Paired In Vitro Susceptibility of Cefiderocol and Colistin in Colistin-Resistant, Extensively Drug-Resistant Gram-Negative Bacilli: A Single-Centre, MIC-Based Analysis of 236 Clinical Isolates.

Extensively drug-resistant (XDR) Gram-negative bacilli increasingly exhaust therapeutic options, leaving colistin and cefiderocol as last-resort agents. We assessed whether cefiderocol retains in vitro activity against colistin-resistant isolates, without inferring clinical efficacy. We analysed 236 non-duplicate XDR Gram-negative isolates from 220 patients. Cefiderocol and colistin MICs were determined via reference broth microdilution with EUCAST-required quality control (iron-depleted broth for cefiderocol; mcr-1-positive Escherichia coli NCTC 13846 for colistin) and interpreted using species-specific EUCAST v16.1 (2026) breakpoints (cefiderocol: Enterobacterales S ≤ 2, I 4, R > 4 mg/L; Pseudomonas aeruginosa S ≤ 2, R > 2 mg/L; colistin: P. aeruginosa S ≤ 4 mg/L). Because EUCAST publishes no clinical cefiderocol breakpoint for Acinetobacter baumannii, the primary analysis comprised species with breakpoints for both agents (n = 150; 126 paired); A. baumannii is reported as an MIC distribution and the pooled 236-isolate comparison as secondary. In the primary analysis, 120/150 isolates (80.0%) were cefiderocol-susceptible at standard or increased exposure. Among paired isolates, cefiderocol covered 77.0% versus 43.7% for colistin; agreement was negligible (Cohen's κ = 0.020), with 54 isolates favouring cefiderocol and 12 the reverse (McNemar p < 0.001). Conditional in vitro susceptibility among 71 colistin-resistant isolates was 76.1% (95% CI 65.0-84.5), giving 33.3 percentage points of incremental coverage and one additional cefiderocol-susceptible isolate per 2.3 matched isolates tested. Secondary analyses were directionally consistent (incremental coverage 21.8-33.3 points). These in vitro findings support paired testing of both agents to identify isolates retaining cefiderocol activity when colistin does not; clinical benefit was not assessed.