Antibody-Dependent Enhancement of Severe Disease Is Mediated by Serum Viral Load in Pediatric Dengue Virus Infections

Abstract

Background

Low preexisting anti-dengue virus (DENV) antibody levels are associated with elevated disease severity. While antibody-dependent enhancement of dengue is thought to be driven by viral load, this has not been conclusively shown. We evaluated the association between preinfection anti-DENV antibody titers, viral load, and disease severity among 133 dengue cases in a Nicaraguan pediatric cohort study.

Methods

Viral load was quantified in acute-phase serum by real-time reverse transcription polymerase chain reaction and analyzed in relation to preinfection antibody titer (measured by inhibition enzyme-linked immunosorbent assay) and dengue severity, categorized using 3 definitions.

Results

Higher viral load was significantly associated with dengue severity; for each increase of 1.0 log₁₀ copies/mL, the odds of severe dengue increased approximately 50%, regardless of severity definition. Viral load at presentation and the odds of severe disease were highest among patients with low to intermediate preinfection antibody titers and lowest among those with the highest antibody titers. We showed the effect of preinfection antibody titer on disease severity was mediated by viral load for each of 3 dengue severity outcomes.

Conclusions

This study demonstrates the association between preinfection anti-DENV antibody titer, serum viral load, and disease severity, and provides evidence for the mechanism of antibody-dependent enhancement in dengue cases.

Low preexisting anti-dengue virus antibody levels are associated with elevated disease severity. The current study demonstrates that this effect is mediated by viral load, providing evidence for the mechanism of antibody-dependent enhancement in dengue.

Dengue is the most common human arboviral disease worldwide. An estimated 50–100 million dengue cases occur annually, resulting from infections with 1 of 4 related dengue virus serotypes (DENV-1-4) [1, 2]. Clinical manifestations range from a mild, self-limited febrile illness to severe disease and death [1–4]. A number of factors may contribute to dengue outcomes; however, host immunity is the most important predictor of symptomatic infection and disease severity [1, 5–9]. Primary DENV infections elicit neutralizing antibody responses that are initially cross-reactive against heterotypic serotypes (approximately 6 months to 1 year), and protection against symptomatic infections is observed for a period of approximately 1–2 years [4, 10]. Heterotypic antibodies at subneutralizing levels are thought to enhance secondary infections with a different DENV serotype. However, antibody-enhanced disease has only recently been conclusively shown to occur in human cases [7, 8].

In 2 large cohort studies in Nicaragua and Thailand, the risk of developing severe dengue was significantly higher among individuals with low to intermediate preinfection anti-DENV antibody titers compared to those with either very low or high titers [7, 8]. In Nicaragua, the greatest risk of developing severe dengue occurred among patients with preinfection antibody titers of 1:21 to 1:80 as measured by inhibition enzyme-linked immunosorbent assay (iELISA) [8]. The effect remained significant regardless of the definition of dengue severity: dengue hemorrhagic fever (DHF)/dengue shock syndrome (DSS), from the 1997 World Health Organization (WHO) guidelines [11]; severe dengue, from the 2009 WHO guidelines [2]; or hospitalization. Additional data in support of antibody-enhanced disease come from trials of the dengue vaccine, Dengvaxia, where dengue-naive vaccine recipients were at increased risk for subsequent hospitalization with virologically confirmed dengue [12].

The mechanism by which specific preinfection anti-DENV antibody titers increase disease severity is termed antibody-dependent enhancement (ADE) [13, 14]. ADE had been demonstrated in cultures of Fcγ receptor-bearing cells [15] and animal models [16–18], and this effect served to explain high rates of severe infection among neonates with primary DENV infections [19]. However, direct evidence of ADE in human dengue cases has remained elusive. ADE is posited to occur through the facilitation of DENV entry and replication in Fcγ receptor-bearing target cells in the presence of non- or subneutralizing antibody titers. In human cases, this should result in higher serum viral loads, leading to immune activation and increased levels of circulating nonstructural protein 1 that ultimately result in plasma leakage and shock [19–21]. High serum viral load has been independently associated with DHF/DSS and severity in previous studies [14, 22–29], but published data have predominantly come from the study of acute infections with little or no access to preinfection specimens or prior infection history. The few reports that have evaluated viral load and disease severity in the context of preinfection antibodies, measured using ADE assays in vitro, have thus far produced differing results [15, 30–32]. To date, it has not been conclusively demonstrated that preinfection antibody titer is associated with disease severity in human dengue cases through increases in viral load.

