Plasma and Cerebrospinal Fluid Pharmacokinetics of Intravenous Oxaliplatin, Cisplatin, and Carboplatin in Nonhuman Primates

The platinum analogues, cisplatin and carboplatin, are chemically reactive anticancer drugs that produce a cytotoxic effect through platination of DNA and the formation of cross-links. These agents have a broad spectrum of antitumor activity, including modest activity against brain tumors (1). Oxaliplatin is a new platinum analogue that contains a 1,2-diaminocyclohexane carrier ligand, which confers enhanced cytotoxicity, an altered drug resistance pattern, and a unique toxicity profile in comparison to cisplatin and carboplatin (2, 3). Oxaliplatin in combination with 5-fluorouracil is approved treatment of metastatic colon cancer, but its potential role in treating primary and metastatic brain tumors has yet to be defined (4, 5).

In addition to their affinity for DNA, platinum analogues undergo spontaneous chemical reactions and irreversibly bind to proteins and low molecular weight nucleophiles (e.g., methionine). These protein-bound and low molecular weight platinum complexes are not cytotoxic to cancer cells, but may contribute to the drugs' toxic effects, such as the nephrotoxicity from cisplatin (2, 5–7).

For pharmacokinetic studies, the inactive protein bound platinum in plasma is separated from low molecular weight platinum species by ultrafiltration. Elemental platinum in the ultrafiltrate is then usually measured by atomic absorption spectrometry. The ultrafiltrate contains the active parent drug but also contains inactive platinum metabolites, which over time account for an increasing proportion of the ultrafiltrable platinum. For cisplatin, which is the most chemically reactive platinum analogue, parent drug accounts for <1% of platinum in plasma ultrafiltrate four hours after an i.p. dose in rats (7). Therefore, elemental platinum concentration in plasma ultrafiltrate may not accurately reflect the concentration profile of active parent drug, especially at later time points, and methods to more specifically estimate parent drug concentration are needed.

The blood-brain barrier limits access of many anticancer drugs to the brain. Identifying drugs that penetrate the blood-brain barrier may be one factor in selecting new agents or specific analogues from a class of agents for clinical development in brain tumors. Directly sampling brain tissue drug concentrations after a systemic dose is not practical; consequently, cerebrospinal fluid (CSF) drug penetration is usually used as a surrogate for blood-brain barrier penetration. The CSF pharmacokinetics of cisplatin and carboplatin have been characterized, (8–13), but the CSF penetration of oxaliplatin has not been studied.

We compared the plasma and CSF pharmacokinetics of oxaliplatin with that of cisplatin and carboplatin in a nonhuman primate model that has proven highly predictive of central nervous system pharmacology of drugs in humans. To more accurately estimate the concentration of active parent drug in plasma ultrafiltrate and CSF, we developed a pharmacokinetic model to distinguish the disposition of the parent drug from the inert (inactive) platinum species that are detected by atomic absorption spectrometry.

Drugs. Drugs were obtained commercially and diluted in 5% dextrose or 0.9% saline for infusion. Oxaliplatin (0.4 mg/mL, Sanofi Synthelabo, Bedford, OH) was given at a dose of 5 mg/kg i.v. over 2 hours (human equivalent, 110 mg/m²) to four animals or 7 mg/kg i.v. over 1.5 hours (human equivalent, 140 mg/m²) to one animal. Cisplatin (1 mg/1 mL, Bristol-Meyers Squibb, Princeton, NJ) was given to three animals at a dose of 2 mg/kg i.v. over 1 hour (human equivalent, 40 mg/m²). Animals receiving cisplatin were prehydrated for 1 hour with 200 mL of sodium chloride with 150 mg/kg sodium thiosulfate (250 mg/mL, American Reagent Laboratory, Shirley, NY). Over 3 hours following the cisplatin infusion, they received 300 mL NaCl mixed with 350 mg/kg sodium thiosulfate. Carboplatin (10 mg/mL, Bristol-Meyers Squibb) was given at a dose of 10 mg/kg i.v. over 1 hour (human equivalent, 200 mg/m²). All drugs were infused through a central venous catheter. Antiemetics and i.v. fluids were given to each animal.

