Right Dose, Right Now: Model Validation

(Of note: This study is a part of the Right dose, Right now project. For the
full rationale of the project, a separate approved project request has been
added to the appendix (gehonoreerde projectaanvraag). Also, additional
information can be found on http://www.autokinetics.eu.)

Sepsis is a major and growing problem. In the Netherlands alone, around 15.000
patients are diagnosed with severe sepsis each year. Despite major scientific
efforts, including many failed clinical trials mainly focusing on inflammatory
mediators and the introduction of care bundles, the mortality rate for severe
sepsis still remains unacceptably high at around 30%.
This is alarming, especially since the incidence of sepsis continues to
increase and now exceeds that of colon cancer, breast cancer and AIDS combined.
Antibiotics are essential for treating sepsis. Their early and appropriate use
has repetitively been shown to reduce mortality rates. However, achieving
adequate antibiotic exposure in critically ill patients is a major challenge
due to markedly different pharmacokinetic (PK) profiles in the critically ill.
Nevertheless, doctors still rely on standard antibiotic dosing schemes, that
were developed based on data from healthy volunteers and non-critically ill
patients. Depending on patient characteristics, clinical course and therapy,
this strategy may result in underdosing and/or drug-related toxicity during the
course of intensive care treatment.
Therefore, we developed AutoKinetics (AutoK) software. AutoK aims to make use
of patient data that is available from the electronic patient records, for
example about fluid balance and renal function. Using this data, AutoK is able
to give fast and precise dosing advice, using published population
pharmacokinetic models of any drug. AutoK runs on the computer at the bedside.
Thus, advice is readily available, even before treatment is started, and is
continuously updated as disease and therapy evolve: true personalised dosing.
We believe that AutoK can improve antibiotic dosing, morbidity and mortality
for severe sepsis.

Eventually, we will test AutoK in multicenter clinical trial. We will randomise
patients with severe sepsis (n=42 per group, per antibiotic), for antibiotic
dosing through AutoK or standard therapy. This will also concern a WMO request.
We will file a separate request accordingly.

This request concerns the validation of pharmacometric models using
prospectively collected data which is not being collected in the context of
regular patient care and / or quality controls.

To select the most appropriate pharmacometric models

This study concerns model validation based on prospectively gathered data that
is not being collected in the context of regular patient care and / or quality
controls.

We will validate existing pharmacometric models which are publically available
from the international literature. We will select these models using a
standardized search strategy. Examples of this strategy can be found in the
attached approved project request (gehonoreerde projectaanvraag). We only
intend to select pharmacometric models of common and relevant antibiotics and
antimycotic agents used at the intensive care unit.
For the validation of the selected pharmacometric models, we aim to use both
retrospectively available data and prospectively collected data. The exact
content of data needed may vary and is dependent on the specific pharmacometric
model used. A large part of the data will be routinely collected in the context
of regular patient care and / or quality controls. In general, the data needed
consists of demographic data like age and gender, data concerning time and
quantity of medication administration and data generated by patient
surveillance monitors and the laboratory. For the retrospective part of this
research a separate non-WMO request has been filed, which has already been
approved by the METc VUmc.

This specific request concerns model validation using data that has not or will
not be collected in the context of regular patient care and / or quality
controls. We aim to collect this data prospectively. The data needed mainly
concerns serum concentrations at different time points of all relevant
antibiotic and/or antimycotic agents commonly used at the intensive care unit.

If the selected pharmacometric models are not accurate enough according to
Bland Altman anaylsis, as judged by > 20% error, we will attempt to either
combine or calibrate the models. If this does not appear to be sufficient, we
will develop new models using a part of the retrospectively and prospectively
collected data. The remaining part of the data will subsequently be used to
validate the newly developed models.

The blood samples are taken from existing lines. Therefore, the burden is
minimal and the risk negligible.

This study is group related. The objective of this study is related to the
prediction of antibiotic plasma concentrations among intensive care patients.
By definition, this cannot be investigated other than within the group of
intensive care patients.