The objective of the current study was to characterize the association between DENV viral load and preinfection antibody titer among dengue cases in the Pediatric Dengue Cohort Study (PDCS), which is a community-based, prospective cohort study established in 2004 and based in Managua, Nicaragua. This research builds upon the aforementioned findings of increased risk for severe infection at low to intermediate anti-DENV antibody titers measured using the DENV iELISA [8]. Here, we use quantitative molecular testing to definitively establish the relationship between preinfection antibody titers, viral load, and dengue severity.

METHODS

Ethics Statement

The study protocol was reviewed and approved by the institutional review boards of the Nicaraguan Ministry of Health, the University of California, Berkeley, and Emory University. Written informed consent was obtained from a parent or guardian, and verbal assent was obtained from children aged 6 years and older.

Study Population and Clinical Information

The database of all acute dengue cases in the PDCS from 2005 to 2014 was queried to identify patients with DHF/DSS and dengue fever (DF). For inclusion in the current analysis, individuals had to have an acute-phase serum sample available and had to have been reverse transcription polymerase chain reaction (RT-PCR)-positive at presentation. All patients who met these criteria and developed DHF/DSS were initially included in the study. One patient, who had indeterminate serostatus, was removed from analysis (n = 38 DHF/DSS cases in final analyses). A subset of DF cases (n = 95) was selected to provide a random distribution of DENV serotypes and a 1:2 ratio of primary to secondary cases (Table 1). The ratio of primary to secondary DF cases was selected to maximize power given the limited number of secondary DF cases for which the viral load could be quantified and provide a comparison group for severe primary dengue cases. Clinical information was obtained from the standardized study forms, which have been described previously [3, 33].

Table 1.

Patient Data and DENV History for 133 Dengue Cases in the PDCS (2005–2014)

Factor	DHF/DSS	DF	DwWS/SD	DwoWS	Hospitalized	Not Hospitalized
Cases, total	38 (100)	95 (100)	60 (100)	73 (100)	65 (100)	68 (100)
Sex, female	19 (50.0)	49 (51.6)	32 (53.3)	36 (49.3)	35 (53.9)	33 (48.5)
Age, y, mean (SD)	9.8 (3.0)	9.0 (3.1)	9.5 (3.3)	9.0 (2.9)	9.5 (3.3)	9.0 (2.9)
Day of fever, mean (SD)	2.0 (1.2)	2.0 (1.1)	2.0 (1.1)	2.0 (1.1)	2.0 (1.2)	2.0 (1.1)
Primary DENV infection	7 (18.4)	35 (36.8)	14 (23.3)	28 (38.4)	17 (26.2)	25 (36.8)
Secondary DENV infection	31 (81.6)	60 (63.2)	46 (76.7)	45 (61.6)	48 (73.9)	43 (63.2)
Number of previous infectionsa
1	24 (80.0)	47 (79.7)	36 (81.8)	35 (77.8)	38 (82.6)	33 (76.7)
2	6 (20.0)	9 (15.3)	8 (18.2)	7 (15.6)	8 (17.4)	7 (16.3)
>2	…	3 (5.1)	…	3 (6.7)	…	3 (7.0)
Serotype
DENV-1	5 (13.6)	24 (25.2)	11 (18.3)	18 (24.7)	13 (20.0)	16 (23.5)
DENV-2	10 (26.3)	19 (20.0)	14 (23.3)	15 (20.6)	14 (21.5)	15 (22.1)
DENV-3	23 (60.5)	50 (52.6)	35 (58.3)	38 (52.1)	38 (58.5)	35 (51.5)
NA	…	2 (2.1)	…	2 (2.74)	…	2 (2.94)

Data are n (%) unless otherwise noted.

