The results of this study showed that the target region of 28S-ITS2 is specific and applicable to the genetic diagnosis of Aspergillus and that the whole blood qPCR assay may be a very useful and reliable method for the early diagnosis of IA in patients with hematological malignancies.
Over the past decade, many researchers have explored and evaluated PCR assays for the early diagnosis of IA [10, 21–24], while the European Aspergillus PCR Initiative (EAPCRI) has performed a lot of meaningful work around PCR standardization [13, 19, 25]. However, PCR is still not included in the EORTC/MSG criteria . Meanwhile, there is a major consensus in favor of qPCR technology due to its sensitivity and speed [13, 16, 24, 26–28], especially after release of the Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines, which has encouraged better experimental practice for more reliable and unequivocal interpretation of qPCR results [14, 29]. A critical issue of qPCR is the choice of the target DNA, which is highly debated since it is directly associated with assay sensitivity and specificity. To date, several genes have been used to assess qPCR assay results in the diagnosis of IA [10, 11, 15], and the ribosomal DNA gene region has been used more and more frequently and shows the greatest potential [16–18, 25, 29, 30]. Ribosomal DNA comprises transcriptional domains (5S, 5.8S, 18S, and 28S), which evolve relatively slowly and have higher conservation rates, and non-transcriptional domains (ITS1 and ITS2), which are more variable . A multicenter study comparing the use of the 28S and 18S primer sets targeting the D1–D2 region of the large ribosomal subunit gene revealed that the former was more specific, probably because when the 18S region is targeted, a portion of the human rDNA gene needs to be amplified, especially in the absence of A. fumigatus DNA [18, 32].
Overall, use of the 28S primer set is more suitable than the 18S primer set for the diagnosis of Aspergillus. The primers in our study were derived from the 28S-ITS2 region in the GenBank database and targeted both the conservative and variable regions as described before, while the probe was modified by minor groove binder, which further confirmed its specificity. The target sequence was aligned using BLAST and tested for analytic sensitivity and specificity using 15 Aspergillus species, 24 Candida species, 10 bacterial species, and serial 10-fold dilutions of A. fumigatus DNA, with the results of 6 copies and 100%, respectively, indicating the applicability of the Aspergillus qPCR assay.
With regard to the best blood fraction to test, specimen choice had critical implications in the extraction methodology and PCR results. Historically, the pathogenesis of IA has been poorly understood, and knowledge regarding the source of Aspergillus nucleic acids in the blood is limited with no consensus having been achieved to date. Testing serum specimens using PCR is dependent on the detection of free circulating Aspergillus DNA, while the use of EDTA-anticoagulated whole blood samples allows for the detection of conidia, hyphal fragments, and freely circulating DNA [12, 18]. In our study, plasma and whole blood were compared for Aspergillus DNA extraction, and the results showed greater numbers of initial DNA copies in whole blood than in plasma. Thus, whole blood was the better specimen for Aspergillus DNA extraction, a finding that was consistent with those of earlier reports and the EAPCRI recommendation [12, 13, 33].
To initially assess the clinical applicability of this Aspergillus qPCR, 86 blood samples from 86 individuals were tested: 72 of these individuals were febrile and had hematological malignancies, while 41 were diagnosed with IA according to clinical criteria. The qPCR results and clinical data were analyzed using ROC . The main obstacle we encountered was that the number of proven cases of aspergillosis was too low to be used as the sole criterion for diagnosis since the blood cultures were almost always negative for Aspergillus species. The probability of confirming the diagnosis histopathologically was also low because patient status was not suitable for invasive diagnostic procedures. According to the qPCR results, the proven and probable cases were defined as being within the reference standard category on the basis of the EORTC/MSG criteria, the only accepted tool, although it has some limitations [20, 34].
The choice of an optimal cut-off value depends on the test purpose. An ROC analysis was conducted that revealed a good AUC value (0.828). In clinical practice, there are few effective clinical approaches for the diagnosis of IA; once a diagnosis is missed, patient prognosis is very poor. Hence, we chose a qPCR index that yielded an adequately high sensitivity, an acceptable false-positive rate (1-specificity), and a better Youden index, which reflects integrated diagnostic ability. In the ROC analysis, a cut-off value of 25.24 seemed to yield the optimal trade-off (Figure 2). The qPCR cut-off value of ≥25 copies/μL was tentatively established as the most suitable for obtaining a high yield in clinical screening for IA, while the Cq value was 36, consistent with that of the EAPCRI report (mean, 35.3; upper limit, 38.8 cycles) . According to this threshold value, the diagnostic sensitivity and specificity were 90.9% and 73.4%, respectively, while meta-analysis showed that these values would be 88% (95% CI, 75–94%) and 75% (95% CI, 63–84%) if only a single positive sample was required . With this threshold, the specificity (73.4%) seemed suboptimal, and 41% (7/17) of the patients with false-positive results were diagnosed as possible IA and 59% (10/17) were diagnosed as no IA, which accounted for the possible IA and no IA rates of 37% (7/19) and 22% (10/45), respectively (Table 3). We analyzed these patients and speculated that their degraded diagnoses were mainly due to the limitations of the EORTC/MSG criteria  and the fact that high sensitivity can also cause false-positive results due to low-level contamination . Due to the heterogeneity of the different studies, there is no way to compare them. Of course, increasing the size of the study population, especially the number of proven cases, may result in a better cut-off index and a more suitable guide for the presumptive diagnosis of IA in future studies.
Another limitation of this study was that we did not collect serial specimens for qPCR since we immediately administered empirical antifungal treatment to patients with neutropenic fever, which was thought to affect the diagnostic detection . Otherwise, no association was found between Aspergillus DNA detection and treatment efficacy or patient outcome [38–41], the probable reason for this being the lower DNA loads in the blood, which made reliable quantification difficult due to the random dispersion of the few available DNA copies . The current use of Aspergillus qPCR as a diagnostic screening tool is recommended instead of evaluation for therapy, and a single PCR-negative result is sufficient to exclude the diagnosis of IA because of this assay’s high sensitivity [10, 16, 42]. The GM test was not performed simultaneously with the one-time qPCR for each patient in this study since Aspergillus GM and DNA release do not occur in a parallel fashion due to IA pathogenesis and host factor changes that occur over time [43–45]. Additionally, several studies have indicated that qPCR was comparable or better to GM for IA diagnosis and that it is more scientific and appropriate to combine GM and qPCR for IA diagnosis [42, 45–47].