276 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES We presently have a more complete understanding of the basis of column selec- tivity than for other kinds of selectivity (solvent strength, solvent type, temperature, etc.). As discussed in Section 5.4.1, column selectivity can be quantitatively defined by five different characteristics: • column hydrophobicity H • column steric resistance S* • column hydrogen-bond acidity A • column hydrogen-bond basicity B • column cation-exchange capacity C For the case of neutral samples, only the first four column parameters (H, S*, A, B) are important. The difference in selectivity of two columns for a neutral sample can be characterized by a selectivity function F s (−C) that is defined by Equation (6.3) below (Section 6.3.6.1; see also Section 5.4.3). For the columns of Figure 6.14, compared to the Symmetry column, values of F s (−C) are equal to 26 for the Alltima column, 12 for the Luna column, and 16 for the Spherisorb column. A value of F s (−C) > 5 can lead to potentially significant changes in relative retention, as observed for this example. Larger changes in relative retention as a result of a change in column are observed for ionic samples because the column parameter C has a much larger effect on column selectivity for such samples. 6.3.5 Isomer Separations Compounds that resemble each other structurally tend to have similar values of k and are often more difficult to separate. The RPC separation of isomeric compounds is therefore commonly regarded as challenging. While this may be somewhat true in general, one review of the RPC separation of 137 different isomer-pairs with alkylsilica columns [37] found that 90% of these samples could be separated with R s > 1.0 by simply optimizing %B (k varied up to by 4-fold) and temperature (T variedbyasmuchas25 ◦ C). An example is shown in Figure 6.15 for the separation of three hydroxytestosterone isomers on a C 18 column as a function of temperature. At 40 ◦ , all three isomers co-elute; at 28 ◦ C, the three isomers are separated with R s = 0.8. Lower temperatures generally favor a better separation of isomers. 3.1 3.3 3.5 Time (min) 3.1 3.3 3.5 3.7 3.9 Time (min) 40 ° C 28 ° C 2b 2a 11b (a)(b) 2a + 2b + 11b Figure 6.15 Separation of three hydroxytestosterone isomers (2α,2β,11β) as a function of temperature. Conditions: 250 × 4.6-mm (5-μm) C 18 column; 40% acetonitrile-water; 2 mL/min. Adapted from [38]. 6.3 SELECTIVITY 277 S CO CH COOH CH 3 S CO CH COOH CH 3 S CO CH COOH CH 3 (1) (2) (3) 1 2 3 0 5 10 (min) 15 Figure 6.16 Separation of isomers with a cyclodextrin-bonded column. Conditions: 250 × 4.6-mm (5-μm) Cyclobond I column; 30% acetonitrile–pH-4.5 buffer; 35 ◦ C; 2.0 mL/min. Adapted from [40]. 6.3.5.1 Enhanced Isomer Selectivity While it is possible to achieve the baseline separation of some isomers by RPC with alkylsilica columns, the use of a cyclodextrin column may be a better choice [39–41], as illustrated by the example of Figure 6.16. The enhanced isomer selectivity provided by cyclodextrin columns may be due to the presence of –OH groups in the cyclodextrin molecule (allowing for hydrogen bonding between solute and column). The greater rigidity of the three-dimensional cyclodextrin molecule, in contrast to flexible alkylsilica ligands, may also provide a more demanding steric fit for one isomer over another. However, even better isomer selectivity is generally possible by the use of normal-phase chromatography with a silica column (Section 8.3.5). Silver-ion complexation of olefins has been found to be a useful means for enhancing the RPC separation of cis-fromtrans-olefin isomers. The addition of Ag + to the mobile phase results in a preferential interaction with cis olefins. The resulting Ag + -olefin complex is more polar and hence prefers the mobile phase (Section 2.3.2), resulting in its decreased retention compared to the trans isomer. Compounds with varying degrees of cis-unsaturation can be separated according to the number of cis double bonds in the molecule. Silver-ion complexation has been used for the separation of unsaturated fatty acids or their derivatives [42], but is also able to create changes in selectivity for nitrogen- or sulfur-containing heterocyclic compounds [43]. Today silver-ion complexation for the enhanced separation of cis- from trans-olefins is in most cases carried out with silver-impregnated ion-exchange columns [42]. 6.3.5.2 Shape Selectivity This special form of column selectivity was discussed in Section 5.4.1.1. Although it has received considerable attention in the literature, shape selectivity has lim- ited practical application. The reader may therefore wish to skip to following Section 6.3.6. 278 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES “slot” C 18 bonded p hase Figure 6.17 ‘‘Slot’’ model of ‘‘shape selectivity.’’ Shape selectivity refers to the preferential retention of planar polycyclic aro- matic hydrocarbons (PAHs) and certain other compounds (carotenes, steroids) on polymeric alkylsilica columns [44–46]. As a result excellent separations of these iso- meric solutes are often observed with polymeric columns. Two solute characteristics favor the increased retention of a PAH isomer on a polymeric column: molecular planarity (or reduced molecular thickness), and a larger ratio of molecular length to breadth (L/B). It is assumed that polymeric stationary phases are only accessible via narrow openings or ‘‘slots’’ so that planar and/or narrower molecules are better able to enter the stationary phase, and are therefore better retained. This ‘‘slot’’ model of retention on polymeric columns is illustrated in Figure 6.17, where a nonplanar PAH is shown being inserted into a narrow ‘‘slot’’ in the stationary phase of a polymeric column. Because of the greater ‘‘thickness’’ of this nonplanar molecule, it will not be able to completely enter the narrow slot. A more planar, less thick PAH would be better able to access the stationary phase—and therefore be retained more strongly. For separation on polymeric columns (which have a higher coverage of alkyl ligands), it is found that isomeric PAHs are much better resolved than is the case for monomeric columns [45]. This was illustrated in Chapter 5 by the separation of a mixture of 12 C 22 PAH isomers on a polymeric column (Fig. 5.23c), compared to separation on a monomeric column (Fig. 5.23d). However, monomeric columns generally perform better than polymeric columns in other respects (Section 5.2.3) and are much more widely used. Because only a few classes of compounds are better separated on polymeric columns, shape selectivity is rarely tried as a means of improving separation during method development. 6.3.6 Other Selectivity Considerations So far our discussion of separation selectivity has emphasized its use during method development—the systematic variation of conditions to maximize resolution and/or shorten run time. Once a satisfactory separation has been achieved in this way, certain adverse possibilities require consideration. Thus after a routine method has been developed, it is likely to be used over an extended period and in different laboratories. Columns degrade during routine use and must therefore be replaced from time to time. As discussed in Section 5.4.2, column selectivity may vary slightly from one manufacturing batch to another, in which case a replacement column of the same kind from the same source may no longer provide an acceptable separation. This is more likely for separations with > 10 peaks, especially when combined 6.3 SELECTIVITY 279 with marginal resolution (R s < 2). When a replacement column exhibits changed selectivity and fails to provide adequate resolution, it is necessary to locate a different column that is equivalent to the original column, or to vary separation conditions so as to minimize the effect of a change in column selectivity (Section 6.3.6.1). An equivalent replacement column may also be required if the original column has been discontinued and is no longer available, or if the original column is unavailable at another site where the separation is to be carried out. Another problem unrelated to column variability is the possibility that a sample component may be overlapped by another peak and therefore missed in the final RPC analysis. The discovery of such hidden peaks can be aided by the use of orthogonal separation (Section 6.3.6.2). 6.3.6.1 Equivalent Separation Two approaches can be used to restore an original separation when column selectivity changes: (1) a change in column source (i.e., part number), or (2) a change in separation conditions (‘‘method adjustment’’). Which procedure is applicable may depend on the recommendations of a regulatory body (Section 12.8) when an HPLC method falls under governmental regulation. A change in column source requires the identification of an alternative, ‘‘equiv- alent’’ column. Most RPC columns can be characterized by five selectivity parameters (H, S*, A, B, and C), as discussed above and in Section 5.4.1. Equivalent columns will have similar values of each of these column parameters. A column comparison function F s has been defined for two columns 1 and 2 (Eq. 5.4) and any sample (ionic as well as neutral). For the case of neutral samples, the column parameter C is not relevant, so Equation (5.4) simplifies to [47] neutral samples only: F s (−C) ={[12.5(H 2 − H 1 )] 2 + [100(S* 2 − S* 1 )] 2 + [30(A 2 − A 1 )] 2 + [143(B 2 − B 1 )] 2 } 0.5 (6.3) where subscripts 1 or 2 refer to values of a specific parameter (H, S*, etc.) for each column. For two columns that are exactly equivalent, F s and F s (−C) = 0. When F s or F s (−C) ≤ 3 for two columns, it is highly likely that they will provide equivalent separation (i.e., critical R s values that differ by < 0.5 units). An example of the application of Equation (6.3) for a gradient method and an ionic sample is shown in Figure 6.18. An original RPC procedure was developed, giving the separation shown in Figure 6.18a for a drug product and 10 impurities (total of 11 peaks, each marked by an *). Two possible replacement columns were selected for which F s ≤ 3 (vs. the column of Fig. 6.18a), giving the separations shown in Figures 6.18b and c. These