Analyst, January 1996, Vol. 121 (61-66) 61 Investigation of On4 i ne Reversed-phase Liq u id C h r o mat og rap h y-Gas Chromatography-Mass Spectrometry as a Tool for the Identification of Impurities in Drug Substances Elise C. Goosensa, Karel H. Stegmana, Dirk de Jonga, Gerhardus J. de Jong" and Udo A.Th Brinkmanb a Solvay Duphar, Analytical Development Department, P.O. Box 900, 1380 DA, Weesp, The Netherlands Boelelaan 1083, I081 HV, Amsterdam, The Netherlands Free University, Department of Analytical Chemistry, De The potential of on-line reversed-phase (RP) LC-GC-MS for the identification of impurities in pharmaceutical products has been investigated. The technical aspects of the system were studied using the potential drug eltoprazine as test compound. After LC separation on a 2 mm id RPLC column, using 5 mmol 1-1 methanesulfonic acid in acetonitrile-water (84 + 16 v/v) as eluent, the eltoprazine-containing fraction of 200 p1 was introduced on-line into the GC-MS, the mass spectrometer being a magnetic sector instrument.Before this introduction, the methanesulfonic acid was removed on-line from the LC eluent via an anion-exchange membrane. Most of the solvent introduced into the GC-MS was evaporated via a solvent vapour exit installed in front of the capillary GC column. As an application, impurity profiling was performed on the drug substance mebeverine which contained mebeverine amine as test impurity. The addition of acetonitrile was necessary before introduction into the GC; therefore only about 10% v/v of the mebeverine amine LC peak was transferred to the GC-MS.Nevertheless, electron impact and chemical ionization spectra of the impurity could be obtained at a level of 0.1% with respect to the drug. Keywords: Impurity profiling; reversed-phase liquid chromatography; gas chromatography-mass spectrometry, drug substance Introduction During the various stages of developing new pharmaceutical products, such as impurity profiling of the drug substance, metabolic studies and stability studies of the drug substance and the formulated product, unknown compounds are often ob- served in liquid chromatography (LC). These impurities have to be characterized and identified if they are present at a level exceeding 0.1 %area with respect to the drug substance. In general, these unknowns are isolated by preparative LC and subsequently characterized by NMR, IR and MS.This is often a time-consuming process, mainly because of the required isolation procedure. Moreover, artefacts can be formed during isolation, which often includes evaporation and storage. It would therefore be highly desirable to couple the LC system directly to these spectroscopic devices. The practicality of LC- MS for drug identification, impurity and metabolite profiling and identification has been demonstrated'-3 and its potential has increased now that several interfaces have become available. LC-MS still has its limitations when the LC eluent contains non-volatile buffer salts or other additives. In addition, sufficient structural information of the compound of interest is not always provided, unless LC-MS-MS is performed.Another approach is to couple LC on-line with capillary gas chromatog- raphy (GC)-MS. The advantage of this approach is that GC-MS is a routine procedure which is highly sensitive and provides adequate structural information when using standard ionization techniques such as electron impact (EI) and chemical ionization (CI). Moreover, the extra separation step introduced in LC-GC- MS increases the over-all efficiency. On the other hand, application of the technique is limited to compounds amenable to direct GC analysis. In 1987 Raglione et al.4 described the use of LC-GC-MS for the analysis of solvent-refined coal. Liquid chromatography was used for ring-size separation, with subsequent GC-MS for isomer separation and identification.On-line LC-GC-MS was succesfully applied by Ostman and Nilssons and Vreuls et a1.6 for the determination of polycyclic hydrocarbons in urban air and vegetable oils, respectively. They all used normal-phase LC systems, i.e., apolar mobile phases. Ogorka and co-workers,7.8 on the other hand, used