NS1 antigen detecting assays for diagnosing acute dengue infection in people living in or returning from endemic countries

  • Protocol
  • Diagnostic

Authors


Abstract

This is the protocol for a review and there is no abstract. The objectives are as follows:

To assess the diagnostic accuracy of commercial NS1 antigen detecting assays for acute dengue infection.

To compare the diagnostic accuracy of LFIs and ELISAs antigen detecting assays.

To explore the following potential sources of heterogeneity: degree of endemicity, continent, patient age, primary versus secondary dengue infection, serotype, type of participant (residents or traveller), and timing of sample collection from onset of symptoms.

Background

 

Target condition being diagnosed

Dengue is an important global health problem. Every year it infects 50 million people, resulting in about 500,000 hospitalized people with dengue haemorrhagic fever and a case fatality rate of 5% to 10%. Dengue is endemic in at least 100 countries in Asia, the Pacific, the Americas, Africa, and the Caribbean. Approximately 2.5 billion people live in areas where they are at risk of dengue infection (Kyle 2008; WHO 2009; Guzman 2010a; Huy 2010; Alves 2011; Amarasinghe 2011; Azami 2011; Blaylock 2011; Brown 2011; Chaaithanya 2011; Cuong 2011; Franco 2011; Gutierrez 2011; Hayden 2011; Lernout 2011; Rai 2011; Tissera 2011; Whitehorn 2011; Hynes 2012; Simmons 2012). The dengue transmission area is continuing to expand due to many direct and indirect factors linked to increased urbanization, increased travel, and global warming (McMichael 2006; Natiello 2008; Banu 2011; Chen 2011; Luz 2011; Amarasinghe 2012; Bouri 2012).

Dengue is a virus transmitted by Aedes mosquitoes, mainly Aedes aegypti. The virus incubation period is between four to 10 days from the bite of a mosquito infected with dengue virus and the start of symptoms in the infected person (WHO 2009).The dengue virus belongs to the Flaviviridae family of viruses and has four serotypes (DENV 1, DENV2, DENV3, and DENV 4). Its genome encodes three structural proteins (the capsid (C), membrane (M) and envelope (E) glycoproteins) and seven non-structural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5). Each dengue virus type shares around 65% genome identity, which is approximately the same degree of genetic relatedness that West Nile virus shares with Japanese encephalitis virus. Despite genomic differences, each dengue virus serotype causes nearly identical syndromes in humans. Infection of a person with one serotype induces life-long immunity against reinfection by the same serotype only, and transient and partial protection against infection with the other serotypes (Gubler 1997; Guzman 2010a; Gordon 2013).

Clinical presentation

Clinically, dengue infections can range from causing no illness through to causing death (Guzman 2010b; Giraldo 2011; Kalayanarooj 2011a; Simmons 2012). People at an early stage of infection may present with sudden onset of fever and non-specific symptoms such as body ache, myalgia, skin erythema, facial flushing, headache and arthralgia that are common to other febrile syndromes. This is defined as the acute phase of the disease and typically lasts for between two to seven days. After this initial phase, the disease usually progresses through a critical phase characterized by defervescence (temperature drops to between 37.5°C to 38°C) and by varying degrees of plasma leakage into the pleural and abdominal cavities that lasts for 24 to 48 hours. At this stage, different scenarios in disease progression may occur. Some people will improve after defervescence while a proportion of people will deteriorate with warning signs. This is known as dengue with warning signs. These people will probably recover with early intravenous rehydration. However, some people will deteriorate and progress to the severe forms of the disease, which can result in death if not managed appropriately. Classification of dengue disease has been recently revised: infected people are now classified as having either dengue or severe dengue (WHO 2009). People who recover without major complications are classified as having dengue, whereas those who have any of the following conditions are designated as having severe dengue: plasma leakage resulting in shock, accumulation of serosal fluid sufficient to cause respiratory distress, or both; severe bleeding; and severe organ impairment. The new classification scheme is currently under assessment and several studies are ongoing to verify its utility in the clinical management of dengue cases (Gregory 2010; Alexander 2011; Chitra 2011; Giraldo 2011;Kalayanarooj 2011b; Kulkarni 2011; Narvaez 2011; Srikiatkhachorn 2011; Akbar 2012).

The mechanisms that lead to severe dengue are not completely understood although there are several hypotheses. Risk factors for the development of severe disease include prior infection with a heterotypic serotype (Kliks 1988, Dejnirattisai 2010; Tricou 2011), the strain of the infecting virus (Vaughn 2000; Wang 2003; de Araújo 2009, Maciel-de-Freitas 2011; Costa 2012), age and gender (Anders 2011; Balasubramanian 2012), nutritional status (Thisyakorn 1993; Kalayanarooj 2005), and the person's genetic background (Nguyen 2008; Appanna 2010). The person's age and immunological status are key factors that influence disease progression (Guzman 2010a). The severity of dengue shock syndrome is age-dependent, and vascular leakage is most severe in young children and is thought to be related to the intrinsic integrity of the capillaries (Gamble 2000). Epidemiological studies have shown that severe disease is more common after secondary infections (for example, when an individual who has been already infected by dengue virus is infected again by a different serotype of the virus) rather than during primary infection (Gubler 1997; Graham 1999; Gordon 2013). This might be due to antibody-mediated immune enhancement, whereby non-neutralizing antibodies developed to a previous infection lead to enhanced viral uptake during a new infection by a different serotype (Kliks 1989; Halstead 2003). In particular, infection with DENV-1 followed by infection with DENV-2 or DENV-3, or DENV-3 infection followed by DENV-2, are associated with dengue shock syndrome (Guzman 1990; Guzman 2000; Alvarez 2006).

