Clinical Study

Tamoxifen paradoxically decreases paclitaxel deposition into cerebrospinal fluid

of brain tumor patients

Johnson Chen1, Casilda Balmaceda2, Jeffrey N. Bruce3, Michael B. Sisti3, May Huang4, Ying Kuen K. Cheung5,Guy M. McKhann3, Robert R. Goodman3 and Robert L. Fine11Experimental Therapeutics Program, Division of Medical Oncology, College of Physicians and Surgeons of ColumbiaUniversity, New York, NY, USA; 2Division of Neuro-Oncology, Department of Neurology, Neurological Institute ofNew York, College of Physicians and Surgeons of Columbia University, New York, NY, USA; 3Department ofNeurological Surgery, Neurological Institute of New York, College of Physicians and Surgeons of ColumbiaUniversity, New York, NY, USA; 4Herbert Irving Center for Clinical Research, GCRC, College of Physicians andSurgeons of Columbia University, New York, NY, USA; 5Department of Biostatistics, Mailman School of PublicHealth, Columbia University, New York, NY, USA

Key words: brain neoplasms, cerebrospinal fluid, drug resistance, P-glycoprotein, tamoxifen

Summary

Background: P-glycoprotein (Pgp) mediates, in part, resistance to natural product chemotherapy drugs whichconstitute over half of the available drugs for cancer treatment. Tamoxifen (TAM) enhances intracellular depositionof natural product chemotherapy in human cell lines by inhibition of Pgp. Pgp is highly expressed in the choroidplexus and is thought to be a key component of the blood–cerebrospinal fluid barrier (BCSFB). We conducted aprospective, randomized study to assess if Pgp inhibition by TAM alters deposition of paclitaxel in cerebrospinalfluid (CSF).

Methods: Ten patients with either primary or metastatic brain tumors were randomized to: paclitaxel alone(175 mg/m2/IV) or a course of TAM (160 mg/m2 PO BID on Days 1–5) followed by paclitaxel (175 mg/m2/IV onDay 5). CSF and plasma samples were obtained following paclitaxel infusion; paclitaxel and TAM concentrationswere measured by high-performance liquid chromatography assays.

Results: Paclitaxel was detected in the CSF of six of the 10 patients. Peak CSF paclitaxel concentrations of thepaclitaxel and paclitaxel–TAM groups ranged between 3.5–57.4 and 2.3–24.6 nM, respectively. Though there was a2.4-fold higher mean CSF paclitaxel concentration and a 3.7-fold higher median peak CSF:plasma paclitaxel ratiofor those who received paclitaxel alone as compared to combined paclitaxel–TAM, it was not statistically significant(P = 0.22). In one patient enrolled to both arms, higher CSF concentrations of paclitaxel and higher paclitaxelCSF: plasma ratios were observed when given paclitaxel alone.

Conclusions: The trend towards lower paclitaxel CSF concentrations when given with TAM is consistent with thepublished finding that Pgp�s localization in the endothelial cells of the choroid plexus works in an opposite directionand keeps drugs in the CSF. Thus, agents which inhibit Pgp, such as TAM, may increase efflux of Pgp substrates outof the BCSFB and may paradoxically lower CSF concentrations of natural product chemotherapy drugs. Concep-tually, this finding implies that the Pgp in the BBB and BCSFB keeps natural toxins such as paclitaxel, from enteringthe brain (BBB) and, if they do enter the brain, keeps them in the CSF (BCSFB) where they may be less harmful thanif they re-entered the brain. Thus, our work supports this novel idea and adds to the understanding of the functions ofthe BCSFB.

Introduction

Protection of the central nervous system against theinflux of xenobiotics is thought to be mediated primarilyby the blood–brain barrier (BBB) and the blood–cere-brospinal fluid barrier (BCSFB). These barriers alsocontribute to resistance against chemotherapy agentsderived from natural product toxins by limiting theirpenetration into the central nervous system (CNS). Acentral mediator in this multidrug resistance (MDR) isP-glycoprotein (Pgp), encoded by the MDR1 gene, anATP-dependent drug efflux pump highly expressed bycerebral capillary endothelium and by choroid plexus

epithelial cells [1,2]. Thus, it is thought that inhibition ofPgp at the BBB and BCSFB may increase the penetra-tion of natural product anticancer agents, such aspaclitaxel, and aid in treating tumors in the brain andleptomeninges.

Tamoxifen (TAM) has been identified as a Pgp inhib-itor capable of inhibiting the MDR phenotype in varioushuman cell lines [3,4]. We demonstrated that incubationof TAM or its metabolites – namely, N-desmethyltam-oxifen (N-DESTAM) and 4-hydroxytamoxifen (4-OH-TAM) – with tritiated vinblastine increased intracellularaccumulation of vinblastine in human renal cancer lines ina dose-dependent manner which was positively correlated

Journal of Neuro-Oncology (2006) 76: 85–92 � Springer 2006DOI 10.1007/s11060-005-4171-7

with the degree of Pgp expression [4]. Using an electrophilicanalogue of TAM, (3H)TAM aziridine ((3H)TAMA), wefound that TAM covalently and specifically bound Pgpin MDR Chinese Hamster lung cells [5]. The binding of(3H)-TAMA to Pgp was inhibited by natural productdrugs associated with the MDR phenotype, such as vin-cristine, vinblastine, and doxorubicin, but not by syntheticantimetabolites such as 5-fluorouracil and Ara-C. Thisdemonstrated that TAM-induced inhibition of MDRoccurred through direct binding to Pgp�s major drugbinding site [5].

