Reply to âComment on 'In Silico Modeling of ... - ACS Publications

Subscriber access provided by UB + Fachbibliothek Chemie | (FU-Bibliothekssystem)

Communication

Reply to “Commentary on "In silico modeling of gastrointestinal drug absorption: Predictive performance of three physiologically based absorption models"” Erik Sjögren, Helena Thörn, and Christer Tannergren Mol. Pharmaceutics, Just Accepted Manuscript • DOI: 10.1021/ acs.molpharmaceut.6b00775 • Publication Date (Web): 17 Nov 2016 Downloaded from http://pubs.acs.org on November 20, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Molecular Pharmaceutics is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Molecular Pharmaceutics

Reply to “Commentary on "In silico modeling of gastrointestinal drug absorption: Predictive performance of three physiologically based absorption models"” Erik Sjögrena, Helena Thörnb, Christer Tannergrenb

a

Department of Pharmacy, Uppsala University, BOX 580, S-751 23 Uppsala, Sweden

b

Pharmaceutical Technology and Development, AstraZeneca R&D Gothenburg, Sweden

Running title: Modeling of gastrointestinal drug absorption in humans

Corresponding author: Christer Tannergren, PhD. Pharmaceutical Technology and Development AstraZeneca R&D Gothenburg Pepparedsleden 1, SE-43183 Mölndal Sweden Telephone : +46 31 7761976 Fax: +46 31 7763700 E-mail: [email protected]

ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Abstract: This is a reply to the commentary on "In silico modeling of gastrointestinal drug absorption: Predictive performance of three physiologically based absorption models" by Turner and other Simcyp associates. In the reply we address the major concerns raised by Turner et al. regarding the methodology to compare the predictive performance of the different absorption models and at the same time ensure that the systemic pharmacokinetic input were exactly the same for the different models; the selection of the human effective permeability value of fexofenadine; the adoption of model default values and settings; and how supersaturation/precipitation was handled. In addition, we also further discuss aspects related to differences in in silico models and the potential implications of such differences. Our original report should be viewed as the starting point in a thorough and transparent review of absorption prediction models with the overall aim of improving their application as validated tools for bridging studies of active pharmaceutical ingredients from various sources and origins in a regulatory context. With this reply we encourage other independent investigators to perform further model evaluations of commercial as well as other existing or recently implemented models. This will boost the overall progression of physiologically based biopharmaceutic models for predicting and simulating intestinal drug absorption both in research and development as well as in a regulatory context.


Page 2 of 10

Page 3 of 10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


We thank the authors, Turner and other Simcyp associates, for the well-articulated, respectful and ample commentary 1 to our study 2 on the comparison of three in silico models for the prediction of human intestinal drug absorption. Turner et al. acknowledge that the study by Sjögren et al. in 2016 2 is the first of its kind and consequently of great importance and interest for the scientific field of biopharmaceutics. The report should be viewed as the starting point in a thorough and transparent review of absorption prediction models, moving towards the overall aim of improving their application as validated tools for bridging studies of active pharmaceutical ingredients (API) from various sources and origins in a regulatory context. Turner el al. agree with several aspects of our report, such as the complexities of supersaturation/precipitation, the lack of knowledge of colonic absorption and the need for further model evaluations and comparisons. Turner et al. also offer a number of criticisms, their main concern being our conclusion that the applied versions of GI-Sim (4.1) and GastroPlusTM (8.0) perform better than Simcyp (13.1) in predicting the intestinal absorption of incompletely absorbed drugs. As Turner et al. acknowledge, this conclusion was based on the lower degree of accuracy of the data predicted by Simcyp (13.1) in relation to the observed clinical data, in combination with the clear trend towards decreased accuracy of the predictions, with lower predicted fraction absorbed (fabs), in comparison to the other models. In our report we elaborated in detail on the model settings, the input data, the adopted approach, the rationale and the results and we will therefore not extensively discuss most of these points in this reply. However, we believe some further discussion and clarifications are warranted with respect to certain aspects of the commentary by Turner et al. First, we would like to re-emphasize that the focus of the study by Sjögren et al. in 2016 2 was the prediction of intestinal drug absorption, based on factors such as solubility, dissolution, permeability, dose and formulation type, and comparison of the outputs from all three models. It is unfortunate therefore that, in their commentary, Turner et al. put emphasis on pharmacokinetics and processes related to systemic disposition, which predominately affect a drug after the absorption process. In our view, this is therefore not a valid criticism of the processes under discussion in the report by Sjögren et al. in 2016. 2 It is also unfortunate that Turner et al. focus very little of their discussion on the results



