Journal of Pharmaceutical and Biomedical Analysis 88 (2014) 519–524
Contents lists available at ScienceDirect
Journal of Pharmaceutical and Biomedical Analysis journal homepage: www.elsevier.com/locate/jpba
Robust method optimization strategy—A useful tool for method transfer: The case of SFC Amandine Dispas ∗ , Pierre Lebrun, Bertyl Andri, Eric Rozet, Philippe Hubert University of Liege (Ulg), Department of Pharmacy, CIRM, Laboratory of Analytical Chemistry, 1 Avenue de l’Hôpital, CHU B36, B-4000 Liège, Belgium
a r t i c l e
i n f o
Article history: Received 28 June 2013 Received in revised form 2 September 2013 Accepted 26 September 2013 Available online 12 October 2013 Keywords: Quality by Design (QbD) Design Space (DS) Robust method optimization Inter-laboratory method transfer Supercritical Fluid Chromatography (SFC)
a b s t r a c t The concept of Quality by Design (QbD) is now well established in pharmaceutical industry and should be applied to the development of any analytical methods. In this context, the key concept of Design Space (DS) was introduced in the field of analytical method optimization. In chromatographic words, the DS is the space of chromatographic conditions that will ensure the quality of peaks separation, thus DS is a zone of robustness. In the present study, the interest of robust method optimization strategy was investigated in the context of direct method transfer from sending to receiving laboratory. The benefit of this approach is to speed up the method life cycle by performing only one quantitative validation step in the final environment of method use. A Supercritical Fluid Chromatography (SFC) method previously developed was used as a case study in this work. Moreover, the interest of geometric transfer was investigated simultaneously in order to stress a little bit more the transfer exercise and, by the way, emphasize the additional benefit of DS strategy in this particular context. Three successful transfers were performed on two column geometries. In order to compare original and transferred methods, the observed relative retention times (RT) were modelled as a function of the predicted relative RT and of the method type (original or transferred). The observed relative RT of the original and transferred methods are not statistically different and thus the method transfer is successfully achieved thanks to the robust optimization strategy. Furthermore, the analytical method was improved considering analysis time (reduced five times) and peak capacity (increased three times). To conclude, the advantage of using a DS strategy implemented for the optimization and transfer of SFC method was successfully demonstrated in this work. © 2013 Elsevier B.V. All rights reserved.
1. Introduction The transfer of a method from a sending laboratory (called sender, i.e., R&D laboratory) to a receiving laboratory (called receiver, i.e., quality control (QC) laboratory) is an important step in an analytical method life cycle. Method transfer can be performed using different methodologies [1,2]. The general decisional flowchart (Quality by Testing (QbT) methodology or trial and error) used to answer an analytical problem is technology selection, then method development followed by method validation and robustness studies. After all these steps, method transfer to a receiving laboratory could be considered before use in routine analysis. Nevertheless, the interest of robust optimization was previously introduced [3]. In the case of method transfer, a robust optimization strategy could help to directly transfer the method after its development without its quantitative validation from the sending lab to the receiving lab. Indeed in many cases the sending laboratory is a R&D unit that will never use the developed analytical
∗ Corresponding author. Tel.: +32 4366 4319; fax: +32 4366 4317. E-mail address: [email protected] (A. Dispas). 0731-7085/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jpba.2013.09.030
method to perform quantitative measurements of a compound in a drug substance or drug product. Hence the initial validation of the quantitative performances of the method at the end of the development phase at the sending lab could be bypassed to speed-up its life cycle. This direct transfer is the first objective investigated in the present work. Obviously, a formal validation phase will always remain necessary before any quantitative analysis in the receiving – QC laboratory, which is definitively its final goal. Robust optimization of analytical methods is fully included into the concept of Quality by Design (QbD). QbD is defined by ICH Q8 R2 [4] as “a systematic approach to development that begins with predefined objectives and emphasises product and process understanding based on sound science and quality risk management”. An analytical method can be considered as a process that must have an output of acceptable quality [5,6]. Borman et al. [7] demonstrated that the QbD concept for manufacturing processes could also be applied to analytical methods [8]. In this context, the Design Space (DS) was introduced as a key component of analytical method development [6,9]. The DS is defined as “the multidimensional combination and interaction of input variables (e.g., material attributes) and process parameters that have been demonstrated to provide assurance of quality” [4]. Thus, the DS is a subspace of the experimental
520
A. Dispas et al. / Journal of Pharmaceutical and Biomedical Analysis 88 (2014) 519–524
domain in which the assurance of quality has been proved. For chromatographic method development, the DS could be defined as the space of chromatographic conditions that will ensure the quality of the separation. Therefore, the method robustness is guaranteed inside the DS limits. To obtain an analytical DS, Designs of Experiments (DoE) play an important role. Nonetheless, method development processes, that do not take into account the prediction error and its propagation in order to manage the risk, cannot be considered as fully QbD compliant and will not allow to compute the DS. Consequently, relying only on mean response surfaces obtained using DoE to define optimal condition does not plainly fulfil the requirements of quality risk management defined by QbD. A DoE-DS strategy taking into account uncertainty of the process is mandatory to perform robust optimization, DoE is not enough in this context. The analytical method transferred in this case study is a Supercritical Fluid Chromatography (SFC) method aiming at separating polar compounds and developed using a DoE-DS methodology compliant with the QbD principles [10]. SFC is a separative technique knowing a resurgence of interest [11,12]. The many advantages of this green technique, such as high throughput or improved chromatographic performances are now worldwide approved. An additional constraint of the analytical method transfer studied in this case study is that the receiver’s equipment is not a conventional SFC but a new SFC system dedicated to perform Ultra High Performance Supercritical Fluid Chromatography (UHPSFC). This difference in equipment between the two laboratories suggests that a geometric transfer could be performed simultaneously with the analytical method transfer. The interest of sub 2 m particles was described in neat CO2 mobile phase [13] or with modifier added in the CO2 [14,15]. Hence, in addition to be the first inter-laboratory SFC method transfer, this work demonstrates the feasibility of geometric transfer in SFC as a case study of the high added value of robust DoE-DS optimization strategy. To our knowledge, direct transfer of a method developed on conventional particles size (e.g., 5 m) to sub 2 m particles was not yet published. In the context of green analytical chemistry, more and more pharmaceutical industries showed their interest for rapid, lowcost and eco-friendly analytical methods. The strategy proposed in this paper enables to transfer existing method (i.e., SFC) to fast method (i.e., UHPSFC) without a time-consuming transfer protocol. Even if some theoretical aspect of SFC geometric transfer remains to be studied in a further work, the proposed methodology could be applied by many industrial users interested by a rapid implementation of the UHPSFC in their laboratories.
