Abstract
Background. Anti–dengue virus (DENV) immunoglobulin M (IgM) seroconversion has been the reference standard for dengue diagnosis. However, paired specimens are rarely obtained, and the interval for this testing negates its usefulness in guiding clinical case management. The presence of DENV viremia and appearance of IgM during the febrile phase of dengue provides the framework for dengue laboratory diagnosis by using a single serum specimen.
Methods. Archived paired serum specimens (n = 1234) from patients with laboratory-confirmed dengue from 2005 through 2011 were used to determine the diagnostic performance of real-time reverse transcription polymerase chain reaction (RT-PCR), for detection of DENV serotypes 1–4, and enzyme-linked immunosorbent assays (ELISAs), for detection of DENV nonstructural protein 1 (NS1) antigen and anti-DENV IgM.
Results. During 1–3 days after illness onset, real-time RT-PCR and NS1 antigen testing detected 82%–69% and 90%–84% of cases, respectively, as viremia levels declined, while anti-DENV IgM ELISA detected 5%–41% of cases as antibody appeared. Over the 10-day period of the febrile phase of dengue, the cumulative effect of using these 3 types of tests in a diagnostic algorithm confirmed ≥90% of dengue cases.
Conclusions. The use of molecular or NS1 antigen tests to detect DENV and one to detect anti-DENV IgM in a single serum specimen collected during the first 10 days of illness accurately identified ≥90% of dengue primary and secondary cases.
Dengue is an important public health problem worldwide, with an estimated 96 million cases annually in tropical and subtropical countries [1, 2]. Infection with each of the 4 dengue virus (DENV) serotypes (DENV-1–4) may cause dengue, and there are no vaccines or drugs to prevent or treat the disease. The clinical presentation of dengue is similar to that of other acute febrile illnesses (AFIs; eg, malaria, leptospirosis, measles, influenza, and chikungunya); therefore, early laboratory diagnosis should improve clinical management and public health surveillance.
Historically, laboratory diagnosis of dengue required demonstration of anti-DENV immunoglobulin M (IgM) seroconversion in 2 serum samples, one collected during the first 5 days of illness and a second collected after 5 days [3]. However, the long interval between specimens made collection difficult, was not useful for patient-care decisions, and introduced an inherent bias in surveillance data.
Patients with dengue or other AFIs usually seek medical attention within several days of fever onset [4, 5], providing the opportunity to diagnose dengue by evaluating a single serum specimen obtained at this time. DENV viremia occurs for 3–5 days prior to fever onset and continues for approximately 5 days into the febrile illness, and it can be detected by molecular assays specific for DENV RNA or immunoassays targeting DENV nonstructural protein 1 (NS1) antigen [6]. In addition, an anti-DENV IgM response becomes detectable by IgM-capture immunoassays 3–5 days after onset of fever, peaks 6–10 days after fever onset, and may persist for up to 90 days [7, 8].
We conducted a retrospective study, using serum specimens prospectively collected from patients with clinically suspect dengue identified 1–10 days after onset of illness to determine the performance of single-specimen diagnostic testing versus that of IgM anti-DENV seroconversion in paired specimens. Results were analyzed to determine which tests provided the best overall diagnostic result on each day after onset of illness among persons with primary and secondary DENV infection. These results were used to design a diagnostic testing algorithm based on when a specimen was obtained after illness onset.
METHODS
Clinical Specimens
Archived paired serum specimens (n = 1234) were obtained from randomly selected patients with suspect dengue who were tested at the Centers for Disease Control and Prevention Dengue Branch in San Juan, Puerto Rico, as a part of 2 surveillance programs, described below, conducted from January 2005 through December 2011. All patients met the 2009 World Health Organization (WHO) dengue case definition and had one specimen collected during the acute febrile phase, 1–5 days after onset of illness, and a second specimen collected during the convalescent phase, 6–14 days after onset of illness. There were 483 patients for whom the CDC IgM antibody-capture enzyme-linked immunosorbent assay (CDC MAC ELISA) [7] revealed anti-DENV IgM seroconversion between collection of the acute- and convalescent-phase specimens.
