Phenotypic Resistance Profiles, Biofilm Formation, and In Vitro Carbapenem-Sparing Antimicrobial Activity in...

Empirical management of acute pyelonephritis in Eastern Europe is increasingly constrained by extended-spectrum β-lactamase (ESBL)-producing Enterobacterales and by uropathogen phenotypes-such as strong biofilm formation-which may further blunt antimicrobial activity. We aimed to characterise resistance mechanisms, minimum inhibitory concentration (MIC) distributions, biofilm-forming capacity, and the in vitro performance of carbapenem-sparing agents and to test whether these microbiological features improve prediction of clinical failure beyond standard bedside risk scores. We retrospectively analysed 102 Enterobacterales isolates recovered from 129 consecutive culture-confirmed adult pyelonephritis admissions at "Victor Babeș" University Hospital, Timișoara (March 2022-March 2025). MIC values were determined by Vitek 2 and interpreted using EUCAST v13 breakpoints; ESBL, AmpC, and carbapenemase phenotypes were confirmed by combination disk and modified carbapenem inactivation methods. Biofilm formation was quantified by the microtiter-plate crystal-violet assay. Mediation, Restricted Mean Survival Time (RMST), and decision-curve analyses were used to assess added clinical value. ESBL was confirmed in 30/102 (29.4%) isolates, AmpC in 9 (8.8%), and carbapenemase in 4 (3.9%). ESBL+ isolates were more often strong biofilm formers (33.3% vs. 12.5%; p = 0.014) and showed a 4- to 16-fold rightward MIC shift for cefepime, piperacillin-tazobactam, and ciprofloxacin. Among carbapenem-sparing agents, ceftazidime-avibactam (96.7% S), fosfomycin (80.0% S), and amikacin (73.3% S) retained the highest activity against ESBL+ isolates. Strong biofilm formation and the ESBL phenotype were independently associated with worse outcomes (adjusted OR 3.5 and 4.7); an exploratory mediation analysis suggested that biofilm formation may explain part of the observed association between the ESBL phenotype and treatment failure and that delayed effective therapy may account for a further portion of this association. A microbiology-enhanced model that added the ESBL phenotype, biofilm strength, and acquisition setting to routine clinical variables improved discrimination over a clinical-only baseline (AUC 0.89 vs. 0.71) and showed a higher net benefit on exploratory decision-curve analysis across the 10-40% threshold range. These predictive findings derive from a single-centre cohort with a small number of events and were only internally validated; they require validation in independent cohorts before any clinical application can be considered. The ESBL phenotype and strong biofilm formation were each independently associated with worse outcomes in pyelonephritis and may help identify candidate isolates for carbapenem-sparing strategies anchored on ceftazidime-avibactam, fosfomycin, and amikacin; given the observational, single-centre design, these associations should be regarded as hypothesis-generating.