- Email: [email protected]
Spectral correlation of high-performance liquid chromatography-diode array detection data from two independent chromatographic runs
Journal of Chromatography A, 1190 (2008) 141–149
Contents lists available at ScienceDirect
Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma
Spectral correlation of high-performance liquid chromatography-diode array detection data from two independent chromatographic runs Peak tracking in pharmaceutical impurity profiling Wei Li a,b , Chang-qin Hu a,∗ a b
National Institute for the Control of Pharmaceutical and Biological Products, Beijing 100050, China Anhui Institute for Drug Control, Hefei 230051, China
a r t i c l e
i n f o
Article history: Received 17 December 2007 Received in revised form 19 February 2008 Accepted 28 February 2008 Available online 6 March 2008 Keywords: Two-dimensional chromatographic spectral correlative map HPLC-DAD Identification Impurities
a b s t r a c t A novel qualitative analytical method for peak tracking in impurity profiling control by the correlation of spectra was established. Two-dimensional (2D) standard spectrochromatographic data produced by high-performance liquid chromatography with diode array detection (HPLC-DAD) were compared with sample data to develop two-dimensional chromatographic spectral correlative maps. Taking full advantage of separation efficiency of HPLC and spectral specificity of the analytes, the method was successfully used to recognize impurities in quinolone antibacterials, when in combination with relative retention times (RRTs). For the comparison of spectra was expanded to three-dimensional space, simultaneous identification of the chromatographic peaks can be obtained rapidly without preparation and injection of a reference solution, even when the mobile phase changed or the peaks of multi-component samples overlapped. © 2008 Elsevier B.V. All rights reserved.
1. Introduction The presence of process impurities and degradation products, even in small amounts in active pharmaceutical ingredients (APIs), may influence the efficacy and safety of the pharmaceutical products. Impurity profiling (i.e., the identity as well as the quantity of an impurity in a pharmaceutical product) is always receiving important critical attention from the regulatory authorities. The International Conference on Harmonization (ICH) has formulated a workable guideline regarding the control of impurities—it was pointed out that an impurity is identified in a drug sample if its content is 0.1% or greater. In pharmacopoeias and ICH guidelines [1–3], high-performance liquid chromatography (HPLC) methods are recommended as the most useful tools for impurity control. Impurity peaks can be identified by correspondence of retention times to peaks of chemical reference standards (CRS). When CRS of the impurities are not available, chromatographic parameters like relative retention times (RRTs) can be used instead. However, identification of each impurity simply by comparison of RRTs in chromatograms is uncertain, because the shift of retention time of particular peaks, due to unavoidable shortcomings such as imperfect reproducibility in chromatography and columns from different
∗ Corresponding author. Tel.: +86 10 67095308; fax: +86 10 65115148. E-mail address: [email protected] (C.-q. Hu). 0021-9673/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.chroma.2008.02.097
sellers despite with the same type of stationary phase, may lead to erroneous assessment of the quality of medical samples in impurity profiling control, especially when these impurities have similar retention times. Advances in computer application techniques and instrumentation for HPLC coupled to diode array detection (DAD) and to mass spectrometry (MS) provide more information and make qualitative analysis accurate and faster. Though HPLC–MS is a powerful tool for identification, the methods based on it are usually more sophisticated so that the laboratories that are not well equipped cannot easily perform them. In contrast, methods based on HPLCDAD such as the maximum absorbance at a certain wavelength [4], multiple absorbance ratio [5], which are not reliable, are employed to identify substances, and it is reported that DAD spectral library [6,7] and spectral correlative chromatography [8] were applied to identification of peaks in chromatograms. A correlation coefficient curve or index can indicate which peak in a chromatogram is likely a known component, yet comparing sample data with only one standard spectrum at a time makes manipulation less efficient and the results appear unclear for two independent chromatograms. Taking this into account, we expanded the comparison of spectra to three-dimensional space and developed two-dimensional chromatographic spectral correlative maps. Different from the conventional spectral matching method mentioned above, in this paper every spectrum data from two independent chromatographic runs were compared with each other. Therefore,
142
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
Fig. 1. The chemical structures of gatifloxacin, moxifloxacin and their related substances.
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
a curved surface was constructed, with the retention times of the two chromatograms as the X and Y axes and the correlation coefficients as the Z axis. From the contour map of the curved surface (viz. two-dimensional chromatographic spectral correlative map), peaks of multi-components in an unknown sample can be recognized simultaneously from the spectral relationship of two chromatograms without preparation and injection of each component solution repeatedly, since the components were once eluted in one chromatogram of a well-known sample. With the method it is easy to find out how many components two samples have in common, and it is also convenient to track peaks in optimization of chromatographic conditions. Furthermore, elution order of each component can be distinguished with one injection, even when peaks overlapped seriously. As an optimal method for separation, HPLC has been used in the analysis of gatifloxacin [9–11] and moxifloxacin [12,13]. The chemical structures of the two quinolone antibacterials and their potential related substances are shown in Fig. 1. Similar retention behaviors of the impurities and changes of chromatographic system make it hard to identify peaks by RRTs in impurity profiling control. As examples, the chromatographic and spectral characteristics were studied in this paper and two-dimensional chromatographic correlation spectroscopy was established as a method for rapid identification of peaks, when combined with RRTs, without recourse to CRS of impurities. 2. Theory Retention time (tR ) or RRT of a given substance in a given HPLC system will not always be identical due to even slight changes in components of the chromatographic system, such as the mobile phase and the column, so that the recognition of chemicals is more reliable using spectral or MS data [14]. In HPLC-DAD systems, suppose that “S” is one spectrum from chromatogram A and “T” is one spectrum from chromatogram B, and the two spectra consist of n absorbance values at different wavelengths. “sk ” is the kth absorbance value of “S” and “tk ” is the kth absorbance value of “T”. “¯s” and “t¯ ” are the average absorbance values of the two spectra, respectively. The correlation coefficients of the two spectra can be calculated according to the formula [15]:
n
RS,T =
n k=1
k=1
(sk − s¯ )(tk − t¯ )
(sk − s¯ )2
n
k=1
(tk − t¯ )
2
1/2
The higher the Rs,t value is, the greater is the similarity between the two spectra. A curved surface can be obtained by comparing every spectrum extracted from chromatogram A with every spectrum from chromatogram B, with the retention times of the two chromatograms as the X and Y axes and the Rs,t value as the Z axis. A contour map is the level surface of Z for a threshold value of Rs,t , viz. a two-dimensional chromatographic spectral correlative map, from which the spectral relationships of the peaks in the two chromatograms are apparent. This enables us to recognize every eluting peak whose spectral information is in accordance with the peaks in a standard chromatogram by injecting a mixed solution once, that is to say, each component is identified simultaneously.