We determined that TAM was transported by Pgp byassessing TAM-induced stimulation of the ATPaseactivity of human Pgp in membrane vesicles derivedfrom Sf9 insect cells stably transfected with the humanMDR1 gene [6]. Using 10 lM verapamil as a positivecontrol (100% stimulation of the Pgp ATPase function),equimolar concentrations of TAM or N-DESTAM werefound to induce 75% stimulation of the Pgp ATPasefunction as compared to verapamil. Thus, TAM inhibitsPgp via direct interaction.

The development of many Pgp inhibitors has beenlimited by the unacceptable toxicity experienced at thedoses required to achieve concentrations necessary tomodulate Pgp activity in vivo. Our Phase I trial byTrump et al. [7] of high-dose TAM over 13 days with a5-day course of vinblastine determined that TAM couldbe safely administered at dosages which achieved plasmaconcentrations equal to those required to inhibit Pgp inpreclinical studies, namely 6–10 lM. Furthermore, themajority of trials with Pgp inhibitors have not shownclinical benefit because of the myriad of drug resistancemechanisms, other than Pgp, in cancer cells. The BBBand BCSFB sites may be ideal areas for interventionwith Pgp inhibitors, such as TAM, because: (1) the BBBcapillary endothelial and choroid plexus Pgp are innormal diploid cells with wild-type p53, thus othermechanism of drug resistance commonly found in can-cer cells (i.e. mutant p53) are unlikely and (2) drugresistance mechanisms in brain tumors are twofold – (a)BBB and BCSFB initially; (b) inherent mechanisms ofresistance within the brain tumor cell; and (3) TAMitself has antiproliferative activity in gliomas. Thus, itis thought that the initial block in the BBB and BCSFBto successful natural product chemotherapy approachesin treating CNS tumors and leptomeningeal carcinoma-tosis could potentially be overcome by MDR inhibitorssuch as TAM.

Paclitaxel has not been shown to be an effectiveagent for primary brain tumors [8], but its wide rangeof utility for many tumor types (i.e. breast, lung) whichfrequently metastasize to the leptomeninges makes itan attractive agent for attempts to increase its deposi-tion into the CSF. To begin addressing this question,we conducted a trial with high-dose TAM to determineits effect on paclitaxel penetration across the BCSFB.Patients with either primary or metastatic brain tumorswere prospectively randomized to receive either pac-litaxel alone or a course of high-dose TAM precedingpaclitaxel, with CSF and plasma drug concentrationsdetermined over the 48 h period following paclitaxelinfusion.

Methods

Patients

Patients, ages 18–75, were considered eligible if they had ahistologically documented primary brain tumor whichrecurred following surgical resection; or an initial brainmetastasis arising from a histologically confirmedsystemic neoplasm (Table 2). Other criteria included:Karnofsky performance status of at least 60%; lifeexpectancy greater than 2 months; no concurrent use oforal contraceptives; no history of deep vein thrombosis orpulmonary embolism in the last 6 months; no history of ahypercoaguable state; and adequate hematologic, renaland hepatic function. Prior chemotherapy, TAM andradiation therapy were allowed as long as there was an8 week hiatus off any therapy. Patients with any seriousmedical illness that was expected to compromise theirability to tolerate paclitaxel were excluded. The Institu-tional Review Board at Columbia University MedicalCenter approved this protocol and written consent wasobtained from all patients. Patient information wasprotected in accordance with HIPAA guidelines.

Chemotherapy regimen

Patients were enrolled from 1998 to 2004 and randomlyassigned to receive either paclitaxel alone (175 mg/m2/IV over 3 h) or to receive 5 days of pretreatment withTAM (160 mg/m2 PO BID Days 1–5) followed bypaclitaxel (175 mg/m2/IV over 3 h on Day 5) 6 h afterthe last dose of TAM (Table 1). The 5-day treatmentplan with TAM, instead of the 13-day treatment pub-lished in Trump et al. [7], was chosen because some ofthese patients required semi-acute medical or surgi-cal intervention. Standard premedications, includingdexamethasone, were administered prior to paclitaxelinfusion.

Pharmacokinetic sampling and analysis

Blood samples were collected from the arm contralateralto that which received paclitaxel. CSF specimens wereobtained from Ommaya reservoirs or lumbar puncture.Blood and CSF samples were obtained concurrently attime points ranging from 0.5 to 48 h following com-pletion of the 3-h paclitaxel infusion. Plasma and CSFsamples were stored at )80 �C for no longer than2 months until analysis. Pharmacokinetic data wereanalyzed using WinNonlin� (Burlington, MA). Thearea under the plasma concentration curve (AUC) for

Table 1. Treatment scheme

D1 D2 D3 D4 D5

Tamoxifena

160 mg/m2 PO BID X X X X X

Paclitaxel

175 mg/m2/IV Xb

a Combined arm only.b Paclitaxel was given 6 h after last dose of TAM.

86

paclitaxel was calculated from 0 to 24 h using thetrapezoidal method.