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

obtained with GastroPlusTM (8.0), as GastroPlusTM is publically available and able to be used in thirdparty comparisons. In addition, one of the main intestinal drug absorption-related aspects/concerns discussed by Turner et al. is precipitation, and the description of precipitation rate is apparently the same in GastroPlusTM (8.0) and Simcyp (13.1). 3, 4 Instead, Turner et al. focus almost entirely on GISim. One of the main concerns that Turner et al.1 have with the study by Sjögren et al.2 is the convolution approach, which was adopted to harmonize the transformation of the output, i.e. drug absorption over time, from the respective absorption model to plasma concentration-time profiles. Briefly, Sjögren et al. used the absorption-time profiles generated directly by each of the three software together with convolution using exactly the same systemic pharmacokinetic parameters for all three models to simulate plasma concentration-time profiles. 2 This approach ensured that any observed differences in the software prediction of AUC and Cmax were a direct function of differences in gastrointestinal drug absorption and not of differences in the systemic pharmacokinetic algorithms. The central point of justification for the convolution approach was hence to exclude any differences in software disposition algorithms to facilitate comparison of the capacity of the models to predict gastrointestinal absorption, based on plasma drug exposure. The numerical convolution of absorption-time profiles into plasma drug concentration-time profiles is a well-established method that is applied in many areas, such as in vitro in vivo correlations of modified-release formulations.5 To clarify, GI-Sim does include the ability to predict plasma drug concentration profiles and associated metrics without need of further manipulation. As pointed out by both Sjögren et al. 2 and Turner et al. 1, the convolution approach is associated with certain drawbacks, e.g. saturation of first-pass metabolism is mechanistically not accounted for, when describing drug disposition after oral administration. However, from the perspective of assessing intestinal absorption, these drawbacks should affect the results of the different models similarly. Moreover, in our opinion, fabs is a finite measure which cannot exceed 100%, and we do not agree with the viewpoint by Turner et al. that “… the drug undergoes repeated cycles of EHR [entero-hepatic recirculation] thus increasing fraction absorbed …”. 1 Furthermore, the complexities of systemic disposition were acknowledged in Sjögren et al. 2016 2 in that the main conclusions were


Page 4 of 10

Page 5 of 10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


based on the predicted AUC and fabs rather than on Cmax and tmax. It should also be noted that the trend for overprediction of short tmax values shown in Sjögren et al. 2016 2 is not readily explained by misspecification of systemic disposition or entero-hepatic recirculation. Turner et al. also point out that the subjects participating in most of the intravenous studies were different from those participating in the oral studies. From a pharmacokinetic understanding and modeling & simulation point of view it would naturally be ideal to base any simulation evaluation study on clinical reference data where the oral dose was administered concomitantly with an intravenous dose of radiolabeled drug, but unfortunately this kind of data is largely unavailable. Finally on this topic, we would like to highlight the importance of always relating the predicted plasma exposure to the corresponding predicted fabs. This provides an assessment of both how realistic the absorption predictions are and the appropriateness of the adopted systemic disposition parameters. Turner et al. also express concerns around the selection of a human effective permeability (Peff) value of 0.07 instead of 0.12 for fexofenadine. 6, 7 Having reported these values ourselves, we are aware of the variability in Peff measurements in the literature. However, we believe that this exemplifies one of the advantages of the bottom-up modeling approach applied in Sjögren et al. 2016 2, namely that the input parameters are selected based on the researcher's best knowledge and experience before any simulation is performed and should only be changed if there is a clear data-driven rationale for it. In the study by Sjögren et al. 2016, all three models underpredicted the plasma exposure of fexofenadine. 2