143
Chromatographic columns were: (1) a Shiseido capcell MG C18 (250 mm × 4.6 mm i.d., 5 m particle size), (2) a Sumipax ODS A212 (150 mm × 6 mm i.d., 5 m particle size), and (3) an Alltech Alltima C18 (250 mm × 4.6 mm i.d., 5 m particle size). A Thermo Finnigan HPLC–MS device that consisted of a Surveyor pump, a Surveyor autosampler, and a Surveyor MS detector MSQ was also used. 3.2. Chemicals and reagents Gatifloxacin (GTFX), gatifloxacin decarboxylate (DC-GTFX), 7-ethylenediamine gatifloxacin (EDA-GTFX), 7-(2-methylethylenediamine) gatifloxacin (MEDA-GTFX), 8-hydroxygatifloxacin (HD-GTFX), gatifloxacin N-methylate (NM-GTFX), and 7-(2methylpiperazine) gatifloxacin (MP-GTFX) were obtained from the Sino-American Shanghai Squibb Pharmaceutical (Shanghai, China). 6-methoxygatifloxacin (MO-GTFX) and 8-ethoxygatifloxacin (EOGTFX) were obtained from the Henan Kangtai Pharmaceutical Group (Xingyang, China). 8-Fluorogatifloxacin (F-GTFX), 8-fluoro3-ethoxycarbonylgatifloxacin (FEC- GTFX) were obtained from the Zhejiang Xinchang Pharmaceutical (Xinchang, China) and gatifloxacin tablets from the Jiangsu Suzhong Pharmaceutical (Jiangyan, China). Moxifloxacin (MXFX), 8-hydroxymoxifloxacin (HD-MXFX), 6,8-dimethoxymoxifloxacin (DM-MXFX), 8ethoxymoxifloxacin (EO-MXFX), 6-methoxy-8-fluoromoxifloxacin (MF-MXFX) were obtained from Bayer AG (Leverkusen, Germany). HPLC-grade acetonitrile was purchased from Fisher Scientific (Fair Lawn, NJ, USA) and other chemicals used were of analytical grade. 3.3. Methods 3.3.1. HPLC-DAD conditions For analysis of GTFX, the mobile phase consisted of 1% triethylamine solution (adjusted to pH 4.3 with phosphoric acid)–acetonitrile (87:13, v/v). For analysis of MXFX, the mobile phase 1 consisted of 1% triethylamine solution (adjusted to pH 2.5 with phosphoric acid)–methanol (70:30, v/v); the mobile phase 2 consisted of phosphate buffer (1 g of potassium dihydrogenphosphate and 10 mmol ammonium tetrabutyl sulphate dissolved in 1000 mL of water, adjusted to pH 2.5 with phosphoric acid)–acetonitrile (85:15, v/v). In both analyses, spectra were recorded in the range of 200–500 nm with a resolution of 1 nm. The flow rate was set at 1.0 mL min−1 and a constant column temperature was maintained at 30 ◦ C. The injected volume was 10 L. 3.3.2. LC–MS condition A mixture of water/acetonitrile 50:50 (v/v) as the mobile phase was delivered at a speed of 0.6 mL min−1 . The MS was operated in the positive ion mode with a cone voltage of 120 V, the electrospray voltage was 4.5 kV, and the mass range m/z was 100–600.
3.1. Instruments
3.3.3. Solution preparation GTFX (reference substance for the calculation of the RRTs) and the 10 related substances were dissolved in the mobile phase to produce a solution containing 0.1 mg mL−1 of each component. A quantity of GTFX tablets sample was ground and weighed and the mobile phase was added to produce a solution of 5 mg mL−1 . MXFX and its 4 related substances were dissolved in mobile phase to produce a solution containing 0.1 mg mL−1 of each component.
HPLC-DAD devices were: (1) a Shimadzu HPLC device that consisted of a LC-10AT pump, an SIL-10AVP autosampler, and a SPDM10AVP DAD system; (2) a Waters HPLC device that consisted of a 2690 separation module and a PDA996 DAD system.
3.3.4. Data analysis The DAD data were converted to ASCII files and transferred to an IBM personal computer. All programs for data analysis and plotting were written and compiled in this laboratory using MATLAB 6.5.
3. Materials and methods
144
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
Table 1 RRTs of GTFX and its 10 related substances with different HPLC columns Brand of column
RRT
Shiseido Alltima Sumipax
DC-GTFX
EDA-GTFX
F-GTFX
MEDA-GTFX
HD-GTFX
GTFX
FEC-GTFX
NM-GTFX
MP-GTFX
MO-GTFX
EO-GTFX
0.29 0.31 0.31
0.45 0.43 0.45
0.67 0.66 0.67
0.72 0.70 0.73
0.78 0.76 0.77
1.00 1.00 1.00
1.10 1.14 1.10
1.15 1.19 1.16
1.21 1.21 1.20
1.30 1.31 1.28
1.56 1.60 1.56
Fig. 2. Chromatogram of GTFX and its 10 related substances. (1) DC-GTFX; (2) EDAGTFX; (3) F-GTFX; (4) MEDA-GTFX; (5) HD-GTFX; (6) GTFX; (7) FEC-GTFX; (8) NMGTFX; (9) MP-GTFX; (10) MO-GTFX; (11) EO-GTFX.
The standard spectrochromatographic data exported from three-dimensional (3D) chromatograms of the 10 related substances were saved to be compared with sample spectrochromatographic data for identification. For two identical spectra, Rs,t equals 1. Nevertheless, in practice the spectra are modified to a small extent by noise. Therefore, it is necessary to define a threshold value of Rs,t above which two spectra are regarded as identical. As can be seen from the correlation coefficients calculated between the spectra of GTFX and its 10 related substances (Table 2), the 11 compounds had their own spectral characteristics. Except that the spectra of EDA-GTFX and MEDA-GTFX were similar and the spectra of GTFX, NM-GTFX and EO-GTFX were likewise similar, the spectra of F-GTFX and FEC-GTFX were the closest (correlation coefficient 0.982731). So the threshold of Rs,t was set to 0.995 in order to discriminate the substances. Since spectral identification of the peaks was performed in each cluster divided by RRTs, it was not affected by the spectral similarities of the chemicals mentioned above because in each spectral region there were no correlation coefficients of substances that exceeded the threshold. 4.2. Investigation on influencing factors of the method
4. Results and discussion 4.1. Establishment of the identification method An optimal chromatogram of the solution mixture consisting of GTFX and the 10 related substances is shown in Fig. 2. The RRT of each substance, using three different brands of C18 column, is listed in Table 1. The separations of the DC-GTFX, EDA-GTFX, and EO-GTFX peaks from those of the other eluates were satisfying and the RRTs of the three compounds could be used for identification. Peaks from F-GTFX, MEDA-GTFX, and HD-GTFX and those from FECGTFX, NM-GTFX, MP-GTFX, and MO-GTFX were in clusters, so these analytes could hardly be recognized from the RRTs when chromatographic conditions changed or when the elution order was useless because only some of the impurities were present in the sample. Therefore, spectral information was an important aid for the identification. In the chromatographic system, the peaks of the impurities could be divided into four clusters according to RRT: RRT < 0.55, 0.55 < RRT < 0.9, 1 < RRT < 1.45, 1.45 < RRT. Identification of the peaks in each cluster was performed with the aid of spectral information.
