The study team measured vaccine‑associated signals using assays that detect yellow fever virus nonstructural protein 1 (YFV NS1), an antigen encoded by Dengvaxia but absent from wild‑type DENV. Baseline samples were collected before vaccination; follow‑up sampling was scheduled at approximately 20 months (T1) and 40 months (T2) relative to baseline. At the T1 time point, 61% of vaccinated children met the study’s predefined criteria for vaccine responders. The investigators then applied a more sensitive Luminex dilution (1:50) for YFV NS1 to search for low‑level responses among children originally classified as non‑responders. This approach revealed additional low‑level YFV NS1 reactivity that the initial binary classification had not captured, and those low signals persisted, at least at low levels, through T2.
The authors used receiver operating characteristic analysis within the study dataset to compare known responders and uninfected unvaccinated controls; the area under the curve was 0.94. To keep classification consistent with conventional serologic practice, a ≥4‑fold increase in YFV NS1 signal (1:50 dilution, T1 relative to baseline) was applied as the working cutoff for vaccine‑associated responses. Within this dataset that cutoff corresponded to 82.7% sensitivity and 92.7% specificity versus the predefined reference groups. The manuscript notes that sensitivity analyses using the ROC‑derived 3.55‑fold threshold produced similar qualitative results.
The cohort for this analysis comprised 186 dengue‑naïve children with complete longitudinal samples: 114 who received a single Dengvaxia dose and 72 who were unvaccinated. Infection risk was assessed in two discrete intervals: baseline→T1 and T1→T2. During the first interval (baseline to T1), vaccination was associated with a reduced risk of wild‑type DENV infection, reported as vaccine effectiveness (VE) = 45%. In the second interval (T1 to T2), the investigators observed no difference in infection risk between vaccinated and unvaccinated children.
Risk ratios were estimated using binomial regression with a log link, comparing each vaccinated subgroup against DENV‑naïve unvaccinated children as the reference group. The manuscript includes both symptomatic and asymptomatic infections in these analyses and infers infecting serotype for many infections using post‑infection neutralization profiles when possible.
When breakthrough wild‑type DENV infection occurred during the baseline→T1 interval, vaccinated children mounted post‑infection antibody responses with greater overall magnitude and broader serotype recognition than unvaccinated children experiencing primary infection. The investigators interpreted these differences as consistent with vaccine‑associated immune priming that altered the character of the subsequent antibody response. By contrast, for infections that occurred in the T1→T2 interval, differences in post‑infection antibody magnitude and serotype breadth between vaccinated and unvaccinated children were no longer detected.
The authors summarized responses using cumulative neutralizing antibody titers across DENV1–4 as a magnitude metric and counted serotypes meeting assay‑specific positivity thresholds to define serotype breadth (0–4 ST). Effect sizes for comparisons used a probabilistic index (Mann–Whitney framework) with bootstrap confidence intervals.
Neutralizing antibodies to DENV1–4 were measured by a focus reduction neutralization test. Binding antibodies to ED3 and NS1 antigens (DENV1–4 and YFV NS1) were quantified with a Luminex multiplex platform. For the 1:500 dilution assays, seropositivity thresholds were calculated from measurements in an independent set of DENV‑naïve children as the mean plus three standard deviations; a ≥2‑fold increase from baseline to T1 was required for classification as seroconversion in that format. For enhanced detection of YFV NS1, a 1:50 dilution and a ≥4‑fold increase from baseline were used for the primary analyses in the manuscript.
The methods section reports how values below assay limits were handled (e.g., assigning titers <20 a value of 5 prior to log10 transformation and assigning an MFI of 1 for ED3 values below the assay positivity threshold) so that cumulative metrics could be calculated consistently across individuals.
Nonparametric paired tests (Wilcoxon signed‑rank) and Friedman tests with Dunn’s post hoc corrections were used for longitudinal comparisons. Categorical differences in serotype breadth were assessed with Fisher’s exact test. For post‑infection comparisons between vaccinated and unvaccinated groups, effect sizes used the Mann–Whitney probabilistic index with bootstrap‑estimated 95% confidence intervals. Risk ratios and their confidence intervals were derived from binomial regression models; vaccine effectiveness was calculated as (1 − RR) × 100 within each study interval. The manuscript distinguishes hypothesis‑driven primary comparisons from additional exploratory subgroup analyses.
The parent longitudinal cohort enrolled nearly 3,000 children, of whom a subset were dengue‑naïve at baseline. For the present analysis, the authors focused on the 186 dengue‑naïve children who had complete paired sera at baseline, T1 (~20 months) and T2 (~40 months), comprising 114 vaccinated and 72 unvaccinated participants. Baseline samples were collected prior to vaccination, which occurred roughly two months after enrollment. Follow‑up sampling windows varied slightly across participants; time points are therefore reported relative to baseline in the manuscript.
Sex was recorded for all participants and both males and females were included in the study. The authors report that sex was not considered as a biological variable in the analyses.
The authors acknowledge that this analysis used an analytic subset of dengue‑naïve children selected for complete longitudinal sampling through T2; children without complete sampling were not included. Sampling windows differed modestly across individuals, which complicates exact timing comparisons. The predefined binary classification of vaccine responders versus non‑responders did not capture the spectrum of low‑level vaccine‑associated signals detected with a more sensitive assay dilution. Sex was not analyzed as a biological variable in these comparisons. The manuscript leaves open several specific questions documented in the text: how long low‑level vaccine‑associated YFV NS1 responses persist beyond 40 months, the clinical relevance of low‑level YFV NS1 signals in those initially labeled non‑responders, and the immunologic mechanisms that explain the loss of the apparent protective association in the second study period. These items remain to be addressed in future analyses or complementary studies.