Transcript for:
HPLC Method Validation Overview

Title: HPLC for Pharmaceutical Scientists URL Source: blob://pdf/de7c9088-432f-46ae-8201-dd8e7bb8ae28 Markdown Content: # 9 # METHOD VALIDATION # Rosario LoBrutto and Tarun Patel 9.1 INTRODUCTION The method validation process is to confirm that the method is suited for its intended purpose. Although the requirements of validation have been clearly documented by regulatory authorities [ICH, USP, and FDA], the approach to validation is varied and open to interpretation. Validation requirements differ during the development process of pharmaceuticals. The method validation methodologies in this chapter will focus on the method requirements for preliminary and full validation for both drug substance and drug product. Preliminary method validation is generally performed in the earlier phases of development up to Phase IIa because at this time ICH Q2A and Q2B [1] are not yet binding. A more extensive validation (full validation) is performed for methods used in later stages of drug development (after Phase IIa) and for methods that will be used to evaluate marketed products. Specific require-ments or methodologies for validation depending on the life cycle of the potential drug candidate in each specific area in the drug development process will be addressed in the corresponding chapter. An analytical method is a laboratory procedure that measures an attribute of a raw material, drug substance, or a drug product. Analytical method vali-dation is the process of demonstrating that an analytical method is reliable and adequate for its intended purpose. Any method that is utilized to deter-mine results during drug substance and formulation development will have to be validated. Reliable data for release of clinical supplies, stability, and setting shelf life can only be generated with appropriate validated methods. 455 > HPLC for Pharmaceutical Scientists , Edited by Yuri Kazakevich and Rosario LoBrutto Copyright 2007 by John Wiley & Sons, Inc. Validation of high-performance liquid chromatography (HPLC) methods focus mainly on the following: > Identification tests > Quantitative measurements of the content of related substances* > Semiquantitative and limit tests for the control of related substances* > Quantitative tests for the assay of major components (e.g., drug substance and preservatives) in samples of drug substance or drug product (assay, content uniformity, dissolution rate, etc.) Moreover, HPLC methods that are described in pharmacopeias may not have to be validated but should be verified, if necessary. Well-characterized refer-ence materials with documented purity should be used throughout the vali-dation study, especially during full development. Validation experiments and analyses must be carried out on fully qualified and calibrated instrumentation, and some references have been published on this subject [26]. Analytical method validation is established through documented evidence demonstrating the accuracy, precision, linearity, selectivity, ruggedness, and/or robustness of that particular test method which will be utilized to generate test results for a drug substance or drug product. Different test methods require different validation parameters. All analytical procedures require some form of method validation, regardless of whether the test method is utilized for the testing of Good Laboratory Practice (GLP) toxicology, shelf-life determina-tion (stability indicating), in-process controls [7], clinical release, or release of products for open market [8]. As development of the project progresses and as more analytical and product-specific information is acquired, the analytical methods evolve and are gradually updated. The extent of validation increases and the documentation is completed. During the early development phase, depending on the analytical labora-tory, generic validation protocols may be used because project-specific pro-tocols are not required. Sometimes an internal Standard Operating Procedure (SOP) suffices and a generic validation protocol does not need to be used. Usually, for Phase I, validation experiments may be carried out concurrently with the analysis of the first batch of clinical supplies or the first delivery of drug substance to be used for clinical supplies. However, depending on the pharmaceutical organization method validation may need to be performed prior to the analysis of material that will be used for clinical supplies. For analytical method validation during full development (after final syn-thesis has been set for drug substance and after final market formulation has been set for drug product) corresponding to the definitive control procedure for new drug application (NDA), a specific validation protocol has to be written. Before start of the experimental work, the protocols must be written > 456 METHOD VALIDATION *Related substances described in this chapter encompass degradation products, and synthetic by-products. by an analytical chemist and approved by a quality assurance department. Some of the items that are necessary to be specified in the validation proto-col are listed below: > The analytical method for a given product or drug substance > The test to be validated > The test parameters for each test, including type and number of solutions and number of injections > The acceptance criteria for each parameter based on an internal SOP (product or method-specific adaptations may be necessary and are accept-able, if justified) > List of batches of drug substance and/or drug products > For a drug product the grade/quality of the excipients used in the formulation > List of reference materials to be used in the validation experiments > Information on the instruments and apparatus to be used > Responsibilities [author, chemists, analytical research project leader, quality assurance (QA), etc.] Depending upon a companys culture, a method validation protocol could be simple (listed items above) or exhaustive (in addition to the listed items above, each parameter to be validated is described in detail): How solutions are going to be made, the experimental design, how the calculations are going to be performed, any software to be utilized (e.g., Excel). If a full-length protocol is required within a particular company, then the writing of this protocol and approval of the protocol would need to be completed prior to the com-mencement of the validation work. Otherwise, there may be many deviations to the protocol which will be needed to be referenced to in the final method validation report. Some companies also have templates for the validation reports, thereby allowing for facile population of the results. Once populated, the file is reviewed to determine if all validation parameters and acceptance criteria were met. If they were not, then a deviation is added and the proper justification must be given. If it is deemed that the justification is not ap-propriate, then an action plan for the specific figure of merit in question is determined (i.e., repeat analysis, change of the analytical procedure, and revalidation). Also, if the analytical method has not yet been approved at the time of writing the validation protocol, it is recommended to attach a final draft of the method to the protocol. The final HPLC method must also be approved with the validation report submission. 9.2 VALIDATION REPORT A validation report is written during early and full development, and approval by QA is required. Existing method validation data from earlier stages of > VALIDATION REPORT 457 development may be used for full development if the HPLC method has not changed. Minor changes such as change in equilibration time may be accept-able, and the preliminary validation performed for early phase may be used. These data can be referred to in the validation report, and reference to the original data must be given. The validation report should contain reference to the analytical methods (specific code number used as identifier within the pharmaceutical organiza-tion) and the corresponding drug substance or product name. Note that for early-phase method validation reports the results maybe filled in a predefined table and compared against the acceptance criteria. However, for late-phase validation, more explicit reports are generated explaining each and every experiment, with detailed steps of sample and standard preparation. The list of reference materials (reference standards with the appropriate certificate of analysis) as well as the list of calibrated and qualified instruments used in the validation experiments should be documented in the report. For drug substances the list of the batches of drug substances, notebook number/reference number for any individual impurities, or solutions or inter-mediates used should be listed. For drug products the list of the batches of drug substances, drug product, and the grade/quality of excipients should be listed. The test parameters and acceptance criteria should be listed together with the results for each test, and the results passed or failed should be indi-cated. The validation report should also contain whether the method valida-tion was successful and if any changes had to be applied to the analytical method, and then the final analytical method must be resubmitted for QA approval. 9.3 REVALIDATION After any major changes in the HPLC method (solution preparation, experi-mental conditions, etc.) or after change in route of synthesis of the drug sub-stance or drug product manufacture (change of process, change of equipment, change of analytical procedure), it must be assessed whether a new validation or a partial validation is required, addressing all the validation parameters that may be affected by the methodological change. If revalidation is not deemed necessary, then the reasons behind the decision must be documented in the revision history of the test method and the proper change control initiated. The revision of the test method and any documents that refer to the original method, such as the analytical specifications, will then be approved by QA. When revalidation is deemed necessary, the reason for change must be docu-mented and any new validation activities must be performed according to an approved, updated HPLC validation protocol. The results would then be doc-umented in an update of the validation report or a supplement to the original validation report. > 458 METHOD VALIDATION 9.4 ASSIGNMENT OF VALIDATION PARAMETERS The type and degree of validation depends on the nature of the test. In par-ticular, methods described in pharmacopeias may not have to be validated but should be verified, if needed. Different test methods require different valida-tion parameters. As development of the project progresses and as more ana-lytical and product-specific information is acquired, the analytical methods evolve and are gradually updated. The extent of validation increases and the documentation is completed. Table 9-1 outlines the validation parameters that are usually required for the early development stage, and Table 9-2 outlines the validation parameters that are usually required for the full development stage. The proposed acceptance criteria in Table 9-3 should be included in the validation protocol, especially for the full development stage. There are numerous method validation examples in the literature [918]. Each company has their own approach and own set of acceptance criteria for different analytical assays, but these approaches must be within the confines of their line unit QA department and be in accordance with any regulatory provisions. In the next section a description for each of the parameters to be validated (figures of merit) are described in detail and examples are given for each. > ASSIGNMENT OF VALIDATION PARAMETERS 459 TABLE 9-1. Early Development Type of Tests to Be Validated Weight Percent/Assay/Content Impurity Testing: Validation Parameters Identity Uniformity/Dissolution Quantitative Test a Specificity Yes Yes Yes Linearity No Yes b Yes b Accuracy No Yes c Yes d Precision (repeatability) No Yes Yes e Limit of detection No No Yes f Limit of quantitation No g No Yes d Stability of the solutions No Yes Yes > aIf impurities not available, with drug substance. > bFour points may be adequate. > cFor drug product only (assay/CU/dissolution). > dA spiking experiment carried out is adequate at this stage (only possible if impurity/impurities are available). > eAt least triplicate analysis. > fNot required, but recommended. > gFor the identity test of a 0-mg formulation (placebo), it may be necessary to document the absence of drug substance, and an LOQ determination will then be required. 