rapid diagnosis of influenza: an evaluation of two commercially available rt-pcr assays
TRANSCRIPT
Journal of Infection (2012) 65, 60e63
www.elsevierhealth.com/journals/jinf
Rapid diagnosis of influenza: An evaluation of twocommercially available RT-PCR assays
Mark Li a,b,*,c, Nigel Brenwald a,b, Sandra Bonigal a,b, Kulvir Chana a,b,Husam Osman a,b, Beryl Oppenheim a,b
aDepartment of Medical Microbiology, City Hospital, Birmingham, United KingdombHPA West Midlands Public Health Laboratory, Birmingham Heartlands Hospital, United Kingdom
Accepted 7 April 2012Available online 17 April 2012
KEYWORDSInfluenza;Commercial;PCR;Comparison
* Corresponding author. Department507 4078; fax: þ44 0121 507 5521.
E-mail addresses: [email protected] (K. Chana), [email protected]
c Birmingham HPA Laboratory, Birm0121 4242240; fax: þ44 0121 7726229
0163-4453/$36 ª 2012 The British Infdoi:10.1016/j.jinf.2012.04.003
Summary Objective: To evaluate the sensitivity and specificity of influenza virus detection bytwo commercial reverse transcriptase PCR methods compared with a reference real-time PCR.Methods: 122 clinical specimens were tested on Xpert� Flu and RealStar� Influenza Screen &Type. A reference real-time RT-PCR, at a specialist laboratory was chosen as the gold standardfor comparison.Results: RealStar� Influenza Screen & Type had higher sensitivity for influenza A and influenza Brespectively (92.3% and 88.2%) when compared to Xpert� Flu (78.8% and 76.5%). Both tests hadexcellent specificity.Conclusions: The simplicity and speed of the Xpert� Flu system could allow it to be used in thenear-patient setting; however in circumstances where excluding a diagnosis of influenza maybe critical, negative specimens may need to be repeated using a more sensitive assay.ª 2012 The British Infection Association. Published by Elsevier Ltd. All rights reserved.
Introduction
In 2009 a novel influenza virus A/H1N1v was identified inMexico and soon reached pandemic status.1 During the sea-sonal influenza period of 2010/11, influenza A/H1N1v and in-fluenza B were the predominant circulating viruses in
of Microbiology, City Hospital, Du
(M. Li), [email protected] (B. Osman), beryl.oppenheiingham Heartlands Hospital, Bor.
ection Association. Published by E
England and Wales.2 Several methods have been used to ob-viate the need for protracted viral cultures in influenza diag-nosis. These include antigen testing, immunofluorescenceand molecular testing. Recent reports have highlighted sen-sitivities of only 10e50% for rapid antigen-based tests.3 Im-munofluorescence requires highly trained laboratory staff
dley Road, Birmingham, B18 7QH, United Kingdom. Tel.: þ44 0121
(N. Brenwald), [email protected] (S. Bonigal), [email protected]@nhs.net (B. Oppenheim).desley Green East, Birmingham B9 5SS, United Kingdom. Tel.: þ44
lsevier Ltd. All rights reserved.
Evaluation of two commercial RT-PCR assays 61
thatmay be overwhelmed by the increase in workload duringan epidemic. Molecular diagnosis of influenza is attractive,offering timely and accurate results which can aid clinicalmanagement and infection control.4 Standardised commer-cial assays would allow in-house testing without the require-ment for referral of specimens to specialist centres, thusspeeding up the availability of results.
Materials
Recent studies have separately evaluated the performanceof Xpert� Flu (Cepheid, Sunnyvale, CA, USA) and RealStar�
Influenza Screen & Type (Alere, Hamburg, Germany).Xpert� Flu is a simple test that runs on GeneXpert� system(Cepheid, Sunnyvale, CA, USA). It simultaneously detectsand differentiates influenza A, influenza B, and influenzaA/H1N1v (A/H1N1v) viruses in about 80 min. It involvesa cartridge-based PCR system for performing nucleic acidextraction, PCR amplification, and real-time detection. Itautomatically sets the temperatures and number of cycleswithout intermediate sample-handling steps. This avoidsthe need to transport specimens onto a new plate on an-other platform. Therefore sampling errors can be reducedand it may be suitable for point of care testing (POCT) bynon-laboratory trained healthcare staff. In addition, all in-ternal controls for Xpert�Flu are included within the car-tridge system.
