Quality by Design Compliant Analytical Method ... - ACS Publications

Nov 22, 2011 - substance or a product as compliant with respect to the corre- sponding product specifications. The aim is similar for content assays t...
1 downloads 8 Views 2MB Size
Download PDF
ARTICLE pubs.acs.org/ac

Quality by Design Compliant Analytical Method Validation E. Rozet,*,†,§ E. Ziemons,† R.D. Marini,† B. Boulanger,‡ and Ph. Hubert† † ‡

Analytical Chemistry Laboratory, CIRM, Institute of Pharmacy, University of Liege, Liege, Belgium Arlenda SA, Liege, Belgium

bS Supporting Information ABSTRACT: The concept of quality by design (QbD) has recently been adopted for the development of pharmaceutical processes to ensure a predefined product quality. Focus on applying the QbD concept to analytical methods has increased as it is fully integrated within pharmaceutical processes and especially in the process control strategy. In addition, there is the need to switch from the traditional checklist implementation of method validation requirements to a method validation approach that should provide a high level of assurance of method reliability in order to adequately measure the critical quality attributes (CQAs) of the drug product. The intended purpose of analytical methods is directly related to the final decision that will be made with the results generated by these methods under study. The final aim for quantitative impurity assays is to correctly declare a substance or a product as compliant with respect to the corresponding product specifications. For content assays, the aim is similar: making the correct decision about product compliance with respect to their specification limits. It is for these reasons that the fitness of these methods should be defined, as they are key elements of the analytical target profile (ATP). Therefore, validation criteria, corresponding acceptance limits, and method validation decision approaches should be settled in accordance with the final use of these analytical procedures. This work proposes a general methodology to achieve this in order to align method validation within the QbD framework and philosophy. β-Expectation tolerance intervals are implemented to decide about the validity of analytical methods. The proposed methodology is also applied to the validation of analytical procedures dedicated to the quantification of impurities or active product ingredients (API) in drug substances or drug products, and its applicability is illustrated with two case studies.

T

he concept of quality by design (QbD) has recently been adopted for the development of pharmaceutical processes to ensure a predefined product quality.13 Focus on applying the QbD concept to analytical methods has increased as it is fully integrated within pharmaceutical processes and especially in the process control strategy.46 In particular, recent publications have highlighted the possibility of applying the QbD concept to method development and to determine method operable design region (MODR).710 In addition, several authors11,12 have emphasized the need to switch from the traditional checklist implementation of, for example, the International Conference on Harmonization (ICH) Q213 or U.S. Pharmacopeia (USP) Æ1225æ14 method validation requirements to a method validation approach that should provide a high level of assurance of method reliability. This switch is needed to ensure adequate measurement of the critical quality attributes (CQAs) of the drug product. It is crucial to evaluate the validity of analytical methods by considering their intended purpose and demonstrating that they meet the requirements that have been predefined in the analytical target profile (ATP).15 The core aim is to define the purpose of the analytical method being validated and its measure of fitness (e.g., accuracy, precision, and so on). In fact, the general objective of validation of an analytical procedure is “to demonstrate that it is suitable for its intended purpose”.13,16 The impact of the results generated by analytical methods on the quality of the products manufactured is r 2011 American Chemical Society

of key importance for selecting the appropriate validation criteria, and hence the CQAs, to assess the validity of these methods. In this work we will focus on two categories of analytical methods: the first is analytical procedures intended to quantify impurities, and the second corresponds to content assays of drug products or drug substances (as defined in ICH Q2).13 Usually, when such methods are validated, their intended purpose is defined as their ability to quantify the impurities or the active substance in its matrix (drug substance or drug product). However, this is not their final intended purpose. This intended purpose is directly related to the final decision that will be made with the results generated by the analytical methods under study. The final aim of quantitative impurity assays is to correctly declare a substance or a product as compliant with respect to the corresponding product specifications. The aim is similar for content assays to the one for impurity assays: to make the correct decision about product compliance with respect to their own specification limits. Thus, it is with these aims that the fitness of these methods should be defined, as they are key elements of the ATP. The ATP of these categories of methods should thus include the maximum acceptable risk to make incorrect decisions. Validation of the analytical method should then judge the ability of these methods Received: July 19, 2011 Accepted: November 22, 2011 Published: November 22, 2011 106

dx.doi.org/10.1021/ac202664s | Anal. Chem. 2012, 84, 106–112

Analytical Chemistry

ARTICLE

Figure 1. Decision profile for assessing the validity of an HPLCUV method for quantification of (R)-timolol impurity in (S)-timolol drug substance. The dashed line is the one-sided 95% β-expectation tolerance interval; the diagonal solid line is the identity line y = x; and λ = Limp is the impurity specification. Open circles represent the concentration levels of the validation standards analyzed to assess the validity of the method.