Abbreviations: DENV, dengue virus; DF, dengue fever; DHF, dengue hemorrhagic fever; DSS, dengue shock syndrome; DwoWS, dengue without warning signs; DwWS/SD, dengue with warning signs/severe dengue; NA, not available; PDCS, Pediatric Dengue Cohort Study.

^aSecondary cases with prior DENV infections captured in the PDCS; 4 patients entered the study non-naive and were excluded from this calculation.

Serological Testing

Healthy annual serum samples were collected in July of 2005–2010 and in March/April of 2011–2014. Annual samples were tested for total anti-DENV antibody using an iELISA, as described previously [8, 34]. Each sample was tested at an initial dilution of 1:10 and was considered positive if the percentage of inhibition was ≥50% relative to negative controls. Positive samples were then tested at four 10-fold serial dilutions from 1:10 to 1:10 000; titers were estimated using the method of Reed and Munch [35]. Preinfection iELISA titers were used for all analyses. These were obtained during annual sampling between dengue seasons and preceding the dengue case in the current study. Dengue cases were categorized as primary or secondary based on: (1) documented DENV infections in the PDCS for a given patient and/or (2) iELISA in the convalescent sample from the DENV infection included in the current analysis (primary infection, convalescent iELISA titer <1:2560; secondary infection, convalescent titer ≥1:2560) [36].

Molecular Testing

All samples had been previously tested with a qualitative DENV RT-PCR [37]. For the current study, total nucleic acids were extracted from 75 µL of archived, acute-phase serum (or all available serum) using an eMAG instrument (bioMérieux) and the manufacturer recommended protocol. Eluates were immediately tested in a quantitative, serotype-specific real-time RT-PCR (rRT-PCR) for DENV, performed as described [38]. Each run included a 4-point standard curve for the expected serotype, positive controls for the remaining serotypes, and a no-template control. Viral load was quantified based on the standard curve and expressed as log₁₀ copies/mL of serum. Samples that tested negative in the serotype-specific DENV rRT-PCR were reflexed to testing in an internally controlled, pan-DENV rRT-PCR. This assay was performed as previously described for the qualitative detection of DENV RNA [39], and each run included positive controls for each DENV serotype, RNase P (internal control), and a negative control. Throughout this manuscript, the term viremia will be used in reference to the qualitative presence/absence of DENV in serum; viral load will be used in relation to the quantification of viremia by rRT-PCR.

Statistical Analysis

GraphPad version 8.0.1 was used to test associations between viral load at presentation, day of fever, primary/secondary infection, and severity by 2-way ANOVA. All other statistical analysis was performed using SAS version 9.4 (SAS Institute). For each dengue case, severity was classified according to: (1) the 1997 WHO guidelines (DF vs DHF/DSS); (2) the 2009 WHO guidelines (dengue without warning signs vs dengue with warning signs/severe dengue [DwWS/SD]); and (3) hospitalization status. A multivariable model for each of these 3 outcomes was developed using a domain-based approach including 10 unique domains organized according to symptoms, signs, and management (see Supplementary Material). The final model for each dengue severity outcome included any variables significant in the individual domains at α = .05 plus the following 5 variables, which were a priori forced into the 3 models: age, sex, DENV serotype, preinfection antibody titer, and viral load.

A linear regression model with viral load as the outcome variable was evaluated using preinfection antibody titer and day of fever as potential predictors. Unless otherwise noted, primary infection was used as the reference population for models that included preinfection titer as a variable. Preinfection titer was included as a categorical variable: <1:21, primary; <1:21, secondary; 1:21 to 1:80; 1:81 to 1:320; and >1:320. The interaction term of preinfection antibody titer and day of fever was considered but was not statistically significant and thus not included in the final viral load model.

