526 QUALITATIVE AND QUANTITATIVE ANALYSIS each peak. However, both procedures rely on the assumption that the detector response for nonstandardized (e.g., unknown) peaks is the same as for peaks for which standards are available; this assumption may or may not be appropriate, depending on the sample composition and choice of detector. UV detection is notorious for order-of-magnitude differences in sensitivity for different compounds. 11.4.1.4 Standard Addition The method of standard addition (‘‘spiking’’) can be useful when a sample blank cannot be obtained, and the sample matrix can affect analyte recovery and/or response. For example, when measuring insulin levels in plasma, it is impossible to obtain plasma without insulin, so standard addition can be used. Standard addition can be based on a single-point or multiple-point calibration. For single-point calibration, the sample is split into two fractions. One fraction is spiked with a known concentration of the standard, and both fractions are analyzed. The calibration factor is obtained as S = area s − area ns conc s (11.7) where area s and area ns are the areas of the spiked and nonspiked samples, respec- tively, and conc s is the concentration of standard added to the spiked sample. Using the data of Table 11.1, we see that for conc s = 2ng/mL,area s = 911 and area ns = 487. S = (911 − 487)/(2 ng/mL) = 212 area units/(ng/mL). Now the con- centration of the non-spiked sample (conc ns ; shown as 0 ng/mL in Table 11.1) can be determined as conc ns = area ns S (11.8) or conc ns = 487/212 = 2.3ng/mL. For multiple-point calibration using standard addition, the sample is split into n + 1 fractions, where n is the number of standards to be used. Then n samples are spiked, each with a different concentration of standard and all samples are analyzed. A calibration curve is plotted, as shown in Figure 11.10 for the data of Table 11.1. Note that the regression line is extended to the left until it intersects the x-axis (arrow in Fig. 11.10). The value of the intercept x corresponds to −conc ns . Linear regression of the data of Table 11.1 gives y = 201x + 496. Solving for x and inserting y = 0 gives x =−2.5, so conc ns = 2.5 ng/mL. The result (<1 SD difference) is the same as that obtained above with Equations (11.7) and (11.8). It should be stressed that the method of standard addition does not correct for baseline variation or other sample interferences. These problems must be handled in the usual way, before the standard addition procedure is applied. In effect, this approach is a form of in situ calibration, and it can be very useful when the more traditional techniques of external or internal standardization cannot be used. As noted in Section 11.3.1, the method of standard addition also can be useful in confirming peak identity, although its value for this purpose is no greater than the use of a retention time. 11.4 QUANTITATIVE ANALYSIS 527 0 5 10 15 20 25 30 −50510 y = 201x + 496 r 2 = 0.9976 Peak Area (x10 −2 ) Concentration Added (ng/mL) Figure 11.10 Use of the method of standard addition to determine analyte concentration at x-axis intercept (arrow); data of Table 11.1. 11.4.1.5 Evaluating Calibration Curves Plots for wide concentration-range calibration curves, such as those of Figures 11.8 and 11.9 can be hard to interpret when a linear x-axis is used, because the low concentration points are crowded together. An alternate way to examine the data is to make a plot of %-error against log concentration (a ‘‘%-error plot’’). %-Error is determined by using the regression equation to calculate the theoretical y-value for each concentration; error is calculated as (experimental value − theoretical value)/theoretical value, and expressed as %-error. The benefit of the %-error plot is shown in Figure 11.11 for a data from a hypothetical method-validation study. Calibrators were injected at 10 concentra- tions: 1, 2, 5, 10, 20, 50, 100, 200, 500, and 1000 ng/mL. The plot of %-error versus (linear) concentration of Figure 11.11a shows that there is much more relative error at lower concentrations, but this figure is hard to interpret because of the crowding of data points at low concentrations. The %-error versus log concentration plot of Figure 11.11b solves the crowding problem, and now the error at each concentration can be easily examined. The data fall into two sets—those above 20 ng/mL with a relatively constant error of ≈±1% (1 SD)—while remaining concentrations show increasing relative error as concentration is reduced (dashed lines of Fig. 11.11b). This pattern is expected, as S/N makes a larger contribution to relative error (Eq. 11.1) at low concentrations. Other than the (normal) increase in error at low concentrations, the data of Figure 11.11 have acceptable regression statistics (r 2 = 0.9999, y-intercept <SE), so a multiple-point calibration curve with forced zero (y = mx; Sections 11.4.1.1, 11.4.1.2) is appropriate. The %-error plot can highlight problems with calibration curves. Data similar to that of Figure 11.11a, b are shown in the example of Figure 11.11c,which emphasizes the