three separations (a–c) are seen to be essentially equivalent, with very similar resolution for each compound. A fourth column (with F s = 10) gave the result of Figure 6.18d, where the last two peaks are seen to overlap (arrow). Consequently this last column is not equivalent to the starting column of Figure 6.18a, as suggested by its larger value of F s . Method adjustment aims at correcting for differences in column selectivity, when locating an equivalent column proves to be impractical or impossible. Instead of changing the column, other separation conditions are varied as a means of counteracting the change in column selectivity. An example of this approach is shown 280 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES (a) (b) (c) (d) Figure 6.18 Example of the use of values of F s to select columns of similar selectivity for pos- sible replacement in a routine HPLC assay. Gradient separations where only the column is changed for the separations of a–d. Asterisks mark peaks of interest, values of F s calculated from Equation (5.5) (ionic [not neutral] sample). Reproduced with permission from [48]. in Figure 6.19. The original separation with column A is shown in Figure 6.19a, with baseline resolution (R s = 1.7). At a later time a new column B (same type, different production lot) was used and the separation of Figure 6.19b resulted; the observed resolution is unacceptable, due to the increased overlap of peak-pairs 2 and 3(R s = 1.3) and 6/7 (R s = 1.2)—the result of a change in column selectivity. When carrying out method adjustment, the first step is to determine how separation changes when one or more experimental conditions are varied. For neutral samples, possible choices in conditions include %B and temperature (preferred), or the variation of the proportions of two organic solvents that together comprise the B-solvent (e.g., mixtures of ACN and MeOH). The effect of a change in temperature or %B for column B is shown in Figure 6.19c,d. We see that an increase in temperature (Fig. 6.19c) reduces the resolution of peak-pair 2 and 3 (a change in R s of −0.8 units) but has an opposite effect on peak-pair 6 and 7 (a change in R s of +0.2 units). A decrease in temperature could therefore restore baseline resolution for peaks 2 and 3 but would further decrease the resolution of peaks 6 and 7 (which are already poorly resolved). 6.3 SELECTIVITY 281 (c) Column B 50% B, 45°C R s = 0.5 0246 Time (min) 1 2 + 3 4 5 6 7 (d) Column B 55% B, 35°C R s = 2.1 024 Time (min) 1 2 4 5 6 7 3 0246 Time (min) (b) Column B 50% B, 35°C R s = 1.2 1 2 3 4 5 6 7 0 246 Time (min) (a) Column A 50% B, 35°C R s = 1.7 2 3 4 5 6 7 1 Figure 6.19 Example of method adjustment for a seven-component mixture of neutral com- pounds. Sample: 1, oxazepam; 2, flunitrazepam; 3, nitrobenzene; 4, 4-nitrotoluene; 5,ben- zophenone; 6, cis-4-nitrochalcone; 7, naphthalene. Conditions: 150 × 4.6-mm C 18 column (B differs from A only in a 10% lower ligand coverage); 2.0 mL/min; acetonitrile-water mobile phases; other conditions shown in figure. Separations a–d recreated from data of [8, 9]. An increase in %B, (Fig. 6.19d), on the other hand, results in improved resolution for each peak-pair, and acceptable resolution (R s = 2.1). The approach of Figure 6.19 is straightforward for this sample, in that a change in only one condition (%B) was able to correct for the change in selectivity of column B. Often this is not the case, and then a simultaneous change in two (or more) conditions may be necessary to adjust for a change in column selectivity. While a selection of two or more new conditions can be made by trial and error, a more efficient approach has been described [50]. Retention data are required for 282 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES four experimental runs, as in Figure 6.19a–d, following which a simple mathematic procedure can predict conditions for the closest possible match to the original separation of Figure 6.19a, using the replacement column of Figure 6.19b.For allowed changes in conditions during method adjustment, see Section 12.8. The problem of restoring an equivalent separation (when column selectivity changes between different production lots) is best anticipated during method devel- opment, rather than addressed after the problem arises. Three different options exist during method development: 1. check the reproducibility of different production lots of the column used for the final method 2. confirm the identity of one or more equivalent replacement columns 3. carry out method adjustment for one or more nonequivalent replacement columns Option 1 should be our first choice. Several different lots of the selected column can be evaluated for equivalent separation. Usually it will be found that all of the column lots tested provide adequate separation. If this is not the case, it may be necessary to replace the original column with a different (more reproducible) column. Option 2 is a useful supplement to option 1, even when option 1 confirms that different production lots of the original column are equivalent (later production lots may not be