|Home | About | Journals | Submit | Contact Us | Français|
Summary: The menu of diagnostic tools that can be utilized to establish a diagnosis of influenza is extensive and includes classic virology techniques as well as new and emerging methods. This review of how the various existing diagnostic methods have been utilized, first in the context of a rapidly evolving outbreak of novel influenza virus and then during the different subsequent phases and waves of the pandemic, demonstrates the unique roles, advantages, and limitations of each of these methods. Rapid antigen tests were used extensively throughout the pandemic. Recognition of the low negative predictive values of these tests is important. Private laboratories with preexisting expertise, infrastructure, and resources for rapid development, validation, and implementation of laboratory-developed assays played an unprecedented role in helping to meet the diagnostic demands during the pandemic. FDA-cleared assays remain an important element of the diagnostic armamentarium during a pandemic, and a process must be developed with the FDA to allow manufacturers to modify these assays for detection of novel strains in a timely fashion. The need and role for subtyping of influenza viruses and antiviral susceptibility testing will likely depend on qualitative (circulating subtypes and their resistance patterns) and quantitative (relative prevalence) characterization of influenza viruses circulating during future epidemics and pandemics.
The extensive menu of diagnostic tools that can be utilized to establish a diagnosis of influenza virus infection is comprised of old and classic virologic techniques as well as an increasing number of more recent and emerging methodologies and systems. Significant advancements in the field of molecular diagnostics have occurred since the first reverse transcription-PCR (RT-PCR) assay for detection of influenza virus was described in 1991 (131). Continual improvements in technology and assay chemistry have allowed development of molecular assays with high sensitivities and specificities, short turnaround times, and increasing levels of multiplexing capability for detecting respiratory viruses (39, 55, 59, 60, 62, 72, 73, 80). A review of how the various existing diagnostic methods were utilized, first in the context of a rapidly evolving outbreak of novel influenza virus and then during the different subsequent phases and waves of the pandemic, demonstrates the unique roles, advantages, and limitations of each of these methods. Wide variation in diagnostic capabilities of existing diagnostic systems with regard to sensitivity, specificity, and other important performance characteristics, such as turnaround time, throughput, and complexity of use, determined the populations (hospitalized patients or outpatients), locations (clinical laboratories or point-of-care [POC] sites), and purpose (clinical care or infection control) that the different test types were utilized for. The 2009 influenza pandemic is the first pandemic in the molecular era and the first to occur in the face of extensive pandemic preparedness activities that have been undertaken globally since the emergence of the highly pathogenic H5N1 influenza A virus. This review provides an opportunity to revisit and learn from the diagnostic challenges involved at various stages of an influenza pandemic and to better understand and define the needs and characteristics of methodologies and platforms that will prove to most useful in future pandemics.
Viral isolation remains an essential and important skill to be maintained and available in reference laboratories for antigenic characterization of circulating and novel influenza viruses. Viral propagation in culture is also required for antiviral susceptibility testing using neuraminidase inhibitor (NAI) assays, which are important for phenotypic confirmation of resistance and for validation of novel markers of resistance identified by sequencing-based assays. Viral culture is also the only modality for vaccine production for influenza viruses at this time.
Viral culture has increasingly been replaced by molecular assays as the modality of choice for influenza diagnosis in most clinical laboratories. Disadvantages of culture include delayed availability of results (3 to 14 days) and a significantly lower sensitivity, particularly for samples with low viral loads. Shell vial culture techniques introduced in the 1990s reduced the detection time to 1 to 2 days, with sensitivities comparable to those of conventional tube cultures (83, 90, 107). Another advantage is the requirement of less technical expertise when pre-cytopathic-effect (pre-CPE) staining is done. A more recent advancement in the last decade has been the use of cocultured cells, where two or more cell lines that support the growth of a variety of respiratory viruses are mixed and grown as monolayers on shell vials (46, 87). A mix of specific monoclonal antibodies is used for viral detection at 24 h, 48 h, or 5 days postinoculation. The advantage of this method over conventional shell vial culture is the ability to simultaneously culture several viruses in the same shell vial without requiring different cell culture setups for individual viruses. Several cocultured cell line systems are currently available that support the growth of multiple respiratory viruses, including influenza A and B viruses, parainfluenza virus types 1 to 3, adenovirus, and respiratory syncytial virus (RSV). The R-Mix shell vial system (Quidel/Diagnostic Hybrids Inc. [DHI], Athens, OH) combines a human adenocarcinoma cell line (A549) and mink lung epithelial cells (Mv1Lu). The R-Mix Too system combines the Madin-Darby canine kidney (MDCK) and A549 cell lines, which are excellent for isolation of respiratory viruses and have the advantage of not supporting the growth of the highly pathogenic severe acute respiratory syndrome coronavirus (SARS-CoV). Although viral culture is playing less of a role in routine clinical diagnosis of influenza by laboratories performing molecular assays, shell vial-based techniques with hybrid cell lines were utilized widely by a large number of laboratories that did not have access to molecular assays during the pandemic (57). Several studies comparing the performances of R-Mix-based systems to those of other molecular and nonmolecular methods demonstrated that R-Mix culture performed very well, with sensitivities ranging from 88.9% to 96.9% and a specificity of 100% for detection of the 2009 H1N1 influenza A virus (47, 49).
Even in laboratories using molecular methods for routine clinical diagnostics, isolation of virus in culture is highly valuable in some clinical situations, such as for (i) further characterization of viral strains from patients with unusual or severe disease; (ii) sequencing of strains with mutations (e.g., shifted melt profiles) in molecular assays; (iii) culture confirmation of suspected false-positive PCR results; (iv) culture confirmation of positive PCR results for uncommon infection sites, such as the myocardium, brain, or other specimen types that molecular assays are not typically validated for; and (v) differentiating prolonged shedding of viral nucleic acid from true viral replication.
Antibodies to influenza virus appear after ~2 weeks and peak 4 to 7 weeks after infection. A variety of serological tests, including hemagglutination inhibition (HAI) assays, complement fixation assays, and enzyme immunoassays (EIAs), exist (71, 83, 90, 107). A ≥4-fold change or increase in influenza virus antibody titers in paired acute- and convalescent-phase samples obtained at least 2 weeks apart establishes the serologic diagnosis of influenza. The requirement for 2 samples obtained weeks apart renders this method not useful for clinical diagnostic testing, but it can be helpful for making the diagnosis retrospectively. Although not useful as a diagnostic test for most clinical scenarios, serology can play an important role in some situations. These include the following: (i) in the absence of previous vaccination or infection with a particular strain, such as when a novel virus emerges to which there is no preexisting cross-reactive immunity, a single specimen with a positive titer is diagnostic (71); and (ii) serology can help to establish a diagnosis of novel or seasonal influenza virus infections beyond the period when culture and/or PCR tests would be positive. Thus, in patients with a history of influenza-like illness (ILI) but who have stopped shedding the virus or in patients with asymptomatic infections, serology may be the only available option to make a diagnosis.
Serology is a valuable tool for conducting seroepidemiological studies, which are essential for estimating the true burden of infection with an antigenically novel strain. Many such studies performed during the recent pandemic not only helped to ascertain the geographic extent and epidemiologic spectrum of the novel virus after its emergence but also elucidated other important aspects of the virus, such as preexisting cross-reactive immunity in older adults, which were relevant for public health recommendations regarding vaccine prioritization (54, 71, 107). An HAI assay titer of ≥40 has been associated with a 50% or more reduction in risk of influenza virus infection or disease in susceptible populations (11, 54, 109, 124). While this titer has been used by the majority of studies until the present for defining seropositivity, a recent study reported a highest sensitivity and specificity of 90% and 96%, respectively, for adults under 60 years of age and a 92% specificity for adults aged ≥60 years, using a combination of HAI and microneutralization (MN) titers (HAI titer of ≥20 and MN titer of ≥40) of a single convalescent-phase serum sample (124). The authors further suggested that although useful for conducting large-scale seroprevalence studies, an HAI titer cutoff of ≥40 alone may therefore underestimate the number of 2009 H1N1 influenza A virus-infected persons.
Another recent study used an MN titer cutoff of ≥10 for assessing exposure to the 2009 H1N1 influenza A virus (1). MN titers of ≥40 have conventionally been used to represent significant levels of antibodies. This is based on the fact that this level is four times the minimum level of antibodies (≥10) that the assay can detect rather an established correlate of protection. However, using more stringent levels of antibody titers in seroprevalence studies may underestimate the extent of exposure, given the reduced serum antibody responses in elderly and immunocompromised subjects. The study demonstrated that seroprevalence in people of 20 to 29 years of age was significantly higher than that among the 40- to 49-year-old and ≥50-year-old age groups with an MN titer cutoff of ≥40, but there was no difference between the age groups with a titer cutoff of ≥10. Additional studies to better define and optimize serologic criteria for use with different age groups and with convalescent-phase samples would be important to facilitate seroepidemiologic investigations of novel influenza viruses.
