Diagnosis of dengue: an update

Abstract

Early diagnosis of dengue, the most common mosquito-borne disease globally, remains challenging. Dengue presents initially as undifferentiated fever, with symptoms becoming more pathognomonic in the later stages of illness. This limits the timeliness in the delivery of appropriate supportive interventions. Laboratory tests are useful for diagnosis although the short-lived viremia and the presence of secondary infection with one of the four heterologous viral serotypes collectively complicate the choice and interpretation of laboratory tests. In this article, the authors review the various approaches for diagnosis of dengue and discuss the appropriate tests to use, including when a dengue vaccine, which is in the late stages of development, is licensed for use. The ensuing reduced dengue prevalence could make diagnosis for vaccine efficacy and escape-mutant monitoring even more challenging.

Keywords::

• dengue
• diagnostics
• early diagnosis
• surveillance
• vaccine

Figure 1. Approximate window of detection for dengue diagnostics.

NS1: Non-structural protein 1.

Data taken from [1].

Figure 2. Positive- and negative- predictive values of the various diagnostic approaches for dengue at different rates of prevalence.

Results were generated from median values of sensitivity and specificity presented in Table 2 for PCR (conventional and real time), IgM (ELISA), IgM (Rapid) and NS1 detection, respectively. Aspecificity of 96% was assumed for conventional PCR and real-time PCR as most studies have limited population sizes. Diagnosis using the WHO 2009 classification was used to represent clinical diagnosis. Data for WHO criteria were obtained from Low etal. [12] and separated into two age groups (<56 and ≥56 years). NPV: Negative-predictive value; NS1: Non-structural protein 1; PPV: Positive-predictive value.

Dengue is endemic throughout the tropical world. The WHO has estimated that approximately 3 billion people live at risk of infection each year. Infection produces a spectrum of clinical manifestation, from mild influenza-like illness to dengue fever (DF) or severe dengue illness. The latter comprise of either plasma leakage, which leads to hypovolemic shock or dengue shock syndrome and internal hemorrhage, or other organ failure, including encephalopathy [1]. The disease is caused by dengue virus (DENV), which belongs to the genus Flavivirus of the Flaviviridae family [2]. DENV is a positive-sense, single-stranded RNA virus (approximately 11 kb in length). The genome is transcribed as a single open reading frame encoding three structural (C, prM and E) and seven nonstructural (NS1, NS2A, NS2B, NS3, NS4A, NS4B and NS5) proteins [3] that are subsequently cleaved into individual components by proteolytic cleavage [4]. The untranslated terminal regions at both ends of the viral genome (3′ and 5′ untranslated terminal regions) are important in the regulation of translation and replication of the viral genome [5].

DENV is composed of four antigenically distinct serotypes. Infection by a specific serotype confers lifelong immunity against the specific serotype but not to the remaining three, although transient crossprotection has been observed within 2–3 months following acute dengue infection [6]. Secondary infection carries a higher risk of plasma leakage that, if not appropriately supported clinically with fluid management, can lead to shock [7]. This association between secondary infection and severe dengue is thought to be mediated by the binding of crossreactive, neutralizing antibodies at subneutralizing concentration that enhances the infection of monocytes and dendritic cells (DCs) via the Fc receptors, a process termed antibody-dependent enhancement (ADE) [8–10]. Besides ADE, other factors that could influence severe clinical outcome in a dengue infection include both human host [11–14] and viral factors [15–18].

DENV is transmitted to humans primarily by infected Aedes aegypti – the predominant epidemic vector – while Aedes albopictus and Aedes polynesiensis have also caused dengue outbreaks [19–21]. Current methods of disease prevention rely on reducing the vector population density. However, given the experience of countries such as Singapore, where periodic epidemics of dengue continue despite concerted public health efforts to control the vector population [22], a cost-effective vaccine remains the most viable option for dengue prevention.

The development of a dengue vaccine has been complicated by the concern that subneutralizing levels of antibodies may paradoxically increase the risk of severe dengue in the form of dengue hemorrhagic fever (DHF) and dengue shock syndrome through ADE [8–10]. Hence, while a dengue vaccine was initially advocated in the 1940s [23], it was not until 1971 that the feasibility of a dengue vaccine in preventing DHF was studied [10,24]; and based on initial data [25–29], a vaccine search was initiated, endorsed and discussed by the SEARO/WHO Research Study Group and experts in the field [24]. Despite initial optimism [30,31], more than three decades have passed without a licensed dengue vaccine in the market. However, current developments are promising and six tetravalent candidate vaccines are in Phase I–III trials. Optimistically, a licensed vaccine can be anticipated in the next 5–7 years [32–34].

