Introduction

Galactomannan (GM) is a fungal antigen produced by several molds including Aspergillus fumigatus. GM release in bloodstream is the rationale for its use for diagnosis [1] and management [2,3] of invasive aspergillosis. The most widely accepted test is the Platelia Aspergillus Ag assay (Bio-Rad Laboratories, Marnes la Coquette, France), an Enzyme Linked Immunosorbent Assay using a rat monoclonal antibody [4]. The test is mainly used as a screening test in serum of patients at risk of invasive aspergillosis and has been implemented in numerous hematological wards since the 90’s [5–7]; it is currently accepted as a microbiological criterion for the diagnosis of invasive aspergillosis [8]. Therefore, a GM positive result is often the only element [9], although concerns have been raised about the accuracy of GM testing [1,10].

Performance of the assay is variable, with metaanalysis showing pooled sensitivity of 0.71 and pooled specificity of 0.89 in proven cases of invasive aspergillosis, i.e. based on biopsy diagnosis [11]. False reactivity has always been observed from 5% in adults to as much as 83% in newborn babies [10]. This false positivity rate has led the manufacturer to recommend retesting of positive samples. This resulted in publications underlying the possibility to observe unreproducible positive results [12–20]. Recent reports have also questioned the role of long sample storage in the lack of reproducibility with different conclusions [13,15,16,20–22].

We therefore performed a formal comparison between the routine laboratory diagnosis and a Bio-Rad referent operator who performed additional testing blind to the routine one. Based on this intra-laboratory reproducibility study, our purpose was to characterize the analytical false positive results occurring during testing and to distinguish them from positive results related to the serum and the patient. Among the possible limitations of the GM assay, little attention has been paid to the initial preanalytical preparation steps of the assay as a cause of false GM positivity leading to imprecise or erroneous conclusions in several previous studies.

Material and Methods

GM assay

GM detection was performed by Platelia Aspergillus Ag Assay (Ref#62794) following the manufacturer's instructions (Bio-Rad Laboratories). Serum handling was performed under safety cabinet until the heating phase of the procedure. Heating at 100°C was performed in water bath (TW 12, Julabo, Seelbach, Germany). The R3, R4 and R5 reagents are defined as negative control serum, cut-off control serum, and positive control serum, respectively. The optical density (OD) results were translated in GM-OD index (GM-ODI) using the ratio OD-sample/OD-R4 (mean of two replicates). The positivity threshold is GM-ODI ≥0.5 as recommended by the manufacturer. An internal control (mix of different R4 reagents of the same batch since the positive control of the kit cannot be used as interlot control since it is not always calibrated at the same level) was assessed in each series to calculate the inter-assay coefficient of variation (CV). The mean of 104 internal control GM ODIs was 0.94+/-0.14, i.e. a CV of 14.5%. The 95% confidence interval (95% CI) for the 0.5 GM-ODI threshold was 0.36–0.64, after the formula 0.5 +/- 1.96 SD with SD = 0.5 CV.

We tested all the consecutive sera over 22 consecutive working days using the #62794 kit from February 17^th to March 10^th 2014. At day 0, each serum sample was tested four times (Fig 1). One test was performed by three different routine technicians (Operator#1 Test#1) over the study period without any change to the usual procedure. The three other tests were performed by a unique technician (Operator#2) blind of routine results; Test#2 was performed on the same GM extraction as Test#1, and Tests #2 and #3 on a different GM extraction.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Schematic representation of the reproducibility study.

Each serum was extracted twice by two different workers. Interpretation was as shown knowing that other combinations were possible but that only those obtained during the study are shown. Results close to the positivity threshold value (GM optical density index = 0.5) were classified as non-conclusive taking into account the interassay coefficient of variation of the kit.

https://doi.org/10.1371/journal.pone.0124044.g001

At day 1, all the samples with at least one of the four tests (Test#1-#4) GM-ODI>0.5 were retested similarly as Day 0. Test#5 was performed by three different routine technicians (Operator#1 Extraction#3 Test#5) following the usual procedure. Three other tests were performed by Operator#2; Test#6 was performed on Extraction#3, and Tests #7 and #8 on Extraction#4.