Although there is no specific antiviral treatment for dengue infection, non-severe dengue infection can be successfully managed with oral fluids and paracetamol. Severe disease can be successfully managed by careful monitoring of warning signs and early initiation of aggressive intravenous rehydration therapy (Nimmannitya 1987). This relies on early diagnosis of dengue infection and close monitoring for clinical signs of plasma leakage. Differentiating dengue from other causes of febrile illness is therefore critical. However, a reliable clinical diagnosis is very difficult to achieve during the early stages of the diseases due to the non-specific symptomatic presentations of dengue infection (Potts 2008).  

Reference standards

Accurate laboratory confirmation of dengue infection currently relies on demonstration of virus presence by the methods described below. All of these methods have limitations and their performance depends on the stage of disease progression when sample collection was performed (Guzmán 2004; Teles 2005; WHO 2009).

Viral isolation

Viral isolation is laborious and requires between seven to 14 days to obtain results. Thus, it is an unsuitable assay for early diagnosis of dengue infection. Cell culture is the most widely used method for viral isolation. Mosquito cell lines of Ae. albopictus C6/36 and Ae. pseudoscutellaris AP61 are the cell lines of choice for routine diagnosis (Race 1978; Tesh 1979; Kuno 1985; Nawa 1987). Virus identification is mostly performed using dengue specific monoclonal antibodies in immunofluorescence (IF) (Henchal 1983; Gubler 1984). The World Health Organization (WHO) recommends that virus isolation samples are collected from suspected dengue cases within the first five days from onset of fever (WHO 2009).

Viral RNA detection by RT-PCR

This technique allows for a faster turnaround time (24 to 48 hours), but is an expensive and often unaffordable option for routine diagnostic purposes in endemic countries (Guzmán 2004). For routine diagnosis of dengue virus infection, the most commonly used nucleic acid amplification tests (NAATs) are based on a single reverse transcriptase polymerase chain reaction (RT-PCR) assay (Morita 1991; Lanciotti 1992), a nested RT-PCR assay (Lanciotti 1992), or a one-step multiplex RT-PCR assay (Harris 1998). The sensitivity of RT-PCR assays compared to virus isolation in mosquito cell culture varies between 25% and 79% (Raengsakulrach 2002). The reported sensitivity is above 90% for several recent NAAT methodologies compared to virus isolation using the C6/36 cell line (Callahan 2001; Houng 2001; Wu 2001; Hapugoda 2010). However, an external quality control assessment of the efficiency and accuracy of dengue NAAT methods applied by expert laboratories, reported low sensitivity of these assays even in laboratories where reliable real-time procedures were used (Domingo 2010). Time of sample collection is crucial for viral genome identification by NAAT and the WHO recommends samples are collected within the first five days from onset of fever (WHO 2009).

In summary, both viral isolation by culture and viral genome identification by NAATs require highly equipped laboratories, well-trained staff, and their use should be restricted to samples collected between five days from fever onset (WHO 2009). The high logistical requirements mean these methods are only suitable for implementation in national laboratories in most of the dengue endemic countries. Moreover, the NAATs often operate on sophisticated and expensive devices and contamination can often undermine their performance in routine clinical settings.

Serological assays based on the detection of IgM and IgG antibodies

Currently these assays are the most widely used laboratory method for diagnosis of dengue infection. The presence of IgM and IgG antibodies in correlation with dengue progression and type of infection (primary versus secondary) has been extensively characterized (Gubler 1984; Innis 1989; Chanama 2004; Wahala 2011). Although less expensive and simpler to perform than viral isolation and RT-PCR, these assays require testing of paired sera (for example, one sample collected during the acute phase and the second collected during the convalescent phase, with the second sample collected ≥ 7 days from onset of fever) to provide a definitive diagnosis (WHO 2009). As results are retrospective, they are unsuitable for making clinical decisions.

IgM antibody-capture enzyme-linked immunosorbent assay (MAC-ELISA) (Innis 1989), E/M-specific capture IgM and IgG ELISA (Burke 1982; Bundo 1985), E/M antigen coated IgM and IgG ELISA and the haemagglutination inhibition test (HI) (Clark 1958) are the most commonly used serological techniques for the routine diagnosis of dengue virus infections (Shu 2004; Guzman 2010a). Traditionally, the HI test was used to detect and differentiate primary and secondary dengue virus infections due to its simplicity and sensitivity. However, this test has recently become less popular and has gradually been replaced by the MAC-ELISA and E/M-specific capture IgM and IgG ELISA (Shu 2004). Over 50 commercial kits are available with variable sensitivity and specificity (Innis 1989; Blacksell 2006; Dussart 2006; Blacksell 2007; Kumarasamy 2007a; Kumarasamy 2007b; Bessoff 2008; Dussart 2008; Lapphra 2008; Hunsperger 2009; McBride 2009; Ramirez 2009; Shu 2009; Zainah 2009; Guzman 2010c; Lima 2010; Osorio 2010; Pok 2010; Tricou 2010; Wang 2010; Blacksell 2011; Chaterij 2011; Ding 2011; Fry 2011; Najioullah 2011; Blacksell 2012b).

