Müller, Anne; Scholz, Markus; Blankenstein, Oliver; Binder, Gerhard; Pfäffle, Roland; Körner, Antje; Kiess, Wieland; Heider, Annegret; Bidlingmaier, Martin; Thiery, Joachim; Kratzsch, Jürgen
Harmonization of growth hormone measurements with different immunoassays by data adjustment.
In: Clinical Chemistry and Laboratory Medicine, Vol. 49, No. 7: pp. 1135-1142
Background: The aim of our study was to evaluate the between-assay variability of commercially available immunoassays for the measurement of human growth hormone (hGH). In addition, we asked whether the comparability of the diagnosis of childhood onset growth hormone deficiency could be improved by adjusting hGH results by statistical methods, such as linear regression, conversion factors, and quantile transformation. Methods: In archived sera from 312 children and adolescents (age: 17 days-17 years) hGH values between 0.01 and 16.5 ng/mL were determined by using the following immunoassays: AutoDELFIA (PerkinElmer), BC-IRMA (Beckman-Coulter), ELISA (Mediagnost), IMMULITE 2000 (Siemens), iSYS (IDS), Liaison (DiaSorin), UniCel DxI 800 Access (BeckmanCoulter) and "In house"-RIA (Tubingen). Results: The assays differed in median hGH concentrations by as much as 5.44 ng/mL (Immulite), and as little as 2.67 ng/mL (BC-IRMA). The mean difference between assays ranged from 0.35 to 2.71 ng/mL, whereas several samples displayed differences up to 11.4 ng/mL. The best correlation (r=0.992) was found between AutoDELFIA and Liasion, the lowest (r=0.864) was between an in-house RIA and iSYS. The between-assay CV (mean +/- SD) of values within the cut-off range was 24.3%+/- 7.4%, resulting in an assay-dependent diagnosis of growth hormone deficiency (GHD) in more than 27% of patients. Yet, adjustment of this data by linear regression or a conversion factor reduced the CV below 14%, and the ratio of assay-dependent diagnoses below 8%. Using quantile transformation, the CV and ratio were reduced to 11.4% and < 1%, respectively. Conclusions: hGH measurements using different assays vary significantly. Linear regression, conversion factors, or particularly quantile transformation are useful tools to improve comparability in the diagnostic procedure for the confirmation of GHD in childhood and adolescence.