For each dengue severity outcome, mediation analysis was conducted to determine if the association between preinfection antibody titer and dengue severity was mediated by viral load. There were 4 steps in the mediation analysis. In the first step, the association between the causal variable (preinfection antibody titer) and the outcome (severity) was evaluated through logistic regression, which is appropriate for a binary outcome variable; in the second step, the association between preinfection antibody titer and the mediating variable (viral load) was evaluated through simple linear regression, which is appropriate as viral load is a continuous outcome measure. If either of the associations in the first 2 steps was nonsignificant, we would conclude no mediation. In the third step, viral load was evaluated as a predictor of dengue severity while controlling for preinfection antibody titer. The fourth step evaluated the significance of preinfection antibody titer in the model in step 3: a nonsignificant effect yielded a conclusion of full mediation by viral load while a significant effect yielded a conclusion of partial mediation by viral load. We reached a conclusion of whether full, partial, or no mediation existed for each of the 3 severity outcomes [40].

RESULTS

One hundred and thirty-three dengue cases were included in the study population (Table 1). Cases occurred from August 2005 to October 2014 and were initially selected based on severity according to the 1997 WHO dengue guidelines [11]: DHF/DSS (38 cases, 28.6%) and DF (95 cases, 71.4%). To evaluate the generalizability of our findings with other systems for classifying disease severity, cases were also categorized based on the 2009 WHO dengue guidelines [2] and hospitalization status. Severity was evaluated as a dichotomous variable for each system; Table 1 shows the breakdown of cases in each category. Forty-two patients (31.6%) had primary DENV infections; 91 (68.4%) had secondary infections, including 18 individuals (13.5%) with 2 or more documented previous DENV infections. DENV serotype was confirmed for 131/133 cases. Two cases did not test positive in the serotype-specific, DENV multiplex assay; DENV viremia was confirmed, but could not be quantified, in the internally controlled pan-DENV rRT-PCR. These 2 cases were removed from statistical analyses that included viral load.

Viral Load and Dengue Severity

Mean viral load at presentation was 7.84 log₁₀ copies/mL serum (SD, 1.70), and the range of quantified values was 2.79–10.90 log₁₀ copies/mL serum. Viral load at presentation was similar for infections with DENV-1, -2, and -3 (Figure 1A), and the distribution of viral loads among primary and secondary cases did not differ (Figure 1B). The highest viral loads were observed among patients who presented on day 1 or day 2, and viral load had declined significantly by day 4–5 (Figure 1C and Supplementary Figure 1).

Figure 1.

Serum viral load of dengue virus (DENV) at presentation based on (A) DENV serotype, (B) primary or secondary infection, and (C) day of fever at presentation. Bars indicate median ±95% confidence interval. *, P < .05; **, P < .01.

Ten domain-based multivariable models, including 80 unique variables, were evaluated for each disease severity outcome. Predictors that remained significant in the final models as well as those variables that were forced into each model are shown in Table 2. Viral load was significantly higher among patients with more severe disease (Figure 2), and this was the only predictor that remained significant in all models of disease severity. For each 1.0 log₁₀ copy/mL increase in viral load, the odds of having a severe case increased 40%–59% depending on the classification system (Table 2). Viral load remained a significant predictor of severity when day of fever at presentation and preinfection antibody titer were forced into the models. To identify a quantitative threshold that may inform patient triage decisions, we evaluated the odds ratio of severe disease among patients with viral loads between 6.0 and 9.0 log₁₀ copies/mL. The odds of a severe case were significantly elevated, regardless of the classification system, among patients who presented with a viral load ≥8.0 log₁₀ copies/mL of serum compared to those who had a viral load below this threshold (odds ratio, 3.4–3.5; Supplementary Table 1).

Table 2.