Drug concentration assays

Plasma and CSF levels of paclitaxel were assayed byHPLC adapted from a method reported by Vergniolet al. [9]. Paclitaxel was extracted from samples byadding 1 ml of 30% acetonitrile and 50 ll of docetaxel(2 ng/ll as an internal standard) to 1 ml of sampleplasma. This mixture was vortexed and applied to a C18solid phase extraction cartridge (100 mg/1 ml). Thesample cartridges were pre-conditioned with 1 ml ofmethanol followed by 1 ml of distilled water (dH2O).Immediately following the addition of sample mixture,the column was washed with 1 ml of dH2O followedby 1 ml of 50% methanol. The adsorbed paclitaxelwas eluted with 250 ll of 90% methanol. The eluantwas analyzed by an HPLC system consisting of a220 · 4.6 mm column packed with Hypersil ODS and amobile phase consisting of 65% methanol that waspumped at 1 ml/min. The eluant was monitored at228 nm. Thismethod has a lower limit of quantitation forpaclitaxel at 2 ng/ml (2.3 nM).

Plasma and CSF concentrations of TAM were assayedby HPLC using a method previously described by Web-ster et al. [10]. In brief, TAMwas extracted by addition of200 ll of acetonitrile and 50 ll of methanol to 50 llsample. This mixture was vortexed and centrifuged withthe resulting supernatant being transferred to an Infrasilquartz cuvette. The sample was photoactivated by ultra-violet light (254 nm) for 2 min. Concentration analysiswas done by an HPLC system consisting of a100 · 4.6 mm column packed with Hypersil ODS and amobile phase consisting of 93% methanol with 0.18%triethylamine pumped at 1 ml/min. Fluorescence wasdetected with an excitation wavelength of 265 nm and anemission wavelength of 375 nm. This method has a lowerlimit for quantitation of TAM at 5 ng/ml (8.9 nM).

Statistical analysis

Differences in the CSF concentration curves of paclit-axel between the two treatment groups were assessed by

the Wilcoxon rank-sum test with intention to treatanalysis. The CSF concentration curves of paclitaxel forall patient observations were assessed and ranked. Thesecond observation when Patient No. 1 received TAM –which occurred following initial randomization andinitial enrollment to the paclitaxel alone arm – wasexcluded from statistical analysis in accordance with theprinciples of intention-to-treat analysis (Table 3). Therank orders for the CSF concentration curves of pac-litaxel were determined according to the followingassumptions: (1) that CSF concentrations of paclitaxeldecrease with time from drug exposure; (2) that con-centration curves consisting solely of samples withundetectable levels of paclitaxel will be assigned a rankof 0; and (3) that CSF concentration curves of paclitaxeldo not cross. If it was observed that CSF concentrationcurves crossed each other for different patients, thosetwo curves were assigned the same rank. Therefore, ahigh rank reflects a high CSF-paclitaxel concentration(Table 3). TheWilcoxon rank-sum test was then performedon the resulting rank lists for the two treatment arms.

Results

Patient characteristics

Nine different patients were enrolled into the study:three patients with primary brain tumors and six withmetastatic brain lesions (Table 2). We included bothprimary and metastatic brain tumor patients into therandomization process because our goal was to assessthe effects of TAM on CSF penetration of paclitaxelacross the BCSFB, not into the tumor type. One of thosepatients, Patient No. 1, was randomized initially topaclitaxel alone and subsequently randomized to thepaclitaxel–TAM arm 9 weeks later. Thus, there were atotal of 10 patient observations in the study. Five of thenine patients were on concurrent phenytoin, a CYP3A-inducing anticonvulsant. Eight of the nine patients wereon daily dexamethasone (also a CYP3A inducer) andall study patients received dexamethasone as part ofthe premedication regimen for paclitaxel infusion. Onlyone patient (Patient No. 1) did not receive a prior

Table 2. Patient characteristics

Patient

No.

Age

(years)/sex

Tumor type Concurrent

CYP3A-inducing

anticonvulsant

Concurrent

chronic

dexamethasone

Prior

brain

radiation

1 34/female Breast No No Yes

2 54/female Breast Yes Yes No

3 57/male Anaplastic

oligodendroglioma

Yes Yes Yes

4 73/male Non-small cell lung No Yes Yes

5 67/male Small cell lung No Yes Yes

6 58/female Glioblastoma

multiforme

Yes Yes No

7 43/male Melanoma Yes Yes Yes

8 43/male Glioblastoma

multiforme

Yes Yes Yes

9 51/male Renal No Yes No

87

CYP3A-inducing agent (e.g. phenytoin, dexamethasone),but did have previous brain irradiation. Six patientsreceived radiation therapy for their brain tumors prior tostudy entry.

No toxicities in renal, hepatic, cardiac, ocular orhematopoietic systems were seen. Grade 1–2 cerebellarataxia was experienced by three patients; these symp-toms resolved within 24–48 h following cessation ofTAM.

Toxicity of paclitaxel

The single dose of paclitaxel resulted in no neurologic,anaphylactic or hematopoietic toxicities above Grade 1.Weekly CBCs for 3 weeks following infusion showedthat all patients had CBC�s with WBC >3.0 · 109/l andplatelets >100 · 109/l.