The samples were classified as confirmed negative (Test#1–4 with ODI <0.5), confirmed positive (Test#1–4 with ODI ≥0.5) and unreproducible positive subdivided in extraction and ELISA unreproducible positive. Extraction unreproducible positive corresponded to a positivity linked to one of the two extractions (Test#1 and Test#2 ODIs ≥0.5, and Test#3 and Test#4 ODIs <0.5, or conversely). ELISA unreproducible positive corresponded to the positivity related to the ELISA performing with only one replicate with GM-ODI ≥0.5 (Fig 1). When at least one Test#1-#4 GM-ODI value was <0.5 but with the mean of the four tests within the 95% CI (0.36–0.64), these samples were classified as non-conclusive. A new classification was performed after retesting at Day 1 using the same criteria.

The remaining serum after the reproducibility study was stored at -20°C for 8 months. After thawing, a new GM extraction was performed and Test#9 performed by a unique routine technician in a single run. Test#9 was compared to Test#1.

Results

We tested 567 sera from 235 patients over 22 consecutive working days in 16 series with the lot # 3L0024 of the kit. Seventeen samples were excluded from the reproducibility analysis since their volume did not allow all the tests.

Among the 550 remaining samples, 520 (94.5%) were confirmed negative upon Test#1-#4 results and 30 sera (5.5%) had at least one test positive (Table 1). Fifteen of these 30 sera (50%) were confirmed positive with upon Test#1-#4positive results, of which two (S203 and S 277, Table 1) were from one patient with probable invasive aspergillosis according to the European Organization for Research and Treatment of Cancer/ Mycoses Study Group [8].; 5 sera (16.7%) were non-conclusive; 10 sera (33.3%) were unreproducible positive divided in 6/10 (60%) extraction and 4/10 (40%) ELISA unreproducible positive (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Galactomannan optical density index (GM-ODI) results of the reproducibility study upon the four tests (Test#1-#4) of the serum samples performed at Day 0, the four tests (Test#5-#8) performed at Day 1, and the GM-ODI after 8 month storage at -20°C (Test#9).

https://doi.org/10.1371/journal.pone.0124044.t001

When retested at Day 1 with four additional tests (Test#5–8) after storage at 4°C, the Day 0 confirmed positives and non-conclusives were similarly classified but for one non-conclusive which turned confirmed positive (S311 Table 1). For the Day 0 unreproducible positive results, all turned confirmed negative but for one which was classified as non-conclusive (S25 Table 1). Table 1 also shows Test#5 compared with Test#1 results after storage at 4°C for <72h as in our routine conditions. Test#5 results of the confirmed positives and non-conclusive were similar to Test#1 GM-ODIs (paired t-test: p = 0.49). Quantitatively, we observed two distinct populations when comparing Day 0 and Day 1 results. On one hand, the confirmed positives and non-conclusives with no GM-ODI variation (slope at 0.99). On the other hand, the ELISA and extraction unreproducible positives represented a distinct group of sera (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Graph plotting initial Test#1 against Test#5 (see Fig 1) performed after serum storage ≤72h at +4°C.

A linear trend line was observed for confirmed positive and unreproducible positive results as in Fig 2.

https://doi.org/10.1371/journal.pone.0124044.g002

Among the 30 sera with at least one GM-ODI>0.5 at Day 0, 27 had enough serum left to be stored at -20°C which were retested (Extraction#5 Test#9) after 8 months with the lot # 4G1027 of the kit. Table 1 shows Test#9 in comparison with Test#1 results. Once again, the confirmed positive and non-conclusive sera had similar GM-ODI (paired t-test: p = 0.63) whereas the unreproducible positives at Day 0 were all retested negative as at Day1.

