Improving In Vitro–In Vivo Correlation (IVIVC) for Lipid-Based Formulations: Overcoming Challenges and Exploring Opportunities

1. Introduction

Besides the chemical and physical approaches applicable to the active substance, the galenic approach is a real opportunity to improve the bioavailability of poorly soluble BCS II or IV active substances. The Biopharmaceutics Classification System (BCS) categorizes drugs based on their solubility and permeability. BCS class II compounds are poorly soluble but highly permeable, while class IV compounds exhibit both low solubility and low permeability, making them particularly challenging to develop [1]. Various approaches exist to solve this problem; salts, co-crystals, or polymorphic forms can be used to improve the solubility of the active substance. The galenic approach can also be chosen; for example, amorphous solid dispersions (ASDs), complexation with cyclodextrins, or the use of lipid-based formulations (LBFs) could enhance bioavailability. The choice of a lipid formulation is based on the use of one or more lipidic excipients mixed with the active molecule. Lipid-based formulations are used to improve the body’s exposure to active substances after oral administration. Although the main advantage of LBFs lies in the increase in apparent gastrointestinal solubility, it has also been shown that these formulations can offer advantages in terms of permeability and, in certain circumstances, promote lymphatic transport and improve bioavailability by partially avoiding hepatic first-pass metabolism [2,3,4]. LBFs can appear as Self-Emulsifying Drug Delivery Systems (SEDDSs), and depending on the droplet size formed upon dispersion, they can be further classified as SMEDDS (Self-Microemulsifying) or SNEDDS (Self-Nanoemulsifying). In the 2000s, they were classified with the aim of grouping them according to their composition. This lipid-based formulation classification system (LFCS) was introduced by Pouton [5] and updated a few years later to add a fourth formulation type [6]. LFCS classifies LBFs into four main types, according to the relative proportions of lipids, surfactants, solvents, and co-solvents (Figure 1). Type I formulations are the simplest, comprising active ingredients dissolved in an oily vehicle, such as triglycerides alone or mixed glycerides. Type II formulations include combinations of glycerides and lipophilic surfactants with a low lipophilic–hydrophilic balance (HLB). Type III formulations include mixtures of glyceride lipids and more hydrophilic surfactants with a higher HLB and may also include co-solvents. Finally, a type IV HLB classification was introduced later in response to the growing use of formulations that do not contain traditional lipids. Type IV formulations include only a combination of surfactants and co-solvents without hydrophobic compounds [6]. Beyond SEDDS and other liquid lipid-based formulations, Solid Lipid Nanoparticles (SLNs) and Nanostructured Lipid Carriers (NLCs) represent additional lipid-based delivery systems [7]. SLNs are produced by replacing the liquid lipid of an emulsion with a solid lipid or a blend of solid lipids, resulting in solid lipids dispersed in an aqueous medium and stabilized with surfactants. NLCs were developed to overcome some limitations of SLNs, as they combine solid and liquid lipids as primary structural components. This allows NLCs to achieve higher drug-loading capacity and increase stability during storage.

Despite evidence of improved oral absorption and bioavailability of poorly water-soluble compounds via LBFs, their formulation design remains complex. Unlike conventional formulations, in the case of LBFs, the behavior of lipids in the gastrointestinal tract, and, in particular, their digestion after administration, must also be taken into account. Under these conditions, there are complex interactions between physiological and physical processes, making it difficult to rationally select an optimal composition of lipid excipients [8]. Moreover, in vitro studies cannot fully mimic these processes, and the predictability of available in vitro tests can sometimes be questioned. Indeed, it is difficult to accurately predict the performance of different lipid formulations. This lack of predictability can lead to difficulties when establishing in vitro/in vivo correlations (IVIVC) in the case of LBFs [9].