4.2.1. Applicable concentration range Samples of 2, 1, 0.5, 0.2, 0.1, 0.05 and 0.025 g of the 10 substances related to GTFX were measured by HPLC-DAD. The correlation coefficients acquired ranged from 0.999983 to 0.966388. So long as the injected amount was above 0.1 g, each analyte could be identified with a correlation coefficient >0.995 (threshold of the Rs,t value). When the amounts of the impurities in the samples were very low, the injection volumes were increased for qualitative analysis first, and then quantitative analysis followed with the normal injection volume. 4.2.2. Effect of pH When the pH values of the triethylamine solution in the mobile phases were adjusted to 3.00, 4.05, 4.44 and 5.00, the correlation coefficients between spectra of the analytes and the standard spectral data ranged from 0.997258 to 0.999944. Since quinolones like GTFX are zwitterions with isoelectric points at pH 7.7 [10], GTFX and its related substances are mostly in the form of cations in the more acidic mobile phases, so the spectra were relatively stable.
Table 2 Correlation coefficients calculated for GTFX and its related substances Substances
DC-GTFX EDA-GTFX F-GTFX MEDA-GTFX HD-GTFX GTFX FEC-GTFX NM-GTFX MP-GTFX MO-GTFX EO-GTFX
Correlation coefficient (R) DC-GTFX
EDA-GTFX
F-GTFX
MEDA-GTFX
HD-GTFX
GTFX
FEC-GTFX
NM-GTFX
MP-GTFX
MO-GTFX
EO-GTFX
1 – – – – – – – – – –
0.518492 1 – – – – – – – – –
0.530916 0.951126 1 – – – – – – – –
0.512996 0.999921 0.953520 1 – – – – – – –
0.797517 0.432514 0.519268 0.430011 1 – – – – – –
0.393811 0.846607 0.956317 0.851837 0.48238 1 – – – – –
0.611522 0.948549 0.982731 0.94892 0.585845 0.903799 1 – – – –
0.413890 0.865245 0.966497 0.870126 0.494723 0.999095 0.918182 1 – – –
0.569833 0.737422 0.874560 0.740431 0.783925 0.907056 0.870795 0.909366 1
0.326260 0.750314 0.868333 0.755440 0.538288 0.954123 0.819657 0.948019 0.914575 1 –
0.357569 0.815527 0.936622 0.821186 0.467248 0.997825 0.879773 0.994451 0.903990 0.964688 1
–
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
4.2.3. Instrumental effects Since a Shimadzu SPDM10AVP DAD system was used for construction of the standard spectrochromatographic data, a Waters PDA996 DAD system was applied to study the spectral differences
145
generated by different instruments. The correlation coefficients acquired ranged from 0.999165 to 0.999997 (>0.995), which meant the instruments did not affect the accuracy of the identifications.
Fig. 3. Flow diagram of the method for identifying related substances in GTFX. (a) Three-dimensional chromatogram of the GTFX sample containing four impurities (A–D). (b) Three-dimensional chromatogram of the 7 impurities of GTFX: (1) F-GTFX; (2) MEDA-GTFX; (3) HD-GTFX; (4) FEC-GTFX; (5) NM-GTFX; (6) MP-GTFX; (7) MO-GTFX. (c) Three-dimensional spectral correlative map constructed, when every spectrum in (a) was compared with every spectrum in (b). (d) Contour map of (c) for a threshold value of R (0.995). Impurities A and B: unknown substances; impurity C: HD-GTFX; impurity D: NM-GTFX.
146
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
4.2.4. Precision and stability of correlation coefficients calculated with spectra The precision was validated by calculating the similarity of spectra obtained by the analysis of five replicate injections of
each analyte, individually. The correlation coefficients ranged from 0.999955 to 0.999997. The stability was monitored by measuring the solutions of each compound over the periods of 1, 4, 8, and 12 h.
Fig. 4. Dependence of the 2D chromatographic spectral correlative maps on the pH value of the mobile phases in analysis of GTFX and its impurities. (a) Two-dimensional chromatographic spectral correlative map obtained when pH value of buffer in mobile phase equaled to 4.01. (b) Two-dimensional chromatographic spectral correlative map obtained when pH value of buffer in mobile phase equaled to 4.50. (c) Two-dimensional chromatographic spectral correlative map obtained when pH value of buffer in mobile phase equaled to 4.60. (1 and A) F-GTFX; (2 and B) MEDA-GTFX; (3 and C) HD-GTFX; (4 and D) FEC-GTFX; (5 and E) NM-GTFX; (6 and F) MP-GTFX; (7 and G) MO-GTFX.
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
The correlation coefficients acquired ranged from 0.999848 to 0.999998. 4.3. Application of the method 4.3.1. Simultaneous identification of multi-components The GTFX sample solution was analyzed and a 3D chromatogram was recorded as Fig. 3a. According to RRTs, the impurities fell in two regions: 0.55 < RRT < 0.9 and 1 < RRT < 1.45. Every spectrum in the data exported from the 3D chromatogram and the standard spectrochromatographic data exported from the two regions in the standard 3D chromatogram shown in Fig. 3b were compared to construct a 3D spectral correlative map of the two chromatograms, which is a curved surface in Fig. 3c, with the retention times of the two chromatograms as the X and Y axes and the R value as the Z axis. A contour map as Fig. 3d is the level surface for a threshold value of R (0.995), viz. a two-dimensional chromatographic spectral correlative map. It is apparent from the map that peak A and B are unknown impurities, peak C corresponds to peak 3 (HD-GTFX), peak D corresponds to peak 5 (NM-GTFX). The two impurities were identified simultaneously, regardless of the existence of unknown peaks. The quasi-molecular ion peaks of C and D were m/z 362 and 390, respectively, by ESI-MS after solutions of the impurities were collected and injected. It validated the results of the identification. 4.3.2. Peak tracing in chromatographic system variation A mixed solution of GTFX and its related substances was analyzed when the pH values of the triethylamine solution in the mobile phase were adjusted to 4.01, 4.30, 4.50, 4.60, and 4.73, sep-
147
arately. The spectrochromatographic data acquired were exported and compared with standard data exported from Fig. 3b, to construct 2D spectral correlative maps shown in Fig. 4. It can be found clearly that the retention behavior of FEC-GTFX, NM-GTFX, MPGTFX, and MO-GTFX was sensitive to the pH value of the mobile phase. With the increase of the pH value, there was a sharp change in retention behavior of NM-GTFX. When the pH value was above 4.60, the whole elution order changed and the retention time of NM-GTFX was longer than that of MP-GTFX. Although some of the impurity peaks overlapped (or even seemed like one peak) most of the time, all the impurity peaks could be traced. The mixed solution of MXFX and its related substances was analyzed to study the efficiency of the method when completely different mobile phases were used. In the upper region of Fig. 5, the mixed solution was analyzed when using mobile phase 1, the chromatogram in the left region of Fig. 5 obtained when using mobile phase 2. It can be found clearly that the elution orders in the two chromatographic systems were totally different. It means that even the mobile phase was changed in the analysis, each component can be easily recognized. Both the examples indicated that the method was powerful in peak tracing in chromatographic system variation, and it would be more efficient in impurity profiling control in pharmaceuticals. 4.3.3. Identification of unresolved peaks Identification of the unresolved peaks is a challenging task for analysts. In the use of DAD spectral library [6,7], when peaks overlap, searching a standard spectra library could produce varying qualitative results due to the extraction of sample spectra with
Fig. 5. 2D chromatographic spectral correlative map of MXFX and its related substances analyzed in two kinds of mobile phases. (1 and A) EO-MXFX; (2 and B) MF-MXFX; (3 and C) DM-MXFX; (4 and D) MXFX; (5 and E) HD-MXFX.