to meet this requirement stated in their ATP. Therefore, validation criteria, corresponding acceptance limits, and method validation decision approach should be settled in accordance with the final use of these analytical procedures. This work proposes a general methodology to achieve this in order to align method validation within the QbD framework and philosophy for the validation of analytical procedures dedicated to the quantification of impurities or API in drug substances or drug products. β-Expectation tolerance intervals are implemented to decide about the validity of analytical methods. The methodology proposed is further applied to one case study for each category of method in order to illustrate its applicability.

underestimate the concentration of impurity. This would allow a substance/ product to erroneously be declared compliant when it was not compliant, that is, the patient risk. This patient risk is crucial and it should be included in the ATP. Therefore, the validation acceptance limit and the validation decision methodology for this section of the concentration range should thus be set to control underestimation. This is illustrated in Figure 1. In order to take into account these elements, the acceptance limits λ for quantitative impurities assays can be generally defined in concentration values (Figure 1): λ ¼ Limp

’ QUANTITATIVE IMPURITY ASSAYS Following ICH Q2, validation of a procedure aimed at quantifying impurities should be realized at least over a range of concentrations covering the reporting level of the impurity to 120% of the specification (Limp).13 ICH Q3A specifies that the quantification limit for the analytical procedure should be no more than (e) the reporting threshold.17 This thus defines the concentration range over which the analytical method must be validated. Keeping in mind that the final objective of a quantitative impurity assay is to correctly declare compliant a drug substance or a drug product, the validation acceptance limits should therefore not be bilateral over the whole concentration range as depicted in Figure 1. Indeed, the method validation acceptance limits for concentration levels smaller than the impurity specification (Limp) should be set in order to control overestimation. The risk for this part of the concentration range studied is to overestimate the concentration of impurity in such a way that a truly compliant substance/product will erroneously be declared as not compliant. This is thus the producer risk. While ATP focuses on what the method should achieve from a patient perspective, this producer risk would only be detailed within the pharmaceutical industry as it only pertains to manufacturability, and it would not be divulged to external parties. Nonetheless, it is important for the industry to know this risk and eventually reduce it if it is found to be unacceptable. Underestimation is not a problem for this part of the concentration range. The problem is the converse for the concentration range above the specification Limp. Indeed it is crucial for these concentration levels that the analytical procedure does not

ð1Þ

where Limp is the specification of the impurity under investigation. Decision Methodology. The two core validation criteria for this category of quantitative methods are method accuracy (bias, sometimes called trueness)18,19 and precision. However, it is not necessary to apply an acceptance limit to each of these criteria separately, as a method with a lack of accuracy could still be fit for its intended use if it has good enough precision. It is the simultaneous combination of both criteria that should be used to judge the validity of the quantitative impurity assays, as advocated by Hubert and co-workers20 and Hoffman and Kringle21 This simultaneous combination is sometimes called total error or results accuracy (by opposition to the method accuracy).12,18,19 Total error approach provides a useful methodology for dealing with biases that may vary, for example, with concentration. However, if there is a known overall method bias, it may be more appropriate in a QbD approach to eliminate it if possible or to correct for it. A convenient way to measure this adequately is by using the statistical methodology of tolerance intervals.22 Depending on the definition used, a tolerance interval can represent an interval that, for instance, covers a certain percentage of a distribution with a given probability, or an interval where each future result has a userdefined probability to fall into. A tolerance interval can be claimed, for example, to contain an expected proportion of 90% of future analytical results. Tolerance intervals should not be confused with confidence intervals, which give only indications about the sampling distribution of a statistical parameter, such as the mean. Therefore, the decision methodology should be the one proposed by Hubert and co-workers20 or Hoffman and Kringle21 107

dx.doi.org/10.1021/ac202664s |Anal. Chem. 2012, 84, 106–112

Analytical Chemistry

ARTICLE

Figure 2. (A) Unreliability region for routine results obtained for the HPLCUV method aimed at quantifying (R)-timolol in (S)-timolol drug substance. (B) Close-up of the unreliability region. Dashed lines are the one-sided 95% β-expectation tolerance intervals; the diagonal continuous line is the identity line y = x; λ = Limp is the impurity specification; and the shaded region corresponds to the unreliability region.