2.2. Standard samples preparation According to their UV absorbance, the compounds were divided into three groups. All stock solutions were prepared in pure methanol. Stock solutions of each compound from group 1 (caffeine and serotonin) were obtained by dissolving 10 mg in a volumetric flask of 10.0 ml. Group 2 contains histamine, dopamine, noradrenalin and paroxetine. Stock solutions of each compound from group 2 were obtained by dissolving 50 mg in a volumetric flask of 10.0 ml. Group 3 consists of pseudoephedrine, prepared by dissolving 100 mg in a volumetric flask of 10.0 ml. All the stock solutions were stored at −27 ◦ C. Working solutions were daily prepared by diluting stock solutions in methanol/2-propanol/heptane 1/4/5 (v/v/v) to obtain concentrations of 20 g ml−1 , 100 g ml−1 , and 250 g ml−1 respectively. 2.3. Instrumentation At the sending site, the analyses were performed on an a FusionTM A5 SFC conversion module (Aurora SFC Systems, Inc., Redwood City, CA) with an Agilent Technologies (Waldbron, Germany) HPLC 1100 series, previously described [10]. At the receiving site, a Waters® Acquity UPC2 equipped with a PDA detector was used to carry out the experiments. This new system is a dedicated SFC instrumentation optimized to reduce void volumes. The dwell volume measured was 1.4 ml for SFC system and 0.4 ml for UPC2 system. The extra-column volume was estimated at 35 l and 60 l, respectively. The injector was equipped with a 10 l loop operating in the partial loop with needle overfill mode, with methanol and methanol/2-propanol/heptane 1/4/5 (v/v/v) as strong and weak needle wash, respectively. A Viridis SFC 2-ethylpyridine column (100 mm × 4.6 mm I.D., particles size: 5 m), Acquity BEH 2-ethylpyridine column (100 mm × 3 mm I.D., particles size: 1.7 m) and Acquity BEH 2ethylpyridine column (50 mm × 3 mm I.D., particles size: 1.7 m), all from Waters® , were used to carry out SFC and UPC2 experiments, respectively. The mobile phase composition was optimized previously [10], binary gradient elution profile was performed using CO2 as mobile phase A and a solution of 25 mM TFA diluted into methanol as mobile phase B. The outlet pressure was set at 200 bar and the oven temperature at 60.5 ◦ C. Data acquisition was made at a wavelength of 220 nm. Chemstation (Agilent® ) and Empower 3 (Waters® ) were used to control the sending and receiving equipment, respectively. HPLC calculator [16,17] was used to calculate the gradients for method transfer. JMP 10.0.0 for Mac was used to carry out the statistical analyses.
2. Materials and methods
3. Results and discussion
2.1. Chemicals and reagents
3.1. Robust method optimization
Histamine dihydrochloride (99.7%), paroxetine hydrochloride hemihydrate (100.5%), cetirizine dihydrochloride (100.5%), and caffeine (100.1%) were purchased from Fagron (Waregem, Belgium). Pseudoephedrine hydrochloride (100.4%) was purchased from Certa (Braine-l’Alleud, Belgium). Serotonin hydrochloride (≥98%), dopamine hydrochloride (batch 0001437030), and DLnorepinephrine hydrochloride (≥97%) were provided by Sigma Aldrich (St. Louis, MO, USA). Methanol (Ultra LC/MS), 2-propanol (LC/MS grade), n-heptane (HPLC solvents) and trifluoroacetic acid (99%, Ultra LC/MS) (TFA) were purchased from J.T. Baker (Deventer, Netherlands). Carbon dioxide 4.5 (99.995%) was purchased from Westfalen (Brussels, Belgium).
In order to be compliant to QbD requirements, the SFC method was first developed at the sending lab using Design Space strategy [10]. In the previous study, a chromatographic screening to select the nature of stationary and mobile phase was performed. Then, the robust method optimization was performed by means of design of experiments and design space computation. A four factors central composite design was tested. The four selected factors were the gradient slope (modified by changing the time to change the proportion of methanol form 5 to 40%), the concentration of TFA dissolved in methanol, the isocratic time before the gradient and the temperature. Then, the retention times (at the beginning, the apex and the end of the peak) were recorded and the retention factors were modelled by an identical polynomial
A. Dispas et al. / Journal of Pharmaceutical and Biomedical Analysis 88 (2014) 519–524
521
3.2. Inter-laboratory method transfer Contrary to Quality by Testing strategy [1], the method was simply transferred and tested from the sender to the receiver, without a quantitative validation of the method by the sender. The transfer was performed on a different equipment (in the receiving lab), on two column geometries at different flow rates in order to assess the robustness of the method obtained with the DoE-DS optimization methodology and to maximize the chromatographic performances. For the following discussions, the initial method (sending lab) is called “SFC” and the transferred methods (receiving lab) are called “UHPSFC”.
Fig. 1. Three dimensions experimental domain of SFC method. Factor ‘isocratic time’ was set at 3 min. White P(S > 0; ttot < 15 min) < 0.3; Blue P(S > 0; ttot < 15 min) = 0.3–0.35; Purple P(S > 0; ttot < 15 min) = 0.35–0.4; Red: P(S > 0; ttot < 15 min) = 0.4–0.45. DS includes purple and red zones. Optimal condition is defined as red space.
equation using a multiple linear equation regression maximizing the adjusted coefficient of determination. Retention times were predicted using the developed model. The predictive distribution of new responses is then used, accounting for both the residual error and the model estimation uncertainty. The errors obtained for the predicted retention times are propagated to the separation (S). Then, Monte–Carlo simulations were used to obtain the distribution of S from the distribution of retention times of the critical peaks pair, for a given chromatographic condition. Finally, considering the distribution of retention times and S, the posterior probability for S to be higher than 0 and analysis time < 15 min (ttot ) was used to compute the DS. A probability of about 0.5 was identified. Experimentally, it led to a very satisfactory chromatographic separation. As shown in Fig. 1, the DS is the area of chromatographic conditions that will ensure the quality of the separation. In this study, the design space is the area in which a separation between peaks (S) is superior than 0 min and the time of analysis (ttot ) is inferior than 15 min. Thus, the DS is a robustness area as previously described [10,18]; indeed, the factors (e.g., gradient time, temperature, etc.) were modified in the limits of DS and the quality of separation was successfully tested. This methodology allowed defining optimal chromatographic conditions; the observed SFC chromatogram is shown in Fig. 2. Extensive discussion of these results has been previously published [10]. Furthermore, to challenge the robustness of the optimal analytical conditions obtained during the SFC method development, inter-laboratory transfer and geometric transfer are simultaneously investigated. The objective of this complex transfer was to stress the analytical procedure by modifying the lab, the equipment and the column geometry in order to demonstrate the added value of the DS optimization strategy. The separation of all compounds in the transferred method is both the demonstration of acceptability of the analytical method transfer and the further confirmation of the method robustness.
3.2.1. Geometric method transfer As the receiver’s SFC equipment is not a conventional one but a UHPSFC a geometric transfer is investigated. Three steps are critical for geometric method transfer [19]. Firstly, the selection of stationary phase chemistry with an identical selectivity will assure the efficient method transfer. For this purpose, the ideal situation is to work with a stationary phase identical in nature but different in geometry. However, the stationary phase chemistry (2-ethylpyridine) used to develop SFC method was not commercially available on sub 2 m particles. Then, it is necessary to find a column that has retention and selectivity properties as close as possible to the original one. Consequently, a BEH (ethylene bridged hybride) 2-ethylpyridine stationary phase was selected to perform the method transfer. The difference between the two stationary phases used could be a potential source of selectivity variability. Secondly, the flow rate and gradient step times have to be correctly recalculated by means of equations derived from chromatography theory [17]: FUHPSFC = FSFC ×
2 dcUHPSFC 2 dcSFC
×
dpSFC dpUHPSFC
(1)
where F is the mobile phase flow rate, dc the column internal diameter and dp the particles diameter. For the longer sub 2 m column (100 mm), the geometric transfer of flow rate was not technically possible for too high backpressure reasons. Indeed, the modifier added in the CO2 mobile phase caused a higher backpressure than conventional neat CO2 [13]. SFC system equipped with ultra high pressure pump should be used to perform separation on longer column with small particles. Anyway, for the demonstration presented in this paper, an optimized transfer gradient profile (non-optimal flow rate) was developed. For the short column, the geometric transfer at the optimal flow rate was performed without any pressure problem. The isocratic and gradient steps were recalculated according to Eqs. (2) and (3): tisoUHPSFC = tisoSFC × SGUHPSFC = SGSFC ×
V0 FSFC × UHPSFC FUHPSFC V0SFC V0SFC
V0UHPSFC
×
FUHPSFC FSFC
(2) (3)
where tiso is the isocratic time, SG is the gradient slope and V0 is the column dead volume. Thirdly, the system used to carry out the experiments has to be optimized to reduce extra-column and dwell volume. Indeed, the system void volume should be reduced to minimize band broadening. These three rules were defined for liquid chromatography method transfer. In the case of SFC method, the impact of pressure should be considered because of its impact on fluid density. Despite this potential effect, classical geometrical rules were firstly used in this work. Most likely due to the high robustness of the method, no significant change of selectivity due to fluid density variation was observed. Thus, no further investigation were carried out in the present study.
