In this study, we report the diagnostic performance of conventional nested aspergillus PCR in the BAL fluid in a well-characterized cohort of 191 immunosuppressed patients. To our knowledge, this is the largest cohort of IFD cases being prospectively analyzed by aspergillus PCR in the BAL. Remarkably and in contrast to previous data, our results fail to support the clinical applicability of aspergillus PCR in the BAL in the immunosuppressed population, mainly due to the lack of sensitivity and specificity of the method.
Our results are in contrast to some but not all previously published studies. While several authors [27–31] suggested the sensitivity of aspergillus PCR in the BAL to range between 80% and 100%, other authors described much poorer results. For instance, Skladny et al. reports a sensitivity of 43% for proven and 33% for probable IFD using BAL PCR [19, 32]. Remarkably, false negative results in proven IFD have been described commonly. Buchheidt et al. detected negative PCR results in the BAL of two out of three patients with positive histological findings for invasive Aspergillosis  and Khot et al. reported a negative PCR result from one patient with a positive biopsy .
The low sensitivity of the assay in our study might have several reasons. Firstly, we have additionally computed the diagnostic performance of the assay by incorporating the group of “possible” IFD as potentially having IFD. As more cases not “formally” suffering from IFD are included as diseased, there is a decrease in the pre-test-probability, thus further reducing the sensitivity of the test from 36% to 26%, as depicted in Table 2. Secondly, previous studies selected patients by definite radiographic findings or pneumonia, thus improving the pre-test probability of the assay [18, 29]. Thirdly, the majority of our patient population (almost 60%) was on chronic anti-fungal therapy at the time of bronchoscopy, which might have decreased the number of organisms present in the pulmonary parenchyma and consequently in the BAL fluid, thus impairing identification by BAL PCR [34, 35]. Importantly, data on previous fungal therapy are missing in most previous reports [25, 29, 30, 36]. As compared to older studies, it is plausible that recently introduced, highly potent anti-fungal therapy such as voriconazol and new prophylactic treatment schemata’s reduce fungal load to undetectable amounts in the extra vascular compartments. Fourthly, as only selected aspergillus primers have been applied, detection of other fungal microorganisms was not possible using the current assay. Although most IFD cases are related to aspergillus infection, other fungi such as mucormycosis or rhizophus with a similar or identical clinical presentation would have been missed with the selected primers. Fifthly, it is tempting to hypothesize that fungal microorganisms are mainly present in the intravascular space, thus precluding identification in the alveolar space. And finally, we cannot exclude a technical failure of the methods, although the consistency of the findings and the broad experience of the investigators with molecular techniques make this possibility unlikely [37–40].
In the current study, only 3 patients full field the criteria of “proven” IFD according to the EORTC/NIAID classification. Nevertheless, all presented a false-negative result showing a negative PCR for Aspergillus in BAL. In the first case, the patient had been on antifungal therapy over 6 month at the time of bronchoscopy. Surprisingly, with the exception of the positive histology obtained by video-assisted thoracoscopic surgery, clinical EORTC/NIAID classification was negative due to the absence of neutropenia, fever, graft vs. host disease and current steroid therapy, e.g. “no” IFD. In the second case, the patient was on cyclosporine, had a suspicion of fungal infection due to the presence of a coin lesion on thoracic CT-scan as well as a positive aspergillus culture in the BAL, thus fulfilling the “probable” clinical criteria for IFD. The third patient with a proven biopsy had a new “fungi” suspect infiltrate in the CT scan, but low clinical signs (“possible”) for IFD according to the EORTC/NIAID classification. This exemplifies the low agreement between the EORTC/NIAID classification, histological findings and the clinical presentation in IFD. In all three cases, the suspicion of fungal disease raised by the radiologist lead to the initiation of antifungal medication.
While our results suggest an unexpected low sensitivity of the aspergillus PCR in the BAL, the specificity of the test is in line with previously reported data. Historical controls have shown specificities ranging between 84% and 100% [18, 29, 30]. This can be explained by the small number of proven invasive aspergillosis.
Our study has a few limitations. The limited number of “proven” and “probable” IFD cases hampers robust inferences on the diagnostic performance of the any test due the low prevalence of the disease (=low pre-test probability). Moreover, the lack of a reliable diagnostic gold-standard for IFD in the absence of histological samples and the heterogeneous clinical presentation of the disease limit the diagnostic certainty in the majority of cases. And finally, despite all efforts, we cannot definitively exclude the possibility of a technical failure in the assay procedure.