by comparing the tolerance intervals computed based on the validation data with the acceptance limits previously defined. In order to fully explain how tolerance intervals are computed, the Supporting Information shows the steps performed to compute β-expectation tolerance intervals (note Hoffman and Kringle21 used β-content γ-confidence tolerance intervals that apply a probability or confidence to the interval) for the quantitative impurity assay used as the example in the next section with the raw data given in Table S-1, Supporting Information. In addition, the interested reader is referred to the following publications to obtain full computational details about tolerance intervals.2225 Nonetheless, a key difference here is that only a one-sided tolerance interval should be used depending on the side of the impurity specification level being tested, as shown in Figure 1. Equation 1 shows that the method has to demonstrate continuously improving quantitative performances when the concentration of impurity gets closer to the impurity specification (see Figure 1). At the limit when the concentration of impurity is exactly at the impurity specification (Limp), the method should be perfect: without any bias and any variability. This is infeasible in practice. Therefore, the method is no longer providing reliable results for concentration levels around the specification limit. It is of core importance to find out if the values of routine results around the specification limit (Limp) are still providing the correct decision about the compliance of the product with adequate guarantee. Therefore, a maximum unreliable region [Ulow; Uup] around the specification limit (Limp ) should be defined and added in the definition of the ATP together with the maximum

probability of making a wrong decision, that is, the maximum false compliance risk. Another risk, the false noncompliance risk, can also be given. However, as false noncompliance decisions are solely a manufacturer’s risk, it is not required to provide this risk in the definition of the ATP. Nonetheless, the noncompliance risk can still form a part of the company’s internal method acceptance criteria (in the remainder of this paper, both risks will be included in the ATP for this last reason). Decision Graph. In order to decide if the analytical method is fit for its purpose, two elements have to be verified. The first element is to evaluate the ability of the method to provide reliable results over the whole concentration range studied, excluding the unreliability region around the impurity specification. This is performed by use of the decision graph. To draw the decision graph, β-expectation tolerance intervals are computed from the analytical results obtained at each concentration level of the validation standards. Then the upper tolerance limits are linearly connected for concentration levels under the validation acceptance limit (λ), while the lower tolerance limits are connected for concentration levels above the validation acceptance limit (λ). For instance, on the basis of this graph, the decision rule could be (1) if all the upper one-sided 95% β-expectation tolerance intervals, computed at the concentration levels of the validation standards smaller than the lower limit of the maximum unreliability region around the impurity specification (Ulow), are under the acceptance limit (λ =Limp) AND (2) if all the lower one-sided 95% β-expectation tolerance intervals, computed at the concentration 108

dx.doi.org/10.1021/ac202664s |Anal. Chem. 2012, 84, 106–112

Analytical Chemistry

ARTICLE

Figure 3. Decision profiles for assessing the validity of an HPLCUV method for quantification of fenofibrate in a drug product. Dashed lines are the 95% β-expectation tolerance intervals (either one-sided or bilateral); the diagonal solid line is the identity line y = x; and λimp and λup are the lower and upper specifications of the active substances in the product, respectively. Open circles represent the concentration levels of the validation standards analyzed to assess the validity of the method.

