Evaluation of commercial ELISA and RIA for measuring porcine C-peptide: implications for research

Pre-clinical demonstration of porcine islet graft function is necessary to support the clinical transplantation of pig islets. C-peptide concentration is an especially useful marker of insulin secretion, because its measurement is not confounded by the presence of exogenous insulin. To measure porcine C-peptide (PCP), researchers in the field exclusively used the Millipore (previously Linco Research) radioimmunoassay (RIA) until 2011, when Mercodia released an alternative enzyme-linked immunosorbent assay (ELISA). (At the end of 2013, the Millipore RIA was withdrawn from the market for commercial reasons.) In our current study, to directly compare these two assays, we performed validation studies on each. We also performed interlaboratory comparisons. Then, to determine the level of agreement between the assays, we analyzed the porcine serum C-peptide concentration measurement results obtained from each assay.

Using pre-established method validation acceptance criteria, we determined and evaluated the detection limit, sensitivity, precision, linearity, and recovery of the two commercially available PCP assays described above (ELISA and RIA). After validation requirements were met, we performed a method comparison by determining C-peptide concentration in 60 serum samples collected from 31 normal, healthy adult Landrace pigs in the fasting state; a subset underwent an intravenous glucose challenge test, to stimulate the typical physiologic range of C-peptide. All analyses were performed according to manufacturer instructions. To compare the assays, we used Deming regression analysis.

Both assays met acceptance criteria. The RIA had a sensitivity of 0.1 ng/ml; it was linear to 2.9 ng/ml. The ELISA had a detection limit of 0.03 ng/ml; it was linear to 1.2 ng/ml. Recovery ranged from 89 to 113% with both assays. The coefficient of variability was 8% in interlaboratory comparisons. Deming regression analysis directly comparing both assays revealed significant correlation between them (before log-transformation: R2=0.9803, P<0.0001; after log-transformation: R2=0.9727, P<0.0001). Measured C-peptide concentration was lower with the ELISA than with the RIA; individual measurements plotted against the averages of the pair demonstrated that the variability from the mean strongly depended on increasing concentration. To transform ELISA data, we used the standard regression equation y=2.191x+0.1119 and the log-transformed regression equation y=0.8101x+0.7502. Both the transformed and the log-transformed (exponential) values approximated the measured RIA levels with a high degree of accuracy in the concentration range of 0 to 1.0 ng/ml.

Porcine C-peptide concentration can be reliably measured in porcine serum samples with either assay (ELISA or RIA). However, the C-peptide results generated by these two assays are not equivalent. Therefore, assay bias must be considered before directly comparing pre-clinical studies that used either of these assays. We determined that harmonization between the assays is appropriate in a specific concentration range. Outside of that range, we do not know whether a linear correction function can be more broadly applied. The variation between the two assays may be related to calibration or reagent factors. To determine which assay is truly more accurate and to effectively compare interlaboratory results, we will need a traceable reference standard.