reversed-phase (RP) LC-GC-MS for the identification of unknown compounds observed in RPLC chromatograms during the development of new drug sub- stances. The authors circumvented the problems arising from the high water content of the LC eluent and the presence of buffer salts by inserting a liquid-liquid extraction step between the RPLC and GC parts of the system. In this way, the impurity of interest was extracted from the aqueous LC fraction into an organic solvent.Subsequently, 500 p1 of the extract were transferred on-line to the GC-MS part of the system. Depending on the analyte and composition of the LC eluent, dichloro- methane, n-pentane or n-hexane was used as the extraction solvent. In earlier papers we reported on the feasibility of the direct coupling of RPLC with GC.9-11 Acetonitrile-water mixtures containing methanesulfonic acid (MSA) as an ion-pair reagent were used as the eluent in a RPLC-GC set-up in which MSA was removed by an anion-exchange membrane prior to introduction into the GC.' The potential drug eltoprazine was used as the test compound, and LC columns with an id of 2 mm and eluent flow rates up to 200 p1 min-1 were used. Insertion of a solvent vapour exit in front of the capillary GC column allowed an introduction volume into the GC of 200 pl at an introduction rate of 200 pl min-1.The technique was found to be limited to acetonitrile-water mixtures in which the percent-62 Analyst, January 1996, Vol. 121 age of water does not exceed that of the azeotrope (84 + 16 v/v) . 9 3 10 In this paper the use of RPLC-GC-MS for the identification of impurities and degradation products of pharmaceutical products was studied using a magnetic sector mass spec- trometer; eltoprazine was selected as the test compound. The drug substance mebeverine and some minor impurities were selected as model compounds for the impurity profiling study. Mebeverine amine was selected as the 'unknown impurity' which had to be identified at a level of 0.1% of mebeverine.Experimental Chemicals Eltoprazine, mebeverine, mebeverine amine, butoverine, verat- ric acid and 3-desmethyl-mebeverine were from Solvay Duphar (Weesp, The Netherlands). Tetrabutylammonium hydroxide 30-hydrate was purchased from Fluka (Buchs, Switzerland). Potassium hydroxide, potassium dihydrogenphosphate and sulfuric acid were from J.T. Baker (Deventer, The Netherlands) and methanesulfonic acid from Merck (Darmstadt, Germany). All other reagents were of analytical-reagent grade. Before use, the eluents and regenerant solutions were filtered through a Millipore (USA) HA membrane (0.45 mm) to remove small particles. Apparatus and Procedures The LC system consisted of a Model 260D syringe pump (ISCO, Lincoln, NE, USA), a 25 cm X 2 mm id Chromspher poly-cls LC column from Chrompack (Bergen op Zoom, The Netherlands), a Rheodyne (Cotati, CA, USA) injection valve with an internal loop volume of 1 pl (Type 7413) and a UV detector from Jasco (Tokyo, Japan).The GC part of the system consisted of two GCs, a Mega GC (Carlo Erba Strumentazione, Milan, Italy) with on-column injector and autosampler (ASSSO), in which the retention gap (10 m X 0.53 mm id CPWax52CB, d f = 0.02 pm, Chrompack) was installed, and a second Mega GC for the capillary analytical column (30 m X 0.32 mm id DB-1, d f = 0.25 pm; J&W, Folsom, CA, USA). The two GCs were connected via a heated interface with a length of 70 cm (Horst, Lindenfels, Germany) to control the temperature of the final part of the retention gap. A solvent vapour exit (SVE) was installed in the second GC between the retention gap and the analytical column via a Y- shaped glass press-fit.A 10 cm X 0.32 mm id fused silica capillary connected this Y-piece with another glass Y-piece which was connected with an on-off pinch solenoid valve (Sirai, Pioltello, Italy) and a restrictor (1 m X 0.05 mm id X 0.36 mm od deactivated fused silica, LC-Service, Emmen, The Netherlands). Apart from the valve, the whole SVE assembly was installed in the GC oven. A cation micromembrane suppressor (CMMS-2 mm, Dio- nex, Sunnyvale, CA, USA) was inserted in the eluent stream between the LC column and the UV detector. The regenerant was delivered