Objectives: To determine the ferritin inter-assay differences between three "Conformité Européenne" (CE) marked tests, the impact on reference intervals (RI), and the proportion of individuals with iron deficiency (ID), we used plasma and serum from healthy blood donors (HBD) recruited in three different Switzerland regions.

Design and methods: Heparinized plasma and serum from HBD were obtained from three different transfusion centers in Switzerland (Fribourg, Geneva, and Neuchatel). One hundred forty samples were recruited per center and per matrix, with a gender ratio of 50%, for a total of 420 HBD samples available per matrix. On both matrices, ferritin concentrations were quantified by three different laboratories using electrochemiluminescence (ECL), latex immunoturbidimetric assay (LIA), and luminescent oxygen channeling immunoassay (LOCI) assays, respectively. The degree of agreement between matrices and between the three sites/methods was assessed by Passing-Bablok and we evaluated the proportion of individuals deemed to have ID per method.

Results: Overall, no difference between serum and heparinized plasma ferritin values was observed according to Passing-Bablok analyses (proportional bias range: 1.0-3.0%; maximum constant bias: 1.84 µg/L). Significant median ferritin differences (p< 0.001 according to Kruskal-Wallis test) were observed between the three methods (i.e., 83.6 µg/L, 103.5 µg/L, and 62.1 µg/L for ECL, LIA, and LOCI in heparinized plasma, respectively), with proportional bias varying significantly between ±16% and ±32% on serum and from ±14% to ±35% on plasma with no sign of gender-related differences. Affecting the lower end of RI, the proportion of ID per method substantially varied between 4.76% (20/420) for ECL, 2.86% (12/420) for LIA, and 9.05% (38/420) for LOCI.

Conclusions: Serum and heparinized plasma are exchangeable for ferritin assessment. However, the order of magnitude of ferritin differences across methods and HBD recruitment sites could lead to diagnostic errors if uniform RI were considered. Challenging the recently proposed use of uniform ferritin thresholds, our results highlight the importance of method- and region-specific RI for ferritin due to insufficient inter-assay harmonization. Failing to do so significantly impacts ID diagnosis.