Effect of tamoxifen on CSF paclitaxel concentrations

A total of 16 CSF samples were obtained from the 9study patients. In the paclitaxel alone arm, there weretwo of eight samples with undetectable levels in the CSF(25%); while in the paclitaxel–TAM combined armthere were four of eight samples with undetectablepaclitaxel levels in the CSF (50%) (Table 3). The highestCSF paclitaxel concentrations were observed at 0.5 h,equivalent to 30 min following the completion of the 3-hpaclitaxel infusion. The CSF concentrations at that timepoint represented only 0.4–1.3% of the concurrentplasma paclitaxel concentration. Patient No. 1, withmetastatic breast cancer to the leptomeninges who wasenrolled in both arms, was the only patient with serialCSF samples with detectable levels of paclitaxel. Theabsolute CSF paclitaxel concentrations for Patient No. 1

were 2–3 times higher at each of the three correspondingtime points when treated with paclitaxel alone ascompared to when the patient received paclitaxel andTAM. In addition, when Patient No. 1 received paclit-axel alone, the ratio of CSF:plasma paclitaxel concen-tration was higher at each time point and increased at ahigher rate over the period from 3 to 24 h when com-pared to the patient�s subsequent paclitaxel–TAM course(Table 3).

Peak paclitaxel concentrations in CSF in the paclit-axel alone and in the paclitaxel–TAM groups rangedfrom 3.5 to 57.4 nM and 2.3 to 24.6 nM, respectively.The mean peak CSF concentration of paclitaxel, calcu-lated from the highest concentration obtained from 0 to48 h after the paclitaxel infusion was 19.1 and 7.9 nMfor the paclitaxel alone and paclitaxel–TAM group,respectively. This represents a 2.4-fold higher mean peakCSF paclitaxel concentration in the paclitaxel alonegroup as compared to the combined paclitaxel–TAMgroup. The median peak CSF:plasma paclitaxel ratioobtained from 0 to 48 h after the paclitaxel infusion was0.101 and 0.027 for the paclitaxel alone and paclitaxel–TAM group, respectively. This translates to a 3.7-foldincrease in median peak paclitaxel CSF:plasma ratio inthe paclitaxel alone group as compared to the paclit-axel–TAM group. Because the mean and median werecalculated based on measurements at various timepoints, it would be difficult to interpret these foldchanges. Therefore, a Wilcoxon test on the ranks of theCSF concentration curves, as described in the StatisticalAnalysis section, was conducted on the paclitaxel alonevs. paclitaxel–TAM groups and gave a P-value of 0.22.Although the test does not indicate statistical signifi-cance, it is likely to be due to the small sample size. Theincreased mean CSF paclitaxel concentration and

Table 3. Effect of TAM on CSF concentrations of paclitaxel

Patient

No.

Received

TAM

Peak Plasma

TAM (lM)aTime after the end of paclitaxel infusion (h) Paclitaxel concentration (nM) (Percent of

paclitaxel in CSF to concurrent plasma sample)

Rank*

0.5 1 3 6 12 24 48

1b No NA 57.4 (1.3) 26.9 (3.7) 18.7 (30.8) 4

2b No NA 3.5 (2.3) 2

3c No NA 16.4 (27.5) 3

4b No NA UD (4.3) 0

5b No NA UD (3.9) 0

6c No NA 32.8 (15.9) 5

1b Yes 1.40 24.6 (0.4) 10.5 (0.8) 5.9 (3.4)

7b Yes 1.03 UD (7.7) 0

8c Yes 0.78 2.3 (1.9) 1

9b Yes 2.34 UD (1.1) UD (1.4) UD (1.4) 0

NA, not applicable; UD, undetectable, below assay�s lower limit of detection (2.3 nM).a All values obtained from plasma sampling at 0.5 h following the end of paclitaxel infusion, except for 12 h for patient 8 (the only plasma sample

available for that patient).b Metastatic tumor to brain.c Primary brain tumor.

* P-value = 0.22 in Wilcoxon analysis between the TAM/paclitaxel and paclitaxel alone groups.

Rank: Patients were ranked from 0 (undetectable) to 5 (highest mean CSF paclitaxel concentration) for Wilcoxon rank-sum statistical analysis.

Patient No. 1 is counted only during their first exposure (paclitaxel alone) and not during their second exposure (paclitaxel–TAM) in accordance

with intention to treat analysis.

88

median peak paclitaxel CSF:plasma ratio (2.4- and3.7-fold, respectively) suggests a trend towards higherCSF paclitaxel concentrations when paclitaxel was givenalone as compared to when given with TAM (Table 3).

Peak plasma TAM concentrations in the paclitaxel–TAM group ranged from 0.78 to 2.34 lM. The plasmaTAM concentration of 0.78 lM was observed in Patient8, for whom only a 12 h time point was available. Thepeak plasma TAM concentrations in the other patientsin the paclitaxel–TAM group were obtained at 0.5 hfollowing the end of the paclitaxel infusion and 6 h afterthe last dose of their 5-day regimen of TAM, which wasthe earliest sampled time (Table 3).

Pharmacokinetic effect of TAM on plasma paclitaxelconcentrations

No statistically significant difference in plasma paclit-axel pharmacokinetics between the study arms werefound upon formal Wilcoxon statistical analysis(P = 0.17) over the 0.5–48 h following the end of thepaclitaxel infusion (Table 4). Notably, the plasma pac-litaxel concentrations for Patient No. 1 were consistentlyhigher at each time point over the sampled time whenthe patient received paclitaxel–TAM compared to pac-litaxel alone. Plasma paclitaxel AUC from 0 to 24 h forPatient No. 1 was 10.062 and 17.756 lg/ml h when thepatient received paclitaxel alone and paclitaxel–TAM,respectively. However, no clear trend towards increasedpaclitaxel concentration AUC�s were observed in theother three patients within the paclitaxel–TAM arm.For example, in Table 4, the patient with the highestpeak plasma TAM concentration had the lowest plasmapaclitaxel concentration (Patient No. 9). Thus, there wasno clear and consistent demonstrable effect of TAMupon the pharmacokinetics of paclitaxel (Table 4).