by an SSI (State College, PA, USA) pump. If necessary, acetonitrile was added to the LC eluent via a Spectroflow 400 pump (Kratos Analytical, Ramsey, NJ, USA).Electron impact and CI spectra were recorded on a Kratos (Manchester, UK) Concept 1 S double focusing mass spec- trometer, which was coupled to a Sun computer with Mach3 software for computing data. Perfluorokerosine was used as the reference compound for calibration. A schematic presentation of the total RPLC-GC-MS system is given in Fig. 1. After sample introduction into the LC system and separation on the LC column, the eluent flows through the CMMS device which contains the anion-exchange membrane. The regenerant solution is continuously pumped along the other side of the membrane in a direction opposite to that of the LC eluent. MSA in the eluent is exchanged with the hydroxide ions from the regenerant. After the passage through the CMMS, acetonitrile can be added to the eluent in order to achieve the azeotropic acetonitrile-water ratio (84 + 16 v/v).The eluent is then flushed through the syringe needle of the on-column autosampler and subsequently led to waste. Transfer of an LC fraction is accomplished by introducing the needle for a few seconds, the transfer time, into the retention gap. The transfer volume, i.e., the volume introduced into the GC, can be calculated from the introduction rate and transfer time. As soon as the transfer is started, the temperature programmes of both GCs are started, the valve of the SVE is opened and most of the solvent is discharged to the atmosphere. Following evaporation of the solvent, the SVE valve is closed and gas flows through the retention gap and the analytical GC column to the MS, purging the remaining solvent in the SVE line via the restrictor.10 By starting the temperature ramp of the retention gap (after 3 min) prior to that of the GC column (after 8 min), a cold trapping effect can be created at the top of the analytical column; as a result, narrow reconcentrated peaks are obtained.gJ0 The mass spectrometer started to acquire data after a typical solvent delay time (4 min for eltoprazine and 2.5 min for mebeverine amine).The mass range mlz 50-500 was scanned at 0.8 s decade-]. The transfer line temperature was 280 OC, the source temperature 100 "C, the electron current 500 pA and the electron energy 70 eV. In the CI mode, ammonia-methane (90 + 10 v/v) was used as the reagent gas. All further RPLC-GC-MS conditions for the determination of eltoprazine and mebeverine amine are given in Table 1.Results and Discussion Development of the RPLC-GC-MS system In a previous paper we described the on-line coupling of RPLC and GC for the determination of eltoprazine." A 2 mm id LC eluent pump UV detector ionisation valve MS Fig. 1 Schematic diagram of the RPLC-GC-MS instrumentation.Analyst, January 1996, Vol. 121 63 column was used and the eluent was acetonitrile-water (84 + 16 v/v) which contained 5 mmoll-1 methanesulfonic acid (MSA), the flow rate being 200 p1 min-1. By using a thin film coated Carbowax retention gap, acetonitrile-water mixtures can be introduced into the GC at introduction rates of up to 200 p1 min-l and introduction volumes up to 200 p,l with a restriction to the water content of the RPLC eluent which should not be higher than about 16% v/v.9JO However, because any addition of buffer components or additives, even volatiles, ruined the retention gap or distorted the peak shape of the analyte," MSA had to be removed from the eluent before introduction of the eltoprazine-containing fraction into the GC.About 99.9% of MSA was removed from the LC eluent via an in-line coupled anion-exchange micromembrane inserted be- tween the LC and GC parts of the system. MSA ions were exchanged with hydroxide ions present in large excess in the regenerant. The regenerant also contained 84% v/v acetonitrile, in order to prevent diffusion of acetonitrile from the eluent to the regenerant side of the membrane, which would disturb the percentage of water in the eluent.A 200 p1 volume of LC effluent, free of MSA, was introduced into the Carbowax- coated retention gap using an SVE in front of the capillary GC