Specimens were submitted for diagnostic testing as part of 2 Puerto Rico dengue surveillance programs: the island-wide, passive dengue surveillance system (PDSS [5, 9]) and a population-based enhanced dengue surveillance system (EDSS [10]). The PDSS obtains diagnostic specimens and demographic and clinical information from all patients with suspected dengue reported from all over the island [5, 9]. The EDSS was a study of AFI in the Ponce health district, located in southern Puerto Rico, and collected the same demographic and clinical information as the PDSS [10]. Patient and diagnostic information was maintained in a secure database.
Ethics Statement
This study was reviewed by the CDC Human Research Protection Office and determined to be exempt from US human subjects' protection regulations (protocol 6273). Specimens used in this study were deidentified according to CDC Institutional Review Board approved protocol 6279.
Dengue Diagnostic Testing Procedures
Selected specimens were previously tested according to the diagnostic algorithm in use for each surveillance system (PDSS and EDSS) and included anti-DENV IgM. Specimens in the final panel were reevaluated according to the manufacturer's instructions by means of the following 6 tests: the CDC DENV-1–4 Real-Time RT-PCR (RT-PCR) Assay [11], performed in the multiplex mode according to the package insert and used to detect DENV RNA (limit of detection, 1 × 103 genome copy equivalents); 2 DENV NS1 antigen-capture assays, the Dengue Early ELISA (Panbio, Melbourne, Australia) and the DENV Detect NS1 ELISA (InBios, Seattle, WA); and 3 anti-DENV IgM-capture ELISAs, the CDC MAC ELISA (serum dilution, 1:400) [7], the Dengue Virus IgM Capture DxSelect ELISA (Focus Diagnostics, Cypress, CA), and the DENV Detect IgM Capture ELISA (InBios). Testing was performed in a Clinical Laboratory Improvement Amendment–certified laboratory, and results were presented according to day after illness onset.
To classify patients according to their DENV infection status, paired specimens were tested by the CDC anti-DENV IgG ELISA [12], which was calibrated on the basis of a hemagglutination inhibition assay. Patients with paired specimens positive for anti-DENV IgG in the acute phase (ie, 1–5 days after illness onset) were defined as having secondary or greater DENV infection; patients with a negative IgG ELISA result were classified as having primary DENV infection.
Analytical Plan
Analysis of Test Sensitivity
Anti-DENV IgM seroconversion detected by the CDC MAC ELISA, defined as a negative test result for the acute-phase specimen (obtained 1–5 days after illness onset,) and a positive test result for the convalescent-phase specimen (obtained 6–10 days after illness onset), was used as the reference to independently determine the sensitivity of each of the other tests. In another analysis, the CDC DENV 1–4 real-time RT-PCR was used as the reference test to determine the sensitivities of the NS1 and anti-DENV IgM ELISAs on days 1–8 after illness onset. In all instances, a positive result was defined by the cutoffs specified in the manufacturers package insert for all commercial tests in this study.
For all analyses, point estimates and 95% confidence intervals (CIs) were calculated using the Wilson score interval, implemented in the binconf function in the Hmisc package in R, version 3.10–1.1 (available at: http://CRAN.Rproject.org/package=Hmisc).
Analysis of Cumulative Effect
To determine the additive contribution of each test to the final diagnosis of dengue at each day after illness onset, all specimens, whether from a seroconversion-positive or -negative pair, were included in the analysis. The additive effect of combining diagnostic tests was determined using a step-wise hierarchy when >1 test was positive for a specimen on a given day after illness onset. Thus, a cumulative effect of RT-PCR and anti-DENV IgM included specimens positive by either RT-PCR and/or anti-DENV IgM. Similarly the cumulative effect of NS1 and anti-DENV IgM included specimens positive by either NS1 and/or anti-DENV IgM.
RESULTS
The Testing Panel
A total of 617 patients (1234 serum specimens) were identified who met the inclusion criteria. The patients' mean age was 27 year, the median age was 18 years, and the ages ranged from 2 months to 84 years; however, age-specific analysis by day of illness onset could not be performed because of inadequate sample sizes. Approximately 52% of patients were male. All specimens were tested to determine DENV infection status; 295 (24%) were primary infections, and 939 (76%) were secondary or greater infections. Although case patients were identified during a period when all 4 DENV serotypes were circulating in Puerto Rico, there were too few specimens to allow for robust serotype-specific analyses (DENV-1, 284 specimens; DENV-2, 81; DENV-3, 30; and DENV-4, 84).
