CELL INTELLIGENCE · 肿瘤细胞治疗研究

After five decades of CHOP-based therapy, the treatment landscape for patients with newly diagnosed large B-cell lymphoma (LBCL) is rapidly evolving. Novel agents, including antibody-drug conjugates, bispecific antibodies, targeted therapies, and chimeric antigen receptor (CAR) T-cell therapies, have demonstrated significant activity, raising the potential for improved outcomes from first-line treatment. However, this progress has introduced substantial complexity without a clear framework for clinical decision-making. Ongoing phase 3 trials employ heterogeneous designs, including differing control arms, eligibility criteria, and endpoints, yet share a common limitation: the inconsistent incorporation of biomarker-driven patient selection. As a result, cross-trial comparisons are challenging, and identifying the optimal regimen for an individual patient remains difficult. Although the POLARIX trial established polatuzumab-based therapy (R-CHP-pola) as superior to R-CHOP, many ongoing studies continue to use R-CHOP as the comparator, further complicating interpretation. Moreover, "R-CHOP + X" strategies evaluated in clinically defined populations without biologic stratification may yield modest benefits across heterogeneous groups, limiting their applicability to individual patients. These trends risk producing a fragmented landscape of effective yet non-comparable regimens without a principled framework for patient selection. We thus propose a harmonized, data-driven strategy to shape the next era of LBCL therapy: (1) pooled patient-level analyses; (2) standardized biomarker and minimal residual disease platforms; (3) AI-enabled predictive modeling integrating trial and real-world data; and (4) regulatory approaches promoting collaborative trial design and shared biomarker analyses. Ultimately, progress in LBCL will depend on aligning data, biology, and clinical decision-making to deliver precise, equitable, and effective care.