ANALYST, JANUARY 1989, VOL. 114 53 Development and Optimisation of a High-performance Liquid Chromatographic Assay for Tioconazole and its Potential Impurities Part 11.* Selection of Detection Conditions for Potential Impurities Adrian G. Wright University of Bradford, Bradford BD7 1 DP, UK John C. Berridget Pfizer Central Research, Sandwich CT13 9NJ, UK Anthony F. Fell University of Bradford, Bradford BD7 1 DP, UK An optimised high-performance liquid chromatographic separation developed for the assay of tioconazole and its potential impurities has been applied to real-world samples where tioconazole is in excess. As severe peak tailing interferes with the assay of two of the impurity peaks, changes in detection wavelength have been examined as a means to discriminate between this interference.The resulting enhancement of resolution has been exploited in the optimisation of analysis time. Keywords: Tioconazole; detection wavelength optimisation; United States Pharmacopeia The resolution, R,, between two adjacent chromatographic peaks is given by: k’ 1 + k’ R =’-. ( a - l) fl.- s 4 a where a is the selectivity factor between the two compounds, N is the column plate count and k‘ is the average capacity factor for the two peaks. For these studies the column is specified and N is, therefore, fixed. The selectivity factor, a, may be optimiscd with respect to column chemistry and mobile phase composition.’ As it is assumed that changing the proportion of the aqueous phase does not affect the selectivity between two peaks, increasing the average capacity factor, k ’ , can be exploited to improve resolution.However, the degree of improvement becomes progressively less significant once k’ exceeds The principal variables influencing the separation of tioco- nazole and its potential impurities (Fig. 1) have been identified and optimised with respect to selectivity in previous work.3 For convenience, the sample used for method develop- ment contained all components (tioconazole and the four potential impurities; each between 40 and 80 pg ml-1) at comparable levels. However, this does not reflect the real- world situation where, under United States Pharmacopeial (USP) regulations,4 the bulk drug must contain a minimum of 97% tioconazole. As the impurities are present at such low levels the detection sensitivity required is high.In fact, real-world samples analysed using the optimised separation conditions, and under these constraints of high sensitivity, yield chromatograms where tailing of the tioconazole peak is so extensive as to interfere with the determination of related compounds B and C (Fig. 2). Clearly the sloping base line would not be acceptable for a routine assay of these compounds. A solution to this interference problem was sought by optimising the detection conditions to increase the discrimination between tioconazole peak tailing and the signals for the two impurity peaks. It is common practice to adopt a single detection wavelength which fulfills the twin requirements of satisfactory sensitivity for all peaks of interest and maximum discrimination against interference due, for example, to peak tailing.However, few reports on the detailed assessment and validation of this strategy have appeared in the applications literature. The possibilities of increasing the average capacity factor for enhancing resolution and overcoming interference of the tioconazole peak were also examined. /-N CH- I cHz-Nd 0 cld \ I R Tioconazole R = \ = cHzo Related compound A 1 - { 2-[ (3-t h ien yl )met hoxyl-2- (2,4-dichlorophenyl)ethyl}imidazole \ Related compound B 1 - { 2 - [ ( 2,5- d i c h I o ro -3-t h i e n y I 1 m e t h ox y 1 -2 - (2,4-d i c h I o ro p h en y I ) et h y I } i m id azo I e R = CI I \ Related compound C 1 -{ 2-[(5-bromo-2-chloro-3-thienyl)methoxy]- Br 2-(2,4-dichlorophenyI)ethyl}imidazole R = cHh cI s Compound D (hydrolysis product) 1 -( 2,4-d i c h I o ro p h e n y 11-24 i mi dazo I - 1 - y I )et h a no I R = H * For Part I of this series see reference 3.