To synthesize the diagnostic performance of inflammatory biomarkers measured in blood or CSF for distinguishing HAM/TSP patients from HTLV-1 asymptomatic carriers and seronegative controls.

HTLV-1-associated myelopathy/tropical spastic paraparesis (HAM/TSP) is a chronic neuroinflammatory disorder lacking validated diagnostic biomarkers for routine clinical use. Although multiple inflammatory mediators have been investigated, no systematic review has specifically synthesized the diagnostic performance of soluble biomarkers measured directly in blood or cerebrospinal fluid (CSF).

We conducted a systematic review following PRISMA-DTA 2018 and SWiM 2020 guidelines. Observational studies quantifying inflammatory biomarkers directly in serum, plasma, or CSF of HTLV-1–infected adults were included, excluding studies limited to ex vivo stimulated samples, gene expression, or anti-HTLV-1 antibodies. Risk of bias was assessed with JBI checklists and QUADAS-2. Due to substantial heterogeneity, results were synthesized narratively. Diagnostic accuracy was a pre-planned secondary sub-analysis.

Thirty two studies including 1,118 HAM/TSP patients, 982 asymptomatic carriers, and 878 seronegative controls were analyzed. Seventeen of 32 studies (53%) had low risk of bias (JBI), none high. The most consistently elevated biomarkers in HAM/TSP versus asymptomatic carriers were CSF CXCL10 (3/3 studies; AUC 0.988 and 0.927) and CSF neopterin (3/3 studies; AUC 0.975 and 0.942), each matching or exceeding proviral load in the same cohorts. CSF CXCL9 also discriminated well (AUC 0.971). Serum soluble IL-2 receptor (AUC 0.965, 0.920–1.000) and plasma CXCL10 (AUC 0.917, 0.844–0.989) emerged as accessible alternatives. Serum IL-6, TNF-α, and IFN-γ showed less consistent elevation. Only 5/32 studies reported formal accuracy metrics, and heterogeneity across assays was substantial.

CSF neopterin and CXCL10 are the most consistently supported candidate diagnostic biomarkers for HAM/TSP, although evidence is limited by heterogeneity, two-gate designs, and data-driven thresholds. Plasma sIL-2R and plasma CXCL10 may provide accessible alternatives. Standardization and prospective validation are needed for clinical implementation.