importance of picking the proper y-intercept. If a multiple-point calibration is chosen, with the curve forced through zero, the %-error plot of Figure 11.11c results. Lower concentrations show increasing error, with an average error ≈50% for 1 ng/mL. For a bioanalytical method, where maximum error allowed at the LLOQ =±20%, the lowest concentration with an average error of ≤20% is 5 ng/mL (≈9% error); this limits the application of the method to a 528 QUALITATIVE AND QUANTITATIVE ANALYSIS –15 –10 –5 0 0 200 400 600 800 1000 5 10 15 % – Error Concentration (ng/mL) –15 –10 –5 0 1 10 100 1000 5 10 15 % – Error% – Error Concentration (ng/mL) Concentration (n g /mL) ±1 RSD (c) (b) (a) 70 60 50 40 30 20 10 0 mean values 1000100101 Figure 11.11 Use of %-error plots to examine calibration-curve data. (a) %-Error versus (lin- ear) concentration; (b)dataof(a) plotted as %-error versus log concentration (y-intercept <SE; curve forced through zero); (c) %-error versus log concentration (y-intercept > SE; curve improperly forced through zero). 11.5 SUMMARY 529 range of 5 to 1000 ng/mL. For this data set the y-intercept > SE, so the proper curve fit is y = mx + b (with b = 0). The resulting %-error plot (not shown) closely resembles Figure 11.11b, with an average error of <6% throughout the curve. Now the calibration curve allows a bioanalytical method to be applied over a range of 1 to 1000 ng/mL. It is interesting to note that for both curve fits (with b = 0or b = 0), r 2 > 0.9999, so r 2 alone is not sufficient to ensure good performance of a multiple-point calibration. Low r-orr 2 -values can indicate that there are problems with a calibration curve, but the converse is not necessarily true—large r 2 -values do not guarantee a well-behaved curve. A plot of %-error against log concentration (%-error plot) is a useful way to make a visual examination of data. We recommend examining all calibration-curve data with a %-error plot as a means of highlighting potential problems with the method. 11.4.2 Trace Analysis HPLC is used for the analysis of samples of widely varying concentration. The term trace analysis often is used to describe small sample concentrations. One way to define trace analysis is to describe samples for which the precision of measurement is affected by the concentration, often with a transition point to trace analysis when S/N < ≈100 (Section 4.3). Other than the problems associated with dealing with low concentrations and small signals, trace analysis is little different from the analysis of more concentrated samples. With trace analysis, peak height measurements may be preferred over peak area. We recommend evaluating both peak height and peak area; choose the final measurement technique based on the one that gives the best precision and accuracy. For additional information, see Sections 2.6.3.2, 4.2.4, and 11.2.5; for discussion related to specific detectors, consult the appropriate detector discussion in Sections 4.4 through 4.16. 11.5 SUMMARY Use of the HPLC as a qualitative or quantitative analytical tool requires that the sys- tem be operating properly and that the data system be set up to accurately determine peak retention times and peak areas or heights. Consideration has to be taken relative to resolution requirements for the peaks of interest, including the influence of relative peak size and shape. Although the HPLC system is not as useful a tool for qualitative analysis as some dedicated instruments (e.g., FTIR, NMR, or high-resolution MS), with the help of the appropriate detector(s) it can provide valuable qualitative infor- mation for many applications. Liquid chromatography shows its strongest assets with quantitative analysis. HPLC can be used for trace analysis of pollutants in river water, drugs and their metabolites in biological systems, or impurities in reagents. It is also useful for determining content uniformity of pharmaceutical products with high precision and accuracy. Quantitative analysis relies on selection of appropriate reference standards and proper calibration so that the results are of high quality and can withstand the scrutiny of review by regulatory agencies. 530 QUALITATIVE AND QUANTITATIVE ANALYSIS REFERENCES 1. N. Dyson, Chromatographic Integration Methods, 2nd ed., Royal Society of Chemistry, Letchworth, UK, 1998. 2. L. R. Snyder and J. J. Kirkland, Introduction to Modern Liquid Chromatography, 2nd ed., Wiley, New York, 1974. 3. V. R. Meyer, J. Chromatogr. Sci., 33 (1995) 26. 4. R. Q. Thompson, J. Chem. Ed., 62 (1985) 866. 5. Guidance for Industry: Part 11, Electronic Records; Electronic Signatures—Scope and Application, USFDA-CDER, Aug. 2003, http://www.fda.gov/ora/compliance ref/ part11. 6. Guidance for Industry: Bioanalytical Method Validation, USFDA-CDER, May 2001, http://www.fda.gov/cder/guidance/index.htm. 7. Reviewer Guidance: Validation of Chromatographic Methods, USFDA-CDER, Nov. 1994, http://www.fda.gov/cder/guidance/index.htm. 8. L. R. Snyder, J. J. Kirkland, and J. L. Glajch, Practical HPLC Method Development, 2nd ed., Wiley-Interscience, New York, 1997, p. 71. 