column to eliminate most of the solvent. No losses of eltoprazine were observed (recovery, 99%) and repeatability was satisfactory (relative standard deviation at the 150 pg ml- level, 3%). In the present study, this RPLC-GC system was coupled with a magnetic sector mass spectrometer instead of a flame ionization detector. Fig. 2 shows a total ion current (TIC) chromatogram as well as an EI spectrum of eltoprazine obtained after a 1 p1 injection of a 150 pg ml-1 eltoprazine standard solution. Of the other peaks shown in the chromatogram, the last two are siloxane-containing compounds due to stripping of the stationary phase of the GC column.The first peak is tributylamine; probably this is an impurity from the regenerant. Obviously, the EI-spectrum of eltoprazine obtained by means of LC-GC-MS, with the molecular ion at m/z 220 and its relevant fragments at m/z 178 and m/z 163, is fully comparable with the library spectrum. Although this is a satisfactory result, there is one main aspect that has to be considered when coupling the MS to the RPLC-GC system: the MS has to tolerate the introduction of large volumes of acetonitrile-water vapour. In our RPLC- GC system, about 95% of the LC eluent is eliminated via the solvent vapour exit. This means that, out of the introduction volume of 200 pl of acetonitrile-water (84 + 16 v/v), 10 pl will be transferred to the MS.Preliminary experiments, however, showed that the maximum volume of solvent that can be introduced into the ion source is about 3 p1. The introduction of more solvent disturbed the electronics and vacuum of the system. The problem was solved by temporarily closing the isolation valve between the ion source and the mass analyser of the MS (see Fig. 1) during solvent introduction, which was indicated by a pressure increase in the ion source. The valve was re-opened when the vacuum had been restored and data acquisition was started after the solvent delay time. In other words, the introduction of 200 yl of acetonitrile-water into the GC-MS apparently does not create problems if an SVE is used and the spectrometer is protected during solvent introduction into the MS. For comparison, Ogorka and c o - w ~ r k e r s ~ ~ ~ transferred LC fractions of 500 p1 apolar solvents to a GC-MS, with the difference that they used a quadrupole-type MS instead of a magnetic sector instrument.In order to avoid malfunctioning of the MS, next to an SVE they also used a GC-MS open-split interface (ratio about 1 : 11) and the filament was switched off for about 10 min. Coupling LC-GC to a magnetic sector MS has some advantages over a quadrupole MS; for example, high resolution MS reveals the exact mass of the mass peaks and can therefore facilitate the elucidation of the structure or identifica- tion of unknown compounds. On the other hand, the scanning rate that can be used with a quadrupole MS is higher compared with a magnetic sector instrument.In our system we have tried to find an optimum between the number of scans per peak and the ion intensity by varying the scanning rate from 0.3 s per decade (decrease in ion intensity) to 2 s per decade (fewer scans per peak, leading to the risk of missing the peak). As a compromise, 0.8 s per decade was selected for our experiments. Application: Impurity Profiling A mixture of mebeverine and some of its possible impurities (butoverine, veratric acid, mebeverine amine and 3-des- methylmebeverine) was selected as the test mixture to study the potential of RPLC-GC-MS as an identification technique for impurity profiling. Mebeverine amine was selected as the test impurity. Because of its low thennostability, mebeverine itself ~~ - ~~ ~~ ~ Table 1 RPLC-GC-MS conditions for the determination of eltoprazine and mebeverine amine Conditions Eltoprazine Mebeverine amine LC System Eluent Injection volume Flow rate UV detection Acetonitrile-water (84 + 16 v/v) + 5 mmol 1-1 Acetonitrile-water (50 + 50 v/v) + 5 mmol 1-1 MSA MSA or 1 mmol 1-1 MSA 1 Pl 1 czl 200 yl min-1 254 nm 200 p1 min-1 220 nm LC-GC Interface Membrane CMMS-2 mm CMMS-2 mm Regenerant Flow rate 2 ml min-1 2 ml min-1 Acetonitrile addition - 500 pl rnin-1 60 mmol 1-l TBAOH in