Development and validation of a capillary zone electrophoresis method for the determination of propranolol and N-desisopropylpropranolol in human urine

Analytica Chimica Acta 559 (2006) 9–14

Development and validation of a capillary zone electrophoresis method for the determination of propranolol and N-desisopropylpropranolol in human urine Juan Jos´e Berzas Nevado a,∗ , Juana Rodr´ıguez Flores b , Gregorio Casta˜neda Pe˜nalvo b , Francisco Javier Guzm´an Bernardo c b

a Regional Institute for Applied Scientific Research, University of Castilla-La Mancha, Avenida de Camilo Jos´ e Cela 10, 13071 Ciudad Real, Spain Department of Analytical Chemistry and Food Technology, Faculty of Chemical Sciences, Avenida de Camilo Jos´e Cela 10, 13071 Ciudad Real, Spain c Department of Analytical Chemistry and Food Technology, Faculty of Environmental Sciences, Avenida de Carlos III s/n, 45071 Toledo, Spain

Received 19 September 2005; received in revised form 16 November 2005; accepted 18 November 2005 Available online 17 January 2006

Abstract A simple, rapid and sensitive procedure using solid phase extraction and capillary zone electrophoresis for the determination of propranolol (a beta-blocker) and one of its metabolites, N-desisopropylpropranolol, has been developed and validated. The optimum separation of both analytes was obtained in a 37 cm × 75 ␮m fused silica capillary using 20 mmol/L phosphate buffer (pH 2.2) as electrolyte, at 25 kV and 30 ◦ C, and hydrodynamic injection for 5 s. Prior to the electrophoretic separation, the samples were cleaned up and concentrated using a C18 cartridge and then, eluted with methanol, allowing a concentration factor of 30. Good results were obtained in terms of precision, accuracy and linearity. The limits of detection were 28 and 30 ␮g/L for Ndesisopropylpropranolol and propranolol, respectively. Additionally, a robustness test of the method was carried out using the Plackett–Burman fractional factorial model with a matrix of 15 experiments. The presented method has been applied to the determination of both compounds in human urine. © 2005 Elsevier B.V. All rights reserved. Keywords: Capillary zone electrophoresis; Propanolol; N-desisopropylpropanolol; Ruggedness; Urine determination

1. Introduction Propranolol (P) is a ␤-adrenoceptor antagonist, which is widely used in the treatment of several diseases such as arrhythmias, thyrotoxycosis, angina pectoris and hypertension. It is also consumed in sport and in other stressing activities as a doping agent. Accordingly, the development of rapid and direct procedures to monitor P intake is interesting. However, P is metabolised after oral administration to some metabolites, one of them being N-desisopropylpropranolol (D). Thus, several authors have proposed procedures to analyse the metabolites of P. For this purpose, liquid and gas chromatography coupled with mass spectrometry [1,2] have been used because of their

∗

Corresponding author. E-mail address: [email protected] (J.J.B. Nevado).

0003-2670/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.aca.2005.11.070

versatility. However, these equipments are very expensive. Alternatively, liquid chromatography with fluorimetric detection has been used as well [3–10]. In these cases, the limits of detection obtained were typically about 24 ␮g/L [11]. In the specific field of capillary electrophoresis (CE), one approach to improve the determination of P has been the on line preconcentration. Accordingly, Li et al. [12] have used field-amplified injection combined with acid stacking for the determination of P and metoprolol, another beta-blocker. In this case, the limit of detection of P was 0.1 ␮g/L. Likewise, a preconcentration–separation system consisting of a preconcentration capillary bonded with carboxyl cation-exchange stationary phase, a separation capillary for capillary zone electrophoresis (CZE) and a tee joint interface of the capillaries was developed by Zhang and He [13] for the determination of the same compounds. The detection limit for P was 0.02 ␮g/L. Nevertheless, no application to biological fluids, namely urine, was carried out in both papers.

J.J.B. Nevado et al. / Analytica Chimica Acta 559 (2006) 9–14

10

The extraction and preconcentration processes were achieved with a Supelco (Bellefonte, PA, USA) vacuum manifold coupled to a vacuum pump. The C18 cartridges were from Waters (Milford, MA, USA). A Crison micro-pH 2002 instrument was used for pH measurements. 2.2. Reagents Fig. 1. Chemical structures of the compounds.