Discussion

This study is the largest reported series of brain tumorpatients in which CSF deposition of paclitaxel has beenassessed. Paclitaxel�s poor accumulation in CSF relativeto the concurrent plasma paclitaxel concentration in thisstudy supports the published literature. Glantz et al.reported ratios of CSF paclitaxel concentration toplasma paclitaxel concentration of 0.005–0.079 at 0.25 hfollowing the end of a 3-h infusion in four patients withprimary and metastatic brain tumors treated with pac-litaxel doses ranging from 90 to 200 mg/m2 [11]. Nopatient in that study had quantifiable levels of paclitaxelat 6 h and beyond; five patients in our study had mea-surable CSF paclitaxel concentrations past the 6 h per-iod. Differences between the Glantz et al. study and thepresent study may be partly due to technological dif-ferences in detection, namely a slightly lower limit ofpaclitaxel detection with our method (5 vs. 2.3 nM,respectively) [11]. However, measurable CSF levels ofpaclitaxel have previously been reported up to 72 hfollowing the end of a 24 h infusion of paclitaxel at315 mg/m2 [12].

The CSF paclitaxel concentrations observed up to24 h in our present study are similar to those whichproduced significant antitumor activity against humanprimary brain tumor cell lines in vitro following a 24-hpaclitaxel exposure [13]. In vitro, tumor cells sensitive topaclitaxel generally have IC50s ranging from 10 to50 nM which many patients achieved in their CSF inour study when treated with paclitaxel alone (Table 3).

No significant increase in CSF accumulation of pac-litaxel was noted with pretreatment with TAM. Thehigh-dose TAM regimen utilized here was partiallybased on the recommendations from our Phase I trialwhich determined the maximum tolerated dose (MTD)

Table 4. Pharmacokinetic effect of TAM on plasma paclitaxel concentrations

Patient No. Received

TAM

Peak plasma

TAM (lM)aTime after the end of paclitaxel infusion (h) Plasma paclitaxel

concentration (nM)

Paclitaxel AUC

0–24 h (lg/ml h)

Rank*

0.5 1 3 6 12 24 48

1b No NA 4386.9 721.4 60.9 10.062 8

2b No NA 154.6 9

3c No NA 59.7 7

4b No NA 1667.6 680.4 55.0 38.7 7.028 6

5b No NA 59.7 1

6c No NA 1644.2 849.1 310.3 206.1 131.2 49.2 35.1 4.021 5

1b Yes 1.40 6087.4 1272.9 173.3 17.756

7b Yes 1.03 502.4 318.5 193.2 30.5 3.029 3

8c Yes 0.78 121.8 4

9b Yes 2.34 257.6 222.5 172.2 163.9 2

NA, not applicable.a All values obtained from plasma sampling at 0.5 h following the end of paclitaxel infusion, except for patient 8 who had only a 12 h sample.b Metastatic tumor to brain.c Primary brain tumor.

*P-value = 0.17 in Wilcoxon analysis between the TAM/paclitaxel and paclitaxel alone groups.

Rank: Patients were ranked from 1 (lowest plasma paclitaxel concentration curve) to 9 (highest plasma paclitaxel concentration curve) for

Wilcoxon rank-sum statistical analysis. Patient No. 1 is counted only during their first exposure (paclitaxel alone) and not during their second

exposure (paclitaxel–TAM) in accordance with intention to treat analysis.

89

of TAM when used in combination with vinblastine [7].Of the 53 patients enrolled in that study, nine patientsreceived an MTD regimen of oral TAM consisting ofa Day 1 loading dose of 400 mg/m2 followed by amaintenance dose of 150 mg/m2 twice daily for 12more days (Days 2–13). This produced mean plasmaconcentrations of TAM and N-DESTAM of approxi-mately 4 and 6 lM, respectively. The plasma concen-tration of N-DESTAM following a high-dose TAMregimen is generally equal or greater than the TAMconcentration [14]. We had previously shown thatN-DESTAM had similar activity as compared to TAMfor inhibiting Pgp in cell lines with MDR when testedat equimolar concentrations [4,6]. Peak plasma TAMconcentrations in the paclitaxel–TAM group in thepresent study ranged from 0.78–2.34 lM; N-DESTAMconcentrations were not measured. However, based onextrapolation, the expected combined plasma concen-tration of TAM and N-DESTAM is 1.56–4.68 lM. Inthe present study, patients did not receive a loading doseof 400 mg/m2 of TAM and received only 5 days at themaintenance dose of 160 mg/m2 BID due to the clinicalconditions of our patients. The lower peak plasma TAMconcentrations observed in this study relative to Trumpet al. may reflect, in large part, the dosage regimen dif-ferences; namely, total doses of TAM were lower in ourcurrent study (total of 4000 mg/m2 in the Trump studyvs. 1600 mg/m2 in the present study) [7].

