Performance Characteristics of the Platelia Aspergillus Enzyme Immunoassay for Detection of Aspergillus Galactomannan Antigen in Bronchoalveolar Lavage Fluid

MATERIALS AND METHODS

RESULTS

Precision and reproducibility were compared in serum and BAL by testing normal human serum or BAL that had been shown to be negative in the Platelia Aspergillus EIA, into which the positive kit control and cutoff control were reconstituted. In serum, the averages ± one standard deviation and the coefficient of variation (indicated in parentheses) of five aliquots of the positive control were 2.93 ± 0.35 (0.12) on day 1 and 3.23 ± 0.25 (0.08) on day 2. For the cutoff control, the results were 0.61 ± 0.12 (0.19) on day 1 and 0.59 ± 0.06 (0.10) on day 2. In BAL, the average of five aliquots of the positive control were 3.40 ± 0.20 (0.06) on day 1 and 3.61 ± 0.09 (0.02) on day 2, while the averages of the cutoff control were 0.86 ± 0.05 (0.06) on day 1 and 0.91 ± 0.03 (0.04) on day 2.

Using the 0.5 or 1.0 cutoff, the sensitivity was 81.8% among patients with IA (Fig. 1). Only one sample per patient was used to calculate the sensitivity and specificity. The two false-negative results occurred in lung transplant patients who were receiving mold-active antifungal agents, itraconazole and voriconazole in one patient each, while the results were positive in all nine IA patients who were not receiving mold-active antifungal agents (P = 0.018; Fisher exact test). The effect of mold-active antifungal agents (amphotericin B administered intravenously or aerosolized, itraconazole, and voriconazole) also was assessed in patients who underwent diagnostic BAL for evaluation of symptoms or chest CT abnormalities or in whom Aspergillus or other mold was isolated from BAL but in which IA was not diagnosed. Among patients whose BALs were classified as diagnostic, GM was detected in 4 of 7 (57.1%) patients who were receiving mold-active antifungal agents compared to 5 of 59 (8.5%) who were not receiving them (P = 0.005). Among patients colonized with mold, GM was detected in 3 of 16 (18.8%) patients who were receiving mold-active antifungal agents compared to 5 of 40 (12.5%) who were not (P = 0.676). GM antigenemia was detected in one of four (25%) IA patients with paired BAL and serum.

• Open in new tab
• Download powerpoint

FIG. 1.

GM detection in BAL in patients with IA and controls. The broken horizontal line at 0.5 is the cutoff for positivity, and the number within the box (n) represents the number of patients with negative results. A dashed line at 1.0 is provided for reference for estimation of the accuracy at a higher cutoff.

Specificity in lung transplant patients who underwent surveillance BAL to monitor for rejection or infection was 95.8% using the 0.5 cutoff and 96.6% using the 1.0 cutoff (Table 1). Among the surveillance controls, GM was detected more often in specimens deemed to be colonized with Aspergillus (15.4%) or other mold (10%) than in specimens with negative cultures (1.2%) (Table 2). Most of the patients were receiving mold-active antifungal prophylaxis at the time of surveillance BAL, as they received prophylaxis for 3 months after transplantation.

In the diagnostic BAL group, positive results were noted in 27.3% at the 0.5 cutoff and 13.6% at the 1.0 cutoff. Positive results were more common after lung transplantation (46.2%) than after transplantation with other organs (15.0%; P = 0.010) (Table 2). In the lung transplant group, positive results were more common in specimens colonized with Aspergillus (60%) than in noncolonized specimens (38.5%; P = 0.414) or those colonized with other mold (33.3%; P = 0.559). Three of nineteen (15.8%) BALs containing Penicillium spp. and one containing Fusarium were positive. The results were negative in all four specimens containing Paecillomyces spp. Four of the eighteen (22.2%) lung transplant patients with positive BALs subsequently developed IA. Positive results were not observed in healthy subjects, none of whom were neutropenic or receiving immunosuppressive medications.

Four patients among the diagnostic controls were receiving piperacillin-tazobactam, of whom GM was detected in the BAL in three. These four were lung transplant patients, and two had positive cultures for Aspergillus or other mold (Paecillomyces). Serum was not tested for GM in these patients.

Among transplant controls from whom Aspergillus or other mold was isolated, 21.4% were positive at the 0.5 cutoff and 14.3% were positive at a 1.0 cutoff (Table 1). Of noncolonized controls, BAL was positive at the 0.5 cutoff in 7.3% compared to 1.3% at a 1.0 cutoff. Some patients underwent several bronchoscopies but were evaluated only once per category, explaining why the total for this parameter is 206.

