|Home | About | Journals | Submit | Contact Us | Français|
We evaluated the ability of direct fluorescent antigen (DFA) influenza tests to identify novel H1N1 influenza. DFA results were compared to PCR. Negative predictive value of DFA testing was at least 96%. Therefore, when performed on specimens of adequate quality, DFA tests can effectively rule out infection with novel H1N1.
In the setting of the novel H1N1 influenza outbreak in the United States, a major barrier to effective control of the outbreak has been the lack of knowledge about the performance of commercially available influenza tests for detection of novel H1N1. Even for seasonal influenza, rapid antigen test formats like enzyme-linked immunosorbent assay EIA), optical immunoassay, or lateral flow have a wide range of sensitivity and specificity (24–95% and 69–100%, respectively(e.g. ), and results can be greatly influenced by specimen type, study population, and duration of symptoms prior to testing. Detection of influenza by direct fluorescent antigen (DFA) testing involves sedimentation of respiratory epithelial cells onto a slide well and subsequent staining with influenza-specific antibodies conjugated to fluorescent dye; the test takes 1–4 hours. The sensitivity of DFA tests for seasonal influenza is higher than that of point-of-care tests e.g. Binax, Directigen(EIA) ([2–5]) but requires technical expertise and a fluorescence microscope. Culture of influenza virus from a respiratory specimen has historically been the gold standard for diagnosis, but can take days to complete. PCR testing is highly sensitive (lower limit of detection 1–10 infectious units; ), but also requires technical expertise and expensive equipment. Also, as in the case of novel H1N1, PCR testing may require rapid production and validation of new primer and probe sets.
At our hospital, where we perform DFA testing as our rapid test of choice for influenza diagnosis, we knew that our test had sufficient sensitivity (≥95% compared to viral culture) to allow us to effectively rule out seasonal influenza, but were uncertain how the test would perform for detection of novel H1N1. Without knowledge of test sensitivity or negative predictive value for diagnosis of novel H1N1, it was at first impossible to interpret a negative test result in a patient with suspected “swine flu.” To assess the performance characteristics of DFA testing for novel H1N1, we therefore compared results of our hospital laboratory's DFA tests to results of PCR testing for novel H1N1 influenza performed by the Massachusetts (Hinton) State Laboratory Institute (HSLI).
Analyses include all specimens that (1) were collected between May 18 and May 29, 2009 and (2) were subjected to both DFA testing for influenza and PCR testing for novel H1N1. This date range was selected because before this time period, no specimens from our institution had tested positive for H1N1 by PCR, and after this time period, we no longer submitted DFA-negative specimens for PCR testing. All specimens which were DFA-positive for influenza A were forwarded to the HSLI for PCR testing (below). In accordance with testing guidelines set up by the HSLI, specimens which were DFA-negative for influenza were only forwarded for PCR if they had been collected from inpatients or health-care workers.
Specimens were collected only from symptomatic health care workers or patients meeting CDC criteria for influenza-like-illness, as part of their routine clinical evaluation. Flocked swabs (Copan Diagnostics) (n=109) or, rarely, nasopharyngeal (NP) aspirates (n=3) were used to collect respiratory epithelial cells from the posterior nasopharynx. Swabs (2 per patient) were immediately placed after collection in either phosphate buffered saline (PBS), for DFA, or M4-RT viral transport medium (Microtest), for PCR. Specimens were put immediately on ice after collection and stored at 4C until processed for testing. Specimens were almost exclusively collected by trained hospital respiratory therapists.
Two commercial DFA kits were used for influenza testing in the Beth Israel Deaconess Medical Center clinical laboratory: 1) Simulfluor influenza A/B (Chemicon/Millipore), and 2) D3 DuetTMDFA RSV/Respiratory Virus Screening Kit (RVP; Diagnostic Hybrids), which tests for influenza A/B, adenovirus, respiratory syncytial virus, and parainfluenza. The Simulfluor kit was used for influenza testing unless the RVP was specifically ordered by the clinician. The presence of ≥30 columnar epithelial cells per test well was required for the specimen to be considered adequate for DFA testing; results from inadequate specimens were not reported unless they were positive (which occurred extremely rarely). Inadequate NP swab specimens were infrequent (3% of total specimens during the time frame of the study); higher rates were observed with other specimen types (NP aspirates, bronchoalveolar lavage) which were infrequently submitted.
