Multicenter validation of a galactomannan chemiluminescence immunoassay for the diagnosis of pulmonary aspergillosis on serum of patients with hematological disease.

An accurate diagnosis of invasive aspergillosis (IA) in patients with underlying hematological malignancies relies heavily on galactomannan detection. In this study, we compared the VirCLIA chemiluminescence immunoassay (CLIA) with the frequently used Platelia enzyme-linked immunosorbent assay (ELISA) on serum from hematology patients with suspected IA. Patients were categorized according to EORTC/MSGERC 2020 definitions into proven/probable IA and possible/no IA. The first cohort included 161 patients at four centers, and the VirCLIA manufacturer's cutoff of 0.200 was evaluated. Next, the optimal cutoff was determined using the Youden's index. In a second independent cohort of 189 patients from four centers, this optimal cutoff was evaluated again. In the first cohort, sensitivities and specificities for probable/proven IA were 21.1% and 100.0% for ELISA (1.0 cutoff) and 36.6% and 95.6% (0.5 cutoff), compared to 11.3% and 97.8% for CLIA (0.200 cutoff). In the second cohort, the sensitivities of ELISA and CLIA were comparable (ELISA ≥ 1.0: 33.3%, CLIA ≥ 0.200: 38.1%). The area under the ROC curve was lower for CLIA than for ELISA in the first cohort (65.0% vs 78.7%, = 0.005) but comparable in the second cohort (79.5% vs 81.3%, = 0.649). Youden's index identified 0.100 as the optimal CLIA cutoff with sensitivities of 35.2% and 61.9% in cohorts 1 and 2, respectively, at slightly reduced specificities of 85.6% and 90.5%. While the sensitivity of both assays was low to moderate at best, in patients with a high pre-test probability, we suggest 0.100 as the cutoff for the VirCLIA assay.IMPORTANCEThis study demonstrates a comparable performance of the novel chemiluminescence immunoassay (CLIA) and the conventionally used enzyme-linked immunosorbent assay for galactomannan serum testing in hematological patients at high risk for invasive aspergillosis. In patients with a high pre-test probability, a lower CLIA cutoff of 0.100 is preferred.