Although simpler and less expensive than viral isolation and NAATs, MAC-ELISA (detecting IgM antibodies) has some important limitations. This test does not allow for early diagnosis of dengue infection as during primary infections IgM antibodies are detectable only in 50% of the patients between day 3 and 5 from onset of fever, increasing to 80% by day 5 and 99% by day 10 (PAHO 1995). Therefore, WHO recommend that this test is run on samples collected after five days from onset of fever (WHO 2009). In patients with secondary dengue virus infections the kinetics of IgM production are more variable and IgM levels are significantly lower than during primary infection (Gubler 1996; Chanama 2004). False-negative test results for dengue-specific IgM antibodies can occur during secondary infection (Chanama 2004). Also, some tests show non-specific reactivity in sera from patients with malaria and leptospirosis (Hunsperger 2009).

The IgG ELISA can be used to confirm a dengue infection if paired sera are collected and to distinguish between primary and secondary infection (Vaughn 1999; Guzmán 2004). In primary dengue infection, IgG is generally detectable at low titres at the end of the first week of illness, increasing slowly thereafter, with serum IgG still detectable after several months (and possibly for life). During a secondary infection, high levels of IgG antibodies are detectable during the acute phase. A fourfold or greater increase in IgG antibodies in acute and convalescent paired sera can be used to document recent infections (Shu 2004). However, IgG ELISA generally lacks specificity within the flavivirus serocomplex groups (Makino 1994). Also, IgG antibodies may persist for life which further complicates the serodiagnosis.

A dengue virus E/M protein-specific IgM/IgG ratio can be used to distinguish between primary and secondary dengue virus infections. IgM capture and IgG capture ELISAs are the most common assays for this purpose (Prince 2011). According to this method, a dengue infection is defined as primary if the IgM/IgG optical density (OD) ratio is higher than 1.2 (using patient sera at 1/100 dilution) or 1.4 (using patient sera at 1/20 dilution). The infection is secondary if the ratio is less than 1.2 (using patient sera at 1/100 dilution) or 1.4 (using patient sera at 1/20 dilution). (Kuno 1991; Shu 2003; Shu 2004). However, ratios may vary between laboratories, indicating the need for better standardization of test performance (Falconar 2006).

Based on WHO recommendations (WHO 2009), dengue virus diagnosis can be considered confirmed when at least one of the following diagnostic methods gives a positive result:

  • Viral isolation by culture;

  • Identification of viral RNA by RT-PCR;

  • IgM seroconversion (from negative to positive) in paired sera ;

  • IgG seroconversion (from negative to positive) in paired sera or at least a fourfold increase of IgG titer in paired sera

However, a thorough diagnostic work-up for dengue infection should include both virus isolation procedures (by virus identification or genome detection) and detection of immune responses (WHO 2009).

Index test(s)

Several studies have suggested that detecting dengue non-structural 1 (NS1) protein could be a useful method for early diagnosis of dengue (Young 2000; Huang 2001; Alcon 2002; Koraka 2003; Xu 2006). NS1 is a 46-kDa non-structural glycoprotein that is highly conserved among flaviviruses, including dengue, Japanese encephalitis, yellow fever and tick-borne encephalitis (Winkler 1989; Alcon-LePoder 2006; Chuansumrit 2011). Dengue NS1 antigen has been detected in acute-phase sera of infected individuals (Young 2000; Alcon 2002; Hu 2011) as early as the first day after the onset of fever in 80% (Alcon 2002) to 87.5% (Hu 2011) of samples. NS1 antigen remains detectable in convalescent sera up to 14 days after onset of fever (Alcon 2002; Hu 2011). Although positivity rates of samples collected between day 8 and 14 decrease steadily (Hu 2011), overall NS1 antigens circulating in blood of infected patients are detectable for a longer period than viral RNA (Alcon 2002; Kumarasamy 2007a; Huhtamo 2010; Singh 2010). It is unclear to what extent NS1 antigen may be a useful diagnostic marker for secondary infection (Duong 2011). Some studies and reports suggest no significant difference in NS1 detection in sera from patients with primary and secondary infections (Alcon 2002; Lapphra 2008; Lima 2010). Other studies reported a significant decrease in the sensitivity of NS1 antigen-based tests in secondary infections (Kumarasamy 2007a; Hang 2009; Osorio 2010; Singh 2010). Given the higher level of anti-NS1 antibodies that characterizes secondary infection, the lower sensitivity of NS1 antigen-based test could be due to the high proportion of NS1 antigen bound to immunocomplexes and the consequent lower concentration of freely circulating antigens in the blood of people infected with dengue (Koraka 2003).

Several commercial kits detecting NS1 antigen are available. These tests use blood samples and include test formats based on lateral flow immunoassays (LFIs), often also called rapid diagnostic tests (RDTs), or enzyme-linked immunosorbent assay (ELISA).

LFIs are simple to perform and require little laboratory infrastructure. Therefore, they are suitable for use in resource-limited settings by lesser-trained health workers or laboratory staff. However, LFIs must be stored in a dry environment and at cool temperatures, possibly also requiring use of cold chain (2°C to 8°C).