Another approach to the determination of P has been reported with end-column amperometric detection in urine [14]. The limit of detection obtained in this work was 0.05 ␮M (13 ␮g/L). Additionally, the same authors reported the determination of P in conjunction with two other beta-blockers by capillary electrophoresis using an end-column electrochemical detection. In this case, a carbon fibre (33 ␮m) electrode was used as the working electrode and the limit of detection obtained was improved up to 10−8 mol/L (2.6 ␮g/L) for P [15]. Unfortunately, no metabolites of P were determined in both papers. The use of laser-induced fluorescence (LIF) detection has been also used for the determination of P [16]. In this case, the limits of detection were improved by combining single capillary isotacophoretic preconcentration and CZE separation in the same capillary together with frequency doubled argon ion LIF detection. As a result, levels of 10−9 mol/L of propranolol and metoprolol, were detected. However, the authors studied no metabolites and provided no application of this method to urine samples. It is evident that the determination of P and D using UV/visible detection cannot provide limits of detection similar to LIF or electrochemical detection in CE. For this purpose, a previous preconcentration is needed. In this way, using a solid phase extraction (SPE) step prior to the analytical separation has been widely used and reported in literature, because it enables the elimination of the interferences present in the urine and the concentration of the analytes. The aim of this work is to develop an analytically simple, sensitive and robust method for the determination of P and D (see Fig. 1 for structures) in urine by capillary zone electrophoresis. 2. Experimental 2.1. Apparatus A Beckman P/ACE 5500 capillary electrophoresis system equipped with a diode array detector was used. The system was controlled by a personal computer provided with P/ACE software. The separations were carried out in a fused-silica capillary of 37 cm total length × 75 ␮m i.d. × 375 ␮m o.d., housed in a cartridge with a 100 ␮m × 800 ␮m detection window. The capillary was conditioned by flushing first with 0.1 mol/L NaOH for 30 min and then with deionised water for 10 min before it was used for the first time.

All solvents and reagents were of analytical grade unless indicated otherwise. All the solutions were prepared with deionised water (Milli-Q quality). Propranolol hydrochloride and N-desisopropylpropranolol hydrochloride were from Tocris (Bristol, UK). Stock solutions (100 mg/L) of both of them were prepared in deionised water. Buffer solutions were prepared by adjusting the pH of a solution of H3 PO4 with HCl or NaOH to the required value. All these reagents were from Panreac (Barcelone, Spain). 2.3. Analytical pre-treatment Fresh human urine samples were obtained from different volunteers who had not taken P, for direct submission to solid phase extraction. The extraction of P and its metabolite, D, from the samples was performed in a reverse-phase C18 cartridge (Waters Sep-Pak Plus). The cartridge was conditioned prior to use with 5 mL of methanol followed by 5 mL of 10−3 mol/L phosphate buffer solution (pH 7.0). Different volumes of urine (not exceeding 6 mL) were loaded into the conditioned cartridge, which was then washed with 8 mL of 10−3 mol/L phosphate buffer (pH 7.0) and 3.5 mL of a methanol:water (50:50, v:v) solution. Then, P and D were eluted with 3.0 mL of methanol. Later on, this extract was evaporated to dryness with a gentle nitrogen stream and finally, it was reconstituted with 200 ␮L of milli-Q water and transferred to the appropriate vials, to be injected into the capillary electrophoresis equipment. 3. Results and discussion 3.1. Preliminary experiments A urine sample was spiked and cleaned-up as explained above, so that the final concentrations of both analytes were 5 mg/L. This solution, which will be called “Z”, was used throughout the optimisation of the electrophoretic separation. It is well known that the pH dramatically influences the ionisation of those compounds having acid or basic attributes and their electrophoretic behaviour. That is to say, the pKa values are a key point to select the best CE mode to be applied because the charge that a particular molecule can have at a particular pH value can be predicted. Thus, when all the analytes are charged (all positively, all negatively or some positively and the rest negatively) the CZE mode can be used. Conversely, if an analyte is not charged at a particular pH value, it is necessary to use the MEKC mode. Otherwise, that compound would never be separated from the electroosmotic flow (EOF). In this case, the pKa

