Aspergillus specific nested PCR from the site of infection is superior to testing concurrent blood samples in immunocompromised patients with suspected invasive aspergillosis

Invasive aspergillosis (IA) is a severe complication in immunocompromised patients. Early diagnosis is crucial to decrease its high mortality, yet the diagnostic gold standard (histopathology and culture) is time‐consuming and cannot offer early confirmation of IA. Detection of IA by polymerase chain reaction (PCR) shows promising potential. Various studies have analysed its diagnostic performance in different clinical settings, especially addressing optimal specimen selection. However, direct comparison of different types of specimens in individual patients though essential, is rarely reported. We systematically assessed the diagnostic performance of an Aspergillus‐specific nested PCR by investigating specimens from the site of infection and comparing it with concurrent blood samples in individual patients (pts) with IA. In a retrospective multicenter analysis PCR was performed on clinical specimens (n = 138) of immunocompromised high‐risk pts (n = 133) from the site of infection together with concurrent blood samples. 38 pts were classified as proven/probable, 67 as possible and 28 as no IA according to 2008 European Organization for Research and Treatment of Cancer/Mycoses Study Group consensus definitions. A considerably superior performance of PCR from the site of infection was observed particularly in pts during antifungal prophylaxis (AFP)/antifungal therapy (AFT). Besides a specificity of 85%, sensitivity varied markedly in BAL (64%), CSF (100%), tissue samples (67%) as opposed to concurrent blood samples (8%). Our results further emphasise the need for investigating clinical samples from the site of infection in case of suspected IA to further establish or rule out the diagnosis.


| INTRODUC TI ON
Invasive aspergillosis (IA) is a severe, life-threatening complication in immunocompromised patients (pts), contributing significantly to morbidity and mortality. 1 Definitive diagnosis of IA remains challenging as present diagnostic standards provide suboptimal accuracy. 2 Aspergillus spp. are the most frequent fungal pathogen in haematological high risk pts and invasive pulmonary aspergillosis is the most common cause of mortality due to mould disease. 3 The highest incidence (10% to 20%) and mortality rates (60% to 90%) of IA have been reported following allogeneic hematopoietic stem cell transplantation (HSCT) and heart or lung transplantation. 4,5 IA typically represents a localised infection with the respiratory tract being the main site of infection 6 leading to invasive pulmonary aspergillosis (IPA). Early diagnosis is crucial to decrease the high mortality, 7 but distinct proof of infection is only defined by the diagnostic gold standard comprising of histopathology and culture which is time consuming and cannot offer early confirmation of IA in the vast majority of cases. As a result, IA is underdiagnosed and, in most cases, antifungal therapeutic efforts are initiated empirically merely based on clinical criteria, e.g. suspicious lung infiltrates on chest computed tomography (CT) scans. The limitations of this approach have been depicted in a study that found only around 50% of pts with CT based IA diagnosis to truly harbour IA after surgical removal of suspected tissue. 8 This diagnostic dilemma consequently leads to overtreatment and high healthcare costs. 9 Sensitivity of culture-based methods has been reported to be as low as 30% 10 and its results are often not congruent with histological results. 11 There is an urgent need for research into diagnostic tools capable of detecting infection before disease manifests. However, the impaired clinical condition of haematological pts at risk for IA often prohibits invasive diagnostic procedures (eg tissue sampling). Thus, the search for improved diagnostic tools has focused on biomarkers such as Galactomannan (GM), Beta-D-Glucan (BDG), Aspergillus-specific polymerase chain reaction (PCR) or the lateral flow device (LFD) (reviewed in 12 20 which may be attributed to the higher fungal burden observed at the site of infection. 21 Besides methodological differences, selection of specimens is another decisive aspect in the context of optimised fungal diagnostics. There is currently limited data that explicitly focuses on the direct comparison of an Aspergillus specific PCR in blood and in samples from the site of infection from immunocompromised pts harvested at the same time. The significance of PCR assays for detection of fungal pathogens is rising and it will supposedly find its way into the upcoming new EORTC/MSG criteria and is already included in the recent ECCMID Guidelines. 22 In this context we further investigated the diagnostic potential of an established Aspergillus PCR assay for diagnosis of IA with special regard to optimal type of specimen and prior antifungal treatment. We directly compared the diagnostic performance of an Aspergillus specific PCR in multiple specimens representing the direct site of infection (BAL, CSF, tissue) with same-day blood samples (obtained within 24 h) from 133 individual pts.

| PATIENTS AND ME THODS
The study was approved by the local Ethics Committee (Ethics Committee of the Faculty of Medicine, University of Heidelberg, Germany) and samples were obtained according to good clinical practice guidelines (GCP) in compliance with the Declaration of Helsinki.
Due to the retrospective nature of the study protocol to retrospectively investigate the clinical data of patients treated at their own local institutions in a scientific intent and data were obtained anonymised concurrently, the need for written informed consent was waived according to German Ethics Committees regulations. 23 The study included 133 pts at risk for IA presenting with early signs of suspected IA and for whom 138 samples from the sites of infection 38 pts were classified as proven/probable, 67 as possible and 28 as no IA according

| Patients characteristics
The majority of pts suffered from malignant haematological diseases. Six pts suffered from solid tumours with consecutive therapy-associated intensive neutropaenia. Twenty-five patients received allogeneic and two patients received autologous stem cell transplantation. Patient characteristics are summarised in Table 1

| Underlying antifungal therapy (AFT)
Seventy-six percent of the patients were under mould-active antifungal prophylaxis (AFP) or therapy (AFT) at the time of diagnostic procedures. Patients must have received a full daily dosage of AFT prior to sampling in order to be counted as having received AFT. The median number of antifungal regimes prior to specimen sampling was 1 (range 0-3).