Serologic studies were also critical for evaluating the immunogenicity of the 2009 H1N1 influenza A vaccines, which was important for formulating recommendations on dosing for different age groups. Serologic assays have been and will continue to be used for antigenic characterization of circulating influenza viruses to determine antigenic drifts or shifts and formulate vaccine recommendations during both pandemics and annual epidemics. Thus, serologic investigations are an important component of the pandemic response, and maintenance of serologic testing capabilities in select laboratories will be one critical component of future pandemic preparedness activities.
Rapid antigen tests played a unique role throughout the 2009 influenza pandemic. The first case of pandemic influenza in the United States was diagnosed using an investigational rapid test device (Meso Scale Diagnostics) being evaluated in a clinical study in San Diego, CA (13). Subsequent to the emergence and spread of the pandemic virus, rapid antigen tests for influenza were utilized in several clinical settings, both before and after the availability of more sensitive molecular assays for specific detection of the new virus. Of the 7 FDA-cleared tests available in April 2009, 2 tests were Clinical Laboratory Improvement Amendments (CLIA) waived, making them valuable for providing rapid diagnosis in physician's offices and emergency rooms, both of which were sites with substantial test volumes during enhanced surveillance performed early in the pandemic. In addition, the tests were also utilized by the majority of laboratories without molecular capabilities.
Although commercially available rapid influenza diagnostic tests (RIDTs) vary widely in their reported sensitivities (51, 121), their high specificities and positive predictive values during peak influenza season allow early confirmation, facilitate timely treatment decisions, and enable improved patient care by limiting additional and often unnecessary diagnostic and therapeutic interventions (including hospitalizations) in these patients. Furthermore, their lower cost and minimal to no technical complexity render them particularly useful in low-complexity laboratories and low-resource and point-of-care settings both in the United States and abroad.
The high specificities of rapid tests were utilized advantageously during the pandemic to facilitate specific detection of influenza virus in patients presenting with ILI at a time when significant amounts of other respiratory viruses were cocirculating with influenza virus in most communities (78, 126). Rapid specific diagnosis of influenza in patients presenting with clinically indistinguishable influenza and noninfluenza ILI can greatly facilitate appropriate allocation of limited supplies of antivirals early in an outbreak, when effective control strategies have the greatest potential to minimize spread. RIDTs are valuable for their unique and singular capability of providing rapid point-of-care detection among all the diagnostic tools currently available for influenza. It should be noted, however, that although they have widely been shown to have high specificities and positive predictive values during peak influenza virus prevalence, three recent studies evaluating rapid antigen tests during the recent pandemic reported an unexpectedly high false-positive rate of 37.8% (116) or low specificities, of 48.1% (112) and 50.7% (88). The first study noted the limitations of being retrospective in nature and including some specimens (throat swabs) that were not approved by the kit manufacturer. The second report attributed low specificity to the testing algorithms employed during the study, and the third did not offer an explanation. These reports reillustrate the importance of consulting detailed protocols provided by public health authorities and of adherence to manufacturers' recommendations for optimizing the performance of RIDTs.
The major limitation of currently available RIDTs is low and widely variable sensitivity. The clinical sensitivities of these tests have been reported to range from 20% to 90%, varying widely with the populations studied and with study methodologies (sample collection, storage and transport, specimen type, swab and transport media used, and degree of adherence to manufacturers' recommendations for test procedures). Other factors that can account for the wide range of sensitivities reported for RIDTs include interstudy differences in the comparator gold standard used (PCR or culture), definitions of ILI used, duration of illness prior to collection of specimens, and differences in subtypes tested and rapid assays used in the different studies. Sensitivities of RIDTs for the 2009 H1N1 influenza A virus subtype have been reported to be lower than those for seasonal influenza virus by the majority of studies (12, 36, 37, 49, 118, 123), although some have reported them to be comparable (63, 126) or higher (2). The suboptimal sensitivities had important implications for how these tests were utilized in various inpatient (emergency rooms, hospitals, and long-term care institutions), outpatient (clinics), and social (e.g., summer camps or cruise ships) settings. Early reports of low clinical sensitivities led to the recommendations that negative test results should not be used to rule out influenza virus infection, that influenza virus-negative specimens should be tested additionally with more sensitive tests if specific confirmation was sought, and that patients should be treated empirically regardless of a negative test result if clinically indicated. Extensive guidance and algorithms for physicians and clinical laboratory directors on how the tests should be used in various clinical and public health scenarios were published (14, 15) by the Centers for Disease Control and Prevention (CDC). Failure to implement infection control precautions for patients hospitalized with influenza, based on a negative rapid antigen test screen in the emergency department, posed significant health care and infection control burdens resulting from extensive contact investigations for health care workers and delays in treatment (26, 27). Recognition of the low negative predictive values of these tests is important to prevent erroneous infection control and treatment decisions in nursing homes, institutionalized settings, and emergency departments, all of which are sites where these tests are utilized for rapid screening and can help to limit the burden of influenza if used appropriately (18). Low negative predictive values should also be taken into consideration when rapid tests are used in settings where significant transmission of influenza virus occurs, such as schools and camps (17, 25). In addition to their low sensitivities, other important limitations of RIDTs are the lack of influenza virus subtyping by any test currently available in the United States or even of influenza virus type differentiation by many kits, a lack of automation, subjective readouts, and the fact that only a few of the available tests are CLIA waived. Improvements in RIDT design that overcome these shortcomings will represent significant advances in POC diagnostics.
Sensitivities of rapid antigen assays have been shown to correlate directly with viral loads in clinical specimens (12, 36, 86). Although the duration and level of viral shedding can vary with the host (being highest in children and immunocompromised populations) or the underlying cross-reactive immunity from previous influenza virus infections or infections with different subtypes, viral loads have been reported to be highest in the first 2 to 3 days of illness (10, 12, 86). A recent study demonstrated a high overall sensitivity of a rapid antigen test (75.6%) compared with RT-PCR for patients with 2009 H1N1 influenza A, with sensitivity peaking at day 3 of symptoms (79.9%) and decreasing to 67.3% on day 5 (86). This study, however, used a test [SD Bioline influenza A/B/A (H1N1) pandemic test kit (Standard Diagnostics, Yongin, South Korea)] that specifically detects the pandemic subtype along with typing influenza viruses (74, 86). Currently, none of the RIDTs available in the United States specifically target the pandemic H1N1 antigens. While investigations to systematically evaluate the RIDTs currently available in the United States in large prospective clinical studies will be important to clarify their performance characteristics in future influenza seasons, a true increase in sensitivity may require inclusion of monoclonal antibodies with better coverage for the novel strain. Indeed, the rapid antigen test reported to have higher clinical sensitivity in the above study specifically detects the pandemic subtype, using a monoclonal antibody against the hemagglutinin (HA) protein of the 2009 H1N1 influenza A virus and antibodies to the nucleoproteins (NP) of the seasonal influenza A and B viruses. Improvements in analytical and clinical sensitivities of RIDTs will likely require several approaches, including (i) optimization of specimen quality by early collection and adherence to manufacturers' recommendations for specimen collection and transport procedures, (ii) improvements in swab material for better viral capture (29), and (iii) improvements in assay chemistry and coverage. Finally, it is important that once a RIDT is approved by the FDA, there is no requirement for the manufacturer to make changes when influenza virus strains undergo genetic drift. Therefore, laboratories need to continually reevaluate the test's performance characteristics before using the assay for clinical testing.
The rapid antigen tests presently available in the United States are designed to detect influenza A and B viruses, and some can further differentiate the two types. None of the assays currently approved for clinical use can discriminate influenza A virus subtypes. Development of rapid antigen assays that can differentiate subtypes of influenza A virus would be an important advancement in POC diagnostics for influenza, and efforts toward this goal by both academic institutions and industry have been supported by federal funding agencies. Indeed, the first case of the 2009 influenza pandemic was identified in a clinical study of an investigational rapid antigen device with subtyping capabilities, developed by Meso Scale Diagnostics in collaboration with the CDC (13, 70). Detection of influenza A virus and the absence of detection of seasonal H1N1, H3N2, and H5N1 subtypes by this test led to subsequent testing that confirmed the unsubtypable strain to be a novel virus. Our laboratory has been involved in the development of multiplex point-of-care assays for detection of several respiratory viruses and influenza virus subtypes, including the 2009 H1N1 influenza A virus (M. L. Van Dyke et al., presented at the 27th Annual Clinical Virology Symposium and Annual Meeting of the Pan American Society for Clinical Virology, Daytona Beach, FL, 8 to 10 May 2011). Recent reports from other countries have also described the development of rapid antigen tests that specifically detect the pandemic subtype (23, 81, 86). Point-of-care assays that have the ability to subtype can clearly play a critical role in timely detection of novel strains and future pandemics given their utilization at the front lines of clinical care. The availability of such assays with multiplex detection of several influenza virus subtypes and other respiratory viruses would also optimize antiviral management of influenza virus infections in outpatient clinics and emergency departments, particularly during cocirculation of subtypes with various antiviral susceptibility patterns in a given season.
