Endocrine Abstracts

Introduction: The determination of basal and stimulated calcitonin (CT) is very important for the precocious diagnosis of medullary thyroid carcinoma (MTC) and for its dynamic follow-up. Unfortunately, monitoring is difficult when using assays with different detection systems. Our objective was to identify the correction factor between the reagents for CT measurement on the LiaisonXL and Cobase 601 assays (immunochemiluminescence, respectively electro-immunochemiluminescence).

Materials and methods: The measurement rangesare 1–2000 pg/ml for LiaisonXL and 0.5–2000 pg/ml for Cobase 601. All samples above the measuring range were diluted. We selected a group of 89 patients, 28 men and 61 women, with baseline CT or calcium-stimulated CT samples dosed on Cobas, with values between 0.5–2812 pg/ml. The samples were aliquoted and stored at –20ºC, and subsequently analysed on Liaison. The agreement was obtained according to ethical norms through informed consent.

Results: After applying various statistical tests, we found a strong Spearman correlation coefficient between the two assays, of 0.992. There is a segmented linear correlation between the two assays. The general linear regression equation is: 1.108 × + 19.337 (P < 0.05), but for higher accuracy we identified the segmented regression equations on two disjoint intervals. The increase of the bias was observed at high CT values: for values between 200 400 pg/ml on Cobas, the linear regression equation becomes 0.807 × + 87.554 (P < 0.05), and for values between 400–2812 pg/ml on Cobas, the linear regression equation becomes 0.997 × + 191.094 (P < 0.05). We also went on and explored an alternative approach for obtaining a higher accuracy prediction model. Proceeding from the obtained data we developed a software prediction model based on Machine Learning (ML) algorithms, which by design is a prediction model that improves continuously in accuracy of the estimates as more and more data becomes available. In addition, empirical tests have shown that the ML prediction model is superior to the statistical method.

Conclusions: For the first time in literature, we identified the correction factors between Liaison and Cobas assays for CT determination. For the dynamic monitoring of a patient with MTC it is very important to quantify CT with the same method of analysis, and to apply correction factors when using different assays. The results may be influenced by the the fact that the analysys was made on stored samples. Continuing the study on a larger number of fresh samples will ensure increasing accuracy of the correction factors.

Keywords: calcitonin, liaison, cobas, correction factor.