A population pharmacokinetic study of alminoprofen penetration into synovial fluid

BIOPHARMACEUTICS & DRUG DISPOSITION, VOL. 16, 627-634 (1995)

A POPULATION PHARMACOKINETIC STUDY OF ALMINOPROFEN PENETRATION INTO SYNOVIAL FLUID M. TOD*It, C. POBELT, V. LE GROSI, K. LOUCHAHI', 0. PETITJEAN*, N. BRIOEJ' AND J. L. GARCIA-MACES *Dipartement de Pharmacotoxicologie, Hbpital Avicenne, 125, route de Stahgrad, 93009 Bobigny Ckdex, France 'Service de Phamtacie. Hbpital de Saintes, 17108 Saintes. France $Versus, 27, rue Fontaine, 75009 Paris, France DLaboratoires du Dr. E. Bouchara, Levallois-Perret, France

ABSTRACT The pharmacokinetics of alrninoprofen in plasma and synovial fluid (SF) at steady state (300mg t.i.d.) was studied in 45 patients with knee effusion. Plasma and SF samples, one each per patient, were obtained. Six groups were made according to the time of sampling after ingestion of the 13th dose: 1h (n= 7), 2 h (n= 7), 4 h (n= 7), 6 h (n= lo), 8 h (n = 6), 12 h (n= 8). A three-compartment model was used to describe alminoprofen kinetics in plasma and SF, with two parameterizations, a 'classical' and a 'physiological' one. The non-linear mixed effect model approach was used to estimate the mean and variance of the pharmacokinetic parameters. The mean f SE of the estimates (coefficient of variation of interindividual variability as a percentage) were volume of distribution, l l - O i 1.71 1 (12%); elimination rate constant, 0.236f0-025h-' (18%); absorption rate constant 2.80f0.31 h-' (464%), clearance of influx into SF, 0.29 fO.l4mLmin-'; clearance of efflux into plasma, 0.56 &0.25mLmin-l. These two clearances were not significantly different, which indicates that passive diffusion occurs in both directions. The mean f SD alminoprofen concentration versus time curve in plasma and SF at steady state was simulated and showed that the meanhSD maximal concentration in SF was 8.1 f 6.3 mg L-' and was obtained 4 h after dose administration. KEY WORDS: alminoprofen; population pharmacokinetics; NONMEM; synovial penetration

INTRODUCTION Alminoprofen is a non-steroidal anti-inflammatory drug (NSAID) that belongs to the arylpropionic acid group. As such, it is used for the treatment of inflammatory and degenerative arthritis. It is usually given per os at 300 mg t.i.d. and the plasma steady state concentration is rapidly reached, owing to its 3 h half-life. However, the clinical response to NSAIDs should correspond better to the drug concentration within inflamed tissues than that "Correspondence to: M. Tod, Phamacie, HBpital Avicenne, 125, route de Stalingrad, 93009 Bobigny a d e x , France.

CCC 0142-2782/95/080627-08 01995 by John Wiley 8c Sons, Ltd.

Received 9 August 1994 Accepted 28 February 1995

628

M. TOD ET AL.

in the vascular system.' In this respect, synovial fluid is an appropriate and accessible reflection of articular interstitial fluid.z The distribution of several NSAIDs in synovial fluid has been ~ t u d i e d ; it ~ ?was ~ found that the peak concentration in synovial fluid was generally lower, occurred later, and decreased more slowly than that in plasma. Several factors (type of pathology, degree of inflammation, protein concentration, pH, interleukin-6, . . .) have been shown to influence the penetration of NSAIDs into synovial but evaluation of the interindividual variability in articular kinetics of NSAIDs has been hindered by limited sampling (typically one sample per patient), irregular sampling times, and multiple-dose protocols. Aarons8 was the first to apply the non-linear mixed effect model (NONMEM) to such studies and showed that means and variances of the pharmacokinetic parameters describing the penetration of an NSAID (flurbiprofen) into synovial fluid could be adequately assessed. Another goal of these pharmacokinetic studies is to better elucidate the mechanisms involved in the transport of NSAIDs across the synovial membrane, and this insight can be achieved by incorporating physiological data into pharmacokinetic models, as shown by Simkin et aL4 Therefore, in this study, the pharmacokinetics of alminoprofen in plasma and synovial fluid were modelled according to a semiphysiological model and the data were subjected to the NONMEM approach to derive the central tendency and the interindividual variability of the corresponding parameters.