Multivariable Analysis of Factors Associated With Dengue Severity Among Pediatric Cases in the PDCS

	DHF/DSS		DwWS/SD		Hospitalization
Effect	OR (95% CI)	P Value	OR (95% CI)	P Value	OR (95% CI)	P Value
Age, y	1.13 (.95–1.35)	.167	1.08 (.92–1.27)	.344	1.06 (.90–1.26)	.489
Sex
Male	Reference		Reference		Reference
Female	1.03 (.42–2.52)	.948	1.14 (.51–2.56)	.747	1.16 (.52–2.59)	.711
Serotype
DENV-1	Reference		Reference		Reference
DENV-2	2.56 (.55–12.00)	.903	0.98 (.26–3.64)	.769	1.21 (.33–4.47)	.819
DENV-3	5.64 (1.40–22.75)	.021	1.33 (.47–3.78)	.488	1.89 (.66–5.45)	.211
Antibody titer (preinfection)
<1:21 Primary	Reference		Reference		Reference
<1:21 Secondary	1.54 (.38–6.21)	.765	1.08 (.33–3.52)	.580	1.28 (.39–4.24)	.945
1:21 to 1:80	3.46 (.93–12.83)	.108	5.31 (1.63–17.28)	.001	3.78 (1.15–12.51)	.009
1:81 to 1:320	1.51 (.30–7.72)	.760	1.25 (.33–4.74)	.832	1.34 (.34–5.27)	.981
>1:320	2.22 (.33–14.88)	.736	0.70 (.13–3.86)	.267	0.63 (.12–3.40)	.222
Viral load, log10 copy/mL	1.59 (1.07–2.36)	.021	1.44 (1.10–1.88)	.007	1.40 (1.07–1.83)	.014
Pharyngeal erythema	3.44 (1.20–9.84)	.021	NE	NE	NE	NE
Receipt of oral rehydration	0.17 (.05–.57)	.004	NE	NE	NE	NE
Heart rate, beats/min	NE	NE	1.04 (1.01–1.08)	.014	NE	NE
Respiratory rate/min	NE	NE	NE	NE	1.32 (1.06–1.65)	.015
Retro-orbital pain	NE	NE	NE	NE	2.75 (1.04–7.23)	.041

Values in bold were significant in multivariable analysis for one or more severity outcomes.

Abbreviations: DHF/DSS, dengue hemorrhagic fever/dengue shock syndrome; DwWS/SD, dengue with warning signs/severe dengue; NE, variable was not eligible for that model.

Figure 2.

Serum viral load of dengue virus is higher at presentation among patients with severe disease. Viral load for severe (▲) and nonsevere (●) cases based on 3 classification systems. Bars indicate medians ±95% confidence interval; *** P < .001 for the indicated pair-wise comparisons. Abbreviations: DF, dengue fever; DHF/DSS, dengue hemorrhagic fever/dengue shock syndrome; DwoWS, dengue without warning signs; DwWS/SD, dengue with warning signs/severe dengue.

DENV-3 was associated with increased odds of DHF/DSS versus DF as compared to DENV-1, but this relationship was not observed for the other classification systems. Additional predictors were identified in the multivariable analysis using specific outcome variables: pharyngeal erythema with DHF/DSS, heart rate with DwWS/SD, and respiratory rate and retro-orbital pain with hospitalization. Receipt of oral rehydration fluid was protective against DHF/DSS. However, none of these factors was associated with severity in more than 1 model (Table 2).

Preinfection Anti-DENV Antibody Titer and Viral Load

The odds of having a severe dengue case were significantly elevated among patients with preinfection anti-DENV titers of 1:21 to 1:80 relative to individuals with a primary infection (Table 2), and viral load was significantly associated with preinfection antibody titer (P = .04, linear regression; Figure 3D). Consistently and uniquely, patients with preinfection antibody titers of 1:21 to 1:80 and DHF/DSS or DwWS/SD had significantly higher viral loads than patients with primary, nonsevere cases (Figure 3A–3C). Among secondary cases, significantly lower viral loads were observed among patients with titers >1:320 (Supplementary Figure 2). This finding remained significant in a separate multivariable analysis performed to control for day of fever at presentation (Supplementary Table 2). Finally, there was no association between preinfection antibody titer and the duration of time from measurement of the titer and day of fever onset (mean, 4.2 months; SD, 2.3; Supplementary Figure 3).

Figure 3.

