Segmented regression to streamline semi-mechanistic pharmacodynamic modelling of antibiotic effects

Segmented regression of 24-hour time-kill data for colistin against Pseudomonas aeruginosa captured kill and regrowth slopes at all concentrations, revealing weaker regrowth drug effects.

• Why it matters: Semi-mechanistic pharmacodynamic models often require inferring subpopulation dynamics from total counts, creating identifiability problems and heavy model-testing and computation.
• What they did: The workflow segmented log-transformed time-kill curves into kill and regrowth phases, used linear regression to estimate phase-specific slopes per concentration and replicate, pooled slopes into concentration-effect relationships, then fit a semi-mechanistic PD model using selected relationships and parameters.
• The result: The semi-mechanistic model described bacterial dynamics across all concentrations, with parameters comparable to those derived in the slope-pooling step, enabling more efficient and robust model development.

AI summary of the abstract · may contain errors · not medical advice · read the paper.