Meanwhile, apart from vector control, the burden of dengue on society can also be reduced through appropriate and timely clinical interventions to prevent severe morbidity and mortality. This relies on early and accurate diagnosis of dengue. Even when a vaccine or an antiviral drug becomes available, the need for accurate diagnosis would not be diminished. Instead, the requirement for accurate diagnosis could become more demanding, as surveillance for dengue in vaccinated individuals would be needed to determine vaccine efficacy and for the early detection of vaccine-escape mutants. The goal of this review is thus to determine the state of the art in diagnosis of dengue and identify areas where improvements through research are needed to prepare for the quality of diagnostics needed in a postvaccination world.

Current status in diagnosis of dengue

Clinical diagnosis

Diagnosis of dengue starts with a clinical suspicion, prompted by the recognition of a collection of presenting symptoms and signs. In the early acute febrile phase of illness, dengue patients often present with a history of sudden onset fever, which is often accompanied by nausea, aches and pains. Unfortunately, these symptoms are not unique to dengue and are reported with other febrile illnesses (OFI). The onset of a maculopapular rash, retro-orbital pain, petechiae or bleeding nose or gums are more pathognomonic of dengue and would more probably trigger a differential diagnosis of dengue, although these symptoms usually appear in the later stages of illness, nearer the phase of fever defervescence, when plasma leakage occurs [1]. Their usefulness for early diagnosis would thus be more limited. A list of the commonly reported symptoms is shown in Table 1.

As each of the individual symptoms cannot accurately differentiate dengue from OFIs, an alternative approach to clinical diagnosis is to use a permutation of a list of symptoms or signs. The WHO guidelines for dengue are such examples [1,35]. Indeed, when applied to prospectively recruited patients who presented with acute febrile illness less than 72 h from fever onset, both the 1997 and 2009 WHO guidelines showed a similar sensitivity of over 95% in young adults, albeit with poor specificities of less than 40% [12]. Consequently, the WHO classification schemes can be useful to trigger a suspicion of dengue. During epidemics, when the prevalence of dengue is high, cases that fit these definitions could be treated as presumptive dengue while awaiting other test results. However, they cannot be used for a confirmatory diagnosis of dengue. Furthermore, caution must be exercised in places where dengue infection occurs in older adults. The same study showed that adults who are 56 years of age and older had greatly reduced sensitivities, from over 95 to 73.7% and 81.6% for the 1997 and 2009 WHO classification schemes, respectively. In such cases, the study suggested that leukopenia in patients with febrile illness should trigger a suspicion of dengue [12].

Besides the WHO classification schemes, others have attempted to develop diagnostic algorithms for dengue. Tanner etal. used a data mining approach to identify a group of symptoms and hematologic measurements to differentiate dengue from OFIs [36]. The resultant algorithm had a sensitivity and specificity of 71.2 and 90.1%, respectively. Another comprehensive multivariable logistic regression model was also developed and validated for distinguishing DHF from DF, DHF from DF or OFIs, dengue from OFIs and severe dengue from nonsevere dengue. The model was found to have a sensitivity that ranged from 89.2 (dengue from OFIs) to 79.6% (DHF from DF) [37]. This model also provides a tool for probability calculation and classification of patients based on readily available clinical and laboratory data. However, the usefulness of such algorithms remains to be tested in different populations with different circulating DENV strains.

An important limitation in the development of useful clinical approaches to diagnosis of dengue is the lack of standardization with regard to study design, diagnostic criteria and data collection [38]. While this is not surprising as these studies were performed by various laboratories in different countries, it limits the development of diagnostic classification or algorithms that can be applied internationally. Indeed, the need for more prospective studies to construct a valid and generalizable algorithm to guide the differential diagnosis of dengue in endemic countries remains urgent [38].

Laboratory diagnosis

As clinical diagnosis lacks specificity, a definitive diagnosis of dengue infection requires laboratory confirmation. A number of different laboratory tools are available for diagnostic use. A summary of laboratory diagnostic methods used in dengue infection is shown in Table 2 and the approximate time from illness onset at which these diagnostic tests should be used is shown in Figure 1.

Virus isolation

Dengue viremia can be detected from 2 to 3 days prior to the onset of fever to up to 5.1 and 4.4 days after the onset of the disease for primary and secondary infection, respectively [39]. During this viremic period, blood, serum or plasma samples can be used for virus isolation.

