Modelling skin permeability with micellar liquid chromatography Laura J. Waters a,⇑ , Yasser Shahzad a , John Stephenson b a School of Applied Sciences, University of Huddersfield, Queensgate, Huddersfield HD1 3DH, UK b School of Human and Health Sciences, University of Huddersfield, Queensgate, Huddersfield HD1 3DH, UK article info Article history: Received 24 May 2013 Received in revised form 16 July 2013 Accepted 2 August 2013 Available online 13 August 2013 Keywords: Permeability Chromatography Modelling Micellar Partition coefficient abstract This study evaluates the potential application of micellar liquid chromatography (MLC) to predict skin permeation with a series of model compounds. MLC has previously been found to be useful in the pre- diction of partition coefficient values (log P) for pharmaceutical compounds, yet has not been incorpo- rated in skin permeability models prior to this work. This article provides statistically supported data that this technique enhances the ability to predict the permeability of similar drugs through the skin (K p ). The replacement of a traditional physicochemical parameter, namely the octanol–water partition coefficient (log P ow ) with a chromatographically determined value (log P mw ), results in a quantitative par- tition–permeability relationship that is robust to variation. MLC offers many benefits compared with the traditional techniques employed to obtain log P values. Ó 2013 Elsevier B.V. All rights reserved. 1. Introduction Drug permeation through human skin is an increasingly impor- tant delivery route, often described by the associated permeability coefficient (K p ). It is widely known that discrepancies often occur between the in vivo reality (where the stratum corneum controls the rate of intercellular passive diffusion) and in vitro based models (Walters, 2002). Moreover, even models based on human in vivo data can experience orders of magnitude of discrepancy from observed data (Farahmand and Maibach, 2009) highlighting the complexity of such phenomena. A significant body of research exists assessing a variety of model based systems as their potential benefits are recognised widely. These models are often described as quantitative structure–permeability relationships (QSPRs) (Moss et al., 2002) with their main shortfall arising from the incon- sistency in the values used to derive them. For example, the heter- ogeneity of the data, in terms of skin origin and experimental conditions, has been shown to contribute to the residual variance in existing models (Geinoz et al., 2004). In addition, drug perme- ation through human skin is significantly influenced by the degree and nature of charge on the drug molecule which is hard to accom- modate within the majority of proposed models (Zhang et al., 2012) and often studies are restricted to one specific series of com- pounds (Hadgraft et al., 2000). This often prevents proposed models from being more widely applicable as they are only valid for a designated group of drugs, rather than all drugs, which would be far more useful for analytical purposes. One particular challenge is attempting to retain the simplicity of a model so that it is widely used within the scientific community yet maintain the validity to explain complex real-world data (Mitragotri et al., 2011); for example, retaining the predictive ability of the model in the pres- ence of transdermal enhancers (Moss et al., 2012) or different vehi- cles (Ghafourian et al., 2010). In some cases it has even been proposed that it may be best to apply both QSPRs and other models such as Gaussian process-based machine learning methods (Brown et al., 2012), with some claims that the latter may have signifi- cantly improved predictivity compared with the former (Lam et al., 2010). Several permeability correlation models have become very well established in the scientific community for the prediction of K p and are regarded as an ideal initial step in the development and characterisation process for pharmaceutical compounds. Often models are based on parameters such as the lipophilicity and molecular weight (or size) of a drug (Magnusson et al., 2004; Ng et al., 2012; Potts and Guy, 1992; Singh and Roberts, 1996), although other, related physicochemical parameters have been investigated in an attempt to improve the predictivity of the mod- els. Another approach employed to predict the ability of a drug to permeate the skin is to employ in vitro based model systems, for example those that have been successfully used to predict other biological permeations; such as those based on the blood–brain barrier (Sinkó et al., 2012), using electrokinetic chromatography (Wang et al., 2009), simple silicone membrane skin models (Khan et al., 2005; Waters et al., 2013) and most recently, chromato- graphic techniques (Hidalgo-Rodríguez et al., 2013). Simple mod- els, such as a silicone membrane, are particularly suited to studies that require comparative values, for example, with varia- tions in the presence of additional excipients or temperature 0928-0987/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.ejps.2013.08.002 ⇑ Corresponding author. Tel.: +44 01484 472190. E-mail address: l.waters@hud.ac.uk (L.J. Waters). European Journal of Pharmaceutical Sciences 50 (2013) 335–340 Contents lists available at ScienceDirect European Journal of Pharmaceutical Sciences journal homepage: www.elsevier.com/locate/ejps