equivalent!). By identifying one or more equivalent (replacement) columns during method development (e.g., Fig. 6.18b,c), any of these columns can serve as a replacement in the event that future lots of the original column exhibit changed selectivity and are no longer suitable [48]. Option 3 can be used whenever option 2 fails (no alternative columns are equivalent to the original column). If the required changes in conditions are minor, the use of an alternative column with method adjustment may be considered as an equivalent method, not requiring complete re-validation of the method (Section 12.8). Another version of method adjustment that avoids the use of changed conditions for a replacement column is to select separation conditions during method development that provide equivalent separations for both the original and one or more nonequivalent replacement columns [51]. 6.3.6.2 Orthogonal Separation The problem of missing or ‘‘hidden’’ peaks can arise during method development when two compounds remain unseparated despite changes in separation conditions. If this situation is overlooked, the final method will be unacceptable because of the missing peak. Even when all the compounds of interest have been separated during method development, later samples may contain additional (unexpected) components that might be missed if overlapped by another peak. Missing peaks are most likely when the overlapped peak is small compared to other peaks in the sample (e.g., a sample impurity), or when the number of possible compounds in a sample is initially unknown. The problem of missing peaks can be addressed by the use of selective detection (mainly mass spectrometry) and/or the development of an orthogonal separation: a separation with very different selectivity that is therefore likely to separate two peaks 6.3 SELECTIVITY 283 that were overlapped in the original (primary) method. For method development where mass spectrometric detection (LC-MS) is often employed, several sets of orthogonal conditions have been proposed that employ changes in the column, B-solvents, and/or mobile phase pH [52–53]. Two or more of these procedures can be used with a given sample to minimize the possibility that two solutes will be overlapped and therefore missed during method development. The specificity of LC-MS, combined with very different separation selectivity, makes it highly unlikely that any sample component will be overlooked. Once method development is completed for a given (hopefully representative) sample, there is always the possibility that future samples may be found to contain additional compounds not present in the original sample. These extra compounds might arise because of changes in a manufacturing process, contamination of the sample during processing or handling, or for other reasons. Routine HPLC assay procedures usually involve UV rather than MS detection; therefore an orthogonal procedure based on UV detection needs to be available for laboratories where the routine assay is performed. A further requirement is the complete separation of known sample components by the orthogonal method. A general procedure for the development of an orthogonal separation that meets the requirements above for routine use has been described and proved useful in several different laboratories [49]. Starting with the primary method used for routine application, a column of very different selectivity is selected, based on a large value of F s (−C) (Eq. 6.3) for the orthogonal versus primary columns. In addition to a change in the column, the B-solvent is changed; if ACN was used for the primary method, MeOH is used for the orthogonal method, and vice versa. Because the compounds separated by the routine procedure may not be fully resolved at this point, separation temperature and %B are next optimized for the maximum resolution of the sample (Section 6.4.1.2). Finally, the orthogonality of the latter method is evaluated from a plot of log k for the orthogonal method against log k for the routine method. If insufficient orthogonality has been achieved at this point, further changes in separation conditions can be explored (use of a different column, change in pH, etc.). When an ionic sample is involved (Chapter 7), a large value is sought of F s (Eq. 5.4) rather than F s (−C). The latter general procedure, which can be extended to gradient elution and used for either neutral or ionic samples, can be further improved [47, 54] for greater orthogonality of the two separations. An example of an orthogonal method that was developed as above is shown in Figure 6.20 for a primary method based on gradient elution. In Figure 6.20a the separation of a sample by the routine method is shown. A major component 3 and four impurities 1, 2, 4, and 5 are separated to baseline. When the routine method was first developed, it was established that only these five components were present in samples manufactured at that time. When the orthogonal method of Figure 6.20b was developed at a later time and applied to a new batch of this active ingredient, another sample impurity 6 was discovered that is overlapped in the routine method by the major component 3 (and therefore would be missed in the primary method of Fig. 6.20a). 