Comparative performance of the Platelia Aspergillus Antigen and Aspergillus Galactomannan antigen Virclia Monotest immunoassays in serum and lower respiratory tract specimens: a “real-life” experience

Diagnosis of proven invasive pulmonary aspergillosis (IPA) requires histological or cytopathological evidence of fungal tissue invasion or recovery of Aspergillus spp. by culture or detection by polymerase chain reaction (PCR) from a normally sterile site by biopsy or needle aspiration from a lesion consistent with an infectious process (1, 2). A wide variety of non-culture-based methods including detection of Aspergillus galactomannan (GLM) or β-d-glucan and Aspergillus-specific DNA by PCR are widely used across centers (3) for supporting a probable IPA diagnosis (4). Aspergillus spp. GLM is a cell wall polysaccharide that is released in tissues during hyphal development and behaves as a surrogate biomarker of fungal invasivity (5). The clinical usefulness of GLM detection in serum or bronchoalveolar fluid (BAL) specimens for the diagnosis of IPA has been shown both in hematological and intensive care unit (ICU) patients (1, 6–9). Since its FDA clearance in 2003, the Platelia Aspergillus Antigen immunoassay (Bio-Rad, CA, USA) (10–12), marketed as a sandwich enzyme-linked immunosorbent assay microtiter plate, has become the “gold standard” for Aspergillus GLM measurement. To avoid incurring high laboratory-associated costs, batch testing is routine practice in many laboratories using the Platelia assay, which may translate into delayed result reporting. Several lateral flow assays (LFA) capable of detecting Aspergillus antigen (including GLM) in serum and BAL within the hour are commercially available. Overall, the clinical diagnostic efficacy of LFAs performed in BAL specimens for proven/probable IPA has been reported to be good in virus-associated IPA (13), although may underperform in other risk population groups compared to ELISA-based assays such as the Platelia test (14–17). A new chemiluminescence-based immunoassay, the Aspergillus GLM antigen Virclia Monotest (Vircell S.L., Granada, Spain) has been recently marketed for GLM quantification in sera and BAL specimens. The operational advantages of this immunoassay compared with the Platelia assay include its monotest format, and automated processing in a random-access platform, which makes it possible to provide “same day” results (in around 1 h) at a relatively low cost. A few studies, mainly including hematological patients, have compared the analytical and clinical performance of both immunoassays (18–20). These studies uniformly showed that, from a qualitative standpoint, the Virclia assay performs at least comparably to the Platelia assay when using sera and BAL specimens. Moreover, the Virclia immunoassay provided quantitative values in sera and BAL that correlated reasonably well with those returned by the Platelia assay (18–20). Hence, the Virclia immunoassay shows promise as to its potential use as an alternative to the Platelia assay for the diagnosis of IPA. Here, we further compared the performance of both GLM immunoassays using consecutive non-cryopreserved serum and lower respiratory tract (LRT) specimens, including BAL, bronchoscopic aspirates (BAs), and tracheal aspirates (TAs), which were prospectively collected from hematological and non-hematological patients and tested in parallel upon clinical request. Moreover, we evaluated whether quantitative values returned by the Virclia immunoassay could be reliably converted into Platelia index values over the linear dynamic range of the latter assay; this is of clinical relevance provided that the Platelia GLM index values, both in serum and BAL, above clinically validated cut-offs, have been established as mycological criteria for the diagnosis of IPA in current consensus guidelines (1, 6–9).

