The accuracy of Influenza A (H1N1) “swine flu” laboratory testing: A systematic review of diagnostic test accuracy.

Abstract

Review question/objective The aim of this systematic review is to comprehensively search the available literature and to synthesise the best available evidence to determine the diagnostic accuracy of currently available laboratory tests for swine flu (H1N1), using viral culture as a reference test. Background The Influenza virus affects mainly the nose, throat, bronchi and, occasionally the lungs. Infection usually lasts for about a week and is characterised by sudden onset of high fever, aching muscles, headache and severe malaise, non-productive cough, sore throat and rhinitis. 1 The virus is transmitted easily from person to person via aerosol droplets and small particles produced when infected people cough or sneeze. Influenza tends to spread rapidly in seasonal epidemics. Most infected people recover within one to two weeks without requiring medical treatment. However, in the very young, the elderly, and those with other serious medical conditions, infection can lead to severe complications of the underlying condition, pneumonia and death. 1 In their simplest form, viruses are composed of an outer protein coat and an inner core of either deoxyribonucleic acid (DNA) or ribonucleic acid (RNA), however, unlike bacteria, no virus has been shown to contain both 2 indicating they are unable to transcribe DNA to RNA, thus requiring host cells. A virus particle is present as either in either a replicating or defective form within the infected cell. The extracellular virus is defined as a virion. Influenza viruses belong to the family of Orthmyxoviridae. This family of viruses have a single stranded, negative sense, eight-segmented RNA genome.3 The eight segments of RNA encode 11 proteins, of which nine are assembled into the infectious virion.4 Three proteins are found on the surface of the virion, haemagglutinin (HA) and neuraminidase (NA), as well as the transmembrane ion channel protein (M2). This family of viruses are enveloped by a lipid envelope that is derived from proteins synthesised by the host. The lipid envelope confers some protection to the virus, not only from host immune cells but also from environmental factors. Influenza A, B and C viruses differ in the host species used and their pathogenicity 5 and can be distinguished on the basis of antigenic differences between proteins on their outer protein coat.6 Of the influenza viruses, only A and B cause frequent (and occasionally severe) diseases in humans 3 There is only one type of Influenza B, whereas Influenza A has multiple subtypes, based on a combination of genes that encode for surface proteins, 3 discussed further below. A single extracellular Influenza A virus particle (virion) has a characteristic spiky appearance when viewed under electron microscopy. 7 The spikes are due to glycoproteins that protrude beneath the lipid envelope. The major glycoproteins are Haemagglutinin and Neuraminidase. A minor component of the lipid envelope is the M2 protein which acts as an ion channel. The matrix protein (M1) lies beneath the lipid envelope and covers the ribonucleoprotein (RNP) complexes. The RNP complexes are comprised of the viral RNA, covered with nucleoprotein and associated with other protein complexes. The viral genome consists of eight separate RNA segments and the segments 1, 2 and 3 code for viral polymerase i.e., PB2, PB1, and PA respectively, segment 4 for haemagglutinin (HA), segment 5 for Nucleoprotein (NP), segment 6 for neuraminidase (NA), segment 7 for matrix protein (M1), and membrane channel proteins (M2), segment 8 for non-structural proteins (NS1 and NS2). 8 The sketch in Figure 1 is a sketch showing the main features of a typical influenza A virus.Figure 1 Schematic influenza A virion: 8The category of Influenza A viruses is further subdivided on the basis of haemagglutinin (H) and neuraminidase (N) proteins expressed on the outer surface of the virus. 5 Both of these proteins are important for the virus to be able to get into the host cell, reproduce and be released to infect other cells - making them potential targets for both diagnosis and therapy. Haemagglutinin allows the virus to enter host cells, such as epithelial cells lining the respiratory tract. This molecule mediates viral attachment to the host cell and its subsequent entry into cells by causing fusion of the cell membranes (endocytosis). Variation in haemagglutinin molecules can result in altered affinity for the different types of the cellular receptor (sialic acid) for binding either the human or the avian variant forms. Because of this, receptor specificity of the haemagglutinin molecule is thought to be able to determine the transmissibility of an influenza virus strain in a given species. 3 Once the virus is inside the host cell (such as epithelial cells), it uses the cell to synthesise copies of itself. Neuraminidase hydrolyses the bonds between haemagglutinin and sialic acid after the virus has reproduced, allowing it to leave the cell to continue infecting other cells. Neuraminidase activity is required for the release of the viral particles/virions and the activity of neuraminidase is the target of the current antiviral drugs, Oseltamivir and Zanamivir. 