IgM Anti-DENV ELISAs
When the 3 anti-DENV IgM ELISAs were used to test the panel (Figure 1A), the sensitivities on day 1 after illness onset were 0% (95% CI, 0%–8%), 7% (95% CI, 2%–18%), and 5% (95% CI, 1%–15%), for the CDC MAC ELISA, the Dengue Virus IgM Capture DxSelect ELISA, and the DENV Detect IgM Capture ELISA, respectively, and beginning >2 days after illness onset, the sensitivities increased and differed by the respective test (Figure 1A and Supplementary Table 1). These same specimens were used to determine the percentage positivity by test type (Figure 1B ). Similar trends were observed as with sensitivity results, indicating that the Dengue Virus IgM Capture DxSelect ELISA had the highest percentage positivity throughout infection, followed by the DENV Detect IgM Capture ELISA and the CDC MAC ELISA.
A, Sensitivities of anti–dengue virus immunoglobulin M (IgM)–capture enzyme-linked immunosorbent assays (ELISAs), by day after onset of illness, when compared to the Centers for Disease Control and Prevention (CDC) real-time reverse transcription polymerase chain reaction (PCR) as the reference test. B, Percentage of specimens testing positive by IgM-capture ELISAs, out of the total number of specimens tested, by day after onset of illness. Whiskers denote 95% confidence intervals.
Sensitivity of CDC DENV-1–4 Real-Time RT-PCR
When the CDC DENV-1–4 Real-Time RT-PCR was compared to anti-DENV IgM seroconversion findings yielded by the CDC MAC ELISA, its sensitivities ranged from 9% to 82% (Figure 2 and Supplementary Table 2). When real-time RT-PCR was compared to seroconversion findings determined by the DENV Detect IgM Capture ELISA or Dengue Virus IgM Capture DxSelect ELISA, there was no measurable difference in test sensitivities (Supplementary Figure 1A and 1B).
Sensitivities of dengue virus (DENV) nonstructural protein 1 (NS1) and DENV nucleic acid detection by the Centers for Disease Control and Prevention (CDC) real-time reverse transcription–polymerase chain reaction (PCR) assay, by day after onset of illness, when compared to detection of immunoglobulin M (IgM) seroconversion by the CDC MAC enzyme-linked immunosorbent assay (ELISA) as the reference test. Whiskers denote 95% confidence intervals.
Sensitivities of DENV NS1 ELISAs
The sensitivities of the NS1 antigen-capture ELISAs were determined by comparison to IgM anti-DENV seroconversion findings yielded by the CDC MAC ELISA, the Dengue Virus IgM Capture DxSelect ELISA, and the DENV Detect IgM Capture ELISA. When compared to positive results of the CDC MAC ELISA (indicating the presence of anti-DENV IgM seroconversion), sensitivities of the DENV Detect NS1 ELISA ranged from 30% to 91% (Figure 2 and Supplementary Table 3). The sensitivities of the Dengue Early ELISA ranged from 13% to 87% (Figure 2 and Supplementary Table 3). When the same comparison was performed for the DENV Detect IgM Capture ELISA or the Dengue Virus IgM Capture DxSelect ELISA, the sensitivities of the NS1 ELISAs, by day after illness onset, did not vary from that observed in comparisons to seroconversion results of the CDC MAC ELISA (Supplementary Figure 1A and 1B).
When the NS1 ELISAs were performed on specimens that tested positive by the CDC DENV-1–4 Real-Time RT-PCR, sensitivities of the DENV Detect NS1 ELISA ranged from 47% to 93% (Figure 3 and Supplementary Table 3). The sensitivities of the Dengue Early ELISA (Panbio) ranged from 33% to 93%. Specimens collected on days 9 and 10 after illness onset were not analyzed because of small sample sizes. The lowest sensitivity obtained was 33%, on days 8 and 10 (Figure 3 and Supplementary Table 3).