522
A. Dispas et al. / Journal of Pharmaceutical and Biomedical Analysis 88 (2014) 519–524
3.2.2. Advantages of SFC sub 2 m In order to compare chromatographic performances of original and transferred methods, peak capacity (P), best measurement of the performance of gradient elution, was calculated by Eq. (4) proposed by Neue [20]: P=
Fig. 2. Chromatograms (top to bottom) RECEIVER: UHPSFC short column transferred method at 2.5 ml min−1 ; RECEIVER: UHPSFC transferred method at 1.8 ml min−1 ; RECEIVER: UHPSFC transferred method at 1 ml min−1 ; SENDER: original SFC method. Elution order: 1, Caffeine; 2, Pseudoephedrine; 3, Paroxetine; 4, Cetirizine; 5, Histamine; 6, Dopamine; 7, Norepinephrine; 8, Serotonin
Two different approaches were investigated to perform the geometric method transfer. Firstly, a low flow rate (1 ml min−1 ) gradient was recalculated for the long column (Fig. 2). After this low pressure transfer, the flow rate was progressively increased (from 1.5 to 1.8 ml min−1 ) while keeping a backpressure lower than 400 bars. Consequently, two transfers on the same column geometry were successfully performed (at 1 and 1.8 ml min−1 ). Secondly, the method was transferred on a short column in order to drastically decrease the analysis time. The gradient profile of the original SFC and transferred UHPSFC methods are shown in Table 1. However, the pressures measured on packed column with 5 m and 1.7 m (and at different flow rates) were different, impacting directly on fluid density and hence retention and selectivity are modified. Nevertheless, the high amount of modifier used slightly decreased the influence of CO2 density on the polarity of mobile phase. This phenomenon should be further studied and considered for method transfers. Despite these changes on fluid density and then on selectivity, the separation of all peaks was kept after method transfer. This study is the first demonstration of the feasibility of SFC method transfer. The influence of pressure would be certainly studied more deeply in further work, focused on theoretical aspect of SFC geometric transfer.
tG (1/n)
n 1
(4)
ω
where P is peak capacity, tG gradient run time, n the number of peaks selected for the calculation and ω is the peak width. Peak capacity was calculated considering the seven peaks eluted during the gradient time. P was estimated at 29 for SFC method and at 40 for UHPSFC method (long column, maximal flow rate); an improvement of 35% was thus achieved by this method transfer, considering the shorter analysis time (reduced twice). Furthermore, the resolution (except for the peaks pair 5–6) was significantly improved by using UHPSFC. The SFC method transfer on short sub 2 m column was highly valuable considering analysis time and peak capacity. Indeed, this method was able to reach a peak capacity of 72 for an analysis time lower than 2 min. Of course, the resolution was lower in the case of very fast separation (UHPSFC short column). Generally, UHPSFC showed better chromatographic performances than SFC, the column length should be selected according to the requirements of the method (fastness, resolution, peak capacity, etc.). As mentioned previously [10] polar compounds are not the easiest case study for SFC separation, higher chromatographic performances should be obtained with less polar compounds. 3.2.3. Assessment of the analytical method transfer In order to compare the elution profile of the methods, relative retention times were calculated by dividing each retention time (RT) by the RT of the last compound (the most influenced by the gradient). As shown in Table 2, the predicted and observed relative RT were close, illustrating the adequacy of the transfer process. Obviously, the tiny differences observed could be easily explained by the slight changes of selectivity discussed above. To confirm these observations, the observed relative RT were modelled as a function of the predicted relative RT and of the method type (original method (SFC) or transferred method (UHPSFC at 1.0 ml min−1 , UHPSFC at 1.8 ml min−1 and UHPSFC short column)). A quadratic effect of predicted relative RT as well as an interaction term between method type and predicted relative RT were 2 = also included into the model. Fig. 3 illustrates this model (Radj 0.9882). First, the method × predicted relative RT interaction is not significant (p-value = 0.4650) as well as the method main effect (pvalue = 0.0668). This highlights that no differences between the original SFC of the sender and the transferred UHPSFC methods have been detected showing the reliability of the transfer methodology applied. In addition, the prediction of the relative RT is shown to be sufficiently accurate even if the predicted relative RT2 term is significant (p-value < 0.0001). This statistically significant effect has no practical implications. The quadratic term merely signifies that the correspondence between the RT predicted by the transfer
Table 1 Gradient profile of SFC and transferred UHPSFC methods. Time (min) Flow rate (ml min−1 )
2 ml min−1
CO2 (%)
MeOH TFA 25 mM (%)
SFC
95 95 60 60 95 95
5 5 40 40 5 5
0 3 12.2 17.2 17.7 30
1 ml min−1
1.8 ml min−1
UHPSFC long column 0 2.75 10.57 14.83 15.25 25.72
0 1.53 5.87 6.25 6.49 14.29
2.5 ml min−1 UHPSFC short column 0 0.47 2.03 2.88 2.97 5.06
Obs rel RT UHPSFC Short column
0.125 0.613 0.675 0.845 0.863 0.889 0.945 1.000
Obs RT UHPSFC Short column
0.213 1.046 1.152 1.441 1.472 1.517 1.612 1.706
523
1.32 4.8 6.4 7.5 8.56 9.71 10.46 11.04
0.12 0.435 0.579 0.679 0.775 0.879 0.948 1.000
1.549 5.064 6.17 7.569 7.784 8.916 9.519 9.732
0.159 0.52 0.634 0.778 0.8 0.916 0.978 1.000
0.884 4.239 5.359 6.526 6.859 7.488 8.012 8.379
0.106 0.506 0.64 0.779 0.819 0.894 0.956 1.000
0.465 2.223 2.873 3.537 3.758 4.099 4.398 4.609
0.101 0.482 0.623 0.767 0.815 0.889 0.954 1.000
Fig. 3. Relative observed retention time versus relative predicted retention time, quadratic model. Red line SENDER: original SFC method; green line RECEIVER: UHPSFC transferred method at 1 ml min−1 ; blue line RECEIVER: UHPSFC transferred method at 1.8 ml min−1 ; brown line RECEIVER: UHPSFC short column transferred method at 2.5 ml min−1 .
CAF PSE PAR CET HIS DOP NOR SERO
Obs rel RT UHPSFC 1.8 ml min−1 Obs RT UHPSFC 1.8 ml min−1 Obs rel RT UHPSFC 1 ml min−1 UHPSFC
Obs RT UHPSFC 1 ml min−1 Obs RT rel
SFC
Obs RT Pred RT rel
Predicted SFC
Pred RT
Compound
Table 2 (Columns 2–3) SFC predicted (pred) retention times, relative (rel) predicted retention times; (columns 4–5) SFC respective observed and relative retention times; (columns 6–9) respective observed and relative retention times for UHPSFC at 1 and 1.8 ml min−1 ; (columns 10–11) respective observed and relative retention times for UHPSFC on short column.