The Platelia Aspergillus Antigen immunoassay is nowadays the “gold standard” for Aspergillus GLM measurements in serum and BAL specimens; it is thus instrumental for the diagnosis of IPA in a variety of clinical settings (1, 6–9). The Virclia GLM immunoassay is rapid, easy to carry out, and can be performed in a monotest format, thus avoiding batch testing, which is a routine practice in most centers using the Platelia assay. Positioning of Virclia immunoassays as a mycological criterium for IPA requires a thorough assessment of its analytical and clinical performance compared to that of the Platelia assay, which was our aim in this study. To our knowledge, only three studies have previously addressed this goal (18–20). Buil et al. (20), enrolled patients with hematological disease and compared the performance of both immunoassays in BAL specimens. In turn, Troncoso et al. (18) and Leyva Calero et al. (19), in their analyses included BAL and sera collected from patients with different clinical conditions. Our study included prospectively collected consecutive specimens from patients with a wide variety of clinical conditions tested for GLM upon clinical request, allowing us to compare the performance of both methods in a “real-life setting”. In contrast, a prospective/retrospective study design was used in previous studies (19, 20), resulting in cohorts “enriched” in proven and probable IA cases. Thus, comparison across these studies (18–20) and between them and ours is not straightforward. Our study was singular; in that, it included off-label LRT samples for both immunoassays (BAS and TA). We found this approach of interest given that TA are frequently sampled in mechanically ventilated patients for microbiological studies and that BAS is a more “concentrated” and rather more homogeneous specimen than BAL, the latter because of the instillation and suction volumens may differ across centers.

Our data indicated that the Virclia assay returns more positive results than the Platelia assay in all specimen types assayed; consequently, the PPA across assays was low overall. As expected, median Virclia index values of specimens testing negative with the Platelia assay were significantly lower than those testing positive. In contrast, the number of specimens testing positive with Platelia and negative with Virclia was negligible. These data concur with that previously reported for on-label specimens (18–20). Of relevance, the degree of qualitative agreement across the assays, as assessed by Kappa statistics, was overall much higher for sera (0.56) than for BAL (≤0.24), as previously reported (10), or for BAS and TA (≤0.22). The same observation was made when hematological patients were analyzed separately. Leyva Calero et al. (19) reported a κ value of 0.72, considering sera and BAL in combination for the analysis. Potentially relevant differences between our study and theirs (19) include the number of patients with proven/probable IPA in the cohorts, patients’ clinical characteristics, the consideration of Virclia indeterminate results (excluded from the analyses herein) for analyses, and, importantly, the number of specimens per patient collected, which was more than one in around two-thirds of patients in our cohort and a single sample in the other studies (19, 20). In turn, the overall agreement reported by Buil et al. (20) for both immunoassays using BAL specimens collected from hematological patients was between 0.62 and 0.77, depending on the cut-off used for positivity (for both Virclia and Platelia). Unfortunately, no conclusion could be derived from the null agreement seen in the current study across both assays for BAL specimens from hematological patients, as there were very few of these in our sample set and none tested positive with the Platelia assay.

We next assessed the clinical diagnostic performance of both GLM assays. We reasoned that for this analysis, possible IPA cases had to be excluded, as some of which would have to be recategorized attending Platelia results; this would have introduced a clear bias skewed toward the Platelia assay. Thus, only patients with proven and probable IPA and controls (those in which IPA was ruled out) were selected for analysis. Unfortunately, this analysis was hampered by the limited number of cases, which also precluded performing subanalyses exclusively including patients of different clinical conditions or excluding off-label specimens. In this setting, the overall clinical sensitivity of the Virclia assay was higher than that of the Platelia assay (100% and 91.7%, respectively); the reverse was true regarding specificity (65% vs 89.4%, respectively). Nevertheless, the areas under the ROC curves (AUCs) were comparable (P = 0.21). Likewise the AUCs for both GLM immunoassays were comparable when the data were analyzed according to the sample type (sera, sera plus BAL or off-label specimens. Unfortunately, due to the very limited number of proven/probable IPA cased in the cohort, the study was not sufficiently powered to compare the clinical performance of both assays. The clinical sensitivity reported for the Virclia and Platelia assays in previous studies was 80.9% and 63.2%, respectively, in Leyva Calero et al. (19), and between 50% and 75% depending upon the criteria set for positivity for both assays in Buil et al. (20). In both studies, the specificity of both assays was high (above 90%) when probable/proven IPA cases were compared to non-IPA cases. Of interest, resetting the Virclia cut-off for positivity did not result in improved clinical performance (data not shown).