Mosquito inoculation remains the most sensitive method for virus isolation. The isolation rate of the four serotypes of DENV is in the range of 71.5–84.2% [40]. Various mosquito species have been found to be useful and sensitive in dengue isolation, including A. aegypti, A. albopictus and Toxorhynchites splendens, where both male and female mosquitoes are susceptible [6,41–44]. These mosquitoes are inoculated intrathoracically with serum or plasma specimens [40–44]. Specimens collected early in the course of illness have a greater isolation rate (85.3% before day 4 of illness) than those collected later (65.4% after day 4 of illness) [40]. Furthermore, the isolation rate in patients with primary dengue (91.0%) was higher than those with secondary dengue (77.6%). This observation could be due to the interference of crossreactive antibodies with virus isolation or a faster rate of viral clearance in patients with secondary DENV infection [39]. Either explanation, however, suggests that the prevalence of primary or secondary DENV infection may influence the overall virus isolation [40,45].

Mouse brain inoculation has also been used to isolate and amplify DENV. Generally, unweaned mice (2–4 days of age) are inoculated intracerebrally with serum or plasma specimens and observed daily. Moribund mice are then sacrificed to harvest the isolate [46,47].

Both mosquito and mouse brain inoculation techniques are not routinely used in day-to-day diagnosis owing to their highly specialized technical, safety and facility requirements as well as high maintenance costs. Instead, virus isolation using cell lines is more widely used. The most commonly used cell line for DENV isolation is C6/36, which was derived from A. albopictus. Alternatively, mammalian cell lines such as Vero, LLC-MK2 or BHK-21 could also be used, although these offer lower sensitivity than C6/36 [41,42,48–53]. Besides diagnosis, virus isolation offers the advantage of providing a virus isolate that may be characterized during subsequent in vitro studies, such as genome sequencing, virus neutralization and infection studies.

A successful isolation of DENV on mosquito or cell culture can be confirmed and serotyped by an immunofluorescence assay using DENV- and serotype-specific monoclonal antibodies, respectively [41,53]. Virus isolation is highly specific and has a theoretical detection limit of a single viable virus, although in practice, the sensitivity is only approximately 40.5% in cell line-based virus isolation [54]. It also requires highly trained operators, a dependence on sample integrity and a short viremia period, thus providing a narrow window of opportunity from illness onset [54]. Therefore, despite its advantages, this approach is not widely used in routine diagnostic laboratories.

Viral RNA detection

Reverse transcriptase PCR (RT-PCR) detection of dengue viral RNA extracted from blood, serum or plasma provides a rapid, sensitive and specific method for dengue infection confirmation. Various primers and protocols have been developed, validated and used in conventional RT-PCR [1,54–61] and real-time RT-PCR, either using SYBR® Green as a fluorescent detection marker or labeled oligonucleotide probes [58,59,62–79]. A technique using a single reaction mixture at a constant temperature (nucleic acid sequence-based amplification [NASBA]) [80] was found to be highly sensitive (98.5%) and specific (100%) [81,82]. NASBA may be highly useful and applicable during outbreak field diagnosis where thermocyclers are not readily available.

Sensitivity of conventional RT-PCR ranges from 48.4 to 98.2% and has a detection limit of 1–50 plaque forming units (PFUs) [54–56,59]. These assays employ primers that bind to known conserved regions of the DENV genome to avoid false negative findings due to spontaneous mutations expected in the replication of the RNA viral genome. The use of in silico methods to develop a cocktail of primers that bind to almost all DENV with known sequences has also been explored [58], although validation in a clinical setting remains to be carried out. The sensitivity of RT-PCR is also highly dependent on the short window of opportunity that coincides with the viremic period, which can last up to 8 days from illness onset (Figure 1). However, RT-PCR is rarely positive in a case of dengue after 6 days from illness onset [1].

Fluorescence-based real-time RT-PCR has a better reported sensitivity (58.9–100%) and detection limit (0.1–3.0 PFUs) owing to the sensitivity of the fluorescence detector within the thermocycler [59,62–67,71,74,75,79]. Multiplex RT-PCRs that differentiate DENV serotypes in a single assay have also been developed [56,59,65,69,76]. The experience with NASBA is more limited compared with RT-PCR, although a study has shown that it can be as sensitive as RT-PCR, with a detection limit of <25 PFUs/ml [81]. RNA extraction from whole blood may be more sensitive (90.0%) than serum or plasma (62.0%) in the same pool of samples [78].