Applying a higher Peff value for a BCS class III drug such as fexofenadine would increase the

predicted fabs and presumably increase the accuracy of all three models, not just of Simcyp. Turner et al. also comment that possible entero-hepatic recirculation of fexofenadine was not accounted for. However, in the work they refer to themselves it was concluded that there was no evidence for enterohepatic recirculation of fexofenadine. 8 The next major concern Turner et al. express in their commentary on the report by Sjögren et al. 2016 2

is the adoption of model default values and settings. The main rationale for this was that it was

assumed that each model was released for distribution with the optimal settings as default. As a consequence, it was anticipated that any changes made to the default settings/input would decrease the



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

overall model performance and that such changes would have been far more likely to be questioned by the model's developers than using the default settings. Indeed, the recently published EMA Draft Guideline on the qualification and reporting of physiologically based pharmacokinetic (PBPK) modelling and simulation states that “Any modification of the default values of the system-dependent parameters supplied in a commercial PBPK platform should be justified…”.9 However, the commentary from Turner et al. indicates that Simcyp (13.1) is not distributed with optimal overall settings/input as the default. In our experience there is significant value in applying robust, validated default parameters as this can minimize errors caused by variability in the experience and knowledge of researchers using the software. One of the few absorption-related processes that Turner et al. highlight in their commentary is the challenging subject of supersaturation and precipitation. Like Turner et al., we acknowledge that these processes have the potential to highly influence the simulations and that particular consideration is needed for certain compounds. However, the focus of the study by Sjögren et al. in 2016 2 was to compare differences in output between the included models. From this perspective it is of particular interest to view the observed dissimilarities in the results obtained with Simcyp (13.1) and GastroPlusTM (8.0) as these two models adopt an apparently comparable precipitation approach (firstorder process) with equal rates of precipitation as the default (900 s or 0.25 h). Now, as a consequence of the differences in luminal volumes and how the level of supersaturation is handled between Simcyp (13.1) and GastroPlusTM (8.0), differences in precipitation behavior will occur even when the same input values for precipitation are used. This is very important information for a researcher, as this will not only be the case for default values but also when measurements are available. The simulation study using Simcyp (14.1) on ketoconazole and posaconazole by Cristofoletti et al. 2016 10, which is referred to by Turner et al., is an illustrative example on this subject. In that study, Cristofoletti et al. reported that the experimental precipitation data could not be successfully used in the Simcyp simulations to accurately replicate observed plasma concentrations. 10 To accomplish a good correlation between the observations and the Simcyp simulations, the precipitation rate had to be adjusted to “very fast” for posaconazole and “negligible“ for ketoconazole. This raises a number of


