Simultaneous automated image analysis and Raman spectroscopy of powders at an individual particle level
Graphical abstract
Introduction
Polymorphism of drug compounds involves the occurrence of different types of packing of the same molecule in a crystal lattice [1]. The expression “solid form” can be used as a broader term to describe not only crystalline single component systems, but also amorphous matter and binary systems, such as salts, solvates, cocrystalline and coamorphous systems [2]. Different solid forms of a drug compound can be critical for the health outcome of a patient because they may affect product performance, especially when exhibiting a different particle size, shape, solubility, dissolution rate and bioavailability [3,4]. These critical material attributes can thus affect product quality, safety and efficacy [5]. Unexpected solid form changes, such as metastable polymorphs [6], elusive crystal forms [7] and unintentional seeding caused by very low amounts of an unwanted polymorph [8], have had negative effects on the availability of otherwise affordable drugs. An example of a drug product with a detrimental uncontrolled solid form change was Norvir® (ritonavir), leaving AIDS patients temporarily without a treatment [8]. To ensure that the polymorphism landscape is properly explored by the industry, the regulatory authorities have issued guidance documents on solid form characterization and control [9]. For reasons such as the ones mentioned above, the solid form diversity of particulate matter, both in solid, semi-solid and liquid products, is of particular interest to the pharmaceutical industry when aiming for a more detailed product and process understanding. Solid form screening has become an industrial practice to cope with these challenges and in this context, different methods for generating the maximal number of new forms as well as high throughput analytical methods for quantification and detection have been developed [10].
Solid form diversity can be also a critical part of the patent portfolio of a given drug compound and its respective products. A number of litigation cases involving solid form issues have affected patent validity. For instance, the presence of forms I and II of ranitidine hydrochloride in anti-ulcer drug products led to litigation between Glaxo and Novopharm [8]. In many cases, such as in a case between Calgene and Dr Reddy's around the anti-cancer drug product Revlimid (lenalidomide), the key question is related to low amounts of a polymorphic impurity in the drug product [11]. It should be pointed out that many of these litigation cases have initiated intensive analytics in searching for very low amounts of a given solid form [8].
Conventional approaches to detect and quantify low levels of solid form impurities are based on the analysis of bulk materials [12]. Here we report an analytical technique at an individual particle level for the detection and quantification of solid form diversity. Crystals tend to grow in a specific crystallographic direction [13] and different polymorphs and crystallization/processing conditions typically result in different crystal morphologies [14]. Quantitative assessment of particle morphology has evolved with technological advancements in instrumentation and computing power. Methods based on automated particle tracking, bright/dark-field imaging and image analysis (IA) are capable of particle size and shape analysis of even hundreds of thousands of particles within a reasonable timeframe [15]. Particle morphology could therefore be used for solid form assessment, potentially resulting in a fast and sensitive analytical technique.
It has been estimated that about 90 % of studies related to polymorphism use at least two solid-state analytical techniques [16]. The use of a combination of characterization techniques enables scientific insight into the complexity of these phenomena with a higher accuracy [17]. Our study evaluates Raman spectroscopy, IA and the combination of Raman spectroscopy and IA at a single particle level in order to detect and quantify a low amount of a solid form impurity of a crystalline drug material. Partial least squares-discriminant analysis (PLS-DA) is a well-established data analytical method that combines dimensionality reduction and high prediction capability. PLS-DA is commonly used for variable selection as well as predictive and descriptive classification modeling. The PLS-DA algorithm is applicable for analyzing high dimensional data and does not assume the data to fit any distribution, making it suitable for imbalanced data with a high number of variables (n>1000) [18]. By using crystalline carbamazepine (CBZ) anhydrate (AH) and dihydrate (DH) as model solid forms, we aim to quantify solid forms in these binary mixtures, ultimately even at a single particle level. Quantification based on IA, Raman spectroscopy or a combination thereof is compared and a strategy is proposed for detection of low amounts of an unwanted solid form.
