PredicDiff™: a computational tool for the prediction of PERLs concentrations based on extractables data

Abstract

PredicDiff™, a computational modeling tool that allows fitting diffusion curves to process equipment related leachables (PERLs) is presented. Based on the measurement of extractables (analytical data), Fick’s second law of diffusion, and a Trust Region Rebounds based fitting algorithm (optimize.curve_fit from Scipy), the Python-based model fits a diffusion curve to each extractable/PERL thus allowing the determination of the PERL amount after any arbitrary contact time and temperature for example, the actual production conditions. In addition, PredicDiff™ delivers the system’s diffusion- and partition coefficients and the equilibrium concentration. Three case studies are presented: 1) interpolation of ε-caprolactam from a polysulfone disconnector from 24h to 2h, 2) adjustment of the diffusion of ε-caprolactam from a polysulfone disconnector from 40°C to 21°C, and 3) extrapolation of 2,4-Di-tert-butylphenol from an ultra-low-density polyethylene (ULDPE) bag from 70 to 90 days.

In addition, the usability of PredicDiff™ for inter- or extrapolation of an unidentified extractable from a silicone tubing is shown. In the first case, after a contact time of 2h, the concentration and hence, also patient exposure to ε-caprolactam is reduced by 70% in comparison to the extractable value after 24h. In the second case, further adjustment based on contact temperature (21°C vs. 40°C) gives a total reduction of 87%. In the third case, the concentration and therefore, also patient exposure to 2,4-Di-tert-butylphenol increases by 2.6% if storage is prolonged from 70 days to 90 days. PredicDiff™ has no limitations on the types of extractables (including those whose identities are not elucidated) or concentration ranges. Based on the remodeling of diffusion curves from literature and the calculation of extractables amounts from studies (analytical data), it is shown that PredicDiff™ provides reliable results within an acceptable range of uncertainty. Inter- and extrapolated PERLs can support the extractables and leachables (E&L) risk management by quickly calculating a more realistic concentration and ultimately, patient exposure.

Introduction

Polymeric materials are widely used in industrial manufacturing. In (bio)pharmaceutical production, polymers are predominantly employed as single-use (filters, bags, tubing, connectors) and in some cases, they are reused multiple times (for example, hoses). One disadvantage of polymeric materials is leaching of chemical impurities. In pharma, those polymer-borne chemical impurities are called extractables / leachables and more specifically in production, PERLs (process equipment related leachables). PERLs may pose a risk to the product quality and to patient safety. As a result, the risk of PERLs must be assessed and managed (21 CFR 600.11; USP <665>; USP <1665>; EU Annex 1). Therefore, depending on the risk level, the polymeric materials are extracted based on defined protocols (USP <665>; Tumambac et al., 2017). Extraction is typically performed in simple solvents such as ethanol/water 50:50 and both acidic (pH 3) and alkaline (pH 10) water, respectively, at 40°C and for different periods (30min, 24h, 7d, 21d). Orthogonal analytical methods including gas- and liquid chromatographic separation, coupled with detectors, and mass spectrometry are used to identify and semi-quantify the PERLs. Extraction is designed to be exaggerated in comparison to actual production conditions, so that no potential leachables go undetected. In the next step of the risk management process, the extractables data is converted into patient exposure and toxicologically assessed.

Exaggerated extractables data may lead to hypothetical patient exposures that exceed the permitted toxicological threshold values. In such cases, the presence and concentration of the target extractables can be verified in the drug product. Such analysis can be cumbersome as a typical E&L assessment of a manufacturing line encompasses hundreds of different extractables including compounds whose identities cannot be fully elucidated. Another possibility is a simulated-use extraction study under consideration of the actual production conditions. However, any additional study has a notable impact on time and costs. At the same time, it appears logical to assume that the amount of an extractable after contact for 2h at 20°C must be lower in comparison to extraction for 24h at 40°C. Due to the underlying diffusion-controlled mechanism, simple linear- or any other random interpolation is out of question. Therefore, the authors have developed a Python-based model (PredicDiff™) for inter- and extrapolation of extractables data based on time and temperature, under consideration of Fickian diffusion.

The mathematical and physical groundwork on Fickian diffusion of chemical compounds has been laid decades ago, resulting in applications in geology (for example, leaching from landfills) and at a later stage, also in food industry, to verify leachables from food packaging. A comprehensive listing of all the published applications of diffusion-based migration modeling is beyond the scope of this work and it would require a separate review article however, in the context of extractables and leachables in the food industry, the development of the migration model in the scope of the FOODMIGROSURE project, the FACET (Flavours, Additivers, and food Contact materials Exposure Task) exposure tool and the MERLIN-Expo tool are particularly noteworthy (Franz and Simoneau, 2008; Oldring et al., 2014; Ciffroy et al., 2022). The Joint Research Centre (JRC) of the European Commission issued a guidance document for users of diffusion-based migration modelling from plastic food contact materials (Hoekstra et al., 2015). The FDA has recognized migration modeling for preparation of premarket submissions for Food Contact Substances (FDA 2007).

Commercial software has been created, most notably SML (Specific Migration Limits Software) and Migratest for multilayer polymers (AKTS SML; Mercea et al., 2018). In their book, (Piringer and Baner 2008) summarize comprehensive information that serves as a basis for modeling diffusion in polymers, including the fundamental mathematical work by Crank (Crank, 1975) and a review of models for diffusion in polymers provided by (Mercea 2008). Diffusion-based migration models for estimation of patient exposure to leachables from polymeric medical devices have also been developed (Morrison et al., 2018; Turner et al., 2020; Saylor and Young, 2023), resulting in open source, Python-based modeling tools (CHRIS) to predict patient exposure to colorants and bulk chemicals from medical devices (FDA CHRIS).

The prerequisite for using diffusion-based migration models is knowledge of the concentration of the compound in the polymer in addition to the compound-, polymer- and solvent specific diffusion- and partition coefficients. Alternatively, if certain polymer-specific coefficients are available, the diffusion- and partition coefficients can be modeled (Brandsch et al., 2002; Gillet et al., 2010; Loschen and Klamt, 2014; Ciffroy et al., 2022; Egert and Langowski, 2022a and b; Elder and Saylor, 2023; Li, 2024; Elder et al., 2024) or approximated based on worst-case assumptions. The CHRIS tools require knowledge about the chemical’s identity and its initial concentration in the polymer. Sartorius was the first to establish an internal extractables modeling tool called ExSim, supporting customers in estimating the PERLs concentrations under their production conditions (Hauk et al., 2021). The latter is not publicly available. The development of PredicDiff™ is based on the same mathematical and physical principles as the existing models. The novelty is the application, i.e. what the model tries to answer “how can I inter- and/or extrapolate extractables amounts to fit my actual production conditions?”.

The aim of this work is presenting the PredicDiff™ model including use cases to demonstrate its usefulness for a rapid, cost-effective, and realistic estimation of PERLs concentrations and subsequently, patient exposure to PERLs.

Download the full article as PDF here: PredicDiff™: a computational tool for the prediction of PERLs concentrations based on extractables data

or read it here

Nicole Heider, Alicja Sobańtka, PredicDiff™: a computational tool for the prediction of PERLs concentrations based on extractables data, European Journal of Pharmaceutical Sciences, 2025, 107108, ISSN 0928-0987, https://doi.org/10.1016/j.ejps.2025.107108.

Visit our Excipient Landscape 2024 here:

Get more information on different excipients and their manufacturers