Page 6 of 10

Page 7 of 10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


central questions. Would the study have come to the same conclusion if another absorption model, e.g. GastroPlusTM, had been used? How should the results be interpreted – and used – if two models were applied and differences in results between the models were acquired? What can be learned from this study for forthcoming simulations/predictions related to compounds for which no clinical reference data are available or for simulations that exclusively have to rely upon in vitro measurements? These are highly relevant questions that model users in the pharmaceutical development field regularly face in their day-to-day practice. Furthermore, as a result of the smaller baseline luminal volumes and the dynamic volume model adopted in Simcyp (13.1), supersaturation/precipitation needs to be considered for a larger number of compounds than required for a model which adopts larger luminal water volumes e.g. GastroPlusTM (8.0). Therefore, one would hope that these processes have been given extra attention in the Simcyp model development evaluations, resulting in a well-grounded selection of default parameters/settings. For example, Turner et al. suggest in their commentary that one of the reasons for the underprediction by Simcyp (13.1) of fexofenadine absorption was that unlimited supersaturation was not allowed for in the simulations. 1 Even though unlimited, stable, intraluminal fexofenadine supersaturation in vivo cannot be dismissed, as intra luminal behavior of fexofenadine has not yet been investigated, it is not the traditional viewpoint with respect to the incomplete absorption of fexofenadine. However, if our report contributes to a future (recommended) default setting in Simcyp (13.1) indicating “unlimited supersaturation” we are glad that our evaluation will be of practical use. Nevertheless, we repeat our recommendation that researchers should perform a systematic evaluation of available models using their selected input data and make their own decisions with respect to best practice. Finally, there are numerous details pertaining to the supersaturation/precipitation process that remain incompletely described by the models; for example the full nucleation process, a description of the particle size of the precipitate, and how/if supersaturation/precipitation is described for compounds that partition into colloidal structures, when the amounts of bile salts decrease.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The final main concern expressed by Turner et al. that we would like to address is the use of nonphysiological system parameter values and models. We agree with Turner et al. that physiological relevance is a key aspect and that it is crucially important that researchers are able to reliably relate data from in silico simulations to the in vivo situation. We also believe that physicochemical and biopharmaceutical relevance are of equal importance to physiological relevance. The goal should be that the model is as correct as possible regarding physiology, physicochemistry and biopharmacy, with preserved prediction accuracy. It is generally acknowledged that there are still too many gaps in our knowledge and too few direct measurements to allow development of a completely physiologically correct model. Every model is also inevitably bound to make simplifications and assumptions of the true physiology at one stage or another. For instance, due to the lack of understanding of colonic drug behavior and associated high risk for inaccuracy in predictions, GI-Sim recommends not to include colonic drug absorption while no such recommendation is given for Simcyp and GastroPlusTM. The presumed misspecification of the larger luminal water volumes in GastroPlusTM (8.0) and GI-Sim (4.1) is one aspect of concern raised by Turner et al. and we agree that reported magnetic resonance imaging and positron emission tomography measurements indicate that the free water volume is less than specified in these two models. 11-13 However, it is likely that the total fluid volume in which the drug can be dissolved is greater than that given by the reported free intraluminal water volumes located in water pockets. For instance, theoretically, a 0.5 mm thick fluid layer lining the intestinal wall would result in a total volume of approximately 100 ml. When intraluminal content, free mucus and adherent mucus layers are considered, this is not unrealistic, although it has yet to be measured. One could also speculate that unabsorbable formulation components could influence the local water balance by increasing luminal osmolarity and resultant water secretion. Furthermore, there is also the general issue of how experimental measurements should be “translated” by the model in order to be representative of the in vivo situation. Finally, the decision on how to describe the physiological model and adopt input data must be made for numerous processes and, as pointed out by Turner et al., there may be additional as-yet uncharacterized or unassessed processes with potential implications for drug absorption in the intestine.


Page 8 of 10

Page 9 of 10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


To summarize, all of the predictive models included in Sjögren et al. 2016 2 have been proven to be very useful and are now established as key tools in drug research and development. However, there is still an opportunity for further improving the use and development of these important models. Given the current uncertainties and limitations of the models described above, it is important in our view that each model is evaluated on how its physiological and biopharmaceutic processes affect the simulations/predictions and how closely this corresponds to clinical observations. We agree with Turner et al. that “… with the given set of API, input parameters and model options (default or otherwise) Simcyp provides the more conservative predictions of exposure of the three platforms.”. We encourage other independent investigators to perform further model evaluations, not only using Simcyp and GastroPlusTM but also using any other existing or recently implemented models. This will boost the overall progression of physiologically based biopharmaceutic models for predicting and simulating intestinal drug absorption both in research and development and in a regulatory context. Finally, as the study under discussion 2 was just one study in an area that has not yet been fully explored, the authors welcome future studies that confirm or challenge the conclusions presented in this report.