METHODS Subjects

Forty-five subjects (34 men, 11 women) with knee effusion were included. They were 41 f 17 years old (mean f SD), range, 18-65, weighed 73 f 12kg (range, 49-108), were otherwise in good health, and their clinical biochemistry tests showed no abnormalities. Nineteen patients had a meniscal lesion, seven had arthrosis, six had a ligamental lesion, five had pain in the knee, two had meniscal and ligamental lesions, one had a transposition of the anterior tibial tuberosity, one had arthritis, one had hydarthrosis, one had undergone a tibial osteotomy, one had evaluation of a traumatic injury, and one had articular lavage. The subjects received no medication other than alminoprofen during the week before and during the study. Study design

The patients received 300 mg of alminoprofen (Minalfene@)t.i.d. at 7 h, 15 h, and 23 h for 4 d and they were sampled after ingestion of the 13th dose with 150 mL of tap water. For each patient, a blood sample (5 mL) and a synovial sample from the knee joint were taken at the same time.

ALMINOPROFEN SYNOVIAL PENETRATION

629

The subjects were divided into six groups according to the time of sampling after taking the 13th dose: 1h ( n = 7 ) , 2 h (n=7), 4 h (n=7), 6 h (n= lo), 8 h (n= 6), 12h (n= 8). All subjects gave written informed consent and the study was approved by the Ethics Committee of La Roseraie Clinic, Aubervilliers. Blood samples, drawn into heparin, were centrifuged and the plasma and synovial fluid were stored at -80°C until tested. No signs of haemarthrosis were detected in the synovial fluid. Drug assay

Alminoprofen was assayed by liquid chromatography using a modified version of the method developed by Paillet et aL9 Plasma samples (1 mL) were added to 0.1 mL of a saturated solution of sodium chloride and 0.1 mL of 0.017 A4 acetic acid. Ketoprofen was used as the internal standard. Alminoprofen and ketoprofen were extracted into 7mL of diethyl ether. After shaking and centrifugation, the organic layer was evaporated under an air stream and mild heating. The dry residue was reconstituted in 0-5mL of mobile phase (methanol/water/acetic acid/dimethyl sulphoxide, 50/46/0.8/3, v/v). The synovial fluid was treated in the same way. A 20 pL aliquot was injected into the chromatographic system. A Nucleosil C18 5 pm column, 250 x 4.6 mm ID, was used. The flow rate was 1 mLmin-l and the drugs were detected by W absorption at 251 nm (0.016 AUFS). The retention times of alminoprofen and ketoprofen were 8-1 and 11.7min, respectively. The calibration was linear in the range 2-50mgL-'. The interassay coefficient of variation was 13.9% at 2mgL-', 10.3% at SmgL-l, 3.7% at 25mgL-', and4.8% at 50mgL-I. The limit of quantification was 2 mg L-1. Data analysis

The analysis was performed with the NONMEM computer program version IV developed by Beal and Sheiner.'O The individual pharmacokinetic parameters were considered to arise from a log-normal distribution characterized by a population mean and interindividual variance. The plasma concentration-time profile in the jth individual was described by the following equation:

where Pi represents the pharmacokinetic parameters of the jth individual, ti is the time of the measurement, and E~ is the residual error. E was assumed to be a random Gaussian variable with a mean of zero and a variance 082, i.e., the residual variability was described by a log-normal distribution. The pharmacokinetic model f was as depicted in Figure 1. Two parameterizations were used. In the first one, the dose was given in compartment 3, which

630

M. TOD ET AL.

I CENTRAL COMPARTMENT

Vd or Vp = 3.15 I

k

-

CP

k PC

or ki

SYNOVIAL FLUID

or kf

Vs = 0.105

represents the gastro-intestinal tract. The dose was absorbed into the central compartment (compartment 1) with volume V , by a first-order process characterized by a rate constant of absorption k, . The drug entered into the synovial fluid compartment (unit volume) with corresponding rate constants k, and kpc.Finally, the drug was excreted with an elimination rate constant k, . The subroutines ADVAN4 and TRANS1 were used for this purpose. This parameterization enabled the estimation of classical parameters of interest, namely V,k, , and k,, . In the second parameterization, physiological data were incorporated into the model as described by Simkin et aL4 The volume of the central compartment was set at a mean value of 3-15L, i.e., the total plasma volume (V,). The observed quantity of alminoprofen in plasma (i.e., the concentration divided by 3.1 5) was multiplied by the fraction 1.92mL min- l / 3250 mL min- (synovial plasma flow divided by the cardiac output of the plasma) to determine the drug mass passing through the vasculature of each knee. The volume of the synovial fluid compartment (V,) for albumin (i.e. the binding protein of alminoprofen) was set at a mean value of 0-105L (corresponding to that in patients with knee effusion) so that the observed concentrations of alminoprofen in synovial fluids were transformed into intrasynovial mass. In this way the true values of the rate constants describing drug influx into the knee (ki) and the efflux from the knee (kf)were obtained, The mean clearance from plasma into joints was calculated as the product of kiand the minute volume of 1-92mL in synovial microvessels, while the mean clearance in the opposite direction was calculated as the product of kfand the intracapsular volume, 105mL.4