Additional advancements in POC diagnostics would include development of more CLIA-waived tests. In April 2009, 2 of the 7 FDA-cleared rapid antigen tests had a CLIA-waived status and were cleared for use in physician's offices. At present, 14 FDA-cleared rapid antigen tests exist in the market (14). Of these, 5 are CLIA waived: QuickVue influenza test (Quidel), QuickVue influenza A+B test (Quidel), BinaxNOW influenza A&B test (Alere), SAS FluAlert A test (SA Scientific), and SAS FluAlert B test (SA Scientific). Only 2 of the 5 CLIA-waived tests (QuickVue influenza A+B test and BinaxNOW influenza A&B test) can detect and distinguish between influenza A and B virus infections, 1 (QuickVue influenza test) can detect both but not discriminate between influenza A and B viruses, 1 (SAS FluAlert A) can detect only influenza A virus, and 1 can detect only influenza B virus (SAS FluAlert B). Development and availability of more CLIA-waived test kits with better resolution of influenza virus types and subtypes will facilitate increased utilization of these tests and improve rapid influenza diagnosis capabilities at the point of care.
A significant disadvantage of currently available RIDTs is the subjective test output, making interpretation of weak positive results difficult for unskilled and even skilled test users. Recent efforts to overcome this shortcoming have focused on developing automated devices that do not require interpretation by users. These devices utilize detection of a fluorescence signal by the reader and an objective qualitative test result (positive or negative) that does not require manual reads. An attractive feature of these platforms is the ability to connect the reader to laboratory information systems with the advantages of rapid reporting, reduced reporting errors, electronic record keeping, and improved quality control. Automated fluorescence-based systems also have potential for increased sensitivity. Currently, only one such system, the 3M Rapid Detection Flu A+B test (3M), is FDA cleared in the United States (28, 47, 49). At least one additional laboratory-developed automated rapid antigen assay has been described (Van Dyke et al., presented at the 27th Annual Clinical Virology Symposium and Annual Meeting of the Pan American Society for Clinical Virology). This test utilizes a small portable battery-operated fluorescence reader and test cartridges that have the capability of detecting up to 10 analytes. Analytical testing has shown significantly improved sensitivities for 20 influenza viruses, including the 2009 H1N1 influenza A virus and other viruses representing all 16 HA subtypes, all 9 NA subtypes, and both of the currently circulating lineages of influenza B virus, in comparison with the QuickVue influenza A+B assay. Successful development of such devices would be a significant advancement toward the goal of developing sensitive, rapid point-of-care devices with the added advantages of automation and high-level multiplex detection of respiratory viruses.
Molecular assays have increasingly been accepted as the gold standard diagnostic method for detection of influenza virus. Although several amplification methods have been described, the majority of current assays, particularly those utilized in clinical laboratories, are based on the PCR amplification format. Advantages of PCR assays over more conventional viral culture-based diagnostics for influenza include significantly higher sensitivities and short turnaround times. Additional valuable features of PCR-based assays include (i) the ability to test for several targets concurrently and thereby provide type and subtype information, detect other respiratory viruses with overlapping seasonality, and detect influenza virus coinfections; (ii) the ability to be implemented using automated and high-throughput platforms that have the potential for testing large sample numbers and requiring minimum technician time; and (iii) the ability to be adapted rapidly for detection of novel targets. These features were key to the critical role that molecular assays played during the influenza pandemic of 2009. The recent pandemic, however, should also serve as an opportunity to reexamine limitations of currently existing PCR assays and platforms and to revisit the challenges that remain to be addressed in making this format implementable by an increasing number of laboratories as well as other potential end-user sites with various levels of technical skill. A discussion of these advantages in the context of the recent influenza pandemic and further improvements in technology that would facilitate increased utilization of molecular assays in subsequent influenza epidemics and pandemics is presented here. Finally, the role of non-PCR nucleic acid amplification methods in the diagnosis of influenza, e.g., loop-mediated isothermal amplification (LAMP), is discussed.
Molecular assays in use in clinical laboratories at the time when the pandemic emerged comprised of assays with various levels of influenza virus subtyping ability, ranging from none to the ability to detect and differentiate between both the seasonal H1N1 and H3N2 subtypes of influenza A virus. None of the other available diagnostic tests for influenza were capable of subtyping. Thus, subtyping of influenza A virus was not performed routinely by the majority of clinical laboratories. Instead, it was largely limited to public health laboratories, where a small, predetermined percentage of influenza A virus samples were subtyped throughout any given influenza season. The goals of this subtyping were (i) to ascertain the relative prevalence of subtypes circulating at the onset of each year's influenza epidemic and define the predominant subtype, if any, in a geographic region; and (ii) to monitor the emergence of a novel subtype or strain. Subsequent to the emergence of the pandemic strain, subtype determination assumed new levels of both clinical and public health relevance and importance, varying with the different phases of the pandemic. Thus, the need and role for influenza virus subtyping can be discussed in the context of (i) its role during the recent and future pandemics and (ii) its role in routine (nonpandemic) clinical situations for optimizing clinical care.
The ability to determine the subtype for subjects testing positive for influenza A virus was critical early during the pandemic. The annual 2008-2009 influenza epidemic was on its decline and influenza virus was circulating at minimal but detectable levels when the novel virus emerged in the United States. Thus, specific detection of the novel virus was important for any subsequent new influenza cases to gain knowledge of the geographic extent of spread and to understand key virologic features, in particular the transmissibility of the novel strain. Once the 2009 H1N1 influenza A virus strain was detected in several communities and states, recognition of its easy transmissibility and pandemic potential further rendered specific diagnosis of the influenza virus subtype critical for individual patients presenting with ILI. This was because prior to elucidation of the clinical spectrum, epidemiology, and high-risk groups, and potential morbidity and mortality associated with the virus, recommendations were made to treat all patients with ILI. Given the cocirculation of multiple subtypes (seasonal H1N1 and H3N2 influenza A and B viruses and 2009 H1N1 influenza A viruses) with various resistance patterns but indistinguishable clinical presentations, subtype determination was critical for directing limited supplies of antivirals to the appropriately infected group of patients. An inability to specifically implement subtype-guided therapeutics would have resulted in misdirection of precious antiviral reserves at a phase when antivirals can serve as critical tools for mitigation of the pandemic. In subsequent but still early phases of the pandemic, subtyping remained useful as a tool to help ascertain the relative prevalences of the 3 cocirculating subtypes. However, once the virus was established in the community as the predominant circulating subtype during the later weeks of the spring wave, subtyping became less important and unnecessary for most non-high-risk outpatient clinical specimens.
Although seasonal influenza virus was previously recognized to circulate at minimal levels between annual epidemics, higher levels were documented to be circulating nationwide in the spring and summer of 2009, likely a function of enhanced surveillance and testing volumes. The role of subtyping in future pandemics will be determined by qualitative (subtypes and antiviral susceptibilities) and quantitative (relative prevalence) characterization of influenza A viruses circulating at the onset and subsequent phases of the pandemic. Just as importantly, and like the case in the 2009 pandemic, public health policies and guidance regarding the need for specific subtype diagnosis will be dictated by the clinical spectrum, morbidity, and mortality associated with a novel virus and likely tailored to relevant epidemiological risk groups.
Note that while subtyping can serve as a surrogate for antiviral resistance once the antiviral susceptibility profiles of circulating subtypes are well characterized, one important caveat is the regional and local variability in antiviral resistance in different subtypes, which can potentially limit extrapolation of national antiviral resistance data for formulating local public health recommendations for antiviral use. For example, while 10.8% of H1N1 influenza viruses were reported nationally to be adamantane resistant in 2007 to 2008, only 1.2% were resistant in Wisconsin and New York (84). Most importantly, subtype-guided antiviral choices should be made carefully in the context of the pathogenicity and clinical profile of a novel virus, which may dictate the level of specificity required for the diagnosis. Different laboratories employ different strategies for subtype determination. One common strategy used by several laboratories is the detection of a conserved target to type influenza virus strains and a more variable target to specifically detect one or all of the circulating subtypes. Should assays have existed that targeted a genetic region conserved in both the seasonal and pandemic H1N1 subtypes or deduced an H1N1 subtype by a positive influenza A virus signal and a negative H3 signal, mere reliance on subtype results would have misguided treatment, with potentially catastrophic consequences. Although this was not the case in the recent 2009 pandemic, particularly because subtyping was not routinely performed, situations similar to this may arise during future novel influenza epidemics or pandemics in the absence of tests that very specifically discriminate amongst all seasonal circulating subtypes. Finally, the success of this strategy relies critically on the ability of the assay to detect the M target. Although this gene is the most conserved gene in the influenza virus genome, mutations in the gene are well described. Strategies that incorporate two conserved targets would likely further minimize failure of surveillance.