Confirmatory PCR testing for novel H1N1 was performed at the HSLI, using PCR reagents and protocols provided by the Centers for Disease Control and Prevention (CDC), Influenza Branch (www.who.int/csr/resources/publications/swineflu/CDCrealtimeRTPCRprotocol_20090428.pdf, ). A specimen was confirmed as “positive” for novel H1N1 if all three targets (panA, SwH1, and SwA) were positive and controls within the kit met the kit specifications. A specimen was confirmed as “negative” for novel H1N1 if all three targets were negative. A specimen that tested positive for only one or two of the three targets was reported as “inconclusive.”
Results for specimens tested by both DFA and PCR were compared. Between May 18 and May 29, 2009, 112 NP specimens (109 swabs and 3 aspirates) were collected from patients with a mean age of 44.1 years; only two patients were <18 years old (1 year and 13 years). 87 (78%) specimens were tested with the Simulfluor kit, 21 (19%) with the RVP kit, and 4 (3%) with both tests. In subset analyses the two assay kits performed comparably, and thus data for the two assays were combined. Data are summarized in Table I. Based on these data, the DFA was calculated to have a sensitivity, specificity, negative predictive value (NPV) and positive predictive value (PPV) of 93% (±8%), 97%(±4%), 96%(±5%), and 95% (±7%), respectively, relative to PCR (95% confidence intervals indicated in parentheses).
The 3 DFA-negative/PCR-positive specimens were reviewed in detail. All were NP swab specimens. In the first, there were <30 columnar epithelial cells in the slide well. However, neutrophils had been incorrectly interpreted as columnar epithelial cells. Therefore, the specimen should have been rejected based on the criteria for specimen adequacy in place at the time (Methods). The second also had high numbers of neutrophils and was similarly inadequate. The third also had high numbers of neutrophils which were misinterpreted as columnar epithelial cells, but there were between 30 and 60 columnar epithelial cells. This review demonstrated that a cutoff for specimen adequacy of ≥60 columnar epithelial cells per well (rather than 30) would have classified all three DFA-negative/PCR-positive specimens as inadequate for DFA testing, resulting in a sensitivity and negative predictive value that approached 100% (see below).
Effective control of any outbreak of a novel influenza strain will rely on the ability of available testing modalities to detect and, critically, to rule-out infection with the novel strain. During the recent epidemic of novel H1N1 disease in Boston, Massachusetts, we found that we were able to effectively rule out novel H1N1 infection in symptomatic adults using DFA testing--a rapid, relatively low-cost, and commercially-available method. Even prior to optimization of criteria for specimen adequacy (below), the NPV of our DFA tests was 96%.
Our specimens were predominantly nasopharyngeal swab specimens collected using flocked swabs (which contain brush-like nylon fibers to improve cell collection), and this collection method produced excellent yield of columnar epithelial cells. Any discrepant results (DFA-negative/PCR-positive) that we did see were explained by one or more of the following factors: 1) borderline insufficient numbers of respiratory columnar epithelial cells in the specimen, or 2) misinterpretation of neutrophils as columnar epithelial cells and therefore incorrect assessment of specimens as having adequate numbers of cells.
In response to these findings, we have now increased our laboratory's cell threshold (for considering a sample adequate for DFA testing) from ≥30 to ≥60 columnar epithelial cells per test well. Past laboratory data (internal data from 2007–8 influenza season, not shown) indicated that use of this more stringent criterion for specimen adequacy would not significantly increase the frequency of specimen rejections. Furthermore, we have emphasized with our technologists the need to reliably distinguish respiratory columnar epithelial cells (the cell type infected by influenza virus) from neutrophils and squamous epithelial cells in assessing specimen adequacy.
We expect that with these changes we will have very few false negative DFA results going forward. Importantly, immunofluorescent staining (e.g., DFA testing) is a standard method for identifying viruses in clinical virology laboratories. Therefore, clinical virology laboratories should be readily able to apply this prior experience to implementation of highly sensitive DFA testing for novel H1N1. It should be noted that our study examined test performance primarily for adult inpatients and healthcare workers (outpatients) who met CDC criteria for influenza-like illness. Therefore, our results may not similarly apply to other patient populations, e.g., pediatric patients or other specific outpatient groups, in whom the performance characteristics of the DFA test may differ. Furthermore, our NPV is specific to the disease prevalence associated with this period of the H1N1 pandemic; if the prevalence of H1N1 in adults increases in the future, the NPV of the DFA tests may decrease to some degree. Nevertheless, our results demonstrate that DFA testing can be used to effectively exclude infection with the novel H1N1 influenza strain in a hospital setting, allowing resources and infection control efforts to be focused on those who test positive.
We thank the members of the BIDMC microbiology laboratory and the Hinton State Laboratory Institute virology laboratory for their tireless efforts during the novel H1N1 outbreak.
There was no financial support for this study.
None of the authors of this manuscript have any conflicts of interest related to this study.