Adapting Rapid Diagnostic Tests to Detect Historical Dengue Virus Infections

Abstract

The only licensed dengue vaccine, Dengvaxia^®, increases risk of severe dengue when given to individuals without prior dengue virus (DENV) infection but is protective against future disease in those with prior DENV immunity. The World Health Organization has recommended using rapid diagnostic tests (RDT) to determine history of prior DENV infection and suitability for vaccination. Dengue experts recommend that these assays be highly specific (≥98%) to avoid erroneously vaccinating individuals without prior DENV infection, as well as be sensitive enough (≥95%) to detect individuals with a single prior DENV infection. We evaluated one existing and two newly developed anti-flavivirus RDTs using samples collected >6 months post-infection from individuals in non-endemic and DENV and ZIKV endemic areas. We first evaluated the IgG component of the SD BIOLINE Dengue IgG/IgM RDT, which was developed to assist in confirming acute/recent DENV infections (n=93 samples). When evaluated following the manufacturer's instructions, the SD BIOLINE Dengue RDT had 100% specificity for both non-endemic and endemic samples but low sensitivity for detecting DENV seropositivity (0% non-endemic, 41% endemic). Sensitivity increased (53% non-endemic, 98% endemic) when tests were allowed to run beyond manufacturer recommendations (0.5 up to 3 hours), but specificity decreased in endemic samples (36%). When tests were evaluated using a quantitative reader, optimal specificity could be achieved (≥98%) while still retaining sensitivity at earlier timepoints in non-endemic (44-88%) and endemic samples (31-55%). We next evaluated novel dengue and Zika RDTs developed by Excivion to detect prior DENV or ZIKV infections and reduce cross-flavivirus reactivity (n=207 samples). When evaluated visually, the Excivion Dengue RDT had sensitivity and specificity values of 79%, but when evaluated with a quantitative reader, optimal specificity could be achieved (≥98%) while still maintaining moderate sensitivity (48-75%). The Excivion Zika RDT had high specificity (>98%) and sensitivity (>93%) when evaluated quantitatively, suggesting it may be used alongside dengue RDTs to minimize misclassification due to cross-reactivity. Our findings demonstrate the potential of RDTs to be used for dengue pre-vaccination screening to reduce vaccine-induced priming for severe dengue and show how assay design adaptations as well quantitative evaluation can further improve RDTs for this purpose.

Keywords: Zika; dengue; dengue vaccine; pre-vaccination screening; rapid diagnostic test.

Figure 1

Sequence alignment of the highly conserved fusion loop region of the E proteins for distinct flaviviruses compared to the Excivion DENV1-4 and ZIKV fusion loop-glycosylated antigens. (A) Sequence homology between ZIKV, DENV1-4, West Nile virus, Japanese encephalitis virus, and yellow fever virus as previously described (15). Sequences of DENV1-4 (B) and ZIKV (C) used for the Excivion Dengue and Zika RDTs, with position of the N-linked glycosylation sequon boxed in black.

Figure 2

Evaluation of the IgG component of the SD BIOLINE Dengue IgG/IgM RDT using samples from the Oregon traveler cohort. (A, B) Visual readings for IgG band positivity evaluated in two independent experiments (average consensus value shown as individual dots) at multiple timepoints. Medians and interquartile ranges are shown as summary statistics. Data are plotted separately for (A) DENV-naïve (flavivirus naïve or yellow fever vaccine recipients) and ZIKV+ (prior ZIKV infection but not DENV infection) samples and (B) DENV+ primary (primary DENV infection with one of the four serotypes) and DENV+ secondary (multiple prior DENV infections or prior DENV and ZIKV infections). Corresponding specificity and sensitivity values (dots) are shown for analyses using all DENV-positive samples (blue) versus only primary DENV-positive samples (pink) with 95% confidence intervals (vertical bars) for each timepoint. Smoothing spline curves are used to visualize trends over time. (C, D) Equivalent to A and B, but with quantitative evaluation of RDT tests performed using R, measured as IgG band pixel intensity. Data again are grouped by infection history and timepoint. Specificity was set to the optimal value (≤98%) or minimum value (≤90%) recommended by GDAC, although for this dataset, the only specificity ≤90% was 100% and thus the same value is shown for both analyses; corresponding sensitivity values are shown.

