BGPT: Paper Review: Efficacy and Long-Term Safety of a Dengue Vaccine in Regions of Endemic Disease

Explore by Goal

Quick Explanation Copied

Dengue vaccine (CYD-TDV) long-term follow-up: protection is clearer in older children, with an unresolved year-3 hospitalization imbalance in younger children

In NEJM’s integrated long-term safety/effectiveness analysis (CYD14, CYD15, CYD23/57), pooled year-3 hospitalization for virologically confirmed dengue shows no statistically clear overall difference (RR 0.84; 95% CI 0.56–1.24), while a substantial apparent benefit in ≥9-year-olds (pooled RR 0.50; 95% CI 0.29–0.86) contrasts with an imbalance in <9-year-olds (pooled RR 1.58; 95% CI 0.83–3.02). The paper emphasizes ongoing monitoring because the younger-age imbalance remains unexplained.

Key study anchor:

Long Explanation

Paper Review (visual-first): CYD-TDV long-term safety & efficacy in endemic regions

Target paper: Hadinegoro et al., NEJM (2015)—long-term follow-up interim analyses + integrated efficacy from CYD14, CYD15, and CYD23/57.

Core scientific question: Does vaccination alter long-term risk of dengue severe outcomes—operationalized via hospitalization for virologically confirmed dengue—while maintaining efficacy against symptomatic dengue?

Key endpoints used

Safety surrogate: hospitalization for virologically confirmed dengue during long-term follow-up years (focus on year 3; year 4 for CYD57).
Severity adjudication: severe dengue classified using independent data monitoring committee criteria; and dengue hemorrhagic fever (WHO 1997 criteria) among hospitalized cases.
Efficacy: vaccine efficacy against symptomatic virologically confirmed dengue during the first 25 months of surveillance; plus serotype- and age/serostatus-stratified patterns.

FIGURE 1 — Year-3 hospitalization counts and relative risk by trial (virologically confirmed dengue)

From the paper’s year-3 counts (vaccine vs control): CYD14 + CYD15 + CYD23/57 (year 3 portion). RR computed as annual incidence rate ratio consistent with paper’s RR framing (note: RRs in the paper are already pooled; here we visualize raw counts and compute approximate RR from incidence-rate denominators reported).

FIGURE 2 — Age-stratified year-3 effect (pooled): <9 years vs ≥9 years

This directly visualizes the paper’s pooled year-3 relative risks for hospitalization for virologically confirmed dengue by age group.

FIGURE 3 — Year-3 severe dengue (IDMC criteria): vaccine vs control (all ages in pooled safety window)

Severe dengue events are rare; visualization helps assess the contribution to the safety-signal discussion.

FIGURE 4 — Integrated symptomatic dengue efficacy during first 25 months (age-stratified)

From the paper’s pooled efficacy estimates: overall symptomatic virologically confirmed dengue efficacy and its age dependence.

FIGURE 5 — Safety surrogate vs efficacy (how the paper’s signals relate conceptually)

A schematic scatter: age group points with (x) efficacy in first 25 months and (y) year-3 hospitalization RR. This is not a formal meta-model; it helps structure tradeoffs the paper reports.

Interpretation (explain second, but still concise): what is known vs uncertain

1) What the trial data support (with the paper’s own emphasis)

Overall: pooled year-3 hospitalization for virologically confirmed dengue shows RR 0.84 (95% CI 0.56–1.24), i.e., not statistically decisive against the null (CI includes 1).
Age effect: in pooled year-3 data, ≥9-year-olds show a lower hospitalization risk (RR 0.50; 95% CI 0.29–0.86), while <9-year-olds show an apparent higher risk (RR 1.58; 95% CI 0.83–3.02). The pattern drives the overall mixed signal.
Severity: severe dengue events are rare; the paper reports severe dengue among year-3 hospitalized cases as 18/22,177 (vaccine) vs 6/11,089 (control) and highlights that classified “severe” cases were mostly IDMC/WHO grades I–II and recovered after supportive care—yet the uncertainty remains because hospitalization is a surrogate and the younger-child imbalance is still “unexplained” within the interim window.
Efficacy: in the earlier integrated efficacy phase (first 25 months), symptomatic virologically confirmed dengue efficacy is ~60% overall and is higher in ≥9-year-olds (65.6%) than in <9-year-olds (44.6%).

2) What remains uncertain / potential failure modes (skeptical checklist)

Hospitalization as a surrogate for severity: the paper uses hospitalization for virologically confirmed dengue to reduce underreporting of severe outcomes during follow-up, but hospitalization can be influenced by healthcare-seeking behavior, local hospital admission practices, and study contact/surveillance intensity. The paper explicitly notes that methods of capturing hospitalization differ between the efficacy surveillance phase and long-term follow-up, which complicates comparisons across phases.
Rare-event instability: year-3 event counts are small enough that age-stratified estimates can swing widely (wide CIs for some subgroups), making “true biology” vs “chance imbalance” hard to disentangle in interim time windows.
Endpoint measurement heterogeneity: the paper reports that long-term follow-up hospitalization capture relies on multiple routes (study contacts + self-report + surveillance at non-study hospitals) and that capture practices vary by country/center; randomization helps, but residual differential misclassification can still occur.
Unexplained age imbalance: the authors consider biological hypotheses (immune mechanisms, waning, age-related physiology/immune maturation), but the interim data include limited severity-marker detail. A “no difference in viremia/clinical pattern” argument helps, yet absence of evidence for a particular mechanistic marker is not evidence of absence of mechanism.

3) Mechanistic plausibility (without overclaiming)

The paper frames the younger-child imbalance as biologically plausible under an immune-enhancement / immune-setting narrative: vaccinated individuals may experience infection in a partially immune context that mimics “secondary infection,” which is classically associated with higher severe-disease risk. However, in this interim report, the authors state that they did not observe a clear corresponding signal in the measured virologic/clinical pattern among hospitalized cases (e.g., similar viremia; no observed changes in clinical picture). That means: plausibility is argued, but mechanistic confirmation is not established by the presented interim markers.

4) Bias/confounding considerations (epistemic humility)

Study oversight/sponsor role: the sponsor designed the trials, performed sample testing and data analyses, and had complete access to trial data; manuscript drafting used a sponsor-paid medical writer, with investigators and non-sponsor authors involved in interpretation. The paper reports procedures intended to preserve blinding of participants and staff for ongoing trials, but sponsor access still warrants careful reading of prespecified vs post hoc handling (some analyses are described as post hoc).
Post hoc age analyses: the paper states prespecified analyses were largely used, with post hoc analyses of younger/older subgroups in some components. Even if reasonable, this can affect interpretability of marginal findings.

What information would disprove or substantially change the paper’s takeaways?

Replication in subsequent follow-up windows: if continued follow-up shows the younger-child hospitalization imbalance disappears (or flips) with additional time, then the year-3 imbalance may reflect chance or transient surveillance/capture artifacts rather than durable predisposition.
Better severity biomarkers: mechanistic clarification would come from severity marker datasets (e.g., objective measures capturing plasma leakage/permeability in a standardized way) showing an immune-enhancement pattern in younger vaccinated children—or conversely, showing no such markers while hospitalization rates change.
Consistent endpoint capture across phases and centers: if updated analyses adjust for differences in hospitalization capture between phases/centers and the age imbalance persists, confidence in a true biological signal increases; if it attenuates after adjustment, measurement bias becomes more likely.

Optional BGPT follow-up actions (bespoke)

Author Reviews (bespoke links)

Note: author names are taken from the provided TEI XML text.

Feedback:

Updated: April 21, 2026