Continuous-Infusion Ceftazidime–Avibactam Dosing in Critically Ill Adults: A Provenance-Aware Pharmacometric Simulation

Background and Objective. Critically ill adults exhibit wide, poorly predictable pharmacokinetic variability that complicates β-lactam/β-lactamase inhibitor dosing, and continuous infusion of ceftazidime–avibactam is used to attain aggressive pharmacokinetic/pharmacodynamic (PK/PD) targets. Because ceftazidime and avibactam are pharmacologically distinct, ceftazidime exposure can appear adequate while avibactam remains limiting, so joint target attainment cannot be inferred from either component alone. We evaluated continuous-infusion ceftazidime–avibactam dosing in critically ill adults without renal replacement therapy across renal-function, pharmacokinetic, PK/PD-target, MIC-distribution, and uncertainty scenarios, using a provenance-aware evidence-composite pharmacometric model to quantify the associated uncertainty. Methods. This was a simulation-based, evidence-composite pharmacometric analysis. No new patient-level clinical data were collected and no new population pharmacokinetic model was fitted; synthetic virtual populations were generated from published clearance equations, reported interindividual variability (67.92% CV ceftazidime, 76.91% CV avibactam), correlated lognormal ceftazidime and avibactam clearance (ρ = 0.94), and published MIC distributions. Ceftazidime and avibactam were modeled as separate pharmacokinetic components and evaluated jointly. Monte Carlo simulation used 100,000 virtual subjects across five EKFC renal-function classes. The ceftazidime target was fCss/MIC ≥ 4 and the avibactam target was fCss/(4 mg/L) ≥ 1, with joint attainment requiring both simultaneously; a total ceftazidime Css > 104 mg/L served as an exposure-based toxicity screen. Probability of target attainment (PTA), MIC-weighted cumulative fraction of response (CFR), calibration against published benchmarks (72 rows), convergence, multi-seed reproducibility, and one-at-a-time and probabilistic sensitivity analyses (300 Latin hypercube draws) were evaluated. Results. Joint target attainment was consistently lower than ceftazidime-only attainment and was constrained by the avibactam component. Among permissible regimens, joint PTA at MIC 4 mg/L ranged from 67.05% to 88.28%, with toxicity-screen probabilities of 7.33–12.32%. For the primary KPC-K. pneumoniae distribution, joint CFR for the best-performing regimens ranged from 73.45% to 88.65%. Calibration met predefined criteria in 69 of 72 rows (3 targeted review, 0 fail; mean ceftazidime PTA bias 0.62 percentage points), and convergence was stable (mean absolute joint PTA difference 0.167 percentage points versus the 100,000-subject reference). Avibactam clearance (|r| = 0.814), avibactam interindividual variability (|r| = 0.499), and avibactam unbound fraction (|r| = 0.249) were the dominant uncertainty drivers; no regimen–distribution pair met the predefined full-robustness criterion (7 conditionally robust, 13 fragile). Conclusions. In this uncertainty-aware simulation, joint continuous-infusion ceftazidime–avibactam target attainment in critically ill adults was limited primarily by avibactam-related exposure and assumptions rather than by ceftazidime pharmacokinetics or MIC distribution alone. Point estimates of PTA and CFR should be interpreted alongside calibration, convergence, and sensitivity analyses, and separate consideration of the two components is warranted. These simulation findings characterize dosing performance and uncertainty; they are not a clinical validation and would require prospective confirmation before informing individualized dosing.