Sensitivities of dengue virus (DENV) nonstructural protein 1 (NS1) antigen-capture enzyme-linked immunosorbent assays (ELISAs), by day after onset of illness, when compared to the Centers for Disease Control and Prevention real-time reverse transcription–polymerase chain reaction as the reference test.
Analysis by DENV Infection Status
Both NS1 ELISAs performed best during the first 10 days of illness among case patients with primary DENV infection, compared with results for patients with secondary or greater DENV infection. This difference in sensitivity was statistically significant (α level, 0.05) for the Dengue Early ELISA on days 3–7 after illness onset and for the DENV Detect NS1 ELISA for days 4, 6, and 7 after illness onset (Figure 4A and Supplementary Table 4).
Difference in sensitivity between specimens from patients with primary and secondary or greater dengue virus (DENV) infection status among nonstructural protein 1 (NS1) antigen-capture enzyme-linked immunosorbent assays (ELISAs; A) and immunoglobulin M (IgM) anti-DENV ELISAs (B), by day after illness onset, when compared to the Centers for Disease Control and Prevention (CDC) real-time reverse transcription polymerase chain reaction as the reference test. Whiskers denote 95% confidence intervals.
There was no statistically significant difference in performance during the first 10 days of illness between the CDC MAC ELISA, the Dengue Virus IgM Capture DxSelect ELISA, and the DENV Detect IgM Capture ELISA (Figure 4B and Supplementary Table 4).
Single-Specimen Dengue Diagnostic Testing Algorithm
The results of the sensitivity analyses for each test indicated that no single test was able to consistently detect DENV infection throughout the course of dengue (days 1–10 after illness onset), owing to time frames (windows) when levels of analytes decrease (for DENV RNA and NS1) or begin to appear (for anti-DENV IgM). Therefore, test combinations were assessed for their ability to accurately diagnosis dengue, using a single serum specimen obtained during this period (Figure 5).
Sensitivities of the cumulative effect for the best-performing dengue virus (DENV) nonstructural protein 1 (NS1) antigen-capture assays and anti-DENV immunoglobulin M (IgM)–capture enzyme-linked immunosorbent assays (ELISAs). A, The Centers for Disease Control and Prevention (CDC) real-time reverse transcription–polymerase chain reaction (RT-PCR), the CDC RT-PCR plus the DENV Detect NS1 ELISA (InBios), and the CDC RT-PCR plus the Dengue Virus IgM Capture DxSelect ELISA (Focus Diagnostics). B, The DENV Detect NS1 ELISA and the DENV Detect NS1 ELISA plus the Dengue Virus IgM Capture DxSelect ELISA. Whiskers denote 95% confidence intervals.
During the first 3 days of illness, tests for the detection of DENV RNA or NS1 antigen alone had the highest chance of detecting a DENV infection (79%–85% for the CDC DENV-1–4 Real-Time RT-PCR Assay and 86%–90% for the DENV Detect NS1 ELISA), and the addition of the Dengue Virus IgM Capture DxSelect ELISA achieved a small increase in the detection of DENV infection (range, 3%–12%). The cumulative effect of the CDC DENV-1–4 Real-Time RT-PCR Assay and the DENV Detect NS1 ELISA increased sensitivity incrementally, compared with each analyte alone. However, between days 4 and 7 of illness, the CDC DENV-1–4 Real-Time RT-PCR Assay detected 82% of DENV infections, declining to 26%, but the addition of the Dengue Virus IgM Capture DxSelect ELISA increased the overall sensitivity of DENV detection to 95%–100% (Figure 6). In a similar manner, the DENV Detect NS1 ELISA detected DENV infection with a sensitivity declining from 81% to 53% between days 4 and 7, but the addition of the Dengue Virus IgM Capture DxSelect ELISA increased the overall detection rate to 97%–99% (Figure 6). After 7 days of illness, the marginal increase in diagnostic detection of the 2-test algorithm was limited because levels of both analytes (viral RNA and antigen) were declining rapidly, compared with the rapid increase in anti-DENV IgM level (Figure 6).
Proposed algorithm of laboratory testing of a specimen from a patient with suspected dengue, by test type and day after onset of illness. Abbreviations: DENV, dengue virus; IgM, immunoglobulin M; NS1, nonstructural protein 1; RT-PCR, real-time reverse transcription–polymerase chain reaction.