148
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
the peaks at different positions in the chromatogram. In the use of spectral correlative chromatography [8], one chromatogram is only compared with one standard spectrum to find the highest points of every curve, so the process has to be performed many times for identification of all peaks in the chromatogram. However, with two-dimensional chromatographic spectral correlative maps in this paper, all unresolved components and their elution order could be recognized in a new chromatographic system, with only a single run. Fixed size moving window evolving factor analysis (FSMWEFA) [16] is a powerful technique for determining the number of components in a chromatogram. The PCA (Principal Component Analysis) was performed on a fixed number of spectra (a fixed size window) and logarithms of the eigenvalues were plotted against the reten-
tion times to form a rank map. The number of curves above the noise level in the rank map was the number of components. FSMWEFA was performed on the spectrochromatographic data exported from the chromatogram in Fig. 4C (retention time 53.8–57.2 min), and Fig. 6a shows the rank map. The 2D chromatographic spectral correlative map in Fig. 4c enlarged (retention time 53.8–57.2 min) was shown in Fig. 6b. The local rank 2 from 54.9 to 56.3 min meant there were two components co-eluting in this region. Only one component eluted in each of the regions before 54.9 min and after 56.3 min. The result of FSMWEFA was in accordance with the 2D chromatographic spectral correlative map. It can be seen that 2D chromatographic correlation spectroscopy do not require good resolution of peaks. With the separation ability of HPLC, most of the peaks have their “pure eluting area” for identifica-
Fig. 6. The FSMWEFA rank map of the chromatogram in Fig. 4c (53.8–57.2 min) and the two-dimensional chromatographic spectral correlative map. (a) The rank map obtained after FSMWEFA was performed on the data exported from the chromatogram in Fig. 4c (retention time 53.8–57.2 min). (b) The two-dimensional chromatographic spectral correlative map in Fig. 4c enlarged.
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
tion. This gave us a hint that the method can test peak purity also. If one peak in sample chromatogram is overlapped with other peaks, the retention times corresponding to the spot in the 2D spectral correlative map would not cover the whole peak range.
149
the solvent exceed the limit, therefore robustness of the method need to be further studied when it is used in the analysis of certain substances. References
5. Conclusion The method based on HPLC-DAD system has an important advantage over HPLC–MS, that is more kind of mobile phases can be chosen to meet the requirement of separation. Making the most of the separation ability of HPLC and spectral specificity, 2D chromatographic correlation spectroscopy is a powerful method for the identification and is successfully applied to a recognition system for peak tracking in impurity profiling control in quinolones. The main advantages of the method are rapidly and simultaneously identifying multi-components without reference substances, accurately distinguishing the elution orders of the unresolved peaks, and testing the peak purity at the same time. However, the method has deficiencies; for example, if there is a small peak completely embedded in a big sample peak, the left and right wings of the big peak can be identified (showing two spots in the map, like a big spot being cut into halves), but the small peak cannot be identified in the map, because it has no “pure elution area”. Sometimes spectra could be misshaped because changes in the pH value of
[1] Q3A(R2): Impurities in New Drug Substances, International Conference on Harmonization of Technical Requirements for the Registration of Pharmaceuticals for Human Use, Geneva, 2006, http://www.ich.org. [2] European Pharmacopoeia, vol. 1, 5th ed., Council of Europe, Strasbourg, 2004, p. 587. [3] The United States Pharmacopoeia, US Pharmacopeial Convention, 27th ed., Rockville, MD, 2004, p. 2510. [4] P.D. Fraser, M.E.S. Pinto, D.E. Holloway, P.M. Bramley, Plant J. 24 (2000) 551. [5] A.C.J.H. Drouen, H.A.H. Billiet, L.D. Galan, Anal. Chem. 56 (1984) 971. [6] M. Herzler, S. Herre, F. Pragst, J. Anal. Toxicol. 27 (2003) 233. [7] F. Pragst, M. Herzler, B.T. Erxleben, Clin. Chem. Lab. Med. 42 (2004) 1325. [8] B.Y. Li, Y. Hu, Y.Z. Liang, L.F. Huang, C.J. Xu, P.S. Xie, J. Sep. Sci. 27 (2004) 581. [9] K. Vishwanathan, M.G. Bartlett, J.T. Stewart, Rapid Commun. Mass Spectrom. 15 (2001) 915. ¨ [10] C.K. Naber, M. Steghafner, M. Kinzig-Schippers, C. Sauber, F. Sorgel, H.J. Stahlberg, K.G. Naber, Antimicrob. Agents Chemother. 45 (2001) 293. [11] N. Sultana, M. Arayne, A. Naz, Pak. J. Pharm. Sci. 19 (2006) 275. [12] Y.R. Kumar, V.V.N.K.V.P. Raju, R.R. Kumar, S. Eswaraiah, K. Mukkanti, M.V. Suryanarayana, M.S. Reddy, J. Pharm. Biomed. Anal. 34 (2004) 1125. [13] J. Sunderland, C.M. Tobin, A.J. Hedges, A.P. Macgowan, L.O. White, J. Antimicrob. Chemother. 47 (2001) 271. [14] Y.G. Li, F. Zhang, Z.T. Wang, Z.B. Hu, J. Pharm. Biomed. Anal. 35 (2004) 1101. [15] Analytical Methods Committee, Analyst 113 (1988) 1469. [16] H.R. Keller, D.L. Massart, Anal. Chim. Acta 246 (1991) 379.