levels of the validation standards higher than the upper limit of the unreliability region around the impurity specification (Uup), are above the acceptance limit (λ =Limp), then the method is declared as valid, as illustrated in Figure 1. However this is not enough, as the unreliability region of the analytical method may be greater than the maximum unreliability region defined in the ATP. Unreliability Graph. To obtain the unreliability graph, the decision graphs have to be interpreted the other way round: for an observed analytical result, what is the corresponding true concentration? It is now the lower one-sided β-expectation tolerance interval for observed result (which corresponds to an upper limit for the true concentration) that must be looked at for concentration levels smaller than the specification limit (Limp), while it is the upper one-sided β-expectation tolerance interval (which corresponds to a lower limit for the true concentration) that has to be computed for concentration levels above the specification limit (Limp), as shown in Figure 2. A region of concentration called the unreliability region is then identified around this target specification limit for the impurity. This unreliability region is defined by the intersection of the lower tolerance interval with the specification limit (Limp) and the intersection of the upper tolerance interval with the specification limit (Limp), as illustrated in Figure 2. Based on the unreliability graph, the decision rule could be as follows: If the unreliability region obtained from the maximum risk of false compliance and false noncompliance (e.g., 5% each) is smaller than or equal to a predefined maximum unreliability region [Ulow; Uup], then the method is declared valid. This unreliability region around the specification limit can also be seen as guard-banding.26 In routine analyses, when results are obtained in this unreliability region, further investigation should be made to ensure that the right decision is taken for the produced batch. Such investigations could involve doing multiple sample preparations and taking the mean of these analyses as an analytical result. These repetitions will provide a more precise estimation of the effective concentration of the analyte in the sample that can be compared to the guard band, in order to increase the confidence about the compliance of the product. Example of Application. In order to illustrate the proposed methodology to assess the validity of quantitative impurity assays, it has been applied to a previously published method concerning the quantification by HPLCUV of (R)-timolol impurity in

(S)-timolol drug substance.27,28 A second example is also detailed in the Supporting Information. The ATP of this method is that the HPLCUV method should be able to quantify (R)-timolol impurity in the presence of (S)-timolol API as well as any other potential impurity or degradation products over a range of 0.1 1.6% relative to the API. The analytical method should be able to perform this determination by demonstrating that each of the probabilities of false compliance and false noncompliance decisions should be at the most 5% outside an unreliability region of maximum 0.1% around the impurity specification. The impurity specification for (R)-timolol corresponds to the concentration level of the validation standards of LR‑tim imp = 0.8%. It corresponds thus to the analytical method validation acceptance limit λ (eq 1). The maximum unreliability region [Ulow;Uup] for this example is then [Ulow = 0.7%; Uup = 0.9%] as defined in the ATP. Figure 1 is the decision graph for the (R)-timolol assay. The one-sided β-expectation tolerance intervals for (R)-timolol were set at a probability level of β = 95%, as required in the ATP.27,28 As can be seen on this graph, the one-sided β-expectation tolerance intervals are not crossing the validation acceptance limit except for concentration levels close to this acceptance limit. Around this validation acceptance limit, the unreliability region shown in Figure 2 is relatively small, ranging from 0.79% to 0.82% for the (R)-timolol assay and fully included into the maximum unreliability region of [Ulow = 0.7%; Uup = 0.9%] defined in the ATP. These results demonstrate two main points: first, that this HPLCUV method is fit for its intended purpose, as there is high probability that each future result will not exceed the acceptance limit; second, that routine results obtained within the unreliability region should not be trusted, as the risks of false compliance and false noncompliance exceed those set initially, that is, 100  95 = 5% for (R)-timolol.

’ CONTENT ASSAYS The methodology previously described for quantitative impurity assays can also be implemented with some adaptations to assays aimed at quantifying active substances in drug substances or drug products. The main difference is that here the product specification is usually bilateral. There is an upper specification limit (e.g., Lup = 105% or 110% of the production target or label claim) and a lower specification (e.g., Llow = 95% or 90% of the 109

dx.doi.org/10.1021/ac202664s |Anal. Chem. 2012, 84, 106–112

Analytical Chemistry

ARTICLE

Figure 4. Unreliability region for routine results obtained for the HPLCUV method for the quantification of fenofibrate in a drug product. Dashed lines are the 95% β-expectation tolerance intervals (either one-sided or bilateral); the diagonal solid line is the identity line y = x; and λimp and λup are the lower and upper specifications of the active substances in the product, respectively.

