Dengue Literature Review

Hertanti2024 - Fatigue and Post-Infectious Fatigue in Dengue

8 min read

Hertanti2024 - Fatigue and Post-Infectious Fatigue in Dengue

Full citation: Hertanti NS, Nguyen TV, Chuang Y-H. Global prevalence and risk factors of fatigue and post-infectious fatigue among patients with dengue: a systematic review and meta-analysis. eClinicalMedicine. 2025;80:103041. (Published online 31 December 2024.) doi:10.1016/j.eclinm.2024.103041. PROSPERO CRD42024543058. Open access (CC BY-NC-ND).

Raw file: [[raw/Hertanti2024.pdf]]

Summary

This is the first systematic review and meta-analysis to quantify the global prevalence and risk factors of fatigue (acute phase) and post-infectious fatigue (PIF; persisting beyond the acute phase, defined here as 2 weeks to ≥6 months) among dengue patients. From 715 records across five databases (inception to 22 June 2024), 40 observational studies were included: 37 contributed to the fatigue prevalence pool (37,790 patients) and 9 to the PIF prevalence pool (5045 patients). A random-effects model (arcsine square-root transformation for fatigue, owing to non-normality) yielded a pooled acute-fatigue prevalence of 59.0% (95% CI 47–70%) and PIF prevalence of 20.0% (95% CI 10–36%).

Both pooled estimates carry extreme heterogeneity (fatigue I²=99.90%, Q=32026.88; PIF I²=98.77%, Q=649.05) and — critically — prediction intervals that span almost the entire possible range (fatigue PI [0.03; 1.00]; PIF PI [0.02; 0.77]). The point estimates are therefore best read as “fatigue is common and PIF is frequent,” not as precise rates: the next dengue cohort could plausibly report anywhere from near-zero to near-universal fatigue. No moderator explained the fatigue heterogeneity; for PIF, study design and mean age were significant moderators (older age → higher PIF).

For PIF, three risk factors reached pooled significance: female sex (OR 1.65), dengue haemorrhagic fever (DHF) (OR 1.80), and pre-existing comorbidities (OR 2.14). The authors list four hypothesized mechanisms for post-viral fatigue carried over from the broader literature — viral persistence, autoimmunity, immune dysfunction, and autonomic dysregulation — and sketch an estrogen→B-cell-activation→autoantibody→fatigue chain to explain the female-sex association, but explicitly state this “requires further confirmation.” The study measures no antinuclear antibodies and no autoantibodies of any kind. Its value to this wiki’s primary thread is precisely that: it is the largest synthesis of the fatigue bank and it names the ANA↔fatigue link as a hypothesis, leaving it untested — it defines the gap rather than filling it.

Study Design

  • Type: Systematic review and meta-analysis of observational studies (PRISMA 2020; PROSPERO-registered). Pooled prevalence (random-effects, DerSimonian-Laird via R meta/metafor v4.3.3); risk-factor pooling (ORs, ≥2 studies required); meta-regression; sensitivity, subgroup, and publication-bias analyses.
  • Sample size: 40 studies / 38,406 dengue patients overall; fatigue pool 37 studies (37,790 patients); PIF pool 9 studies (5045 patients). Risk-factor pools: 2–3 studies each.
  • Setting: Global. Included studies predominantly Asian (n=25, 62.5%), then South America (n=7, 17.5%), Africa (n=4, 10.0%), Europe (n=3, 7.5%), North America (n=1, 2.5%). PIF pool heavily Sri Lanka– and Singapore-dominated.
  • Population: Dengue patients across all phases; 48.2% female, mean age 40.9 (SD 18.5) overall; PIF subset younger (mean 29.5, SD 13.7) and 50.8% female. Fatigue measured by clinical symptoms in 35/40 (87.5%) studies and by a validated fatigue questionnaire in only 5/40 (12.5%).
  • Risk of bias: Hoy’s tool — 3 studies low (7.5%), 28 moderate (70.0%), 9 high (22.5%).