Besides blood samples, RT-PCR can also be used to detect DENV RNA in tissues, including formalin-fixed specimens [82]. Although RT-PCR usually requires expertise in molecular techniques and expensive equipment [83], modified protocols using fast-ramping thermocyclers can be used in conjunction with newly trained operators during emergency settings, such as in differentiating dengue from SARS during an outbreak [62], provided a strict standard operating procedure is followed.

Dengue viral RNA can also be detected in urine [68,72] and saliva [72] samples using real-time RT-PCR. In urine, samples collected between day 6 and day 16 after illness onset were found to have higher rates of detection (50–80%) compared with day 1 to 3 samples (25–50%) [68]. RT-PCR for DENV in urine may thus extend the window of opportunity for viral RNA detection compared with blood specimens (up to day 8). However, the level of viral RNA in urine and saliva samples is low (1 × 101–5 × 101 PFUs/ml) compared with the corresponding serum samples (7.9 × 102–1.9 × 105 PFUs/ml) [72].

Antigen detection

Dengue NS1 is a highly conserved glycoprotein essential for DENV viability and is secreted from infected cells as a soluble hexamer [84,85]. Serum or plasma DENV NS1 level has been found to correlate with viremia titer and disease severity [86–88]. It can be found in the peripheral blood circulation for up to 9 days from illness onset [89–91], but can persist for up to 18 days from illness onset in some cases [92]. NS1 detection thus offers a larger window of opportunity for diagnosis of dengue compared with virus isolation, RT-PCR or NASBA [68,72]. Commercially available NS1 capture-based detection kits with sensitivities that ranged from 54.2 to 93.4% have been comprehensively evaluated (Table 2) [66,79,91,93–103] and found to be able to confirm dengue infection in serum specimens that were RT-PCR negative and secondary dengue infection [97]. However, NS1 detection is less sensitive in secondary dengue infection (67.1–77.3%) compared with primary dengue cases (94.7–98.3%) [93,94,96,103], probably owing to the presence of crossreactive anti-NS1 antibodies that impedes the detection of free NS1 proteins in the serum or plasma [86,89].

Anti-NS1 antibodies can also be used to detect infection in other sample sources, such as tissues, including liver, lung and kidney [104], through immunohistochemistry. This could be useful in postmortem studies. Although highly conserved, serotype-specific NS1 monoclonal antibodies have been raised and applied for NS1-based dengue serotype identification assays [105–109]. A study by Puttikhunt etal. has shown an overall sensitivity of 76.5% and specificity of 100% for diagnosis of dengue while having serotyping sensitivities of 100% for DENV 1, 3 and 4 and 82.4% for DENV2 [109]. However, the sensitivity of these tests may differ with different strains of DENV as the magnitude of NS1 secretion appears to be strain dependent [110].

Antibody detection (IgM & IgG)

Detection of antidengue antibodies (IgM and IgG) is the most widely used test in diagnosis of dengue [111]. These kits are either in the form of Ig capture or direct Ig detection and are configured to detect IgM, IgG or both simultaneously [102,112–115]. There are two versions of these tests: ELISA or strip format (rapid test). While ELISA provides greater sensitivity, the strip format is amenable for bedside use [116].

Antibody response in the form of antidengue IgM can be detected as early as 3–5 days after illness onset. Levels of IgM continue to increase for approximately 2 weeks thereafter and may persist for approximately 179 and 139 days following primary and secondary infection, respectively [117]. Thus, while a single IgM raises the likelihood that a febrile patient has dengue, a definitive diagnosis may require the use of paired sera to demonstrate rising IgM titers.

A multinational and multicenter study of ten IgM kits has concluded that ELISA-based detection kits have higher sensitivities (61.5–99.0%) compared with the rapid test formats (20.5–97.7%). The specificities are in the range of 79.9–97.8% and 76.6–90.6% for ELISA and rapid tests, respectively [115]. Other evaluation studies have also reported similar sensitivities and specificities [62,102,116]. The wide ranges of these values are probably due to the timing of sample collection [118].

Early antidengue IgM response (<2 months) has been found to be crossreactive to all four DENV serotypes [119] and other flaviviruses [115]. Hence, epidemiological information on the prevalence of other flaviviruses would be useful to guide the interpretation of a positive IgM finding. False positives have also been observed in patients with previous dengue or malaria infection [116]. However, more efficient algorithms can be developed to mitigate this problem, as shown by Prince etal. [120].

During primary infection, IgG can only be detected after 10days from illness onset, making it less useful for early diagnosis. However, the rapid increase of IgG levels during secondary infection (as early as day 4 from illness onset) [1] can be suggestive of dengue when the ratio of IgM and IgG is used [62,102,115–117].