284 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES 0 10203040 Original (primary) method 12 4 5 * 3 + 6 (a) (min) 0 10 20 30 (min) 1 2 6 3 5 4 “Orthogonal” method (b) * Figure 6.20 Comparison of separation by an original versus ‘‘orthogonal’’ method. Gradient separations where the column and organic solvent are changed (mobile-phase pH = 6.5 for both a and b). Asterisks mark gradient artifacts (not solute peaks). Reproduced with permis- sion from [49]. 6.4 METHOD DEVELOPMENT AND STRATEGIES FOR OPTIMIZING SELECTIVITY A general approach to method development was presented in Section 2.5.4 and is summarized in Figure 6.21. Seven method-development steps are listed in Figure 6.21a. Step 1 consists of a review of initial information about the sam- ple and a preliminary set of goals for the final separation (required resolution, e.g., R s ≥ 2; desired separation time, e.g., ≤10 min; etc.). A specialized approach to method development may be indicated for some samples (large biochemical or polymeric molecules, Chapter 13; enantiomeric isomers, Chapter 14; inorganic ions, not discussed in present book), or if preparative separation is intended (Chapter 15). Step 2 (sample pretreatment) is required for some samples, those that cannot be injected without damaging the column or that contain interfering substances that are likely to overlap peaks of interest. Step 3 involves an initial choice of chro- matographic mode; RPC will be selected in most cases, but this decision can be modified after initial experiments where separation conditions are varied (step 3 of Fig. 6.21b). The choice of detector (usually UV and/or MS) is made in step 4. The selection of separation conditions (step 5) is usually considered the main part of method development, is detailed in Figure 6.21b, and discussed further below. Step 6 deals with some common problems that can arise after a method is developed, when it is used for routine application. Step 7, which deals with method validation and system suitability, represents another critical part of method development. The selection of chromatographic conditions is examined further in Figure 6.21b. Based on what is known about the sample, initial conditions are selected for the separation (e.g., as in Table 6.1) and an initial run is carried out (step 1 of Fig. 6.21b). For an initial isocratic separation, %B is then varied for adequate retention: 1 ≤ k ≤ 10 (step 1a). Alternatively, a single gradient separation can be used (step 1b; see Section 9.3.1). The initial separation(s) may exhibit 6.4 METHOD DEVELOPMENT AND STRATEGIES FOR OPTIMIZING SELECTIVITY 285 various problems (step 2): tailing peaks, poor retention of the sample (even for an aqueous mobile phase), excessive retention of the sample (even for 100% B), or too many peaks (a ‘‘complex’’ sample). Whatever problem might be encountered in the initial separation(s), the problem should be resolved before proceeding further (see Section 6.6). In some cases a change in separation conditions or chromatographic mode may be indicated (step 3); in this case, return to step 1 and start over. Before beginning experiments for the optimization of selectivity (step 5), the presence in the sample of acidic, basic, and/or neutral compounds should be confirmed (step 4). When the composition of the sample is known before method development starts, step 4 can be omitted. For other samples, individual peaks can be identified as acids, bases, or neutrals by a 0.5-unit change in mobile-phase pH (see Section 7.2 and examples of Figs. 7.2 and 7.3). The all-important selection of conditions for optimized selectivity (step 5) will vary for different samples and chromatographic modes; this topic is addressed below and in individual chapters that focus on sample type and/or chromatographic mode (Chapters 6–9, 13, and 14). Finally, when the optimization of selectivity is completed, column conditions Method Development (a) 1. Assessment of sample composition and separation goals (Section 2.5.4.1) 2. Sample pretreatment (Chapter 16) 3. Selection of chromatographic mode (usually RPC) 4. Detector selection (Chapter 4) 5. Choice of separation conditions (Fig. 6.21b) 6. Anticipation, identification, and solution of potential problems (Section 2.5.4.6) 7. Method validation and system suitability (Chapter 12) Figure 6.21 General method-development approach for use in this and following chapters. . changed selectivity and fails to provide adequate resolution, it is necessary to locate a different column that is equivalent to the original column, or to vary separation conditions so as to minimize the. has lim- ited practical application. The reader may therefore wish to skip to following Section 6.3.6. 278 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES “slot” C 18 bonded p hase Figure 6.17. developed, it is likely to be used over an extended period and in different laboratories. Columns degrade during routine use and must therefore be replaced from time to time. As discussed in