9.4.1 Accuracy The test for accuracy is intended to demonstrate the closeness of agreement between the value found and the value that is accepted either as a conven-tional true value or as an accepted reference value [19]. Therefore, accuracy can be defined as the agreement between the result obtained with method being validated and an accepted reference value. The accuracy can be inferred from precision, linearity, and specificity. The results for the method being validated can be compared to the results with those of a well-characterized, independent method. These results may be compared to an alternate reversed-phase HPLC method (phenyl versus C18 or separation run at dif-ferent pH using the same column) using the same detection scheme. In some, cases an orthogonal method is used to demonstrate accuracy. The methods should differ with respect to separation mode and therefore provide orthog-onal information concerning related substances and degradation products. For example, one method would use reversed-phase (RP) separation mode on a C18 column, and the second method would use a strong cation exchange (SCX) column [20]. The orthogonal methods may show different selectivities toward the degradation products, thereby demonstrating the orthogonal nature of the two separation techniques. The accuracy would be demonstrated 460 METHOD VALIDATION TABLE 9-2. Full Development Type of Tests to Be Validated Weight Percent/Assay/Content Impurity Testing: Validation Parameters Identity Uniformity/Dissolution Quantitative Test Specificity a Yes Yes Yes Linearity No Yes Yes Accuracy No Yes Yes Precision (repeatability) No Yes Yes Precision (intermediate No Yes Yes precision) b Precision (reproducibility) No c c Range No Yes Yes Limit of detection No No Yes e Limit of quantitation No d No Yes Stability of the solutions No Yes Yes Robustness f Yes Yes > aLack of specificity of one analytical procedure may be compensated for by other supporting analytical procedures. > bIn cases where reproducibility has been performed, intermediate precision not needed. > cIn exceptional cases. > dFor the identity test of a 0-mg formulation (placebo) it may be necessary to document the absence of drug substance and an LOQ determination will then be required. > eNot required by ICH, but recommended. > fMay be required, depending on the nature of the test. ASSIGNMENT OF VALIDATION PARAMETERS 461 TABLE 9-3. Proposed Acceptance Criteria for Drug Product (DP) and Drug Substance (DS) Quality Characteristics Parameter to be Validated Acceptance Criteria Identity Selectivity/specificity All known peaks are separated. Major (API) peak is pure [Peak purity angle peak threshold angle]. {DS and DP} For the identity test of a 0-mg formulation (placebo), it may be necessary to document the absence of drug substance, and an LOQ determination will then be required. {only DP} Dissolution Accuracy (mean) (drug Recovery 95105% product) Srel for recovery 2.5% Precision > Repeatability Srel 2.0%, n 6 {at Q time} > Intermediate precision Project specific. Linearity n 6 > Correlation coefficient r 0.990 > y-intercept (absolute value) 5% > Residual standard deviation 2.5% Stability of solutions > Sample 2.0% change over specified time > Reference standard 2.0% change over specified time Specificity > HPLC No interference from placebo solution at the retention time of API. Range (basket/paddle) IR: 30% of specified range MR,SR: From 50% of Q-value to 130% of label claim. Content Precision As defined in assay uniformity Accuracy (CU) Stability of solutions Drug product Specificity Chromatographic peaks are separated. No indication of interference from placebo solution at the retention time of API. Linearity n 6 > Correlation coefficient r 0.990 > y-intercept 5.0% > Residual standard deviation 2.0% Range At least 70130% of declared content Assaydrug Accuracy (mean)DP product Recovery 98.0102.0% > Srel for recovery 2.0%, n 9 (at least three concentrations) 462 METHOD VALIDATION TABLE 9-3. Continued Quality Characteristics Parameter to be Validated Acceptance Criteria Weight AccuracyDS percent Comparison of methods % difference of the mean of two drug (i.e., titration, DSC, PSA) methods 2.0% substance Precision > Repeatability Srel 2.0%, n 6, DP Srel 1.0%, n 6, DS > Intermediate precision Srel 2.0%, n 4 [when combined from two analysts] Linearity n 6 > Correlation coefficient r 0.998 > y-intercept 2.0% > Residual standard deviation 2.0% Stability of solutions > Sample 2.0% change over specified time (DP) > Reference standard 2.0% change over specified time (DP) > Sample 1.0% change over specified time (DS) > Reference standard 1.0% change over specified time (DS) Specificity > HPLC Chromatographic peaks are separated. No indication of interference from placebo solution at the retention time of API. No indication of another peak under the API peak. Range At least 80120% of declared content (100% = concentration X of final sample stock solution) Ruggedness/robustness 1.0% difference for a defined range of intentionally altering sensitive parameters (pH of mobile phase, column, temperature, flow rate, wavelength, etc.) Drug product- Precision Related Repeatability Level < 0.1%, Srel 30%, n 6substances Level 0.1 <0.2%, Srel 20%, n 6(degradation Level 0.2 <0.5%, Srel 10%, n 6products) Level 0.5 <5%, Srel 5%, n 6Level 5%, Srel 2.5%, n 6Drug substance Intermediate precision Level < 0.1%, Srel 40%, n 4(synthetic by- [all replicates combined Level 0.1 <0.2%, Srel 30%, n 4products and from two analysts] Level 0.2 <0.5%, Srel 15%, n 4degradation Level 0.5 <5%, Srel 7.5%, n 4products) Level 5%, Srel 4.0%, n 4Specificity Known peaks are separated. > HPLC No indication of interference from placebo solution at the retention time of API. No indication of another peak under the API peak. ASSIGNMENT OF VALIDATION PARAMETERS 463 TABLE 9-3. Continued Quality Characteristics Parameter to be Validated Acceptance Criteria Linearity n 6 > Correlation coefficient r 0.990, DP and r 0.998, DS > y-intercept Level < 0.5%: 25% Level 0.5 <1%: 10% Level 1%: 5.0% > Residual standard deviation Level < 0.2%: 20% Level 0.2 <0. 5%: 10% Level 0.5 <5%: 5.0% Level 5%: 2.5% Range LOQ to 120% of specification limit of largest impurity or related substance LOD Peak signal/noise ratio 3 : 1 LOQ Peak signal/noise ratio 10 : 1 and Srel 10%, n 5Accuracy (mean) Level < 0.2%: 70130% > Recovery Level 0.2 <0.5%: 80120% Level 0.5 <5%: 90110% Level 5%: 95105% > Srel for recovery Level < 0.5%: 10%, Level 0.5 <5%: 5% Level 5%: 2.5% For all, n = 9 (at least three concentrations), a weighted average maybe used based on the level and the Srel .Stability of solutions [report two decimal places] > Reference standard Level < 5% of theoretical 100% Change 10% concentration over specified time Level 5% of theoretical 100% Change 2.0% concentration over specified time > Sample Related substances (impurities) Level < 0.5% Change 20% over specified time Level 0.5 <5% Change 10% over specified time Level 5% Change 5% over specified time No new peak reporting level Ruggedness/robustness Defined based on an experimental design and data (sensitive parameters and a range for each parameter in the final test method) if the overall purity in both of the methods would still be the same according to a predefined set of acceptance criteria. Different types of separation methods could also be used to show accuracy. For example, if normal-phase chromatography was used as the parent method, this could be compared to a separation obtained using supercritical chromatography. In another example, an electrophoretic method using capillary electrochromatography or capillary electrophoresis could be compared to an HPLC separation. Also, the HPLC weight percent (assay) method of the drug substance can be compared to nonchromatographic methods such as nuclear magnetic resonance (NMR) [21], Phase Solubility Analysis (PSA), and DSC [22] and to nonspecific titration and spectrophotometric assay methods that may have been used in early development before the qualification of a reference stan-dard. Potentiometric titration methods using nonaqueous or aqueous titrations are only amenable to ionizable compounds and are nonspecific because the impurities may contain the same ionizable functionality as the parent com-pound being titrated. Titration is a nonspecific method because synthetic by-products in drug substance may have a p Ka similar to that of the main component (the endpoints for the by-products and the drug substance may overlap in this case) and results may be biased, leading to a higher weight percent of the material. However, these titration methods can be used in early development when a reference standard is not available. Also, the spectrometric-based assay methods such as ultraviolet (UV) may be nonspecific because most of the drug substance impurities contain a similar chromophore as the parent molecule. If UV is used, UV absorption is measured at one or more wavelengths and the absorbance value is recorded for a particular concentration. Sandor Gorog has critically evaluated the difference between specific and nonspecific assay methods in the European and US Pharmacopoeias [23]. The difference between the mean and the accepted true value with a defined confidence inter-val should be reported in the acceptance criteria. The accuracy can also be demonstrated by recovery of drug substance spiked into a placebo for a drug product. The accuracy can also be demon-strated by recovery of the impurity spiked to a drug substance or into a placebo with drug substance. The percentage recovery with the certain accep-tance criteria at each defined level is reported. Accuracy should be assessed using a minimum of nine determinations at a minimum of three concentration levels covering the specified range (e.g., 3 concentrations/3 replicate preparations of each in the total analytical proce-dure) within the ranges shown in Table 9-4. Accuracy is performed to determine recovery of an active or degradation products from a drug product or recovery of related substances from a drug substance. The experiment is designed to recover the total amount of active or degradation product from a drug product or a specific impurity or impuri-ties from a drug substance. For recovery of the active for assay and CU, a known amount of drug substance in solution is spiked into the placebo blend. The influence of sample preparation steps for tablets must be taken into con- > 464 METHOD VALIDATION sideration such as grinding, sonication, and extraction. For assay determina-tion, the experiment setup is straightforward. A minimum of three concentra-tions (centered around the target concentration) and three replicates are prepared at each concentration (one injection each) to make a total of nine determinations. The minimum three concentrations should be 70%, 100%, and 130% of the target concentration. If this is used, then these accuracy solutions could be used for both content uniformity and assay method if they are indeed the same method. Usually in early development the same method is used. In later development if a new fast CU method is developed ( <5 min), it would have to be revalidated for recovery. In the following theoretical example, for a lyophilized drug substance, a placebo solution is made with the excipients and is diluted with the drug substance X stock solution (2.67 mg/mL) and further diluted to the desired concentration with the proper diluent. The nominal 100% level solution without placebo is used as an external calibra-tion standard. The range explored is shown in Table 9-5. An example of the preparation is shown in Table 9-6. The percent recovery of drug substance X from the nominal 70%, 85%, 100%, 115%, and 130% sample solutions was then determined by an external calibration standard (shown in Table 9-7). The average percent recovery is 99.3% ( n = 15), and at each level the recovery is found to be within the acceptance criteria of 98.0102.0%. For validation of HPLC analytical methods during the early phase, an acceptance criteria of 95105% for recovery may be acceptable if the specifications are outside the range of 95105% (i.e., 90110%).The specifications may never be tighter than the acceptance criteria. Moreover, the Srel [%RSD (relative standard devia-tion)] determined from the overall percent recovery is 0.18% ( n = 15) and passes the acceptance criteria of Srel 2.0%. For degradation products, the same procedure as the assay can be followed, except the amount should be from reporting level to at least 120% of the > ASSIGNMENT OF VALIDATION PARAMETERS 465 TABLE 9-4. Minimal Concentration Ranges for Accuracy Test (Wider Range May Be Used) > Type of Analytical Procedure Range, at Minimum, to Be Covered Assay (content) 80120% of declared content Assay (CU) 70130% of declared content Assay (dissolution) 30% of specified range (for immediate release dosage form). If the specification for a controlled release product (modified release or sustained release) covers a region from 20% (after 1 hr) to 90% (after 24 hr), the validated range would cover 50% of 1-hr limit (20% 50% =10%) to 130% of the label claim (label claim 1.3). Degradation products/impurities Reporting level to at least 120% of specification limit. 