Discussion

The present reproducibility study allowed us to clearly delineate on initial testing of the positive GM results between those which were confirmed positive and the unreproducible positive results. For the first time, we quantified the extraction and ELISA causes of these unreproducible positives, showing that both can intervene at a global rate of 33% (10/30), within the range of previous reports: 39% [16], 40% [14], 48% [12], 55% [18], and 64% [15] of the tested samples. Therefore, in contrast to hypotheses which favor a particular component and/or storage of the serum to explain the unreproducible positive results [13,15,16], we highlighted that most of them, if not all, were due to technical variation either during the extraction process or in the ELISA plate itself. Therefore two populations of sera can be differentiated; the first one with a positive result related to the sample for which the GM-ODI remains unchanged upon retesting, either after storage at 4°C for <72h or at—20°C for at least 8 months, and the second one with an initial positive result related to technical variation for which a second testing after storage at 4°C for <72h is negative.

These results lead us to reconsider all articles that have been recently published on GM reproducibility and stability during storage where these two populations were not identified [13,15,16,20–22]. In contrast to the present study, the previous studies performed only one test on first determination and one repeat after storage [13,15,16]. Consequently they could not distinguish the initial positivity related to samples from that related to any technical problem. Interestingly, a lot of the figures from these studies look like ours with two groups of sera. One group for which there is no or low difference between tests and the other one for which the samples completely turned negative. Therefore, the conclusion that freezing reduces the risk of a false positive result [15] cannot be maintained without confirming that the initial positivity is a confirmed positive. In contrast, some studies like ours showed little impact of long-term storage at—20°C on GM-ODI values and that storage did not affect the sera from patients with invasive aspergillosis [20]. One can hypothesize that sera from patients with invasive aspergillosis were confirmed positive, hence the lack of impact of freezing on GM-ODIs. Similarly, several reports [12,14,16,17] noticed that the samples that turned negative on repeated test corresponded to patients with a low probability of invasive aspergillosis, which is also consistent with our results.

Unfortunately, we could not ascribe unreproducible positive results to a specific procedure failure or to a specific technician. The manual processing of the test is very sensitive to small variations of the heat extraction procedure for instance. An external source independent to the operator can occur. One can suspect GM contamination from the water-bath, which can be regularly checked, or from tips or tubes for treatment, which could make the use of antigen-free material compulsory. The various origins can explain the different rates of unreproducible positive results between laboratories. Whatever the source of this lack of reproducibility, it strongly supports re-analysis of a positive sample as soon as possible [1]. Of course, to perform a second test on the same serum does not completely exclude the possibility to obtain a second unreproducible positive result. Therefore, obtaining serial samples in parallel is also recommended [1], acknowledging that the risk of false positive cannot be definitively excluded. The non-conclusive category we created for the reproducibility study is questionable, since we did not classify non-conclusive sera with all GM tests positive but within the CV of the GM assay. However, in case of possible false positive results and non-conclusive results, given the risk of misidentifying an invasive aspergillosis, to provide a positive answer to the clinicians seems preferable. That is why we have not created a non-conclusive category in our routine procedure and insist on the GM-ODI kinetics upon serial samples for final correct interpretation.

The present work was focused on the analytical performances of the Platelia Aspergillus Ag kit and was not designed for deciphering between false positives due to an exogenous source of GM in serum as a consequence of blood contamination with antibiotics [23] or of other invasive fungal infections such as histoplasmosis, cryptococcosis, or penicilliosis [24]. Nor do we address the issue of false positive of digestive origin [10,25]. We think that confirmed GM positives, as a general rule, must be interpreted with care using all the information needed to make a final diagnosis of invasive aspergillosis.

In conclusion, the reasons why some sera turned totally negative on retesting are operator-dependent. We also showed the good stability of GM for at least 72h at +4°C and over 8 months at -20°C when dealing with the confirmed positive sera only. As a consequence, analytical false positive results must be eliminated in performing a second evaluation of the sample as soon as possible to avoid useless investigations or treatments and for correct interpretation when re-testing after storage. Further research on unreproducible positive results is warranted to decrease their occurrence, knowing that full automation of the assay is unlikely for technical reasons.

Supporting Information

Acknowledgments

We thank all the staff of the Saint Louis mycology laboratory for their technical involvement in the study and especially Mrs. Véronique Miquelis, Isabelle Cadic, and Marie-Claude Lambert. Disclaimer: Presented in part at 54th Interscience Conference on Antimicrobial Agents and Chemotherapy September 5–9, 2014 | Washington (e-poster#1863).