It is important that while specific resolution of seasonal subtypes would be valuable for detecting novel subtypes, continued detection of known subtypes during an influenza season does not rule out emergence and cocirculation of a novel strain. Novel influenza virus subtypes resulting from reassortment of currently circulating and novel strains that acquire at least one (e.g., H1N7) or both (e.g., H1N2) surface protein genes from the currently circulating strains may continue to be detected by existing assays.
Subtyping can be valuable in several routine nonpandemic scenarios as well, such as (i) for directing therapeutic choices when subtypes of various susceptibilities are cocirculating during a season in relatively substantial and proportionate amounts and (ii) for characterizing strains that cause unusual, severe, or prolonged disease. This is important from an epidemiologic standpoint both to associate unusual disease with particular strains or subtypes and to gain incremental knowledge of the evolving clinical behavior of a subtype. Thus, unusually prolonged clinical illness and persistent viral shedding in non-high-risk group patients could indicate increasing virulence and/or development of resistance, and subtype determination is recommended for specific situations such as during institutional outbreaks and in infection control investigations.
Automation and integration of extraction, amplification, detection, and high throughput are additional features that are highly desirable for molecular platforms in hospital-based clinical laboratories with large sample testing volumes. Partial or fully automated and integrated extraction, amplification, and detection systems have the advantages of enhanced speed and efficiency while eliminating risk of contamination and carryover during cDNA manipulation. Automation decreases hands-on technician time and steps and technical errors and facilitates additional runs and higher throughput. These features proved to be of immense value in helping laboratories meet surge capacity demands early during the pandemic and throughout the first wave. The enhanced surveillance recommended by national and local public health laboratories led to high volumes of testing, with some laboratories receiving up to hundreds of samples a day. Clinical laboratories with automated systems were well positioned to provide the throughput required in early phases. Despite significant advancements in the field, few systems exist that allow fully automated and integrated molecular detection.
The most pertinent attribute for any given diagnostic system to serve as a useful tool during a pandemic is the ability to be adapted rapidly to provide specific detection of the novel pathogen. Thus, open system platforms that allow rapid development and implementation of new assays or modification of preexisting singleplex or multiplex assays to include the novel target are highly valuable in any clinical or public health laboratory or in laboratories involved in research and development in academia or industry. Specific assays for detection of the pandemic subtype were developed and launched in several clinical laboratories with preexisting expertise for molecular testing, within weeks of public availability of viral sequences. While conventional PCR and real-time PCR systems both are examples of such open platform technologies, real-time systems have several advantages due to detection of amplicons in real time. These include the ability for quantitative detection, a shorter turnaround time, decreased technician time, and lowered risk of contamination due to detection without the need for postamplification steps.
Open platform systems are of particular relevance for influenza diagnostics given the continual drifts in the influenza virus genome and the high potential for reassortment not only among different subtypes but also between different lineages of influenza viruses circulating at any given time. Even the use of conserved targets is not immune to drifts, with several studies showing accumulation of mutations in internal conserved genes of influenza virus within a few months into the pandemic (96, 104). Mutations in the primer or probe binding areas can have significant impacts on sensitivity (76) and result in false-negative results. Indeed, false-negative results and shifted melt curves and/or cycle threshold (CT) values in association with multiple single nucleotide polymorphisms (SNPs) in highly conserved regions of the matrix gene were described as early as the first wave (133) and, subsequently, in the second wave as well (34, 35, 66). Therefore, molecular assays used in clinical laboratories should undergo periodic evaluation to detect any changes in performance characteristics for detection of recently and currently circulating strains.
Laboratory-developed assays using open platforms have the advantage of rapid adaptability for prompt updates of assays in the event of such drifts in the genome. Indeed, such a loss of sensitivity for detection of the seasonal H3N2 strain circulating in February 2011 was observed with the xTAG respiratory viral panel (RVP) FDA-cleared assay. Absence of detection of H1 and H3 subtypes in samples with a positive matrix signal was initially presumed to be indicative of the pandemic subtype, as was the case in many laboratories using this combination of conserved and subtyping targets. However, this lack of H1/H3 signal was subsequently discerned to be a result of mutations in the H3N2 strain resulting in failure of detection. Such problems can be addressed in a significantly more useful timeline with a laboratory-developed assay than with a manufacturer-offered one. Furthermore, influenza virus strains may vary geographically such that mutated strains in one particular region may not reflect what is circulating elsewhere, which is taken into account before an FDA-cleared assay will be changed by a manufacturer, and that can result in delays in changing an assay that may last longer than a given influenza season.
Interestingly, although the xTAG RVP test's package insert states that it can detect the matrix gene of 2009 H1N1 influenza A virus but cannot identify the hemagglutinin gene of this virus in clinical specimens, it has since been observed and recently reported (48, 92) to occasionally detect the H1 gene of the pandemic virus as well. A retrospective review of raw data for all specimens positive for influenza A virus in clinical laboratory runs over a 7-month period (September 2009 to April 2010) demonstrated that a small but significant number of patients (5%) with the 2009 pandemic virus had a positive or no-call result for the H1 gene, as opposed to the negative result for the H1 gene that was previously observed in clinical studies to correlate with the 2009 pandemic influenza virus. This unreliable detection of the 2009 influenza virus H1 gene by the assay may have led to some cases being misidentified as the seasonal H1 virus, with implications for therapy and now, in the context of disappearance of the seasonal strain, implications for surveillance. Seasonal H1 influenza virus strains, however, were shown to have significantly higher fluorescence than pandemic H1 influenza virus for the HA target, and all of the pandemic influenza virus strains that falsely tested positive for the H1 target gave low positive results, such that manual examination of the raw data clearly distinguished this group from the true seasonal H1 positive strains. Although the seasonal virus was minimal in circulation in comparison to the pandemic strain during those months, this report underscores the importance of extensive clinical evaluations of preexisting assays that do not specifically target a novel virus to discover such issues.
It is also important that while open platform technologies are valuable for timely development of new and rapid modification of existing laboratory-developed assays, not all laboratory-developed assays are as thoroughly validated by individual laboratories as required by the FDA. Limited availability of strains of the novel agent can be a significant problem. Evaluation of a small number of strains and/or small number of clinical samples may fail to bring out important performance characteristics of an assay, as observed in the above study. Furthermore, few laboratories have the capabilities and resources to develop their own molecular assays. These laboratories rely on using existing FDA-cleared assays, and a process must be developed with the FDA to allow manufacturers to modify and optimize these assays for detection of novel strains in a timely fashion. Suboptimal validation of FDA-cleared assays as well as laboratory-developed assays when novel influenza virus strains emerge can pose significant risks, particularly in the context of strains associated with significant morbidity and mortality. The consequences of false-negative results include failure to implement infection control precautions and failure to treat high-risk subjects. Likewise, false-positive results can misinform diagnosis and result in unnecessary and costly therapeutic interventions, such as antivirals and hospitalizations.
Various other non-PCR-based amplification technologies, such as nucleic acid sequence-based amplification (NASBA), loop-mediated isothermal amplification (LAMP), and, most recently, simple amplification-based assay (SAMBA), have been described (91, 105, 129, 132).
LAMP is a nucleic acid amplification method, developed in 2000 (98), that has been utilized successfully for detection of several viral agents, including seasonal influenza viruses (108), H5N1 influenza virus (69), and the 2009 pandemic influenza virus (77, 95). Several studies using clinical samples have shown comparable sensitivity and specificity to those of RT-PCR assays for influenza virus (95, 108). The method utilizes 2 or 3 pairs of primers that bind to 6 to 8 sites on viral cDNA and amplify nucleic acid under isothermal conditions (60 to 65°C), thereby needing simple equipment such as a heat block or water bath. The total assay time ranges from 1 to 1.5 h. Sequence-specific molecular detection with sensitivities and specificities comparable to those of the current gold standard RT-PCR assays without the need for complex and expensive equipment and technical expertise are valuable features of this method. Several recent modifications have further simplified its use and have the potential to facilitate application in nonlaboratory venues. These modifications include the use of dried primers and reagents (95), special extraction reagents for direct extraction of nucleic acid from clinical specimens without extraction equipment, newer detection strategies that permit visual detection without additional technology (89), and, most recently, the use of technologies such as pocket warmers (56) to even obviate the need for heat blocks for amplification. The high level of specificity conferred with the use of multiple primers that target 6 to 8 different sites in the gene have made this useful for SNP detection. The implications of this for influenza diagnosis include the necessity of simultaneous reactions targeting the M gene along with the H gene, given the higher likelihood of mutations in the H gene and the potential for false-negative results if using only the H gene. On the other hand, the ability to detect SNPs makes it valuable not only as a diagnostic tool (with subtyping abilities) but also as one that facilitates surveillance for novel types and subtypes as well as ongoing mutations in the hemagglutinin gene.