RealStar� Influenza Screen & Type is a batch based assaythat can be run on a range of PCR platforms. It would not besuitable for near-patient testing as it requires a number ofspecimen handling steps which would need to be performedby a trained biomedical scientist. The entire process fora complete set of results takes 4 h. Both these assayshave shown specificity of up to 100% but with variable sen-sitivity in recent validation studies.5e8 However none ofthese studies directly compared the two commercial as-says. We therefore sought to compare these two commer-cially available RT-PCR assays, using as a reference real-time RT-PCR (reference method) offered by a specialist lab-oratory and widely used for the diagnosis of influenza in theUnited Kingdom.
Method
All specimens sent to our laboratory for influenza detectionbetween 13th and 27th December 2010 were included in thestudy. Only 122 of the 150 specimens had sufficient samplefor the evaluation as each had to be tested on both thecommercial assays and the reference method. The 122specimens comprised 97 nasopharyngeal swabs, nine naso-pharyngeal aspirates, nine endotracheal aspirates, fournasal swabs and three unknown specimen type. Specimenswere originally tested on receipt using the Xpert� Flu systemand were then stored and frozen in MicroTest� M4RT� (Re-mel, Lenexa, KS, USA) at �20 �C. These specimens werethen thawed and subsequently used for further testing inthe other two assays. 71 specimens were from female pa-tients and 51 frommales. Patients’ ages ranged from1monthto 91 years with a median age of 34 (mean 37).
RealStar� Influenza Screen & Type was run on an ABIPrism� 7500 PCR System (Applied Biosystems, Cheshire,
UK). RNA was extracted using the Arrow NA (NorDiag,Oslo, Norway) prior to amplification. Specimens were bar-code labelled and tested in batches; each batch had a neg-ative control and positive controls for influenza. Each testwas performed according to manufacturer’s instructions,by one of two biomedical scientists blinded to previous re-sults. The following thermal profile was used: 50 �C for10 min, 95 �C for 2 min followed by 45 cycles at 95 �C for15 s, 55 �C for 45 s and 72 �C for 15 s.
The reference method’s total nucleic acid extractionwas performed using the MDx automated biorobot (Qiagen,D€usseldorf, Germany). The RT-PCR was performed on ABI7500 fast real-time PCR (Applied Biosystems, Cheshire, UK)using a multiplex format. The one-step RT-PCR thermalprofile was as follows: 50 �C for 15 min, 95 �C for 2 min,95 �C for 15 s for 45 cycles, followed by one cycle at 60 �Cfor 40 s. Internal controls were also included throughoutthe process to control for extraction and PCR. Barcodeidentification was included throughout the referencemethod process to reduce the chance of incorrect specimentesting.
Results
For diagnosis of influenza A, the reference method identi-fied 52 cases, of which Xpert� Flu detected 41(78.8%).There were no cases detected by Xpert� Flu that werenot positive in the reference method. RealStar� InfluenzaScreen & Type detected 51 positives of which only 48were identified by the reference method (92.3%). Fig. 1shows the concordance of the different tests for influenzaA for illustrative purposes. The reference method con-firmed all 52 influenza A positive cases as influenza A/H1N1v on a separate A/H1N1 specific PCR, while Xpert�
Flu confirmed 39 of its 41 positive cases. RealStar� Influ-enza Screen & Type identified 51 cases, one of which failedto confirm as influenza A/H1N1v and was negative by thereference method.
For influenza B, the reference method identified 17cases, of which Xpert� Flu detected 13 (76.5%). RealStar�
Influenza Screen & Type detected 15 cases all of whichwere confirmed by the reference method. Table 1 showsthe sensitivity, specificity, positive and negative predictivevalues for the two commercial assays using the referencemethod for comparison.
