A-049 Inter-assay Variation of Insulin-like Growth Factor-1 Remains an Issue

Abstract

Abstract Background Human insulin-like growth factor-1 (IGF-1) is the main mediator of the somatotropic effects of growth hormone (GH). Accurate measurement of IGF-1 is required for the diagnosis and management of GH secretion disorders. Several commercial vendors and laboratories have standardized their IGF-1 assays to the WHO International Standard 02/254. This standardization does not prohibit the existence of significant inter-assay variability in IGF-1 measurements. Such concentration variance can lead to inaccurate patient case decision making. The IGF-1 testing performed in our laboratory was recently migrated from an Immulite 2000 platform to a CobasPro system. This shift in vendor necessitated an investigation of the inter-assay variability between these standardized tests. This relative variance was established through a validation of the analytical performance of the CobasPro IGF-1 assay. Methods Reportable range, intra-day precision and accuracy of IGF-1 measurements using the CobasPro (Roche Diagnostics) system were determined. Patient correlation studies between the CobasPro and an Immulite 2000 (Siemens Healthcare Diagnostics) instrument was also performed. The concentration of IGF-1 in a subset of these correlation study specimens was also quantified by liquid chromatography mass spectrometry (LC-MS) assays from two different reference laboratories. Results The CobasPro assay had a verified linear range of 7–1443 ng/mL (R2 = 0.9998, slope = 0.9915). At IGF-1 concentrations of 54 and 345 ng/mL the intra-day precision (%CV, N = 10) of the CobasPro assay did not exceed 0.9%. Patient specimen correlations between the CobasPro and Immulite 2000 assays showed significant bias (Deming regression: y = 1.543x − 48.61, N = 128, R = 0.9641). This method bias was exacerbated at Immulite 2000 derived IGF-1 concentrations >100 ng/mL (Deming regression: y = 1.662x − 84.29, N = 93, R = 0.9492, CobasPro range = 77–910 ng/mL, Immulite 2000 range = 103–539 ng/mL). Patient specimen correlations between the CobasPro and LC-MS assay from reference laboratory A showed a positive bias (Deming regression: y = 1.375x − 64.5, N = 19, R = 0.9210). An analogous correlation with reference laboratory B demonstrated a higher level of concordance (Deming regression: y = 1.005x + 10.2, N = 16, R = 0.9499). The CobasPro and Immulite 2000 assays both produced acceptable result when challenged by multiple external quality assurance (EQA) specimens provided by the College of American Pathologists (CAP). However, the observed relative trends in IGF-1 concentrations from these challenges were opposite to our observations with authentic patient specimens. The CobasPro assay consistently produced lower IGF-1 concentrations relative to the Immulite 2000 assay for these EQA specimens. Conclusion The CobasPro IGF-1 assay offered acceptable analytical performance. Although both assays were traceable to a reference standard, a significant positive concentration bias was observed between the CobasPro and Immulite 2000 assays. Standardization efforts have not eliminated inter-assay IGF-1 variability. A careful evaluation of the impact of this variance on the diagnosis and management of GH deficiency and acromegaly must be considered when IGF-1 testing platforms are changed.