A recent study described a SAMBA using isothermal amplification and visual detection on a dipstick. The method was evaluated using clinical samples and reported to have a sensitivity and specificity of 95.3% and 99.4%, respectively, using quantitative real-time PCR as the gold standard (130), and it has similar potential for resource-limited settings to that of LAMP assays. However, it requires extraction of nucleic acid prior to amplification and detection, which adds to its complexity. Another study described a nucleic acid lateral flow assay that uses PCR amplification but detection on a lateral flow strip; the sensitivity and specificity of this test were determined to be 88% and 94% in comparison with real-time PCR using clinical samples, but the test takes 2 to 3 h (106).
Even with the remarkable expansion in molecular testing capacity throughout the first wave of the pandemic, the majority of hospitals in the country had limited access to molecular testing. Availability of point-of-care tests with high sensitivities would have further optimized health care at several levels, particularly in clinical settings with limited resources. Presently, rapid antigen assays are the only such tests that allow rapid diagnosis at the point of care, but with significantly lower sensitivities than those of PCR or even culture. The low sensitivities of these tests have been well documented to lead to significant health care burdens when the tests are used inappropriately. Molecular assays that can be used at the point of care would significantly improve health care in such settings. In recent years, few PCR-based systems have been developed and received FDA clearance for moderate complexity settings. The Cepheid GeneXpert Dx system, the FilmArray RP system, the Verigene RVNAT+ system, and the Liat Analyzer system (Table 1), while offering the possibility of molecular testing at the point of care, have a significantly limited throughput and require substantial investment up front. With additional modifications to reduce complexity and if successfully validated in large analytical and clinical validation studies, LAMP or other such less complex and less costly amplification-based assays appear to have the potential to fill a long-needed niche for rapid point-of-care detection without compromising sensitivities and specificities.
At the time when the pandemic emerged, only 3 FDA-cleared molecular assays for influenza existed: (i) the CDC real-time RT-PCR assay, which is comprised of a series of singleplex assays for typing and subtyping of influenza virus and is available to public health laboratories for influenza surveillance; (ii) the ProFlu+ test (Prodesse Inc., Waukesha, WI), a real-time PCR multiplex assay that detects influenza A virus, influenza B virus, and RSV; and (iii) the xTAG RVP assay (Luminex Molecular Diagnostics, Toronto, Ontario, Canada), a conventional PCR assay that detects 12 analytes, including the seasonal H1 and H3 subtypes of influenza virus. Of these, only the last 2 were commercially available, and only 1 of these 2 assays, the Luminex xTAG RVP assay, typed and subtyped influenza viruses.
On 26 April 2009, the Secretary of the Department of Health and Human Services determined and declared that “there is a public health emergency that affects, or has a significant potential to affect, national security, in this case, 2009 H1N1 influenza virus,” justifying the emergency use authorization (EUA) of in vitro devices (IVDs) for detection of the 2009 H1N1 influenza A virus. Subsequently, a large number of assays (n = 16), the majority of which were manufacturer developed, gained EUA status from the FDA for use as IVDs for detection of the 2009 H1N1 influenza A virus (45). The first of these EUAs, granted on 27 April 2009 for the use of the previously FDA-cleared CDC RT-PCR assay for detection of the novel virus, detected the influenza A virus M gene and specifically detected the HA genes of the seasonal H1 and H3 subtypes but did not specifically detect the novel viral genes. The CDC assay did subsequently receive EUAs for modifications of the assay that specifically detected the 2009 H1N1 influenza A virus nucleoprotein and HA genes, but these were granted much later, on 18 December 2009. The first EUAs for assays that specifically detected and discriminated the novel virus subtype were granted on 9 October 2009 (Diathrix Laboratories 2009 H1N1 test) and 27 October 2009 (Prodesse ProFlu-ST influenza assay). The majority of EUAs for commercially available tests subsequently came through in the period from October 2009 to March 2010 (http://www.fda.gov/MedicalDevices/Safety/EmergencySituations/ucm161496.htm), and the most recent EUA was granted to the IQuum Liat influenza A/2009 H1N1 assay, on 4 May 2010. Thus, a significant share of the surge capacity for the specific diagnosis of this virus in the first two waves was met by laboratory-developed assays. Furthermore, several of these laboratory-developed assays reported equal or better sensitivities than those of the available FDA-cleared assays (3, 5, 58, 66, 103).
FDA-cleared assays remain an important element of the diagnostic armamentarium for pathogen detection. Several laboratories do not have the resources and infrastructure to develop, validate, and comply with the extensive quality control and regulatory requirements that need to be met before clinical implementation of laboratory-developed molecular assays. The extensive analytic and clinical validation that underlies an FDA approval process, however, is not permissive to prompt availability of such assays. The EUA process put in place to facilitate a more expeditious validation certainly helped expand molecular testing capacity during the pandemic, particularly to laboratories with no expertise to develop their own assays. However, as of 23 June 2010, all of the EUAs were terminated, and only 4 of the 16 EUA tests obtained FDA clearance for use as an IVD for detection of the 2009 H1N1 influenza A virus. The first of these four assays was a commercially available assay, the Simplexa influenza A H1N1 test (2009), by Focus Diagnostics Inc., Cypress, CA (FDA cleared on 24 May 2010). The second assay was the CDC rRT-PCR 2009 A(H1N1) pdm Flu panel (granted FDA clearance on 22 June 2010) (45), available only to public health laboratories. In addition, the Cepheid Xpert Flu A panel and the IQuum Liat influenza A/B assay received FDA clearance on 24 April 2011 and 4 August 2011, respectively. It should be noted that the Prodesse ProFlu-ST EUA was terminated along with the other EUAs. However, another subtyping assay from Gen-Probe Prodesse, Inc. (the ProFAST assay), obtained FDA clearance (Table 1).
Importantly, even with the availability of FDA-cleared assays, many laboratories do not have the expertise to launch molecular assays. Of all the FDA-cleared molecular assays available for detection of influenza viruses at this time, only four (Table 1) have a CLIA classification of moderate complexity. These include the Verigene RVNATSP test (Nanosphere Inc., Northbrook, IL), the FilmArray Respiratory Panel (RP) test system (Idaho Technology, Inc., Salt Lake City, UT), the Cepheid Xpert Flu assay (Cepheid Inc., Sunnyvale, CA), and the Liat influenza A/B assay (Iquum Inc., Marlborough, MA). These systems fully integrate and automate all steps of sample preparation, nucleic acid amplification, and detection, with minimal hands-on technician time. However, in using a single-test format with 1 sample per instrument, with turnaround times of 3 to 3.5 h on Verigene RVNAT systems and ~1 h on FilmArray and Cepheid systems, they are suitable only for small- or medium-volume testing. The Liat Analyzer system has a turnaround time of 20 min per sample; however, it does not specifically detect the 2009 H1N1 influenza A virus.
Clinical and research laboratories in academia and industry with preexisting expertise, resources, and infrastructure for developing and implementing molecular tests played an unprecedented role in providing the laboratory surge capacity required during this pandemic (3, 5, 8, 24, 49, 57, 58, 78, 103). In fact, such laboratories were able to develop and launch their own in-house assays well in advance of receiving centrally distributed kits. Furthermore, although the CDC rapidly developed and disseminated real-time PCR protocols for detection of the novel virus (128), their utilization by end users required expertise to implement and adapt their assays to local PCR systems and platforms, again made possible only by the existence of local molecular expertise.
The ability to launch laboratory-developed assays in a clinically useful time frame will be critically dependent on the availability of open platform systems. Likewise, the ability to meet laboratory surge capacity during pandemics will require advancements in automation, multiplexing capability, and throughput. While open platforms have been available for over a decade, advancements such as full automation and integration of extraction, amplification, and detection steps are fairly recent. The first two systems that fully integrated and automated these steps, the GeneXpert Dx system and the Liat Analyzer system, became available as recently as 2004, and the Verigene RVNATSP and FilmArray RP systems received FDA-cleared status just this year (Table 1). However, these systems are not “open platforms” in that they do not lend themselves to in-house development by laboratories, although development could be done in collaboration with the manufacturers.