ELISA tests are more complex to perform, have a longer turn-around time to result (approximately two to three hours), and require a minimum of laboratory infrastructure and training of personnel. Despite these limitations, they represent a valuable alternative when LFIs are not available. 

Clinical pathway

To significantly advance the clinical management of dengue cases, a diagnostic test should ideally enable rapid identification of dengue infection during the early stages of the disease (turnaround time < one hour) and be simple enough to use that it can be implemented at peripheral health facilities.

Prior test(s)

LFIs can be rapid and provide results within 30 minutes, and ELISA-based tests within a day. Serological assays based on the detection of IgM and IgG antibodies are the most widely used laboratory method for diagnosis of dengue infection but they are not suited to guide timely clinical decisions on most appropriate patient management. IgM and IgG based tests require testing of paired sera (for example, one sample collected during the acute phase and the second collected during the convalescent phase) in order to provide a definitive diagnosis (WHO 2009). Thus, when using those tests, a laboratory confirmed diagnosis can only be established retrospectively.

The other assays that have a role in the diagnostic work-up for dengue fever are viral isolation by culture and viral RNA isolation by RT-PCR. Viral isolation is laborious and the laboratory turnaround time to results is seven to 14 days. Viral RNA detection by RT-PCR allows for a faster turnaround time (24 to 48 hours) but in endemic countries is an expensive and often unaffordable option for routine diagnostic purposes. While both these techniques are in principle useful for early diagnosis of dengue virus infection, the turnaround time to result to the clinics or primary health centre (where the majority of patients are seen) is lengthy due to cumbersome laboratory procedures (in particular for viral detection by culture) and because these assays are suitable for implementation only in national reference laboratories in most endemic countries.

Role of index test(s)

NS1 antigen-based tests can detect a dengue infection during the acute phase of the disease (day one to seven) and do not require paired sera. In addition, the lateral flow format of the assay is suitable for implementation in decentralized settings. Thus, these tests have great potential to provide timely diagnosis of dengue viral infection and significantly improve patient management. High sensitivity and high specificity are both desirable to support individual patient management, particularly in resource-limited settings where prompt referral of suspected false positive or false negative results for further diagnostic work-up is often inadequate (Peeling 2010; Blacksell 2012a). While a very high sensitivity represents an essential characteristic for an ideal RDT, diagnostic tests with a lower specificity (75% to 95%) could be of added value provided they support differential diagnosis of diseases with very similar clinical symptoms (for example, malaria, leptospirosis, typhoid, typhus, and chikungunya) (Peeling 2010).

Alternative test(s)

Detection of IgM or IgG antibodies, or both antibodies, is most commonly used for confirmation of probable dengue infection. However, these serodiagnostic tests have considerable limitations, as described above. More accurate methods such as virus isolation are not widely available in high disease burden settings.

Rationale

Early diagnosis of dengue infection is critical for differential diagnosis of patients presenting with undifferentiated fever who are living in, or returning from, endemic areas. Some patients progress rapidly from infection to severe disease (dengue haemorrhagic fever with the risk of developing dengue shock syndrome). Although there is no specific treatment for dengue, severe disease can be successfully managed by careful monitoring and early initiation of rehydration therapy.

LFIs are both simple and rapid to perform, facilitating correct diagnosis and proper follow-up. ELISA tests have the advantage of allowing the rapid processing of multiple samples in settings where a laboratory with adequate equipment, power supply and skilled laboratory staff are available.

According to the WHO, NS1 antigen-detecting LFIs should be made available at the primary care level and NS1 antigen-detecting ELISAs at the district level (WHO 2009). However, the benefit of an NS1 antigen-detecting test is uncertain. Sensitivity and specificity vary with endemicity. In evaluations carried out to date, study authors have mainly focused on either one (Dussart 2006; Kumarasamy 2006; Kumarasamy 2007a; Blacksell 2008; Chuansumrit 2008; Lapphra 2008; Phuong 2009; Shu 2009; Zainah 2009; Bessoff 2010; Huhtamo 2010; Wang 2010; Chaterij 2011; Watthanaworawit 2011) or two tests (Bessoff 2008; Chaiyaratana 2009; Hang 2009; McBride 2009; Ramirez 2009; Guzman 2010a; Tricou 2010; Najioullah 2011), and on limited geographic areas (Blacksell 2008; Hang 2009; Lima 2010; Osorio 2010). In several studies, they performed evaluations on samples from returning travellers, thus local endemicity of other diseases or immunity did not play a role in test performance (Lindegren 2005; Shu 2009; Huhtamo 2010).

Objectives

To assess the diagnostic accuracy of commercial NS1 antigen detecting assays for acute dengue infection.

Secondary objectives

To compare the diagnostic accuracy of LFIs and ELISAs antigen detecting assays.

To explore the following potential sources of heterogeneity: degree of endemicity, continent, patient age, primary versus secondary dengue infection, serotype, type of participant (residents or traveller), and timing of sample collection from onset of symptoms.