Discussion
When evaluating molecular tests for the diagnosis ofinfectious disease it is difficult to choose a gold standardwith which to compare results. In this case we chose thereference test used by the Health Protection AgencyLaboratories which was developed in June 2009.9 Usingthis as a standard we found that both RealStar� InfluenzaScreen & Type and Xpert� Flu had reduced sensitivity in de-tecting A/H1N1v and influenza B when compared to the ref-erence method. Sensitivity was greater for RealStar�
Influenza Screen & Type than Xpert� Flu in detecting influ-enza A (92.3% cf 78.8%) and influenza B (88.2% cf 76.5%).
Another major difficulty with this type of study is theproblem of lack of simultaneous testing. However this is
Figure 1 Venn diagram showing number and concordance ofinfluenza A detected by each assay. Venn diagram showingnumber and concordance of influenza A detected by RealStar�
Influenza Screen & Type (RealStar�), Xpert� Flu and the refer-ence method.
62 M. Li et al.
unsuitable in situation where specimens are submitted forurgent testing. We are aware of other studies comparingassays during the H1N1 influenza pandemic/epidemicwhere simultaneous testing was also not possible.10,11
Our study confirms the high specificity of Xpert� Flu forinfluenza A found in other studies,5e7 however sensitivitywas found to be lower (78.8% cf 91.2e93%). AlthoughXpert� Flu automatically interprets the result of the assayit is possible that late amplification of the targets could oc-cur but this would be interpreted as negative by the GeneX-pert system. We checked the assay graphs of those testsshowing discrepant results to the reference laboratory.The graphs of the two specimens which had tested influ-enza B positive by the reference laboratory but negativeby Xpert� Flu were found to show late amplification ofthe influenza B target. A further two Xpert� Flu specimens,which were influenza A positive, A/H1N1v negative, but ref-erence laboratory A/H1N1v positive, also showed late
Table 1 Characteristic of Xpert� Flu and RealStar� Influenza Sca reference method assay.
Xpert� Flu
Influenza A A/H1N1v Influenza B
Sensitivity %(95% CI)
78.8 (64.9e88.4) 75 (60.8e85.5) 76.5 (49.8e
Specificity %(95% CI)
100 (93.5e100) 100 (93.5e100) 100 (95.6e
PPV % (95% CI) 100 (89.3e100) 100 (88.8e100) 100 (71.7eNPV % (95% CI) 86.4 (76.6e92.7) 84.3 (74.3e91.1) 96.3 (90.3e
PPV Z positive predictive value, NPV Z negative predictive value, C
amplification of the H1N1v target. RealStar� InfluenzaScreen & Type showed higher sensitivity for influenza Athan Xpert� Flu but had lower sensitivity (92.3% cf 100%)and specificity (95.7% cf 100%) than was found in anothervalidation study.8
A total of four specimens were negative on bothcommercial assays but the reference method found threeof these to be influenza A positive and the other influenza Bpositive. All four specimens were oropharyngeal swabscollected from patients ranging from 16 to 50 years ofage. Three of the four were collected outside of laboratoryworking hours, therefore were tested the day after collec-tion which may have been relevant to the result.
The reduced sensitivity of the RealStar� InfluenzaScreen & Type and Xpert� Flu compared to the referencemethod may be a reflection of their sensitivity of detectionin specimens containing low virus quantities. Looking at in-fluenza A detection, the majority of specimens missed byboth commercial assays had high Ct (Cycle Threshold)values by the reference assay which is an indirect measureof viral load in the specimens (Fig. 2).
Our results show that Xpert� Flu is highly specific, withall positives in all components of the test being confirmedand suggesting that positive results from this assay do notrequire further confirmation. RealStar� Influenza Screen& Type showed specificities of 95.7% and 97.1% for influenzaA and A/H1N1v respectively and high specificity of 100% forinfluenza B (Table 1).
