Dengue Literature Review

Shih2023 - Autoimmune Disease Risk After Dengue

6 min read

Shih2023 - Autoimmune Disease Risk After Dengue

Full citation: Shih H-I, Chi C-Y, Tsai P-F, Wang Y-P, Chien Y-W (2023) Re-examination of the risk of autoimmune diseases after dengue virus infection: A population-based cohort study. PLoS Neglected Tropical Diseases 17(3): e0011127. https://doi.org/10.1371/journal.pntd.0011127

Raw file: [[raw/Shih2023.pdf]]

Summary

This paper re-examines a prior claim (Li et al. 2018) that dengue infection substantially increases the risk of a broad spectrum of autoimmune diseases. Using Taiwan’s National Health Insurance Research Database (NHIRD) linked to the Notifiable Disease Dataset of Confirmed Cases (NDDCC), the authors assembled a cohort of 63,814 laboratory-confirmed dengue patients diagnosed between 2002 and 2015 and 255,256 matched non-dengue controls. Matching was done 1:4 by age, sex, area of residence, and calendar month of the index date. The mean follow-up was approximately 4.57 years per group. Multivariate Cox proportional hazard regression with Bonferroni correction for multiple comparisons was used to assess risk across 14 specific autoimmune disease categories.

Contrary to Li et al., this study finds that dengue infection is associated with only a modestly elevated overall autoimmune disease incidence (aHR 1.16, P = 0.0002) and, after Bonferroni correction, with one specific disease: autoimmune encephalomyelitis (ADEM; aHR 2.72, P < 0.0001). All 16 dengue patients who developed ADEM (0.025% of the cohort) did so within the first month after symptom onset; zero controls developed ADEM in the same window (HR >9999). The risk was not statistically elevated beyond the first month, supporting a transient, acute autoimmune mechanism rather than a chronic autoimmune predisposition.

The authors attribute the discrepancy with Li et al. to three methodological flaws in the prior study: (1) use of clinically diagnosed rather than laboratory-confirmed dengue cases — only 51.4% of hospitalized “dengue” patients in the Li et al. era were lab-confirmed; (2) failure to correct for multiple comparisons (20 outcomes tested at α = 0.05); and (3) restriction to hospitalised patients, which biases toward severe presentations and allows misdiagnosis of autoimmune diseases as dengue during outbreaks.

Study Design

  • Type: Retrospective population-based cohort study
  • Sample size: 63,814 dengue cases; 255,256 controls (1:4 matching); 319,070 total participants
  • Setting: Taiwan (nationally representative via NHIRD); dengue patients concentrated in southern Taiwan (Tainan, Kaohsiung, Pingtung); 2002–2015 exposure window; follow-up to August 31, 2018
  • Population: All ages (mean 44.6 years); ~50/50 sex; all newly diagnosed, laboratory-confirmed dengue cases

Key Findings

  • Overall autoimmune disease incidence: 27.73 per 10,000 person-years (dengue) vs. 23.47 (controls); aHR 1.16 (95% CI 1.07–1.26), P = 0.0002.
  • After Bonferroni correction (threshold P < 0.0012), only autoimmune encephalomyelitis remained significant: aHR 2.72 (95% CI 1.84–4.02), P < 0.0001.
  • 16 dengue patients (0.025%) vs. 0 controls developed ADEM in the first month (HR >9999, P < 0.0001); risk not significant thereafter (aHR 1.76, P = 0.012 — above Bonferroni threshold).
  • By sex: ADEM risk elevated in males (aHR 32, P = 0.0010) and females (aHR 3.44, P = 0.0001).
  • Autoimmune thyroid disease (aHR 1.18, P = 0.0295) and uveitis (aHR 1.23, P = 0.0190) showed nominal elevation but did not survive Bonferroni correction.
  • SLE (aHR 1.86, P = 0.0817), rheumatoid arthritis (aHR 0.99), Sjögren’s (aHR 0.87), Guillain–Barré syndrome (aHR 1.29), and post-infectious arthritis (aHR 0.97) were all non-significant.
  • Among the prior Li et al. associations that Shih2023 cannot replicate: primary adrenocortical insufficiency, multiple sclerosis, systemic vasculitis, myasthenia gravis, SLE, post-infectious arthritis — all non-significant here.
  • Subgroup analysis restricted to hospitalised dengue patients (more comparable to Li et al.) still shows only ADEM as significant.
  • Only 51.4% of hospitalised patients with dengue discharge diagnosis (2000–2010) were ultimately lab-confirmed in the NDDCC — the key misclassification figure undermining the Li et al. results.
  • Population-based ADEM rate in dengue (0.025%) is substantially lower than prior hospital-based meta-analysis estimate (~0.4%), reflecting inclusion of mild/outpatient cases.