J.J.B. Nevado et al. / Analytica Chimica Acta 559 (2006) 9–14

values for these compounds were estimated by using the ACD labs software and they are shown in Fig. 1. The pKa1 for both molecules corresponds to the deprotonation of the amine group, whereas the pKa2 corresponds to the deprotonation of the alcoholic group. Accordingly, both compounds should be positively charged for pH < pKa1 , so CZE should be the appropriate mode to be used all along this wide pH range. This was confirmed by running several separations of standard solutions from pH 2.2 up to 9 because in all these cases the migration times of the peaks were lower than that of the EOF. However, there is another factor to be considered to select the optimum pH. This is the matrix effect. According to our experience, the CE separations in urine should preferably be in acid media because this eliminates possible interferences coming from the urine. For these reasons, a pH value of 2.2 was selected as a start point in the development of the method. This pH value was provided by a phosphate buffer solution. Thus, a preliminary separation of a “Z” sample was performed under the mentioned conditions and all the explained above was proved right.

11

Fig. 2. Influence of the voltage on the separation. The black line corresponds to D and the grey line to P.

studied because the temperature regulation of our instrument is not efficient beyond 4 ◦ C under room temperature. As expected, an increase in temperature resulted in a decrease of migration times due to the electrolyte viscosity increase. Regarding to resolution, run time and current generated (85.4 ␮A), 30 ◦ C was selected as suitable.

3.2. Effect of pH

3.6. Selected conditions

The effect of the pH on the separation was evaluated by separating a “Z” solution using phosphate buffer (30 mM) adjusted at pH 1.7, 2.2 and 2.7 as electrolyte. All these separations were performed at 15 kV and 25 ◦ C. As a result, the migration times and the resolutions showed no variations along the pH range studied because there were no changes in the ionisation of the analytes. Then, a pH of 2.2 was selected as suitable because it provided the best buffering capacity, as explained above.

From the studies carried out before, we can state that the procedure summarised below is convenient for an efficient separation of the studied compounds: a fused-silica capillary of 37 cm × 75 ␮m inner diameter; 20 mmol/L phosphate buffer (pH 2.2) as electrolyte; 30 ◦ C and 25 kV (147 kV/min in 0.17 min); auto-zero at 1 min and a detection window of 100 ␮m × 800 ␮m. A typical electropherogram obtained under these conditions is shown in Fig. 3. As can be seen, D migrates faster than P, which is in agreement with the CE theory. Thus, when positively charged compounds are separated, they will migrate towards the cathode, where the detector is often placed, according to their mass to charge ratio. Obviously, when two compounds have the same charge, the lightest will migrate faster than the heaviest. In this particular case, both compounds have the same charge, but D is smaller than P, so that is why D migrates faster than P. The capillary was rinsed with NaOH 0.5 mol/L for 1 min and then with electrolyte for 2 min between two consecutive separa-

3.3. Influence of the buffer concentration The optimisation of the buffer concentration was carried out by preparing a set of three electrolytes, containing pH 2.2 phosphate buffer at concentrations ranging from 5 to 40 mmol/L. The separation of the “Z” solution was carried out at 15 kV and 25 ◦ C. As a result, a 20 mmol/L concentration was selected as suitable because it was enough for buffering properly and because the current generated (38.0 ␮A) did not produce a significant Joule heating. 3.4. Influence of the voltage on the separation The effect of varying the voltage from 5 to 30 kV was investigated under the conditions selected above. The influence of the voltage on the migration time of the compounds and on the current generated is plotted as Fig. 2. A voltage of 25 kV provided the best compromise in terms of run time, current generated and linearity between voltage and current. This voltage was used for subsequent stages of the method optimisation. 3.5. Effect of the temperature on the separation The effect of the temperature on the separation was tested between 20 and 35 ◦ C. Temperatures lower than 20 ◦ C were not

Fig. 3. Electropherogram corresponding to a type “Z” standard solution, recorded at 214 nm.