The peak plasma TAM concentrations achieved in thepresent study are notable for being lower than those wehad previously reported to increase the cytotoxicity ofdoxorubicin in a Pgp-expressing hepatocellular carci-noma cell line and of vinblastine in a Pgp-expressingrenal cell cancer line [4,15]. In a clinical trial of patientswith inoperable, advanced hepatocellular cancer, wedemonstrated a response rate of 33% following a 7-dayhigh-dose regimen of TAM akin to that used in thecurrent study with doxorubicin (60 mg/m2 every3 weeks) [16]. Traditionally, doxorubicin at similardoses has a 10–15% response rate as a single agent ininoperable hepatocellular carcinoma. In that study,patients did not receive CYP3A inducers such as anti-convulsants or dexamethasone. However, plasma con-centrations of TAM do not necessarily reflect tissuedeposition of TAM accurately. For example, concen-trations of TAM and N-DESTAM in brain tumors andin surrounding brain tissues were up to 46-fold higherthan that in concurrent plasma for three metastaticbreast cancer patients treated with daily doses of TAMranging from 30 to 60 mg, implying high deposition ofTAM in tissues [17]. Thus, it is possible that tissueaccumulation of TAM and N-DESTAM in the presentstudy may have reached the levels needed to inhibit Pgpat the choroid plexus epithelium despite relatively lowplasma concentrations.

The ability to discern TAM�s potential impact onpaclitaxel penetration across the BCSFB is complicatedby the study population�s treatment history. Themajority of patients in the present study received priorbrain irradiation, which has been shown by severalstudies to disrupt the BBB and the BCSFB and decreasePgp expression by 40% in both animal and human

studies [18]. The significant prior history of brain irra-diation in three of the four patients in the TAM pre-treatment arm may have similarly increased baselineBCSFB permeability by decreasing Pgp expression inirradiated brain tissues, thus hindering TAM�s effect onthe BCSFB.

Plasma and CSF levels of paclitaxel and TAM in ourpresent study may also have been influenced by the ef-fect of CYP3A-inducing medications. Dose rangingclinical trials of paclitaxel in the treatment of gliomasfound that plasma paclitaxel concentrations and plasmapaclitaxel AUC from 0 to 24 h in patients on concurrentCYP3A-inducing anticonvulsants were approximatelyhalf of that detected in patients receiving equivalentpaclitaxel doses but no CYP3A-inducing anticonvul-sants [8,19]. An increase in CYP3A activity may alsoaccount for why we did not detect significant hemato-logic toxicities from 175 mg/m2 of paclitaxel. Catabo-lism of TAM also occurs via the CYP3A pathway andinducers such as anticonvulsants and dexamethasonecan lower the levels of TAM. The Trump et al. studyspecifically excluded patients with known gliomasand brain metastases and thus none of the studypatients were on the aforementioned CYP3A-inducingmedications [7].

Although the concurrent use of CYP3A-inducingmedications may have affected paclitaxel and TAM�spharmacokinetics, TAM did not appear to alter plasmapaclitaxel AUC pharmacokinetics in this study. Also, inour Phase I study by Trump et al., hematologic toxicitywas no greater than what would be expected by usingvinblastine alone [7]. This stands in contrast to otherPgp inhibitors – such as cyclosporine, valspodar (PSC833) and biricodar (VX-710) – which inhibited thecatabolism of natural product anticancer agents andrequired major dose reductions of these agents becauseof toxicity in clinical trials [20–22]. Thus, in concert withthe low toxicity profile observed in our study, TAM mayhold potential advantages over other known Pgpinhibitors because of its apparent lack of effect upon thepharmacokinetics of paclitaxel. In addition, anotherpotential benefit to TAM is its ability to induce stabil-ization and tumor regression in subsets of gliomapatients [23,24]. This is thought to be due to TAM�sability to inhibit protein kinase C activation and trans-location from the cytosol to the membrane which wehad previously demonstrated [15].

The trend towards decreased accumulation of paclit-axel in the CSF of the paclitaxel–TAM group alsosuggests the intriguing possibility that Pgp inhibition atthe choroid plexus by TAM either decreased paclitaxeltransport across its epithelium into the CSF or increasedpaclitaxel elimination out of the CSF. The group led byPiwnica-Worms has shown localization of Pgp to thesubapical region in primary cultured rat choroid plexusepithelial cells [2]. This differs from where Pgp is nor-mally found in the BBB, organs of detoxification andtumor cells, where it is in the apical region of the cell.They demonstrated Pgp in the cells of the normal ratchoroid plexus can function in an opposite manner,whereby it would pump substrates such as paclitaxel intothe CSF or prevent Pgp substrates from leaving the CSF.

90

The end result of uninhibited Pgp function in the cho-roid plexus would be to increase Pgp substrates (i.e.paclitaxel) in the CSF. In the presence of the Pgpinhibitor GF120918 in the same study, transepithelialtransport of paclitaxel paradoxically decreased in thebasal to apical direction and increased in the apical tobasal direction. These data led the authors to proposethat the subapical Pgp in the BCSFB choroid plexusnormally blocks drug substrate elimination out of theBCSFB and subsequently blocks drug efflux out ofthe CSF [2]. Therefore, in our study, TAM inhibitionof the BCSFB�s Pgp could potentially facilitate efflux ofpaclitaxel out of the CSF. This could explain our findingwhere TAM-treated patients had lower CSF paclitaxellevels. Although this was opposite to the anticipatedeffect we expected, high-dose TAM may potentiallyrepresent an interesting clinical intervention for treatingpatients who were inadvertently given natural productchemotherapy drugs, like paclitaxel or vincristine,directly into their CSF. Thus, TAM�s inhibition ofBCSFB–Pgp could potentially decrease levels of paclit-axel in the CSF or decrease other natural chemotherapydrugs which bind to Pgp.