9. Validation of Analytical Procedures: Text and Methodology Q2(R1), International Con- ference on Harmonization, Nov. 2005, http://www.ich.org/LOB/media/MEDIA417.pdf. 10. Impurities in New Drug Substances Q3A(R2), International Conference on Harmoniza- tion, Oct. 2006, http://www.ich.org/LOB/media/MEDIA422.pdf. 11. Specifications: Test Procedures and Acceptance Criteria for New Drug Substances and New Drug Products: Chemical Substances Q6A, International Conference on Harmonization, Oct. 1999, http://www.ich.org/LOB/media430.pdf. 12. J. C. Miller and J. N. Miller, Statistics for Analytical Chemistry, Halsted Press-Wiley, New York, 1984, secs. 4.9–4.10. CHAPTER TWELVE METHOD VALIDATION with Michael Swartz 12.1 INTRODUCTION, 532 12.2 TERMS AND DEFINITIONS, 534 12.2.1 Accuracy, 535 12.2.2 Precision, 536 12.2.3 Specificity, 539 12.2.4 Limit of Detection and Limit of Quantification, 539 12.2.5 Linearity and Range, 540 12.2.6 Robustness, 540 12.3 SYSTEM SUITABILITY, 542 12.4 DOCUMENTATION, 543 12.4.1 Validation Protocol, 544 12.4.2 Test Method, 544 12.4.3 Validation Report, 545 12.5 VALIDATION FOR DIFFERENT PHARMACEUTICAL-METHOD TYPES, 546 12.5.1 Category 1 Methods, 546 12.5.2 Category 2 Methods, 547 12.5.3 Category 3 Methods, 547 12.5.4 Category 4 Methods, 548 12.6 BIOANALYTICAL METHODS, 548 12.6.1 Reference Standard Preparation, 549 12.6.2 Bioanalytical Method Development and Validation, 549 12.6.3 Routine Application of the Bioanalytical Method, 552 12.6.4 Bioanalytical Method Documentation, 553 12.7 ANALYTICAL METHOD TRANSFER (AMT), 554 12.7.1 Analytical Method-Transfer Options, 555 12.7.2 Essentials of AMT, 556 12.7.3 Potential AMT Pitfalls, 558 12.8 METHOD ADJUSTMENT OR METHOD MODIFICATION, 561 12.8.1 pH Adjustments, 563 Introduction to Modern Liquid Chromatography, Third Edition, by Lloyd R. Snyder, Joseph J. Kirkland, and John W. Dolan Copyright © 2010 John Wiley & Sons, Inc. 531 532 METHOD VALIDATION 12.8.2 Concentration of Buffer Salts, 563 12.8.3 Ratio of Components in the Mobile Phase, 563 12.8.4 Wavelength of the UV-Visible Detector, 564 12.8.5 Temperature Adjustments, 564 12.8.6 Column Length, Diameter, and Particle-Size Adjustments, 564 12.9 QUALITY CONTROL AND QUALITY ASSURANCE, 564 12.9.1 Quality Control, 565 12.9.2 Quality Assurance, 565 12.10 SUMMARY, 565 12.1 INTRODUCTION Quality is a commonly used word in the world of analytical chemistry. Quality encompasses many aspects of the laboratory; in this chapter it refers to the develop- ment and application of HPLC methods. Our primary focus will be the validation of HPLC methods, a process that underlies the quality of the method and laboratory results. Other aspects of quality in the HPLC laboratory are also discussed in this chapter, especially quality control and quality assurance (Section 12.9). Method validation establishes, by means of laboratory studies, that the per- formance characteristics of the test method meet the requirements of the intended analytical application. Method validation provides an assurance of reliability during normal use, and this process is sometimes referred to as providing documented evidence that the method does what it is intended to do. Regulated laboratories must carry out method validation in order to be in compliance with governmental or other regulatory agencies. A well-defined and documented method-validation process not only satisfies regulatory compliance requirements but also provides evidence that the system and method are suitable for their intended use, and aids in method transfer. In 1987, the US Food and Drug Association (FDA) first designated the specifications listed in the current edition of the United States Pharmacopeia (USP) as those legally recognized to determine compliance with the Federal Food, Drug, and Cosmetic Act [1–2]. More recently, new information has been published that updates the previous guidelines and provides more detail and harmonization with International Confer- ence on Harmonization (ICH) guidelines [3–4]. The inclusion and/or definition of some terms differs for the FDA, USP, and ICH, but harmonization on a global basis has provided much more detail than was available in the past. So it may be useful to downplay any differences between global regulatory requirements. An HPLC method may be referred to as an ‘‘analytical procedure,’’ ‘‘analyt- ical method,’’ ‘‘assay procedure,’’ ‘‘test method,’’ or just ‘‘method.’’ In the present discussion, we generally will use ‘‘test method’’ or (less often) ‘‘HPLC method’’ for 12.1 INTRODUCTION 533 any of these methods. The largest number of HPLC methods in use today is carried out in the pharmaceutical industry; for this reason we will describe method valida- tion for pharmaceutical applications. Other regulated industries have well-defined processes in place for method validation as well. For example, environmental moni- toring laboratories are under the oversight of the Environmental Protection Agency (EPA) [5], whereas some other organizations rely on directives of the International Organization for Standardization (ISO) [6]. Validation and other laboratory practices are regulated by the FDA, USP, ICH, EPA, and related organizations. Because HPLC methods developed by an industrial analytical laboratory often are used in a manufacturing department, the analytical laboratory may be constrained by manufacturing practices and regulations. Two of the most common references to these practices are cGMP (current Good Manufac- turing Practice, e.g., [7–8]) and the ISO 9000 Global Management Standards [9] and related ISO (International Organization on Standardization) documents. These aspects of laboratory regulation are not discussed further in this chapter. In nonregulated industries and academic laboratories, there also is a need for high-quality test methods that provide reliable data. The use of good scientific practices is often assumed but is not always the case, so method validation is strongly recommended even where it is not required by regulation. The reader should be able to take the information presented here for pharmaceutical applications and use it as a basis for other areas of application. Method validation can be regarded as just one part of an overall validation process that encompasses at least four distinct steps: (1) software validation, (2) analytical instrument qualification or validation (AIQ; Section 3.10.1), (3) method validation, and (4) system suitability. The overall validation process begins with validated software and a qualified instrument; then a test method is developed and subsequently validated using the qualified system. Finally, the performance of the test method on a given day can be confirmed by means of a system suitability test. Each of these four steps is critical to method performance. Two guidelines are important for any method validation process: USP Chapter 1225, Validation of Compendial Methods [2], and the International Conference on Harmonization (ICH) Guideline, Validation of Analytical Procedures: Text and Methodology Q2 (R1) [4]. Although the subject of the current discussion is HPLC, both the USP and ICH guidelines apply to any analytical procedure, technique, or technology used in a regulated laboratory. It should be noted that the USP publishes official test methods, often called compendial methods, that are accepted by the USP as already validated. The USP also publishes guidelines that should be applied to the validation of test methods not developed by the USP (it is assumed that the USP published methods were subject to the same guidelines). This chapter concentrates on the application of the USP (and other regulatory agency) guidelines to HPLC methods developed by independent laboratories (i.e., not by the agencies themselves). Even though the USP is the sole legal document in the eyes of the FDA, this chapter draws from both USP and ICH guidelines, as appropriate, for definitions and methodology. For the most part the FDA, USP, and ICH guidelines agree. Where the guidelines disagree, it is up to the user to decide on an appropriate interpretation of the guidelines. Often this is the responsibility of the user’s quality assurance unit (Section 12.9), and may be aided by review of the latest regulatory actions (e.g., FDA-issued Form 483 Inspectional Observations). For the present discussion, the 534 METHOD VALIDATION regulatory publications will be referred to generically as ‘‘guidelines.’’ In addition to these guidelines, sometimes a regulatory body publishes other information that can be useful for interpretation of the guidelines. One of these is a ‘‘reviewer guidance’’ [10] published by the FDA. This document is intended to help FDA auditors determine what comprises a good test method, so many users try to adhere to the suggestions of this document to ensure that their test methods will pass regulatory scrutiny. In addition to the general process of method validation, we will discuss terms, definitions, and related topics, and—where possible—provide examples to illustrate how these general guidelines apply to HPLC. A major difference between this chapter and other chapters in this book is the regulatory oversight of validated methods in the pharmaceutical, environmental, and certain other industries. Rules pertaining to method validation are described in official documents that originate at different times, are written by different people (with different writing skills), and are released by different agencies (with varying internal policies). This bureaucratic process inevitably results in documents that can be ambiguous, inconsistent, and difficult to interpret. In this chapter we try to impart some unity to the requirements and guidelines contained in these various regulatory pronouncements. We also try to make the discussion practical for the average user. Nevertheless, the bureaucratic language of regulatory recommendations and requirements could not entirely be masked. Consider this limitation as good practice for dealing with the official documents. Method validation may seem to have a vocabulary of its own. This chapter therefore begins with a discussion of important terms and definitions (Section 12.2). A procedure that ensures that a test method can provide valid