Section snippets
Materials
Carbamazepine (CAS 298−46 - 4) was purchased from three different commercial sources: Tokyo Chemical Industry, Co., LTD (Tokyo, Japan), Hawkins, Inc. (Minneapolis, MN, USA) and Carbosynth (Berkshire, UK). Methanol 99.8 % (67−65-1) was purchased from Sigma Aldrich Co. (St. Louis, Missouri, MO, USA). All chemicals used were of analytical reagent grade or higher. Highly purified water (Milli-Q, Millipore Inc., Denver, Massachusetts, USA) was used in all of the studies. Hydrophilic PTFE filters,
Raman model
The Raman PLS-DA model was optimal with three LVs that cumulatively used 76.9 % of the variation in the data to classify CBZ AH and CBZ DH (Table S2, Supporting Information). Separation of CBZ AH and CBZ DH classes is visualized with the scores plot of the three LVs (Figure S4, Supporting Information). The loadings plot (Figure S5, Supporting Information) indicates that the Raman wavenumber range between 1400 cm−1 and 1600 cm−1 is the most significant for classification, in agreement with the
Conclusion
In this study three different classification models were developed for solid form characterization at a single particle level. The IA model was the best for fast screening and comparison of different samples according to their hydrate content. The IA of at least 200 particles was sufficient for a classification error of less than 5 %. This can be relevant for a number of applications where time and cost of analysis are the highest priority. The Raman model could assess solid form composition
CRediT authorship contribution statement
Andrea Sekulovic: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing - original draft, Writing - review & editing, Visualization, Supervision, Project administration, Funding acquisition. Ruud Verrijk: Conceptualization, Resources, Writing - review & editing, Project administration, Funding acquisition. Thomas Rades: Conceptualization, Writing - review & editing. Adam Grabarek: Investigation, Writing - review & editing. Wim
Declaration of Competing Interest
Dr Reddy’s IPDO Leiden (Leiden, Netherlands) has financed the PhD project of Andrea Sekulovic. Andrea Sekulovic and Ruud Verrijk are employed at Dr Reddy’s. Jukka Rantanen and Thomas Rades have not received any consulting fees for this work.
References (30)
- et al.
Variability in commercial carbamazepine samples - impact on drug release
Int. J. Pharm.
(2011) - et al.
Role of nanoparticle size, shape and surface chemistry in oral drug delivery
J. Control. Release
(2016) - et al.
Comparison of macro and micro Raman measurement for reliable quantitative analysis of pharmaceutical polymorphs
J. Pharm. Biomed. Anal.
(2018) - et al.
An overview of recent studies on the analysis of pharmaceutical polymorphs
J. Pharm. Biomed. Anal.
(2011) - et al.
Assessing feature relevance in NPLS models by VIP
Chemometr. Intell. Lab. Syst.
(2013) - et al.
Qualitative in situ analysis of multiple solid-state forms using spectroscopy and partial least squares discriminant modeling
J. Pharm. Sci.
(2007) - et al.
Using terahertz pulsed spectroscopy to quantify pharmaceutical polymorphism and crystallinity
J. Pharm. Sci.
(2005) - et al.
Assessing the detection limit of a minority solid-state form of a pharmaceutical by H double-quantum magic-angle spinning nuclear magnetic resonance spectroscopy
J. Pharm. Sci.
(2017) - et al.
Application of image-based particle size and shape characterization systems in the development of small molecule pharmaceuticals
J. Pharm. Sci.
(2015) - et al.
Comparison of the four anhydrous polymorphs of carbamazepine and the crystal structure of form I
J. Pharm. Sci.
(2003)
Process variables of instant granulator and spheroniser: I. Physical properties of granules, extrudate and pellets
Int. J. Pharm. (Amsterdam, Neth.)
A platform for preparing homogeneous proteinaceous subvisible particles with distinct morphologies
J. Pharm. Sci.
Deep convolutional neural network analysis of flow imaging microscopy data to classify subvisible particles in protein formulations
J. Pharm. Sci.
Theory and origin of polymorphism
Polymorphism in Pharmaceutical Solids
Investigation of the solid forms of deferasirox: solvate, co-crystal, and amorphous form
RSC Adv.
Cited by (4)
Solid-state analysis for pharmaceuticals: Pathways to feasible and meaningful analysis
2023, Journal of Pharmaceutical and Biomedical AnalysisMulti-sensors data fusion for monitoring of powdered and granule products: Current status and future perspectives
2023, Advanced Powder TechnologyA top-down spectroscopic approach for correlating coating thickness distributions with the dissolution profiles of enterically coated pellets
2023, Journal of Pharmaceutical and Biomedical AnalysisCitation Excerpt :The variance in the scores seen within each distribution can likely be attributed to pellet to pellet coating uniformity. While the spectra acquisition in this work was performed manually, future work incorporating machine learning assisted identification of particles or other small features and particle imaging with Raman spectroscopy [34,35] could pave the way for development of integrated and automated online/at-line analytical tools that can be used in advanced manufacturing of pharmaceuticals. In this study, a top-down approach using Raman microscopy and multivariate data analysis was taken to characterize the coating properties of small, functionally-coated pellets from five 300 mg diltiazem HCl ER capsules with different manufacturers (Products 1–5).
Topologically directed confocal Raman imaging (TD-CRI): Advanced Raman imaging towards compositional and micromeritic profiling of a commercial tablet components
2022, Journal of Pharmaceutical and Biomedical AnalysisCitation Excerpt :Raman imaging has been utilized by a community of researchers for a variety of purposes. To name some, this includes determining physical changes (crystalline and amorphous state) [9], polymorphism [12], size analysis of API in tablets [3,7,13,14], quantification (0.044% w/w and 0.8% w/w) of low dosage drugs [15], compositional distribution of pharmaceutical ingredients [3,5,16], physical morphology like pore network and polymer rearrangements [17] etc. Concisely, Raman imaging may be regarded a superior technique for the extraction of relevant information about API and excipients, manufacturing processes as well as for reverse engineering of commercial product [3,11].