Contents lists available at ScienceDirect
Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma
Spectral correlation of high-performance liquid chromatography-diode array detection data from two independent chromatographic runs Peak tracking in pharmaceutical impurity profiling Wei Li a,b , Chang-qin Hu a,∗ a b
National Institute for the Control of Pharmaceutical and Biological Products, Beijing 100050, China Anhui Institute for Drug Control, Hefei 230051, China
a r t i c l e
i n f o
Article history: Received 17 December 2007 Received in revised form 19 February 2008 Accepted 28 February 2008 Available online 6 March 2008 Keywords: Two-dimensional chromatographic spectral correlative map HPLC-DAD Identification Impurities
a b s t r a c t A novel qualitative analytical method for peak tracking in impurity profiling control by the correlation of spectra was established. Two-dimensional (2D) standard spectrochromatographic data produced by high-performance liquid chromatography with diode array detection (HPLC-DAD) were compared with sample data to develop two-dimensional chromatographic spectral correlative maps. Taking full advantage of separation efficiency of HPLC and spectral specificity of the analytes, the method was successfully used to recognize impurities in quinolone antibacterials, when in combination with relative retention times (RRTs). For the comparison of spectra was expanded to three-dimensional space, simultaneous identification of the chromatographic peaks can be obtained rapidly without preparation and injection of a reference solution, even when the mobile phase changed or the peaks of multi-component samples overlapped. © 2008 Elsevier B.V. All rights reserved.
1. Introduction The presence of process impurities and degradation products, even in small amounts in active pharmaceutical ingredients (APIs), may influence the efficacy and safety of the pharmaceutical products. Impurity profiling (i.e., the identity as well as the quantity of an impurity in a pharmaceutical product) is always receiving important critical attention from the regulatory authorities. The International Conference on Harmonization (ICH) has formulated a workable guideline regarding the control of impurities—it was pointed out that an impurity is identified in a drug sample if its content is 0.1% or greater. In pharmacopoeias and ICH guidelines [1–3], high-performance liquid chromatography (HPLC) methods are recommended as the most useful tools for impurity control. Impurity peaks can be identified by correspondence of retention times to peaks of chemical reference standards (CRS). When CRS of the impurities are not available, chromatographic parameters like relative retention times (RRTs) can be used instead. However, identification of each impurity simply by comparison of RRTs in chromatograms is uncertain, because the shift of retention time of particular peaks, due to unavoidable shortcomings such as imperfect reproducibility in chromatography and columns from different
∗ Corresponding author. Tel.: +86 10 67095308; fax: +86 10 65115148. E-mail address: [email protected] (C.-q. Hu). 0021-9673/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.chroma.2008.02.097
sellers despite with the same type of stationary phase, may lead to erroneous assessment of the quality of medical samples in impurity profiling control, especially when these impurities have similar retention times. Advances in computer application techniques and instrumentation for HPLC coupled to diode array detection (DAD) and to mass spectrometry (MS) provide more information and make qualitative analysis accurate and faster. Though HPLC–MS is a powerful tool for identification, the methods based on it are usually more sophisticated so that the laboratories that are not well equipped cannot easily perform them. In contrast, methods based on HPLCDAD such as the maximum absorbance at a certain wavelength [4], multiple absorbance ratio [5], which are not reliable, are employed to identify substances, and it is reported that DAD spectral library [6,7] and spectral correlative chromatography [8] were applied to identification of peaks in chromatograms. A correlation coefficient curve or index can indicate which peak in a chromatogram is likely a known component, yet comparing sample data with only one standard spectrum at a time makes manipulation less efficient and the results appear unclear for two independent chromatograms. Taking this into account, we expanded the comparison of spectra to three-dimensional space and developed two-dimensional chromatographic spectral correlative maps. Different from the conventional spectral matching method mentioned above, in this paper every spectrum data from two independent chromatographic runs were compared with each other. Therefore,
142
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
Fig. 1. The chemical structures of gatifloxacin, moxifloxacin and their related substances.
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
a curved surface was constructed, with the retention times of the two chromatograms as the X and Y axes and the correlation coefficients as the Z axis. From the contour map of the curved surface (viz. two-dimensional chromatographic spectral correlative map), peaks of multi-components in an unknown sample can be recognized simultaneously from the spectral relationship of two chromatograms without preparation and injection of each component solution repeatedly, since the components were once eluted in one chromatogram of a well-known sample. With the method it is easy to find out how many components two samples have in common, and it is also convenient to track peaks in optimization of chromatographic conditions. Furthermore, elution order of each component can be distinguished with one injection, even when peaks overlapped seriously. As an optimal method for separation, HPLC has been used in the analysis of gatifloxacin [9–11] and moxifloxacin [12,13]. The chemical structures of the two quinolone antibacterials and their potential related substances are shown in Fig. 1. Similar retention behaviors of the impurities and changes of chromatographic system make it hard to identify peaks by RRTs in impurity profiling control. As examples, the chromatographic and spectral characteristics were studied in this paper and two-dimensional chromatographic correlation spectroscopy was established as a method for rapid identification of peaks, when combined with RRTs, without recourse to CRS of impurities. 2. Theory Retention time (tR ) or RRT of a given substance in a given HPLC system will not always be identical due to even slight changes in components of the chromatographic system, such as the mobile phase and the column, so that the recognition of chemicals is more reliable using spectral or MS data [14]. In HPLC-DAD systems, suppose that “S” is one spectrum from chromatogram A and “T” is one spectrum from chromatogram B, and the two spectra consist of n absorbance values at different wavelengths. “sk ” is the kth absorbance value of “S” and “tk ” is the kth absorbance value of “T”. “¯s” and “t¯ ” are the average absorbance values of the two spectra, respectively. The correlation coefficients of the two spectra can be calculated according to the formula [15]:
n
RS,T =
n k=1
k=1
(sk − s¯ )(tk − t¯ )
(sk − s¯ )2
n
k=1
(tk − t¯ )
2
1/2
The higher the Rs,t value is, the greater is the similarity between the two spectra. A curved surface can be obtained by comparing every spectrum extracted from chromatogram A with every spectrum from chromatogram B, with the retention times of the two chromatograms as the X and Y axes and the Rs,t value as the Z axis. A contour map is the level surface of Z for a threshold value of Rs,t , viz. a two-dimensional chromatographic spectral correlative map, from which the spectral relationships of the peaks in the two chromatograms are apparent. This enables us to recognize every eluting peak whose spectral information is in accordance with the peaks in a standard chromatogram by injecting a mixed solution once, that is to say, each component is identified simultaneously.