production target or label claim). Figure 3 illustrates the decision profile for a content assay of active substance in a pharmaceutical formulation. It is an HPLCUV assay for the quantification of fenofibrate in a drug product.29 The ATP of this method is that the HPLCUV method should be able to quantify fenofibrate API in the presence of any potential impurity or degradation products as well as in the presence of the excipients present in the drug product over a range of 40120 μg/mL fenofibrate. The analytical method should be able to perform this determination by demonstrating that the probabilities of false compliance and false noncompliance decisions should be at the most 5% outside the unreliability regions of maximum 1.5% around the API specifications. The specifications for fenofibrate were (5% around the production target, which corresponds to limits of λlow = 76 and λup = 84 μg/mL. The two maximum unreliability regions around the product specifications are then [U low = + 74.9 μg/mL; U up = 77.2 μg/mL] and [U low = 82.7 μg/mL; U+up = 85.3 μg/mL]. The quality level probability was set at β = 95% to fulfill the requirement stated in the method ATP. The challenge is to avoid overestimation for concentration values smaller than the lower specification (λlow). Thus only upper one-sided β-expectation tolerance intervals need to be compared to this specification. Underestimation should be controlled for concentration levels above the upper specification limit (λup). Lower one-sided β-expectation tolerance intervals should be compared to this specification. However, simultaneous control of both over- and underestimation should be done for concentration levels within the specifications [λlowλup]. This requires the computation of two-sided β-expectation tolerance intervals and their comparison to the bilateral specification limits. Figure 3 illustrates the decision profile obtained for this HPLC UV method. As can be seen in Figure 3, the β-expectation tolerance intervals are not crossing the specification limits except for concentration levels close to these last limits. However, when approaching concentration levels corresponding to the specifications, the method no longer provides results within these specifications with the selected probability. This range of routine results that will be obtained by this content assay over which the method is no longer reliable defines the method’s unreliability regions, which have to be compared with the maximum unreliability regions defined in the ATP. This is achieved by looking at the reliability profile in a same way as done for the impurities quantitative assays.

To obtain the reliability profiles, it is necessary to compute the lower one-sided β-expectation tolerance intervals for concentration levels smaller than the lower specification limit (λlow), the upper one-sided β-expectation tolerance intervals for concentration levels higher than the upper specification limit (λup), and the two-sided β-expectation tolerance intervals for concentrations within these specification limits. The unreliability regions for the content assay of fenofibrate in a drug product are illustrated in Figure 4. For the HPLCUV method used to quantify fenofibrate, the unreliability regions range from 75.4 to 77.2 μg/mL around the lower specification limit (λlow) and from 83.6 to 85.1 μg/mL around the upper one (λup; Figure 4). These unreliability regions are included into the maximum unreliability regions defined in the ATP, showing that the method is suitable for its intended use. Nonetheless, results obtained during routine analyses within these unreliability ranges cannot be trusted, as the risk to erroneously declare a product as compliant exceeds the maximum risk tolerated.

’ DISCUSSION Deciding on an analytical method’s validity should focus on the final aim of the analytical method under study. This work focuses on quantitative impurity assays and content assays. The challenge when deciding about a method’s validity is ensuring that erroneous decisions based upon the results obtained from it will not be excessive. This is essential in order to limit the patient risk (the risk of erroneously accepting batches as compliant) as well as the producer risk (the risk of erroneously rejecting batches). The approach proposed here uses the statistical methodology of tolerance intervals to allow better understanding and control of theses risks.24,25,30 Indeed, the quality level β allows these two risks to be managed. In the two case studies, a value of β = 95% has been set arbitrarily. However, this value can be modulated, either increased or decreased, on the basis of scientifically sound judgments, knowledge of the pharmaceutical process involved, and information gathered through the QbD development approach. The quality level can be different for the patient or producer risks, allowing mitigation of the risks of erroneously accepting/rejecting batches. For drugs with a narrow therapeutic window or for highly toxic impurities, this quality level should probably be set higher than 95%. Further requirements on this quality level could valuably be provided by regulatory bodies such 110