The first fully integrated and automated open platform system available for in-house development was the Jaguar system (HandyLab Inc., Ann Arbor, MI), which became available in 2008. The technology was bought and renamed the BD Max system by Becton, Dickinson and Company in 2009. A laboratory-developed multiplex assay for influenza A and B viruses and RSV developed on this system was used to screen hundreds of samples a day in a large clinical laboratory early during the pandemic (3). Full automation requiring minimal hands-on time, a 24-sample run capacity, and a turnaround time of ~2.5 h significantly facilitated throughput. The open platform format of this system made rapid adaptation of this assay for specific subtyping of influenza A virus feasible, such that the assay was available to meet diagnostic demands during the first and second waves of the pandemic. This system is a 2-color system that allows moderate-level multiplexing using its melt detection feature. A 2nd generation of this technology, the BD Max6 system, which uses 5 colors, was recently released and is being made available to some laboratories for assay development. This system has the potential to allow significantly greater levels of multiplexing, again through its melt feature and more channels. There are other emerging technologies capable of high-level multiplexing (7, 91), but these are high-complexity and labor-intensive systems involving separate pieces of equipment and long turnaround times of 6 to 10 h. Finally, fully automated benchtop systems that have a small footprint allow expansion of molecular testing to facilities with small laboratory space. The availability of technologies that can be used to develop and modify laboratory-developed assays for novel pathogen detection in a clinically useful time frame will be of increasing importance to future public health emergency laboratory responses.
The first case of the novel influenza virus infection was diagnosed by an investigational rapid antigen test during a clinical study in Southern California, not at a sentinel PCR surveillance site. Although the novel pandemic strain would have eventually been detected via the CDC or regional virologic surveillance sites, the fact that it escaped earlier detection underscores the importance and impact of public health policy in optimal utilization of diagnostics for meeting public health goals. Thus, testing smaller numbers of samples or suboptimal frequencies of surveillance testing would potentially decrease and/or delay detection of a novel strain, particularly in earlier phases prior to its widespread dissemination. Early detection is critical to containment and reduction of morbidity and mortality during a pandemic. Despite nationwide surveillance being in place, the 2009 pandemic strain is estimated to have circulated unrecognized for weeks (96) prior to its detection. Public health policy on the intensity of routine surveillance for influenza should be reevaluated in light of lessons from the recent pandemic. An increased level of sampling to improve geographic representation may be required to optimize surveillance.
Close collaboration between public health laboratories and private and academic laboratories is critical during public health emergencies. Such collaboration is essential for expeditious validation of laboratory-developed assays developed in academic research or clinical laboratories. While pandemic influenza preparedness activities have greatly increased the capacity of public health laboratories to perform detailed molecular subtyping of influenza virus for routine surveillance, assays developed in local hospital laboratories and FDA-cleared assays provide the majority of surge capacity needed during a public health emergency.
The recent pandemic demonstrated the excellent and mutually beneficial collaboration between several private and public health laboratories, in accordance with the provisions of the Laboratory Response Network (LRN) established by the CDC. Assays developed in local sentinel laboratories were validated by state laboratories within days after the emergence of the pandemic in communities. This subsequently allowed accurate reporting of numbers of positive patients by public health officials by including results from private local laboratories. Delays in testing and reporting results can lead to significant underrepresentation of influenza activity at a regional level, particularly during outbreaks of a novel virus, as observed to be the case during the recent pandemic by us and others (78, 99). Ten days after the first case of infection with the novel virus was detected in our community, only 26 cases were reported to have been identified in Wisconsin by the CDC data, when local laboratories were reporting 268 cases by that time (78), and only 3 cases were being reported for the state of Illinois when local laboratories had reported 39 cases of probable novel H1N1 infection (99). Public health decisions for a given geographic population are likely better informed using data that reflect local activity of an emergent pathogen in real time rather than national data early in an outbreak. The unprecedented collaborative sharing of data between private and public health laboratories assisted public health authorities in assessing the burden of disease in the community, providing frequent updates in real time to the public and making informed public health recommendations about testing and antiviral treatment.
Rapid incorporation of private clinical or research diagnostic laboratories with preexisting resources for molecular diagnostic testing into the LRN provides the diagnostic laboratory “surge capacity” required during an outbreak or pandemic (24, 78, 79). Note that while laboratory-developed assays are highly useful during such public health emergencies, there can be significant interlaboratory variation in test characteristics among assays developed by individual laboratories. Existing FDA-cleared influenza assays can also vary in their performance characteristics for detection of a novel strain. The validation of both laboratory-developed assays and FDA-cleared assays by public health laboratories has the added advantage that the performance characteristics of the assay can be determined in comparison with a standardized set of reagents.
Antivirals are important components of the anti-influenza virus arsenal, second only to vaccination, which is the most effective measure for reducing the morbidity and mortality from influenza. However, the efficacy and impact of influenza vaccination can be limited due to vaccine mismatch, host-related factors such as immunosuppression and being elderly, the time required to develop immunity after vaccination (up to 2 weeks in adults and 6 weeks in children), lack of compliance, and, rarely, contraindications or precautions for the vaccine. Furthermore, a vaccine may not be available for periods of up to months after emergence of a novel strain. Antivirals may be the only option available not only for treatment but also for prevention of influenza under these circumstances. Four antiviral drugs are currently licensed and approved for prophylaxis and treatment of influenza virus infections. Amantadine and rimantadine (adamantanes) are active only against influenza A virus, whereas oseltamivir and zanamivir (neuraminidase inhibitors [NAIs]) are active against both influenza A and B viruses. The drugs differ in their routes of administration, age groups they are approved for, and resistance profiles (40). Oseltamivir is the only drug available in oral formulation and the only currently available neuraminidase inhibitor that can be used in infants and young children under 5 years of age.
Of critical importance to the use of antivirals for influenza is that antiviral resistance patterns have the potential to evolve significantly between consecutive epidemics and even during a given influenza season, with resultant implications for empirical treatment and prophylaxis recommendations. Thus, rapid and widespread emergence of resistance to adamantanes among H3N2 strains in the 2005-2006 season led to the recommendation that these drugs no longer be used for prophylaxis or treatment of influenza (19). Subsequently, a sudden and unexplained increase in H1N1 resistance to oseltamivir, from 0.7% in the 2006-2007 season to 10.9% in the 2007-2008 and >99% of strains circulating in the 2008-2009 influenza season, necessitated reinstitution of this class in the 2008-2009 season to combination treatment regimens (adamantane and oseltamivir) for unsubtyped influenza A virus-positive infections when zanamivir could not be used (e.g., patients of <7 years of age, those with chronic lung disease, ventilated patients, and those with inability to use the zanamivir inhalation device). Despite the 10.9% resistance to oseltamivir in H1N1 viruses during the 2007-2008 season, a comparably substantial contemporaneous adamantane resistance rate of 10.7% in H1N1 viruses and continued high levels of resistance (99.8%) in H3N2 viruses, the predominant subtype in circulation that year, precluded adamantanes from being recommended in 2007 to 2008. The decline in adamantane resistance in H1N1 viruses to 0.4%, along with an increase in oseltamivir resistance in H1N1 viruses, to 99.4%, in the subsequent year, however, led to the recommendation to add an adamantane when using oseltamivir for patients with influenza A virus or if the type was unknown. Rates of antiviral resistance can also vary regionally, in some cases with substantial deviation from national rates. This was exemplified by a 17.4% H1N1 oseltamivir resistance rate (versus 10.9% nationally) and 1.2% H1N1 adamantane resistance rate (versus 10.7% nationally) in isolates obtained from Wisconsin in the 2007-2008 season (84).
This significant temporal and regional variation in prevalence of resistance to the limited antiviral medications available for influenza has important public health and clinical ramifications. Antiviral susceptibility testing is not routinely available in clinical laboratories. In the absence of availability of such testing in clinical laboratories, consulting regional and national influenza surveillance data on types and subtypes of influenza viruses circulating at a given time and resistance patterns thereof becomes important in order to make informed treatment decisions, particularly when empirical use of antivirals is complicated by circulation of multiple subtypes with different susceptibilities. Currently, antiviral resistance surveillance is conducted by the CDC, using specimens submitted from various national sites that are tested at the CDC Influenza Laboratory, as well as using specimens that are tested locally at some state public health laboratories. Although of great importance to public health, this surveillance has its limitations due to (i) variable participation of state laboratories and submission of specimens and (ii) the delays involved in laboratory testing and reporting results that are inherent to a surveillance program. Thus, weekly antiviral surveillance reports may have some states not represented at all, particularly early in a season, which can result in major gaps in data given the significant geographic variation in circulating subtypes. Similarly, there can be important cumulative changes with inclusion of more representative specimens over time, making it important for physicians and public health officials to keep current with weekly influenza surveillance reports published by the CDC, while realizing potential pitfalls associated with the reports. Lastly, the antiviral resistance testing at the CDC is performed on either original clinical samples that are tested for a single known mutation in the virus that confers oseltamivir resistance in currently circulating H1N1 strains or influenza virus isolates that are tested using a neuraminidase inhibition assay that determines the presence or absence of neuraminidase inhibitor resistance, followed by neuraminidase gene sequence analysis of resistant viral isolates. Testing of viral isolates alone without matching clinical specimens has the potential to overestimate resistance, as discussed below.