A. Dispas et al. / Journal of Pharmaceutical and Biomedical Analysis 88 (2014) 519–524
and the true RT are not perfectly proportional. Second, the slope and the intercept are also both significant. In case of perfect transfer prediction, one would expect the slope to be one (here it is estimated at 0.87 ± 0.05 at 95% confidence level) and the intercept to be zero (i.e., the intercept should be non-significant, here the intercept is 0.17 ± 0.03 at 95% confidence level). These two facts can be explained by many reasons, the most prominent ones being the change of stationary phase and the application of the classical geometric transfer computation for an SFC method. More research on this subject is necessary to improve the transfer prediction. These results demonstrated and confirmed the high interest of robust method optimization using a DoE-DS strategy. Indeed, the robustness was evaluated during the method development at the sending laboratory contrary to more common strategies where the robustness is evaluated after the quantitative validation step at the sending site before its transfer to the QC laboratory. To evaluate the robustness during development step is now a clear recommendation of the U.S. Pharmacopeial Convention [21]. A method transfer could be successfully achieved even after having bypassed the quantitative validation study and the conventional robustness study, at the condition that no formal quantitative validation is required at the sending laboratory (Fig. 4). Thus, if a method is developed in a laboratory and used in routine in a second laboratory, it is no longer required to validate the quantitative performances of the method before the lab transfer thanks to the robust optimization strategy. Obviously, a formal full quantitative validation of the method remains essential before performing quantitative analysis in routine. However, this one could only be considered after its transfer, i.e., at the QC laboratory were the method will effectively be used in routine analysis. Furthermore, considering the QbT methodology, the method transfer consists of a formal step involving assessment of the quantitative performances of the method. To evaluate the quantitative performances at the receiving lab, several strategies are used [1] but a partial or even more a full validation is generally performed, meaning that two validations are performed during the life cycle of the method. The present strategy (QbD) allows to speed up the life cycle of an analytical method by focusing on the real aim of the R&D laboratory and of the QC laboratory (Fig. 4). Hence, time, resources and money are saved without impairing the quality of the analytical method released to the QC site. In the context of QbD methodology, the DoEDS strategy is the critical step providing the guarantee of successful transfer of an optimal method from the sending to the receiving lab. Therefore, the quality of transfer is assessed at the beginning of analytical method life cycle during robust optimization phase.
524
A. Dispas et al. / Journal of Pharmaceutical and Biomedical Analysis 88 (2014) 519–524
Fig. 4. Analytical method Life Cycle flowchart: Quality by Testing (QbT) versus Quality by Design (QbD) methodology.
4. Conclusion The advantage of robust method optimization was definitively demonstrated in the present paper. Indeed, the high value of the DS strategy implemented for the optimization of SFC separation was successfully established in this work. In order to speed its life cycle, only a robust method optimization, including the concept of DoE and DS, was required to guarantee the success of the transfer bypassing the usual method validation and robustness study steps performed into the R&D laboratory (sender) (Fig. 4). Indeed, the chromatographic separation was not degraded by the variability induced by changing the lab, the equipment and by the geometric transfer thanks to the high reliability of the DS methodology. Furthermore, in the context of QbD methodology, the guarantee of a successful method transfer is already assessed during the robust optimization step performed at the sending lab contrary to the QbT methodology. However, it is clear that a formal quantitative validation phase will always remain mandatory before any quantitative analysis. The main benefit of the present strategy is to limit or focalize the full quantitative validation of the method in the environment in which it will effectively be used in routine analysis. Performing method validation in the QC lab environment allows to minimize the risk to obtain Out Of Specifications results. This work is also the demonstration of the feasibility of SFC geometric method transfer. Despite the selectivity changes (stationary phases and fluid density) these transfers were successful, showing the interest of SFC and UHPSFC for the future of analytical chemistry even if the study of fluid density implication in transfer calculations should be considered as future work. Finally, using sub 2 m stationary phase with SFC presents huge advantages such as the possibility to increase peak capacity and reduce analyze times. Ultra High Performance Supercritical Fluid Chromatography (UHPSFC) is clearly a very promising technique in the framework of green analytical chemistry. References [1] E. Rozet, W. Dewé, E. Ziemons, A. Bouklouze, B. Boulanger, Ph. Hubert, Methodologies for the transfer of analytical methods: a review, J. Chromatogr. B 877 (2009) 2214–2223. [2] E. Rozet, B. Mertens, W. Dewe, A. Ceccato, B. Govaerts, B. Boulanger, P. Chiap, B. Streel, J. Crommen, Ph. Hubert, The transfer of LC–UV method for the
[3] [4]
[5] [6] [7] [8]
[9]
[10]
[11] [12] [13]
[14] [15]
[16]
[17]
[18]
[19]
[20] [21]
determination of fenofibrate and fenofibric acide in Lidoses: Use total error as decision criterion, J. Pharm. Biomed. Anal. 42 (2006) 64–70. B. Dejaegher, Y. Vander Heyden, Robustness tests, LC–GC Europe 19 (2006) 418–423. The international Conference of Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use; Q8 R2 Pharmaceutical Development, http://www.ich.org/fileadmin/Public Web Site/ICH Products/ Guidelines/Quality/Q8 RI/Step4/Q8 R2 Guidelines.pdf (accessed June 2011). P. Nethercote, J. Ermer, Quality by design for analytical methods, Pharm. Technol. 36 (2012) 52–55. E. Rozet, P. Lebrun, B. Debrus, B. Boulanger, P. Hubert, Design Spaces for analytical methods, TrAC 42 (2013) 157–167. P. Borman, P. Nethercote, M. Chatfield, D. Thompson, K. Truman, The application of quality by design to analytical methods, Pharm. Technol. 31 (2007) 142–152. S. Orlandini, S. Pinzauti, S. Furlanetto, Application of quality by design to the development of analytical separation methods, Anal. Bioanal. Chem. 405 (2013) 443–450. P. Lebrun, B. Govaerts, B. Debrus, A. Ceccato, G. Caliaro, P. Hubert, B. Boulanger, Development of a new predictive modelling technique to find with confidence equivalence zone and design space of chromatographic analytical methods, Chemom. Intell. Lab. Syst. 91 (2008) 4–16. A. Dispas, P. Lebrun, P. Sassiat, E. Ziemons, D. Thiébaut, J. Vial, Ph. Hubert, Innovative green supercritical fluid chromatography for the determination of polar compounds, J. Chromatogr. A 1256 (2012) 253–326. L.T. Taylor, Supercritical fluid chromatography, Anal. Chem. 80 (2008) 4285–4294. L.T. Taylor, Supercritical fluid chromatography, Anal. Chem. 82 (2012) 4925–4935. C. Sarazin, D. Thiébaut, P. Sassiat, J. Vial, Feasibility of ultra high performance supercritical neat carbon dioxide chromatography at conventional pressures, J. Sep. Sci. 34 (2011) 2773–2778. T.A. Berger, Demonstration of high speeds with low pressure drops using 1.8 m particles in SFC, Chromatographia 72 (2010) 597–602. A. Grand-Guillaume Perrenoud, J.-L. Veuthey, D. Guillarme, Comparison of ultra-high performance supercritical fluid chromatography and ultra-high performance liquid chromatography for the analysis of pharmaceutical compounds, J. Chromatogr. A 1266 (2012) 158–167. D. Guillarme, D.T.-T. Nguyen, S. Rudaz, J.-L. Veuthey, Method transfer for fast liquid chromatography in pharmaceutical analysis: application to short column packed with small particle. Part I: Isocratic separation, Eur. J. Pharm. Biopharm. 66 (2007) 475–482. D. Guillarme, D.T.-T. Nguyen, S. Rudaz, J.-L. Veuthey, Method transfer for fast liquid chromatography in pharmaceutical analysis: Application to short column packed with small particle. Part II: Gradient experiments, Eur. J. Pharm. Biopharm. 68 (2008) 430–440. P. Lebrun, B. Boulanger, B. Debrus, Ph. Lambert, Ph. Hubert, A Bayesian design space for analytical methods based on multivariate models and predictions, J. Biopharm. Stat. (2013) (in press). B. Debrus, E. Rozet, Ph. Hubert, J.-L. Veuthey, S. Rudaz, D. Guillarme, UHPLC in Life Sciences, The Royal Society of Chemistry, Cambridge, 2012, pp. 67–103. U.D. Neue, Theory of peak capacity in gradient elution, J. Chromatogr. A 1079 (2005) 153–161. U.S. Pharmacopeial Convention, new chapter 1224, 1225, 1226, USP panel expert.
