A Systematic Evaluation of IgM and IgG Antibody Assay Accuracy in Diagnosing Acute Zika Virus Infection in Brazil: Lessons Relevant to Emerging Infections

Abstract

Accurate diagnostics underpin effective public health responses to emerging viruses. For viruses, such as Zika virus (ZIKV), where the viremia clears quickly, antibody-based (IgM or IgG) diagnostics are recommended for patients who present 7 days after symptom onset. However, cross-reactive antibody responses can complicate test interpretation among populations where closely related viruses circulate. We examined the accuracy (proportion of samples correctly categorized as Zika positive or negative) for antibody-based diagnostics among Brazilian residents (Rio de Janeiro) during the ZIKV outbreak. Four ZIKV enzyme-linked immunosorbent assays (ELISAs; IgM and IgG Euroimmun, IgM Novagnost, and CDC MAC), two dengue ELISAs (IgM and IgG Panbio), and the ZIKV plaque reduction neutralization test (PRNT) were evaluated. Positive samples were ZIKV PCR confirmed clinical cases collected in 2015-2016 (n = 169); negative samples (n = 236) were collected before ZIKV was present in Brazil (≤2013). Among serum samples collected ≥7 days from symptom onset, PRNT exhibited the highest accuracy (93.7%), followed by the Euroimmun IgG ELISA (77.9%). All IgM assays exhibited lower accuracy (<75%). IgG was detected more consistently than IgM among ZIKV cases using Euroimmun ELISAs (68% versus 22%). Anti-dengue virus IgM ELISA was positive in 41.1% of confirmed ZIKV samples tested. The Euroimmun IgG assay, although misdiagnosing 22% of samples, provided the most accurate ELISA. Anti-ZIKV IgG was detected more reliably than IgM among ZIKV patients, suggesting a secondary antibody response to assay antigens following ZIKV infection. Antibody ELISAs need careful evaluation in their target population to optimize use and minimize misdiagnosis, prior to widespread deployment, particularly where related viruses cocirculate.

Keywords: Zika antibody assays; Zika virus; diagnosis; immunoassays; plaque reduction neutralization tests (PRNTs); serology; serology evaluation.

FIG 1

Zika antibody detected in serum samples collected during the acute (1-6 days after onset), early-convalescent phase (7–13 days) and late convalescent-phase (≥14 days) from PCR positive Zika cases measured by IgM (A) or IgG (B) NS1 anti-ZIKV ELISAs (Euroimmun). (C) IgM NS1 anti-ZIKV ELISA measurements for acute (1-6 days after onset) and convalescent (≥7 days) samples from PCR positive ZIKV cases. (D) IgG NS1 anti-ZIKV antibody levels in paired serum samples from PCR positive ZIKV positive cases. Dotted horizontal lines represent the cutoff value used in each assay. Data points above the cutoff are considered positive. Trend-line in Panels C and D represent the median antibody levels for acute and convalescent samples. Statistically significant differences between two groups were measured by Mann-Whitney U test (***P = 0.0001). Figure shows antibody ratios calculated as per manufacturers’ instructions (antibody ratio = OD sample/OD calibrator).

FIG 2

The change in (A) IgM and (B) IgG NS1 Euroimmun anti-ZIKV antibody levels between paired serum samples from PCR positive Zika cases by day of collection (days postsymptom onset) of the first (acute) sample. Based on the first sample (collected 0–7 days) and second sample (median interval between samples was 7 days). The highest IgG fold change (change in antibody level between first and second samples) was observed among paired samples collected on days 2 and 7 postsymptom onset.

FIG 3

Anti-ZIKV antibody levels in sequential serum samples collected from Zika PCR positive cases on different days postsymptom onset (0-54 days) measured in (A) IgM Euroimmun NS1 and (B) IgG Euroimmun NS1 anti-ZIKV ELISAs. Dotted line shows assay cutoffs. The figure shows more consistent detection of anti-Zika antibodies (level above the cutoff) among convalescent samples when measuring IgG compared to IgM. Ratios calculated as per manufacturers’ instruction; first collection (acute sample [closed circles]); second collection (convalescent samples [open squares]); third collection (late convalescent samples [closed triangles]).

FIG 4

Correlation between anti-ZIKV and anti-DENV antibody levels in individual sera samples. (A) IgG anti-DENV ELISA (Panbio) versus IgG anti-ZIKV NS1 (Euroimmun) ELISA. Anti-DENV and anti-ZIKV IgG antibody levels showed a positive correlation. When a patient exhibited a relatively high anti-DENV IgG antibody response they also tended to exhibit a relatively high anti-ZIKV IgG antibody response (P < 0.001; r² = 0.258; n = 168); (B) IgM DENV ELISA and IgM ZIKV NS1 ELISA antibody levels. Again, the measurements showed a positive correlation (P = 0.015; r² = 0.015; n = 166). Dotted lines show assay cutoffs. Dashed line shows the best-fitting line (Spearman rank correlation [r²]). Correlation was more significant between anti-ZIKV and anti-DENV IgG antibody measurements. The correlation in antibody measurement between ZIKV and DENV ELISAs suggests a degree of overlap in patient responses and/or cross-reaction in antibody detection.

FIG 5

The change in anti-ZIKV IgM antibody levels by day postsymptom onset in sequential sera from ZIKV PCR positive cases (n = 4). Plots A–D represent four different patients. Each patient had at least five sequential sera samples collected. Anti-ZIKV IgM was measured by NS1 (Euroimmun; shown as squares) and μ-capture N (Novagnost; shown as circles) ELISAs. Dotted lines represent cutoff values for each assay. Ratios calculated as recommended by the manufacturers. The plots display a unique pattern of ZIKV IgM antibody response over time for each patient.

FIG 6

Receiver operating characteristic (ROC) curve comparing sensitivity and specificity at different cutoff values for the anti-ZIKV IgG NS1 ELISA (n = 294 sera); (A) ROC curve; (B) specificity and sensitivity at each cutoff. The dotted line in B indicates the cutoff recommended by the manufacturer (ratio of 1.1). The accuracy of IgG NS1 ELISA was 77.9% using the manufacturer’s cutoff. Higher accuracy was observed when the cutoff was increased to 1.5 (where the curves intersect on plot B). Using this cutoff, the ELISA had an accuracy of 81.0%. Sensitivity and specificity were 78.9 and 82.2%, respectively.

FIG 7

Diagram representing the different patterns of antiviral IgG and IgM antibody responses and viral RNA detection observed in sera from flavivirus infected individuals over days from symptom onset among (A) virus naive and (B) previously exposed individuals. The cartoon exhibits a more prominent IgG response compared to IgM among individuals previously exposed to the virus. In our current study, we observed a more prominent anti-ZIKV IgG compared to IgM response (see Fig. 3).