The negative predictive values (NPV) ranged from 97.3 to 99.0%, depending upon the cutoff for positivity and prevalence (Table 3). The NPV for culture or histopathology were slightly lower, ranging from 96.4 to 98.4%. Positive predictive values (PPV) for antigen detection were higher with the 1.0 cutoff than with the 0.5 cutoff, especially if the diagnostic BALs were included in the calculation. The PPV also was higher for antigen detection than for culture or histopathology in all of the comparisons except for diagnostic BALs at the 0.5 cutoff.

The reproducibility was assessed by comparing the results of the first and second test for all positive specimens. The reproducibility was 93.3% for specimens between 0.5 and 0.9 (n = 15) and 100% for those ≥1.0 (n = 17) (Fig. 2).

• Open in new tab
• Download powerpoint

FIG. 2.

Linear regression analysis of positive results as determined on consecutive days. The horizontal axis represents the results of initial testing (test 1), and the vertical axis represents the results of repeat testing (test 2).

DISCUSSION

Our findings show that precision and reproducibility were comparable in BAL and serum. Sensitivity was 82% in patients with IA, both at the 0.5 and at the 1.0 cutoff. The two false-negative cases both were receiving mold-active antifungal therapy, a known cause for false-negative antigenemia (8). Diagnostic BAL and specimens colonized with Aspergillus or other molds were positive more often in patients receiving mold-active therapy, however. The reasons for a higher positivity rate in those receiving mold-active therapy are unknown but could include undiagnosed invasive disease, which led to the initiation of therapy. Antigenemia was detected in only one of four IA patients. The experience during clinical testing at MiraVista Diagnostics supports the higher sensitivity in BAL than serum. For example, antigenemia was detected in only 19.4% of patients in whom a simultaneous BAL specimen was positive. The higher sensitivity of BAL also has been noted by others (1, 9). The sensitivity was higher for antigen detection than was the histopathology or culture of BAL specimens.

The specificity was 95.8% in surveillance BAL specimens from lung transplant patients. Increasing the cutoff from 0.5 to 1.0 increased the specificity by 0.8%. The specificity for culture and/or histopathology was 89.1% in surveillance BALs. The PPV was higher for antigen detection than for culture or histopathology at either cutoff. For example, assuming 5% prevalence, the PPV for antigen detection was 50.6% at the 0.5 cutoff and 55.9% at the 1.0 cutoff compared to 26.0% for culture or histopathology. If the surveillance and diagnostic BALs are combined, the PPV for antigen detection was 25.8% at the 0.5 cutoff and 39.8% at the 1.0 cutoff compared to 22.1% for culture or histopathology. Thus, antigen detection appears to be more accurate than culture or histopathology when BAL is used for the diagnosis of IA. Although a greater accuracy may be achieved using lung biopsy specimens, surgical biopsy is rarely performed for evaluation of symptoms or CT abnormalities in transplant patients.

Positive results were more common in transplant controls who underwent diagnostic BAL (27.3%), those colonized with Aspergillus or other mold (21.4%), and especially those who had undergone lung transplantation (46.2%). Raising the cutoff to 1.0 reduced the positivity rate in the diagnostic or colonized groups from 24.5 to 13.9%. Positive results in some of these controls may have been caused by undiagnosed IA or other mold infection, as evidenced by subsequent diagnosis of IA in four of our patients with positive diagnostic BALs. Positive results were not observed in healthy individuals not at risk for IA or airway colonization with mold.

Piperacillin-tazobactam may have caused a positive result in a few patients. Although three of four lung transplant patients receiving it had positive results, two also had positive cultures for Aspergillus or Paecillomyces. Others have also reported positive results in BAL from patients receiving piperacillin-tazobactam (9). Two of the three BAL specimens from our patients receiving piperacillin-tazobactam also were positive at the 1.0 cutoff.

In conclusion, the performance characteristics of the Platelia Aspergillus EIA in BAL were similar to those in serum, but the sensitivity was higher in BAL. Use of a 1.0 cutoff improved the specificity, but results between 0.5 and 0.9 should not be disregarded without careful consideration of the diagnosis of IA, since some of these patients may have IA or develop it later. Antigen detection should be added to culture and histopathology in the evaluation of pulmonary infiltrates in organ transplant patients.