Animals. Five adult male rhesus monkeys (Macaca mulatta) ranging in weight from 8.1 to 15.7 kg were used in the study. The experimental protocol was approved by the National Cancer Institute Animal Care and Use Committee. Oxaliplatin, cisplatin and carboplatin were each given on separate occasions to three of the animals, and oxaliplatin was studied in two additional animals. All animals were fed NIH Open Formula Extruded Nonhuman Primate Diet twice daily and group housed in accordance with the Guide for the Care and Use of Laboratory Animals (14). Blood samples were drawn through a temporary saphenous vein catheter, placed contralateral to the site of drug administration. CSF was drawn from a temporary lumbar catheter (n = 1, oxaliplatin) or from a chronically indwelling Pudenz catheter implanted in the fourth ventricle, attached to an implanted Ommaya reservoir (ref. 15; n = 4).

Experiments. Blood (3 mL) and CSF (0.3 mL) samples were obtained before infusion, at 30 and 60 minutes (and 120 minutes for oxaliplatin) after the start of infusion, and then 15 and 30 minutes and 1, 2, 4, 6, 8 hours (10 hours for cisplatin and carboplatin) and 24 hours after the end of infusion. Whole blood was immediately centrifuged, and the plasma was centrifuged through a Microcon 10K MWCO filter (Millipore Co., Bedford, MA) at 12,000 rpm for 40 minutes at 10°C. The plasma ultrafiltrate and CSF samples were immediately frozen at −70°C.

Sample Analysis. Elemental platinum concentrations in plasma ultrafiltrates and CSF were measured with a Perkin-Elmer AAnalyst 800 Atomic Absorption Spectrometer (Perkin-Elmer Co., Norwalk, CT) with an AS autosampler and HGA-800 graphite furnace. Twenty microliters of sample was injected and the furnace was heated slowly to 2,550°Celsius. The absorbance of atomized platinum was measured at 265.7 nm. The assays for cisplatin, carboplatin and oxaliplatin were validated according to the Food and Drug Administration guidelines (16). The standard curves were linear over a range 0.1 to 5 μm. CSF samples were reliably concentrated up to 5-fold, and the lower limit of quantification for CSF was 0.02 μmol/L. The intraday and interday coefficients of variation were <10%, and precision and accuracy exceeded 80%.

Pharmacokinetic Analysis. A compartmental model (Fig. 1) was developed to distinguish active parent drug from inactive metabolites in plasma ultrafiltrates and CSF. The model was fit to plasma ultrafiltrates and CSF platinum concentration-time data in 2 stages using MLAB (Civilized Software, Bethesda, MD). Initially the systemic drug compartments (C_o and C_m) were fit to the plasma ultrafiltrates platinum concentration-time data for each animal separately. The systemic model variables (V_o, V_m, k_om, k_om, k_oe, k_me) were then fixed using best-fitted values, and the CSF variables (k_oc, k_mc, k_oce, k_mce) were fit to the CSF platinum concentration-time data. CSF volume (V_csf) was fixed at 15 mL. This two-stage approach was based on the assumption built into the model that the transfer of drug into and out of the CSF compartment has no significant impact on systemic concentrations of parent drug or metabolite because of the small amount of drug and metabolite in this compartment.

Areas under the concentration versus time curves (AUC) were calculated using the linear trapezoidal method. AUCs were also derived from model-simulated curves for ultrafiltrable platinum and the active parent drug and inactive metabolite components of ultrafiltrable platinum in plasma and CSF. These results were compared with the noncompartmental AUCs measured for ultrafiltrable platinum for each drug. The fraction of drug or metabolite penetrating the CSF was derived from the ratio of the AUCs in CSF to plasma. The half-lives for total platinum and active drug were calculated by dividing 0.693 by the terminal rate constant.