Methods

Criteria for considering studies for this review

Types of studies

We will include: 

  • Paired comparative studies in which participants receive more than one index test and a reference standard;

  • Cohort studies in which a series of patients are recruited and receive one or more index test and a reference standard;

  • Case-control studies comparing a group of patients with laboratory confirmed dengue infection (positive reference standard) and a control group of patients without dengue infection (negative reference standard).

Participants

Patients presenting with clinical suspicion of acute dengue infection, or specimen bank samples from patients with suspicion of acute dengue infection, who either returned from or live in an endemic area. Based on available evidence on the presence of detectable NS-1 circulating antigens in the blood of infected patients, samples should be collected < 10 days after onset of symptoms.

Index tests

NS1 antigen-detecting assays, irrespective of format (ELISA or LFI) applied on human blood samples.

There are three LFIs commercially available:

  • Dengue NS1 Ag STRIP: Bio-Rad Laboratories, Marnes LaCoquette, France (catalogue number: 70700);

  • SD BIOLINE Dengue NS1: Standard Diagnostics Inc., Kyonggi-do, South Korea (catalogue number: 11FK50).

Note: SD BIOLINE Dengue Duo Combo (Standard Diagnostics Inc., Kyonggi-do, South Korea) for NS1, IgM and IgG simultaneously with catalogue number: 11FK45 (10 tests) or 11FK46 (25 tests) is the same test as SD BIOLINE Dengue NS1 catalogue number: 11FK50;

  • Dengue Early Rapid: Panbio®, South Korea (catalogue number: R-DEN01P).

and three ELISA based tests:

  • Dengue NS1 Ag ELISA: Standard Dignostic Inc., Kyonggi-do, South Korea (catalogue number:11EK50);

  • Platelia Dengue NS1 Ag test: Bio-Rad Laboratories, Marnes LaCoquette, France (catalogue number:72830);

  • Dengue Early ELISA: Panbio®, South Korea. Two versions of this test are reported in the literature: the first generation Dengue Early ELISA, with catalogue numberE-DEN01P (version discontinued on 28.05.2009); and the second generation Dengue Early ELISA, with catalogue number E-DEN02P. This latter test is now produced by Standard Diagnostic and has a catalogue number 1PE40. E-DEN02P and 1PE40 are exactly the same test. The E-DEN02P test differs from the first generation E-DEN01P only for some minor sample preparation steps. Studies assessing the performance of the two versions of the Dengue Early ELISA test will be included.

Target conditions

Acute dengue infection, defined as the phase of disease during which the virus is present in the blood (WHO 2009), a period lasting four to five days after the onset of illness. According to current WHO guidelines the acute febrile phase usually lasts for seven days and is often accompanied by facial flushing, skin erythema, generalized body ache, myalgia, arthralgia and headache.

Reference standards

A composite reference standard is usually used, based on a combination of the following diagnostic approaches (Peeling 2010):

  1. Virus isolation (culture);

  2. Virus genome detection (RT-PCR amplification of viral RNA);

  3. Serological response

    1. IgM seroconversion (from negative to positive) in paired samples (collected ≥ 3 days apart, with the last sample collected ≥ 7 days after illness onset (Guzman 2010c) or two-fold increase in IgM titre (15 U to 30 U) in paired samples accordingly to Armed Forces Research Institute of Medical Sciences (AFRIMS) algorithm (Pan-ngum 2013)

    2. IgG seroconversion (from negative to positive) or a ≥ fourfold increase of IgG titre in paired samples (collected ≥ 3 days apart, with the last sample collected ≥ 7 days after illness onset (Guzman 2010c)

    3. IgM (DEN MAC ELISA) < 40 U on the admission sample and ≥ two-fold increase in IgG titre (DEN GAG ELISA) on paired sera, with the a titer ≥100 U in the convalescent sample, accordingly to the AFRIMS algorithm for dengue infection (Pan-ngum 2013).

There are two grades of reference standard: we will define a Grade 1 reference standard as a study using all three methods as reference standard; and a Grade 2 reference standard as a study using any two of the above methods as reference standard. We will consider patients positive by any of the methods as positive for dengue virus infection.

Search methods for identification of studies

Electronic searches

We will attempt to identify all relevant articles by performing a systematic search of articles from 1990, the first time dengue antigens were identified in infected people, onwards.

We will search the following databases using the search terms and strategy identified in Table 1: Cochrane Infectious Disease Specialized Register; MEDLINE via PubMed; EMBASE via Dialog Datastar; MEDION (http://www.mediondatabase.nl/); Science Citation Index; Conference Proceedings Citation Index-Science (CPCI-S); Database of Systematic Reviews and Meta-Analyses in Laboratory Medicine (http://www.ifcc.org/); the Cochrane Central Register of Controlled Trials (CENTRAL) published in The Cochrane Library; LILACS; IndMED; and African Index Medicus.