dx.doi.org/10.1021/ac202664s |Anal. Chem. 2012, 84, 106–112

Analytical Chemistry as USP, Food and Drug Administration (FDA), or ICH, especially regarding patient risk. Concerning quantitative impurity assays, when multiple impurities must be quantified, eventually across different concentration ranges, the same methodology using a decision profile and an unreliability graph is used for each impurity. In such situations, validation standards covering the concentration range of each impurity are prepared and analyzed. Tolerance intervals are then computed for each concentration level of each impurity, and a decision graph as well as an unreliability graph is constructed for each impurity. We acknowledge that the proposed methodology is resource-intensive or practically difficult to apply when many impurities are quantified. To lighten the workload, impurities and degradation products could be ranked following a risk analysis such as failure mode and effects analysis (FMEA). The proposed validation approach could then be applied to key impurities or degradation products, for instance, those being the most toxic or the most frequently observed during the manufacturing process. Relative response factors (RRFs) could also be used to quantify several impurities and the proposed validation approach could then be implemented, using these RRFs to assess the reliability of the results generated for selected impurities or degradation products. For API assay methods, the drug substance specification is a narrow range such as 98.0102.0% or 97.0103.0%. The key difficulty for these cases is that the unreliability regions generally obtained may be excessive and could cover the whole API specification. For example, Hofer et al.31 have shown that the precision of HPLC assays generally cover a large portion of the API specification, therefore the risk to make inadequate decision concerning the batches produced are high. Other assays may be more suitable than HPLC ones such as titration assays, by using a mass balance assay31 or by reducing the variability of HPLC assays.31,32 The proposed methodology to assess the validity of analytical methods could then be applied to these more precise assays to ensure that they will provide reliable decisions with unreliability regions of reasonable width. An additional point highlighted is that the closer concentrations levels of samples are to the specification levels, the higher the patient and producer risks are, as analytical methods are never perfect. Indeed, Chatfield and Borman33 have stated that “there will always be an uncertainty region around a specification limit applied to measured results”. The proposed methodology takes this into account by defining an unreliability region around these specification limits: a region of analytical results values that cannot provide the desired control of the patient and producer risks. This also links into the requirements of the International Organization for Standardization (ISO) to include measurement uncertainty in compliance assessment by realizing what is called “guard-banding”.26 The specification limits are no longer a single result value but a range of values (concentrations, amounts, percentages, etc.). The analytical method's convenience can then be judged by the magnitude of the unreliability region. Unreliability regions that are too large imply a useless method and thus an invalid one. In such situations, it will sometimes be necessary to go back to method development, to review the ATP requirements, or to reassess the analytical MODR. Another possibility is to consider that decisions about a batch’s compliance/noncompliance will be made by using the mean value of a prespecified number of repeated analyses and not just with a single analytical result. The quality of validation data acquired is also of core importance to ensure that the conclusions made about the method validity and about the size of the unreliability regions

ARTICLE

are correct. This implies that repeated experiments should start from the sample preparation involved in the analytical procedure and are not simply repeated injections of the same sample solution. Additionally, for intermediate precision, sufficient sources of variation should be included, such as operators, equipments, days. These sources could be valuably selected from a risk analysis such as FMEA performed during the method development phase.46,34 Finally, the sample size used to carry out method validation should be adequately defined and, for example, not always the minimum ICH Q2 requirements of three repetitions and two or three series frequently performed. The sample size required to compute tolerance intervals at several concentration levels is probably greater than what is currently performed. Nonetheless, the proposed approach provides the guarantee that the analytical method will provide reliable results. This is valuable information that is not obtained by other classical validation approaches, generally involving small sample sizes in order to fulfill the validation checklist.20 Method validation is generally seen as laborious and as a burden. Failing to correctly validate analytical methods leads to increased day-to-day difficulties for the quality control laboratory, such as recurrent out-of-specification (OOS) results or issues when analytical methods are transferred from one laboratory to another. This leads to investigations to identify the root causes that may results in wasted resources that could have been saved by adequately validating analytical methods. This is in contradiction with the concept of QbD. QbD is an opportunity to put back method validation as a useful step in the life cycle of analytical methods.

’ CONCLUSION Analytical methods validation is essential in order to ensure trustworthiness and reliability of the results that will be generated by them during their routine application. Critical decisions are made by use of these results: compliance of batches of products, patient health monitoring, and so on. This work has shown how method validation can be used to give guarantees that analytical methods will provide daily results that can be used efficiently to make adequate decisions. Method validation is here directly linked to the effective purpose of these analytical methods. The proposed method validation approach should allow analytical methods to fully fulfill their control tasks in the framework of QbD for pharmaceutical products development. When the decision profile and unreliability region graph are used, the validity of the method is assessed by controlling the risk of false compliance, and routine decisions regarding products’ compliance are more reliable. This methodology has been discussed in relation to analytical methods for the quantification of impurities and active substances in drug substances or drug products. However, the methodology proposed can be applied to any quantitative analytical method. This work has highlighted how method validation can be switched from the regrettably too frequent checklist validation approach to a QbD compliant approach while still fitting within the current validation regulatory framework. ’ ASSOCIATED CONTENT

bS

Supporting Information. Additional text and equations and one table as described in the text. This material is available free of charge via the Internet at http://pubs.acs.org.