143
Chromatographic columns were: (1) a Shiseido capcell MG C18 (250 mm × 4.6 mm i.d., 5 m particle size), (2) a Sumipax ODS A212 (150 mm × 6 mm i.d., 5 m particle size), and (3) an Alltech Alltima C18 (250 mm × 4.6 mm i.d., 5 m particle size). A Thermo Finnigan HPLC–MS device that consisted of a Surveyor pump, a Surveyor autosampler, and a Surveyor MS detector MSQ was also used. 3.2. Chemicals and reagents Gatifloxacin (GTFX), gatifloxacin decarboxylate (DC-GTFX), 7-ethylenediamine gatifloxacin (EDA-GTFX), 7-(2-methylethylenediamine) gatifloxacin (MEDA-GTFX), 8-hydroxygatifloxacin (HD-GTFX), gatifloxacin N-methylate (NM-GTFX), and 7-(2methylpiperazine) gatifloxacin (MP-GTFX) were obtained from the Sino-American Shanghai Squibb Pharmaceutical (Shanghai, China). 6-methoxygatifloxacin (MO-GTFX) and 8-ethoxygatifloxacin (EOGTFX) were obtained from the Henan Kangtai Pharmaceutical Group (Xingyang, China). 8-Fluorogatifloxacin (F-GTFX), 8-fluoro3-ethoxycarbonylgatifloxacin (FEC- GTFX) were obtained from the Zhejiang Xinchang Pharmaceutical (Xinchang, China) and gatifloxacin tablets from the Jiangsu Suzhong Pharmaceutical (Jiangyan, China). Moxifloxacin (MXFX), 8-hydroxymoxifloxacin (HD-MXFX), 6,8-dimethoxymoxifloxacin (DM-MXFX), 8ethoxymoxifloxacin (EO-MXFX), 6-methoxy-8-fluoromoxifloxacin (MF-MXFX) were obtained from Bayer AG (Leverkusen, Germany). HPLC-grade acetonitrile was purchased from Fisher Scientific (Fair Lawn, NJ, USA) and other chemicals used were of analytical grade. 3.3. Methods 3.3.1. HPLC-DAD conditions For analysis of GTFX, the mobile phase consisted of 1% triethylamine solution (adjusted to pH 4.3 with phosphoric acid)–acetonitrile (87:13, v/v). For analysis of MXFX, the mobile phase 1 consisted of 1% triethylamine solution (adjusted to pH 2.5 with phosphoric acid)–methanol (70:30, v/v); the mobile phase 2 consisted of phosphate buffer (1 g of potassium dihydrogenphosphate and 10 mmol ammonium tetrabutyl sulphate dissolved in 1000 mL of water, adjusted to pH 2.5 with phosphoric acid)–acetonitrile (85:15, v/v). In both analyses, spectra were recorded in the range of 200–500 nm with a resolution of 1 nm. The flow rate was set at 1.0 mL min−1 and a constant column temperature was maintained at 30 ◦ C. The injected volume was 10 L. 3.3.2. LC–MS condition A mixture of water/acetonitrile 50:50 (v/v) as the mobile phase was delivered at a speed of 0.6 mL min−1 . The MS was operated in the positive ion mode with a cone voltage of 120 V, the electrospray voltage was 4.5 kV, and the mass range m/z was 100–600.
3.1. Instruments
3.3.3. Solution preparation GTFX (reference substance for the calculation of the RRTs) and the 10 related substances were dissolved in the mobile phase to produce a solution containing 0.1 mg mL−1 of each component. A quantity of GTFX tablets sample was ground and weighed and the mobile phase was added to produce a solution of 5 mg mL−1 . MXFX and its 4 related substances were dissolved in mobile phase to produce a solution containing 0.1 mg mL−1 of each component.
HPLC-DAD devices were: (1) a Shimadzu HPLC device that consisted of a LC-10AT pump, an SIL-10AVP autosampler, and a SPDM10AVP DAD system; (2) a Waters HPLC device that consisted of a 2690 separation module and a PDA996 DAD system.
3.3.4. Data analysis The DAD data were converted to ASCII files and transferred to an IBM personal computer. All programs for data analysis and plotting were written and compiled in this laboratory using MATLAB 6.5.
3. Materials and methods
144
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
Table 1 RRTs of GTFX and its 10 related substances with different HPLC columns Brand of column
RRT
Shiseido Alltima Sumipax
DC-GTFX
EDA-GTFX
F-GTFX
MEDA-GTFX
HD-GTFX
GTFX
FEC-GTFX
NM-GTFX
MP-GTFX
MO-GTFX
EO-GTFX
0.29 0.31 0.31
0.45 0.43 0.45
0.67 0.66 0.67
0.72 0.70 0.73
0.78 0.76 0.77
1.00 1.00 1.00
1.10 1.14 1.10
1.15 1.19 1.16
1.21 1.21 1.20
1.30 1.31 1.28
1.56 1.60 1.56
Fig. 2. Chromatogram of GTFX and its 10 related substances. (1) DC-GTFX; (2) EDAGTFX; (3) F-GTFX; (4) MEDA-GTFX; (5) HD-GTFX; (6) GTFX; (7) FEC-GTFX; (8) NMGTFX; (9) MP-GTFX; (10) MO-GTFX; (11) EO-GTFX.
The standard spectrochromatographic data exported from three-dimensional (3D) chromatograms of the 10 related substances were saved to be compared with sample spectrochromatographic data for identification. For two identical spectra, Rs,t equals 1. Nevertheless, in practice the spectra are modified to a small extent by noise. Therefore, it is necessary to define a threshold value of Rs,t above which two spectra are regarded as identical. As can be seen from the correlation coefficients calculated between the spectra of GTFX and its 10 related substances (Table 2), the 11 compounds had their own spectral characteristics. Except that the spectra of EDA-GTFX and MEDA-GTFX were similar and the spectra of GTFX, NM-GTFX and EO-GTFX were likewise similar, the spectra of F-GTFX and FEC-GTFX were the closest (correlation coefficient 0.982731). So the threshold of Rs,t was set to 0.995 in order to discriminate the substances. Since spectral identification of the peaks was performed in each cluster divided by RRTs, it was not affected by the spectral similarities of the chemicals mentioned above because in each spectral region there were no correlation coefficients of substances that exceeded the threshold. 4.2. Investigation on influencing factors of the method
4. Results and discussion 4.1. Establishment of the identification method An optimal chromatogram of the solution mixture consisting of GTFX and the 10 related substances is shown in Fig. 2. The RRT of each substance, using three different brands of C18 column, is listed in Table 1. The separations of the DC-GTFX, EDA-GTFX, and EO-GTFX peaks from those of the other eluates were satisfying and the RRTs of the three compounds could be used for identification. Peaks from F-GTFX, MEDA-GTFX, and HD-GTFX and those from FECGTFX, NM-GTFX, MP-GTFX, and MO-GTFX were in clusters, so these analytes could hardly be recognized from the RRTs when chromatographic conditions changed or when the elution order was useless because only some of the impurities were present in the sample. Therefore, spectral information was an important aid for the identification. In the chromatographic system, the peaks of the impurities could be divided into four clusters according to RRT: RRT < 0.55, 0.55 < RRT < 0.9, 1 < RRT < 1.45, 1.45 < RRT. Identification of the peaks in each cluster was performed with the aid of spectral information.