| Bronchoalveolar lavage
Bronchoscopy with BAL was performed as described elsewhere [8], and BAL samples were obtained in a sterile tube without conservation media. The mean sample volume was 10 mL.

| Same day (ie concurrent) blood samples
Blood samples included in this trial had to be drawn within 24 h relative to the sampling from the infectious site in order to be included in this study. For patients with repeated blood sampling the frequency of serial blood sampling was between 1-6 days for all cases according to clinical data record sheets.

| Nested Aspergillus PCR
Total DNA was extracted from 1.5 mL of the leucocyte pellet of BAL samples and was processed by an experienced technical assistant unin-formed of the clinical data, to a previously described DNA extraction and nested PCR protocol. 25 During the initial implementa-

| Statistical analysis
Patients' data were collected and inserted into a database  For determining diagnostic performance proven and probable IA patients (n = 38) served as positive population while patients classified as "noIA" (for whom IA was highly unlikely based on insufficient host factor criteria or compatible radiological findings) (n = 28) were used as negative population.

| RE SULTS
Results for patients with possible IA (n = 67) were described, but were excluded from the analysis of PCR diagnostic performance.  Table 2.
Comparison of PCR sensitivity in each type of sample is depicted in Figure 1 and significance is compared by Dunn's Multiple Comparison Test. PCR in BAL and PCR in CSF were significantly more sensitive compared to PCR in the concurrent corresponding blood samples (P < .005).

| Influence of mould active antifungal treatment
In proven and probable IA patients 76% received at least one dosage of mould-active antifungal treatment prior to sampling. Sensitivity of Aspergillus PCR from the site of infection for patients with underlying AFT was 64% compared to 8% for same day blood samples.
Proven/probable IA patients having received more than one antifungal (n = 6) prior to sampling showed a further decrease in sensitivity of 50% for BAL PCR. For patients without underlying AFT sensitivity was 75% for samples from the site of the infection compared to 12% from a single concurrent blood sample.

| Repeated blood sampling and testing
Data on serial blood PCR testing was available for 25 patients.
Analysing serial blood samples led to an increase in blood PCR sensitivity (15% [95%CI 7%-29%]), but was still significantly inferior to PCR sensitivity in specimens from the site of infection (P < .01).

| D ISCUSS I ON
Although Aspergillus-specific PCR has been widely utilised in clinical practice for patients suspected for IA, there is still ongoing debate on the optimal type of specimen, especially as sampling from the site of the infection (such as BAL) is more invasive and potentially associated with a higher risk than analysing blood samples.
Our data demonstrate that the diagnostic performance of PCR is significantly higher in samples representing the infectious site, in most cases BAL. Sensitivity of the PCR in all samples from the site of the infection combined (BAL, CSF, tissue samples) was 70% and thus strikingly higher than in concurrent blood samples obtained directly before or after the diagnostic procedure (sensitivity 8%), but in line with other studies. 26 By obtaining same day blood samples confounding factors such as differences in stage of disease or type and duration of antifungal therapy are equally adjusted. The observed impaired sensitivity biomarkers such as PCR from blood specimens compared to concurrent BAL from the same patients has recently been demonstrated and had been attributed to the high prevalence of underlying AFT and single blood testing in a recent trial. 27 Furthermore, other studies such as the AMBILOAD Trial 28 have resulted in similar values for diagnostic performance of IA PCR in blood samples. The main reason for the poor diagnostic performance of the blood PCR is most likely previous antifungal treatment or prophylaxis, which has been shown to greatly impair blood PCR performance. 15,16 In our study 76% of the patients were under mould-active antifungal prophylaxis (AFP) or therapy (AFT) at the time of diagnostic procedures. This high rate is due to common clinical practice nowadays, particularly in the context of acute leukaemia as shown by survey data. 29 The results of our study further outline this susceptibility of blood PCR to antifungal treatment and direct comparison illustrates that PCR in BAL, but also in CSF or tissue samples seems to be less affected by AFT, very likely due to higher fungal burden at the site of the infection, which is also confirmed in studies with preclinical animal models. 30 The sensitivity values we observed of 60 to 70% are comparable to other studies including patients under AFT, but are nevertheless worse compared to most of the results obtained from untreated patient cohorts. 31 In addition, many studies addressing the issue of fungal diagnostics are hampered by low numbers of pts with proven or probable IA. 32 Among the patients included within this study were 38 patients classified as proven/probable during the study period of more than 10 years which is an ample number with regard to other studies. An essential advantage of the blood PCR over the more invasive BAL, CSF or tissue samples is the possibility of easy serial sampling.
We therefore also analysed the diagnostic performance of the blood PCR in the context of repetitive sampling which has been shown to confer higher sensitivity, at least for GM. 37