DISCUSSION
This analytical study was a comparison of the performance of readily available tests for detection of DENV viremia and anti-DENV IgM, as well as the performance of these tests when used in a testing algorithm for a single serum specimen obtained during the febrile phase of dengue. While a retrospective study, the specimens were obtained from case patients who met the WHO dengue case definition and had been prospectively confirmed as dengue, based on anti-DENV IgM seroconversion.
The individual performance of the types of tests evaluated in this study has been evaluated in a number of other studies. DENV molecular diagnostic tests have advanced from gel-based to real-time RT-PCR analysis [11, 13, 14], which, although considered a complex test, can be performed in a relatively short time in laboratories in many dengue-endemic countries. NS1 antigen ELISAs provide DENV detection in a less complex format but have lower sensitivity and cannot determine DENV serotype [15, 16]. While the CDC MAC ELISA is probably the most widely used dengue diagnostic test, it has low sensitivity in the first days after illness onset, compared with tests that detect DENV [17, 18]. In addition, all of these immunoassays have been shown to have substantial variation in test performance, depending on the manufacturer [16, 19].
This study showed that, except for the first 3 days after illness onset, 2 tests, one for DENV (molecular or NS1 antigen) and the other for anti-DENV IgM antibody, are required to achieve maximum diagnostic accuracy when performed using a single serum specimen obtained during the first 10 days after illness onset. This study also showed that this testing scheme can be used for individuals with either primary or secondary DENV infection. The best test combination was the CDC DENV-1–4 Real-Time RT-PCR Assay and an anti-DENV IgM ELISA or an NS1 antigen ELISA and an anti-DENV IgM ELISA. While a combination of NS1 and DENV real-time RT-PCR increased laboratory confirmation of dengue during the first 2 days after illness onset, this was not different from the overall 2-test combination used throughout the 10-day period. This approach ensures that the window periods of low concentrations of DENV analytes, which occur during the first 10 days of illness, are filled in by using overlapping tests with high sensitivities, resulting in >90% detection of cases. The study shows that a single diagnostic specimen obtained in the febrile phase of the illness can be analyzed with confidence to make the diagnosis of dengue, a goal of dengue experts since 2006 [8] and a practice in use by the CDC Dengue Branch and Puerto Rico Department of Health for the PDSS for the past 5 years (CDC, unpublished data).
Sensitivities of the IgM ELISAs were most variable at day 3 after illness onset, ranging from 41% (for the Dengue Virus IgM Capture DxSelect ELISA) to 8% (for the CDC MAC ELISA). This difference is probably due to the design of the CDC MAC ELISA, which was originally calibrated for higher specificity for differential diagnosis between flaviviruses, rather than for high sensitivity. Importantly, when the sensitivities of the anti-DENV IgM ELISAs were compared by DENV infection status, there was no significant difference, which is in contrast to previous studies of lower sensitivities in secondary DENV infections [3] and probably reflects manufacturer improvements in test reagents (eg, antigen or monoclonal antibodies).
Use of the CDC DENV-1–4 Real-Time RT-PCR assay to analyze specimens with confirmed IgM seroconversion ≤5 days after illness onset yielded negative results in approximately 10% of cases. This sensitivity is lower than expected, given that, the CDC DENV-1-4 real-time RT-PCR assay, using prospectively collected specimens from days 1–5 after illness onset, yielded positive results in 98% of specimens in which IgM seroconversion was confirmed. Whether the difference in sensitivity between our 2 studies is due to RNA integrity from freeze-thaw cycles, test sensitivity, or some other factor could not be determined. Whether this degree of nondetection can be reduced needs to be determined, since no other studies have made this type of comparison to evaluate the performance of molecular diagnostic assays [11]; most have determined the equivalency or noninferiority of their tests to other RT-PCR tests or virus isolation [20–23].