Dengue neutralizing antibody detection

Neutralizing antibodies inhibit DENV infection and can thus provide greater specificity in distinguishing antibodies to DENV from other crossreactive flavivirus antibodies [121,122]. These antibodies can be measured by using plaque reduction neutralizing tests (PRNTs), first developed by Russell etal. [122] based on the protocol from Dulbecco etal. [123]. To date, PRNT remains the most widely used assay for immunity studies [124,125]. However, it is labor intensive, time consuming and has low throughput [124], and is therefore not routinely used in dengue diagnostics.

New tests such as the ELISA-based microneutralization test (ELISA-MN) [126], the fluorescent antibody cell sorter-based Dendritic Cell-Specific Intercellular adhesion molecule-3-Grabbing Non-integrin expressor DC assay [127] and the enzyme-linked immunosorbent spot microneutralization assay [128] have been developed to overcome the limitations associated with PRNT. These new tests have been separately validated, compared and evaluated [124,126–130] against PRNT and found to have good agreement (false-positive rate <10%) in cases with primary DENV infection [124,130]. However, Putnak etal. reported poor agreement among the tests in association with vaccination or secondary infection [124]. This result could have been influenced by the use of different cell lines or different strains of DENV [129]. Others have suggested the use of Fcγ receptor (FcγR)-positive cells for these assays since DENV infects monocytes and DCs that express such receptors [131,132]. The use of such cells may also provide information on whether the antibodies were able to neutralize DENV intracellularly or whether neutralization was only mediated through the coligation of FcγRIIB, which inhibits FcγR-mediated phagocytosis and hence DENV entry into monocytes [133]. A limited observation suggests that this could provide greater specificity on the DENV serotype responsible for the infection [133]. However, detailed validation is required for all of these assays before they can be used clinically.

Combined antigen/antibody detection

Given the dynamic nature of the NS1 antigen, antidengue IgM and IgG antibody levels during the course of acute illness, efforts have been made to combine all three tests into a single reaction for ease of use. Some of these have been evaluated and shown to have promising diagnostic sensitivity (89–93%) and specificity (75.0–100%) [91,100–102,134].

Advances in rapid diagnostic tools

While rapid bedside diagnosis formats are available for antigen or antibody detection or both simultaneously, the sensitivities and specificities of the available tests have been uniformly lower than the equivalent laboratory-based assays. A complete review of these assays is provided elsewhere [102,118]. These limitations may be due to the use of the lateral flow dipstick approach. The new lab-on-a-chip platform could offer a way to improve the performance of these bedside diagnostic tools [135–137]. This platform makes use of a number of new technologies; some in combination, such as microfluidics [135–137] and grating couplers [137] to improve multiplexing, accommodate better mixing of reagents with test samples as well as achieving greater sensitivity for detecting positive signals [138]. This platform could feature prominently in dengue diagnostics in the coming years.

Disease prognostication

Progression of mild dengue infection to severe dengue (DHF/dengue shock syndrome) is difficult to predict owing to an incomplete understanding of disease pathogenesis. Symptoms and signs of severe dengue have a sudden onset at the time of defervescence [1,19]. Careful monitoring of hematocrit as well as signs of circulatory failure or internal hemorrhage needs to be carried out for at least 2 days after fever defervescence. Specifically, patients should be observed for signs such as severe abdominal pain, passage of black stools, bleeding into the skin, nose or gums, sweating or cold skin, which could indicate the development of DHF [7]. Depending on disease progression, should DHF, occurs, oral rehydration therapy is sufficient for milder DHF while intravenous fluid therapy is suggested for more severe manifestations, and blood transfusion is suggested for critical cases [7].

However, hospitalization for close monitoring of all patients in dengue-endemic countries is often not feasible, particularly during outbreaks, as it stresses the limited medical healthcare resources. Under such circumstances, an ability to predict the development of severe dengue at the early stages of illness could thus be useful for triaging patients. Various clinical markers have been proposed as warning signs of severe dengue progression, as shown in Table 3. How well these clinical symptoms/signs perform in predicting the onset of severe dengue remains to be fully determined.

Besides monitoring individual symptoms or signs, several groups have also evaluated the usefulness of combining these into an algorithm for predicting severe dengue. Lee etal. explored the use of a probability equation that combines four simple clinical laboratory observations, including bleeding, lymphocyte proportion, increased serum urea and low total serum protein [139]. They reported a sensitivity of 100% and specificity of 46%. They estimated that 43.9% of the mild dengue cases could have been prevented from hospitalization in 2004. The authors followed up this study with a prospective validation of their equation in the same hospital and found similar levels of accuracy [140].