5 Treatment prevents the action of neuraminidase, resulting in a failure of the release the new virions being released and therefore prevents further spread of the virus. Influenza A viruses are given the denotation HxNy, where x can be 1-16, as there 16 known subtypes of haemagglutinin and y can be 1-9 as there are 9 known neuraminidase subunits, however only H 1, 2, 3 and N 1 and 2 are commonly found in humans. 5 The 2009 pandemic influenza H1N1 consists of genetic material from viruses from swine, human and avian origins. 9 Globally, at any one time there are likely to be several influenza viruses circulating in seasonal patterns. Influenza viruses gradually change over time by a process called antigenic drift. This process is characterised by sporadic point mutations in either the haemagglutinin or neuraminidase protein. 6 Occasionally, a major antigenic change in either the haemagglutinin or neuraminidase protein occurs by a process called antigenic shift, resulting in a virus that mutates much faster and with the ability to spread much more widely, quickly and replace the prevailing seasonal influenza strain e.g. the 2009 pandemic influenza H1N1. This may result in a new strain of virus. Antigenic shift occurs under rare conditions where a host is infected with a combination of two or more influenza A viruses of different origins or strains that are able to become mixed (or ressorted), leading to the generation of a new viral strain. The resulting virus has the ability to spread rapidly within a population (and also to other populations and subsequently globally), as people are likely to be immunologically naive to this new virus. The 2009 H1N1 virus was a triple ressortment influenza virus with swine human and avian genetic material that rapidly became pandemic. 3, 9, 10 However, although the 2009 H1N1 strain was novel, it shared features with other H1N1 strains which aided in its identification. 11 Pandemic influenza viruses are thought to arise when there is close and frequent contact between humans and other animal species that can be infected by Influenza viruses. The virus develops the ability to jump the species barrier to be able to infect humans. This “crossing” is made possible by certain genetic mutations that permit the binding of the animal viruses to surface proteins in the human respiratory tract. 3 Prior to the 2009 H1N1 pandemic, there have been three well-documented Influenza A pandemics in recent history. 6, 12 Table 1 summaries some of the major details of recent Influenza A pandemics, with the aim of framing the 2009 H1N1 pandemic in a historical context. 4, 5, 12Table 1: Recent Influenza A pandemicsTable: No Caption available.The first of these pandemics was caused by a H1N1 virus and coincided with the outbreak of World War I. It is estimated that roughly a third of the world's population was infected (approximately 500 million people) and resulted in approximately 50 million deaths. 3 The 1957 and 1968 pandemics resulted in a lower number of cases and fatalities; however, they still resulted in a significant global effect. The 2009 pandemic was caused by a H1N1 virus and although the number of fatalities was much lower than any of the previous pandemics, concern was raised by how quickly the virus spread. The outbreak began in the state of Veracruz, Mexico in March/April 2009 and spread rapidly across the world. Only a matter of weeks later (in June 2009), the World Health Organization (WHO) and US Centers for Disease Control (CDC) stopped counting cases (30,000 confirmed cases) and the WHO declared the outbreak a pandemic. 13 Subsequent estimates have reported millions of cases and at least 16,813 deaths. 5 Due to the availability of epidemiological records and preserved archival material, the viruses responsible for the recent pandemics have been identified, allowing patterns of infectivity to be modelled. 11 The haemagglutinin gene of 2009 H1N1 is derived from “classical swine H1N1” virus, which likely shares a common ancestor with the human H1N1 virus that caused the pandemic in 1918. By using homology modelling, Igarashi 11showed that the two viral strains exhibit such a high degree of structural homology that prior exposure to the 1918 strain is likely to confer specific immunity against H1N1. Techniques that model the structure and homology of haemagglutinin and neuraminidase proteins are useful in identifying potential therapeutic targets for drug and/or vaccination development, 14, 15 as well as potential targets for diagnostic tests. H1N1 was first described in the 1918 pandemic and made a resurgence in April 2009 in the form of a triple-reassortant influenza A virus, which is composed of a combination of human, swine, and Eurasian avian strains. 16What was so unusual about this Influenza A strain was that it appeared to be comprised of several separate types of known flu strain and was described as “..an unusually mongrelised mix of genetic sequences” 10 In June 2009, the World Health Organisation declared an H1N1 pandemic - the first global pandemic since the 1968 Hong Kong flu. Infections with influenza H1N1 viruses are usually less severe in their impact than those with Influenza B viruses; however they appear to have more pandemic potential, possibly as a result of the type of haemagglutinin protein present. 