Methods Used

  • RT-PCR (laboratory confirmation criterion: real-time RT-PCR)
  • IgM-IgG Serology ELISA (laboratory confirmation criterion: dengue-specific IgM/IgG)
  • NS1 Antigen Detection (confirmation criterion for 2015 epidemic; NS1 RDT)
  • Cox proportional hazard model (multivariate, Bonferroni and Benjamini–Hochberg corrections for multiple testing)
  • Kaplan–Meier cumulative incidence curves

Entities Mentioned

  • DENV-2 (direct CNS invasion, especially DENV-2 and DENV-3)
  • DENV-3 (direct CNS invasion)
  • NS1 Protein (NS1 rapid antigen test expanded diagnostic capacity from 2015)
  • Aedes aegypti (primary vector; implied in dengue transmission context)

Concepts Addressed

Relevance & Notes

This paper is the most methodologically rigorous large-scale study of post-dengue autoimmune risk to date. Its significance for the wiki is twofold:

First, it substantially revises the claim from Li et al. (2018) that dengue causes a broad spectrum of autoimmune diseases. The Garcia2009 Cuban cohort (in this wiki) found elevated autoimmune markers (ANA, IC, CRP) in symptomatic post-dengue patients, and the Lin group papers established NS1 molecular mimicry as a mechanism for acute autoimmune pathology. Shih2023 now closes the loop at the population level: despite strong biological plausibility, the clinical incidence of most autoimmune diseases after dengue is not epidemiologically elevated. The autoimmunity in dengue appears largely transient, consistent with the convalescence-phase decline of autoantibody titers described in the mechanistic literature.

Second, the methodological critique is broadly applicable to the post-infectious autoimmunity literature: misclassification of the exposure (non-lab-confirmed dengue), failure to correct for multiple comparisons, and restriction to hospitalised patients are common problems that inflate autoimmune risk estimates.

The one robust signal — ADEM — is consistent with the neuropathogenesis literature: DENV-2 and DENV-3 directly invade the CNS; ADEM (also known as postinfectious encephalomyelitis) is understood to result from transient molecular mimicry or non-specific activation of autoreactive T-cell clones against myelin; the first-month concentration of cases matches reported day 3–19 onset of neurological symptoms post-dengue. Shih2023 also contextualises dengue ADEM within the broader flavivirus class (West Nile, Zika, Chikungunya are similarly associated with autoimmune encephalomyelitis).

Limitations: No data on DENV serotype or primary vs. secondary infection status (could not stratify ADEM risk by infection number). Some non-catastrophic autoimmune disease diagnoses rely on ICD codes rather than specialist review. Controls may include some undetected mild/asymptomatic dengue cases (though dengue was non-endemic in most of Taiwan, limiting this bias).

Questions Raised

  • Does ADEM risk after dengue vary by serotype? DENV-2 and DENV-3 are specifically implicated in direct CNS invasion, making them candidate higher-risk serotypes.
  • Does primary vs. secondary dengue infection differentially affect ADEM risk, as it does for severe dengue outcomes?
  • What is the true incidence of ADEM in dengue, accounting for under-diagnosis in outpatient settings? The gap between 0.025% (this study) and 0.4% (hospital-based meta-analysis) suggests significant ascertainment variation.
  • Does the elevated but non-Bonferroni-significant signal for autoimmune thyroid disease (aHR 1.18) or uveitis (aHR 1.23) represent a true modest risk or a type I error? Replication in an independent large cohort is needed.
  • The Garcia2009 cohort showed elevated ANA, IC, and CRP at 2 years post-dengue — how does persistent autoimmune marker elevation reconcile with Shih2023’s finding of no elevated autoimmune disease incidence? Is the ANA elevation clinically inert?

Graph View

Backlinks