acetonitrile-water 100 mmol 1-1 KOH in acetonitrile-water (84 + 16 v/v) (50 + 50 v/v) GC System Introduction rate 200 pl min-1 Temperature GC 1 Temperature GC2 SVE vent-time 90 s Inlet pressure 150 kPa He Introduction volume 200 p1 Temperature interface 200 "C 95 "C (3 min), 30 "C min-1,200 "C 80 "C (8 min), 30 "C min-1,280 "C 700 pl min-1 60 pl 95 "C (3 rnin), 30 "C min-I, 280 "C 80 "C (8 rnin), 30 "C min-',280 "C 150 "C 15 s 150 kPa He64 Analyst, January 1996, Vol.121 cannot be analysed by GC. Therefore, LC acts as a pre- separation step, separating the main product and the impurities from each other. The first step was to develop an LC procedure involving the use of an eluent that is compatible with the GC system. In the existing LC procedure, acetonitrile-50 mmol 1-1 potassium phosphate buffer (pH 6) (40 + 60 v/v) was used as the eluent. An attempt was made to remove this non-volatile additive from the eluent by coupling an anion- and cation-exchange membrane in series between the LC and GC parts of the system.However, the attempts failed because the phosphate ions were not completely removed by the CMMS device (a removal of 75% of the phosphate ion, determined by ion chromatography). Therefore, we preferred to add MSA to the eluent and to use the CMMS for anion removal only. Fig. 3 shows the LC separation of the mixture before and after passage through the CMMS, using acetonitrile-water (50 + 50 v/v) containing 5 mmol 1-1 MSA as eluent. The peak areas of both veratric acid (a) and 3-desmethylmebeverine (c) are seen to be distinctly smaller after passage through the membrane. For the former analyte, with its acidic nature, this is according to expectation. However, the loss observed for the tertiary amine (c) can not easily be explained. The peak area of the test impurity, mebeverine amine (b), remained the same.Passing the membrane device did cause some peak broadening, as was already observed in a previous study." t 6.44 9.29 12.13 14.58 17.43 Time - 178 i 9 m/z 80 io 80 80 loo lb lu, Is0 leo 2.00 820 m/z Fig. 2 (a) TIC chromatogram and (b) EI spectrum of eltoprazine obtained by LC-GC-MS of a standard solution containing 150 pg ml-1 of eltoprazine; (c) library spectrum of eltoprazine. *, Siloxane-containing compounds; t tributylamine. LC injection volume, 1 p1; transfer volume, 200 p1. For further conditions, see Table 1. Since the maximum percentage of water in an acetonitrile- water mixture that can be introduced into the GC system is 16% v/v, our aim was to minimize the percentage of water in the eluent, while still maintaining sufficient separation between mebeverine and each of the impurities. By using a Chromspher Poly C18 LC column, the minimum percentage of water was found to be 50% v/v.Because acetonitrile can diffuse through the membranes,ll 90% v/v of acetonitrile was added to the regenerant solution to provide acetonitrile diffusion to the eluent side. Unfortunately, the increase of the acetonitrile content of the mobile phase was only minor (2% v/v). Consequently, after passage through the CMMS, acetonitrile still had to be added to the LC eluent in order to achieve the required 84 + 16 (v/v) acetonitrile-water ratio. By adding 500 pl min-1 of acetonitrile to the LC eluent [acetonitrile-water (50 + 50 v/v); flow rate, 200 pl min- 13, the required 84 : 16 ratio was obtained. A main consequence of this addition was that the introduction rate into the retention gap increased from 200 to 700 yl min-1 , while the evaporation rate during introduction into the GC (175 pl min-I), of course, did not change. The rather large difference between the introduction and the evaporation rate caused a serious reduction of the maximum introduction volume, as is shown in Table 2.From the maximum introduction volume of 58 pl, 2/7 (16 pl) originates from the LC eluent. As the mebeverine amine- containing LC fraction is about 160 yl, only 10% v/v of the LC peak can be introduced into the GC-MS. If the 16 p1 heart-cut is taken at the peak maximum, the mass percentage of impurity that is transferred will be about 20%. Fig. 4 shows a TIC of the LC-GC-MS transfer of mebeverine amine obtained with a 0.2 mg ml-1 mebeverine