111

dx.doi.org/10.1021/ac202664s |Anal. Chem. 2012, 84, 106–112

Analytical Chemistry

ARTICLE

’ AUTHOR INFORMATION

General principles and definitions. International Organization for Standardization (ISO), Geneva, Switzerland, 1994. (20) Bouabidi, A.; Rozet, E.; Fillet, M.; Ziemons, E.; Chapuzet, E.; Mertens, B.; Klinkenberg, R.; Ceccato, A.; Talbi, M.; Streel, B.; Bouklouze, A.; Boulanger, B.; Hubert, Ph. J. Chromatogr. A 2010, 1217, 3180–3192 (http://hdl.handle.net/2268/29471). (21) Hoffman, D.; Kringle, R. Pharm. Res. 2007, 24, 1157–1164. (22) Hahn, G. J.; Meeker, W. Q. Statistical Intervals: A Guide for Practitioners; Wiley: New York, 1991. (23) Hubert, Ph; Nguyen-Huu, J.-J.; Boulanger, B.; Chapuzet, E.; Cohen, N.; Compagnon, P.-A.; Dewe, W.; Feinberg, M.; Laurentie, M.; Mercier, N.; Muzard, G.; Valat, L.; Rozet, E. J. Pharm. Biomed. Anal. 2007, 45, 82–96 (http://hdl.handle.net/2268/6187). (24) Rozet, E.; Wascotte, V.; Lecouturier, N.; Preat, V.; Dewe, W.; Boulanger, B.; Hubert, Ph. Anal. Chim. Acta 2007, 591, 239–247 (http://hdl.handle.net/2268/6022). (25) Hoffman, D.; Kringle, R. J. Biopharm. Stat. 2005, 15, 283–293. (26) ASME B89.7.3.1-2001. Guidelines for decision rules: considering measurement uncertainty in determining conformance with specifications. (27) Marini, R. D.; Chiap, P.; Boulanger, B.; Rudaz, S.; Rozet, E.; Crommen, J.; Hubert, Ph. Talanta 2006, 68, 1166–1175 (http://hdl. handle.net/2268/18790). (28) Marini, R. D.; Groom, C.; Doucet, F. R.; Hawari, J.; Bitar, Y; Holzgrabe, U.; Gotti, R.; Schappler, J.; Rudaz, S.; Veuthey, J. L.; Mol, R.; Somsen, G. W.; de Jong, G. J.; Thanh, H. P.T.; Zhang, J.; Van Schepdael, A.; Hoogmartens, J.; Bri^ one, W.; Ceccato, A.; Boulanger, B.; Mangelings, D.; Vander Heyden, Y.; Van Ael, W.; Jimidar, I.; Pedrini, M.; Servais, A. C.; Fillet, M.; Crommen, J.; Rozet, E.; Hubert, Ph. Electrophoresis 2006, 27, 2386–2399 (http://hdl.handle.net/2268/9012). (29) Rozet, E.; Mertens, B.; Dewe, W.; Ceccato, A.; Govaerts, B.; Boulanger, B.; Chiap, P.; Streel, B.; Crommen, J.; Hubert, Ph. J. Pharm. Biomed. Anal. 2006, 42, 64–70 (http://hdl.handle.net/2268/6027). (30) USP 33 NF 28 S1, U.S. Pharmacopeia, 2007. USP-NF General Chapter Æ1010æ. (31) Hofer, D. J.; Olsen, B. A.; Rickard, E. C. J. Pharm. Biomed. Anal. 2007, 44, 906–913. (32) Dejaegher, B.; Jimidar, M.; De Smet, M.; Cockaerts, P.; Smeyers-Verbeke, J.; Vander Heyden, Y. J. Pharm. Biomed. Anal. 2006, 42, 155–170. (33) Chatfield, M. J.; Borman, P. J. Anal. Chem. 2009, 81, 9841–9848. (34) Borman, P. J.; Chatfield, M. J.; Damjanov, I.; Jackson, P. Anal. Chim. Acta 2011, 703, 101–113.