Another algorithm using platelet count, crossover threshold of PCR-positive results (viremia in blood) and pre-existing antidengue IgG (secondary infection) measured during the first 72 h of illness was shown to predict hospitalization with a sensitivity and specificity of 78.2 and 80.2%, respectively [36]. Likewise, as discussed earlier, the algorithm that distinguishes DHF from dengue infection is able to achieve a sensitivity of 79.6% [37]. The ability to calculate the probability of development of severe dengue based on routinely performed clinical tests could also be useful to guide prognostication [37]. However, further prospective clinical studies are needed to validate their usefulness.

Quality assurance

While the authors have reviewed the sensitivities and specificities of the various tools for diagnosis of dengue, how these tests actually perform can be affected by a number of variables that differ from laboratory to laboratory, or from region to region. Thus, quality assurance programs should be instituted in all diagnostic and reference laboratories that offer services in diagnosis of dengue. This ensures that the tests perform at the expected levels in different laboratories and in different hands. Details on such quality assurance programs have been reviewed elsewhere [141].

Diagnosis of dengue in a vaccinated population

While an effective and safe vaccine against dengue is anticipated, its introduction could also provide fresh challenges for diagnosis of dengue. Even though the goal of vaccination is to eliminate dengue cases entirely, there are currently no data that indicate that a complete elimination of DENV is feasible with vaccination programs. On the contrary, there remains a concern that the antibodies generated by vaccination may enhance dengue, particularly when antibody levels wane in the years following vaccination. Furthermore, vaccination may not prevent infection against all strains [142] or may drive the emergence of vaccine-escape mutants [143], as encountered with other infections, such as hepatitis B [144]. A comprehensive surveillance of dengue among cases of febrile illness would thus be needed to determine the true efficacy of vaccination and to monitor for vaccine failure.

Epidemiologically, vaccination would reduce DENV transmission and hence the prevalence of dengue. Under such circumstances, diagnostic approaches or tests with high sensitivity but poor specificity would result in a high false-positive rate. However, a low sensitivity could lead to false-negative findings, which could result in an inability to detect the emergence of vaccine failure or escape mutants early enough to trigger the necessary public health responses.

Given these requirements, clinical diagnosis using symptoms, signs and standard routine hematological or biochemical tests is unlikely to provide sufficient specificity (Figure 1). Furthermore, vaccinated individuals may also present with a milder illness than classical dengue infection, making approaches such as the use of the WHO dengue classification schemes less sensitive. Diagnosis of acute DENV infection must thus rely even more on the laboratory.

Serologically, DENV infection in vaccinated individuals would also resemble that of a secondary infection, where a rise in IgM titers is not a consistent feature but a rapid rise in IgG titers or the ratio of IgM and IgG could be suggestive of acute DENV infection [117]. In this respect, collection of a convalescent serum sample to demonstrate rising antibody titers would be very useful in interpreting these serological tests. Caution will need to be exercised in places where another flavivirus, such as West Nile or Japanese encephalitis virus, circulates. Overall, however, serological approaches will probably lack the specificity required for a definitive diagnosis of dengue in the low prevalence setting expected in vaccinated populations (Figure 2).

Detection of DENV or components of DENV are likely to provide the necessary sensitivity and specificity needed (Figure 2). While NS1 antigen detection is easy to use and is suited to point-of-care diagnosis, the presence of vaccine-induced antidengue IgG antibodies, as with secondary DENV infection, could lower the overall sensitivity of this test. Likewise, while virus isolation may be highly specific, it lacks sufficient sensitivity, especially since in most places, a suitable insectary for mosquito inoculation is not likely to be available and laboratories will have to rely on cell cultures. However, virus isolation will not be redundant and would need to be done in all RT-PCR-positive specimens, as isolation of vaccine-escape mutants would be needed to characterize these viruses. Such information could be useful in updating vaccine composition through the development or selection of appropriate vaccine strains or even updating the primers and probes used in molecular diagnostic assays [143].

Nucleic acid detection offers the highest sensitivity and specificity, and would thus be the most appropriate approach for acute diagnosis of dengue in vaccinated populations with low disease prevalence (Figure 2). Emphasis should be on those assays that have been carefully validated in different laboratories serving different populations. The availability of panels of standardized positive and negative controls, along with an internationally coordinated quality assurance program, would be needed to ensure consistency in the performance of the diagnostic assays. Presently, such a molecular diagnostic assay is lacking. RT-PCR method used in different laboratories differ in terms of primers/probes, enzymes and buffers as well as cycling conditions. The method of detection of the RT-PCR amplicon, whether as an end point or real-time assay, is also likely to be different, as with the method of viral RNA extraction from clinical specimens. These limitations need to be addressed urgently if we are to be prepared for diagnosis of dengue and surveillance in a postvaccination world.

