The Heterogeneous Landscape of Frontline Treatment of Peripheral T-Cell Lymphomas: A Systematic Review and Network Meta-Analysis

Peripheral T-cell lymphomas (PTCLs) are rare, heterogeneous malignancies with poor prognosis. The optimal frontline strategy remains undefined due to small trial sizes and a lack of direct head-to-head comparisons. A systematic search identified randomized controlled trials evaluating frontline PTCL therapies. Primary Bayesian random-effects and sensitivity frequentist network meta-analyses were conducted to derive HRs, ORs, RRs, and Surface Under the Cumulative Ranking (SUCRA) probabilities across endpoints. Five RCTs comprising 1286 patients across 7 treatment regimens were included: BV-CHP, Ro-CHOP, A-CHOP, GEM-P, the alternating intensified regimen CEOP/IVE/GDP, and standard anthracycline-based backbones (CHOP and CEOP aggregated as the reference arm). In the primary Bayesian random-effects model, sequential CEOP/IVE/GDP demonstrated the highest cumulative likelihood of efficacy across OS, PFS, and ORR, alongside modest tolerability. BV-CHP also demonstrated notable efficacy across response and survival outcomes, paired with the highest safety profile among experimental regimens. In the CD30-unselected LOO sensitivity analysis, CEOP/IVE/GDP maintained leading overall performance, whereas GEM-P and Ro-CHOP formed a closely packed intermediate cluster. Histological subgroup analyses revealed clear therapeutic divergence: BV-CHP showed the highest rank for PTCL-NOS and ALK-negative ALCL, whereas Ro-CHOP showed top ranking for angioimmunoblastic T-cell lymphoma (AITL/TFH). Frontline management of PTCL benefits from histology-driven perspectives. Our data suggest that BV-CHP may represent a favorable backbone for CD30-positive histologies (ALK-negative ALCL and PTCL-NOS) due to an optimal efficacy-safety balance. Intensive sequential regimens (CEOP/IVE/GDP) provide robust disease control in broad fit populations, while targeted approaches such as Ro-CHOP show promising exploratory signals in AITL/TFH subtypes.