Skip to main content

TDM Example in R Shiny

The dose of beta-lactam antibiotics is currently based on clinical trials from 20 years ago. These studies determined the population-recommended dose based on efficacy, but did not undertake any pharmacokinetic (PK) modeling. The efficacy target did not take Emergence of Drug Resistance (EDR) into account. Currently, the dose of beta-lactams is adjusted empirically on patients with augmented renal clearance or renal impairment.

For a European research project, SGS Exprimo was asked to develop an algorithm for individualized dosing recommendations of beta-lactams, based on observed meropenem serum concentrations. To communicate the Therapeutic Drug Monitoring (TDM) algorithm easily to physicians, some interactive presentations were created.

In a first presentation, we have provided a very simple dosing algorithm, assuming we know the patient pharmacokinetic (PK) parameters exactly. We first pick these parameters, and then explain how we can predict concentrations. In the next slides, we provide population predictions of the recommended dose, ensuring we reach our PK target (Conc > 4*MIC for 100% of dosing interval): first for a variable dose, then for discrete doses. Finally, we present a restricted table with only a few options, and we pick the best treatment for our patient among these options.

In a second presentation, we can now repeat the exercise with an unknown patient. We can predict PK from the (covariate-adjusted) population distribution, but we can also amend this information by looking at serum concentration in blood samples. We can then do the exact same exercise as in the previous paragraph, but now we will have a statistical distribution of predicted values, based on the statistical distribution of estimated PK parameters. We pick a dosing regimen by calculating the probability of reaching efficacy and safety, and only select a regimen if the probability is high enough.

We hope you enjoy this material. Please get in touch if you want to discuss this further!


First presentation
TDM using known PK values



Second presentation
TDM for bayesian estimated PK


How can we help you?