3 Currently, there is no known cure for the influenza virus, therefore strategies have focussed on prevention (such as the use of vaccination programs) and reduction of symptoms (such as with antiviral drugs such as Oseltamivir and Zanamivir). The lack of a cure relates to the success of the influenza virus to adapt in order to evade the immune system. Distinguishing an influenza virus from other circulating pathogens on the basis of signs and symptoms alone is unreliable due to the large degree of similarity in presentation. For example, Mycoplasma pneumonia - causative agent of primary atypical pneumonia (also known as pleuro-pneumonia like organism or PPLO) causes similar signs and symptoms of those of respiratory tract infection. Laboratory testing of respiratory specimens offers a much improved method of identifying causative agents and therefore improving effectiveness of management. Diagnostic tests Diagnostic tests aid clinicians in making a diagnosis and a subsequent treatment plan for the patient based on the test results.17 Diagnostic test interpretation involves comparing features of the patient against known features of the suspected condition and generating a likelihood of the presence or absence of the condition. In order to determine how accurate a diagnostic test is, the test should be compared to a reference test/standard that has been proven as being reliable for the particular condition of interest and patient population. Systematic reviews of diagnostic tests are a recent development and aim to inform the clinician by critically appraising and synthesising the best available evidence concerning the accuracy of diagnostic tests, ideally in comparison with alternate techniques. 18 However, this type of review can be challenging as determining what is meant by accuracy in this context is more complex than with other study designs, such as studies of therapeutic interventions. 19 The accuracy of a test is determined by posing two questions: How often does the test give positive results for patients that have the condition? (sensitivity) and How often does the test give negative results for patients that do not have the condition? (specificity). As with any procedure, there are benefits and limitations. One of the major benefits is that (generally) diagnostic tests are relatively straightforward to conduct and provide the clinician with evidence on which to base a diagnosis within a useful timeframe. When conducting a diagnostic test, a patient sample (e.g. blood, urine, scan etc.) is compared against a reference or test standard, therefore test accuracy can be compromised if the wrong type of patient sample or reference test/material is used. Diagnostic testing for Swine flu (H1N1) Given that there are often different strains of influenza virus circulating at any one time, as well as other organisms eliciting very similar symptoms, identifying a viral strain responsible for an outbreak can be difficult. Often the diagnosis of influenza is based on presenting symptoms 20 instead of the more accurate laboratory testing. 13 One reason for this could be the amount of time taken to obtain laboratory test results. The reference or definitive test to identify a virus is viral culture which can take up to 10 days - by which time the patient may have potentially infected many others. Considering the importance of a making correct diagnosis and the time taken to make this diagnosis, there is a need for a diagnostic test that can be used to correctly identify influenza in a clinically relevant timeframe of hours and not days. Therefore, diagnostic tests that can accurately identify a causative agent in clinically relevant/useful timeframe are extremely useful in clinical practice. In laboratory testing for influenza, respiratory tract samples (e.g. fluid, swabs or sputum) are collected from patients in order to characterise the agent causing the infection. Definitive diagnosis of swine flu H1N1 largely requires a test that is able to distinguish between influenza A subtypes. 16 Currently there are several methods available, each of which relies on particular features of the virus. Broadly, the types of test available to influenza viruses are: Infectivity based assays Protein binding based assays Nucleic acid based assays Infectivity Assays Viruses require host cells to infect and to be able to replicate. Once a cell has been infected with a virus, the cell responds is a predictable way. Infectivity assays are in vitro systems that aim to identify a virus on the basis of the cell type it infects and characteristics of the infection, such as changes in the cell and how the infection progresses. The major infectivity assay used in hospital laboratories in the diagnosis of viruses is viral culture. Viral culture is often considered the most sensitive method for diagnosis of type A influenza virus infection is isolation of the virus in cell culture. 