466 METHOD VALIDATION TABLE 9-5. Range for Recovery Experiment for CU and Assay Method to Be Defined in Method Validation Protocol Target Concentration Number of of Solutions Target Concentration Amount Preparations/Number (%) of Solutions (mg/mL) Injected ( g) of Injections 130 1.30 13.0 3/1 115 1.15 11.5 3/1 100 1.00 10.0 3/1 85 0.85 8.5 3/1 70 0.70 7.0 3/1 TABLE 9-6. Example of Actual Sample Preparation in Method Validation Protocol Milliliters of Stock Actual Final Target Concentration Solution Placebo Volumetric Concentration of Solutions (2.67 mg/mL) Added Flask Used of Solution (%) used (mL) (mg) (mL) (g/mL) 130 12 85.1 25 1281.6 115 11 85.1 25 1174.8 100 a 10 85.1 25 1068 85 8 85.1 25 854.4 70 13 170.3 50 694.2 > a Nominal 100% level without placebo is used as a calibration standard. TABLE 9-7. Recovery Results for Assay and Content Uniformity Method Actual Average Actual Concentration Concentration Average Sample Name (g/mL) (g/mL) % Found % Found % Recovery 70%-Recovery1 694.2 694.2 69.4808 69.5036 99.29 70%-Recovery2 694.2 69.5307 70%-Recovery3 694.2 69.4992 85%-Recovery1 854.4 854.4 84.3520 84.5353 99.45 85%-Recovery2 854.4 84.5062 85%-Recovery3 854.4 84.7478 100%-Recovery1 1068 1068 99.2699 99.1599 99.16 100%-Recovery2 1068 99.0610 100%-Recovery3 1068 99.1488 115%-Recovery1 1174.8 1174.8 114.566 114.4453 99.52 115%-Recovery2 1174.8 114.3211 115%-Recovery3 1174.8 114.4488 130%-Recovery1 1281.6 1281.6 128.6508 128.8411 99.11 130%-Recovery2 1281.6 128.8216 130%-Recovery3 1281.6 129.0509 % Average recovery = 99.31 SD 0.18 %RSD 0.18 specification limit. Note that the reporting level can never be lower than the limit of quantitation (LOQ) of the method. However, during early-phase val-idation for drug products, if authentic degradation products are not available, then low amounts of API are added (LOQ to 120% specification limit of largest impurity) to the placebo and the recovery experiment is performed. Once degradation products of known purity become available (isolated or synthesized), a spiked recovery experiment should be performed. For drug substances these may be directly spiked into the drug substance (DS), and for the drug product (DP) these may be spiked into the DS + placebo. This spiked experiment is conducted to determine whether a sample preparation proce-dure is able to completely extract active and degradation products from the sample matrix. For drug substances, a known amount of spiked impurities (authentic samples) is added to the active pharmaceutical ingredient and the recovery experiment is performed. The purity (A% water residual solvents inorganic impurities) of the impurity that will be spiked must be known as well in order to calculate the actual amount added to the respective DS so that the theoretical amount of the impurity that would be in the solution can be determined. This is because the API may have some amounts of the same known degradation products may already present in the API drug substance. This must be accounted for in the calculation. Therefore, the amount of the impurity that is present in the matrix (DS) must be known. The total of the spiked amount of impurity and the amount of impurity that is present in the drug substance must be used to determine the overall amount of impurity. Also, for drug products when authentic degradation products are added to placebo in presence of API, the purity factor of the isolated degradation product that is spiked needs to be taken into account. In the example shown in Table 9-8 and Table 9-9, a recovery experiment is performed for a drug substance that has 0.2% (area percent normalization) of > ASSIGNMENT OF VALIDATION PARAMETERS 467 TABLE 9-8. Spiked Recovery Experiment > Target concentration 1100 mg in 100 mL of DS (mg/mL) diluent Stock solution of 0.1001 11 mg in 100 mL impurity A (mg/mL) diluent Purity of impurity A 91% Concentration Spiked % of Spiked of DS Impurity target Amount DS (mg) mL Stock A mL Diluent (mg/mL) (1 mg/mL) 100 199 0.001001 0.1001 100 298 0.002002 0.2002 100 496 0.004004 0.4004 impurity A. The specification limit for this impurity is 0.3%, and 120% of the specification limit is 0.36%. Therefore a recovery experiment will be pre-formed where 0.1%, 0.2%, and 0.4% of impurity A will be spiked into the DS in solution. A stock solution of impurity A can be made. Depending on the desired percent of the impurity to be spiked in the DS, an aliquot of stock solution A is added to the 100 mg of the DS and then diluted to volume with the diluent. The purity factor of impurity A must be taken into account. In the following example, 11 mg was multiplied by the 0.91 (purity factor of the impu-rity) to give a total of 10 mg of impurity A. This spiked amount is added to the total (known as the theoretical overall total) as shown in Table 9-9. The actual overall is the percent of the impurity determined by HPLC analysis. Then the percent recovery can be determined (actual overall/theoretical overall) 100. 9.4.1.1 Filter Check. If for the drug product the sample preparation proce-dure (recovery procedure) requires filtering the sample solution prior to the solution being injected into an HPLC system, then a check for adsorption of the components onto the filter membrane must be performed. The experiment should be set up to conduct the filter step and centrifuge on the same solu-tion. So, for the same solution (reference standard solution as well as a sample solution), an aliquot of solution is passed through a membrane filter and col-lected after 2, 4, 6, 8, and 10 mL. In addition, the same solution (not filtered) is centrifuged and supernatant is collected. All solutions are then injected on a HPLC system. Since the identical solution has gone through different paths, the peak areas from the chromatogram should be identical (with some accept-able variability due to injection precision of the analytical instrumentation). If there is no change in peak areas between centrifuged and filtered solutions, then it can be stated that the membrane for that particular filter does not cause adsorption of the analyte(s). If the peak areas between centrifuged and fil-tered solution are different (filtered solution shows smaller peak areas than the centrifuged solution), then it can be stated that the membrane in that par-ticular filter is adsorbing the analyte(s). However, sometimes an increase in peak areas is observed as greater volumes are passed through the filter (e.g., 26 mL). Therefore, the minimum volume that needs to be passed through the filter to get constant peak areas that are comparable to the centrifuged peak areas must be determined. If those areas are within (2.0%), then the desig-nated amount of volume that is needed to pass through the filter before the > 468 METHOD VALIDATION > TABLE 9-9. Spiked Recovery Results > Impurity A Impurity A Theoretical Actual Recovery % in Matrix % Spiked Overall Overall Overall 0.2 0.1 0.3 0.31 103.3% 0.2 0.2 0.4 0.42 105.0% 0.2 0.4 0.6 0.59 98.3% solution can be collected for HPLC analysis must be noted. These types of filter experiments must be performed every time the drug product formula-tion has changed (change in excipients and/or ratios of excipients/DS). Also, even the same membrane filter type from different vendors can give different results due to changes in the housing of that particular membrane filter, and these should also be investigated. 9.4.1.2 Completeness of Extraction. For drug products containing con-stituents that are insoluble in the extraction medium used in the analytical procedure, it may be deemed adequate to perform a separate test for com-pleteness of extraction (in addition to recovery experiments as described above). The completeness of extraction can be evaluated two ways: kinetically (over some elapsed time t) and thermodynamically (change in volume). For time experiments, when the initial recovery experiment is completed (when using a real drug product, not from spiked experiments), the solution that is left over is set aside for time t (usually 24 hours at a temperature con-dition known not to affect the inherent stability of the solution). After time t,the solution is re-shaken by hand and then re-injected to determine assay value. For the volume experiment, the initial recovery experiment is repeated using the actual drug product using an increased volume of sample solvent. For example, if the procedure is stated to extract the content of a drug product with 50 mL of solvent, then further experiments would dictate the use of 75 mL or 100 mL of solvent for this experiment. A generalized procedure to evaluate both kinetic and thermodynamic factors is provided in Table 9-10. This would require that an actual drug product sample is extracted and analyzed as per procedure described in the analytical method with the extraction time of t0 and extraction volume of V0 stated in the method. Six more experiments would be conducted such that longer extraction times t 1, t 2 and t 3 are used and higher extraction volumes V1, V2, and V3 are used. The extraction volumes employed should use the same extraction time specified in the method ( t0). > ASSIGNMENT OF VALIDATION PARAMETERS 469 TABLE 9-10. Generalized Procedure to Evaluate Both Kinetic and Thermodynamic Factors > Experiment Number Extraction Time Extraction Volume 1t0V0 > 2t1V0 > 3t2V0 > 4t3V0 > 5t0V1 > 6t0V2 > 7t0V3 By performing these two experiments (time and volume), it will show whether the initial extraction procedure is sufficient or has any shortcomings. If any shortcomings are observed, then a new extraction procedure must be included in the method (i.e., longer time and/or higher amount of extraction solvent). Most likely, for modified release drug products, time is essential (higher recovery is observed over time). Change in volume will usually have an impact in cases where the solubility of an API is on the edge for that par-ticular sample preparation solvent with the excipients present in the matrix. If this is the case, then the procedure should be modified to extract with higher volume of sample preparation solvent. Lastly, to really prove that the sample preparation procedure or the recov-ery procedure is completely extracting the API and degradation products, uti-lization of a homogenizer must be considered. In general, homogenizers are utilized for automated sample preparation procedures in workstations such as TPWII (Caliper Life Sciences, 68 Elm Street, Hopkinton, MA 01748 or www.caliperls.com). Homogenizers are made up of a stainless steel blade that rotates up to 20,000 rpm. The following example illustrates why homogenizers play a vital role in sample preparation. A modified release (MR) product under development gives a drug release profile for at least 8 hours. This corresponds to the release rate of drug sub-stance or the API from the dosage form within 8 hours at 37C in the disso-lution media (apparatus I as described in the United States Pharmacopoeia (USP) 711 at 100 rpm). For example, when a sample preparation procedure was developed for assay and degradation products for a modified release drug product, it was determined that a sonication time of about 30 minutes with about 4 hours of mechanical shaking provided adequate extraction efficiency of the drug from the dosage form. In contrast, if a stand-alone homogenizer was utilized for the same procedure for this dosage form, then the total sample extraction time was found to be about 5 minutes with intermittent stops required by the system software (e.g., TPWII takes about 2 to 5 seconds from end of one pulse to the start of another pulse). The homogenizer provides the energy needed to break the dosage form and to extract the API very efficiently when compared to conventional sonication and mechanical shaking. The final sample preparation procedure was finalized as follows: Two pulses at 8000 rpm for 10 seconds each Six pulses at 15,000 rpm for 15 seconds each Soak/settle time for 2 minutes (to allow all particles to settle to the bottom of the vessel, which allowed for a facile filtration step) 9.4.2 Precision Precision provides an indication of random errors and can be broken down into repeatability and intermediate precision. This procedure should only be performed when the entire analytical method procedure is finalized. > 470 METHOD VALIDATION Repeatability represents the simplest situation and involves analysis of replicates by the same analyst, generally one injection after the other. Repeat-ability tests are mandatory for all tests delivering numerical data. Repeatabil-ity is divided into two parts: injection repeatability and analysis repeatability (multiple preparations) [24]. Validation of the precision of an HPLC method occurs at three stages. The first stage is injection precision (injection repeatability) based on multiple injections of a single preparation of a sample on a particular sample on a given day. The set of criteria is given for area (% area normalization) methods (DS and DP) based on %RSD of peak area. The second stage is analysis repeat-ability where multiple preparations and multiple injections of a sample are analyzed by the same chemist on the same day. The third stage is intermedi-ate precision and is usually performed by different analysts, on a different system, on a different day on the same DP or DS batch to determine the vari-ability of the analytical test. The intermediate precision test may give indica-tions to potential issues that may arise during method transfer. Relative standard deviation or coefficient of variation ( Srel or %RSD) is used to assess if the adequate precision has been obtained. If automation is utilized, then an intermediate precision test is required to compare results obtained through manual testing versus automated testing (if all solvent composition and analyte concentrations of all actives are identical in both methods). 