The data from Patient No. 1 is consistent with thishypothesis in that despite higher plasma paclitaxel levelsoccurring with the course of paclitaxel–TAM, the cor-responding CSF concentrations of paclitaxel were lessthan 50% of that achieved when the patient receivedpaclitaxel alone (Table 3). Although the sample sizeswere small, the mean peak CSF concentration of pac-litaxel and the median peak paclitaxel CSF:plasma ratiowere 2.4 and 3.7 times higher, respectively, in the groupthat received paclitaxel alone compared to the paclit-axel–TAM group. Although these were not statisticallysignificant by Wilcoxon rank-sum tests (P = 0.22) be-cause of small sample size, they are suggestive of a trend.If the Pgp localization finding posited by the Piwnica-Worms group in rat choroid plexus proves true in hu-mans, then Pgp inhibition may paradoxically decreasepaclitaxel accumulation in the CSF, while possiblyincreasing paclitaxel accumulation in the brain tumortissue via inhibition of the BBB–Pgp efflux pump. Thus,our hypothesis is that Pgp in the BBB acts to preventinflux of toxins into the brain, such as natural productchemotherapy drugs including paclitaxel. However, Pgpin the BCSFB keeps these toxins in the CSF and pre-vents their efflux out into the blood or brain tissue.Essentially, the BCSFB Pgp keeps toxins in the CSFwhere they may be less toxic than if they were trans-ported back into the brain or blood. Thus, conceptuallythe Pgp of the BBB and BCSFB keeps natural toxinsfrom entering the brain and, if they did enter, keepsthem in the CSF and blocks their efflux back into thebrain. The sum effect of these two Pgp barriers would beto lower brain exposure to natural toxins, includingchemotherapy drugs such as paclitaxel. Our findingssupport this novel concept and add to the understandingof the role of the BCSFB–Pgp.

Further study of the role of Pgp and Pgp inhibitors inmodulating drug delivery to the CSF will requiredetermining Pgp�s in situ localization in the humanBCSFB and utilizing higher dosage regimens of TAM

and paclitaxel to compensate for the increased catabo-lism induced by CYP3A-inducing medications, such asphenytoin, phenobarbital, carbamazepine and dexa-methasone. The therapeutic role of treatment with a Pgpinhibitor, like TAM, in altering paclitaxel and othernatural product chemotherapy drugs, ability to cross theBCSFB remains to be determined and warrants furtherstudy.

Acknowledgements

This work was supported by the Herbert Irving Com-prehensive Cancer Center Award, Herbert IrvingScholar Award, and Vinzini Brain Tumor Award toRLF, Doris Duke Clinical Research Fellowship to JCand NIH/NCI RO1 CA089395 grant to JNB. BarrPharmaceuticals generously donated the tamoxifen20 mg tablets used in this study. We thank SusanBronson for expert technical support for the manuscript.

References

1. Cordon-Cardo C, O�Brien JP, Casals D, Rittman-Grauer L,

Biedler JL, Melamed MR, Bertino JR: Multidrug-resistance gene

(P-glycoprotein) is expressed by endothelial cells at blood–brain

barrier sites. Proc Natl Acad Sci USA 86: 695, 1989

2. Rao VV, Dahlheimer JL, Bardgett ME, Snyder AZ, Finch RA,

Sartorelli AC, Piwnica-Worms D: Choroid plexus epithelial

expression of MDR1 P glycoprotein and multidrug resistance-

associated protein contribute to the blood–cerebrospinal-fluid

drug-permeability barrier. Proc Natl Acad Sci USA 96: 3900, 1999

3. Berman E, Adams M, Duigou-Osterndorf R, Godfrey L, Clarkson

B, Andreeff M: Effect of tamoxifen on cell lines displaying the

multidrug-resistant phenotype. Blood 77: 818, 1991

4. Fine RL, Sachs CW, Albers ME, Safa A, Rao US, Scarborough

G, Burchette J, Jordan C, Trump DL: Tamoxifen inhibition of

MDR: laboratory to clinical studies. In: Miyazaki (ed) The

Mechanism and New Approach on Drug Resistance of Cancer

Cells. Elsevier, Amsterdam, pp 323–332, 1993

5. Safa AR, Roberts S, Agresti M, Fine RL: Tamoxifen aziridine,

a novel affinity probe for P-glycoprotein in multidrug resistant

cells. Biochem Biophys Res Commun 202: 606, 1994

6. Rao US, Fine RL, Scarborough GA: Antiestrogens and steroid

hormones: substrates of the human P-glycoprotein. Biochem

Pharmacol 48: 287, 1994

7. Trump DL, Smith DC, Ellis PG, Rogers MP, Schold SC, Winer

EP, Panella TJ, Jordan VC, Fine RL: High-dose oral tamoxifen,

a potential multidrug-resistance-reversal agent: phase I trial in

combination with vinblastine. J Natl Cancer Inst 84: 1811, 1992

8. Fetell MR, Grossman SA, Fisher JD, Erlanger B, Rowinsky E,

Stockel J: Preirradiation paclitaxel in glioblastoma multiforme:

efficacy, pharmacology, and drug interactions. New Approaches to

Brain Tumor Therapy Central Nervous System Consortium.

