Probenecid Interacts with the Pharmacokinetics of Morphine-6-glucuronide in Humans

Five men and five women with an average age, weight, and height of 27.2 yr, 68.9 kg, and 175.7 cm, respectively, were recruited for this study. The study was open with respect to M6G administration and double blind with respect to the administration of transporter inhibitors. M6G (Lipomed, Arlesheim, Switzerland) was administered as an infusion over 30 min. On the basis of simulations using pharmacokinetic23and pharmacokinetic–pharmacodynamic24models, a dose of 0.5 mg/kg was chosen to produce clearly noticeable but not maximum pupil constriction, which was considered to be best suited to identify a modification in the central nervous effects of M6G by coadministration of transporter inhibitors. One hour before M6G administration, subjects ingested in a double-blind, randomized, cross-over fashion 800 mg quinidine, 500 mg probenecid, or placebo. A washout interval of at least 7 days was observed. In a previous study in a similar population, we observed quinidine concentrations of approximately 2 μg/ml after 800 mg oral quinidine,25which produced clinically relevant effects on the QTc time of the electrocardiogram26and which had been found to reverse P-glycoprotein–mediated tumor resistance in humans to a relevant degree.27Probenecid dosing was based on the report that 500 mg probenecid every 6 h resulted in an increase of the plasma concentrations and reduction of the urinary excretion of zidovudine.28,29 

The study was conducted according to the Declaration of Helsinki on biomedical research involving human subjects (Somerset West amendment). The University of Frankfurt Medical Faculty Ethics Review Board approved the study protocol, and written informed consent was obtained from all subjects before the study.

Blood samples (4 ml) were drawn before the administration of quinidine or placebo, at 15 and 30 min after M6G administration, and then every half hour until the end of the 7-h observation period. Plasma concentrations of M6G, M3G, and morphine were assayed by liquid chromatography tandem mass spectrometry.30The lower limit of quantification was 1 ng/ml for morphine and 0.5 ng/ml for M6G and M3G. The coefficient of variation over the calibration range of 0.5 (or 1) to 1,250 ng/ml was less than 5%.

Pupil diameter was assessed using a pupillograph (CIP; Amtech GmbH, Weinheim, Germany). After four initial baseline measurements, pupil size was measured with the start of the M6G infusion every 30 min for approximately 7 h as described previously.24,26,30 

A two-stage data analysis was performed with outcome measures being the areas under the M6G plasma concentration and pupil diameter change–versus –time curves (AUCs). In addition, the clearance of M6G (CL^M6G) was obtained from Dose/AUC, using the AUC extrapolated to infinity. The average contribution of the extrapolated part of the AUC to the total AUC was 5.8 ± 2.5%, with a maximum of 11%. Maximum M6G plasma concentrations, C^max, were read from the data, and the terminal half-life was calculated as ln²divided by the slope of a regression line through the terminal linear segment of the log-transformed M6G concentration–versus –time curve. The influence of the comedications on C^max, AUC, and CL^M6Gwas evaluated by repeated-measures analysis of variance (rm-ANOVA; between-subjects factor medication ), followed by Student-Newman-Keuls tests for multiple comparisons (Sigma Stat 3.00; SPSS, Inc., Chicago, IL). In addition, M6G morphine plasma concentrations and pupil diameter changes were evaluated by 2-way rm-ANOVA (factors medication  and session , i.e ., observation number), followed by the Student-Newman-Keuls test for multiple comparisons. The α level was set at 0.05. All results are reported as mean ± SD.

Plasma concentration–versus –time data were also analyzed using a parametric modeling approach with NONMEM (GloboMax LLC, Hanover, MD). The modeling process repeatedly used goodness-of-fit procedures, which were (1) the NONMEM objective function being −2 times the log likelihood and the χ²approximation with the number of degrees of freedom equal to the difference in the number of terms between two models (α level, 0.05); (2) the median absolute weighted residuals, calculated as (measured − predicted)/predicted, and the mean of the individual mean absolute weighted residuals; and (3) visual inspection of the fits versus  observed data.