21-23 and is traditionally used as the reference standard for comparison test. 24 Viruses require host cells to survive and reproduce, therefore cell lines are inoculated with a sample from the patient and grown in culture medium. Cells infected with the virus are visually identified under standard microscopy. Growth, detection and identification may require several days (often 3 - 10 days) to complete. Inevitably, this results in a delay in diagnosis, administration of antiviral agents and increases the risk of the virus spreading. Major disadvantages with this method include the requirement for specialised laboratory conditions, considerable technical expertise and is time consuming. Furthermore, only viable viruses are able to be detected in such assays, in contrast to techniques that establish the presence of a virus based on particular proteins or nucleic acids. This could potentially result in lower levels of detection of a virus and therefore potentially a false negative result for a patient. There is a need, therefore for a method accuracy that is comparable to viral culture but is much less time, resource and skill intensive. For the purposes of this review and due to its high level of accuracy and reliability, viral culture will be used as the reference test, to which more recently developed tests will be compared to. Protein binding based assays The aim of a protein binding assay is to identify a virus by picking out specific proteins (antigens) that are likely to be expressed by that virus alone. One such technique is immunofluorescence whereby a labelled antibody raised against all or part of the viral protein is used. An antibody can be labelled in several ways, either directly with a fluorescent reporter dye or indirectly - first with an intermediate antibody, which in turn is then detected with a labelled antibody - depending on laboratory requirements and preferences. The amount of the labelled antibody is used as a measure of the amount of viral protein present in the patient specimen, but is unable to give an indication of whether or not the virus is viable and therefore able to cause infection. Antigen detection using immunofluorescence techniques was pioneered in the 1970s, and commercial reagents are now widely used for the detection of influenza viruses. 25These assays require less technical expertise than viral culture techniques and have the advantage of allowing direct evaluation of specimen quality. This technique aims to identify the virus by exploiting antigenic properties of the proteins it expresses on its outer protein coat. Antibodies raised against virus-specific protein sequences are raised in an animal, then are fluorescently labelled (or tagged). The patient sample is incubated with the antibodies, the antibodies bind to the proteins on the virus and the protein of interest becomes visible under a fluorescent microscope. Synthesis of specific antibodies can take a substantial amount of time, depending on the animal species. Immunofluorescence results are generally available within hours but this type of technique can be less sensitive than viral culture as it relies on the specificity of the detection antibodies. Rapid diagnostic tests for viral proteins Rapid diagnostic test (RDT) kits signal the presence of a protein of interest with a chromatic change of membrane bound reagents. Such kits have become increasingly more commercially available and claim to identify viral proteins with a good degree of accuracy. 21, 23 These diagnostic test kits can be produced as dipsticks, cassettes, or cards and they contain internal positive and negative controls. Many of these tests can be automated and can be carried out in a matter of minutes (5 - 40). 25 The sensitivity and/or specificity may be greatly reduced and is thought to vary with virus type or subtype, timing of specimen collection, specimen type, patient age, and the test comparator. This reduced accuracy (compared to viral culture) may, however, still be useful in situations where a result is needed urgently - such as a pandemic situation or in resource poor situations. With the emergence of the pandemic influenza A (H1N1) 2009 virus, RDTs have been widely used for patient triaging, although there are limited data available on their clinical accuracy. 25 To correctly interpret results of diagnostic tests such as RDTs that have relatively low sensitivity and/or specificity, the prevalence of influenza in a community must be considered. 26 During peak disease activity, positive predictive values are highest, but false-negative results more likely. The opposite is true during times of low disease activity. 13When the disease prevalence is low or unknown, RDT results become difficult to interpret and of limited use. 13 Nucleic acid based assays Nucleic acid techniques aim to identify a virus on the basis of short, specific sequences of genetic material (nucleic acids). For influenza A viruses, targets for nucleic acid based assays are conserved regions of the haemagglutinin, neuraminidase and matrix genes. Polymerase chain reaction (PCR) is one such group of techniques. Nucleic acid techniques possess several advantages over infectivity-or antigen-based techniques for the detection of influenza. These tests have improved sensitivity for detecting organisms that are fastidious, no longer viable, or only present in small amounts. 25They are also able to provide rapid genetic information regarding sequence evolution, geographic variation, or the presence of virulence factors or antibiotic resistance. 25Their rapid turnaround times (relative to viral culture) allow them a more prominent role in patient management, and the ability to be able to test for multiple pathogens simultaneously has aided in the diagnosis of nonspecific respiratory syndromes, such as in outbreak settings. 