Invasive aspergillosis (IA) is the most common life-threatening opportunistic invasive mould infection in immunocompromised people. Early diagnosis of IA and prompt administration of appropriate antifungal treatment are critical to the survival of people with IA. Antifungal drugs can be given as prophylaxis or empirical therapy, instigated on the basis of a diagnostic strategy (the pre-emptive approach) or for treating established disease. Consequently there is an urgent need for research into both new diagnostic tools and drug treatment strategies. Newer methods such as polymerase chain reaction (PCR) to detect fungal nucleic acids are increasingly being investigated.
To provide an overall summary of the diagnostic accuracy of PCR-based tests on blood specimens for the diagnosis of IA in immunocompromised people.
We searched MEDLINE (1946 to June 2015) and EMBASE (1980 to June 2015). We also searched LILACS, DARE, Health Technology Assessment, Web of Science and Scopus to June 2015. We checked the reference lists of all the studies identified by the above methods and contacted relevant authors and researchers in the field.
We included studies that: i) compared the results of blood PCR tests with the reference standard published by the European Organisation for Research and Treatment of Cancer/Mycoses Study Group (EORTC/MSG); ii) reported data on false-positive, true-positive, false-negative and true-negative results of the diagnostic tests under investigation separately; and iii) evaluated the test(s) prospectively in cohorts of people from a relevant clinical population, defined as a group of individuals at high risk for invasive aspergillosis. Case-control studies were excluded from the analysis.
Authors independently assessed quality and extracted data. For PCR assays, we evaluated the requirement for either one or two consecutive samples to be positive for diagnostic accuracy. We investigated heterogeneity by subgroup analyses. We plotted estimates of sensitivity and specificity from each study in receiver operating characteristics (ROC) space and constructed forest plots for visual examination of variation in test accuracy. We performed meta-analyses using the bivariate model to produce summary estimates of sensitivity and specificity.
Eighteen primary studies, corresponding to 19 cohorts and 22 data sets, published between 2000 and 2013 were included in the meta-analyses, with a median prevalence of IA (proven or probable) of 12.0% (range 2.5 to 30.8 %). The majority of people had received chemotherapy for a haematological malignancy or had undergone a hematopoietic stem cell transplant. Several PCR techniques were used among the included studies. The sensitivity and specificity of PCR for the diagnosis of IA varied according to the interpretative criteria used to define a test as positive. The mean sensitivity and specificity were 80.5% (95% CI; 73.0 to 86.3) and 78.5% (67.8 to 86.4) for a single positive test result, and 58.0% (36.5 to 76.8) and 96.2% (89.6 to 98.6) for two consecutive positive test results.
PCR shows moderate diagnostic accuracy when used as screening tests for IA in high-risk patient groups. Importantly the sensitivity of the test confers a high negative predictive value (NPV) such that a negative test allows the diagnosis to be excluded. Consecutive positives show good specificity in diagnosis of IA and could be used to trigger radiological and other investigations or for pre-emptive therapy in the absence of specific radiological signs when the clinical suspicion of infection is high. When a single PCR positive test is used as diagnostic criterion for IA in a population of 100 people with a disease prevalence of 13.0% (overall mean prevalence), three people with IA would be missed (sensitivity 80.5%, 19.5% false negatives), and 19 people would be unnecessarily treated or referred for further tests (specificity of 78.5%, 21.5% false positives). If we use the two positive test requirement in a population with the same disease prevalence, it would mean that six IA people would be missed (sensitivity 58.0%, 42.1% false negatives) and three people would be unnecessarily treated or referred for further tests (specificity of 96.2%, 3.8% false positives). Galactomannan and PCR have good NPV for excluding disease but the low prevalence of disease limits the ability to rule in a diagnosis. The biomarkers are detecting different aspects of disease and the combination of both together is likely to be more useful.