At present, antiviral susceptibility testing is not performed routinely in most clinical laboratories in the United States. The decision to implement routine influenza virus antiviral resistance testing in clinical laboratories would best be made in the context of local molecular epidemiology of influenza due to significant geographic and temporal variation in circulating subtypes and resistance patterns annually. Some epidemiologic and clinical scenarios that would warrant a closer look at the advantages and role of resistance testing are discussed below.
In the 2008-2009 prepandemic season, cocirculation of oseltamivir-resistant, adamantane-sensitive H1N1 influenza A viruses with oseltamivir-sensitive, adamantane-resistant H3N2 and influenza B viruses led to the recommendation to use an adamantane along with oseltamivir for treating infections with unsubtyped influenza virus or viruses of the H1N1 subtype. However, antiviral recommendations based on national surveillance data may need to be optimized to reflect variation in resistance patterns noted on further analyses. For example, while the majority of seasonal H1N1 viruses were sensitive to adamantanes, a small number were resistant (~11% in 2007 to 2008 and 1% in 2008 to 2009). Thus, two distinct clades of this virus were circulating: clade 2B, which was adamantane susceptible and oseltamivir resistant (the majority), and clade 2C, which was adamantane resistant and oseltamivir susceptible. Also, some H1N1 viruses were resistant to both classes of drugs. A recent study reported an increasing prevalence of dual resistance in seasonal H1N1 isolates, from <1% (n = 1,753) to 28% (n = 25) of isolates studied during 3 consecutive influenza seasons from 2007 to 2009 (113). The use of an adamantane for prophylaxis or treatment in the subset of patients who were infected with such strains with adamantane or dual antiviral resistance would be ineffective and have the potential to further select resistant subpopulations. Clearly, the dynamics of influenza virus resistance patterns will influence the debate on antiviral susceptibility testing at any given time, but some epidemiologic patterns can be envisioned to better indicate routine testing than others.
Although the current circulation of 2009 H1N1, H3N2, and influenza B viruses with similar susceptibilities to antivirals (with nearly all being susceptible to oseltamivir and resistant to adamantanes, with only sporadic resistance of the 2009 H1N1 influenza A virus to oseltamivir) makes routine resistance testing unnecessary, the need for such testing certainly arises on a case-by-case basis and is often unmet by the currently existing system, which necessitates sending samples out to a central national laboratory or select clinical laboratories, with turnaround times of several days. Clinical situations that would greatly benefit from timely antiviral susceptibility results include cases of severely ill patients with poor clinical responses and prolonged viral shedding while on antivirals and immunocompromised patients, who are at higher risk for developing resistance (9, 67). This is particularly so given that (i) NAIs are the only antiviral class available for currently circulating strains, (ii) the use of inhaled zanamavir (the only drug to which resistance has not been documented thus far in the currently circulating strains) is often precluded by severe disease and/or mechanical ventilation in the most complicated patients, (iii) a lack of response to oral oseltamivir may be confounded by poor drug absorption in some patients with concurrent gastrointestinal illness, (iv) the use of intravenous (i.v.) peramavir (available in select countries but not in the United States, since the EUA expired on 23 June 2010) is of limited value in the presence of oseltamivir resistance, and (v) i.v. zanamavir is currently available only via a compassionate care protocol, and timely resistance results would facilitate early initiation of requests by physicians. A recent study (9) showed a remarkably high prevalence of oseltamivir resistance (33%) in immunocompromised patients (n = 12), with more than half of the resistant viruses being detected at early time points in the course of infections, including prior to antiviral exposure. Emergence of resistant mutants in mixed populations was noted as early as 4 days after antiviral exposure in 1 patient with phenotypic evidence of resistance by day 7. In the absence of routine antiviral susceptibility testing in clinical laboratories, immunocompromised patients often receive prolonged courses of oseltamivir and/or other antivirals before antiviral resistance testing is sought and results become available. Early detection of resistance would greatly facilitate timely changes in antiviral therapy in such cases. Currently, some places use CT values from real-time PCR runs as a surrogate to follow viral loads, with a decrease of 3 CT values indicating an increase of ~1 log in viral load. Thus, persistently low or declining CT values on serially obtained specimens have been used as potential evidence of development of resistance in a given patient on antivirals (93). However, currently available influenza virus PCR assays are not validated for quantitative purposes, and the viral loads measured can vary due to a lack of standardized sample collection procedures. Differences in types of samples (nasal swabs, washes, or nasopharyngeal swabs) collected and interpersonal variation in sampling technique can lead to differences in amounts of respiratory secretions and virus in samples. Therefore, the availability of sensitive and specific antiviral susceptibility tests with rapid turnaround would be of great value to clinicians for optimizing management of influenza in these select clinical populations.
Currently available surveillance-driven antiviral resistance data may have some limitations due to suboptimal local representation in national surveillance data or suboptimal sampling, considering the evolutionary dynamics of this virus with high potential for de novo emergence or selection of resistance in the face of antiviral pressures. Enhanced local surveillance may be considered in some situations to overcome these shortcomings. Testing a small, predefined subset of all influenza virus-positive samples over the course of a season in clinical or local public health laboratories would likely better reflect local resistance trends and optimize antiviral usage. Also, antiviral resistance in influenza virus has been well described to evolve through acquisition of mutations (de novo or as a result of drug selection) or by reassortment resulting in acquisition (65, 68, 97, 101, 120) or loss (41, 42) of new or previous mutations that confer resistance. Neuraminidase inhibitor resistance is particularly prone to such evolution given its association with a surface protein. Enhanced sampling on a local level would help to monitor any evolution in resistance patterns and provide an opportunity to detect and control the spread of any emergent resistant strains in a timely fashion.
Currently existing methods for resistance testing include (i) the neuraminidase (NA) enzyme inhibition assay and (ii) assays to detect molecular markers of antiviral resistance, such as conventional Sanger sequencing, pyrosequencing, and RT-PCR-based assays. Pertinent characteristics of the different methodologies are discussed here in the context of influenza virus antiviral resistance.
The NA enzyme inhibition assay is the traditional method for detecting influenza virus resistance and uses cell culture-grown virus. Since it is a functional assay, the endpoint outcome (reduced susceptibility or resistance) can be ascertained regardless of the molecular pathogenesis of such resistance. Elevations in the 50% inhibitory concentration (IC50), however, are not sufficient for defining resistance and are used in combination with sequencing to identify new or detect previously recognized molecular markers of resistance (101, 114). One caveat associated with using this method is that the inherently higher basal IC50s of NAIs for influenza B viruses (38, 94) can lead to erroneous interpretation of influenza A virus isolates as NAI resistant for patients with mixed influenza A and B virus infections. Although reports of mixed influenza virus infections have previously been uncommon (38, 94), recent studies using deep sequencing have estimated high-level mixed influenza virus infections that include diverse lineages of the same influenza virus subtype, viruses of different types (A and B), or quasispecies with different antiviral resistances (43, 44). Thus, confirmation of the presence of types, subtypes, and even minor nondominant quasispecies that could affect IC50s in functional assays would be important in using this assay. Other limitations of using this assay are the need for viral isolation and propagation and potential overestimation of resistance due to selection or expansion of resistant mutants during viral propagation.
The method remains an essential technique for determining the role of any novel mutations identified by sequencing of samples obtained from patients suspected to have antiviral resistance. A recent study identified the presence of the I22V mutation, previously known to confer resistance in seasonal influenza A and B viruses, for the first time in the 2009 pandemic virus (31); however, confirmation of its role as a marker of resistance would require phenotypic studies. Therefore, although an increasing number of molecular determinants of resistance in the NA gene are being recognized, accurate identification and validation of such markers require phenotypic correlation, with demonstration of elevated IC50s. Note that the sensitivities of phenotypic assays are lower than those of molecular assays, with ~25% resistant mutants required for a significant increase in IC50. To address this issue, improvements in the sensitivities of NA enzyme inhibition assays themselves have been described recently, through improvements in curve fitting analyses that can lead to sensitivities comparable to that of pyrosequencing (30).
Assays for detection of molecular markers of resistance utilize sequencing or PCR technologies to identify known or novel markers of resistance. They offer several advantages over the traditional NAI assay, including the ability to be used directly on clinical specimens, including those with nonviable viruses. Viral isolation is time-consuming, may not be successful for patients on antivirals, and has the potential to alter the proportion of resistant mutants, but all of these limitations can be overcome by assays that can be used directly on clinical specimens.