631

ALMINOPROFEN SYNOVIAL PENETRATION

In both parameterizations, the option of estimating the covariances between parameters was omitted. Linear relationships between pharmacokinetic parameters and covariables (height, weight, albumin and creatinine in serum, interleukin-6 in synovial fluid) were explored and tested by comparing the log likelihoods of the fits against a chi-square statistic, the standard error of the estimates, and the scatter plots. An estimation of the mean fSD profile of alminoprofen kinetics in plasma and synovial fluid was obtained by using the simulation ability of the ADAPT I1 software package." The model was implemented according to the first parameterization and the simulation was produced for over 100 subjects with the population parameters found by NONMEM.

RESULTS AND DISCUSSION Experimentally determined concentrations in plasma and synovial fluid are reported in Table 1. The results of the NONMEM analysis of both parameterizations are summarized in Table 2. All mean parameters were moderately well estimated, i.e., their SEs were low. It should be noted that only an apparent volume of distribution ( V / F ) was obtained, because the bioavailability was unknown and therefore could not be taken into account. Since the drug mass penetrating into the tissular compartment was negligible, plasma kinetics could be considered to be monocompartmental, and the slope of the elimination phase represented by k, . The plasma alminoprofen half-life corresponding to the mean value of k, was 0.69310.236 = 2.94 h, which is in accord with previous estimates. Other than for k,, the interindividual variability of the pharmacokinetic parameters was within the usual range. For k, , the interindividual coefficient of variation was poorly estimated, but values obtained for this parameter usually vary widely. The poor precision of the k, variability estimate is attributable to the poor data, which, in turn, is attributable to the sampling schedule: rmaxwas about 1 h and there were no samples in the absorption phase in most patients. It is noteworthy that after Table 1. Mean f SD concentrations of alminoprofen in plasma and synovial fluid Time (h)" 1

2 4 6 8 12

Number of samples

Plasma (mg L-I)

Synovial fluid (mg L-I)

7 7 7 10 6 8

24.2 f 12.8 19.0f 9.4 15.8 f 7.3 8.6 f 3.6 4.0 f 2.7 2.1 *2-0

4.7* 1.8 9.3 f 4.8 7.2 f 4.9 5.6 f 3.2 4.7f 1.4 2.3 f 3.3

~~

"Time of sampling after ingestion of the last (13th) dose.

632

M. TOD ET AL.

Table 2. Population pharmacokinetic parameters of alminoprofen in plasma and synovial fluid according to classical parameterization Population mean Parameter

Interindividual variability"

Estimate

SEb

Estimate

SEb

W)

0.236 0.0102 0.249 11.0 2.80 4.05

0.025 04048 0.097 1.71 0.309 6.97

18 37 32 12 464

14 29 23 30 430

Physiological ki (h-I) ki (h-9 v p (L) vs (L)

9.08 0.321 3.15 0.105

4.54 0.141

Classical ke101-9 kcp @-I) kpc(h - I ) V(L) ka W') 0s

-

-

-

38 36 42 3.3

-

31 24 22 8.3

"Values Listed under interindividual variability are the coefficients of variation as percentages. bStandard error of the estimate.

accounting for the interindividual variability associated with the pharmacokinetic parameters, a residual variability of only 4% remained, which is of the same order of magnitude as the analytical precision. It was not possible to introduce separate residual error terms for plasma and synovial concentrations. Also, the fit was not significantly improved by the inclusion of the covariates, and no significant correlations were found between the pharmacokinetic parameters and the covariates. This finding does not imply that such correlations do not exist, but the ranges of variation of the covariates and the parameters were too narrow to yield significant correlations. Attempts to estimate individual pharmacokinetic parameters by a Bayesian mean a posteriori estimatorlo were unsuccessful because no convergence could be met for some subjects. Therefore, direct examination of the relationships between individual parameters and covariates was not possible. Incorporation of physiological data into the model yielded an estimate of the true values of the first-order rate constant describing the influx and efflux of alminoprofen between plasma and synovial fluid. As expected, the interindividual variabilities of ki and kf were very similar to those observed for kcp and kpc respectively. Wide interindividual variability was found in the plasma volume, but not in the synovial fluid volume. Variability in plasma volume in healthy subjects is about 12-13%, and taking into account several covariates (age, weight, or body surface area) reduces interindividual variability only to 9-10%.l2 Therefore, the variability of V,,measured in this model might appear surprisingly high, but its confidence interval is also very large (SE = 22%) and consequently its estimate is uncertain. The mean f SE