data on a given day is the system suitability test, described in Section 12.3. Without documentation, there is no proof of method validity; some aspects of method documentation are described in Section 12.4. Validation of test methods for drug substance (pure drug) and drug product (formulated drug) have requirements (Section 12.5) that are distinctly different from bioanalytical methods that measure drugs in biological matrices (Section 12.6). Once a test method has been validated, it often must be transferred to another laboratory for routine application; some of the principles of analytical method transfer (AMT) are discussed in Section 12.7. Many times when test methods are transferred, they do not work exactly as they did in the original laboratory, and over time, most methods require some adjustment to meet system suitability requirements. Methods can be adjusted to meet system suitability, but if they require more substantial changes (are modified), they must undergo as least some re-validation; the topic of adjustment vs. modification is covered in Section 12.8. Finally, a good test method and its application require strong quality control and quality assurance programs, as described in Section 12.9. 12.2 TERMS AND DEFINITIONS Several analytical performance characteristics may be investigated during any method validation protocol: • accuracy • precision/ruggedness 12.2 TERMS AND DEFINITIONS 535 • specificity • limit of detection • limit of quantitation • linearity • range • robustness Although most of these terms are familiar and are used daily in any regulated HPLC laboratory, they sometimes mean different things to different people. For example, ruggedness, which forms a part of any well-designed precision study, is often confused with robustness. The following standard definitions (Sections 12.2.1–12.2.6), of applications in the pharmaceutical industry should clarify any confusion. In this context, drug substance refers to the pure chemical drug, also called the active pharmaceutical ingredient (API). The drug product refers to the product that is sold to the consumer; it usually contains one or more drug substances plus excipients—(other chemicals, fillers, colors, etc.). Several types of test methods are used to measure the API and/or impurities, related substances, excipients, and so forth. The major method types discussed in this chapter are assay, impurity (also related substances), dissolution, and bioanalytical methods. A test method used for assay is one that measures the active ingredient concentration in a drug product or substance. A content uniformity method is similar to an assay method, but it targets the measurement of the variability in drug concentration within a batch of samples. An impurity test measures the (generally unintentional) minor components present in the substance or product that originate from raw material manufacturing, product manufacturing, or degradation during storage or processing. A stability-indicating method is used to quantify the presence of impurities (degradants) generated through a forced degradation of the API; it is assumed that this test will enable measurement of any impurities generated during normal or accelerated shelf-life testing of a drug substance or product. Any degradants found in this way may be included in the impurity test. A dissolution assay measures the concentration of API in a solution designed to simulate release of the drug from a formulation under the conditions of administration of the drug (e.g., in simulated stomach fluids). Whereas the preceding test methods are for drug product or drug substance, a bioanalytical method (Section 12.6) is used to determine the concentration of a drug in a biological system, most commonly plasma. 12.2.1 Accuracy Accuracy is the measure of exactness of an analytical method, or the closeness of agreement between an accepted reference value and the value found in a sample. Established across the range of the method, accuracy is measured as the percentage of analyte recovered by the assay. For the drug substance, accuracy measurements are obtained by comparison of the results to the analysis of a standard reference material, or by comparison to results from a second, well-characterized method. For the assay of the drug product, accuracy is evaluated by the analysis of synthetic mixtures spiked with known quantities of the analytes. For the quantification of impurities, accuracy is determined by the analysis of samples (drug substance or drug . 556 12.7.3 Potential AMT Pitfalls, 558 12.8 METHOD ADJUSTMENT OR METHOD MODIFICATION, 561 12.8.1 pH Adjustments, 563 Introduction to Modern Liquid Chromatography, Third Edition, by Lloyd R. Snyder, Joseph. UK, 1998. 2. L. R. Snyder and J. J. Kirkland, Introduction to Modern Liquid Chromatography, 2nd ed., Wiley, New York, 1974. 3. V. R. Meyer, J. Chromatogr. Sci., 33 (1995) 26. 4. R. Q. Thompson,. intended to help FDA auditors determine what comprises a good test method, so many users try to adhere to the suggestions of this document to ensure that their test methods will pass regulatory scrutiny.