J Clin Oncol 15: 3121, 1997

9. Vergniol JC, Bruno R, Montay G, Frydman A: Determination of

taxotere in human plasma by a semi-automated high-performance

liquid chromatographic method. J Chromatogr 582: 273, 1992

10. Webster LK, Crinis NA, Stokes KH, Bishop JF: High-perfor-

mance liquid chromatographic method for the determination of

toremifene and itsmajor humanmetabolites. J Chromatogr 565: 482,

1991

11. Glantz MJ, Choy H, Kearns CM, Mills PC, Wahlberg LU,

Zuhowski EG, Calabresi P, Egorin MJ: Paclitaxel disposition in

plasma and central nervous systems of humans and rats with brain

tumors. J Natl Cancer Inst 87: 1077, 1995

91

12. O�Boyle KP, Wang Y, Schwartz EL, Regl DL, Einzig A, Dutcher

JP, Wiernik PH, Horwitz SB: Development of two radio-

immunoassays to detect paclitaxel in sera and in cerebrospinal,

ascitic, and pleural fluids. Cancer 79: 1022, 1997

13. Tseng SH, Bobola MS, Berger MS, Silber JR: Characterization of

paclitaxel (Taxol) sensitivity in human glioma- and medulloblas-

toma-derived cell lines. Neuro-Oncology 1: 101, 1999

14. Ducharme J, Fried K, Shenouda G, Leyland-Jones B, Wainer IW:

Tamoxifen metabolic patterns within a glioma patient population

treated with high-dose tamoxifen. Br J Clin Pharmacol 43: 189, 1997

15. Cheng AL, Chuang SE, Fine RL, Yeh KH, Liao CM, Lay JD,

Chen DS: Inhibition of the membrane translocation and activation

of protein kinase C, and potentiation of doxorubicin-induced

apoptosis of hepatocellular carcinoma cells by tamoxifen. Biochem

Pharmacol 55: 523, 1998

16. Cheng AL, Yeh KH, Fine RL, Chuang SE, Yang CH, Wang LH,

Chen DS: Biochemical modulation of doxorubicin by high-dose

tamoxifen in the treatment of advanced hepatocellular carcinoma.

Hepatogastroenterology 45: 1955, 1998

17. Lien EA, Wester K, Lonning PE, Solheim E, Ueland PM:

Distribution of tamoxifen and metabolites into brain tissue and

brain metastases in breast cancer patients. Br J Cancer 63: 641, 1991

18. Vulpen Mvan, Kal HB, Taphoorn MJ, El-Sharouni SY: Changes

in blood–brain barrier permeability induced by radiotherapy: impli-

cations for timing of chemotherapy? (Review).OncolRep 9: 683, 2002

19. Chang SM, Kuhn JG, Rizzo J, Robins HI, Schold SC Jr., Spence

AM, Berger MS, Mehta M, Pollack IF, Rankin C, Prados MD:

Phase I study of paclitaxel in patients with recurrent malignant

glioma: a North American Brain Tumor Consortium report. J Clin

Oncol 16: 2188, 1998

20. Lum BL, Kaubisch S, Yahanda AM, Adler KM, Jew L, Ehsan

MN, Brophy NA, Halsey J, Gosland MP, Sikic BI: Alteration of

etoposide pharmacokinetics and pharmacodynamics by cyclo-

sporine in a phase I trial to modulate multidrug resistance. J Clin

Oncol 10: 1635, 1992

21. Advani R, Saba HI, Tallman MS, Rowe JM, Wiernik PH, Ramek

J, Dugan K, Lum B, Villena J, Davis E, Paietta E, Litchman M,

Sikic BI, Greenberg PL: Treatment of refractory and relapsed acute

myelogenous leukemia with combination chemotherapy plus the

multidrug resistance modulator PSC 833 (Valspodar). Blood 93:

787, 1999

22. Rowinsky EK, Smith L, Wang YM, Chaturvedi P, Villalona M,

Campbell E, Aylesworth C, Eckhardt SG, Hammond L, Kraynak

M, Drengler R., Stephenson J Jr, Harding MW, VonHoff DD:

Phase I and pharmacokinetic study of paclitaxel in combination

with biricodar, a novel agent that reverses multidrug resistance

conferred by overexpression of both MDR1 and MRP. J Clin

Oncol 16: 2964, 1998

23. Couldwell WT, Hinton DR, Surnock AA, DeGiorgio CM, Weiner

LP, Apuzzo ML, Masri L, Law RE, Weiss MH: Treatment of

recurrent malignant gliomas with chronic oral high-dose tamoxi-

fen. Clin Cancer Res 2: 619, 1996

24. Chamberlain MC, Kormanik PA: Salvage chemotherapy with

tamoxifen for recurrent anaplastic astrocytomas.ArchNeurol 56: 703,

1995

Address for offprints: Robert L. Fine, MD, College of Physicians and

Surgeons of Columbia University, 650 West 168th Street, Room 20-05,

New York, NY 10032; Tel.: +1-212-305-1168; Fax: +1-212-305-7348;

E-mail: rlf20@columbia.edu

92