25 PCR - polymerase chain reaction techniques PCR aims to reveal the presence of viral genetic material using small fragments of complimentary sequences (primers). A small sequence of DNA is chosen to be specific to a gene of interest (computer software is available to help identify target sequences), so that when it binds to the viral DNA it can be amplified and identified. There are three main types of PCR: traditional, real time and multiplex. Regardless of the type of PCR used, all PCR assays require good primer design taking into consideration gene target, gene number, mobility of genes between species, stability of gene, and the presence of mutations. 25 In addition to optimal primer sequences, PCR requires high purity of target nucleic acid fragments in order to have minimal interference with this test. Traditional PCR PCR utilises a pair of primers, which are complementary to a defined sequence on each of the two strands of the cDNA. The cDNA is used as this is the DNA sequence that would be directly translated into protein after transcription. The primers are then extended by a DNA polymerase and a copy of the strand is made after each cycle, leading to exponential amplification. The exponential amplification via reverse transcription polymerase chain reaction provides for a highly sensitive technique in which a very low copy number of RNA molecules can be detected. RT-PCR is widely used in the diagnosis of genetic diseases and, semi-quantitatively, in the determination of the abundance of specific different RNA molecules within a cell or tissue as a measure of gene expression. 27Reverse transcription polymerase chain reaction (RT-PCR) is a variation that has an additional step of generating (by reverse transcribing) the DNA sequence prior to its amplification. RT-PCR is commonly used in studying the genomes of viruses whose genomes are composed of RNA, such as Influenza virus A and retroviruses like HIV. 27RT-PCR includes three major steps. The first step is reverse transcription (RT), in which RNA is reverse transcribed to complementary DNA, or cDNA, using reverse transcriptase. This step is very important in order to perform PCR since DNA polymerase can act only on DNA templates. The RT step can be performed either in the same tube with PCR (one-step PCR) or in a separate one (two-step PCR). The next step involves separation of the double stranded DNA (dsDNA), so that the primers can bind to the separate strands. This is done by using high temperature (such as 95°C) so that the bonds between the DNA strands denature. Then, the temperature is decreased until it reaches the annealing temperature which can vary depending on the set of primers used, their concentration, the probe and its concentration (if used), and the cation concentration. 27The final step of PCR amplification is DNA extension from the primers. The length of the incubation at each temperature, the temperature alterations, and the number of cycles are controlled by a programmable thermal cycler. In conventional PCR, the products are detected using agarose gel electrophoresis and ethidium bromide (or other dyes that bind to nucleic acids). Real time PCR can also be used to visualise products of PCR using a fluorescent reporter dye as the reaction occurs and the fluorescence increases as the amplification progresses and the instrument. Real time PCR Real time PCR (rt RT PCR) is another PCR technique and allows visualisation of the RNA of interest as the binding occurs. Fluorescently labelled primers can be detected using a spectrophotometer and fluorescent signal intensity increases proportionally as the number of RNA copies is amplified. Real time RT PCR has several features that are considered improvement over the traditional agarose gels. Firstly, the two steps of amplification and detection are combined in one reaction, increasing the speed and efficiency of testing and reducing the risks of operator error and cross-contamination of samples. Secondly, as the fluorescent signal is proportional to the amount of RNA amplification, this technique is semi-quantitative and can allow estimates of the amount of starting nucleic acid material. Reliability of PCR depends upon the specificity of the primer sequence and that the labelled RNA probes only bind to regions of viral RNA specific to that individual virus strain. As with protein based assays, nucleic acid assays are unable to distinguish between viable and non-viable viruses within a sample. 27 Multiplex PCR Multiplex PCR systems are used to search for multiple PCR targets at the same time in one reaction. This has the advantage of increasing the number of pathogens tested for, without increasing the amount of reagents, technician time or specimen material required. 25Multiplex assays have broadened the scope of respiratory surveillance studies, and have also led to the increasing recognition of dual or triple infections in the same individual. 25 There are some technical considerations when attempting to identify multiple PCR products in the same reaction tube. In the amplification step, all multiplex platforms must balance the competing optimal PCR conditions for each individual target, and must overcome problems of competition and inhibition among the various primers and probes. The detection step is also complex, as the multiple PCR products need to be distinguished from one another. This can be done on several bases, such as the use of different colour reporter dyes, or by differences in their size (molecular weight), using resolution techniques such as agarose gel electrophoresis to differentiate by weight, and capillary-based auto-sequencers that identify targets by length and sequence. 