Table 1. Search strategy
Search set MEDLINE (OvidSP) EMBASE (OvidSP)
1exp Dengue/dengue/
2Dengue Virus/exp Dengue virus/
3(dengue or dengues or DENV* or antidengue* or “break bone fever” or “breakbone fever” or DHF or DSS).tw,ot.(dengue or dengues or DENV or antidengue* or “break bone fever” or “breakbone fever” * or DHF or DSS).tw,ot.
4or/1-3or/1-3
5Viral Nonstructural Proteins/immunoassay/ or double antibody sandwich enzyme linked immunosorbent assay/ or enzyme immunoassay/ or enzyme linked immunosorbent assay/ or enzyme linked immunospot assay/ or enzyme multiplied immunoassay technique/ or microparticle enzyme immunoassay/
6Antigens, Viral/virus protein/ or nonstructural protein 1/
7"NS1*".tw,ot,nm.virus antigen/
8"NS-1*".tw,ot,nm."NS1*".tw,ot.
9((nonstructural or "non structural") adj3 (protein* or glycoprotein* or antigen*)).tw,ot."NS-1*".tw,ot.
10((lateral flow or immunosorbent or immunospecif*) adj3 assay*).tw,ot.((nonstructural or "non structural") adj3 (protein* or glycoprotein* or antigen*)).tw,ot.
11(sandwich* adj5 assay*).tw,ot.((lateral flow or immunosorbent or immunospecif*) adj3 assay*).tw,ot.
12ELISA or ELISAs or EIA or EIAs.tw,ot.(sandwich* adj5 assay*).tw,ot.
13Immunoassay/ELISA or ELISAs or EIA or EIAs.tw,ot.
14exp Immunoenzyme techniques/immunoaffinity chromatography/
15Immunosorbent techniques/(immunoassay* or immuno-assay* or immuno-chromatograph* or immuno-chromatograph*).tw,ot.
16Immunochromatography/Assay* ti, ab((rapid or “point of care” or “near patient” or poc or poct or bedside) adj5 (test* or diagnos* or serodiagnos* or assay* or kit or kits or detect* or capture or classif* or identif*)).tw,ot.
17

(immunoassay* or immuno-assay* or immuno-chromatograph*

or immunochromatograph*).tw,ot.

(RDT or RDTs or RADT or RADTs).tw,ot.
18Point-of-Care Systems/(antigen* adj5 (capture or detect*)).tw,ot.
19

((rapid or “point of care” or “near patient” or poc or poct or bedside) adj5

(test* or diagnos* or serodiagnos* or assay* or kit or kits or detect* or capture or classif* or identif*)).tw,ot.

diagnostic kit/
20RDT or RDTs or RADT or RADTs.tw,ot.test strip/
21(antigen* adj5 (capture or detect*)).tw,ot.or/5-20
22exp Reagent Kits, Diagnostic/4 and 21
23or/5-22((dengue adj3 strip*) or Bioline Dengue or Platelia Dengue or Platelia DENV or Dengue Early Rapid or Dengue Early Capture or Panbio Dengue or Panbio Early ELISA or Pan-E Dengue or Pan-E Early ELISA or Platelia NS1 or NS1 Ag Strip or R-DEN01P or E-DEN02P or E-DEN01P).tw,ot.
244 and 2322 or 23
25

((dengue adj3 strip*) or Bioline Dengue or Platelia Dengue or Platelia DENV

or Dengue Early Rapid or Dengue Early Capture or Panbio Dengue or Panbio Early ELISA or

Pan-E Dengue or Pan-E Early ELISA or Platelia NS1 or NS1 Ag Strip or R-DEN01P or

E-DEN02P or E-DEN01P).tw,ot.

limit 24 to yr="1990 -Current"
2624 or 25

 

(Animal/ or Nonhuman/ or Animal Experiment/) and Human/

27limit 26 to yr=”1990 –Current”Animal/ or Nonhuman/ or Animal Experiment/
28exp animals/ not humans/27 not 26
2927 not 2825 not 28

Searching other resources

We will handsearch reference lists of included articles, and any relevant review articles we identify through the search, for eligible articles. We will contact test manufacturers to identify any unpublished studies and search reports of the WHO for additional studies. In addition we will search the WHO International Clinical Trials Registry Platform (ICTRP), Clinical Trials.gov, Current Controlled Trials metaRegister and manufacturers' websites. If required, we will contact authors of relevant articles in order to obtain missing data.

Data collection and analysis

Selection of studies

We will retrieve the full text articles of all potentially relevant studies and two authors, Martina Casenghi (MC) and Cara Kosack (CSK), will independently assess each article for inclusion. We will resolve any discrepancies by discussion. If no agreement can be reached, we will seek the opinion of a third author, Nathan Ford (NF).

Data extraction and management

Two authors (MC and CSK) will independently extract a standard set of data from each study article using a pre-defined data extraction form (see Table 2). We will resolve any discrepancies through discussion, and if necessary we will consult a third author, NF. Where multiple index tests or reference standards are applied, we will extract and present data for each test. In studies where only a subgroup of participants is included in the review, we will only extract and present data for that particular subgroup. We will extract the number of non-interpretable or invalid test results.