The pharmacokinetics of M6G were described by a two-compartment model:

where A1 and A2 denote the amount of drug in compartments 1 and 2, respectively; C1 and C2 the respective concentrations obtained by dividing the amount of drug by the volume of distribution of the respective compartment, V1 and V2; I(t) is the input function for M6G characterized by rate and duration of the infusion; CL is the total body clearance, and Q the intercompartmental clearance. Alternative one- and three-compartment models were rejected on the basis of goodness-of-fit judgments.

The influence of quinidine or probenecid administration on the model parameters was assessed as follows: First, it was checked whether a particular parameter varied interindividually. Second, it was checked whether there was an intraindividual interoccasion variability on this parameter. If this was found to be the case, it was determined whether this interoccasion variability was partly explained by the comedication. This was done by multiplying the model parameters found to exhibit intraindividual interoccasion variability with 1 in case of placebo coadministration, with a factor θ¹in case of quinidine coadministration and with a factor θ²in case of probenecid coadministration. For example, the clearance CL in equation 1 was replaced by CL · 1 for the placebo condition, CL ·θ¹for the quinidine condition, and factors CL ·θ²for the probenecid condition. If introduction of θ¹or θ²significantly improved the fit, judged by the likelihood ratio test, then an influence of the respective comedication on M6G distribution was demonstrated. This assumed that the influence of a transporter inhibitor on a model parameter was constant throughout the observation period, thus neglecting that the inhibitor’s concentrations varied as a consequence of the inhibitor’s own concentration-versus -time profile. A log-normally distributed interindividual variance of the model parameters was used:, where Pⁱis the value of the parameter of the individual, θ^i,TVis the typical value of this parameter in the population, and η is a variable accounting for the interindividual variability, with mean zero and variance ω². The intraindividual interoccasion variability was modeled analogously, as described previously.31An assumption of normal rather than log-normal distribution of parameter values resulted in worse fits and was therefore rejected. The residual error ε was modeled using a proportional error model. Calculations were performed using first-order conditional estimation and η–ε interaction. Covariates checked for relation with pharmacokinetic structural parameters were age, weight, height, body surface area,32and sex.

Probenecid coadministration increased the M6G plasma concentrations (fig. 1A), with significant differences from placebo starting at 1 h after M6G dosing (two-way rm-ANOVA: P < 0.05 for factor medication , P < 0.001 for factor session ). The AUC was increased 1.26 ± 0.2 times with probenecid coadministration (rm-ANOVA: P = 0.001, Student-Newman-Keuls: P < 0.05 for probenecid vs . placebo) as compared with placebo coadministration, whereas it remained unchanged by quinidine (factor 1.05 ± 0.1). The M6G plasma clearance was significantly influenced by the comedication (rm-ANOVA effect medication : P = 0.001). It decreased from 8.3 ± 1 l/h under placebo by a factor of 0.82 ± 0.2 to 6.7 ± 1.3 l/h under probenecid coadministration, which was statistically significant (Student-Newman-Keuls: P < 0.05 vs . placebo; 95% confidence interval [CI] for differences, −2.5 to −0.7 l/h). With quinidine, the clearance of M6G was 8 ± 1.2 l/h, which did not differ statistically significantly from placebo coadministration (95% CI for differences, −1 to 0.4 l/h). The M6G plasma half-life increased from 1.6 ± 0.2 h under placebo coadministration to 1.8 ± 0.3 h under probenecid coadministration (rm-ANOVA: P < 0.01; 95% CI for differences, probenecid to placebo, 0.1–0.4 h; Student-Newman-Keuls: P < 0.05), whereas it remained unaltered by quinidine coadministration (1.6 ± 0.2 h; P = 0.4 vs . placebo; 95% CI for differences, −0.1 to 0.2 h). Morphine and M3G were found only in traces in plasma as it had been previously observed.30 