Expert commentary

DENV and its mosquito vectors have expanded geographically throughout the tropical world and are now encroaching into subtropical regions. These trends make dengue a global health concern. In the absence of either a licensed vaccine or antiviral drug, reduction of the disease burden relies on early clinical recognition of dengue and the timely initiation of supportive therapy. As differentiation between dengue and other causes of febrile illnesses is difficult based on presenting symptoms and signs, laboratory tests are needed for a confirmatory diagnosis. This review summarizes the current knowledge on clinical as well as laboratory diagnosis of dengue. It reveals that clinical approaches generally have high sensitivities but poor specificities and discusses the various decision algorithms that have been designed to improve the specificity of clinical diagnosis. For confirmatory diagnosis, a range of laboratory tools are available and the main consideration on which tool to use is the time from illness onset. A central theme of this review is the need for a systematic validation of the performance of both the decision algorithms and laboratory assays in different populations and diagnostic laboratory settings, respectively. This need for quality-assured standardized performance could, paradoxically, become more acute when a dengue vaccine or antiviral drug becomes available. The consequent reduction in dengue prevalence necessitates the use of the most sensitive and specific method to derive useful positive and negative predictive values to support clinical decisions in treatment and public health responses.

Five-year view

We speculate that a dengue vaccine will be near licensing in 5years and that potential antiviral drugs against dengue will also enter late stages of clinical trials. The implementation of either countermeasure against dengue would shift the emphasis of diagnosis of dengue from serological to virological. Tools that detect either the viral genome or antigen, particularly at the bedside, would gain favor. These tools are better able to distinguish dengue from other flaviviral infections and are also useful in the early phases of illness, when initiation of antiviral therapy would probably exert its maximal effect. Furthermore, definitive diagnosis of dengue in vaccinated populations would become even more important as it could herald waning immunity or emergence of vaccine-escape mutants; either scenario would trigger a public health emergency. Hence, the need for improvements to existing approaches for the diagnosis of dengue would not be diminished with the advent of either vaccination or antiviral drug therapy, but rather the demand for tests that achieve near-perfect sensitivity and specificity will increase in the next 5 years.

Table 1. Symptoms differentiating dengue infection from other febrile illnesses.

Symptoms	Den|OFIs	p-value	Children (<15 years)|adults (>15 years)	p-value	Ref.
Nausea	50.0|28.9% 68.0|49.0%^† 51.3|30.5%	<0.00001 <0.05 <0.001	50.2|76.4%	<0.001	[12,103,145,146]
Vomitting	16.4|8.4% 16.2|8.5% 57.0–64.0|31.0–46.0%^† 70.0|52.0%^†	<0.00001 0.03 <0.01 <0.05	50.2|76.4%	<0.001	[12,103,145–148]
Retro-orbital pain	26.0|15.9% 26.6|13.5% 10.01^§	<0.00001 0.003 0.001	8.7|29.1%	<0.001	[12,103,146,148]
Aches/pains	1.4^§	<0.0001	20.3|36.4%	0.012	[12,146]
Rash	11.2–41.2|3.0–6.4%	<0.003/0.007	NA	NA	[12,103,134,149]
Tourniquet test positive	34.0|19.0% 42.0|5.0%^† 43.0–65.0|21.0–39.0% 1.86^§	0.02 <0.01 <0.1 <0.001	NA	NA	[12,103,134,149]
Leukopenia	3.8 × 10^3|7.3 × 10^3/µl 4.5 × 10^3|8.1 × 10^3/μl <4.5 × 10^3/μl: 72.1|11.5%	<0.0001 <0.1 <0.001	NA	NA	[12,37,103,145]
Thrombocytopenia (platelet/mm^3)	16|4% (≤100,000)^† 16|82%^‡ (≤100,000) 66|95%^‡ (≤100,000) 14.9|1.5% (≤100,000) 32,000|96,500 163,500|239,000 70,000|104,000^¶	NA NA <0.01 <0.001 <0.001 <0.0001 NA	NA	NA	[12,103,145,147,150]

^†Studies perfomed in children younger than 15 years.

^‡Dengue and severe dengue comparison, performed in children younger than 15 years.

^§Risk ratio.