Detection of DENV NS1 antigen correlated with the presence of DENV RNA, although there were some discrepancies that were not associated with any particular day after illness onset or the presence of anti-DENV IgM. The presence of NS1 beyond DENV RNA detection, which occurred during days 6–10 after illness onset, may be due to the longer half-life of protein antigens, compared with RNA, or to degradation of RNA during freezer storage and freeze-thaw cycles, which is especially a concern in retrospective studies [24]. Other discordant results included specimens positive by real-time RT-PCR but negative for NS1, findings that may be due to NS1–IgG anti-NS1 immune complexes that occur in persons with secondary or greater DENV infections and affect test sensitivity [25–28]. This is also the most likely cause of the higher test sensitivity observed in case patients with primary as compared to secondary DENV infections for both ELISAs. The best sensitivity of NS1 tests was observed on day 1 and 2 after onset of illness (93%) and these results were presumed to be due to low levels of IgG typically observed during that time of illness, including among persons with secondary or greater infection. Previous studies have reported much lower sensitivities for the Dengue Early ELISA, but they often reported aggregate results throughout the course of infection [29–31]. When anti-DENV IgM seroconversion was used as the reference test, both NS1 tests had higher sensitivities than real-time RT-PCR. The DENV Detect NS1 ELISA, which has not been previously evaluated, had the best overall performance for DENV detection over the 10 days of illness.
This study has several limitations. First, although the specimens were collected prospectively and initial anti-DENV IgM testing was performed at that time, final diagnostic testing was not performed until 2–8 years later, which may have introduced specimen degradation due to freeze-thaw cycles. However, given the high sensitivities observed for the respective days after infection throughout the study, this did not appear to be a significant limitation. A second limitation was our inability to conduct serotype-specific analysis of the final results, owing to the insufficient number of cases in each group. Last, this study was conducted in Puerto Rico, where there is one other known circulating flavivirus (West Nile virus) [32]. While this pattern of flavivirus cocirculation reflects the dengue situation in much of the world, this study does not quantify the degree of anti-DENV IgM test cross-reactivity and potential patient misclassification that might occur in patients with suspected dengue who only test positive for this analyte and live in or traveled from areas where other flavivirus infections occur with high frequency (eg, Japanese encephalitis in Asia). The only way to address this limitation is to conduct a similar study in these areas. Until such data become available, the interpretation of the results of single-specimen dengue diagnostic testing should account for the degree of flavivirus cocirculation in the area where the patient likely acquired their infection.
The dengue diagnostic tests used in the evaluation are widely available commercially or from the CDC in dengue-endemic countries, as well as in areas of nonendemicity where travelers return with suspect dengue. Thus, these tests should be available in diagnostic laboratories for combination testing. However, we recognize that single-specimen dengue diagnostic testing will require changes in routine procedure in many countries or regions where dengue is endemic. Furthermore, multiple tests can vary in test cost per country, and this has to be individually evaluated by country. Undertaking similar evaluation studies in these areas will ultimately establish the validity of this approach and identify problems or issues not evident in the present study. We also recognize that, in many dengue-endemic areas, the type of instrument-dependent diagnostic testing described in this study is presently not possible, which underscores the need for development of point-of-care diagnostic assays with acceptable performance characteristics [15, 19].
In conclusion, this study demonstrates the performance and potential utility of widely available dengue diagnostic tests in accurately diagnosing clinically suspected dengue in patients during the febrile and convalescent phases of the illness. This new testing paradigm eliminates the need to collect acute- and convalescent-phase specimens, simplifies the choice of diagnostic tests, and should provide timely information for clinical case management. In addition, greater case confirmation will improve public health surveillance and allow for assessments of the effectiveness of dengue prevention tools such as vaccines, vector control, improved case management practices, and antivirals. The widespread adoption and further evaluation of this single-specimen dengue testing algorithm will ultimately improve prevention and control of this major public health problem.
Notes
Acknowledgments. We thank Oscar Padro and Luis Manuel Santiago for their assistance in data acquisition, selection of specimens, and database expertise for the design of this study.
Financial support. This work was supported by the Centers for Disease Control and Prevention Dengue Branch, Division of Vector-Borne Diseases.
Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
References
Author notes
Presented in part: American Society for Tropical Medicine and Hygiene Conference, Atlanta, Georgia, 12 November 2012. Poster 1097; Third International Conference on Dengue and Dengue Hemorrhagic Fever, Bangkok, Thailand, 22 October 2013.