Pyrosequencing has emerged as a useful sequencing technique in recent years. This technology has been used successfully to identify and detect molecular markers of antiviral resistance in seasonal H1N1 and H3N2 influenza viruses (6, 32), H5N1 virus (33), 2009 H1N1 influenza A virus (31), and influenza B viruses (115). The method is employed by the CDC for detection of antiviral resistance directly in clinical samples and is used in combination with NAI assays to detect novel or previously known molecular markers of resistance. Advantages of pyrosequencing-based assays include (i) a rapid turnaround, with results available in less than 5 h; (ii) help in discovering previously unknown markers of resistance by identifying new mutations within the area of the genome that is sequenced by a given assay; and (iii) increased sensitivity. Pyrosequencing can reliably, accurately, and quantitatively detect mutants present at as low as 5 to 10% prevalence in mixed viral populations, in contrast to Sanger sequencing, which requires mutants to be present at levels of ≥50% to be detected (82). Rapid sensitive quantitative detection makes this technology particularly valuable for antiviral resistance testing in a clinical context. Antiviral-resistant mutants can vary in prevalence from being minor to being dominant in clinical specimens over the course of prolonged infections. Sensitive and timely detection of resistant mutants directly in clinical specimens, which are often obtained serially from immunocompromised patients, who are at highest risk for emergence of such mutants while on sequential or combination antivirals, would enable timely therapeutic changes. Additional advantages include (iv) automation and the capacity for high throughput, as the ability to be run in a 96-well format allows the throughput needed for testing large numbers of samples as well as for analysis of several mutations if qualitative breadth is desired; and (v) use of the technology to help gain an understanding of the clinical relevance of minor resistant variants in well-designed clinical studies. If minor resistant populations identified early in the course of infections are shown to have clinical and phenotypic significance, further studies to optimize drug doses for prophylaxis and treatment would have the potential to advance the current knowledge and management of antiviral resistance in influenza virus. Pyrosequencing-based assays have some limitations as well. (i) While markers of adamantane resistance in influenza virus are well established (mutations at residues 26, 27, 30, 31, and 34 in the M2 protein), molecular determinants of NAI resistance are not as well characterized. Pyrosequencing assay design requires prior knowledge of the locations of mutations, thus restricting its use to detecting previously elucidated markers of resistance. (ii) Primers for sequencing assays for NAI resistance would require continual updates with cumulative knowledge of new markers of resistance as well as due to the natural variation well known to occur within surface proteins of influenza virus. (iii) Design of these primers would also require prior knowledge of the influenza virus subtype, given that neuraminidase resistance is type and subtype specific. (iv) These assays require specialized equipment and training and are not readily available in clinical laboratories. (v) Although pyrosequencing has demonstrated sensitive detection of resistant mutants present at ~10% in viral mixtures, the significance of this limit of detection remains to be determined. The clinical relevance of various proportions of sensitive and resistant viral populations in any given patient is unknown and requires further study. While minor variants may not be associated with clinical and/or phenotypic resistance, early detection of emergence of such minor variants in serially obtained specimens can signify resistance at later time points during the course of prolonged infections. On the other hand, resistant viral populations at levels as low as 1% or lower have been detected in clinical samples by use of other methodologies, and if these populations are shown to have clinical significance, this would render even the current sensitivity of pyrosequencing-based assays suboptimal.
Several recent studies (4, 22, 61, 102, 110, 111, 122; Van Dyke et al., presented at the 27th Annual Clinical Virology Symposium and Annual Meeting of the Pan American Society for Clinical Virology) have described the development of real-time RT-PCR assays for detection of the H275Y (in N1 numbering) mutation that is associated with oseltamivir resistance and reported to be present in all resistant 2009 H1N1 influenza A viruses analyzed so far. A valuable advantage of these assays is that many clinical laboratories with molecular capabilities already have the infrastructure and expertise required for development and implementation of assays using an RT-PCR-based format. A short turnaround time of 3 to 4 h, automation, and high throughput are added benefits. A potentially important feature of real-time RT-PCR assays is their increased sensitivity over that of pyrosequencing assays. A recent study described the development of a real-time PCR assay that reliably detected mixed populations at percentages down to 0.1% H274Y variant component or 1% wild-type component in viral nucleic acid extracts. Assays with such excellent analytical sensitivities offer the advantage of detecting mutations in samples with viral loads that are too low to be detected by pyrosequencing (21, 102). Another advantage of assays with high analytical sensitivities is the ability to confirm or refute the role of mutations identified in cell culture-propagated isolates but not in original samples. A novel Q136K mutation in the NA gene, reported to confer zanamivir and peramivir resistance, was identified in 9 prepandemic H1N1 virus isolates but not in their original clinical specimens (64). Similar detection of mutations (D151G/N) that potentiated the effect of the H274Y mutation on NAI susceptibility and conferred cross-resistance to all 4 NAIs tested was noted by pyrosequencing for viruses propagated in MDCK cells but not in matching clinical specimens (101). This discrepant detection in isolates but not clinical specimens suggests that the variants either occurred in very small proportions in the primary clinical specimens, below the limits of detection of pyrosequencing (~5 to 10%), or reflected genetic variability introduced during passage in culture (100). Detection of mutants in viral isolates should be evaluated in the context of the number of passages, which is often not known or may be high in surveillance or research studies but would be important for interpreting results for clinical purposes.
Detection of resistant mutants in clinical samples is important. However, the need for development of sensitive assays that detect the presence of minor resistant populations has to be balanced by elucidation and understanding of the relevance of such variants for antiviral treatment. Comprehensive studies of virologic resistance at the genotypic and phenotypic levels, along with clinical correlation, would be required to facilitate accurate validation of novel molecular determinants as well as help determine accurate assay cutoffs for clinically relevant resistance.
NAI resistance in the 2009 H1N1 influenza A virus remains rare (<1%), with the majority of this being observed in immunocompromised patients with prior exposure to the drug (16, 20, 50). All resistant isolates analyzed to date have been reported to have the H275Y mutation. However, the recent precedence of unexplained emergence and rapid global spread of oseltamivir resistance in the prepandemic H1N1 virus, unrelated to drug usage, should serve as a reminder to maintain constant vigil for similar trends in currently circulating and any future novel subtypes. Zanamivir resistance has been reported rarely, with just one resistant influenza B virus reported for a zanamivir-treated immunocompromised patient (52, 53). To date, none of the 2009 H1N1 influenza A viruses have been reported to be resistant to this drug.
The role of antiviral susceptibility testing in routine clinical testing would likely benefit from continued discussion, particularly in the context of local influenza epidemiology, annual and geographic trends in resistance, and the composition of patient populations primarily served by clinical laboratories. One approach may be to do targeted enhanced surveillance of resistance in high-risk populations (children, immunocompromised patients, and those with underlying respiratory morbidity, all of which have prolonged viral shedding and have been reported to have higher rates of antiviral resistance [75, 117, 119, 125, 127]). Sensitive assays that can accurately detect resistant mutants in clinical specimens will be valuable. However, diagnostic criteria will need to be developed and standardized for interpreting test results for clinical use, particularly for tests with excellent sensitivity for minor resistant variants.
Swati Kumar is an Assistant Professor in Pediatrics at the Medical College of Wisconsin. After receiving her M.B.B.S. and M.D. degrees in pediatrics from the Institute of Medical Sciences, Varanasi, India, she completed a residency in pediatrics and a fellowship in pediatric infectious diseases at the State University of New York, Downstate Medical Center, Brooklyn, NY. Her postdoctoral studies were supported by a MedImmune Pediatric Research Fellowship grant. She is board certified in pediatrics and pediatric infectious diseases. She has authored and coauthored papers on clinical, epidemiologic, and molecular studies of respiratory pathogens and served as an ad hoc reviewer for several journals in her field. Her interests include studying the clinical, epidemiologic, and genetic spectrum of respiratory viruses and their molecular diagnostics and evolution.
Kelly J. Henrickson is a Professor of Pediatrics and Microbiology at the Medical College of Wisconsin, Milwaukee, WI. He is board certified in pediatric infectious diseases and cares for children and teaches residents and fellows at the Children's Hospital of Wisconsin. In addition to his academic and clinical work, Dr. Henrickson has served as an editor and reviewer in various national and international capacities and has published over 134 articles, abstracts, and book chapters, primarily on respiratory viruses. His research interests include molecular epidemiology and evolution and development of multiplex molecular and rapid point-of-care assays for respiratory viruses. Dr. Henrickson's recent work includes studies to investigate the epidemiology and evolution of the 2009 H1N1 influenza virus and paramyxoviruses at the whole-genome level.