From a quantitative standpoint, by using the Aspergillus Galactomannan VIRCLIA sample set, we showed that the index values returned by both GLM immunoassays were optimally correlated (rho = 0.97). The degree of correlation was lower for on-label clinical samples (ρ = 0.73), a figure similar to that reported by Leyva Calero et al. (19) using sera and BAL in combination (ρ = 0.71) and by Buil et al. (20) employing just BAL specimens (ρ = 0.72). Importantly, the degree of correlation between index values returned by both assays was reported to be higher in sera than in BAL (0.88 vs 0.63) (20). Here, the correlation between index values obtained from both GLM assays substantially decreased when only BAS/TA specimens were considered for analysis (ρ = 0.52).

Interestingly, based upon GLM measurements from both immunoassays made with the Aspergillus Galactomannan VIRCLIA sample set, we built a regression model that allowed the conversion of Virclia index values to Platelia index values, defined by the formula y = (11.97 * X)/3.62 + X. Data on either GLM spiked or GLM-positive clinical serum and BAL specimens (considered in combination) fitted well to the model, whereas that of BAS/TA did not fit so well, perhaps due, at least in part, to the apparent lack of linearity of GLM measurements in these matrices above a certain GLM concentration.

The real-life prospective design is a major strength of the current study, which also had several limitations. First, the scarce number of proven and probable IPA cases in our cohort limited the soundness of the clinical sensitivity and specificity analyses. Second, we lacked comprehensive data collection regarding the use of drugs that may cause false-positive GLM results or antifungal prophylaxis/treatment, which may result in false-negative results. This is a drawback shared by the other studies mentioned above (18–20). Third, one could argue against our approach of eliminating possible IPA cases for clinical sensitivity/specificity analyses, and excluding the results of the Platelia assay for patient categorization. Nevertheless, Buil et al. (20) showed that the clinical sensitivity/specificity of both assays was quite comparable, irrespective of whether or not the Platelia GLM results were taken into consideration for patient categorization. We felt that by exclusively including proven/probable IPA cases we would increase the robustness of the analysis. Fourth, we included off-label (for both assays) BAS/TA specimens in our sample set. Leaving aside the potential clinical relevance of a positive result in BAS/TA, we proved these specimen types to be suitable for qualitative GLM detection; in fact, the limit of detection of both assays in TA, as deduced from experiments using the Aspergillus Galactomannan sample set, appeared to be even lower than in serum and BAL samples. Data on clinical specimens seemed to support this assumption, as the rate of positive results obtained with both GLM immunoassays in BAS compared with paired BAL specimens was much higher. Further studies are nevertheless required to assess the performance of the GLM immunoassays in these specimen types. Fifth, a non-validated sample set (Aspergillus Galactomannan VIRCLIA sample set) was used as a standard for certain experiments. Sixth, indeterminate results returned by the Virclia assay were eliminated rather than interpreted as positive or negative from the data set for analyses. Seventh, serial samples from more than half the patients in our cohort were available for analysis. We assume that these were collected under different clinical and therapeutic circumstances (i.e., pre- or post-IPA diagnosis or pre- or post-administration of antifungal drugs), which could have a differential impact on the performance of the two GLM immunoassays.

Because of the consolidated use of the Platelia assay in clinical practice, and to certain reluctance of clinicians to modify long-established guidelines, our model allowing the conversion of Virclia index values into Platelia values may contribute toward positioning the Virclia assay within the diagnostic algorithm of IPA. Our data warrant further prospective well-powered studies designed to assess the clinical performance of the Virclia GLM immunoassay in homogeneous clinical settings, not only using classic specimens, such as serum and BAL but also samples such as TA, which are commonly processed for microbiological studies in ICU patients undergoing mechanical ventilation.

Eliseo Albert holds a Juan Rodés Contract (JR20/00011) funded by the Carlos III Health Institute (co-financed by the European Regional Development Fund, ERDF/FEDER). Vircell S.L. (Granada, Spain) provided the reagents free of charge and had no role in study design, data collection, and interpretation, or the decision to submit the work for publication. The authors declare no conflicts of interest.

E.A., M.J.A., E.G., M.Á.C., I.T., J.C., and B.O.: Validation, investigation, data curation, manuscript review and editing. M.T., J.L.P., R.O., J.S.-C., N.C., and C.S.: Investigation, data curation, manuscript review and editing. D.N.: Methodology, resources, supervision, manuscript writing.