Corresponding Author

*Telephone: +32-4-3664320. Fax: +32-4-3664317. E-mail: Eric. [email protected]. Notes §

FRS-FNRS postdoctoral researcher (Belgium).

’ ACKNOWLEDGMENT We are very grateful to the anonymous reviewers for providing important comments that led to significant improvements of this article. A research grant from the Belgium National Fund for Scientific Research (FRS-FNRS) to E.R. is gratefully acknowledged. ’ REFERENCES (1) Topic Q8 (R2): Pharmaceutical Development. International Conference on Harmonization (ICH) of Technical Requirements for Registration of Pharmaceuticals for Human Use, Geneva, Switzerland, 2009. (2) U.S. Food and Drug Administration (FDA), Department of Health and Human Services. Pharmaceutical Quality for the 21st Century: A Risk-Based Approach Progress Report, May 2007. http:// www.fda.gov/AboutFDA/CentersOffices/CDER/ucm128080.htm. (3) Yu, L. X. Pharm. Res. 2008, 25, 781–791. (4) Vogt, F. G.; Kord, A. S. J. Pharm. Sci. 2011, 100, 797–812. (5) Gavin, P. F.; Olsen, B. A. J. Pharm. Biomed. Anal. 2008, 46, 431–441. (6) Borman, P. J.; Nethercote, P.; Chatfield, M. J.; Thompson, D.; Truman, K. Pharm. Technol. 2007, 31, 142–152. (7) Peterson, J. J. J. Qual. Technol. 2004, 36, 139–153. (8) Debrus, B.; Lebrun, P.; Ceccato, A.; Caliaro, G.; Rozet, E.; Nistor, I.; Oprean, R.; Ruperez, F. J.; Barbas, C.; Boulanger, B.; Hubert, Ph. Anal. Chim. Acta 2011, 691, 33–42 (http://hdl.handle.net/2268/ 88084). (9) Debrus, B.; Lebrun, P.; Mbinze Kindenge, J.; Lecomte, F.; Ceccato, A.; Caliaro, G.; Mavar Tayey Mbay, J.; Boulanger, B.; Marini, R. D.; Rozet, E.; Hubert, Ph. J. Chromatogr. A 2011, 1218, 5205–5215 (http://hdl.handle.net/2268/93241). (10) Borman, P.; Roberts, J.; Jones, C.; Hanna-Brown, M.; Szucs, R.; Bale, S. Sep. Sci. 2010, 2, 1–8. (11) Nethercote, P.; Borman, P. J.; Bennett, T.; Martin, G.; McGregor, P. Pharm. Manuf. 2010, No. April, 37–48 (http://www. pharmamanufacturing.com/articles/2010/060.html). (12) Rozet, E.; Ceccato, A.; Hubert, C.; Ziemons, E.; Oprean, R.; Rudaz, S.; Boulanger, B.; Hubert, Ph J. Chromatogr. A 2007, 1158, 111–125 (http://hdl.handle.net/2268/3646). (13) Topic Q2 (R1): Validation of Analytical Procedures: Text and Methodology. International Conference on Harmonization (ICH) of Technical Requirements for Registration of Pharmaceuticals for Human Use, Geneva, Switzerland, 2005. (14) USP 33 NF 28 S1, U.S. Pharmacopeia, 2007. USP-NF General Chapter Æ1225æ. (15) Schweitzer, M.; Pohl, M.; Hanna-Brown, M.; Nethercote, P.; Borman, P.; Hansen, G.; Smith, K.; Larew, J. Pharm. Technol. 2010, 34, 52–59. (16) ISO/CEI 17025: General requirements for the competence of testing and calibration laboratories. International Organization for Standardization (ISO), Geneva, Switzerland, 2005. (17) Topic Q3A (R2): Impurities in New Drug Substances. International Conference on Harmonization (ICH) of Technical Requirements for Registration of Pharmaceuticals for Human Use, Geneva, Switzerland, 2006. (18) JCGM 200:2008, International Vocabulary of Metrology  Basic and general concepts and associated terms (VIM), 2008. (19) ISO 5725-1, Application of the statistics: Accuracy (trueness and precision) of the results and methods of measurement, Part I: 112

dx.doi.org/10.1021/ac202664s |Anal. Chem. 2012, 84, 106–112