J.J.B. Nevado et al. / Analytica Chimica Acta 559 (2006) 9–14

12

Table 1 Calibration curves, coefficients of determination, LODs and LOQs (␮g/L) Compound

Equationa

D P

A = −0.34(±0.64) × 103

+ 176.7(±5.2) × 102

c

A = 0.4(±1.2) × 103 + 174.9(±9.9) × 102 c

a

r2

LOD

LOQ

0.9974 0.9904

28 30

93 77

Concentration (c = mg/L) vs. Corrected peak area (A).

tions to avoid adsorption processes in the internal surface of the capillary. 3.7. Quantitative aspects The analytes were monitored and quantified at their maximum absorption wavelength, 216 nm, in order to obtain the maximum signal-to-noise ratio. 3.8. Limits of detection and quantification The limits of detection (LODs) and quantification (LOQs) were estimated in accordance with the baseline noise from an urine extract. The baseline noise was evaluated by recording the detector response over a period of about ten times the peak width. The LOD was obtained as the concentration of a sample that provided a peak with a height three times the baseline noise level and the LOQ was calculated as ten times the baseline noise level. As a result of the cleanup process, the samples were concentrated up to 30 times (6 mL urine to 0.2 mL of reconstituted extract). Thus, the LODs and LOQs obtained are summarised in Table 1. 3.9. Linearity range and calibration curves The calibration curves were obtained by spiking urine extracts at five levels, ranging from the LOQ up to 2 mg/L. The equations, including standard deviations for intercepts and slopes and determination coefficients for the calibration curves, are summarised in Table 1. As a result, the linearity of the calibration curves was satisfactory within the studied range. Likewise, in all cases, the intercepts were estimated as negligible by using Student’s t-test (α = 0.05). 3.10. Precision The repeatability was assessed by running 12 replicates of a urine, which was spiked with P and D so that the concentrations at the end of the analytical pre-treatment were 5 mg/L. The results showed that the relative standard deviation of the corrected peak area for each compound was under 4.5% in all cases. Likewise, the relative standard deviations obtained for the migration times were under 1.5%. The reproducibility was assessed by doing the same as above over two consecutive days. The variances of migration time and corrected peak area for each compound were compared by using the Snedecor’s F-test. As a result, no significant differences were found (P = 0.05).

3.11. Robustness The purpose of a robustness test is to identify possible sources of error when changes occur in the specified method conditions [17]. Fractional factorial designs developed by Plackett and Burman [18], based on balanced incomplete blocks were used for this test. Mean effects and standard errors were calculated according to the procedures described by Youdner and Steiner [19]. The Plackett–Burman design for seven components or variables, as (−1, 0, 1), was set up as follows: • • • • • • •

A: Voltage (23, 25, 27) (kV) B: Injection time (4, 5, 6) (s) C: Phosphate concentration (18, 20, 22) (mmol/L) D: Separation temperature (29, 30, 31) (◦ C) E: pH (2.0, 2.2, 2.4) F: Electrolyte rinsing time (1.8, 2.0, 2.2) (min) G: Measurement wavelength (215, 216, 217) (nm)

The mean effects and standard errors (DA, DB, DC, . . .) were calculated using the procedures described by Youden and Steiner. The robustness was determined in our case from triplicate injections of urine extracts containing 3 mg/L of each compound. The results of the effect of each factor level on the resolution, efficacy, migration time, peak width and corrected peak areas were calculated. The efficacy was expressed in terms of the number of theoretical plates (N):
 Migration time 2 N = 16 Width Taking into account the deviations calculated for the different results, the selected operating factors were tested using Plackett–Burman experimental design and Youden–Steiner statistical analysis. As a result, this method has proved to be robust for all the variations tested in this study. The effects of these factors on the efficacy of both peaks is shown in Fig. 4, as an example. In this particular case, the values of the variations of the seven factors on the efficacy were always within the range calculated using the Youden–Steiner statistical model, which means that the method is robust in terms of efficacy. Likewise, similar results were obtained for resolution, migration time, peak width and corrected peak areas. 3.12. Accuracy Three samples of urine were spiked at three different levels (U1, U2 and U3) and cleaned up as described earlier on. Like-

J.J.B. Nevado et al. / Analytica Chimica Acta 559 (2006) 9–14

13

Fig. 5. Electropherogram corresponding to a urine sample analysed 12 h after the drug intake, recorded at 214 nm.