The results of the pharmacokinetic modeling analysis are given in table 1. The analysis obtained a significant effect of probenecid administration on the total body clearance of M6G. The NONMEM objective function decreased highly significantly (P < 0.001) by 31 when assigning θ²to the total clearance of M6G, and the intraindividual interoccasion variance decreased from 16.9% to 7.6%, given as percent coefficient of variation. Assigning covariates to additional pharmacokinetic parameters did not further improve the fit. The factor θ²was estimated to be 0.8 (95% CI of the estimate, 0.73–0.82), i.e ., the clearance of M6G decreased by approximately 20% when probenecid had been coadministered. Therefore, in equation 1, the parameter CL was replaced with CL ·θ²in case of probenecid coadministration with M6G. In contrast, θ¹accounting for quinidine coadministration did not significantly improve the fit, and the intraindividual interoccasion variance decreased only from 16.9% to 15.7% coefficient of variation. Finally, the body surface area was a covariate to the central volume of distribution V1, whereas the other covariates tested, including sex, had no influence on the pharmacokinetics of M6G. Figure 1Ashows the fits along with the observed data.

Coadministration of probenecid but not of quinidine resulted in a more pronounced miosis (fig. 1B), which was statistically significant at three data points covering the time of maximum miotic effects around 4.5 h (two-way rm-ANOVA: P < 0.05 for factor medication , P < 0.001 for factor session ; Student-Newman-Keuls for probenecid vs . placebo: P < 0.05). With probenecid coadministration, the AUCs increased by a factor of 1.25 ± 0.3 as compared with placebo coadministration (rm-ANOVA factor medication: P < 0.05; post hoc  Student-Newman-Keuls: P < 0.05; 95% CI for differences from placebo coadministration, −4.2 to −0.3 mm · h), whereas AUCs remained unchanged with quinidine coadministration (factor of 1.1 ± 0.3, 95% CI for differences to placebo, −1.5 to 1.7 mm · h). Therefore, the increase in the AUCs of the miotic effects paralleled the observed effect in the AUCs of the plasma concentration data. This suggested that probenecid had an exclusive effect on the plasma concentrations, with the result that miotic effects just followed the increase in plasma M6G.

The results indicate that transmembrane transporter function may affect M6G disposition. The effect of 500 mg probenecid coadministration consisted of a 20% decrease of the M6G clearance. Because M6G is almost exclusively eliminated via  the kidney (92% of the total amount of M6G33), involving glomerular filtration, tubular secretion, and reabsorption,34a likely effect of probenecid was decreasing the renal tubular secretion of M6G, similarly as demonstrated for the action of probenecid on renal excretion of zidovudine28,29or acamprosate.35 

The significant effects on the area under the miotic effects–versus –time curves indicate that the changes in M6G disposition bear the potential of enhancing the clinical effects of M6G, although currently, the effects were seen only in three data points, i.e ., during a short period when maximum miotic effects were observed. Therefore, from the current observations, an important clinically relevant drug interaction cannot be established. The finding is currently more important to attract attention to probenecid sensitive transporters as a possible modulator of the clinical effects of M6G to take a closer look at possible pharmacokinetic interactions of M6G with other drugs for which probenecid was shown to alter their pharmacokinetics, such as zidovudine,28,29adefovir,36ibuprofen,37or cephalosporin antibiotics.38The clinical setting best qualifying for such investigations is morphine therapy in patients with renal failure, in whom M6G accumulates and produces central nervous side effects.2,39–41 

