SÃlvia M. Illamola(1), Ju-Hee Oh(2), Angela K. Birnbaum(1), Minjee Kim(2), Ann C. Tuma(3), Terry C. Burns(3), Cecile Riviere-Cazaux(3), Ian Parney(3), William F. Elmquist(2), Jann N. Sarkaria(3)
(1) Department of Experimental and Clinical Pharmacology, College of Pharmacy, University of Minnesota, Minneapolis, Minnesota. (2) Pharmaceutics, University of Minnesota, United States. (3) Department of Radiation Oncology, Mayo Clinic, Rochester, MN.
Objectives: The blood-brain barrier (BBB) is a critical obstacle to delivery of cytotoxic chemotherapeutics to the central nervous system (CNS). The integrity of the BBB may influence the treatment efficacy for primary brain tumors by limiting the brain penetration of chemotherapy agents [1]. Magnetic resonance imaging (MRI) along with gadolinium contrast allows noninvasive characterization BBB integrity (contrast (CON)), perfusion abnormalities (fluid-attenuated inversion recovery (FLAIR)), and detection of necrotic areas (NEC). In this study, we employed a brain penetrant drug (levetiracetam (LEV)) along with a brain impenetrant drug (cefazolin (CEFA)) to characterize BBB integrity in individuals with a CNS malignancy. The aim of this study was to define the heterogeneity of the BBB by assessing drug distribution in patients with CNS malignancy, and to establish a correlation of BBB permeability with radiographic characteristics for these individuals.
Methods: Prospective observational study in individuals with clinical and radiographic evidence suggesting CNS malignancy. Individuals (>18 years old) were included if they, (1) had a suspected newly diagnosed, local, or intracranial recurrence of primary brain tumor or previously untreated or treated brain metastasis; (2) were able to have MRI with gadolinium contrast; and (3) had an estimated glomerular filtration rate ≥ 30 mL/min/1.73m2. Exclusion criteria included: (1) unable to undergo a biopsy of CNS lesion; (2) documented drug allergy to LEV or CEFA; (3) pregnant or nursing.
Participants had a MRI imaging just prior to resection, which provided MRI characteristics (i.e., CON, FLAIR, NEC) and the localization of the brain tumor. Tissue biopsies were taken from 3-4 tumor locations. Individuals received LEV and CEFA pre-operatively in the operative room by intravenous injection. During surgery, blood samples were collected every 20-30 minutes beginning at the time of skin incision until all tumor samples were removed. The time of removal for each brain specimen, time of administration of each drug, and time of collection of blood samples were recorded. Plasma and brain concentrations of LEV and CEFA were analyzed by LC-MS/MS.
A population pharmacokinetic model was developed for plasma samples. Plasma covariates (i.e., weight, age, sex, eGFR) were tested in the model. Once the plasma model was developed, we fixed all the pharmacokinetic parameters from plasma and we estimated the penetration factor (Fcns) of each compound to the brain, considering brain concentrations as a fraction of plasma concentrations. We evaluated M1 and M3 methods [2] to treat CEFA brain concentrations below the limit of quantification (BLQ). Brain covariates (i.e., MRI imaging, RAD) were tested on the Fcns. Models were assessed using typical methods including goodness of fit diagnostics, objective function, and bootstrap. Pharmacokinetic analyses were conducted using NONMEM® (version 7.4.3) [3] and R version 4.2.1.
Results: A total of 39 (202 plasma and 95 brain concentrations) and 42 individuals (196 plasma and 92 brain concentrations) were included for LEV and CEFA analyses, respectively. For LEV and CEFA, plasma data were described as a one-compartment model with proportional error, and including inter-individual variability on clearance and volume of distribution. We did not find any significant covariate in any of the plasma models. The M1 method performed better than M3 for CEFA, and we retained it as a method to treat CEFA brain concentrations BLQ. After including brain data, the Fcns was estimated as 0.667 (6.08%) and 0.116 (17.63%) for LEV and CEFA, in the base model, respectively. All covariates were significant on Fcns for LEV, and only CON and FLAIR were significant for CEFA. However, only CON was included in the final LEV and CEFA models. Fcns was estimated as 0.800 (7.82%) for CON + and 0.573 (6.65%) for CON – in the LEV model and as 0.248 (18.51%) for CON + and 0.044 (26.52%) for CON – in the CEFA model.
Conclusions: Our analysis indicates that LEV and CEFA are feasible markers for BBB permeability and heterogeneity, and imaging covariates can correlate, depending on the drug molecule, with BBB permeability. The results from our work can be used to further guide treatment decisions in individuals with diagnosed CNS malignancies.
References:
[1] Sarkaria JN et al. Neuro Oncol. 2018 Jan 22;20(2):184-191.
[2] Beal, S.L. J Pharmacokinet Pharmacodyn, 2001.28(5): p. 481-504.
[3] Boeckmann A, Sheiner L, Beal S. 2001. NONMEM Users Guide. University of California, San Francisco, CA.
Reference: PAGE 32 (2024) Abstr 11092 [www.page-meeting.org/?abstract=11092]
Poster: Drug/Disease Modelling - CNS