4. Conclusion

Fig. 4. Variation effects of the levels (−1, 0, 1) of the seven selected operating factors on the efficacy of: (a) Propranolol and (b) N-desisopropylpropranolol.

wise, standard solutions of the compounds were prepared for quantification purposes. Triplicate separations of each sample were carried out before and after aqueous standards of similar concentrations were run under the optimised conditions. The corrected peak area, measured at the maximum absorption wavelength, 216 nm, was selected for quantification purposes. The results of these analyses, shown in Table 2, provided recoveries close to 100%. 3.13. Application In order to demonstrate the applicability of the extraction, preconcentration and the developed CE method, several urine samples were collected from a healthy volunteer who had taken 20 mg and analysed for P and D. The samples were collected every hour during 24 h. The results showed a concentration of 24 ␮g/L of P 12 h after the intake (see Fig. 5 for electropherogram). However, D could only be detected, but not quantified. For the rest of the samples, the concentrations of P and D were below the limit of detection. Table 2 Recoveries obtained in fortified samples of human urine Sample

U1 U2 U3

D

P

Added (␮g/L)

Recovery (%)

Added (␮g/L)

Recovery (%)

200 400 20

97.0 95.3 100.5

210 210 210

104.3 109.0 99.7

The results show that the presented procedure, involving both the analytical pre-treatment and the electrophoretic separation, is adequate for the detection and/or determination of P and D, one of its metabolites, in human urine. The presented pre-treatment improves the limits of detection and quantification expected for the CE techniques with UV/visible detection, because the sample is concentrated up to 30 times. The limits of detection obtained are similar to those reported previously in literature for LIF and electrochemical detectors, which are more sensitive than UV/visible, but suffering from major drawbacks, such as high cost (LIF and electrochemical), interferent adsorption processes occurring on the electrode surface and/or need of home-made manufacturing (electrochemical). The CZE mode complies with the requirements of clinical and/or forensic analysis in terms of reproducibility, accuracy and robustness, and it is also useful for routine analysis. Additionally, it offers advantages such as simplicity of operation, flexibility and economy. References [1] K. Hartonen, M.L. Riekkola, J. Chromatogr. B 676 (1996) 45. [2] H. Kataoka, S. Narimatsu, H.L. Lord, J. Pawliszyn, Anal. Chem. 19 (1999) 4237. [3] E.C. Kwong, D.D. Shen, J. Chromatogr. 414 (1987) 365. [4] P.M. Harrison, A.M. Tonkin, C.M. Cahill, A.J. McLean, J. Chromatogr. 343 (1985) 349. [5] K.A. Smith, S. Wood, M. Crous, Analyst 112 (1987) 407. [6] S.A. Qureshi, H.S. Buttar, J. Chromatogr. 431 (1988) 465. [7] V.G. Beliolipetskaja, V.K. Piotrovskii, V.I. Metelitsa, S.A. Pavlinov, J. Chromatogr. 491 (1989) 507. [8] H.A. Semple, F. Xia, J. Chromatogr. B 655 (1994) 293. [9] N.E. Basci, A. Temizer, A. Bozkurt, A. Isimer, J. Pharm. Biomed. Anal. 18 (1998) 745. ´ [10] I. Rapado-Mart´ınez, R.M. Villanueva-Camu˜nas, M.C. Garc´ıa-AlvarezCoque, Anal. Chem. 71 (1999) 319. ´ [11] M.J. Ruiz-Angel, P. Fern´andez-L´opez, J.A. Murillo-Pulgar´ın, M.C. ´ Garc´ıa-Alvarez-Coque, J. Chromatogr. B 767 (2002) 277–283.

14

J.J.B. Nevado et al. / Analytica Chimica Acta 559 (2006) 9–14

[12] Y. Li, Y.Z. He, Y.Y. Hu, L. Wang, Fenxi Huaxue 33 (2005) 248–250. [13] Z.X. Zhang, Y.Z. He, J. Chromatogr. A 1066 (2005) 211–218. [14] X.X. Bai, T.Y. You, H.W. Sun, X.R. Yang, E.K. Wang, Electroanalysis 12 (2000) 535–537. [15] X.X. Bai, T.Y. You, H.W. Sun, X.R. Yang, E.K. Wang, Electroanalysis 12 (2000) 1379–1383.

[16] A.M. Enlund, S. Schmidt, D. Westerlund, Electrophoresis 19 (1998) 707–711. [17] M. Mulholland, J. Waterhouse, J. Chromatogr. 395 (1987) 539–551. [18] R.L. Plackett, J.P. Burman, Biometrika 33 (1946) 305–325. [19] W.J. Youden, E.H. Steiner, Statistical Manual of the AOAC, AOAC International, Washington D.C., 1975, pp. 33–41.