Due to the lack of existing information on baricitinib's potential to cross the blood–brain barrier in humans, a permeation assay was performed by Eurofins in the MDCKII cell line. The mean permeability of baricitinib from the apical to the basolateral side (A to B) was 4.5 × 10
−6 cm/s, and the B to A was 5.5 × 10
−6 cm/s. According to Palmer and Alavijeh,
25 (
link) this moderate permeability value falls within the acceptable range for a desired target profile of a central nervous system (CNS) drug candidate.
A PBPK model, with seven compartments, among them brain vasculature and brain tissue, for baricitinib was developed in Berkeley Madonna 10.
26 The PBPK model structure is presented in Figure
S1 in supporting information, and the modeling script, including annotations of parameters, is provided as supporting information (Modelling Script). Physiological parameters included organ volumes and blood flow rates for a standard human male.
27 (
link),
28 (
link),
29 (
link),
30 (
link) Blood‐to‐tissue partition coefficients were estimated in silico from Rodger & Rowland's algorithm based on log
K,
pKa, and molecular weight.
31 (
link) Absorption rates and clearance values were from a previous baricitinib model.
32 (
link) Administration was modeled as a single oral dose of 4 mg (maximum recommended daily dose) baricitinib. From the gut, uptake to the liver was modeled with a first‐order rate constant determined in a previous study,
28 (
link) then distributed to systemic circulation. Estimates of baricitinib concentrations in blood plasma over time were validated with a previous model.
32 (
link) Other organs were categorized as slowly perfused (i.e., muscles, adipose, bone, skin) or rapidly perfused (i.e., heart, lung, spleen, kidneys) tissue. Urinary excretion was modeled based on previously established clearance values.
32 (
link)
Concentrations of baricitinib in the brain were computed by different approaches (Figure
S1). First, it was estimated to be 0.91% of blood concentration, as suggested by the Quantitative Structure–Activity Relationship (QSAR) tool of the PreADMET webserver).
33 (
link),
34 This estimation is further referred to as “Prediction 1 [QSAR].” The second approach, “Prediction 2 [Mouse exp.]”, relied on a brain‐to‐plasma concentration ratio of 20%, which was experimentally observed in mice.
35 (
link) Finally, the third approach involved modeling of the blood–brain barrier (BBB) permeation using the quantitative in vitro–in vivo scaling methodology developed by Ball et al.
30 (
link) This last estimation of baricitinib concentration in the brain tissue is herein mentioned as “Prediction 3 [QIVIVE BBB].”
The impact of parameter deviation on the model's predictions was assessed by a sensitivity analysis based on the method by Evans and Andersen
36 (
link) (see Table
S2 in supporting information). For this, each model parameter was individually increased by 5% and the associated impact on maximal brain tissue concentrations (Prediction 1 [QSAR], Prediction 2 [Mouse exp.], and Prediction 3 [QIVIVE BBB]) was computed. Oral administered dose was maintained at 4 mg. Normalized sensitivity coefficients (SC) were determined by using Equation
1:
C and C’ refer to the maximal concentration of baricitinib in brain tissue (Prediction 1 [QSAR], Prediction 2 [Mouse exp.], or Prediction 3 [QIVIVE BBB]) with unchanged parameters or one elevated parameter, respectively, P and P’ to the value of the unchanged or elevated parameter of interest.
To address the impact of parameter uncertainty on these predicted concentrations of baricitinib in the brain, their calculation has been iteratively repeated 1000 times, with the most sensitive parameters (having the greatest influence on the results) being re‐sampled in each iteration. Monte Carlo simulations were performed with Berkeley Madonna 10 associated functions for all the parameters found with an absolute value of normalized sensitivity coefficient > 0.1 for at least one brain concentration (Prediction 1 [QSAR], Prediction 2 [Mouse exp.], or Prediction 3 [QIVIVE BBB]). Parameter simulated distributions were determined according to literature.
32 (
link),
37 (
link),
38 (
link) One thousand simulations were performed, and results were analyzed by comparing first quartile, median, and third quartile values for each time point for each of the brain concentrations (Prediction 1 [QSAR], Prediction 2 [Mouse exp.], and Prediction 3 [QIVIVE BBB]).
Faquetti M.L., Slappendel L., Bigonne H., Grisoni F., Schneider P., Aichinger G., Schneider G., Sturla S.J, & Burden A.M. (2024). Baricitinib and tofacitinib off‐target profile, with a focus on Alzheimer's disease. Alzheimer's & Dementia : Translational Research & Clinical Interventions, 10(1), e12445.