CELL INTELLIGENCE · 肿瘤细胞治疗研究

Newer antibiotics have expanded treatment options for drug-resistant Gram-negative infections, but registration evidence is dominated by syndrome-based trials and post-approval evidence by heterogeneous observational cohorts. Translation to immunocompromised patients remains uncertain because profound neutropenia, transplantation, cell-targeted therapy, organ dysfunction, impaired source control, and limited immune-mediated clearance change both the probability and consequences of treatment failure. This critical narrative Review integrates 27 pivotal-trial protocol identifiers included in a predefined evidence map, immune-phenotype-specific cohorts, contemporary guidance, pharmacokinetic/pharmacodynamic evidence, and treatment-emergent resistance reports for a bounded core set of newer agents first authorized in the United States or European Union between 2014 and 2025, including newer β-lactam/β-lactamase-inhibitor combinations, cefiderocol, sulbactam-durlobactam, aztreonam-avibactam, and selected non-β-lactams. Fifteen of 27 pivotal protocols explicitly excluded at least one major immune phenotype or threshold; 11 had unresolved immune-host enrollment, and one enrolled immunocompromised patients but pooled distinct immune phenotypes. None reported comparative outcomes resolved to a defined immune phenotype. Phenotype-specific post-approval evidence was concentrated in hematological malignancy/profound neutropenia and solid-organ transplantation, whereas direct treatment-outcome evidence was sparse or absent for solid tumors, cellular therapies including CAR-T, advanced HIV infection, inborn errors of immunity/primary immunodeficiencies, and pediatric immunocompromised patients. We therefore separate phenotype-specific transportability from methodological credibility rather than treating directness as an overall certainty grade. The Review then synthesizes treatment by resistance mechanism and applies a host-pathogen-drug-context framework to empirical selection, diagnostics, exposure optimization, source control, de-escalation, relapse, and resistance. The resulting evidence map supports mechanism-active therapy while making the limits of transfer auditable and defining priorities for standardized phenotype reporting and pragmatic enrollment.