Contents lists available at ScienceDirect
Journal of Pharmaceutical and Biomedical Analysis journal homepage: www.elsevier.com/locate/jpba
Robust method optimization strategy—A useful tool for method transfer: The case of SFC Amandine Dispas ∗ , Pierre Lebrun, Bertyl Andri, Eric Rozet, Philippe Hubert University of Liege (Ulg), Department of Pharmacy, CIRM, Laboratory of Analytical Chemistry, 1 Avenue de l’Hôpital, CHU B36, B-4000 Liège, Belgium
a r t i c l e
i n f o
Article history: Received 28 June 2013 Received in revised form 2 September 2013 Accepted 26 September 2013 Available online 12 October 2013 Keywords: Quality by Design (QbD) Design Space (DS) Robust method optimization Inter-laboratory method transfer Supercritical Fluid Chromatography (SFC)
a b s t r a c t The concept of Quality by Design (QbD) is now well established in pharmaceutical industry and should be applied to the development of any analytical methods. In this context, the key concept of Design Space (DS) was introduced in the field of analytical method optimization. In chromatographic words, the DS is the space of chromatographic conditions that will ensure the quality of peaks separation, thus DS is a zone of robustness. In the present study, the interest of robust method optimization strategy was investigated in the context of direct method transfer from sending to receiving laboratory. The benefit of this approach is to speed up the method life cycle by performing only one quantitative validation step in the final environment of method use. A Supercritical Fluid Chromatography (SFC) method previously developed was used as a case study in this work. Moreover, the interest of geometric transfer was investigated simultaneously in order to stress a little bit more the transfer exercise and, by the way, emphasize the additional benefit of DS strategy in this particular context. Three successful transfers were performed on two column geometries. In order to compare original and transferred methods, the observed relative retention times (RT) were modelled as a function of the predicted relative RT and of the method type (original or transferred). The observed relative RT of the original and transferred methods are not statistically different and thus the method transfer is successfully achieved thanks to the robust optimization strategy. Furthermore, the analytical method was improved considering analysis time (reduced five times) and peak capacity (increased three times). To conclude, the advantage of using a DS strategy implemented for the optimization and transfer of SFC method was successfully demonstrated in this work. © 2013 Elsevier B.V. All rights reserved.
1. Introduction The transfer of a method from a sending laboratory (called sender, i.e., R&D laboratory) to a receiving laboratory (called receiver, i.e., quality control (QC) laboratory) is an important step in an analytical method life cycle. Method transfer can be performed using different methodologies [1,2]. The general decisional flowchart (Quality by Testing (QbT) methodology or trial and error) used to answer an analytical problem is technology selection, then method development followed by method validation and robustness studies. After all these steps, method transfer to a receiving laboratory could be considered before use in routine analysis. Nevertheless, the interest of robust optimization was previously introduced [3]. In the case of method transfer, a robust optimization strategy could help to directly transfer the method after its development without its quantitative validation from the sending lab to the receiving lab. Indeed in many cases the sending laboratory is a R&D unit that will never use the developed analytical
∗ Corresponding author. Tel.: +32 4366 4319; fax: +32 4366 4317. E-mail address: [email protected] (A. Dispas). 0731-7085/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jpba.2013.09.030
method to perform quantitative measurements of a compound in a drug substance or drug product. Hence the initial validation of the quantitative performances of the method at the end of the development phase at the sending lab could be bypassed to speed-up its life cycle. This direct transfer is the first objective investigated in the present work. Obviously, a formal validation phase will always remain necessary before any quantitative analysis in the receiving – QC laboratory, which is definitively its final goal. Robust optimization of analytical methods is fully included into the concept of Quality by Design (QbD). QbD is defined by ICH Q8 R2 [4] as “a systematic approach to development that begins with predefined objectives and emphasises product and process understanding based on sound science and quality risk management”. An analytical method can be considered as a process that must have an output of acceptable quality [5,6]. Borman et al. [7] demonstrated that the QbD concept for manufacturing processes could also be applied to analytical methods [8]. In this context, the Design Space (DS) was introduced as a key component of analytical method development [6,9]. The DS is defined as “the multidimensional combination and interaction of input variables (e.g., material attributes) and process parameters that have been demonstrated to provide assurance of quality” [4]. Thus, the DS is a subspace of the experimental
520
A. Dispas et al. / Journal of Pharmaceutical and Biomedical Analysis 88 (2014) 519–524
domain in which the assurance of quality has been proved. For chromatographic method development, the DS could be defined as the space of chromatographic conditions that will ensure the quality of the separation. Therefore, the method robustness is guaranteed inside the DS limits. To obtain an analytical DS, Designs of Experiments (DoE) play an important role. Nonetheless, method development processes, that do not take into account the prediction error and its propagation in order to manage the risk, cannot be considered as fully QbD compliant and will not allow to compute the DS. Consequently, relying only on mean response surfaces obtained using DoE to define optimal condition does not plainly fulfil the requirements of quality risk management defined by QbD. A DoE-DS strategy taking into account uncertainty of the process is mandatory to perform robust optimization, DoE is not enough in this context. The analytical method transferred in this case study is a Supercritical Fluid Chromatography (SFC) method aiming at separating polar compounds and developed using a DoE-DS methodology compliant with the QbD principles [10]. SFC is a separative technique knowing a resurgence of interest [11,12]. The many advantages of this green technique, such as high throughput or improved chromatographic performances are now worldwide approved. An additional constraint of the analytical method transfer studied in this case study is that the receiver’s equipment is not a conventional SFC but a new SFC system dedicated to perform Ultra High Performance Supercritical Fluid Chromatography (UHPSFC). This difference in equipment between the two laboratories suggests that a geometric transfer could be performed simultaneously with the analytical method transfer. The interest of sub 2 m particles was described in neat CO2 mobile phase [13] or with modifier added in the CO2 [14,15]. Hence, in addition to be the first inter-laboratory SFC method transfer, this work demonstrates the feasibility of geometric transfer in SFC as a case study of the high added value of robust DoE-DS optimization strategy. To our knowledge, direct transfer of a method developed on conventional particles size (e.g., 5 m) to sub 2 m particles was not yet published. In the context of green analytical chemistry, more and more pharmaceutical industries showed their interest for rapid, lowcost and eco-friendly analytical methods. The strategy proposed in this paper enables to transfer existing method (i.e., SFC) to fast method (i.e., UHPSFC) without a time-consuming transfer protocol. Even if some theoretical aspect of SFC geometric transfer remains to be studied in a further work, the proposed methodology could be applied by many industrial users interested by a rapid implementation of the UHPSFC in their laboratories.