Table 2. Data extraction
Study IDFirst author, year of publication
Clinical featuresPresenting signs and symptoms
Duration of reported fever in days at time of sample collection
Index of suspicion: suspected dengue cases versus unselected febrile patients
Alternative diagnosis for dengue negative cases
SettingCountry and continent
Endemicity
Endemic serotypes
ParticipantsSample size
Age
Gender
Comorbidities including pregnancy
Study designSampling methods: consecutive versus random
Study type (such as case-control or cohort)
Type of enrolment: prospective versus retrospective
Allocation of index tests to patient if more than one NS1 test was enrolled in the study. Did all individuals receive all index tests?
Target conditionAcute dengue fever
Type of infection (primary versus secondary)
Reference standardType of reference standard applied (Grade 1 versus Grade 2)
Level of education of personnel who performed the reference standard test
Location where reference standard test was performed
Number of observers
Method of inter-observer discrepancy resolution
Number of repeated test
Blinding of the operator(s) to index test results
Establishment of laboratory quality control system
Index testsCommercial name
Format of the test (ELISA versus LFI)
Level of education of personnel who performed the test
Location where test was performed
Number of observers
Method of inter-observer discrepancy resolution
Number of repeated tests
Blinding of the operator to reference standard test(s) results
SampleFresh versus stored samples
Sample transport
Storage conditions of samples
Sample type (serum, plasma or whole blood)
DataNumber of true positive, true negative, false positive, false negative
Number of unclear or invalid results
Note(s)Source of funding
Other relevant information

For each comparison of index test with reference test(s), we will extract data for the index test on the number of true positives, true negatives, false positives and false negatives. We will present results of LFIs and virus isolation as binary figures (positive or negative). We will consider ELISA results positive at any level of NS1 antigen detected. If study authors present PCR results and IgM and IgG levels as continuous variables, we will convert to binary results using the cut-off points given by manufacturers if presented by the study authors.

Assessment of methodological quality

Two authors (MC and CSK) will independently assess the quality of each individual study using the QUADAS-2 tool (Whiting 2011) (see Table 3). Each quality indicator will be answered with a 'yes', 'no', or 'unclear' if insufficient information is available. QUADAS-2 evaluates four domains: patient selection, index tests, reference standard, and time and flow. We will give the reasons for the judgments we make.

Table 3. QUADAS 2
Item Judgement

Risk of bias (RoB) and

concerns of applicability (CoA) - (high/low/unclear)

Domain 1 - Patient selection

Describe methods of patient selection.

Describe included patients (previous testing, presentation, intended use of index test, and setting).

1. Was a consecutive or random sample of patients enrolled? 'Yes' if a consecutive or random sample of patients was enrolled;
'No' if a selected group of patients was enrolled;
'Unclear' if there is insufficient information on enrolment.
-
2. Was a case-control design avoided?

'Yes', if no case-control design was used;

'No', if a case-control design was used;

'Unclear', if there is insufficient information on the study design.

-
3. Did the study avoid inappropriate exclusions? 'Yes' if there were no inappropriate exclusions;
'No' if there were inappropriate exclusions;
'Unclear' if there is insufficient information on exclusions.
-
Could the selection of patients have introduced bias (RoB)? 
4. Was the spectrum of included patients representative of the patients who will receive the test in practice?

'Yes' if patients had clinically suspected acute dengue infection or presented with acute febrile illness and samples were < 10 days from onset of symptoms; and if these patients were consecutively recruited or random samples were taken from consecutive series;

'No' if patients did not have clinically suspected acute dengue infection or did not present acute febrile illness and samples were collected ≥ 10 days after onset of symptoms; or if patient groups without dengue infection were recruited separately (such as healthy controls) or if patients were recruited from non-endemic areas;

'Unclear' if patients' characteristics, time of sample collection and sampling methodology were not adequately described.

-
Are there concerns that the included patients and setting do not match the review question (CoA)? 

Domain 2 - Index test

Describe the index test and how it was conducted and interpreted.

1. Were the index test results interpreted without knowledge of the results of the reference standard?

'Yes' if the operator performing and interpreting  the index test did not know the results of the reference test(s);

'No' If the operator performing the index test knew the results of the reference test(s);

'Unclear' If it is unclear whether blinding of reference standard results took place.

-
2. If a threshold was used, was it pre-specified?

'Yes' if a threshold was pre-specified (only applicable for ELISA tests);

'No' if a threshold was not pre-specified (only applicable for ELISA tests);

'Unclear' if it is unclear whether a threshold was pre-specified (only applicable for ELISA tests).

-
Could the conduct or interpretation of the index test have introduced bias (RoB)? 
Are there concerns that the index test, its conduct, or interpretation differ from the review question (CoA)? 

Domain 3 - Reference standard

Describe the reference standard and how it was conducted and interpreted.

1. Were the reference standards likely to correctly classify the target condition?

'Yes' If an appropriate reference standard was used (Grade 1 or 2);

'No' If the reference standards used do not correspond to the definition of Grade 1 or Grade 2 reference standard;

'Unclear' if not enough information are available on the reference standard used.

-
2. Were the reference standard results interpreted without knowledge of the results of the index tests?

'Yes' If the operator performing and interpreting the reference tests does not know the results of the index test;

'No' If the operator performing and interpreting the reference tests knows the results of the index test;

'Unclear' If it is unclear whether blinding of index test results took place.

-
Could the reference standard, its conduct, or its interpretation have introduced bias (RoB)? 
Are there concerns that the target condition as defined by the reference standard does not match the question (CoA)? 

Domain 4 - Flow and timing

Describe any patients who did not receive the index tests or reference standard or who were excluded from the 2 x 2 table (refer to flow diagram).

Describe the interval and any interventions between index tests and the reference standard.