The effect of preinfection antibody titer on dengue severity is mediated by viral load. A–C, Serum viral load at presentation for severe (▲) and nonsevere (●) dengue cases based on status of infection (primary/secondary) and preinfection antibody titer (secondary cases). Results are displayed for 3 categorization systems: (A) DF versus DHF/DSS; (B) DwoWS versus DwWS/SD; and (C) not hospitalized versus hospitalized. Bars indicate medians ±95% confidence interval. Primary, nonsevere cases were used as the control group for all comparisons. * P < .05; ** P < .01 for the indicated pair-wise comparisons. D, Results of mediation analysis. The effect of preinfection anti-DENV antibody titer on disease severity was fully or partially mediated by viral load. Arrows demonstrate significant relationships; black line represents the previously documented association between preinfection antibody titers and antibody disease severity [9]. Abbreviations: DENV, dengue virus; DF, dengue fever; DHF/DSS, dengue hemorrhagic fever/dengue shock syndrome; DwoWS, dengue without warning signs; DwWS/SD, dengue with warning signs/severe dengue; Hosp, hospitalized.

Mediation of Preinfection Antibody Titer by Viral Load

Mediation analysis was performed for each severity outcome to test our primary hypothesis that higher viral loads represent a mechanism by which anti-DENV antibody titers increase severity risk (Figure 3D and Supplementary Tables 3–5). The significance of the relationship between preinfection antibody titer and dengue severity in the PDCS has been previously established with a larger number of cases [8]. In the smaller sample of cases tested in this manuscript using logistic regression (mediation analysis step 1), the relationship between preinfection antibody titer and dengue severity was significant for both DwWS/SD (P = .01) and hospitalization (P = .04), and near significance for DHF/DSS (P = .08) (Supplementary Tables 3–5). Together, these results were considered clinically significant and subsequent steps in mediation analysis were performed. In regression analyses (steps 2 and 3), the relationships between (1) preinfection antibody titer and viral load (Figure 3D, left-sided red arrow) and (2) viral load and dengue severity (Figure 3D, right-sided red arrow) were significant (P = .04 and P ≤ .02, respectively, for each dengue severity outcome). After adjusting for viral load (step 4), the relationship between preinfection antibody titer and severity was no longer significant for DHF/DSS (P = .17) or hospitalization (P = .14), and the relationship was of borderline significance for DwWS/SD (P = .05). Taken together, these data are consistent with mediation of the effect of preinfection antibody titer on dengue severity by viral load.

DISCUSSION

The current study provides further evidence for ADE in human dengue cases and demonstrates that preinfection anti-DENV antibody titers correlate with viral load and in turn with dengue severity. Consistent with previous reports, patients in our study population with low to intermediate preinfection antibody titers (1:21 to 1:80) had a significantly higher risk of severe disease compared to those with very low or high preinfection antibody titers [7, 8]. The highest viral loads were detected among patients with severe disease and low to intermediate preinfection antibody titers, and the lowest viral loads were observed among those with the highest preinfection titers. Viral load was significantly and independently associated with disease severity among all cases. Finally, formal mediation analyses were performed to evaluate the model in which the effect of preinfection antibody titer (the causal variable) on disease severity is fully or partially mediated by the viral load (the mediator or process variable). We showed that the effect of antibody titer on disease severity was fully mediated by viral load for 2 of 3 severity outcomes and partially mediated for the third.

Prior data exist from a Thai pediatric dengue cohort on the relationship between quantitative preinfection antibody titer, dengue viremia, and disease severity [30, 31]. In 2004, Endy et al reported that higher preinfection neutralizing antibody titers against DENV-3 were protective against DHF during a subsequent DENV-3 infection and associated with lower viral loads [30]. This relationship was not observed in DENV-2 infections [30]. However, this study had a small number of DHF cases with quantitative viral load results (n = 7), which may have led to these differing results. In a subsequent study, Laoprasopwattana et al studied preexisting ADE activity in cell culture (K562 cells) among children who developed subsequent DENV-2 or DENV-3 infections [31]. Although ADE in K562 cells was not associated with disease severity or viremia, this cell culture system has not been shown to predict ADE for human infections in any study [15, 31, 32, 41, 42]. Previously, Kliks et al had shown that ADE of DENV-2 infection using neat sera in human monocytes was associated with severe symptomatic secondary DENV infections [15].