-: To whom correspondence should be addressed. Fig. 1. and hydrolytic degradation product D Structures for tioconazole, related compounds A, B and CANALYST, JANUARY 1989, VOL. 114 A 0 20 timin Fig. 2. Chromatogram for bulk tioconazole run with optimum mobile phase and a detection wavelength of 220 nm. Interference between tioconazole peak tail and related cornpounds B and C is extensive. Eluent, (methanol - acetonitrile (70 + 30)] - pH 4 triethylamine phosphate buffer (0.05 M) containing 1-octanesulphonic acid (0.025 M) (54 + 46 VW) (this is the optimum eluent from reference 3); flow-rate, 1 .5 ml min- 1 The examination of wavelength to enhance discrimination between tioconazole and compounds B and C was primarily concerned with locating the best compromise wavelength which provided good discrimination while still being able to detect <1% mim of each of the potential impurities.Experimental A Hewlett-Packard 1040A diode array detector, a Hewlett-Packard 8SB microcomputer, a Hewlett-Packard 7470A plotter (Hewlett-Packard, Wokingham, UK) and an LDC constaMetric 3000 pump (LDC UK, Stone, UK) were used. The sample injection valve was a Rheodyne 7010 fitted with a 20-pl loop (Alltech Associates, Carnforth, UK). The column was 5-pm Hypersil phenyl (150 x 4.6 mm i.d.) (Technicol, Stockport, England). Mobile phases were pre- pared from HPLC-grade solvents (Rathburn Chemicals, Peebles.UK), HPLC-grade 1-octanesulphonic acid (Fisons, Lough borough, UK) and reagent-grade trieth ylamine (Hop- kin and Williams. Chadwell Heath, UK). The buffer pH was adjusted using reagent-grade phosphoric acid (BDH, Poole, UK). The flow-rate used throughout was 1.5 ml min-1. Tioconazole, the three related impurities and the degradation product (Fig. 1) were supplied by Pfizer Central Research, Sandwich, UK. Results and Discussion The UV - visible spectra for tioconazole and related com- pounds B and C (Fig. 3) were assessed for differences which could be exploited. The spectra for B and C were essentially identical and therefore only one spectrum was compared with that of tioconazole. Differences between the spectra were assessed by calculating the ratio of absorbance for B to that for tioconazole at the same wavelength.Plotting this ratio over the wavelength range of interest (220-270 nm) permitted recognition of the wavelength producing maximum discrimi- nation between compounds B and C and tioconazole (Fig. 4). A graph of ratio against wavelength revealed improved discrimination between tioconazole and compounds B and C above 260 nm, while still retaining a significant absorbance for 190 240 300 190 240 300 Wavelengthinm Fig. 3. (2), B (3) and C (4), and hydrolytic degradation product D ( 5 ) UV - visible spectra for tioconazole ( I ) , related compounds A 220 230 240 250 260 270 Wavelength/nm Fig. 4. Ratio of absorbance for compound B and absorbance tor tioconazole plotted against wavelength.Maximum discrimination is revealed by an increase in ratio above 250 nm these potential impurities. A study of the UV - visible spectra for potential impurities A and D (Fig. 3) indicated only low UV absorbance at 260 nm, which presented difficulties for the quantification of the components. Detection wavelengths above 260 nm were observed to yield greater discrimination between tioconazole and compounds B and C, but also caused a further reduction in the signals for A and D, which were already very low. As a result 260 nm was investigated as the detection wavelength of choice. An overriding consideration in any final decision on detection wavelength was the ability to detect and measure low levels of all four potential impurities. It was considered undesirable to develop a method which minimised the interference from tioconazole so that B and C could be measured, if as a consequence A and D could not be detected.On transferring the separation to the 15-cm column it was found that the optimum separation conditions reported in the earlier study7 yielded a longer analysis time than was desirable. The analysis time was therefore adjusted by decreasing the proportion of aqueous buffer. The proportions of methanol and acetonitrile were kept constant relative to each other (70 + 30) to avoid any loss of selectivity. Studies revealed that changes in the proportion of buffer ranging between 40 and 55% had only a marginal effect on the selectivity between tioconazole and compounds B and C. Two mobile phases were studied to assess the enhancement of resolution between tioconazole and compounds B and C that could be achieved by increasing the average capacity factor.ANALYST. JANUARY 1989, VOL.114 55 Table 1. Regression data for spiked tioconazole standards and external standards for the two mobile phases studied Correlation coefficient Intercept Mobile phase* Standard+ Impurity 220 nm 260 nm 220 nm 260 nm 1 ES A B C D 1 BULK A B C D 2 ES A B C D 2 BULK A B C D 3 . 