Key Findings

  • Acute fatigue is common but the rate is imprecise: pooled prevalence 59.0% (95% CI 47–70%), but I²=99.90% and the prediction interval is [0.03; 1.00] (Fig 2). The 95% CI describes the precision of the mean; the prediction interval describes where a new study would fall — and here it is essentially uninformative (≈3–100%). Every downstream use of “59%” must carry this caveat.
  • PIF is frequent but equally heterogeneous: pooled 20.0% (95% CI 10–36%), I²=98.77%, prediction interval [0.02; 0.77] (Fig 3). Sensitivity analyses (excluding 9 high-risk-of-bias studies → 58.0%; excluding small studies → 61.0%; PIF leave-one-out 18.0–27.0%) confirm the point estimate is robust to study removal — but robustness of a central estimate does not narrow the prediction interval.
  • Significant PIF risk factors (pooled):
    • Female sex — OR 1.65 (95% CI 1.27–2.14), p=0.0002, I²=0.00 (3 studies: Seet2007, Perera2023, Abeysena2019). Zero heterogeneity; a modest effect.
    • DHF — OR 1.80 (95% CI 1.02–3.16), p=0.042, I²=18% (2 studies). Borderline; carried by Perera2023 (1.58, 1.04–2.40, significant); Sigera2021 (3.43, 0.92–12.76) is individually non-significant.
    • Pre-existing comorbidities — OR 2.14 (95% CI 1.36–3.38), p=0.001, I²=0.00 (2 studies).
  • Non-significant PIF risk factors (pooled): older age (2.03, 0.58–7.14; I²=92.5%), post-discharge myalgia (1.82, 0.66–5.05; I²=81%), post-discharge headache (1.16, 0.41–3.31; I²=82%). Individual studies are sometimes significant in opposite directions; the pools are null.
  • Acute-fatigue × DHF is null and contradictory: pooled OR 1.29 (0.43–3.88), p=0.65, I²=93.5% — driven by Ferreira2018 (2.31, harmful) vs Recker2024 (0.75, protective), two large studies pointing opposite ways. This is the single sharpest illustration that dengue severity has no consistent relationship to fatigue in the current literature.
  • ⚠ Table 2 internal error (verified against the PDF): the female-sex row prints Seet2007’s OR as “9.69” but with CI “0.78–4.01” — a point estimate that cannot lie inside its own interval, and incompatible with the pool’s I²=0.00 across {Seet, Perera 1.82, Abeysena 1.50}. The reliable figure is the pooled 1.65; Seet’s unadjusted contribution is ≈1.7 (95% CI 0.78–4.01, non-significant alone). The “9.69” is a transcription of Seet2007’s own adjusted multivariate OR (9.687, CI 1.546–60.684), a sparse-data artifact — see Seet2007 - Post-Infectious Fatigue Syndrome in Dengue. This corrects the prior wiki framing of 9.687 as a robust “outsized” female effect.
  • PIF measurement depends on the instrument: clinical-symptom checklists detected PIF in only 12.6% of cases vs 29.5% with validated fatigue questionnaires (subgroup difference not statistically significant, but directionally large) — mirroring post-COVID data (47.5% by questionnaire vs 43.2% by checklist). Fatigue prevalence is plausibly underestimated in the 87.5% of studies relying on clinical symptoms.
  • Benchmarking: PIF 20.0% is similar to post-Q-fever chronic fatigue, and lower than post-COVID fatigue in the first 6 months (41.0–46.6%).

Methods Used

  • Prevalence Meta-Analysis under Heterogeneity — random-effects pooling of prevalence under extreme between-study heterogeneity; this paper is the worked example for why the prediction interval, not I² alone, governs interpretation.

Entities Mentioned

(None at the entity level. The meta-analysis is serotype-agnostic — it pools across all DENV serotypes and reports no serotype-specific fatigue data, so no serotype entity pages are updated.)