In conclusion, we have evaluated two commercial assaysfor the detection of influenza, one of which, the Xpert�
Flu, is particularly simple, using a cartridge-based systemwhich reduces sampling errors by performing extractionand amplification in the same cartridge. The sensitivity be-ing around 15% lower, in absolute terms, than in other pub-lished evaluations,5e7 supports the approach of usinga more sensitive reference test to confirm negative results,especially in critically ill patients. The high specificity ofXpert� Flu could allow testing of large numbers of patientswhen isolation rooms are under pressure during an epi-demic. This would allow patients who are shown to be pos-itive to be cohort nursed. Negative cases could be nursed inside rooms and await a more sensitive test. This makes theassay potentially suitable for local or POCT. However issuessuch as interpretation of specimens with low levels of RNAand further research with a non PCR gold standard need tobe overcome before implementation of POCT. Although
reen & Type for influenza A, A/H1N1 and B in comparison to
RealStar� influenza Screen & type
Influenza A A/H1N1v Influenza B
92.2) 92.3 (80.6e97.5) 92.3 (80.6e97.5) 88.2 (62.3e97.9)
100) 95.7 (87.2e98.9) 97.1 (89.1e99.5) 100 (95.6e100)
100) 94.1 (82.8e98.5) 96 (85.1e99.3) 100 (74.7e100)98.8) 94.4 (85.5e98.2) 94.4 (85.7e98.2) 98.1 (92.8e99.7)
I Z confidence interval.
Figure 2 Number of specimens positive for influenza A, detected by the three assays at different reference laboratory assay Ctvalue ranges. Number of specimens positive for influenza A, detected by the RealStar� Influenza Screen & Type (RealStar�), Xpert�
Flu and the reference method assays at different reference laboratory assay Ct ranges.
Evaluation of two commercial RT-PCR assays 63
both tests provided an acceptable degree of specificity, thesomewhat lower sensitivity means that they may be unsuit-able for use as standalone tests to exclude influenza in pa-tients where an accurate diagnosis is critical.
Appendix A. Supplementary data
Supplementary data related to this article can be foundonline at doi:10.1016/j.jinf.2012.04.003.
References
1. Anonymous. Emergence of a novel swine-origin influenza A(H1N1) virus in humans. N Engl J Med 2009;360:2605e15.
2. Health Protection Agency UK. Surveillance of influenza andother respiratory viruses in the UK. report, http://www.hpa.org.uk/web/HPAwebFile/HPAwebC/1296687414154;2010e2011.
3. Drexler JF, Helmer A, Kirberg H, Reber U, Panning M,M€uller M, et al. Poor clinical sensitivity of rapid antigentest for influenza A 2009 (H1N1) 2009 virus. Emerg InfectDis 2009;15:1662e4.
4. Takahashi H, Otsuka Y, Patterson BK. Diagnostic tests for influ-enza and other respiratory viruses: determining performancespecifications based on clinical setting. J Infect Chemother2010 Jun;16(3):155e61 [Epub 2010 Feb 27].
5. Popowitch EB, Rogers E, Miller MB. Retrospective and prospec-tive verification of the cepheid xpert influenza virus assay. JClin Microbiol 2011 Sep;49(9):3368e9.
6. Jenny SL, Hu Y, Overduin P, Meijer A. Evaluation of the XpertFlu A Panel nucleic acid amplification-based point-of-caretest for influenza A virus detection and pandemic H1 subtyping.J Clin Virol 2010 Oct;49(2):85e9.
7. Sambol AR, Iwen PC, Pieretti M, Basu S, Levi MH, Gilonske KD,et al. Validation of the Cepheid Xpert Flu A real time RT-PCRdetection panel for emergency use authorization. J Clin Virol2010 Aug;48(4):234e8.
8. Panning M, Baumgarte S, Laue T, Bierbaum S, Raith S,Drexler JF, et al. Singleplex real-time RT-PCR for detectionof influenza A virus and simultaneous differentiation ofA/H1N1v and evaluation of the RealStar influenza kit. J Clin Vi-rol 2011 Feb;50(2):171e4.
9. Health Protection Agency UK. One-step influenza multiplexreal-time RT PCR. National Standard Methods VSOP 50 June2009.
10. Lalle E, Bordi L, Meschi S, Selleri M, Valli MB, Patella E, et al.Comparison of two real-time RT-PCR-based systems for the de-tection and typing of the pandemic influenza A virus, 2009. JVirol Methods 2011 Nov;177(2):193e6.
11. Miller S, Moayeri M, Wright C, Castro L, Pandori M. Comparisonof GeneXpert FluA PCR to Direct Fluorescent Antibody and re-spiratory viral panel PCR assays for detection of 2009 novelH1N1 influenza virus. J Clin Microbiol 2010 Dec;48(12):4684e5.