4.2.1. Applicable concentration range Samples of 2, 1, 0.5, 0.2, 0.1, 0.05 and 0.025 g of the 10 substances related to GTFX were measured by HPLC-DAD. The correlation coefficients acquired ranged from 0.999983 to 0.966388. So long as the injected amount was above 0.1 g, each analyte could be identified with a correlation coefficient >0.995 (threshold of the Rs,t value). When the amounts of the impurities in the samples were very low, the injection volumes were increased for qualitative analysis first, and then quantitative analysis followed with the normal injection volume. 4.2.2. Effect of pH When the pH values of the triethylamine solution in the mobile phases were adjusted to 3.00, 4.05, 4.44 and 5.00, the correlation coefficients between spectra of the analytes and the standard spectral data ranged from 0.997258 to 0.999944. Since quinolones like GTFX are zwitterions with isoelectric points at pH 7.7 [10], GTFX and its related substances are mostly in the form of cations in the more acidic mobile phases, so the spectra were relatively stable.
Table 2 Correlation coefficients calculated for GTFX and its related substances Substances
DC-GTFX EDA-GTFX F-GTFX MEDA-GTFX HD-GTFX GTFX FEC-GTFX NM-GTFX MP-GTFX MO-GTFX EO-GTFX
Correlation coefficient (R) DC-GTFX
EDA-GTFX
F-GTFX
MEDA-GTFX
HD-GTFX
GTFX
FEC-GTFX
NM-GTFX
MP-GTFX
MO-GTFX
EO-GTFX
1 – – – – – – – – – –
0.518492 1 – – – – – – – – –
0.530916 0.951126 1 – – – – – – – –
0.512996 0.999921 0.953520 1 – – – – – – –
0.797517 0.432514 0.519268 0.430011 1 – – – – – –
0.393811 0.846607 0.956317 0.851837 0.48238 1 – – – – –
0.611522 0.948549 0.982731 0.94892 0.585845 0.903799 1 – – – –
0.413890 0.865245 0.966497 0.870126 0.494723 0.999095 0.918182 1 – – –
0.569833 0.737422 0.874560 0.740431 0.783925 0.907056 0.870795 0.909366 1
0.326260 0.750314 0.868333 0.755440 0.538288 0.954123 0.819657 0.948019 0.914575 1 –
0.357569 0.815527 0.936622 0.821186 0.467248 0.997825 0.879773 0.994451 0.903990 0.964688 1
–
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
4.2.3. Instrumental effects Since a Shimadzu SPDM10AVP DAD system was used for construction of the standard spectrochromatographic data, a Waters PDA996 DAD system was applied to study the spectral differences
145
generated by different instruments. The correlation coefficients acquired ranged from 0.999165 to 0.999997 (>0.995), which meant the instruments did not affect the accuracy of the identifications.
Fig. 3. Flow diagram of the method for identifying related substances in GTFX. (a) Three-dimensional chromatogram of the GTFX sample containing four impurities (A–D). (b) Three-dimensional chromatogram of the 7 impurities of GTFX: (1) F-GTFX; (2) MEDA-GTFX; (3) HD-GTFX; (4) FEC-GTFX; (5) NM-GTFX; (6) MP-GTFX; (7) MO-GTFX. (c) Three-dimensional spectral correlative map constructed, when every spectrum in (a) was compared with every spectrum in (b). (d) Contour map of (c) for a threshold value of R (0.995). Impurities A and B: unknown substances; impurity C: HD-GTFX; impurity D: NM-GTFX.
146
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
4.2.4. Precision and stability of correlation coefficients calculated with spectra The precision was validated by calculating the similarity of spectra obtained by the analysis of five replicate injections of
each analyte, individually. The correlation coefficients ranged from 0.999955 to 0.999997. The stability was monitored by measuring the solutions of each compound over the periods of 1, 4, 8, and 12 h.
Fig. 4. Dependence of the 2D chromatographic spectral correlative maps on the pH value of the mobile phases in analysis of GTFX and its impurities. (a) Two-dimensional chromatographic spectral correlative map obtained when pH value of buffer in mobile phase equaled to 4.01. (b) Two-dimensional chromatographic spectral correlative map obtained when pH value of buffer in mobile phase equaled to 4.50. (c) Two-dimensional chromatographic spectral correlative map obtained when pH value of buffer in mobile phase equaled to 4.60. (1 and A) F-GTFX; (2 and B) MEDA-GTFX; (3 and C) HD-GTFX; (4 and D) FEC-GTFX; (5 and E) NM-GTFX; (6 and F) MP-GTFX; (7 and G) MO-GTFX.
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
The correlation coefficients acquired ranged from 0.999848 to 0.999998. 4.3. Application of the method 4.3.1. Simultaneous identification of multi-components The GTFX sample solution was analyzed and a 3D chromatogram was recorded as Fig. 3a. According to RRTs, the impurities fell in two regions: 0.55 < RRT < 0.9 and 1 < RRT < 1.45. Every spectrum in the data exported from the 3D chromatogram and the standard spectrochromatographic data exported from the two regions in the standard 3D chromatogram shown in Fig. 3b were compared to construct a 3D spectral correlative map of the two chromatograms, which is a curved surface in Fig. 3c, with the retention times of the two chromatograms as the X and Y axes and the R value as the Z axis. A contour map as Fig. 3d is the level surface for a threshold value of R (0.995), viz. a two-dimensional chromatographic spectral correlative map. It is apparent from the map that peak A and B are unknown impurities, peak C corresponds to peak 3 (HD-GTFX), peak D corresponds to peak 5 (NM-GTFX). The two impurities were identified simultaneously, regardless of the existence of unknown peaks. The quasi-molecular ion peaks of C and D were m/z 362 and 390, respectively, by ESI-MS after solutions of the impurities were collected and injected. It validated the results of the identification. 4.3.2. Peak tracing in chromatographic system variation A mixed solution of GTFX and its related substances was analyzed when the pH values of the triethylamine solution in the mobile phase were adjusted to 4.01, 4.30, 4.50, 4.60, and 4.73, sep-
147
arately. The spectrochromatographic data acquired were exported and compared with standard data exported from Fig. 3b, to construct 2D spectral correlative maps shown in Fig. 4. It can be found clearly that the retention behavior of FEC-GTFX, NM-GTFX, MPGTFX, and MO-GTFX was sensitive to the pH value of the mobile phase. With the increase of the pH value, there was a sharp change in retention behavior of NM-GTFX. When the pH value was above 4.60, the whole elution order changed and the retention time of NM-GTFX was longer than that of MP-GTFX. Although some of the impurity peaks overlapped (or even seemed like one peak) most of the time, all the impurity peaks could be traced. The mixed solution of MXFX and its related substances was analyzed to study the efficiency of the method when completely different mobile phases were used. In the upper region of Fig. 5, the mixed solution was analyzed when using mobile phase 1, the chromatogram in the left region of Fig. 5 obtained when using mobile phase 2. It can be found clearly that the elution orders in the two chromatographic systems were totally different. It means that even the mobile phase was changed in the analysis, each component can be easily recognized. Both the examples indicated that the method was powerful in peak tracing in chromatographic system variation, and it would be more efficient in impurity profiling control in pharmaceuticals. 4.3.3. Identification of unresolved peaks Identification of the unresolved peaks is a challenging task for analysts. In the use of DAD spectral library [6,7], when peaks overlap, searching a standard spectra library could produce varying qualitative results due to the extraction of sample spectra with
Fig. 5. 2D chromatographic spectral correlative map of MXFX and its related substances analyzed in two kinds of mobile phases. (1 and A) EO-MXFX; (2 and B) MF-MXFX; (3 and C) DM-MXFX; (4 and D) MXFX; (5 and E) HD-MXFX.