9 9 4 >0.997 0.010 0.013 0.004 -0.023 H . 9 9 9 >0.999 -0.016 -0.244 -0.239 0.011 >0.998 >o. 998 0.020 0.016 '0.01 1 0.040 >0.999 >0.998 -0.002 -0.280 -0.247 -0.016 ' 1 = 42% aqueous buffer and 2 = 48"h aqueous buffer. + ES = external standards (containing no tioconazole) and BULK = spiked bulk tioconazole. -0.038 0.010 0.006 -0.035 -0.034 -0.275 -0.266 0.007 -0.039 0.012 0.007 0.02 1 - 0.025 -0.304 -0.276 - 0,020 The first eluent consisted of [methanol - acetonitrile (70 + 30)j - triethylamine phosphate buffer (0.05 M) containing l-octane- sulphonic acid (0.025 M) and adjusted to pH 4 with phosphoric acid (58 + 42 VW).This eluent resulted in a retention time of 11 min for the last peak and an average capacity factor of 9.7 for B and C. The second eluent contained a higher buffer content (52 + 48 ViV) and resulted in a retention time of 13.6 min with an average capacity factor of 11.7 for B and C. The purpose of this assay was to detect and measure accurately low levels of the four potential impurities in the bulk tioconazole. Two sets of standards were run for both mobile phases to establish whether this could be achieved with a detection wavelength of 260 nm compared with the higher sensitivity available at 220 nm.One set of standards consisted of tioconazole (3.6 mg ml-'), spiked with known amounts of the four potential impurities (&55 pg ml-I), equivalent to 0-1.5% of tioconazole concentration, and made up in mobile phase. The second set of standards consisted of external standards containing the four potential impurities at identical concentrations (0-55 pg ml- I ) , but omitting tioconazole. The samples were dissolved in mobile phase to avoid the base-line disturbances, which resulted if another solvent were injected. It was assumed that these deviations resulted from disturbance of the columnhon-pairing agent equilibrium. Therefore, it was necessary to make up two sets of standards for both mobile phases.The linearity of response was assessed for standards with and without tioconazole for both eluents, at both detection wavelengths (260 and 220 nm). The 220-nm wavelength was selected for convenience, on the basis that any slight differ- ence in performance at the official USP detection wavelength (219 nm) would be marginal. Regression analysis of the results (Table 1) yielded good linearity for the two detection wavelengths with both sets of standards. The intercepts for the two detection wavelengths were not significantly different. A study of the error intervals associated with the intercepts revealed that, for any one set of standards, the overlap between the 95% error intervals for the two detection wavelengths was at least 83%, and in most instances, 100%.The results also revealed that A and D both retained a sufficient absorbance at 260 nm to be detected and quantified. A significant improvement in the discrimination between tioconazole and potential impurities B and C was observed with a detection wavelength of 260 nm compared with 220 nm (Fig. 5 ) . It was therefore concluded that a detection wavelength of 260 nm was suitable for this assay. I ' Tioconazole D B 5 10 timin Fig. 5. Overlay of chromatographic traces for detection wavelengths of (-) 220 and (--) 260 nm. Enhancement of discrimination between tioconazole and compounds B and C with a detection wavelength of 260 nm is clear, as is the loss of signal for compounds A and D.Eluent: [methanol - acetonitrile (70 + 30)] - pH 4 triethylamine hosphate buffer (0.05 M) containing 1-octanesulphonic acid (0.025 MY (58 + 42 ViV); flow-rate, 1.5 ml min-1 A comparison of the separations achieved for the two mobile phases revealed only marginal improvement in resolu- tion between related compounds B and C. The enhancement in resolution was calculated to be only 2% at the expense of an increase of 2.6 min in analysis time. As analysis time was the principal consideration during these studies, it was decided that the small increase in resolution did not compensate for the increased analysis time. As a result, the mobile phase producing the shorter analysis time was selected. The final, fully optimised assay conditions were: column, Hypersil phenyl 5-pm, 150 x 4.6 mm i.d.; mobile phase, [methanol - acetonitrile (70 + 30)] - triethylamine phosphate buffer (0.05 M) containing 1-octanesulphonic acid (0.025 M) and adjusted to pH 4 with phosphoric acid (58 + 42 V/V); flow-rate, 1.5 ml min-1; temperature, 40 "C; and