^¶Average.

Den: Confirmed dengue cases; NA: Not applicable; OFI: Other febrile illness.

Table 2. Laboratory diagnostics for dengue: sensitivity and specificity.

Category	Technique	Parameters			Ref.
		Sensitivity (%)	Specificity (%)	Detection limit
Viral detection	Virus isolation (mosquitoes)	71.5–84.2	100	NA	[6,40–42,44]
	Virus isolation (mouse intrecerebral inoculation)	NA	NA	NA	[46,47]
	Virus isolation (cell culture)	40.5	100	≥1 viable virus	[54]
	Viral RNA RT-PCR (conventional)	48.4–100	100	1–50 PFUs	[54–57]
	Viral RNA RT-PCR (real-time detection)	58.9–100	100	0.1–3.0 PFUs	[54,57,63–67,79]
	Viral RNA RT-PCR (NASBA)	98.5	100	<25 PFUs/ml	[81]
	Viral antigen detection (NS1 detection)	54.2–93.4	92.5–100	0.2 ng/ml^†	[79,91–103,110]
Antibody detection	IgM detection	61.5–100	52.0–100	NA	[54,102,115,116]
	IgG detection	46.3–99.0	80.0–100	NA
	Rapid IgM detection (strips)	20.5–97.7	76.6–90.6	NA	[115]
Antigen/antibody combined detection	NS1 and IgM	89.9–92.9	75.0–100	NA	[91,100–102]
	NS1 and IgM/IgG	93.0	100	NA	[101,151]

NA: Not applicable; NASBA: Nucleic acid sequence-based amplification; PFU: Plaque forming unit; RT-PCR: Reverse transcriptase PCR.

^†Data taken from [110].

Table 3. Warning signs and symptoms leading to potential severe dengue (dengue hemorrhagic fever/dengue shock syndrome).

Warning signs for severe disease progression
Signs	Den|OFIs	p-value	Ref.
Abdominal pain & tenderness	63.3|22.6% 78.0|22.0%^†	<0.05 0.03	[150,152]
Persistent vomiting	NA	NA	[153]
Clinical fluid accumulation	46.4|7.1%	0.002	[154]
Mucosal bleed/ spontaneous bleeding	84.1|65.9% 26.6|10.4%	0.008 0.05	[150,153]
Lethargy, restlessness	12.7|2.2% 54.0|20.0%^†	0.05 0.002	[150,152]
Liver enlargement >2 cm	91.6|72.4%^‡	0.01	[146]
Severe dengue (DHF/DSS)
Symptoms	Mild|severe	p-value	Ref.
Severe plasma leakage leading to shock and fluid accumulation with respiratory distress	92.9–100% in severe den	NA	[154]
Severe bleeding (evaluated by clinician)	10.0–35.7% in severe den Gums: 5.7–6.0|65.0–67.8% Nose: 0.9|12.0–16.9%	<0.01	[139,154]
Severe organ involvement  Liver: increased AST/ALT  CNS: impaired consciousness  Heart and other organs	AST: 1293|196 IU/l ALT: 309|132 IU/l AST: 76|12 U/l ALT: 30|20 U/l	0.015 0.075 <0.01 <0.01	[37,145,153,155]
Thrombocytopenia ≤100,000/mm^3	16|82%	NA	[145]

^†Dengue and severe dengue comparison, performed in children younger than 15 years of age.

^‡Children (younger than 15 years of age) versus adults (older than 15 years of age).

ALT: Alanine aminotransferase;

AST: Aspartate aminotransferase; Den: Confirmed dengue cases; DHF: Dengue hemorrhagic fever; DSS: Dengue shock syndrome; IU/l: International units per liter; NA: Not applicable; OFI: Other febrile illness.

Key issues

• • Clinical diagnosis using the 1997–2009 WHO dengue classification schemes has high sensitivity but lacks specificity.

• • Decision algorithms for diagnosis have been proposed but lack prospective validation.

• • Choice of diagnostic assays should be guided by the time from illness onset.

• • The presence of pre-existing antibodies from a previous heterologous dengue virus infection or a previous flavivirus infection can affect the sensitivity or specificity of many diagnostic assays.

• • Postvaccination surveillance would face the same challenges for diagnostics as currently encountered with secondary dengue infection.

• • Despite the above, diagnosis of dengue in vaccinated individuals is critical for the surveillance of vaccine failure and escape mutants.

• • Diagnostic assays with high sensitivity and specificity will be in particular demand in the low dengue prevalence setting following vaccination.

Financial & competing interests disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.