9.4.3 Linearity The purpose of the test for linearity is to demonstrate that the entire analyt-ical system (including detector and data acquisition) exhibits a linear response and is directly proportional over the relevant concentration range for the target concentration of the analyte. It is recommended to perform the linear-ity of the API and related substances independently; and once linearity has been demonstrated, another linearity could be performed containing both API and specific related substance if necessary. For this reason, a stock solution of each substance (API, degradation products, synthetic by-product) must be pre-pared separately (one per solution), and a serial dilution from this stock solu-tion must be injected into an HPLC system (constant injection volume). There are two major reasons to perform a linearity test on each solution indepen-dently. First, each substance may not be pure and the linearity test for each component may become confounded. This is especially true if the active drug substance contains the impurity that linearity is being performed on and/or if the impurity contains the active drug substance as an impurity. Second, when each substance is studied independently, the calculation of relative response factor (RRF) is much easier to determine. The ranges that should be covered for the linearity test are described in Table 9-11. At least five concentrations within the range specified above for the linearity test should be used. When a linearity test is needed for an assay and > ASSIGNMENT OF VALIDATION PARAMETERS 471 degradation products test, it is generally recommended that five or more con-centrations should be utilized to cover the entire range. The focus should be on both (a) the lower limit (LOQ to 1.0%) for the degradation products and (b) the higher limit (80120%) for the assay of the active. If the assay method is also used for content uniformity, the range 80120% should be expanded accordingly to 70130%. As stated in the recovery section, if authentic degradation products or impurities are not available, then an API may be utilized to perform the lin-earity test at the lower concentration range (reporting level to 1.0%). In addi-tion for drug products, if assay and degradation products are calculated from a single 100% reference standard solution (mass percent) or area percent nor-malization, then two independent linearity tests must be generated to demon-strate linearity at the degradation product level as well as at the assay level. Hence, if the 100% standard is used to quantitate the levels of degradation products, the slopes of the low-level linearity and the high-level linearity curves must be compared. If the criteria for agreement between the two slopes are not met, then for quantitation of the degradation products a lower con-centration of standard (usually between 0.5% and 5.0%) is used to calculate the degradation products (related substances). This will be further discussed in the next section. Acceptability of linearity data is often judged by examining the correlation coefficient and y-intercept of the linear regression line for the area response versus concentration plot and residual standard deviation (standard error compared to the calculated y-value at a certain target % level). Correlation coefficients of >0.990 (DP) or 0.998 (DS) are generally considered as evidence of acceptable fit of the data to the regression line. The y-intercept and %RSD acceptance criteria for DP and DS depend on the linearity range being tested, > 472 METHOD VALIDATION > TABLE 9-11. Recommended Ranges for Linearity Tests > Type of Analytical Procedure Range to be Covered > Drug Substance > Weight percent 80120% of target concentration Impurities LOQ or reporting level to at least 120% of specification > Drug Product > Assay (content) 80120% declared content Assay (CU) 70130% declared content Assay (dissolution rate) 30% of specified range Impurities/degradation products LOQ or reporting level to at least 120% of specification Assay and degradation products LOQ or reporting level to 120% of assay content [only if 100% reference standard is utilized to calculate low level of degradation products] and the proposed criteria are shown in Table 9-3.Although these are very prac-tical ways of evaluating linearity data, they are not true measures of linearity [25, 26]. The coefficient of correlation can be subject to misinterpretation and may give a misrepresentation of linearity, since different datasets can yield identical regression statistics [27, 28]. The parameters, correlation coefficients, y-intercept, and %RSD by themselves can be misleading and should not be used without a visual examination of the response versus concentration plot [29]. A more statistically sound approach to examine linearity would include examining the residuals from a linear regression. The residuals are the dis-tances of the experimental points from the fitted regression line, measured in a direction parallel to the response axis. Analysis of the residuals provides further support that the calibration curve would be deemed linear if the resid-ual response shows a normal distribution with a zero mean [30]. Although correlation coefficients of the linear regression can be >0.99, the plots of the response factor versus the concentration can shed light if there are any appar-ent deviations from linearity.A slope close to zero (response factor versus con-centration) would indicate that a linear response is obtained over the specified concentration range. An additional acceptance criterion that could be consid-ered is that the response factor will show %RSD of 2.0% across all concen-tration levels between 80% and 120% of the target concentration (assay). Also, this %RSD acceptance criterion could be applied to the low-level lin-earity regions such that the response factor will show %RSD of 10% across all concentration levels between LOQ to 120% of impurity specification level. Additionally, this comparison of the response factor can be used to help justify the LOQ above and beyond the typical S/N >10, injection precision require-ments, and low-level linearity requirements.A simple test would be to compare the response factor difference between the proposed LOQ and the 5 LOQ value concentration and also to compare the response factor difference between the proposed LOQ and the maximum concentration tested in the low-level linearity experiment. Both of these percent difference values should show 10% difference to provide additional support for qualifying the pro-posed LOQ as the official LOQ for the method. 9.4.3.1 Linearity Example (Assay and Content Uniformity). An example for linearity for Assay and Content Uniformity is given. The target concen-tration is 1.0 mg/mL for this particular drug substance D. Table 9-12 shows the table that could be included in a method validation protocol stating the con-centrations that will be tested from 50130% of the target concentration. The sample preparation procedure is indicated in Table 9-13. The linearity results and the relative response factors at each concentration are shown in Table 9-14. The response factor is calculated by peak area divided by concentration at each concentration level. A typical graph for linearity is obtained such that the concentration is plotted on the x axis and the area counts are plotted on the y axis Figure 9-1. The %RSD of 0.51% for the calculated response factors at all concentrations is reported in Table 9-14. In general, %RSD should be > ASSIGNMENT OF VALIDATION PARAMETERS 473 less than 2.0% for assay methods (80120% of target). A plot of response factor versus concentration is shown in Figure 9-2. The near-zero slope (0.03) of the response factor plot indicates that a linear response is obtained over this concentration range. Table 9-15 shows the regression analysis performed by Excel using the available Add-In functionality ToolPak. In Table 9-15, the y value using the 100% standard is calculated ( y = mx + b), where x is the con-centration of the 100% standard. The %RSD (0.51%) is calculated as well and is defined as the standard error/ y. In order to identify if there are significant deviations from the assumed linearity, an investigation of the residuals should 474 METHOD VALIDATION TABLE 9-12. Linearity for Assay for Drug Product in Method Validation Protocol Prepare and analyze solutions at the following levels: Target Concentration of Solutions Concentration (%) (mg/mL) Amount Injected ( g) Number of Injections 130 1.30 13.0 2115 1.15 11.5 2100 1.00 10.0 285 0.85 8.5 270 0.70 7.0 250 0.50 5.0 2 > Criteria: Linearity Correlation coefficient: r 0.998. > y-intercept: 2.0% when compared to the calculated y-value at the 100% level. Residual standard deviation: 2.0% (standard error compared to the calculated y-value at 100% level). TABLE 9-13. Example of Actual Sample Preparation for Compound D in Method Validation Protocol Stock Target Concentration Solution a Added Sample Final Concentration Solution of Solutions Used Solvent of Solution Number (%) (mL) (mL) (g/mL) 1 130 12 25 1281.6 2 115 11 25 1174.8 3 100 b 4 10 1068 4 85 8 25 854.4 5 70 13 50 694.2 6 50 9 50 480.6 > aStock solution concentration of active X is 2.67 mg/mL. > b100% level without placebo is used as a calibration standard. ASSIGNMENT OF VALIDATION PARAMETERS 475 TABLE 9-14. Linearity Results (Assay and Content Uniformity) Concentration Response Factors (g/mL) Peak Area Average (RF) 480.6 5,040,447 10,487.8 480.6 5,044,555 5,042,501 10,496.4 694.2 7,282,317 10,490.2 694.2 7,269,300 7,275,808.5 10,471.5 Avg RF: 10495.5 854.4 8,910,381 10,428.8 854.4 8,926,904 8,918,642.5 10,448.2 SD (RF): 53.6 1,068 11,323,083 10,602.1 1,068 11,319,630 11,321,356.5 10,598.9 %RSD (RF): 0.51 1,174.8 12,345,022 10,508.2 1,174.8 12,309,278 12,327,150 10,477.8 Slope: 0.03 1,281.6 13,427,476 10,477.1 1,281.6 13,404,485 13,415,980.5 10,459.2 Figure 9-1. Assay and CU linearity from 50% to 130% of the target. Figure 9-2. Response factor versus concentration ( g/mL) at the assay level. be performed; this is demonstrated in Figure 9-3. The residuals should be ran-domly distributed around a true mean of zero. The plot of residuals is usually plotted versus the x values (concentration in this example). If a linear, unweighted regression model is used, the residual plot should show random behavior in a constant range, and it should not be biased in one direction or the other or should not show a systematic or curved pattern of the residuals 476 METHOD VALIDATION TABLE 9-15. Regression Analysis Using Analysis ToolPak Add-In Functionality from Excel Summary Output Regression Statistics Multiple R: 0.999838398 R square: 0.999676821 Adjusted R square: 0.999644503 Standard error: 57705.10759 Observations: 12 ANOVA df SS Regression: 1 1.03002E +14 Residual: 10 33298794422 Total: 11 1.03035E +14 Coefficients Standard Error Intercept = b 16218.49184 57793.36424 X Variable 1 = m 10515.47644 59.78890334 x = 1068 g/mL Y = 11214310.34 y = mx + b at 100% level Y-intercept (abs. value) = 0.14 (Intercept/ y)*100 %RSD = 0.51 (Std error/ y)*100 Figure 9-3. Residual plot where X Variable 1 is the concentration of the API in solution. [3133]. The following acceptance criteria have been met as shown in Table 9-16 for the linearity part of this validation. 9.4.3.2 Linearity Example (Related Substances, Low-Level Linearity). An example for linearity of related substances is given. The target concentration is 1.0 mg/mL for compound D (drug substance). The maximum specification limit of a particular impurity is 7.0%. Therefore the linearity range should be from LOQ to at least 120% of maximum specification limit, which would be from 0.4% to 9.0% of target concentration. Table 9-17 is the table that could be included in a method validation protocol stating the concentrations that will be tested are from 0.4% to 9.0% of the target concentration with the > ASSIGNMENT OF VALIDATION PARAMETERS 477 TABLE 9-16. Comparison of Results to Acceptance Criteria Parameter Acceptance Criteria Results Linearity > Correlation coefficient r 0.9998 0.999 (Table 9-15) > y-intercept 2.0% 0.14% (Table 9-15) > Residual standard deviation 2.0% 0.51% (Table 9-14) Range At least 70130% a 50130% (100% is 1.0 mg/mL) > aThis range is used because this linearity is for both assay and content uniformity. TABLE 9-17. Linearity for Related Substances for Compound D Target % of Target Concentration Target Amount Injected Number of Solutions (mg/mL) (g) Injections 9.0 0.09 0.9 27.0 0.07 0.7 25.0 0.05 0.5 23.0 0.03 0.3 22.0 0.02 0.2 21.0 0.01 0.1 20.4 0.004 0.04 6 > aNote that in this theoretical example the specification limit for compound D is set to 7.0% and since no isolated impurity exists the active pharmaceutical ingredient is used for the low level linearity. Criteria: Linearity Correlation coefficient: r 0.998 > y-intercept: 5.0% when compared to the y-value at the 1.0% level. Residual standard deviation: (0.49.0%): 4.6% a; calculation based on 1.0% level standard LOQ: S/N 10:1, Srel 10%, n5 > aJustification for acceptance criteria is set based on the weighted value from 7 levels: Criteria set at %RSD 2.5% for 5.0, 7.0 and 9.0% Criteria set at %RSD 5% for 1.0, 2.0 and 3.0% Criteria set at %RSD 10% for 0.4% proposed acceptance criteria. The linearity results and the relative response factors at each concentration are shown in Table 9-18. The graph for linearity where area response was plotted versus analyte concentration is shown in Figure 9-4. The response factors for all concentrations in the low-level region is calculated and reported in Table 9-18 (%RSD = 3.8%). In general, %RSD should be less than 10.0% for related substance methods in the low-level lin-earity region.A response factor versus concentration plot is provided in Figure 9-5. The slope ( 10) of the response factor plot indicates that there is some deviation of linear response toward the lower end of the concentration range studied. An additional test was conducted where the response factor differ- 478 METHOD VALIDATION TABLE 9-18. Linearity Results for Related Substances % of Target Concentration Response (1.0 mg/mL) (g/mL) Peak Area Factors (RF) 0.3738 3.738 35,860 9,593.365436 0.3738 3.738 35,941 9,615.034778 0.3738 3.738 35,749 9,563.670412 0.3738 3.738 35,210 9,419.475655 0.3738 3.738 35,183 9,412.252541 Avg. RF (0.4%) 0.3738 3.738 35,237 9,426.698769 LOQ: 9,505.1 1.068 10.68 107,181 10,035.67416 