Concepts Addressed

  • Post-Dengue Syndrome — quantifies the fatigue/PIF bank of post-dengue sequelae and its risk factors.
  • Autoimmunity in Dengue — names autoimmunity as one of four hypothesized PIF mechanisms; supplies the meta-analytic correction to the Seet2007 female-sex OR.
  • Antinuclear Antibodies — gap marker: the largest fatigue synthesis in dengue measures no ANA.

Relevance & Notes

This paper is pivotal for the wiki’s primary thread — is there a correlation between ANA and chronic fatigue in dengue? — but its contribution is to define the gap, not close it:

  1. It quantifies the fatigue bank. Before this, the fatigue side of the bridge rested on single cohorts (Seet2007 24.4% at 2 months; Garcia2009 asthenia 23.0% at 2 years). Hertanti2024 establishes that acute fatigue (~59%) and PIF (~20%) are common across 40 studies — but with prediction intervals so wide that the rate is almost uninformative; what is solid is the qualitative claim that fatigue is a major post-dengue burden.
  2. It corrects the “shared confounder” framing. The wiki’s ANA↔fatigue logic leans on female sex being a strong driver of both outcomes (the “shared signal = shared confounder” argument). Hertanti2024 deflates the fatigue half: the cross-study female→PIF effect is a modest, precise 1.65 (I²=0.00) — not the dramatic 9.687 that the single Seet2007 multivariate model produced. Sex remains a confounder to control, but it is a smaller lever than the prior framing implied. The mandatory-control-for-sex lesson from the methodology thread (see Methodology Critique - ANA IIF Abstract Draft) still holds, but the expected effect size is recalibrated downward.
  3. It names autoimmunity but tests nothing. Autoimmunity is 1 of 4 mechanisms the authors list, and the estrogen→B-cell→autoantibody→fatigue chain is flagged “requires further confirmation.” Zero ANA, zero autoantibodies are measured. So even the field’s largest fatigue synthesis leaves the ANA↔fatigue correlation untested — confirming this is genuinely open territory rather than a settled negative.
  4. It reinforces severity-independence. The null/contradictory DHF×acute-fatigue result (Ferreira 2.31 vs Recker 0.75, pooled NS) is consistent with the wiki’s existing cross-cohort finding that DF-vs-DHF severity does not predict post-dengue sequelae (see Post-Dengue Syndrome, Seet2007 - Post-Infectious Fatigue Syndrome in Dengue, Garcia2009 - Long-term Clinical Symptoms Post-Dengue). The borderline DHF→PIF signal (1.80) is a tension to hold, not a reversal — it rests on 2 studies, one non-significant.

Limitations (author-stated and curator-noted): (1) extreme, largely unexplained heterogeneity; (2) 87.5% of studies used non-standardised clinical-symptom assessment, likely under-detecting fatigue; (3) most studies moderate risk of bias; (4) risk-factor pools are tiny (2–3 studies) and dominated by a small set of Sri Lankan cohorts (Abeysena2019, Sigera2021, Perera2023, Umakanth2018), limiting generalisability and raising a single-setting-cluster concern; (5) no inverse-probability weighting for population differences; (6) the Table 2 transcription error noted above.

Questions Raised

  • Does ANA-positivity track PIF at the individual level once sex and age are controlled? Hertanti2024 cannot say — it pools study-level prevalences and measures no ANA.
  • Which fatigue instrument should the wiki privilege? The 12.6% (clinical symptoms) vs 29.5% (validated questionnaire) gap means cross-study fatigue rates are not comparable without harmonising the measure — directly relevant to aligning Seet2007 (FQ), Gawali2021, and Garcia2009.
  • Is the borderline DHF→PIF signal (1.80, 2 studies) real, given that DHF→acute-fatigue is null and contradictory? A cohort co-measuring acute severity and PIF trajectory would resolve whether severity acts only on the persistent phase.
  • The PIF risk-factor evidence is geographically narrow (Sri Lanka–heavy). Do the female/DHF/comorbidity associations replicate outside South Asia?

Graph View

Backlinks