1. Was there an appropriate interval between index test and reference standard?

'Yes' if the index test and the reference standard were performed on the same acute-phase sample (fresh or stored) or samples were collected within 24 hours. However, index and references test samples must both have been collected during the acute phase of illness (except paired samples for IgM/IgG serology);

'No' If index test and reference standard tests were performed on samples taken more than 24 hours apart;

'Unclear' If information on timing of collection of sample for index and reference test were not provided.

-
2. Did all patients receive the same reference standard?

'Yes' if the same reference standard was used in all patients;

'No' if the choice of the reference standard varied between individuals;

'Unclear' if it is unclear whether different reference standards were used.

-
3. Were all patients included in the analysis?

'Yes' if all patients were included in the analysis;

'No' if not all patients were included in the analysis;

'Unclear' if it is unclear whether all patients were included in the analysis.

-
4. Were uninterpretable results reported?

'Yes' if the number of participants in the 2 x 2 table matched the number of participants recruited into the study or if sufficient explanation is provided for discrepancy (including reporting of uninterpretable results);

'No' if the number of participants in the 2 x 2 table did not match the number of participants recruited into the study, and insufficient explanation is provided for any discrepancy;

'Unclear' if it is not possible to work out whether uninterpretable results occurred.

-
Could the patient flow have introduced bias (RoB)? 

Statistical analysis and data synthesis

We will summarize data from each study in 2 x 2 tables of true positives, true negatives, false positives, and false negatives. We will exclude studies that provide insufficient data to construct 2 x 2 tables. In the description of studies we will report the number of non-interpretable or invalid tests results.

In the primary analysis, we will group studies by study design, reference test grade, and index test type (LFIs or ELISAs), and if there are enough studies, by the commercial name of index tests. Data from the same study may contribute to different comparisons (for example, LFI or ELISA versus Grade 1 reference standard; LFI or ELISA versus Grade 2 reference standard). To demonstrate the variation in accuracy between studies, we will plot estimates of sensitivities and specificities in forest plots and receiver operating characteristics (ROC) space. We will analyse the data from case-control studies separately.

We will perform meta-analyses by fitting hierarchical summary ROC (HSROC) models (Rutter 2001) using the NLMIXED procedure in SAS (SAS Institute Inc., Cary, North Carolina, USA). The HSROC model includes random-effects to allow for variation in accuracy and threshold between studies, and also a shape parameter that allows for dependency between accuracy and threshold. If the number of studies available is too few to estimate all parameters, we will simplify the HSROC model by assuming a symmetrical shape to the summary ROC curve or removing one or both variance parameters for the random effects, or both, if there is no indication of heterogeneity in either the accuracy or threshold parameters. If studies report common thresholds, we will derive summary sensitivities and specificities from the parameters of the HSROC model, and present SROC plots with operating points and 95% confidence regions.

We will compare the test accuracy of commercial brand names within each index test type before making comparisons between index test types. Where more than one commercial test of the same index test type (LFI or ELISA) is tested on the same patients against the same reference standard, we will select one commercial brand at random for the analysis comparing test types, to avoid bias due to inclusion of the same participants more than once in the analysis. Our analyses will first include all studies with relevant data, and will then be restricted to studies that made direct comparisons between tests with the same participants. Where adequate data are available, we will undertake meta-analyses to estimate and compare the performance of the tests. We will make comparisons between index test types (or commercial name) by adding a covariate for index test type (or commercial name) to the HSROC model to investigate the effect of test type (pr commercial name) on accuracy, threshold or shape of the SROC curve, and threshold and shape of the SROC curve. Also, we will assess the statistical significance of the differences using likelihood ratio tests to compare models with and without the covariate terms.

Investigations of heterogeneity

We will assess heterogeneity by visual inspection of study results plotted in ROC space and in the forest plots. Also, we will explore potential sources of heterogeneity by including covariate terms in the HSROC model. We will examine the following

potential sources of heterogeneity: degree of endemicity (low, medium, high), continent (Africa, Asia, South America), patient age (adults, children, mixed), primary versus secondary dengue infection, serotype (1, 2, 3, or 4), participants (returning travellers versus people living in endemic area), and timing of sample collection (≤ seven days from onset of symptoms versus > seven days from onset of symptoms).

Sensitivity analyses

If sufficient evaluations are available, we will assess the robustness of the meta-analyses by conducting sensitivity analyses using components of the quality assessment. In particular, we will examine those relating to the proportion of participants not included in the analysis for any reason, including non-interpretable results.

Assessment of reporting bias

We will not attempt to assess reporting bias.

Acknowledgements

We are grateful to our affiliated institutions and organizations. We acknowledge the referees for their comments. The editorial base for the Cochrane Infectious Diseases Group is funded by the UK Department for International Development (DFID) for the benefit of developing countries.

Contributions of authors

MC and CK wrote the study protocol and conducted the literature research. RL conducted the literature research and performed data extraction. MB supported the writing of the data extraction, statistical analysis and data synthesis. NF was involved in the development and concept of the review and commented critically on the study protocol.

Declarations of interest

The authors have no conflicts of interest and have received no financial support for the development of this protocol.

Sources of support

Internal sources

  • Médecins sans Frontières, Amsterdam, Netherlands.

  • Médecins sans Frontières, Geneva, Switzerland.

  • World Health Organization, Geneva, Switzerland.

  • University of Pennsylvania, Philadelphia, USA.

External sources

  • Cochrane Collaboration, UK.

Ancillary