In multivariable models from the current study, viral load remained a consistent predictor of severe disease when preinfection antibody titer or DENV infection history (primary/secondary) were included in the models, and the risk increased approximately 50% for each 1.0 log₁₀ copy/mL increase in serum viral load, regardless of the definition of severity (Table 2). In addition, there was a consistent and significant difference in acute-phase viremia between severe and nonsevere cases (Figure 2). The relatively narrow differential between viral loads in severe and nonsevere cases (difference between medians ≤1.0 log₁₀ copy/mL) may indicate a threshold for the development of severe disease in human cases, as has been seen in animal models [43]. Taken together, these findings suggest that there is a conserved pathway for the development of severe dengue, and while the likelihood of severe disease is greater among patients with low to intermediate preinfection antibody titers, severe cases occur in relation to the viral burden in a given infection, whether or not the infection was enhanced by preexisting antibodies.

This study provides a systematic evaluation of the viral load among dengue cases in the PDCS. Previously, our group quantified viral load among febrile children with DENV infections who were clinically diagnosed with a nondengue illness. Among those children, we observed 2 populations: (1) patients with low viral loads (4.49 log₁₀ copies/mL serum; SD, 1.64) who did not develop long-lived antibody responses following the infection; and (2) patients with viral loads similar to those observed in the current study (7.63 log₁₀ copies/mL serum; SD, 1.54), who developed the expected increase in antibody titers at the annual sample [3]. The range of viral load among dengue cases in the PDCS, including those from the current study and the latter population from our previous research, is similar to published data from large cohort studies in Thailand and Vietnam [26, 30, 44]. Previous observations have indicated that incidence of dengue severity and hospitalization is greater in Southeast Asia compared to Latin America [45, 46]. Our analysis demonstrates that the apparent difference in severity between geographic regions cannot be explained by differences in the distribution of viral loads between regions. The force of infection, age, time interval between DENV infections, and the introduction of new or fitter DENV strains may impact disease outcomes [4, 47, 48]. In addition, cellular host immune responses may help control DENV infections, leading to decreased viral load and the prevention of ADE or symptomatic infections altogether [49, 50]. These factors could not be evaluated in the current study but understanding their relative contributions to disease outcome is necessary for developing risk estimates for a population.

One limitation was that sequential samples were not available from individual patients for this study. As such, viral load was evaluated in aggregate over the days after fever onset and peak values could not be determined. The absence of a peak viral load may have limited the power of this study to fully characterize the interplay between preinfection antibody titer, viral load, and severity, as peak values are significantly higher among severe cases and secondary infections [14, 30, 44].

In conclusion, this study demonstrates the association between preinfection anti-DENV antibody titer, serum viral load, and disease severity. Viral load remained a strong predictor of severe disease in all models and mediated the effect of preinfection antibody titer on severity. These data are consistent with a contribution of ADE to dengue severity in human cases and the existence of subneutralizing antibody titers that confer increased risk for more severe disease.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. We thank members of the study teams at the Centro de Salud Sócrates Flores Vivas, the National Virology Laboratory in the Centro Nacional de Diagnóstico y Referencia, and the Sustainable Sciences Institute in Nicaragua for their dedication and high-quality work, and we are grateful to the study participants and their families.

Financial support. This work was supported by the National Institutes of Health (NIH) and National Institute of Allergy and Infectious Diseases (NIAID) (grant numbers K08AI110528 to J. J. W. and P01AI106695 to E. H.); the Pediatric Dengue Cohort Study was supported by the NIH (grants P01AI106695 and U19AI118610 to E. H. and R01AI099631 to A. B.); and the Bill and Melinda Gates Foundation (Pediatric Dengue Vaccine Initiative grant VE-1 and FIRST grant to E. H.).

Potential conflicts of interest. All authors: No reported conflicts of interest. 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.