Probenecid inhibits several transporters of the MRP, OATP, and organic anion transporter families, which was shown in vitro  by decreasing the transport at the blood–brain barrier for morphine42or for M3G.43From interaction of transmembrane transport of M6G with digoxin or the P-glycoprotein inhibitor PSC833 (valspodar44) in P-glycoprotein knockout mice,13from the fact that digoxin has been found to be a substrate of Oatp245,46and from the observation that PSC833 also interacted with MRP2,47it is concluded that Oatp2 (as already hypothesized13) and MRP2 are possible candidate transporters involved in the current effects of probenecid on M6G disposition. OATP-C, MRP2, and P-glycoprotein, which hold an important role for drug disposition, were found to be colocalized in organs such as the kidney, liver, or duodenum,48–50but MRP2 not at the blood–brain barrier.51Because they form outward transporters,48–50their inhibition would result in retention of M6G in the body as currently observed after probenecid administration. To obtain a hypothesis regarding which transporter was likely to be involved in decreasing the clearance of M6G by probenecid, we genotyped the participants (with approval of the ethics committee and after having obtained written informed consent) for single nucleotide polymorphisms (SNPs) in the ABCB1  gene that codes for P-glycoprotein (SNPs 2677G>T(A) and 3435C>T), in the ABCC2  gene that codes for MRP2 (SNPs -24C>T, 1249G>A, and 3972C>T), and in the SLCO1B1  gene that codes for OATP-C (SNPs 388A>G and 521T>C). The three carriers of the SLCO1B1  SNP 521T>C, which causes an amino acid exchange Val174Ala in the OATP-C transporter, had a smaller M6G clearance with probenecid coadministration (CL^M6G= 5.3 ± 0.7 l/h) than the seven noncarriers of this particular mutation CL^M6G= 7.3 ± 1.1 l/h). This difference was statistically significant (t  test: P = 0.025; 95% CI for differences, 0.3–3.5 l/h). Moreover, the clearances of the carriers and noncarriers of that mutation did not overlap. The three carriers had M6G clearances under probenecid cotreatment of 5.3, 4.6, and 6.1 l/h, while the lowest clearance in a noncarrier was 6.3 l/h. None of the other SNPs or resulting haplotypes (the latter in the case that a linkage disequilibrium between single SNPs could be found by means of the EH computer program: J. Ott, Ph.D., Rockefeller University, New York, New York52) influenced any of the parameters of M6G disposition or effects, with all P  values being well above 0.1. This result could mean that OATP-C is involved in the transmembrane transport of M6G in humans. That an effect of the OATP-C mutation on M6G clearance was only evident with probenecid coadministration does not contradict the hypothesis. It could be seen as a challenge test required to observe the pharmacogenetic effect, analogously to a glucose tolerance test needed for diagnosis of diabetes in individuals without permanently increased blood glucose. Here, the transporter function may have been at the lower range of the distribution, but the differences between genotypes were not great enough to form separate groups. When challenged with probenecid, the compensation capacity of the mutated transporters may have failed earlier so that the mutation carriers were then separated from the nonmutated subjects. However, an involvement of OATP-C in M6G disposition is merely a hypothesis fitting to previous observations from experiments in laboratory animals and supported by the current preliminary pharmacogenetic analysis and by the fact that functional consequences were described for the 521T>C SNP.53The hypothesis must be tested in a new study using probenecid coadministration with M6G in subjects recruited according to their SLCO1B1  genotype.

The lack of a quinidine effect given to block P-glycoprotein20–22,27indicated that P-glycoprotein played a minor role in M6G disposition, which contrasts with the expectations raised on the basis of experiments in vitro .8,54However, it seems to be in accord with the negative results obtained in P-glycoprotein knockout mice.11–13It could also mean that at the blood–brain barrier P-glycoprotein played no important role in the current context. This is similar to the failure to demonstrate P-glycoprotein effects on opioid brain distribution localized at the blood–brain barrier.25,55,56In addition, it could mean that the inhibiting effect of quinidine on P-glycoprotein at the kidney or at the blood–brain barrier was not big enough to alter the disposition of M6G. Quinidine coadministration increased plasma concentrations only of orally and not of systemically administered morphine,26,56which may be related to the much higher local quinidine concentrations to which intestinal P-glycoprotein was exposed in contrast to the comparably lower systemic quinidine concentrations to which P-glycoprotein at other localizations was exposed. It remains to be seen whether more potent P-glycoprotein modulators than quinidine (for review see Tan et al .57) interact with the systemic disposition of M6G.

In conclusion, we found that probenecid interacts with the pharmacokinetics of M6G in humans, possibly involving OATP-C transporters.