detection wavelength, 260 nm.The proposed method was validated by comparison of assay results obtained with the USP and proposed procedures for a number of bulk tioconazole batches. Quantification of impur- ity levels was achieved by comparison with an external standard containing the four impurities, in eluent, at the concentrations given in Table 2. The assay results for the proposed method agreed very closely with those for the USP56 ANALYST, JANUARY 1989, VOL. 114 Table 2.Comparison o f assay results for the few potential impurities determined by the USP and the proposed methods USP method, Proposed method, Batch” Impurity % rn!m % rnirn 1 A B C D 2 A B C D 3 A B C D 4 A B C D (0.3 0.5 0.3 <0.3 0.4 0.5 4 . 3 0.7 0.7 <0.3 0.6 0.4 - - - - (0.2 0.5 0.3 <0. 1 <0.2 0.4 0.5 <0. 1 <0.2 0.8 0.6 <0. 1 <0.2 0.7 0.4 <o. 1 * Solutions of tioconazole were prepared in eluent to give a concentration in the range 6.7-7 mg ml-1. The external standard used for the quantification contained the potential impurities at the following concentrations, dissolved in eluent: A = 40, B = 60, C = 59 and D = 39 pg ml-1. procedure currently prescribed. Note that the degradation product D is not covered in the USP monograph. Conclusions Following the earlier optimisation of selectivity by modifica- tion of mobile phase composition and selection of column chemistry,j a second strategy was adopted to optimise resolution and analysis time.In the present work, small increases in the proportion of aqueous buffer in the mobile phase (while keeping the proportions of the organic modifiers, and therefore selectivity, constant relative to each other) were found to increase the average capacity factor for peaks B and C, without significantly enhancing resolution. The improve- ment in resolution achieved in this way was only 2% for these peaks, whereas the analysis time increased by 24%, making this approach impractical. The detection wavelength selected (260 nm) permitted increased discrimination between the peak tail of tioconazole and the peaks for B and C, with consequently enhanced resolution.Therefore, the final assay conditions were opti- mised fully with respect to both selectivity and detection conditions. Good resolution was achieved between all peaks within 12 min. The assay was shown to yield good linearity for each of the potential impurities, both in the presence and absence of tioconazole, over the concentration range of interest. The proposed method and the recommended USP procedure gave comparable results for the three potential impurities that are controlled, in all four batches of bulk tioconazole examined. The assay method finally developed fulfilled all the addi- tional requirements specified. Moreover, on a purely practical level, for real-world samples the mobile phase was non- destructive towards the column.the assay time was half that for the USP assay, separation was improved (compared with a rioconazole B 0 25 timin Fig. 6. Typical chromatogram of bulk tioconazole recorded under USP condition^.^ Detection wavelength, 219 nm; eluent, acetonitrile - methanol - water (containing 2 ml of ammonia solution) (39 -t 36 + 25 V/V/V); flow-rate, 1.5 ml min-1 the USP method, see Fig. 6) and all four potential impurities could be accurately quantified simultaneously. It should perhaps be noted that with the increasing availability of multi-channel detection systems5 it would be possible, in practice, to employ two, or more, detection wavelengths simultaneously, in order to optimise further the detection sensitivity for the earlier eluting peaks of A and D. Current practice in the international compendia, however, is to recommend the use of a single detection wavelength for regulatory purposes. A. G. Wright is grateful to Pfizer Central Research for financial support. References 1. 2. Berridge, J . C., in “Techniques for the Automated Optimiza- tion of HPLC Separations,” Wiley, Chichester, 1985. Schoenmakers, P. J., in “Optimization o f Chromatographic Selectivity-a Guide to Method Development,” Journal of Chromatography Library, Elsevier, Amsterdam, 1986, Volume 35. Wright, A. G., Fell. A. F., and Berridge, J. C.. 1. Chrorn- atogr. ~ in the press. “United States Pharmacopeia XXI,” United States Pharma- copeial Convention, Rockville. MD. 1985, Suppl. 2. p. 1895. Fell, A. F.. and Clark, B. J . , Eur. Chromatogr. News, 1987, 3. 4. 5 . 1(1), 16. Paper 8102258E Received June 6th, 1988 Accepted October Ilth, 1988