Mechanistic constraints in dengue severity: a systematic review with evidence stratification and agent-based evaluation of logical sufficiency.

Background: Severe dengue is a vascular immunopathology characterized by plasma leakage, thrombocytopenia, hemorrhage, and, in its most critical form, dengue shock syndrome. Although NS1-mediated endothelial injury, glycocalyx disruption, inflammatory myeloid activation, and coagulation/platelet abnormalities have all been implicated, it remains unclear which mechanisms are most consistently supported and whether they form a coherent functional architecture capable of explaining vascular decompensation. This review asks two linked questions: which dengue mechanisms are supported by the contemporary evidence base, and whether the strongest supported components are logically sufficient, when coupled, to generate a synthetic analog of connected endothelial-barrier failure. Methods: We conducted a PubMed-indexed systematic review of dengue mechanistic studies published from 2020 to 2025 under a dengue-only eligibility policy. Full texts were assigned to six mechanism families, graded on a five-tier evidence scale, and classified using predefined claim ceilings: C0 empirical restriction, C1_conditional regularity, or C2 exploratory evidence. Meta-analysis readiness was assessed using PICOS criteria. A constraint-first agent-based model (ABM) was then used as a permanent C2 logical sufficiency evaluator to test whether the three strongest evidence families could jointly generate a synthetic analogue of connected endothelial-barrier failure under explicit assumptions. Results: Of 200 retrieved records, 59 were included after full-text adjudication. Three mechanism families reached C1_conditional evidence: NS1-linked vascular permeability (DENV-M01, n=23), endothelial glycocalyx/barrier disruption (DENV-M02, n=17), and myeloid effector activation (DENV-M03, n=12). Receptor gating, coagulopathy/platelet dysregulation, and therapeutic mechanistic targets remained C2 evidence-gap families. Two null randomized trials imposed C0 restrictions: rupatadine did not significantly reduce plasma leakage (RR = 0.68, 95% CI 0.41-1.12), and oseltamivir did not improve time to defervescence (MD =+ 0.1 days, p=0.055). No mechanism family was eligible for quantitative pooling because CI-bearing estimates were sparse and outcome definitions were insufficiently harmonized. In the ABM, the review-supported NS1-barrier-myeloid set generated a spatially connected endothelial-barrier failure analog. This analog emerged when upstream viral/NS1 pressure and myeloid collateral cost exceeded barrier reserve and repair capacity. The regime remained stable under changes in update rule, rule form, and spatial patch scale, indicating that it was not a single implementation artefact. Boundary location was more stable than local execution timing, whereas high heterogeneity intensity produced only bounded boundary displacement. Minimality ablation showed partial, not complete, minimality: upstream pressure and barrier fragility were load-bearing, whereas the myeloid arm was phase-dependent and counter-directional, consistent with a dual role in early containment and late collateral damage within the model. Clinical trial registration: The current evidence supports a minimum-range organizational account of severe dengue vascular decompensation centered on the NS1-barrier-myeloid unit. This account is best interpreted as a competing-constraint model: viral/NS1 pressure, endothelial/glycocalyx barrier preservation, repair capacity, and myeloid effector control can become difficult to maintain within the same physiological window during progression toward vascular leakage. The ABM provides C2-level in silico support for logical sufficiency by showing that these review-supported components can generate a connected endothelial-barrier failure analog under explicit assumptions. It does not establish causal mechanistic validation, molecular equivalence, or patient-level prediction. Claim escalation now requires longitudinal cohorts measuring NS1/viraemia, endothelial barrier injury markers such as SDC1 or Ang-2, and myeloid effector proxies such as sTREM-1 or CXCL10, together with orthogonal functional perturbation assays reporting CI-bearing outcomes.