25 A detailed search of all the major databases, as well as The Joanna Briggs Library of Systematic Reviews and the Cochrane Collaboration library was conducted to establish that there were no existing or underway systematic reviews on this topic. Aims of the review The aim of this systematic review is to comprehensively search the available literature and to synthesise the best available evidence to determine the diagnostic accuracy of currently available laboratory tests for swine flu (H1N1), using viral culture as a reference test. Inclusion Criteria Types of participants Studies will be considered for inclusion in the review if the participants are human patients exhibiting influenza-like symptoms who have been tested for swine flu (H1N1) using both a diagnostic and reference test. There will be no exclusion based on age, gender or co-morbidities, but subgroup analysis will be conducted if there is sufficient data and if appropriate. Phenomenon of interest The phenomenon of interest is the correct identification of H1N1 infection. This review will consider studies that compare the accuracy of laboratory tests aiming to diagnose swine flu (H1N1) in patients presenting with influenza-like symptoms, in comparison to viral culture. Outcome measures The accuracy of diagnostic tests will be determined by how well the test correctly gives positive results for patients with confirmed swine flu (H1N1) and negative results for patients without, as determined by viral culture. It is anticipated that data will be presented as raw test results or sensitivities and specificities of the test compared with viral culture will be presented. Other statistics to elaborate on diagnostic test accuracy may include likelihood ratios and/or predictive values. Types of studies This review will consider any quantitative study that examines the diagnostic accuracy of laboratory tests utilised for swine flu (H1N1) where the participant undertakes both the index and reference test. Exclusion criteria Studies that utilise non-human participants, focus on the analytical sensitivity of a test, do not distinguish between influenza A subtypes, or do not use viral culture as the reference test will be excluded. Search strategy The search strategy aims to find both published and unpublished literature. Initial search terms and databases were chosen in discussion with a research librarian, with the aim of identifying the maximum number of articles. Databases will be searched from their inception, to 31st May 2010. A three-step search will be used. Initially, a limited search of Pubmed and Cinahl will be undertaken in order to identify appropriate keywords. These keywords are presented in Appendix I. Analysis of the text words and Mesh heading identified by the search to describe relevant articles will then be used to identify additional search terms which will then be used to search across all included databases. The databases to be searched are listed in Appendix II. Thirdly, the reference list of identified papers will be searched for additional studies. The search will not be limited by year but due to a lack of translational resources, included studies will be limited those published in the English language. Assessment of methodological quality The methodological quality of each paper will be critically appraised by two reviewers independently, in order to limit potential reviewer bias. Several quality criteria need to be considered when evaluating studies for potential inclusion. These include: • the clinical spectrum of included patients • blinded interpretation of test and reference standard results • potential for verification bias • patient sampling, prospective design • adequate description of the index test, reference standard, and study population. In order to assess the methodological quality of papers included in the review, the checklist developed by the QUADAS initiative checklist 28, 29 will be used as a critical appraisal instrument. Studies will not be excluded on the basis of quality and issues relating to study quality will be explored. The checklist is presented in Appendix III. In situations where study features may not be reported in the primary studies, the reviewers might need to contact authors or seek additional information for clarification. Data extraction Data will be extracted from included studies using the STARD 25 item checklist 30 (Appendix VI) and will include details pertinent to: population, setting, test details and the sensitivity and specificity for the index test, as compared with viral culture. Data analysis and synthesis Where appropriate, the sensitivities and specificities from individual studies will be combined to generate a summary estimate of the accuracy of PCR techniques, as compared with viral culture methods. Revman 5 (Cochrane Collaboration) and Microsoft Excel computer software will be used for data management and analysis. If meta-analysis is not possible, the data will be discussed in narrative summary and the reasons why the data was unable to be combined will be explored. Conflicts of interest None Acknowledgments As this systematic review will form part of a Masters of Clinical Sciences thesis, a secondary reviewer (TS) will be used for critical appraisal only.