1.068 10.68 107,194 10,036.89139 2.136 21.36 214,144 10,025.46816 Avg. RF (2.1%) 2.136 21.36 213,809 10,009.78464 5XLOQ: 10,017.6 3.204 32.04 328,167 10,242.41573 3.204 32.04 328,013 10,237.60924 %Difference (Avg. RF) 5.1264 51.264 530,457 10,347.55384 (0.4%:2.1%): 5.4 5.1264 51.264 532,297 10,383.44647 6.408 64.08 662,721 10,342.08801 6.408 64.08 661,735 10,326.701 8.544 85.44 890,138 10,418.28184 Avg. RF (8.5%): 8.544 85.44 889,655 10,412.62875 10,415.5 Overall Avg. SD (RF): %RSD (RF): %Difference (Avg. RF) RF: 9,991.6 381.1 3.8 (0.4%:8.5%): 9.6 Figure 9-4. Low-level linearity from 0.4% to 9% of the target. ence between the proposed LOQ (0.4%) and 5 the proposed LOQ value con-centration (2.1%) was determined, and a value of 5.4% was obtained which was deemed acceptable ( 10% is acceptable). The response factor difference between the proposed LOQ (0.4%) and the maximum concentration tested (8.5% of target) in the low-level linearity experiment was determined, and a value of 9.6% was obtained which was deemed acceptable. Both percent dif-ference values of 5.4% and 9.6%, respectively, showed less than a 10% dif-ference and could be considered to be acceptable (Table 9-18). However, it would be recommended to use a low-level standard of 1.0% to quantitate impurities in the region of 0.4% to 9.0% because dilute standards of greater than 1.0% would lead to underestimation of impurities below 1.0%. Table 9-19 shows the regression analysis performed by Excel using the available Add-In functionality ToolPak. In Table 9-19, the y value of 4.6% was calculated using the 1.0% standard ( y = mx + b), where x is the concentration of the 1.0% standard and acceptance criteria given in Table 9-17 of 5.0% was met. In Table 9-19, the %RSD (standard error/ y) was calculated as 2.1%, and this had met the acceptance criterion of 4.6% given in Table 9-17. In order to identify if there are significant deviations from the assumed linearity, an investigation of the residuals should also be performed; this is demonstrated in Figure 9-6. The residuals were randomly distributed around a true mean of zero. The acceptance criterion for the % RSD is calculated as follows: 9.4.3.3 Low-Level Linearity Versus High-Level Linearity: When Do You Use a Dilute Standard to Calculate Impurity Levels? There are two ways to determine the related substances: one in terms of mass percent using a refer-ence standard (at 100% level of API or dilute reference standard) and the 3 2 5 3 5 1 10 77 5 15 10 732 5 7 4 6 ( ) + ( ) + ( ) = + + = = . % % % . . . % > ASSIGNMENT OF VALIDATION PARAMETERS 479 > Figure 9-5. Response factor versus concentration ( g/mL) at the related substance level. second using area percentages (area normalization). The percent of each impurity using either approach should be the same, assuming that the impu-rities have the same response factor as the API and there is no deviation from linearity throughout the whole concentration range from LOQ to 130% of API. The impurities can be calculated against a 100% standard of the API in 480 METHOD VALIDATION TABLE 9-19. Regression Analysis Using Analysis ToolPak Add-In Functionality from Excel Summary Output Regression Statistics Multiple R: 0.99997499 R Square: 0.999949981 Adjusted R square: 0.999946855 Standard error: 2216.504059 Observations: 18 ANOVA df SS Regression: 1 1.57145E + 12 Residual: 16 78,606,243.91 Total: 17 1.57153E + 12 Coefficients Standard Error Intercept = b 4,871.508798 770.5743113 X Variable 1 = m 10,443.38585 18.46542508 x = 10.68 for the 1% level 1% Y = 106,663.8521 y = mx + b at 1% level Pass Y-intercept = 4.57 (Intercept/ y)*100 Pass RSD = 2.08 (Std. error)*100 Figure 9-6. Residual plot where X Variable 1 is the concentration of the API in solution. terms of mass percent only if there is no deviation from linearity throughout the whole concentration range; otherwise, impurities can be overestimated or underestimated. Two separate linear regression calculations for the two dif-ferent concentration ranges should be performed: (1) lower range from report-ing level to 120% of impurity specification and (2) upper range from 70% to 130% of assay specification. If impurities are calculated with respect to a 100% reference standard solu-tion or based on area percentages, an additional criterion should be set such that the difference in slopes of upper range and slopes of lower range of the linearity curves is 4.0%. Also, the response factors from the lower-range lin-earity curve and the upper-range linearity curve should be compared and the overall %RSD should be 5.0% (this can be dependent on weighted average of the acceptance criteria for %RSD of the response factor at the low- and high-level linearity regions). In the event that this criterion is not met and there is deviation from linearity in the low-level region the impurities should be calculated against an appropriate dilute reference solution of the drug sub-stance that is comparable to the level of impurities that are typically present. Consider the two examples previously described for linearity at the assay and related substance levels shown in Tables 9-12 to 9-19 and in Figures 9-1 to 9-6. Comparison of the slopes shows a percent difference of less than 1%. The %RSD of all the response factors from 0.4% to 130% of the target concentration is 3.8% (Table 9-20), which is less than the 5.0% proposed acceptance criteria for %RSD of the response factors across the whole concentration range. Also, Figure 9-7 shows that when the response factor was plotted against all the analyte concentrations used in both linearity curves from 0.4% (3.738 g/mL) to 130% (1282 g/mL) of target, there is a deviation from linearity in the low-level analyte concentration region. Even though all acceptance criteria were met for linearity in both regions, it would be recom-mended not to use the 100% standard to quantitate impurities and not to use area normalization but instead to use a dilute standard to calculate the mass % of impurities. It would be recommended to use a low-level standard of 1.0% to quantitate impurities because standards of greater than 1.0% would lead to underestimation of impurities below 1.0%. The use of a low-level standard is generally recommended because it is an excellent way to determine if there is a mass balance discrepancy. This is especially important if the method will be used to release new batches of clinical supplies or when the method is used for supportive stability studies. If a discrepancy in mass balance is observed, this would trigger an investigation to elucidate the root cause for the mass balance issue (i.e., new impurity with different response factor, new impurity with no UV chromophore). 9.4.4 LOD/LOQ Limit of detection (LOD) and limit of quantitation (LOQ) are determined for all impurity tests (including residual analysis during cleaning verification). > ASSIGNMENT OF VALIDATION PARAMETERS 481 482 METHOD VALIDATION TABLE 9-20. Response Factors from 0.4130% of Target Concentration Concentration ( g/mL) Peak Area Response Factors (RF) 3.738 35,860 9,593.4 3.738 35,941 9,615.0 3.738 35,749 9,563.7 3.738 35,210 9,419.5 3.738 35,183 9,412.3 3.738 35,237 9,426.7 10.68 107,181 10,035.7 10.68 107,194 10,036.9 21.36 214,144 10,025.5 21.36 213,809 10,009.8 32.04 328,167 10,242.4 32.04 328,013 10,237.6 51.264 530,457 10,347.6 51.264 532,297 10,383.4 64.08 662,721 10,342.1 64.08 661,735 10,326.7 85.44 890,138 10,418.3 85.44 889,655 10,412.6 480.6 5,040,447 10,487.8 480.6 5,044,555 10,496.4 694.2 7,282,317 10,490.2 694.2 7,269,300 10,471.5 854.4 8,910,381 10,428.8 854.4 8,926,904 10,448.2 1,068 11,323,083 10,602.1 1,068 11,319,630 10,598.9 1,174.8 12,345,022 10,508.2 1,174.8 12,309,278 10,477.8 1,281.6 13,427,476 10,477.1 1,281.6 13,404,485 10,459.2 Avg. overall RF: 10,193.2 SD overall RF: 386.4 %RSD: 3.8 Figure 9-7. Response factor versus concentration ( g/mL) from the related substance level to 130% assay level. The definition of LOD is defined by the USP as a parameter of limit tests. It is the lowest concentration of analyte that can be detected, but necessarily not quantititated, under the stated experimental conditions [USP 1225 Vali-dation of Compendial Methods]. In contrast, LOQ is defined as a parame-ter of quantitative assays for low levels of compounds in sample matrices, such as impurities in bulk drug substances and degradation products in finished pharmaceuticals. The LOQ is the lowest concentration in a sample that may be measured with an acceptable level of accuracy and precision, under the stated experimental conditions [USP 1225 Validation of Compendial Methods]. The LOQ must be lower than or equal to the reporting level or reporting threshold (RTH), as defined in ICH Q3B (R1) guidelines, which is based on maximum daily dose (MDD) for any drug product. For example, if MDD for any drug product is >1 g, then RTH = 0.1%. If MDD is 1 g, then RTH = 0.05%. If the test procedure is not able to attain this limit, the reasons should be discussed on a pure scientific basis with the appropriate health authorities. Each separation method, including HPLC, has a minimal concentration at which it will be able to detect and quantitate with some level of confidence. Statistically speaking, limit of detection (LOD) is defined for a peak that gives a signal-to-noise ratio of about 3 : 1, and limit of quantitation (LOQ) is defined for a peak that gives a signal-to-noise ratio of about 10 : 1. Some analytical chemists may prefer to inject five or six consecutive injections of these solutions and then calculate %RSD of peak area for the peak of inter-est. LOD giving %RSD of 20% and LOQ giving %RSD of 10% are accept-able values throughout the industry and are accepted by health authorities as well. To determine LOD and LOQ, progressive dilutions of the analyte are pre-pared and analyzed. For HPLC, the normal range for LOD/LOQ is between 0.01% and 0.2% for non-peptide/protein-related products; however, for protein/peptides the LOD/LOQ range is typically between 0.1% and 0.5%. From the injected series, a peak is selected whose height hs is about 3 to 10 times larger than the signal noise hn , as defined below: hn = largest deviation (positive or negative) of the detector signal from the average baseline level measured over a span of at least 10 peak widths from the retention time of the target analyte. hn must be measured in the same units as hs . hn may be obtained from the same chro-matogram as hs or from a blank injection (Figure 9-8A) (if other peaks elute within 10 peak widths of target analyte). hs = peak height of the analyte, measured from the average baseline level of the top of the peak, measured in the same units as hn as shown in Figure 9-8B. Wh = peak width at half-height. For HPLC, the LOD and LOQ can be defined (in mass or concentration unit): > ASSIGNMENT OF VALIDATION PARAMETERS 483 where S/N = hs/2 hn and Cs is the amount or concentration of injected analyte. It is also recommended to express the LOD and LOQ in percent of the declared amount of drug substance. Additional criteria for LOQ should be that this concentration be included in the low-level linearity curve and that all criteria for linearity must be met. 9.4.5 Relative Response Factors The relative response factor (RRF) of 1 defines that an impurity and active at identical concentrations have the same analytical response. Generally, no cor-rections for RRF need to be performed if the RRF of particular impurities are between 0.8 and 1.2. If the relative response factors are outside this region, the impurities can be overestimated or underestimated. For example, if impu-rity X has an RRF of 0.5 compared to the active, then it would be underesti-mated; and if impurity Y has an RRF of 1.5 compared to the active, then it can be overestimated. In these cases the RRF must be taken into account when determining the percent of related substances during evaluation of the purity of the DS or the DP. The following equation could be used. (9-1) where PAA i is the peak area of the individual related substance in the test solution, PAA is the peak area of active in the drug product test solution, and RRF is the relative response factor of impurity to active. Also, if the there are two actives in the drug product and the impurities in the drug product from one active are being quantitated versus the other active C PAA PAA PAA RRF i = + # > i 100 1 LOD S N LOQ S N == CC > SS 310 > 484 METHOD VALIDATION > Figure 9-8. (A) Signal noise from blank chromatogram. (B) Peak height measurement for calculation of LOQ. the variant ratio (DS1 versus DS2) must be taken into account. Assume DS1 is the active that the DS2 impurities are being quantitatied against. Therefore, if the DS1 : DS2 ratio was 1 : 2 the variant ratio (VR) of 0.5 would be used. If it was a 1 : 4 variant, then a factor of 0.25 would be used. The following equation can be used to take into consideration the RRF and the VR: (9-2) where PAA i is the peak area of the individual related substance in the test solution, PAA is the peak area of active in the drug product test solution, VR is the variant ratio of DS1 to DS2 (e.g., 1 for the 1 : 0 variant, 0.5 for the 1 : 2 variant, 0.25 for the 1 : 4 variant), and RRF is the relative response factor of DS2 to DS1. 