrated. Results were calibrated with in-house calibrators in horse serum (A, E) or a single-point calibrator (B, F). Bland-Altman plots of the same analyses are included, as well (C, D, G, H) The individual labs are CSMC = Black, UW = Open circles, and PNNL = Orange circles. One extreme outlier sample was removed from the insulin comparison.

Figure 4.. Comparison with reference method.

(A) Mean C-peptide concentrations for single-donor reference materials measured across laboratories (TaMADOR) were plotted vs. the measurements performed by University of Missouri Standardization Program. The linear regression is plotted (blue dashed line) as is the line of identity (black dotted line). (B) The percent bias was also calculated for each sample and the mean bias is plotted as a dashed line.

Figure 5.. Comparison of insulin and C-peptide concentrations in different sample types.

The mean concentration of insulin observed in the TaMADOR laboratories is plotted vs. the mean observed C-peptide concentration.