Hughes [Thomson] et al. 2021

A hybrid machine learning/pharmacokinetic approach outperforms maximum a posteriori Bayesian estimation by selectively flattening model priors

CPT: Pharmacometrics & Systems Pharmacology · 10.1002/psp4.12684

Structure

A two-compartment model, driven by weight and creatinine clearance.

Study population

General patients — United States and North America. Fitted to 2,700 patients and 5,153 concentration observations.

Covariate As reported
Age 64.2 (IQR 31.2–87.5) yr
Weight 84.2 (IQR 52–153.4) kg
BMI 28.1 (IQR 19–47) kg/m²
SCr 0.9 (IQR 0.5–2.6) mg/dL

Reported as the source publication gives them — mean ± SD, or median with range or interquartile range.

Equations

The model's structural parameters as Chorus implements them.

CL=2.42(1+0.0126(CrCl66))CL = 2.42 \cdot (1 + 0.0126 \cdot (CrCl - 66))V1=0.667WeightV_1 = 0.667 \cdot WeightV2=0.737WeightV_2 = 0.737 \cdot WeightQ=2.82Q = 2.82
LaTeX source
CL = 2.42 \cdot (1 + 0.0126 \cdot (CrCl - 66))
V_1 = 0.667 \cdot Weight
V_2 = 0.737 \cdot Weight
Q = 2.82

Implementation notes

Re-estimated Thomson 2009 model. CrCl capped at 150 mL/min (capped Thomson).

Simulation settings

Models

selected

Dose candidates (mg)

Interval candidates (h)

Demographics

Physical characteristics.

kg
cm
yr