Figure 3

Evaluation of the IgG component of the SD BIOLINE Dengue IgG/IgM RDT using samples from the EcoDess cohort in Esmeraldas, Ecuador. (A, B) Average visual readings for IgG band positivity evaluated in one experiment by three separate analysts at multiple timepoints (dots show individual points, groups are summarized by medians and interquartile ranges). Data are plotted separately for (A) DENV-naïve and ZIKV+ only samples and (B) DENV-positive only with either primary or secondary DENV immunity, or DENV+ and ZIKV+ samples. Corresponding specificity and sensitivity values (dots) are shown for analyses including all DENV-positive samples and only primary DENV-positive samples, with 95% confidence intervals (vertical bars) for each timepoint. Specificity and sensitivity values are shown to the right, with estimates fitted using smoothing splines to visualize trends. (C, D) Equivalent to (A, B), but with quantitative evaluation of RDT tests using R, measured as IgG band pixel intensity. Specificity was set to optimum (≥98%) and minimum (≥90%) values, and corresponding sensitivity values are shown.

Figure 4

Evaluation of the Excivion Dengue IgG RDT using samples from the Nicaragua Pediatric Dengue Cohort and Hospital-based Studies. (A, B) Visual evaluation of the IgG band positivity in one experiment by two separate analysts (individual dots). Data are grouped by DENV and ZIKV infection history (medians and interquartile ranges), separated by (A) DENV-naïve and ZIKV+ samples and (B) DENV+ primary and DENV+ secondary samples (including multiple prior DENV and/or ZIKV infections). Corresponding specificity and sensitivity values (dots) are shown for both analyses of all DENV-positive samples versus only primary DENV-positive samples. Vertical bars show 95% confidence intervals for each timepoint. (C, D) Equivalent to A and B, but with quantitative evaluation of RDT tests using the ChemBio Cube lateral flow reader, measured as IgG band pixel intensity at multiple timepoints. Simple linear regressions are used to visualize trends over time. Optimal (≥98%) and minimum (≥90%) specificity and corresponding sensitivity values are shown.

Figure 5

Evaluation of the Excivion Dengue IgG RDT using samples from the ECUAVIDA Study. (A, B) Quantitative evaluation of RDT tests using the ChemBio Cube lateral flow reader, measured as IgG band pixel intensity (medians and interquartile ranges are shown as summary statistics). Data are plotted separately for (A) DENV-seronegative and (B) DENV-seropositive samples as classified by Dengue Abcam ELISA. Specificity and sensitivity values (dots) with 95% confidence intervals (vertical bars) are shown at a single timepoint, with specificity set to either ≥98% or ≥90%. Corresponding sensitivity values are shown.

Figure 6

Evaluation of the Excivion Zika IgG RDT using samples from the Nicaragua Pediatric Dengue Cohort and Hospital-based Studies. (A, B) Average visual readings for IgG band positivity evaluated in one experiment by two separate analysts (individual dots) at single timepoint (medians and interquartile ranges are shown). Analyses are grouped by (A) DENV+ samples (primary DENV infection with one of the four serotypes or multiple prior DENV infections) and (B) ZIKV+ samples (prior ZIKV infection but not DENV infection). Corresponding specificity and sensitivity values (dots) with 95% confidence intervals (vertical bars) are shown. (C, D) Equivalent to A and B, but with quantitative evaluation of RDT tests using the ChemBio Cube lateral flow reader, measured as IgG band pixel intensity at multiple timepoints. Optimal (≥98%) and minimum (≥90%) specificity and corresponding sensitivity values are shown. Simple linear regressions show temporal trends.