9.4.6 Stability of Solution The objective of this experiment is to demonstrate that the sample and refer-ence standard solutions prepared according to the respective method are stable at least during normal duration of an analytical sequence (it is recom-mended usually to do solution stability at 24, 48, and 72 hours). The accep-tance criteria are given in Table 9-21. It should have already been determined during the method development stage if the diluent is suitable for the sample preparation and the diluent does not react with the active and/or excipients in the matrix. It is recommended to analyze samples and a reference standard solution at intervals between initial timepoint and time t (in hours or days) against freshly prepared reference standard solutions. A sample solution and a reference standard solution are prepared and analyzed immediately. The same solutions are then stored at normal ambient laboratory conditions and at 5C. After time t (hours or days), the solutions initially prepared (bring the 5C solution to room temperature or whatever temperature set in the method prior to injection) are reanalyzed against freshly prepared reference standard solution. The solutions (sample or reference standard) are said to be stable C ii = + # PAA PAA PAA VR RRF 100 1 > ASSIGNMENT OF VALIDATION PARAMETERS 485 TABLE 9-21. Acceptance Criteria for Stability of Solution > Test Type Acceptance Criteria Assay, CU or dissolution 2.0% Weight percent (DS) 1.0% Related substances/area% 1.0% Degradation product/ Level <0.5% Change 20% over specified time impurities Level 0.5 <5% Change 10% over specified time level 5% Change 5% over specified time No new peak reporting level for time t when stored at certain conditions when the acceptance criteria are met. For drug substances it is acceptable to compare the peak area response at initial timepoint and at the defined time t (generally 24, 48, and 72 hours), assuming that the HPLC system is running, system suitability is run daily, and original mobile phase is used for all analysis (i.e., to avoid any potential change in the analyte ionization due to slight variation of mobile-phase pH). The change in response is reported and compared to initial. For early-phase vali-dation for drug product a similar approach can be used such that the refer-ence solution (standard) and the drug product are re-injected at initial timepoint and then re-injected at time t. The change in response is reported and compared to initial. The driver for determining solution stability is to determine the length of time the samples are stable in the respective diluents. For example, if a run sequence is for 24 hours and the solutions are only stable for 12 hours, the later samples analyzed after 12 hours will lead to inaccurate results. At times the solution stability in the respective diluent will be greater at 5C than at room temperature; therefore, solution stability at room temperature and refrigerator conditions must be evaluated. This may be beneficial in the case when there are instrument problems and the entire sequence needs to be rerun. With proper determination of the solution stability, the original solu-tions prepared could be run on a suitable HPLC without having to re-prepare the solutions. Also, for some compounds the solutions may be stable for x number of days regardless of temperature in which they are stored; however, they may be prone to photocatalyzation in the diluent, so proper storage con-ditions may need to be defined. In this case, three solutions may be used for the solution stability experiment: control (room temperature/protected from light, amber volumetric flask), room temperature (clear flask), refrigerator (clear flask). Depending upon the outcome, a comment should be added to the analyti-cal method stating the duration within which the solutions remain stable (e.g., the time within which the acceptance criteria are met). 9.4.7 Ruggedness/Robustness The definition in regard to ruggedness given by USP (USP 1225 Validation of Compendial Methods) is as follows: The ruggedness of an analytical method is the degree of reproducibility of test results obtained by the analy-sis of the same samples under a variety of normal test conditions, such as different laboratories, different analysis different instruments, different days, etc. Ruggedness is normally expressed as the lack of influence on test results from the operational and environmental variables of the analytical method. Ruggedness is a measure of reproducibility of test results under normal expected operational conditions from laboratory-to-laboratory and from analyst-to-analyst. Practically speaking, ruggedness is another name for inter-mediate precision, where two analysts, from two different laboratories, on two > 486 METHOD VALIDATION different days, utilizing different instrumentations, lot numbers of columns, reagents, solvents, and chemicals, follow the identical test method to test the identical sample. Generally, it is not considered necessary to study these effects individually, and the use of an experimental design (matrix) is encouraged. The obtained results from both analysts are grouped together to determine whether this additive precision is acceptable or not. For example, if each analyst prepared two sample preparations API at target concentration for intermediate precision, then a total of four values are pooled together (addi-tive precision) as stated in Table 9-3 of Assay, Precision; Srel 2.0%, n 4. In addition to an additive precision requirement, some laboratories also include an acceptance criterion (for example, absolute mean difference 2%) for mean value. For example, if analysts 1 and 2 prepare three sample preparations each, then additive precision is calculated from a total of six values (three from each analyst). In addition, the mean value obtained by analyst 1 ( n = 3) is compared against the mean value obtained from analyst 2 ( n = 3), in which it must pass an absolute difference (between the two means) of 2.0%. Robustness tests examine the effect that operational parameters have on the analysis results. The following definition of Robustness is taken from the ICH glossary [34]: > The evaluation of robustness should be considered during the development phase and depends on the type of procedure under study. It should show the reli-ability of an analysis with respect to deliberate variations in method parameters. If measurements are susceptible to variations in analytical conditions, the analytical conditions should be suitably controlled or a precautionary statement should be included in the procedure. Robustness is performed on a given test method, but each parameter is deliberately modified one at a time to determine its effect on the final result (e.g., total % degradation products, assay level, etc.). Prior to evaluation of robustness testing, a set of system suitability parameters must be set. This system suitability is a set of acceptance criteria for a particular method. Typical system suitability criteria include ranges for retention time (R t), RRT (relative retention time), apparent efficiency ( N, determined from the peak width at certain peak height; i.e., 5%, 10%, or 50% peak height are the most typical), resolution ( Rs), tailing factor ( Tf, at a certain peak height), asymme-try factor, and so on.Typically, these system suitability criteria and their respec-tive ranges are defined during method development. During the evaluation of robustness testing, the series of system suitability requirements must be met to ensure that the validity of the analytical procedure is maintained whenever used. For the determination of a methods robustness, chromatographic parame-ters are varied within a realistic range and the quantitative influence of the variables is determined. If the influence of the parameter is within a previ-ously specified tolerance, the parameter is said to be within the methods robustness range. Obtaining data on these effects will allow to judge whether > ASSIGNMENT OF VALIDATION PARAMETERS 487 a method needs to be revalidated when one or more of the parameters are changed [35]. For HPLC methods, the parameters that can be deliberately modified are mobile-phase composition, pH of the mobile phase (if applicable), ionic strength of the aqueous portion of the mobile phase, concentration of mobile-phase additive (chaotropic, ion-pairing), gradient slope (if applicable), initial hold time for gradient (if applicable), flow rate, column (different lots and suppliers), column temperature, injection volume, autosampler temperature (if applicable), and wavelength. After changing these parameters, it must be assessed whether the system suitability requirements can be met for the par-ticular HPLC method. Consider the following two examples shown below. 9.4.7.1 Effect of Modifier Concentration. For example, during the method validation for a particular drug product the concentration of the acidic mod-ifier (phosphoric acid) was varied. The target concentration of the acidic mod-ifier was 0.6 v/v% phosphoric acid. The concentration of the acidic modifier was varied from 0.4 v/v% to 0.8 v/v%. At all concentrations studied except for 0.4 v.v% and 0.8 v/v% phosphoric acid, the recommended system suitability requirements were met (Table 9-22). In Figure 9-9, significant changes in reten-tion were obtained for impurity C, which is the penultimate intermediate of the drug substance. It is basic in nature, and this impurity is greatly influenced by the phosphoric acid concentration. This may be attributed to the chaotropic effect described in Chapter 4, Section 4-10. Since the target phosphoric acid concentration is 0.6 v/v%, minor changes in the acid concentration of 0.1 v/v% from the target have no impact on any of the system suitability cri-teria. However, larger changes than 0.1 v/v% of phosphoric acid could lead to system suitability failures. 9.4.7.2 Effect of Variation of Temperature. Another variable that should be studied is the effect of temperature on the separation, especially if the method is sensitive to temperature and/or will be transferred to a manufac-turing facility (receiving laboratory) that resides in an environment that leads > 488 METHOD VALIDATION > TABLE 9-22. System Suitability Parameters Determined as a Function of Concentration of H 3PO 4(v/v%) > System Suitability 0.4% 0.5% 0.6% 0.7% 0.8% Rs >4 between compound A 6.6 6.5 6.2 6.3 6.2 and compound B Rt for compound B 8.29.2 min 8.9 8.9 8.8 8.8 8.8 Tailing factor of compound B 2.3 2.2 2.1 2.0 1.9 (5% peak height) 2.3 RRT compound C =2.12.3 1.9 2.1 2.2 2.3 2.4 to a higher ambient temperature than the transfer laboratory. The target tem-perature of this example method was 21C. It was deemed necessary to choose this temperature in order to obtain the desired resolution between the criti-cal pairs of synthetic by-products/degradation products and meet the other system suitability criteria. A temperature study was conducted from 17C to 25C at 1C temperature intervals. Table 9-23 shows the retention times of a sample spiked with various synthetic by-products and potential forced degradation products analyzed at various temperatures from 17C to 25C. Co-elution of impurities F and G were observed at 25C.At 1718C the reten-tion of compound B was outside of the system suitability requirements. At 1923C, minor changes in selectivity were observed, but all system suitability criteria were met. Also, within the temperature range of 19C to 23C, > ASSIGNMENT OF VALIDATION PARAMETERS 489 Figure 9-9. Effect of acidic modifier concentration on the analyte retention. TABLE 9-23. Retention Times of Compound B, Synthetic By-products and Degradation Products Peak Name 17C 18C 19C 20C 21C 22C 23C 24C 25C A 5.5 5.6 5.8 5.8 5.9 6.0 6.0 6.1 6.2 B 8.0 8.1 8.2 8.4 8.5 8.6 8.7 8.8 9.0 D 10.7 10.8 11.0 11.1 11.3 11.4 11.6 11.7 11.9 E 11.2 11.4 11.6 11.7 11.9 12.1 12.3 12.5 12.7 F 12.4 12.6 12.9 13.0 13.2 13.5 13.7 14.0 14.6 a G 14.3 14.3 14.4 14.4 14.4 14.5 14.6 14.6 14.6 a H 16.2 16.3 16.5 16.5 16.5 16.7 16.9 16.9 17.0 C 18.7 18.7 18.7 18.6 18.5 18.4 18.4 18.4 18.3 I 25.3 25.4 25.6 25.7 25.8 25.9 26.0 26.1 26.2 > aDenotes co-elution of components. acceptable resolution of synthetic by-products/degradation products from main compound, as well as resolution of critical pairs of impurities, was obtained. When dealing with methods that are temperature-sensitive, addi-tional criteria should be set for resolution between critical pairs of impurities, and the selectivities/retention of the active and related substances should be closely examined. Only within the temperature range of 19C to 23C could all the system suitability criteria shown in Table 9-24 be met.Additional system suitability requirements for resolution between critical pairs of impurities could also be added, especially for impurities F and G (if the authentic impurities are available). The degree of ruggedness and robustness of a particular test method depends on the stage of the drug development process. For drug substance (active pharmaceutical ingredient) and formulation development, both ruggedness and robustness experiments to study a new molecular entity (NME) in very early stages (Phase I clinical trials) would be much more limited compared to later stages in development when the final API synthesis is set and the final market image has been defined. For the first case (phase I clinical trials), very minimal ruggedness/robustness testing may be required. However, this really depends on a companys culture, since it is not a requirement from any health authorities. However, for the second case (after Phase IIa), an exhaustive study varying many critical parameters and multi-laboratory studies would be required. 