2.2. Standard samples preparation According to their UV absorbance, the compounds were divided into three groups. All stock solutions were prepared in pure methanol. Stock solutions of each compound from group 1 (caffeine and serotonin) were obtained by dissolving 10 mg in a volumetric flask of 10.0 ml. Group 2 contains histamine, dopamine, noradrenalin and paroxetine. Stock solutions of each compound from group 2 were obtained by dissolving 50 mg in a volumetric flask of 10.0 ml. Group 3 consists of pseudoephedrine, prepared by dissolving 100 mg in a volumetric flask of 10.0 ml. All the stock solutions were stored at −27 ◦ C. Working solutions were daily prepared by diluting stock solutions in methanol/2-propanol/heptane 1/4/5 (v/v/v) to obtain concentrations of 20 g ml−1 , 100 g ml−1 , and 250 g ml−1 respectively. 2.3. Instrumentation At the sending site, the analyses were performed on an a FusionTM A5 SFC conversion module (Aurora SFC Systems, Inc., Redwood City, CA) with an Agilent Technologies (Waldbron, Germany) HPLC 1100 series, previously described [10]. At the receiving site, a Waters® Acquity UPC2 equipped with a PDA detector was used to carry out the experiments. This new system is a dedicated SFC instrumentation optimized to reduce void volumes. The dwell volume measured was 1.4 ml for SFC system and 0.4 ml for UPC2 system. The extra-column volume was estimated at 35 l and 60 l, respectively. The injector was equipped with a 10 l loop operating in the partial loop with needle overfill mode, with methanol and methanol/2-propanol/heptane 1/4/5 (v/v/v) as strong and weak needle wash, respectively. A Viridis SFC 2-ethylpyridine column (100 mm × 4.6 mm I.D., particles size: 5 m), Acquity BEH 2-ethylpyridine column (100 mm × 3 mm I.D., particles size: 1.7 m) and Acquity BEH 2ethylpyridine column (50 mm × 3 mm I.D., particles size: 1.7 m), all from Waters® , were used to carry out SFC and UPC2 experiments, respectively. The mobile phase composition was optimized previously [10], binary gradient elution profile was performed using CO2 as mobile phase A and a solution of 25 mM TFA diluted into methanol as mobile phase B. The outlet pressure was set at 200 bar and the oven temperature at 60.5 ◦ C. Data acquisition was made at a wavelength of 220 nm. Chemstation (Agilent® ) and Empower 3 (Waters® ) were used to control the sending and receiving equipment, respectively. HPLC calculator [16,17] was used to calculate the gradients for method transfer. JMP 10.0.0 for Mac was used to carry out the statistical analyses.
2. Materials and methods
3. Results and discussion
2.1. Chemicals and reagents
3.1. Robust method optimization
Histamine dihydrochloride (99.7%), paroxetine hydrochloride hemihydrate (100.5%), cetirizine dihydrochloride (100.5%), and caffeine (100.1%) were purchased from Fagron (Waregem, Belgium). Pseudoephedrine hydrochloride (100.4%) was purchased from Certa (Braine-l’Alleud, Belgium). Serotonin hydrochloride (≥98%), dopamine hydrochloride (batch 0001437030), and DLnorepinephrine hydrochloride (≥97%) were provided by Sigma Aldrich (St. Louis, MO, USA). Methanol (Ultra LC/MS), 2-propanol (LC/MS grade), n-heptane (HPLC solvents) and trifluoroacetic acid (99%, Ultra LC/MS) (TFA) were purchased from J.T. Baker (Deventer, Netherlands). Carbon dioxide 4.5 (99.995%) was purchased from Westfalen (Brussels, Belgium).
In order to be compliant to QbD requirements, the SFC method was first developed at the sending lab using Design Space strategy [10]. In the previous study, a chromatographic screening to select the nature of stationary and mobile phase was performed. Then, the robust method optimization was performed by means of design of experiments and design space computation. A four factors central composite design was tested. The four selected factors were the gradient slope (modified by changing the time to change the proportion of methanol form 5 to 40%), the concentration of TFA dissolved in methanol, the isocratic time before the gradient and the temperature. Then, the retention times (at the beginning, the apex and the end of the peak) were recorded and the retention factors were modelled by an identical polynomial
A. Dispas et al. / Journal of Pharmaceutical and Biomedical Analysis 88 (2014) 519–524
521
3.2. Inter-laboratory method transfer Contrary to Quality by Testing strategy [1], the method was simply transferred and tested from the sender to the receiver, without a quantitative validation of the method by the sender. The transfer was performed on a different equipment (in the receiving lab), on two column geometries at different flow rates in order to assess the robustness of the method obtained with the DoE-DS optimization methodology and to maximize the chromatographic performances. For the following discussions, the initial method (sending lab) is called “SFC” and the transferred methods (receiving lab) are called “UHPSFC”.
Fig. 1. Three dimensions experimental domain of SFC method. Factor ‘isocratic time’ was set at 3 min. White P(S > 0; ttot < 15 min) < 0.3; Blue P(S > 0; ttot < 15 min) = 0.3–0.35; Purple P(S > 0; ttot < 15 min) = 0.35–0.4; Red: P(S > 0; ttot < 15 min) = 0.4–0.45. DS includes purple and red zones. Optimal condition is defined as red space.
equation using a multiple linear equation regression maximizing the adjusted coefficient of determination. Retention times were predicted using the developed model. The predictive distribution of new responses is then used, accounting for both the residual error and the model estimation uncertainty. The errors obtained for the predicted retention times are propagated to the separation (S). Then, Monte–Carlo simulations were used to obtain the distribution of S from the distribution of retention times of the critical peaks pair, for a given chromatographic condition. Finally, considering the distribution of retention times and S, the posterior probability for S to be higher than 0 and analysis time < 15 min (ttot ) was used to compute the DS. A probability of about 0.5 was identified. Experimentally, it led to a very satisfactory chromatographic separation. As shown in Fig. 1, the DS is the area of chromatographic conditions that will ensure the quality of the separation. In this study, the design space is the area in which a separation between peaks (S) is superior than 0 min and the time of analysis (ttot ) is inferior than 15 min. Thus, the DS is a robustness area as previously described [10,18]; indeed, the factors (e.g., gradient time, temperature, etc.) were modified in the limits of DS and the quality of separation was successfully tested. This methodology allowed defining optimal chromatographic conditions; the observed SFC chromatogram is shown in Fig. 2. Extensive discussion of these results has been previously published [10]. Furthermore, to challenge the robustness of the optimal analytical conditions obtained during the SFC method development, inter-laboratory transfer and geometric transfer are simultaneously investigated. The objective of this complex transfer was to stress the analytical procedure by modifying the lab, the equipment and the column geometry in order to demonstrate the added value of the DS optimization strategy. The separation of all compounds in the transferred method is both the demonstration of acceptability of the analytical method transfer and the further confirmation of the method robustness.
3.2.1. Geometric method transfer As the receiver’s SFC equipment is not a conventional one but a UHPSFC a geometric transfer is investigated. Three steps are critical for geometric method transfer [19]. Firstly, the selection of stationary phase chemistry with an identical selectivity will assure the efficient method transfer. For this purpose, the ideal situation is to work with a stationary phase identical in nature but different in geometry. However, the stationary phase chemistry (2-ethylpyridine) used to develop SFC method was not commercially available on sub 2 m particles. Then, it is necessary to find a column that has retention and selectivity properties as close as possible to the original one. Consequently, a BEH (ethylene bridged hybride) 2-ethylpyridine stationary phase was selected to perform the method transfer. The difference between the two stationary phases used could be a potential source of selectivity variability. Secondly, the flow rate and gradient step times have to be correctly recalculated by means of equations derived from chromatography theory [17]: FUHPSFC = FSFC ×
2 dcUHPSFC 2 dcSFC
×
dpSFC dpUHPSFC
(1)
where F is the mobile phase flow rate, dc the column internal diameter and dp the particles diameter. For the longer sub 2 m column (100 mm), the geometric transfer of flow rate was not technically possible for too high backpressure reasons. Indeed, the modifier added in the CO2 mobile phase caused a higher backpressure than conventional neat CO2 [13]. SFC system equipped with ultra high pressure pump should be used to perform separation on longer column with small particles. Anyway, for the demonstration presented in this paper, an optimized transfer gradient profile (non-optimal flow rate) was developed. For the short column, the geometric transfer at the optimal flow rate was performed without any pressure problem. The isocratic and gradient steps were recalculated according to Eqs. (2) and (3): tisoUHPSFC = tisoSFC × SGUHPSFC = SGSFC ×
V0 FSFC × UHPSFC FUHPSFC V0SFC V0SFC
V0UHPSFC
×
FUHPSFC FSFC
(2) (3)
where tiso is the isocratic time, SG is the gradient slope and V0 is the column dead volume. Thirdly, the system used to carry out the experiments has to be optimized to reduce extra-column and dwell volume. Indeed, the system void volume should be reduced to minimize band broadening. These three rules were defined for liquid chromatography method transfer. In the case of SFC method, the impact of pressure should be considered because of its impact on fluid density. Despite this potential effect, classical geometrical rules were firstly used in this work. Most likely due to the high robustness of the method, no significant change of selectivity due to fluid density variation was observed. Thus, no further investigation were carried out in the present study.