amine standard solution (1 pl injected on the LC) and the pertinent EI spectrum.Because of extensive d - e 1 9 b a I. Time/min 10 Fig. 3 of mebeverine (d) and 25 pg ml-l of mebeverine amine (b) and some other impurities [veratric acid (a) 3-desmethylmebeverine (c) butoverine (e)]; (a) before CMMS passage, (h) after CMMS passage. LC injection volume, 1 pl; eluent, acetonitrile-water (50 + 50 v/v) containing 5 mmol 1-1 of MSA; flow rate eluent, 200 pl min-I; regenerant, 100 mmol 1-1 KOH in acetonitrile-water (50 + 50 v/v); flow rate regenerant, 2 ml min- l . LC-UV chromatogram of a standard mixture containing 1 mg ml- Table 2 Effect of raising the introduction rate into the GC Introduction rate 700 p1 min-' Evaporation rate 175 pl min-' Maximum volume of liquid in 10 m retention gap (9) 45 p1 Maximum transfer time 5 s Maximum introduction volume 58 plAnalyst, January 1996, VoE.121 65 50 I fragmentation into fragments with m/z 149, 121 and 72, the molecular ion of the compound (m/z 193) could not be detected. In a second run, a CI spectrum was therefore also recorded. This distinctly shows the protonated molecular ion m/z 194 (Fig 4). Obviously, using both ionization modes, EI and CI, in tandem, is an interesting tool for obtaining structural information. The mixture shown in Fig. 3 contained 2.5% m/m of mebeverine amine (25 pg ml-1) compared with mebeverine. After transfer of the mebeverine amine fraction to the GC-MS, about 5 ng of this compound is finally detected, which amount was considered to be close to the identification limit (see below). Since it was our aim to identify impurities as low as 0.1 % m/m of the main compound, the mebeverine concentration in the mixture had to be increased from 1 mg ml-1 to 25 mg ml-1.Maintaining the same mebeverine amine concentra- tion (25 pg ml-I), a 0.1% m/m level was then obtained. Introduction of this mixture into the LC system gave a mebeverine peak that was strongly overloaded; consequently, mebeverine amine co-eluted with the main peak. Decreasing the 100 50 0 11 ' I CH, Mebeverine amine I_ 501 I 121 ,., , . , , / ' , , , ;, . 60 80 100 120 140 160 180 200 220 240 0 1 J , I , , , , , , , , , , , I , m/z 194 100 3 I 1 72 I MSA concentration from 5 to 1 mmol 1-1 improved the separation of mebeverine amine from mebeverine sufficiently [(Fig.5(a)]. On-line transfer of the mebeverine amine fraction (about 5 ng) to, and analyses by, GC-MS yielded the EI and CI spectra shown in Fig. 5(b), which shows the two most important fragments (EI: m/z 72 and 121) and the protonated molecular ion (CI: m/z 194), respectively. The total and the extracted ion chromatograms in Fig. 5(a) show that this amount is very close to the identification limit of the system. If necessary, single ion monitoring at m/z 121 and 72 will increase sensitivity and therefore will contribute to confirm the presence of these ions in the peak. Obviously, an identification level of 0.1 % with respect to the drug substance can be achieved, despite the relative inefficiency of the LC separation, the unfavourable degree of fragmentation of the test impurity and the relatively high water percentage of the LC eluent.b I 0 28 Ti me/m in 5.1 3 7.57 10.42 13.26 16.11 18.55 Time + '"1 50 4 72 121 60 80 100 120 140 160 180 200 0 m/z 194 50 60 80 100 120 140 160 180 200 220 01 m/z Fig. 5 (a) LC-UV chromatogram of a mixture containing 25 mg ml-1 of mebeverine (d) and 25 pg ml-I (0.1% m/m) of mebeverine mine (b) and veratric acid (a) after CMMS passage with a TIC and extracted ion chromatograms (mlz 72 and mlz 121) of the transferred mebeverine amine fraction (about 5 ng) obtained by LC-GC-MS. (b) The pertinent EI and CI spectra of the transferred mebeverine amine fraction. LC injection volume, 1 y1; eluent, acetonitrile-water (50 + 50 v/v) containing 1 mmol I-* MSA; flow rate eluent, 200 yl min-1; regenerant, 100 mmol 1-1 KOH in acetonitrile-water (50 + 50, v/v); flow rate regenerant, 2 ml min-l; acetonitrile addition, 500 p.1 min-'; transfer volume, 58 pl.For further conditions, see Table 1.66 Analyst, January 1996, Vol. 121 Conclusions The present work gives a fair idea of the potential and limitations of on-line