633

ALMINOPROFEN SYNOVIAL PENETRATION SlMULATlON OVER 100 SUBJECTS

20 -

10 -

-

0

2

I

I

I

I

f

4

6

8

10

12

TIME AFTER THE LAST DOSE (HOURS) Figure 2. The simulation of the mean*SD alminoprofen concentration versus time curve in plasma and synovial fluid at steady state

clearance from plasma into joints was 0.29 f 0.14mLmin-l, while that in the opposite direction was 0.56 f0.25 mL min-l. These values are not significantly . eight ~ different and are very similar to those derived by Simkin et ~ 1 for different NSAIDs. These results indicate that the kinetics of alminoprofen penetration into synovial fluid are comparable to those of other NSAIDs: the penetration proceeds by simple diffusion in both directions and binding to albumin is not restrictive, i.e., does not impair the ability of the drug to enter into the extravascular space.13Therefore, protein binding has little effect on the kinetics of alminoprofen penetration into synovial fluid, but should influence the steady state level of the drug in synovial fluid compared to plasma. However, no further calculations in this respect were made, because alminoprofen, like most NSAIDs, is administered as a racemic mixture and each enantiomer is known to have specific pharmac~kinetics.'~ In particular, differences in protein binding account, in part, for the stereoselective disposition of ibuprofen enantiomers in synovial fluid at eq~ilibrium.'~ The mean fSD alminoprofen concentration versus time curve in plasma and synovial fluid at steady state obtained by simulation for over 100 subjects is shown in Figure 2. Compared to the experimental data in Table 1, two features are notable: the simulated SDs are lower than their experimental counterparts and the time to peak concentration in synovial fluid is longer than that observed in the raw data. The first feature can be explained, in part, by the fact that the simulation was done without output noise, whose variance is added to the interindividual variance of real data. The second element is probably a

634

M. TOD ET AL.

consequence of the relatively imprecise estimation of the experimental mean alminoprofen concentration at each sampling time, owing to the small number of subjects (i.e., six to 10). In contrast, population pharmacokinetic parameters were estimated from all the raw data, i.e., 90 concentration measurements. Therefore, we have more confidence in the simulated output to represent the true kinetics of alminoprofen and its variability. In summary, application of the NONMEM to a pharmacokinetic model incorporating physiological data enabled us to fully describe the interindividual variability of alminoprofen kinetics in plasma and synovial fluid and to show that diffusion was the most probable mechanism of drug penetration into the synovial fluid. REFERENCES 1. 2. 3. 4.

5. 6.

7. 8. 9. 10. 11. 12. 13. 14. 15.

S. B. Abramson and G. Weissmann, Arthritis Rheum.,32, 1 (1989). P. A. Simkin and K. L. Nilson, Clin.Rheum. Dis., 7 , 99 (1981). W. J. Wallis and P. A. Simkin, Clin. Pharmacokinet., 8, 496 (1983). P. A. Simkin, M. P. Wu and D. M. Foster, Clin. Pharmacokinet., 25, 342 (1993). P. Netter, B. Bannwarth, C. Monot, R. J. Royer and A. Gaucher, Presse Med., 12,2049 (1983). R. Rau, G. Berner, H. H. Wagener and U. Vogtle-Junkert, Arzneim. Forsch., 39, 1166 (1989). A. M. M. Miltenburg, J. M. Van Laar, R. De Kurper, M. R. Daha and F. C. Breedveld, Br. J . Rhewnatol., 30, 186 (1991). L. Aarons, J. Phurmucokiner. Biopharm., 19,265 (1991). M. Paillet, H. Merdjan, A. Brouard, D. Doucet, H. Barreteau and G. Fredj, J. Chromotogr., 343, 455 (1985). S . L. Real and L. B. Sheiner, NONMEM Users Guide, University of California, San Francisco, 1992. D. Z. DArgenio and A. Schumitzky, ADAPT ZI Users Guide, University of Southern California, Los Angeles, 1992. C. Lentner (Ed), Geigy Scientific Tubles, 8th edn, Vol. 3, Ciba-Geigy, Bade, 1984, pp. 65-67. F. Hervi, S. Urien, E. Albengres, J. C. Duchk and J. P. Tillement, Clin. Pharmucokiner., 26,44 (1994). N. Muller, E. Payan, F. Lapicque, B. Bannwarth and P. Netter, Fundam. Clin. Pharmucol., 4, 617 (1990). R. 0. Day, K. M. Williams, G. G. Graham, E. J. Lee, R. D. Knihinicki and G. D. Champion, Clin.Pharmacol. Ther., 43, 480 (1988).