9.4.8 Specificity An investigation of specificity should be conducted during the validation of identification test, the determination of impurities, and the assay. The proce-dures used to demonstrate specificity will depend on the intended objective of the analytical procedure. It is not always possible to demonstrate that an analytical procedure is specific for a particular analyte (complete discrimina- > 490 METHOD VALIDATION > TABLE 9-24. System Suitability Parameters Determined as a Function of Temperature > System Suitability 17C 18C 19C 20C 21C 22C 23C 24C 25C RS >4 between 6.4 6.6 6.3 6.4 6.4 6.4 6.7 6.6 6.5 compound A and compound B Tailing factor of 2.1 2.1 2.2 2.1 2.1 2.2 2.1 2.1 2.1 compound B (5% peak height) > 2.3 RRT compound 2.35 2.31 2.28 2.22 2.18 2.14 2.11 2.04 2.03 C=2.12.3 RT for compound 8.0 8.1 8.2 8.4 8.5 8.6 8.7 8.8 9.0 B 8.29.2 min tion). In this case a combination of two or more analytical procedures is rec-ommended to achieve the necessary level of discrimination (e.g., for optically active substances, in addition to an achiral HPLC method, a chiral HPLC method may be added). Identity is a general requirement for dosage forms. When determining specificity for identity, the assay and related substances or the content unifor-mity methods can be used. Assay and content uniformity methods are quan-titated by external reference standard. This identity test confirms that the correct active ingredient (s) is present and is present in correct ratio if multi-ple variants are available. The method could also be used for post-packaging analysis. The general requirements are that the sample and standard chro-matograms should correspond in retention time and normalized peak area within 10%. The easiest way to perform specificity for any HPLC method is to perform this test in conjunction with a forced decomposition study. The utilization of mass spectrometry (MS) detector (in series) after a Photo Diode Array (PDA) detector to obtain more information is encouraged (in terms of mass-to-charge ratio of parent ions, initial fragmentation pattern, and peak purity). Specificity is confirmed when an API peak is pure (confirmed by PDA and/or MS) and there is no interference from placebo solution (placebo dis-solved in sample preparation solvent) at the retention time of an API peak. 9.4.9 Forced Degradation Studies (Solid State and Solution)Drug Substance and Drug Product Forced degradation studies are usually performed during the salt selection process for the drug substance (DS). In drug product (DP) development, the forced degradation studies of DS in the presence of excipients are first per-formed during the pre-formulation stage to assist in the selection of the most formidable compounds and excipients. This may lead to the development of more suitable formulations, packaging, and change in storage and manufac-turing conditions as the optimal formulation is defined to be used in clinical studies. Forced degradation testing is often repeated when the final drug substance route and market formulation is defined and when the compound enters phase 3 clinical trials. A good overview of forced degradation testing according to the regulatory guidance documents, with emphasis on what should be considered for late clinical phases and for registration application dossiers (i.e., marketing authorization applications or new drug applications), is provided by the Impurity Profiling Group [36]. Forced degradation studies (sometimes referred to as stress testing) are also performed in order to demon-strate specificity during the development and validation of stability-indicating methods. These studies are usually performed at conditions exceeding that of accelerated storage conditions. Forced degradation studies may provide information in regard to degrada-tion pathways and degradation products that could form during storage of the > ASSIGNMENT OF VALIDATION PARAMETERS 491 drug substance or the drug product. The main goal of forced degradation studies is to effectively produce samples containing representative and realis-tic degradation products. These degradation products should be assessed if they are (a) related to the drug substance or the excipients or (b) due to drug substanceexcipient interactions under certain forced degradation conditions. A delicate balance of efficiency and severity/duration of stress conditions is needed. Overstressing can destroy relevant compounds or generate irrelevant compounds. Understressing may fail to generate important degradation prod-ucts. The extent of degradation targeted should be approximately anywhere from 5% to 10%. The other goal is that the potential degradation products that are generated should be resolved from the active component during development of a stability-indicating HPLC method. The assessment of peak purity using diode array and LC-MS detection are usually employed. These degradation products that are generated during the forced degradation studies can be identified, and the determination of degradation pathways and mech-anisms for the drug substance and drug product can be elucidated. Forced degradation studies are carried out either in the solution state and/or in the solid state. Usually the forced degradation testing is carried out on one batch of drug substance and/or one formulation blend (capsules and tablets). This forced degradation testing should not be part of a formal stability program. The temperature/humidity conditions used may be more severe than the typical accelerated stability testing conditions in order to generate potential degradation products in a reasonable time.The typical forced degradation con-ditions include thermolytic, hydrolytic, oxidative, photolytic (in excess of ICH conditions), high pH (alkaline conditions), and low pH (acidic conditions). Outlined in Table 9-25 and Table 9-26 are some solid-state and solution forced degradation studies, respectively, that could be conducted. In the following > 492 METHOD VALIDATION > TABLE 9-25. Solid-State Forced Degradation Studies > Stress Test Conditions Duration Thermal 50C and 80C 1 wk and 2 wks (closed container) (ambient RH) Thermal/oxidative 50C and 80C 1 wk and 2 wks (open container) (ambient RH) Thermal/humidity 40C/75% RH 1 wk and 2 wks (open container) Light Ambient Maximum 1.2 million lux (closed container) hours and 200 watt hours/square meter Light/oxidative Ambient Maximum 1.2 million lux (open container) hours and 200 watt hours/square meter section we will elaborate on oxidative, hydrolytic, and photolytic forced degra-dation (stress testing). For oxidation, different stressing schemes can be used, and this depends generally on the structure of the drug substance (active component): auto-oxidation, metals, peroxide-mediated, peroxy-mediated, bubbled oxygen, and pressurized oxygen. Auto-oxidation involves a free radical initiator such as AIBN (2,2 -azobisisobutyronitrile) or AMVN (2,2 -azobis(2,4-dimethylvaleronitrile) to initiate oxidation [37] and has been used to mimic long-term room temperature degradation related to oxidation. The concentration > ASSIGNMENT OF VALIDATION PARAMETERS 493 TABLE 9-26. Solution Forced Degradation Studies Test Factor Test Conditions Duration pH 10 mg in 2 mL water 1 day and 3 days a 10 mg in 2 mL of 0.1 N HCl 10 mg in 2 mL of 0.1 N NaOH All in amber volumetric flasks and at room temperature. Oxidation (H 2O2) 10 mg/2 mL 3% H 2O2 1, 2, and 3 days At 5C and room temperature in amber volumetric flasks. If DS is not soluble, then pH modification may be necessary. Oxidation (metal ion 10 mg/2 mL water containing 1, 2, and 3 days catalyzed) 100 ppm Fe 3+, Ni 2+, Cu 2+ saturated with bubbled oxygen in amber volumetric flasks. Oxidation (saturated 10 mg/2 mL saturated with bubbled 1, 2, and 3 days with oxygen) oxygen in amber volumetric flasks. Oxidation (free radical Ratio of DS to AIBN (1 : 10) in 4 hr, 1 day, 2 day initiator : AIBN) MeCN and/or MeCN/water (50 : 50) at 40C, in amber flask. Oxidation (pressurized 10 mg/2 mL with pressurized 1, 2, and 3 days oxygen) oxygen in closed headspace vial (80300 psi), protect from light. Oxidation + light Depending on which oxidative 6 hr, 24 hr stress causes most oxidation, a new solution is prepared and is put in light chamber. Light 50 mg/10 mL water Maximum 1.2 million Ambient lux hours and 200 watt hours/square meter, 6 hr, 1 day, 2 days Heat 10 mg in 2 mL water at 50C 6 hours, 1 day, 2 days > aIf no degradation occurs, consider storing solution in GC oven at 50C for 4 hr. of AIBN that could be explored includes 1 : 1, 1 : 10, and 1 : 100 ratio of DS to AIBN. The typical solvents used include 50% acetonitrile: 50% water and 100% acetonitrile, and the stress study is usually conducted at 40C. Some compounds are very susceptible to transition-metal-catalyzed oxidation, and the API solution can be stressed with various amounts of different transition metals. For peroxide-mediated forced degradation, dilute hydrogen peroxide (0.33%), can be used, but hydroxyl radicals are very reactive but unselective and may overstress the sample. For peroxy-radical-mediated forced degrada-tion, peroxy radicals react with low-energy CH bonds or C C bonds and have been deemed as a more realistic oxidative system. Amines, sulfides, and carboncarbon double bonds are susceptible to oxidation using hydrogen peroxide to give corresponding N-oxides, sulfoxides (or sulfones), and epox-ides, respectively. These functional groups are prone to electrophlilic attack, and the reactions are ionic and do not involve free radicals [38]. Another, more predictive oxidative system would include forced degradation by purging the headspace with oxygen. The typical pressure that could be applied is 80 300 psi. Bubbled oxygen can be used to generate 1O2. Using this type of oxida-tion would be advisable if API possesses certain functionalities such as sulfides, dienes/polyenes, imidazoles, purines, furans, and other heterocycles, since they are susceptible to reaction with 1O2 [3941]. For formulation development, forced degradation studies with 1O2 should be considered, especially for liquid-based formulations (suspensions and solutions). Also, excipients should be carefully evaluated for their ability to generate 1O2 upon h, especially flavors and dyes [42]. More detailed examples about potential degradation pathways related to structure are given in the formulation chapter and in reference 38. Another common pathway of degradation is hydrolysis, which is sometimes called solvolysis because the reaction may involve the pharmaceutical cosol-vent such as ethyl alcohol or polyethylene glycol. The solvents such as water can act as nucelophile and attack the electropositive centers on the active pharmaceutical ingredient or excipients. Most compounds that are susceptible to hydrolysis include compounds that have a labile carbonyl func-tionality such as esters, lactones, and lactams. Also, the substituents near the carbonyl moieties of these compounds can have a dramatic effect on the reac-tion rates, and the substituent groups may exert electronic (inductive and resonance), steric, and/or hydrogen bonding effects that can affect the inher-ent stability of the compounds in the solid and solution state [43]. In order to determine potential hydrolysis degradation products of the API or the drug product, stress testing in water, acid, and base is conducted usually at room temperature protected from light. Note that the reaction rates of hydrolysis reactions can be further catalyzed under the acidic or alkaline conditions. If degradation is not observed under these conditions, the forced degradation studies using the same solutions can be performed at accelerated temperatures (i.e., 4060C). Photolysis is another common degradation pathway because light-sensitive drugs can be affected by either sunlight or artificial light sources such as > 494 METHOD VALIDATION fluorescent light. The different reaction types that can be initiated photo-chemically include, reduction, N-dealkylation, hydrolysis, oxidation, isomer-ization, ring alteration, polymerization, or removal of various substituents like halogens or carboxyl groups (Figure 9-10) [44]. Note the forced degradation studies should be performed both with the API and the drug product (blend) because excipients can initiate, propagate, or participate in photochemical reactions. A control sample (sample protected from light) should always be run in parallel during the photodegradation study. Further details on photo-stability stress testing can be found in reference 45. Generally the reporting of degradation studies are not required for an investigational new drug (IND) application. These studies are usually per-formed to develop confidence that the HPLC method used to analyze the drug substance and the drug product is stability-indicating. However, forced degra-dation studies are required at the submission of a new drug application (NDA). Any significant degradation products should be isolated and/or char-acterized by the appropriate analytical techniques (LC-MS, LC-NMR, NMR, UV, etc.), and a full report of these degradation studies would need to be per-formed [46]. Procedures for the preparation and/or isolation (if performed) and the methodology for the structural determination should be reported even in the event that the impurity cannot be identified. The physical and chemical properties of the isolated degradation product (where applicable) should also be assessed. Also, the mechanism and kinetics of the formation of each degra-dation product should be assessed as well as the order of the reaction. This may be challenging in some cases if the degradation process occurs by various types of oxidation [4749]. Also, for drug products a distinction must be made between degradation products that are related to the drug substance itself, drug substanceexcipient interactions, and those related solely to excipients. 