RPLC-GC-MS. No technical problems are encountered when 200 pl LC eluent fractions are used for analysis, and additives such as MSA can be removed by means of a membrane-based anion exchange. If precautions are taken, such as solvent venting in the GC part of the system and protection of the MS analyser during solvent introduction, MS detection does not meet with any serious problems, even when using a magnetic sector instrument.Finally, the sequential recording of EI and CI spectra adds to the identification power of the total LC-GC-MS set-up. The main limitation of the present system is the maximum percentage of water in the LC eluent of 16% v/v. As most RPLC separations require the use of distinctly higher water contents, post-LC addition of acetonitrile is necessary which adversely affects the maximum allowable introduction volume into the retention gap and, thus, the sensitivity of the total procedure. The use of LC columns with smaller internal diameter (0.32-1 mm) will not really be beneficial because of the negative influence on the loading capacity. One recommendation should be to select an LC stationary phase which enables separations of the analytes of interest with an LC eluent which is rich in organic modifier.Another aspect of some concern is the ion- exchange membrane device, which gets clogged rather easily. It is therefore important to filter all solutions prior to use. Besides, one should always be aware that analyte losses may occur as a result of (largely unknown) analyte-membrane interactions. As regards a comparison with LC-MS, both techniques can obviously play a complementary role in impurity profiling. LC- MS is a powerful technique for the determination of molecular ions of relatively polar and high relative molecular mass compounds that are not amenable to direct analysis by GC. However, for all compounds that can be subjected to direct GC analysis, LC-GC-MS is the method of choice because of the improved separation efficiency, the higher sensitivity as well as the much more versatile and powerful identification.It should be noted that the GC behaviour is unknown prior to the identification. Finally it should be admitted that, because of the in-between GC step, two typical problems of LC-GC-MS, viz., the presence of buffer salts and the percentage of water in the eluent, are less stringent in LC-MS. Here, volatile buffer constituents like ammonium acetate and fonnate can often be tolerated or are even necessary, and water does not create problems at all. In conclusion, the present study illustrates that RPLC-GC- MS can serve highly useful purposes in areas of applications such as impurity profiling, where qualitative (structural in- formation) rather than quantitative analysis is of primary importance. Further optimization should mainly be directed at improving RPLC-GC interfacing for aqueous mobile phases. The authors like to thank the students W. Salburg and L. Maslam for their experimental work and P. Scherpenisse from the AM1 group at Solvay Duphar for stimulating discussions. References 1 2 3 4 5 6 7 8 9 10 11 Emi, F., J . Chromatogr., 1982, 251, 141. Tomer, K. B., and Parker, C. E., J . Chromatogr., 1989, 492, 189. Qin, X., Ip, D. P., Chang, K. H.-C., Dradransky, P. M., Brooks, M. A., and Sakuma, T., J. Pharm. Biomed. Anal., 1994,12,221. Raglione, J. T. V., Troskosky, A., and Hartwick, R. A., J . Chromatogr., 1987, 409, 213. Ostmann, C., and Nilsson, U., J . High Resolut. Chromatogr., 1992, 15, 745. Vreuls, J. J., de Jong, G. J., and Brinkman, U. A. Th., Chromato- graphia, 1991, 31, 113. Ogorka, J., Schwinger, G., Bruat, G., and Seidel, V., J . Chromatogr., 1992, 626, 87. Wessels, P., Ogorka, J., Schwinger, G., and Ulmer, M., J. High Resolut. Chromatogr., 1993, 16, 708. Goosens, E. C., de Jong, D., van den Berg, J. H. M., de Jong, G. J., and Brinkman, U. A. Th., J . Chromatogr., 1991, 552,489. Goosens, E. C., de Jong, D., de Jong, G. J., and Brinkman, U. A. Th., J . Microcol. Sep., 1994, 6, 207. Goosens, E. C., Beerthuizen, I. M., de Jong, D., de Jong, G. J., and Brinkman, U. A. Th., Chromatographia, 1995,40, 267. Paper Sl05009J Received July 28, 1995 Accepted September 15, I995