Polymerase Chain Reaction in the Diagnosis of Invasive Aspergillosis in Hematopoietic Stem Cell Transplant Recipients: A Meta-Analysis of Diagnostic Performance

Background: Invasive aspergillosis (IA) is associated with significant morbidity and mortality among hematopoietic-stem cell transplant (HSCT) recipients. Timely and accurate diagnosis of an active infection is needed in order to initiate targeted antifungal therapy. Although previous studies have assessed the performance of Aspergillusspp. PCR with promising findings, PCR has not yet been incorporated in the diagnostic algorithms by the guideline issuing organizations.

Aim: Our goal was to evaluate PCR as a screening or a confirmatory tool for the diagnosis of IA among HSCT patients and to determine the different parameters that could contribute to contradictory reports from the literature regarding PCR performance. Thus, we performed a meta-analysis of clinical trials that evaluated the accuracy of PCR for IA performed on serum and whole blood of HSCT patients.

Methods: We performed a bivariate meta-analysis of diagnostic data for Aspergillus spp. PCR on blood specimens across HSCT patients. We included all studies involving human subjects that assessed the performance of any PCR assay for IA on whole blood or serum and that used as reference standard the EORTC/MSG criteria. We performed 2 separate analyses either including or excluding the cases that were defined as “possible”. When possible cases where included they were classified as negative. Out of 37 total studies, 25 met strict quality criteria and were included.

Results: 25 studies with 2,595 patients were analyzed. Six studies performed PCR on serum (23%) and the remaining used whole blood assays (77%). The pooled diagnostic performance of whole blood and serum PCR was moderate, with a sensitivity of 84% (95% CI 75%-91%). Specificity was 76% (95% CI 65%-84%) when possible cases were included and 85% (95% CI 74%-91%) when excluded, suggesting that a positive or negative result is unable on its own to confirm or exclude a suspected infection. In subgroup analysis, the performance of PCR on serum was not significantly different from whole blood when possible cases were included. Interestingly, exclusion of possible cases revealed higher specificity of serum PCR (94% vs. 80%, meta-regression p-value=0.05) over whole blood without altering the sensitivity (79% vs. 85%, meta-regression p-value=0.4) and high positive likelihood ratio (LR+=12.6) suggesting that serum PCR could be used as a confirmatory test. No methodological parameter significantly affected the performance of PCR. Importantly, PCR specificity was greatly increased when 2 positive PCR results were used to define positivity (73% vs. 95%, respectively), giving a high positive likelihood ratio of 12.8, whereas sensitivity decreased (85% vs 64%, respectively). Importantly, the European AspergillusPCR Initiative (EAPCRI) recommendations improved even further the performance of PCR when at least two positive specimens were used to define ‘PCR positivity’.

Conclusions: Two positive PCR results should be considered highly indicative of an active Aspergillus spp. infection in HSCT patients. Use of the EAPCRI recommendations by clinical laboratories can further enhance PCR performance.