Model-independent plasma ultrafiltrate and CSF pharmacokinetic variables for oxaliplatin, cisplatin and carboplatin are presented in Table 1. The mean (±SD) peak plasma ultrafiltrable platinum concentrations were 10.6 ± 2.6, 12.5 ± 3.9, and 93.5 ± 14.6 μmol/L for oxaliplatin (5 mg/kg), cisplatin, and carboplatin, respectively. Platinum concentrations peaked in CSF within 30 minutes of the end of i.v. infusion and were substantially lower than plasma ultrafiltrate concentrations for all three platinum analogues. Mean (±SD) peak CSF concentrations were 0.21 ± 0.11, 0.19 ± 0.03, 0.89 ± 0.12 μmol/L for oxaliplatin (5 mg/kg), cisplatin, and carboplatin, respectively. The mean (±SD) AUC of plasma ultrafiltrate was 61 ± 22 μmol/L hour for oxaliplatin (5 mg/kg), 18 ± 6 μmol/L hour for cisplatin and 211 ± 64 μmol/L hour for carboplatin; and the mean (±SD) AUC of CSF was 1.2 ± 0.4 μmol/L hour for oxaliplatin (5 mg/kg), 0.56 ± 0.08 μmol/L hour for cisplatin, and 8 ± 2.2 μmol/L hour for carboplatin. The mean CSF penetrations (CSF AUC to plasma ultrafiltrate AUC ratios) for oxaliplatin, cisplatin, and carboplatin were 2.0%, 3.6%, and 3.8%, respectively.

In an attempt to estimate the fraction of plasma ultrafiltrate and CSF platinum that represents parent or active drug for each agent, we developed and fit a compartmental model (Fig. 1) to the plasma ultrafiltrate and CSF platinum concentration-time data, which was measured using the standard nonspecific atomic absorption spectrometry method. For the three platinum analogues, the plasma concentration-time profiles were biexponential (Fig. 2) with the more prolonged terminal elimination phase presumably representing inactive (nonreactive) low molecular weight platinum complexes (metabolites) that are formed through spontaneous chemical reactions with low molecular weight nucleophiles. The model was fit to plasma ultrafiltrate and CSF platinum concentration-time data for each animal with each agent individually in two stages (plasma ultrafiltrate then CSF). Pharmacokinetic model variables from these fits are listed in Table 2, and measured plasma ultrafiltrate and CSF platinum concentrations and simulated curves from the model variables from representative animals are shown in Fig. 2.

Pharmacokinetic variables were derived from the model variables for the parent drug (C_o) and low molecular weight metabolites (C_m) and are shown in Table 3. In plasma ultrafiltrate, the simulated AUC for the active form of oxaliplatin was 17.6 ± 4.2 μmol/L hour, which is 29% of measured platinum AUC in plasma ultrafiltrate; for the active form of cisplatin the AUC was 14.3 ± 4.4 μmol/L hour, which is 79% of measured platinum AUC, and for the active form of carboplatin the AUC was 170 ± 46 μmol/L hour, which is 81% of measured platinum AUC. In CSF, the simulated AUC for active oxaliplatin was 0.3 ± 0.15 μmol/L hour, which is 25% of measured platinum AUC in CSF, for active cisplatin the AUC in CSF was 0.5 ± 0.12 μmol/L hour which is 89% of measured platinum AUC in CSF, and for active carboplatin the AUC was 4.5 ± 1.5 μmol/L hour which is 56% of measured platinum AUC in CSF. The CSF to plasma ratios for active drugs

The compartmental model seems to be predictive of the pharmacology of the platinum analogues. Based on the fitted model variables, the conversion of parent drug to inactive metabolites accounts for >90% of the clearance of cisplatin and oxaliplatin, but <20% of carboplatin clearance, which is known be primarily renally excreted. The clearance of oxaliplatin parent drug is 2-fold greater than that of cisplatin, indicating that oxaliplatin is more reactive than cisplatin. However, if the clearance of oxaliplatin and cisplatin are calculated using the AUC of the measured plasma ultrafiltrate platinum concentrations (from Table 1), the clearance of cisplatin (4.04 ± 0.26 L/h) exceeds that of oxaliplatin (2.40 ± 0.48 L/h), because inactive metabolites account for a larger proportion of ultrafiltrable platinum in the plasma after oxaliplatin infusion than after cisplatin. The plasma half-life of active oxaliplatin is 0.2 hour, which is comparable with the previously reported α half-life (distributive phase) of 0.28 hour (17) and shorter than the plasma half-life of active cisplatin (0.3 hour).