522
A. Dispas et al. / Journal of Pharmaceutical and Biomedical Analysis 88 (2014) 519–524
3.2.2. Advantages of SFC sub 2 m In order to compare chromatographic performances of original and transferred methods, peak capacity (P), best measurement of the performance of gradient elution, was calculated by Eq. (4) proposed by Neue [20]: P=
Fig. 2. Chromatograms (top to bottom) RECEIVER: UHPSFC short column transferred method at 2.5 ml min−1 ; RECEIVER: UHPSFC transferred method at 1.8 ml min−1 ; RECEIVER: UHPSFC transferred method at 1 ml min−1 ; SENDER: original SFC method. Elution order: 1, Caffeine; 2, Pseudoephedrine; 3, Paroxetine; 4, Cetirizine; 5, Histamine; 6, Dopamine; 7, Norepinephrine; 8, Serotonin
Two different approaches were investigated to perform the geometric method transfer. Firstly, a low flow rate (1 ml min−1 ) gradient was recalculated for the long column (Fig. 2). After this low pressure transfer, the flow rate was progressively increased (from 1.5 to 1.8 ml min−1 ) while keeping a backpressure lower than 400 bars. Consequently, two transfers on the same column geometry were successfully performed (at 1 and 1.8 ml min−1 ). Secondly, the method was transferred on a short column in order to drastically decrease the analysis time. The gradient profile of the original SFC and transferred UHPSFC methods are shown in Table 1. However, the pressures measured on packed column with 5 m and 1.7 m (and at different flow rates) were different, impacting directly on fluid density and hence retention and selectivity are modified. Nevertheless, the high amount of modifier used slightly decreased the influence of CO2 density on the polarity of mobile phase. This phenomenon should be further studied and considered for method transfers. Despite these changes on fluid density and then on selectivity, the separation of all peaks was kept after method transfer. This study is the first demonstration of the feasibility of SFC method transfer. The influence of pressure would be certainly studied more deeply in further work, focused on theoretical aspect of SFC geometric transfer.
tG (1/n)
n 1
(4)
ω
where P is peak capacity, tG gradient run time, n the number of peaks selected for the calculation and ω is the peak width. Peak capacity was calculated considering the seven peaks eluted during the gradient time. P was estimated at 29 for SFC method and at 40 for UHPSFC method (long column, maximal flow rate); an improvement of 35% was thus achieved by this method transfer, considering the shorter analysis time (reduced twice). Furthermore, the resolution (except for the peaks pair 5–6) was significantly improved by using UHPSFC. The SFC method transfer on short sub 2 m column was highly valuable considering analysis time and peak capacity. Indeed, this method was able to reach a peak capacity of 72 for an analysis time lower than 2 min. Of course, the resolution was lower in the case of very fast separation (UHPSFC short column). Generally, UHPSFC showed better chromatographic performances than SFC, the column length should be selected according to the requirements of the method (fastness, resolution, peak capacity, etc.). As mentioned previously [10] polar compounds are not the easiest case study for SFC separation, higher chromatographic performances should be obtained with less polar compounds. 3.2.3. Assessment of the analytical method transfer In order to compare the elution profile of the methods, relative retention times were calculated by dividing each retention time (RT) by the RT of the last compound (the most influenced by the gradient). As shown in Table 2, the predicted and observed relative RT were close, illustrating the adequacy of the transfer process. Obviously, the tiny differences observed could be easily explained by the slight changes of selectivity discussed above. To confirm these observations, the observed relative RT were modelled as a function of the predicted relative RT and of the method type (original method (SFC) or transferred method (UHPSFC at 1.0 ml min−1 , UHPSFC at 1.8 ml min−1 and UHPSFC short column)). A quadratic effect of predicted relative RT as well as an interaction term between method type and predicted relative RT were 2 = also included into the model. Fig. 3 illustrates this model (Radj 0.9882). First, the method × predicted relative RT interaction is not significant (p-value = 0.4650) as well as the method main effect (pvalue = 0.0668). This highlights that no differences between the original SFC of the sender and the transferred UHPSFC methods have been detected showing the reliability of the transfer methodology applied. In addition, the prediction of the relative RT is shown to be sufficiently accurate even if the predicted relative RT2 term is significant (p-value < 0.0001). This statistically significant effect has no practical implications. The quadratic term merely signifies that the correspondence between the RT predicted by the transfer
Table 1 Gradient profile of SFC and transferred UHPSFC methods. Time (min) Flow rate (ml min−1 )
2 ml min−1
CO2 (%)
MeOH TFA 25 mM (%)
SFC
95 95 60 60 95 95
5 5 40 40 5 5
0 3 12.2 17.2 17.7 30
1 ml min−1
1.8 ml min−1
UHPSFC long column 0 2.75 10.57 14.83 15.25 25.72
0 1.53 5.87 6.25 6.49 14.29
2.5 ml min−1 UHPSFC short column 0 0.47 2.03 2.88 2.97 5.06
Obs rel RT UHPSFC Short column
0.125 0.613 0.675 0.845 0.863 0.889 0.945 1.000
Obs RT UHPSFC Short column
0.213 1.046 1.152 1.441 1.472 1.517 1.612 1.706
523
1.32 4.8 6.4 7.5 8.56 9.71 10.46 11.04
0.12 0.435 0.579 0.679 0.775 0.879 0.948 1.000
1.549 5.064 6.17 7.569 7.784 8.916 9.519 9.732
0.159 0.52 0.634 0.778 0.8 0.916 0.978 1.000
0.884 4.239 5.359 6.526 6.859 7.488 8.012 8.379
0.106 0.506 0.64 0.779 0.819 0.894 0.956 1.000
0.465 2.223 2.873 3.537 3.758 4.099 4.398 4.609
0.101 0.482 0.623 0.767 0.815 0.889 0.954 1.000
Fig. 3. Relative observed retention time versus relative predicted retention time, quadratic model. Red line SENDER: original SFC method; green line RECEIVER: UHPSFC transferred method at 1 ml min−1 ; blue line RECEIVER: UHPSFC transferred method at 1.8 ml min−1 ; brown line RECEIVER: UHPSFC short column transferred method at 2.5 ml min−1 .
CAF PSE PAR CET HIS DOP NOR SERO
Obs rel RT UHPSFC 1.8 ml min−1 Obs RT UHPSFC 1.8 ml min−1 Obs rel RT UHPSFC 1 ml min−1 UHPSFC
Obs RT UHPSFC 1 ml min−1 Obs RT rel
SFC
Obs RT Pred RT rel
Predicted SFC
Pred RT
Compound
Table 2 (Columns 2–3) SFC predicted (pred) retention times, relative (rel) predicted retention times; (columns 4–5) SFC respective observed and relative retention times; (columns 6–9) respective observed and relative retention times for UHPSFC at 1 and 1.8 ml min−1 ; (columns 10–11) respective observed and relative retention times for UHPSFC on short column.