148
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
the peaks at different positions in the chromatogram. In the use of spectral correlative chromatography [8], one chromatogram is only compared with one standard spectrum to find the highest points of every curve, so the process has to be performed many times for identification of all peaks in the chromatogram. However, with two-dimensional chromatographic spectral correlative maps in this paper, all unresolved components and their elution order could be recognized in a new chromatographic system, with only a single run. Fixed size moving window evolving factor analysis (FSMWEFA) [16] is a powerful technique for determining the number of components in a chromatogram. The PCA (Principal Component Analysis) was performed on a fixed number of spectra (a fixed size window) and logarithms of the eigenvalues were plotted against the reten-
tion times to form a rank map. The number of curves above the noise level in the rank map was the number of components. FSMWEFA was performed on the spectrochromatographic data exported from the chromatogram in Fig. 4C (retention time 53.8–57.2 min), and Fig. 6a shows the rank map. The 2D chromatographic spectral correlative map in Fig. 4c enlarged (retention time 53.8–57.2 min) was shown in Fig. 6b. The local rank 2 from 54.9 to 56.3 min meant there were two components co-eluting in this region. Only one component eluted in each of the regions before 54.9 min and after 56.3 min. The result of FSMWEFA was in accordance with the 2D chromatographic spectral correlative map. It can be seen that 2D chromatographic correlation spectroscopy do not require good resolution of peaks. With the separation ability of HPLC, most of the peaks have their “pure eluting area” for identifica-
Fig. 6. The FSMWEFA rank map of the chromatogram in Fig. 4c (53.8–57.2 min) and the two-dimensional chromatographic spectral correlative map. (a) The rank map obtained after FSMWEFA was performed on the data exported from the chromatogram in Fig. 4c (retention time 53.8–57.2 min). (b) The two-dimensional chromatographic spectral correlative map in Fig. 4c enlarged.
W. Li, C.-q. Hu / J. Chromatogr. A 1190 (2008) 141–149
tion. This gave us a hint that the method can test peak purity also. If one peak in sample chromatogram is overlapped with other peaks, the retention times corresponding to the spot in the 2D spectral correlative map would not cover the whole peak range.
149
the solvent exceed the limit, therefore robustness of the method need to be further studied when it is used in the analysis of certain substances. References
5. Conclusion The method based on HPLC-DAD system has an important advantage over HPLC–MS, that is more kind of mobile phases can be chosen to meet the requirement of separation. Making the most of the separation ability of HPLC and spectral specificity, 2D chromatographic correlation spectroscopy is a powerful method for the identification and is successfully applied to a recognition system for peak tracking in impurity profiling control in quinolones. The main advantages of the method are rapidly and simultaneously identifying multi-components without reference substances, accurately distinguishing the elution orders of the unresolved peaks, and testing the peak purity at the same time. However, the method has deficiencies; for example, if there is a small peak completely embedded in a big sample peak, the left and right wings of the big peak can be identified (showing two spots in the map, like a big spot being cut into halves), but the small peak cannot be identified in the map, because it has no “pure elution area”. Sometimes spectra could be misshaped because changes in the pH value of
[1] Q3A(R2): Impurities in New Drug Substances, International Conference on Harmonization of Technical Requirements for the Registration of Pharmaceuticals for Human Use, Geneva, 2006, http://www.ich.org. [2] European Pharmacopoeia, vol. 1, 5th ed., Council of Europe, Strasbourg, 2004, p. 587. [3] The United States Pharmacopoeia, US Pharmacopeial Convention, 27th ed., Rockville, MD, 2004, p. 2510. [4] P.D. Fraser, M.E.S. Pinto, D.E. Holloway, P.M. Bramley, Plant J. 24 (2000) 551. [5] A.C.J.H. Drouen, H.A.H. Billiet, L.D. Galan, Anal. Chem. 56 (1984) 971. [6] M. Herzler, S. Herre, F. Pragst, J. Anal. Toxicol. 27 (2003) 233. [7] F. Pragst, M. Herzler, B.T. Erxleben, Clin. Chem. Lab. Med. 42 (2004) 1325. [8] B.Y. Li, Y. Hu, Y.Z. Liang, L.F. Huang, C.J. Xu, P.S. Xie, J. Sep. Sci. 27 (2004) 581. [9] K. Vishwanathan, M.G. Bartlett, J.T. Stewart, Rapid Commun. Mass Spectrom. 15 (2001) 915. ¨ [10] C.K. Naber, M. Steghafner, M. Kinzig-Schippers, C. Sauber, F. Sorgel, H.J. Stahlberg, K.G. Naber, Antimicrob. Agents Chemother. 45 (2001) 293. [11] N. Sultana, M. Arayne, A. Naz, Pak. J. Pharm. Sci. 19 (2006) 275. [12] Y.R. Kumar, V.V.N.K.V.P. Raju, R.R. Kumar, S. Eswaraiah, K. Mukkanti, M.V. Suryanarayana, M.S. Reddy, J. Pharm. Biomed. Anal. 34 (2004) 1125. [13] J. Sunderland, C.M. Tobin, A.J. Hedges, A.P. Macgowan, L.O. White, J. Antimicrob. Chemother. 47 (2001) 271. [14] Y.G. Li, F. Zhang, Z.T. Wang, Z.B. Hu, J. Pharm. Biomed. Anal. 35 (2004) 1101. [15] Analytical Methods Committee, Analyst 113 (1988) 1469. [16] H.R. Keller, D.L. Massart, Anal. Chim. Acta 246 (1991) 379.