We used a nonhuman primate model that is predictive of the CSF pharmacology of anticancer drugs in humans to study the CSF penetration of the new platinum analogue, oxaliplatin, and compare it with the CSF penetration to cisplatin and carboplatin. Multiple CSF samples were obtained from the animals after a short drug infusion to characterize the time course of drug concentration in the CSF. Drug exposure (AUC) in CSF was calculated and the CSF penetration was derived from the ratio of the AUC in CSF to AUC in plasma. This is a more accurate means of assessing CSF penetration under nonsteady state conditions, because the shape of the plasma and CSF drug concentration-time curves are different, and the ratio of individual CSF to plasma concentrations will be highly dependent on the timing of the samples relative to the dose. For example, the ratio of individual, simultaneous CSF to plasma carboplatin concentrations ranged from <1% to 52% in the three animal studied.

Using this method, the CSF penetrations of the three platinum analogues using both compartmental and noncompartmental methods were comparable and limited (<5%). The CSF penetration of cisplatin in our model is similar to a prior study (18), but the CSF penetration of carboplatin is lower than previous reports (8–10). The disparity between our study and these previous reports with carboplatin may be related to the limited number CSF samples measured in prior studies. Oxaliplatin and cisplatin are neurotoxic, but the toxicity is primarily expressed as peripheral neuropathy. The limited penetration of these agents into the central nervous system may explain why they do not produce central nervous system toxicity.

The pharmacokinetic model developed for this study provided interesting insights into the pharmacokinetic behavior of the platinum analogues and may have broader use in analyzing pharmacokinetic data generated with the standard but nonspecific atomic absorption spectrometry assay. Oxaliplatin parent drug was more reactive than cisplatin in our animals, based on the pharmacokinetic model. The rate constant describing metabolic conversion to less reactive metabolites (k_om) was 3.6 h⁻¹ for oxaliplatin and 1.9 h⁻¹ for cisplatin. The rapid metabolic conversion of oxaliplatin (parent drug) is consistent with prior studies (2, 5, 6) and with the large number of low molecular weight platinum complexes in plasma ultrafiltrates from patients (19). Although renal excretion has been said to account for 50% of the elimination of oxaliplatin (17, 20), when fractioned by high-performance liquid chromatography, the urinary platinum consists of the same low molecular weight platinum metabolites that are seen in plasma ultrafiltrates rather than parent drug (19). Our model findings that predict a minor role for renal excretion are consistent with the finding that oxaliplatin dose reduction is not necessary in patients with renal insufficiency (21).

According to the pharmacokinetic model, parent drug accounts for 79% and 81% of platinum in plasma ultrafiltrates for cisplatin and carboplatin, respectively, but only 29% for oxaliplatin. This suggests that oxaliplatin is more reactive with small molecular weight nucleophiles, which remain in the plasma ultrafiltrates, than cisplatin, which binds to proteins that are removed by the ultrafiltration step in sample preparation. Measuring elemental platinum in plasma ultrafiltrate is assumed to be equivalent to measuring active drug (5, 17, 22). This is a reasonable assumption for calculating the AUC of carboplatin or cisplatin, but measurement of platinum concentration in plasma ultrafiltrate by atomic absorption spectrometry seems to be a poor surrogate for oxaliplatin concentration in plasma. Our pharmacokinetic model does not account for drug that reacts with plasma proteins, because we did not measure total platinum in plasma or platinum in the protein fraction after ultrafiltration. It seems that the model compensates for this elimination pathway by estimating very large volumes of distribution (V_m) for the metabolites. The 4-fold larger V_m for cisplatin (219 L) compared with oxaliplatin (54 L) would be consistent with the greater degree of protein binding with cisplatin. Furthermore, the model predicts that the parent drug fraction of the platinum is lower in CSF for carboplatin (56%) and oxaliplatin (25%) than the fraction of active drug in plasma ultrafiltrate.

In conclusion, at doses approximating those used in humans, CSF concentrations of the platinum analogues peaked at 0.2 μmol/L for cisplatin and oxaliplatin and 0.9 μmol/L for carboplatin and drug exposures (AUC) in CSF were <5% of the exposures achieved in plasma for all three analogues.