Declaration of interest HT and CT are employees of AstraZeneca. GI-Sim has been developed by AstraZeneca for internal use. AstraZeneca has ongoing license agreements for both Simcyp and GastroPlusTM.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

References: 1. Turner, D. B.; Liu, B.; Patel, N.; Pathak, S. M.; Polak, S.; Jamei, M.; Dressman, J. B.; Rostami-Hodjegan, A. Commentary on "In silico modeling of gastrointestinal drug absorption: Predictive performance of three physiologically based absorption models". Mol Pharm 2016. 2. Sjögren, E.; Thörn, H.; Tannergren, C. In Silico Modeling of Gastrointestinal Drug Absorption: Predictive Performance of Three Physiologically Based Absorption Models. Mol Pharm 2016, 13, (6), 1763-78. 3. Agoram, B.; Woltosz, W. S.; Bolger, M. B. Predicting the impact of physiological and biochemical processes on oral drug bioavailability. Adv Drug Deliv Rev 2001, 50 Suppl 1, S41-67. 4. Jamei, M.; Turner, D.; Yang, J.; Neuhoff, S.; Polak, S.; Rostami-Hodjegan, A.; Tucker, G. Population-based mechanistic prediction of oral drug absorption. AAPS J 2009, 11, (2), 22537. 5. U.S. Food and Drug Administration, Center for Drug Evaluation and Research (CDER) Guidance for Industry: Extended release oral dosage forms: Development, evaluation, and application of in vitro/in vivo correlations. 1997. 6. Tannergren, C.; Knutson, T.; Knutson, L.; Lennernas, H. The effect of ketoconazole on the in vivo intestinal permeability of fexofenadine using a regional perfusion technique. Br J Clin Pharmacol 2003, 55, (2), 182-90. 7. Tannergren, C.; Petri, N.; Knutson, L.; Hedeland, M.; Bondesson, U.; Lennernas, H. Multiple transport mechanisms involved in the intestinal absorption and first-pass extraction of fexofenadine. Clin Pharmacol Ther 2003, 74, (5), 423-36. 8. Lappin, G.; Shishikura, Y.; Jochemsen, R.; Weaver, R. J.; Gesson, C.; Houston, B.; Oosterhuis, B.; Bjerrum, O. J.; Rowland, M.; Garner, C. Pharmacokinetics of fexofenadine: evaluation of a microdose and assessment of absolute oral bioavailability. Eur J Pharm Sci 2010, 40, (2), 125-31. 9. European Medicines Agency, Committee for Medicinal Products for Human Use (CHMP) Draft guideline on the qualification and reporting of physiologically based pharmacokinetic (PBPK) modelling and simulation. 2016. 10. Cristofoletti, R.; Patel, N.; Dressman, J. B. Differences in Food Effects for 2 Weak Bases With Similar BCS Drug-Related Properties: What Is Happening in the Intestinal Lumen? J Pharm Sci 2016, 105, (9), 2712-22. 11. Mudie, D. M.; Murray, K.; Hoad, C. L.; Pritchard, S. E.; Garnett, M. C.; Amidon, G. L.; Gowland, P. A.; Spiller, R. C.; Amidon, G. E.; Marciani, L. Quantification of gastrointestinal liquid volumes and distribution following a 240 mL dose of water in the fasted state. Mol Pharm 2014, 11, (9), 3039-47. 12. Schiller, C.; Frohlich, C. P.; Giessmann, T.; Siegmund, W.; Monnikes, H.; Hosten, N.; Weitschies, W. Intestinal fluid volumes and transit of dosage forms as assessed by magnetic resonance imaging. Aliment Pharmacol Ther 2005, 22, (10), 971-9. 13. Shingaki, T.; Takashima, T.; Wada, Y.; Tanaka, M.; Kataoka, M.; Ishii, A.; Shigihara, Y.; Sugiyama, Y.; Yamashita, S.; Watanabe, Y. Imaging of gastrointestinal absorption and biodistribution of an orally administered probe using positron emission tomography in humans. Clin Pharmacol Ther 2012, 91, (4), 653-9.


Page 10 of 10

Reply to âComment on 'In Silico Modeling of ... - ACS Publications

Recommend Documents

Reply to âComment on 'In Silico Modeling of ... - ACS Publications