9.5 DISTINGUISHING DRUG-RELATED AND NON-DRUG-RELATED DEGRADATION PRODUCTS It is very important to distinguish which degradation products are related to the placebo or the drug substance. Therefore, the placebo should undergo similar degradation protocol as the drug product and the drug substance. The following questions should be asked when performing the studies. > Is the new peak that is generated an impurity from the drug substance, a degradation product, or the placebo itself? > At what level is the impurity/degradation product present? > Is it a process-related impurity and, if so, at what step of the process is it formed? > Is it a degradation product? If so, under what degradation condition is it formed? > DISTINGUISHING DRUG-RELATED 495 496 METHOD VALIDATION Figure 9-10. Examples of photoinduced reactions in drug molecules. (I) Photoinduced isomerization (Ib), cyclization and enolketo isomerization (Ic) of stilboestrol (Ia). (II) Photoinduced reactions of ketoprofen (IIa) and its degradation product (IIb); decar-boxylation (IIb), reduction (IIc), and dimerization (IIc) products of the drug. (III) N-dealkylation of methotrexate followed by oxidation. (IV) Dehalogenation (IVb) and photohydrolysis (IVc) of frusemide (IVa) [44]. 9.5.1 Drug Product Stress Stress testing should be conducted on final qualitative formulation.This should be a collaborative study with the analytical team. A gradient method should always be employed with an isocratic hold at high organic to ascertain if any hydrophobic degradation products are present. 9.5.1.1 Combination Products. For combination products, additional ques-tions need to be asked such as, Are there any potential chemical reactions between actives? and What is the potential impact on physical properties of the product (e.g., hygroscopicity)? Therefore, forced degradation of the com-bined actives in solution and the solid state should be explored to probe all potential modes of degradation. This, in essence, will lead to more stable formulations. 9.5.1.2 Experimental Approaches. Forced degradation studies (solid state and solution state) should be conducted for defining a suitable stability indi-cating method. Outlined in Table 9-25 and Table 9-26 are some solid state and solution forced degradation studies respectively that could be conducted. An initial forced degradation scouting study, to help define the time frames for a particular forced degradation study, should be conducted. For example in Figure 9-11 a probe forced degradation study of a heated solution at 80C was conducted at 1, 4, and 8 hours. As can be seen in Table 9-27, 8 hours was sufficient to generate about 10% degradation products. The results of the forced degradation scouting study can also be used if adequate degradation is observed (510% total degradation products). The forced degradation should be carried out on the drug substance, the formulation blends (oral drug prod-ucts), and the placebo blends all in parallel to determine if the excipients or > DISTINGUISHING DRUG-RELATED 497 > Figure 9-11. Forced degradation solution heat stress study: 1 mg/mL of API in solu-tion. Top to bottom: initial, 1, 4, and 8 hours at 80C. (stress conditions) 498 METHOD VALIDATION TABLE 9-27. Heat Stress Study of a 1 mg/mL Solution of API at 80C RT (min): 4.100 4.133 4.400 4.933 5.017 5.183 6.733 6.883 7.850 8.283 8.367 RRT: 0.394 0.397 0.422 0.474 0.482 0.498 0.646 0.661 0.754 0.795 0.803 Initial: <0.05 <0.05 <0.05 0.09 1 hr: <0.05 0.06 <0.05 <0.05 0.10 0.08 4 hr: 0.29 0.45 0.19 <0.05 <0.05 <0.05 <0.05 <0.05 0.49 <0.05 0.10 8 hr: 0.62 0.88 0.53 0.08 0.15 0.07 0.07 <0.05 0.95 <0.05 0.07 Active RT (min): 8.667 8.883 9.233 9.617 9.950 10.417 10.683 10.750 11.700 12.383 Total RRT: 0.832 0.853 0.886 0.923 0.955 1.000 1.026 1.032 1.123 1.189 Impurities (A%) Initial: 0.38 <0.05 0.72 <0.05 97.06 0.38 0.98 2.6 1 hr: 0.09 <0.05 <0.05 0.71 <0.05 97.11 0.00 0.90 0.05 2.7 4 hr: 0.38 <0.05 0.31 0.73 <0.05 94.87 0.37 0.91 0.39 0.08 4.7 8 hr: 0.36 <0.05 0.63 0.72 <0.05 91.50 0.36 0.92 1.43 0.36 8.2 impurities in the excipients are participating in generating additional forced degradation products. 9.6 CONCLUDING REMARKS This chapter described the fundamentals and figures of merit for HPLC method validation in pharmaceutical analysis. Special considerations for addressing linearity, recovery, and setting system suitability requirements were presented. The validation process is to confirm that the method is suited for its intended purpose and to prove the capabilities of the test method. The def-initions of method validation parameters are well explained by health author-ities. Although the requirements of validation have been clearly documented by regulatory authorities, the approach to validation is varied and open to interpretation, and validation requirements differ during the development process of pharmaceuticals (early to late phase). However, the requirements also vary by different health authorities. This point is very well demonstrated by Shabir [50], shedding light on the differences and similarities between validation requirements of the US Food and Drug Administration, the US Pharmacopeia, and the International Conference on Harmonization. REFERENCES 1. www.ICH.gov 2. C. C. Chan, H. Lam, Y. C. Lee, X.-M. Zhang, (eds.), Analytical Method Validation and Instrument performance verification , Wiley Interscience, New York, 2004. 3. D. Paroitt, Performance verification testing of high performance liquid chro-matography equipment, LC-GC 12 (1994), 135. 4. V. Grisanti and E. Zachowski, Operation and performance qualification, LC-GC 20 (2002), 355. 5. J. W. Dolan, Performance qualification of LC systems, LC-GC 20 (2002), 842. 6. M. Zooubair El Fallah, Performance verification: a regulatory burden or an analytical tool? LC-GC 17 (1999), 343. 7. P. W. Wrezel, I. Chion, and M. S. Hussain, Validation and implementation of in-process control HPLC assays for active pharmaceutical ingredients, LC-GC North Am. 22 (2004), 10061009. 8. S. Ahuja and S. Scypinski (eds.), Handbook of Modern Pharmaceutical Analysis ,Academic Press, London, 2001. 9. M. Bouma, B. Nuijen, M. T. Jansen, G. Sava, F. Picotti, A. Flaibani, A. Bult, and J. H. Beijnen, Development of a LC method for pharmaceutical quality control of the antimetastatic ruthenium complex NAMI-A, J. Pharm. Biomed. Anal. 31 (2003), 215228. 10. C. Akay and S. A. Ozkan, Simultaneous LC determination of trimethoprim and sulphamethoxazole in pharmaceutical formulations J. Pharm. Biomed. Anal. 30 (2002), 12071213. > REFERENCES 499 11. D. Ciutaru, I. Badea, L. Lazar, D. Nicolescu, and A.Tudose,A HPLC validated assay of paclitaxels related impurities in pharmaceutical forms containing Cremophor EL, J. Pharm. Biomed. Anal. 34 (2004), 493499. 12. Y. Shao, R. Alluri, M. Mummert, U. Koetter, and S. Lech, A stability-indicating HPLC method for the determination of glucosamine in pharmaceutical formula-tions, J. Pharm. Biomed . Anal. 35 (2004), 625631. 13. N. H. Zawilla, M. A. Mohammad, N. M. El kousy, and S. M. El-Moghazy Aly, Determination of meloxicam in bulk and pharmaceutical formulations, J. Pharm. Biomed. Anal. 32 (2003), 11351144. 14. K. R. Tambwekar, R. B. Kakariya, and S. Garg, A validated high performance liquid chromatographic method for analysis of nicotine in pure form and from formula-tions, J. Pharm. Biomed. Anal. 32 (2003), 441450. 15. A. M. Y. Jaber, H. A. Al Sherife, M. M. Al Omari, and A. A. Badwan, Determina-tion of cetirizine dihydrochloride, related impurities and preservatives in oral solution and tablet dosage forms using HPLC, J. Pharm. Biomed. Anal. 36 (2004), 341350. 16. R. K. Verma, S. Garg, A validated high performance liquid chromatographic method for analysis of isosorbide mononitrate in bulk material and extended release formulations, J. Pharm. Biomed. Anal. 30 (2002), 583591. 17. J. Milano, L. M. Morsch, and S. G. Cardoso, LC method for the analysis of oxiconazole in pharmaceutical formulations, J. Pharm. Biomed. Anal . 30 (2002), 175180. 18. Y. Sun, K. Takaba, H. Kido, M. N. Nakashima, and K. Nakashima, Simultaneous determination of arylpropionic acidic non-steroidal anti-inflammatory drugs in pharmaceutical formulations and human plasma by HPLC with UV detection, J. Pharm. Biomed. Anal . 30 (2003), 16111619. 19. ICH Q2A: Validation of analytical methods (definitions and terminology), 1994. 20. W. Demond, R. A. Kenley, J. L. Italien, D. Lokensgard, G. Weilersbacher, and K. Herman, Orthogonal HPLC methods for quantitating related substances and degradation products of pramlintide, AAPS PharmSciTech 1 (2000), article 6. 21. G. Maniara, K. Rajamoorthi, S. Rajan, and G. W. Stockton, Method performance and validation for quantitative analysis by 1H and 31P NMR spectroscopy. Appli-cations to analytical standards and agricultural chemicals, Anal. Chem. 70 (1998), 49214928. 22. D. Giron, Thermal analysis of drugs and drug products, in J. Swarbick and J. C. Boylan (eds.), Encyclopedia of Pharmaceutical Technology , Marcel Dekker, (2002), 27662793. 23. S. Gorog, The sacred cow: The questionable role of assay methods in characteriz-ing the quality of bulk pharmaceuticals, J. Pharm. Biomed.Anal . 36 (2005), 931937. 24. J. D. Orr, I. S. Krull, and M. E. Swartz, Validation of Impurity methods, Part II, LC-GC North Am . 21 (2003), 11461152. 25. R. Cassidy and M. Janoski, LC-GC 10 (1992), 692. 26. C. A. Dorschel, J. L. Ekmanis, J. E. Oberholtzer, F. V. Warren, and B. A. Bidlingmeyer, Anal. Chem. 61 (1989), 951A. 27. S. Burke, Regression and calibration, LC-GC Europe Online Supplement Statis-tics and Data Analysis (2001), 1318. 500 METHOD VALIDATION 28. M. M. Kiser and J. W. Dolan, Selecting the best curve fit, LC-GC Europe March (2004), 138143. 29. J. M. Green, A practical guide to analytical method validation, Anal. Chem. 68 (1996), 305A309A. 30. Analytical Methods Committee: Is my calibration linear? AMC technical brief, No. 3, December 2000. 31. J. Erner, Chapter 2.4, in J. Ermer and J. H. McB. Miller (eds.), Method Validation in Pharmaceutical Analysis , Wiley-VCH, New York, 2005. 32. EURACHEM: The fitness for purpose of analytical methods. A Laboratory Guide to Method Validation and Related Topics. http://www.eurachem.ul.pt, 1998. 33. K. Baumann, Regression and calibration for analytical separation techniques. Part II: Validation, weighted and robust regression, Process Control and Quality 10 (1997), 75112. 34. ICH guidelines Q2B, Validation of Analytical Procedures: Methodology, Novem-ber 1996. 35. http://www.labcompliance.com/methods/meth_val.htm, accessed May 16, 2006. 36. S. Klick, P. G. Muijselaar, J. Waterval, T. Eichinger, C. Korn, T. K. Gerding, A. J. Debets, C. Sanger-van de Griend, C. van den Beld, G. W. Somsen, and G. J. De Jong, Toward a generic approach for stress testing of drug substances and drug product, Pharm. Technol. February (2005), 4866. 37. G. Boccardi, Autoxidation of drugs: prediction of degradation impurities from results of reactions with radical chain initiators, Il Farmaco 49 (1994), 431. 38 G. Boccardi, Oxidative susceptibility texting, Chapter 7, in S. W. Baertschi (ed.), Pharmaceutical Stress Testing: Predicting Drug Degradation , Vol. 153, Drugs and the Pharmaceutical Sciences, Taylor and Francis, 2005. 39. M. Wasserman (eds.), Singlet Oxygen , Academic Press, New York, 1979. 40. A. A. Frimer (ed.), Singlet O 2, Vols. II and III, CRC Press, Boca Raton, FL, 1985. 41. Foote, Pryor, (eds.), Free Radicals in Biology , Vol. II, Academic Press, New York, 1976, 85133. 42. R. Seburg, Photosensitized Singlet Oxygen-Mediated Degradation of an Angiotensin II Antagonist, Losartan Potassium, Merck Research Laboratories, http://www.iirusa.com/forceddegradation/files/IIR_p1038_Seburg.pdf#search= AIBN%20and%20forced%20degradation%20and%20HPLC , accessed May 15, 2006. 43. J. K. Guillory and R. I. Poust, Chemical kinetics and drug stability, Chapter 6 in G. S. Banker and C. T. Rhodes, Modern Pharmaceutics , 3rd ed., Marcel Dekker, New York, 1996, pp. 139165. 44. H. H. Tnnesen, Formulation and stability testing of photolabile drugs, Int. J. Pharm. 225 (2001), 114. 45. E. Fasani and A. Albini, Photostability stress testing, Chapter 10 in Pharmaceuti-cal Stress Testing: Predicting Drug Degradation , edited by S.W. Baertschi, 1 st edition, Vol. 153, Drugs and the Pharmaceutical Sciences Series (Taylor and Francis), 2005. 46. FDA, Center for Drug Evaluation and Research, Submitting Documentation for the Stability of Human Drugs and Biologics, Rockville, MD, February 1987. > REFERENCES 501 47. S. W. Hovorka and C. Schoneich, Oxidative degradation of pharmaceuticals: Theory, mechanisms and inhibition, J. Pharm. Sci . 90 (2001), 253269. 48. K. C. Waterman, R. C. Adami, K. M. Alsante, J. Hong, M. S. Landis, F. Lombardo, and C. J. Roberts, Stabilization of Pharmaceuticals to Oxidative Degradation, Pharm. Dev. Tech . 7 (2002), 132. 49. S. R. Byrn, R. R. Pfeiffer, and J. G. Stowell, Solid State Chemistry of Drugs , 2nd ed., SSCI Inc., West Lafayette, IN, 1999. 50. G. A. Shabir, Validation of high-performance liquid chromatography methods for pharmaceutical analysis: Understanding the differences and similarities between validation requirements of the US Food and Drug Administration, the US Phar-macopeia and the International Conference on Harmonization, J. Chromatogr. A 987 (2003), 5766. 502 METHOD VALIDATION