A. Dispas et al. / Journal of Pharmaceutical and Biomedical Analysis 88 (2014) 519–524
and the true RT are not perfectly proportional. Second, the slope and the intercept are also both significant. In case of perfect transfer prediction, one would expect the slope to be one (here it is estimated at 0.87 ± 0.05 at 95% confidence level) and the intercept to be zero (i.e., the intercept should be non-significant, here the intercept is 0.17 ± 0.03 at 95% confidence level). These two facts can be explained by many reasons, the most prominent ones being the change of stationary phase and the application of the classical geometric transfer computation for an SFC method. More research on this subject is necessary to improve the transfer prediction. These results demonstrated and confirmed the high interest of robust method optimization using a DoE-DS strategy. Indeed, the robustness was evaluated during the method development at the sending laboratory contrary to more common strategies where the robustness is evaluated after the quantitative validation step at the sending site before its transfer to the QC laboratory. To evaluate the robustness during development step is now a clear recommendation of the U.S. Pharmacopeial Convention [21]. A method transfer could be successfully achieved even after having bypassed the quantitative validation study and the conventional robustness study, at the condition that no formal quantitative validation is required at the sending laboratory (Fig. 4). Thus, if a method is developed in a laboratory and used in routine in a second laboratory, it is no longer required to validate the quantitative performances of the method before the lab transfer thanks to the robust optimization strategy. Obviously, a formal full quantitative validation of the method remains essential before performing quantitative analysis in routine. However, this one could only be considered after its transfer, i.e., at the QC laboratory were the method will effectively be used in routine analysis. Furthermore, considering the QbT methodology, the method transfer consists of a formal step involving assessment of the quantitative performances of the method. To evaluate the quantitative performances at the receiving lab, several strategies are used [1] but a partial or even more a full validation is generally performed, meaning that two validations are performed during the life cycle of the method. The present strategy (QbD) allows to speed up the life cycle of an analytical method by focusing on the real aim of the R&D laboratory and of the QC laboratory (Fig. 4). Hence, time, resources and money are saved without impairing the quality of the analytical method released to the QC site. In the context of QbD methodology, the DoEDS strategy is the critical step providing the guarantee of successful transfer of an optimal method from the sending to the receiving lab. Therefore, the quality of transfer is assessed at the beginning of analytical method life cycle during robust optimization phase.
524
A. Dispas et al. / Journal of Pharmaceutical and Biomedical Analysis 88 (2014) 519–524
Fig. 4. Analytical method Life Cycle flowchart: Quality by Testing (QbT) versus Quality by Design (QbD) methodology.
4. Conclusion The advantage of robust method optimization was definitively demonstrated in the present paper. Indeed, the high value of the DS strategy implemented for the optimization of SFC separation was successfully established in this work. In order to speed its life cycle, only a robust method optimization, including the concept of DoE and DS, was required to guarantee the success of the transfer bypassing the usual method validation and robustness study steps performed into the R&D laboratory (sender) (Fig. 4). Indeed, the chromatographic separation was not degraded by the variability induced by changing the lab, the equipment and by the geometric transfer thanks to the high reliability of the DS methodology. Furthermore, in the context of QbD methodology, the guarantee of a successful method transfer is already assessed during the robust optimization step performed at the sending lab contrary to the QbT methodology. However, it is clear that a formal quantitative validation phase will always remain mandatory before any quantitative analysis. The main benefit of the present strategy is to limit or focalize the full quantitative validation of the method in the environment in which it will effectively be used in routine analysis. Performing method validation in the QC lab environment allows to minimize the risk to obtain Out Of Specifications results. This work is also the demonstration of the feasibility of SFC geometric method transfer. Despite the selectivity changes (stationary phases and fluid density) these transfers were successful, showing the interest of SFC and UHPSFC for the future of analytical chemistry even if the study of fluid density implication in transfer calculations should be considered as future work. Finally, using sub 2 m stationary phase with SFC presents huge advantages such as the possibility to increase peak capacity and reduce analyze times. Ultra High Performance Supercritical Fluid Chromatography (UHPSFC) is clearly a very promising technique in the framework of green analytical chemistry. References [1] E. Rozet, W. Dewé, E. Ziemons, A. Bouklouze, B. Boulanger, Ph. Hubert, Methodologies for the transfer of analytical methods: a review, J. Chromatogr. B 877 (2009) 2214–2223. [2] E. Rozet, B. Mertens, W. Dewe, A. Ceccato, B. Govaerts, B. Boulanger, P. Chiap, B. Streel, J. Crommen, Ph. Hubert, The transfer of LC–UV method for the
[3] [4]
[5] [6] [7] [8]
[9]
[10]
[11] [12] [13]
[14] [15]
[16]
[17]
[18]
[19]
[20] [21]
determination of fenofibrate and fenofibric acide in Lidoses: Use total error as decision criterion, J. Pharm. Biomed. Anal. 42 (2006) 64–70. B. Dejaegher, Y. Vander Heyden, Robustness tests, LC–GC Europe 19 (2006) 418–423. The international Conference of Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use; Q8 R2 Pharmaceutical Development, http://www.ich.org/fileadmin/Public Web Site/ICH Products/ Guidelines/Quality/Q8 RI/Step4/Q8 R2 Guidelines.pdf (accessed June 2011). P. Nethercote, J. Ermer, Quality by design for analytical methods, Pharm. Technol. 36 (2012) 52–55. E. Rozet, P. Lebrun, B. Debrus, B. Boulanger, P. Hubert, Design Spaces for analytical methods, TrAC 42 (2013) 157–167. P. Borman, P. Nethercote, M. Chatfield, D. Thompson, K. Truman, The application of quality by design to analytical methods, Pharm. Technol. 31 (2007) 142–152. S. Orlandini, S. Pinzauti, S. Furlanetto, Application of quality by design to the development of analytical separation methods, Anal. Bioanal. Chem. 405 (2013) 443–450. P. Lebrun, B. Govaerts, B. Debrus, A. Ceccato, G. Caliaro, P. Hubert, B. Boulanger, Development of a new predictive modelling technique to find with confidence equivalence zone and design space of chromatographic analytical methods, Chemom. Intell. Lab. Syst. 91 (2008) 4–16. A. Dispas, P. Lebrun, P. Sassiat, E. Ziemons, D. Thiébaut, J. Vial, Ph. Hubert, Innovative green supercritical fluid chromatography for the determination of polar compounds, J. Chromatogr. A 1256 (2012) 253–326. L.T. Taylor, Supercritical fluid chromatography, Anal. Chem. 80 (2008) 4285–4294. L.T. Taylor, Supercritical fluid chromatography, Anal. Chem. 82 (2012) 4925–4935. C. Sarazin, D. Thiébaut, P. Sassiat, J. Vial, Feasibility of ultra high performance supercritical neat carbon dioxide chromatography at conventional pressures, J. Sep. Sci. 34 (2011) 2773–2778. T.A. Berger, Demonstration of high speeds with low pressure drops using 1.8 m particles in SFC, Chromatographia 72 (2010) 597–602. A. Grand-Guillaume Perrenoud, J.-L. Veuthey, D. Guillarme, Comparison of ultra-high performance supercritical fluid chromatography and ultra-high performance liquid chromatography for the analysis of pharmaceutical compounds, J. Chromatogr. A 1266 (2012) 158–167. D. Guillarme, D.T.-T. Nguyen, S. Rudaz, J.-L. Veuthey, Method transfer for fast liquid chromatography in pharmaceutical analysis: application to short column packed with small particle. Part I: Isocratic separation, Eur. J. Pharm. Biopharm. 66 (2007) 475–482. D. Guillarme, D.T.-T. Nguyen, S. Rudaz, J.-L. Veuthey, Method transfer for fast liquid chromatography in pharmaceutical analysis: Application to short column packed with small particle. Part II: Gradient experiments, Eur. J. Pharm. Biopharm. 68 (2008) 430–440. P. Lebrun, B. Boulanger, B. Debrus, Ph. Lambert, Ph. Hubert, A Bayesian design space for analytical methods based on multivariate models and predictions, J. Biopharm. Stat. (2013) (in press). B. Debrus, E. Rozet, Ph. Hubert, J.-L. Veuthey, S. Rudaz, D. Guillarme, UHPLC in Life Sciences, The Royal Society of Chemistry, Cambridge, 2012, pp. 67–103. U.D. Neue, Theory of peak capacity in gradient elution, J. Chromatogr. A 1079 (2005) 153–161. U.S. Pharmacopeial Convention, new chapter 1224, 1225, 1226, USP panel expert.
