A provenance-aware methodological pilot for reconstructing fragmented antimicrobial evidence into simulation-ready components
antimicrobial resistance, antibiotic PK/PD, antimicrobial pharmacometrics, Monte Carlo simulation, virtual populations, ceftazidime, avibactam, reproducible research
Antimicrobial Pharmacometrics · Active Research
A provenance-first methodology for converting fragmented ceftazidime–avibactam pharmacokinetic and pharmacodynamic evidence into traceable, simulation-ready components — extracting compatible evidence at the component level and recording its origin, transformation, uncertainty, and review state.
💊 Antibiotic PK/PD 🦠 Antimicrobial Resistance 📊 Pharmacometrics 🎲 Monte Carlo Simulation 👥 Virtual Populations 🔁 Reproducible Research
2 PK Models
100k Subjects
PTA & CFR
PSA Uncertainty
Active research · Associated manuscript under journal submission
This project develops a provenance-first methodology for converting fragmented ceftazidime–avibactam pharmacokinetic and pharmacodynamic evidence into traceable, simulation-ready components. Rather than treating each publication as wholly usable or unusable, the workflow extracts compatible evidence at the component level and records its origin, transformation, uncertainty, and review state.
Published antimicrobial PK/PD evidence is often scientifically valuable but not directly reusable. One source may report clearance equations without MIC weights, another may report trough concentrations without a complete simulation model, and another may provide in vitro or target-site pharmacodynamic information that is useful but incompatible with the clinical scenario of interest.
Ceftazidime–avibactam is a particularly demanding case because the two agents cannot be collapsed into a single exposure metric. Ceftazidime supplies the antibacterial beta-lactam exposure, while avibactam protects ceftazidime from susceptible beta-lactamases. A regimen may appear adequate for ceftazidime while avibactam exposure remains limiting, so the two components must be modeled separately and evaluated jointly.
Instead of an all-or-nothing include/exclude decision, each source is decomposed into reusable scientific components — clearance relationships, interindividual variability, protein binding or unbound fraction, PK/PD target definitions, MIC distributions, calibration anchors, and uncertainty inputs.
Each component carries:
Missing data are not silently imputed. They are retained as explicit evidence gaps unless a compatible donor component is substituted with clear labeling.
The primary scenario models critically ill adults without renal replacement therapy receiving continuous-infusion ceftazidime–avibactam. This population is clinically important, exposure-sensitive, and well suited to demonstrating why component-level provenance and joint PK/PD evaluation matter.
The framework integrates the following analytical components:
Separate ceftazidime and avibactam clearance models with renal-function-dependent exposure and correlated interindividual variability.
A 100,000-subject virtual population propagates correlated variability through both component models.
Joint PK/PD target attainment, probability of target attainment (PTA) across MIC values, and MIC-weighted cumulative fraction of response (CFR).
Toxicity-oriented exposure screening to flag regimens whose exposure profiles warrant caution.
Full component list:
Model behavior is checked against external evidence and stress-tested for stability:
The public repository contains model configuration, provenance controls, deterministic tests, scientific audit documentation, and reproducibility materials associated with the methodological manuscript.
This project is a methodological research framework. It is not a clinical dosing tool, treatment recommendation, or validated decision-support system.
The complete framework — configuration, provenance controls, tests, and documentation — is available on GitHub.
The associated manuscript is currently under submission. This page will be updated with citation details once a formal reference is available. Please treat all materials as a methodological research framework rather than peer-reviewed, published findings.
For related work in antimicrobial resistance and computational microbiology, see the Publications page.