Those correlations are commonly used for formulations such as controlled-release systems, where the relationship between in vitro dissolution and in vivo drug absorption is more straightforward. In these cases, dissolution is the limiting factor, which means that the prediction of in vivo performance is directly linked to the prediction of dissolution kinetics. However, for complex formulations like LBFs, the establishment of a reliable IVIVC becomes more challenging. LBFs involve dynamic processes that are not easily captured by traditional in vitro dissolution tests or in silico pharmacokinetic models. As a result, predicting in vivo performance from in vitro or in silico data for LBFs requires more sophisticated models and a deeper understanding of the physiological factors that influence drug absorption.

2. IVIVCs in the Context of Lipid-Based Formulations

In recent years, the concept and application of in vitro–in vivo correlation for pharmaceutical dosage forms has received considerable attention from the pharmaceutical industry, academia, and regulators. Correlations between in vitro and in vivo data are often used during pharmaceutical development to reduce development time and optimize formulation [10]. From a regulatory perspective, IVIVC enables dosage form optimization while minimizing the number of clinical trials in humans. In addition, they can be used to establish dissolution acceptance criteria and serve as surrogates for additional bioequivalence studies [10]. According to the United States Pharmacopeia (USP), an IVIVC is “the establishment of a rational relationship between a biological property, or a parameter derived from a biological property, produced by a test method or a parameter derived from a biological property produced by a pharmaceutical form, and a physicochemical property or characteristic of the same pharmaceutical form” [11]. For the Food and Drug Administration (FDA), an IVIVC is “a predictive mathematical model describing the relationship between an in vitro property of a dosage form and a relevant in vivo response” [12]. Typically, the in vitro property is the rate or extent of drug dissolution or release, while the in vivo response is the plasma drug concentration or amount of drug absorbed. In vitro–in vivo correlation (IVIVC) is a valuable biopharmaceutical tool used in drug development to predict in vivo drug performance.

However, IVIVC has several limitations that affect its development, accuracy, and applicability [13]. One major limitation is physiological variability; when a correlation is established using preclinical data, translation from animal to human physiology can challenge its validity. In addition, inherent variability between human gastrointestinal tracts can further complicate predictions. Pharmacokinetics also presents challenges for IVIVC, as drugs exhibiting nonlinear absorption or metabolism, significant first-pass effects, or involvement in active transport mechanisms are more difficult to correlate. Another important limitation is that IVIVC is formulation-dependent; a correlation established for one type of formulation may not necessarily be valid for another. Finally, the level of correlation is an important consideration; level A correlations are the most informative but can be difficult to establish for drugs with complex absorption mechanisms, while level B and C correlations provide less precise predictions and may not satisfy regulatory requirements without supplementary data. Nevertheless, for the purpose of supporting formulation design, level B and C correlations are often sufficient.

There are different levels of in vitro–in vivo correlation, allowing relationships to be established between the dissolution properties and its in vivo performance in the case of dosage forms such as sustained release, where API release determines bioavailability [10,14]. Level A represents the most precise correlation, directly linking in vitro dissolution rate to drug entry into the body as shown in Figure 2, enabling bioavailability to be predicted without further human studies. Level B compares average dissolution and residence times in vivo, without matching plasma concentration profiles point by point. Level C links a dissolution time point to an average pharmacokinetic parameter (such as AUC or Cmax) but offers a less complete correlation. Multiple level C extends this approach to several dissolution time points, enabling certain formulation modifications to be justified. Finally, level D is a qualitative analysis or ranking with no regulatory value, mainly used to guide formulation development [10,14].

LBFs are designed to enhance the solubility and absorption of poorly water-soluble drugs, using lipid-based excipients such as oils, surfactants, and co-surfactants. However, these excipients introduce a level of complexity in the correlation between in vitro dissolution profiles and in vivo pharmacokinetics. While traditional formulations may rely on simple dissolution release testing to predict bioavailability [15], LBFs require more sophisticated models to account for the dynamics of lipid digestion, micelle formation, and lymphatic transport, for instance, which are not directly observable in basic in vitro tests. The lack of standardized, universally accepted in vitro and in silico methods that capture the full complexity of lipid-based systems can result in discrepancies between in vitro and in vivo data and between different studies.

There are notable case studies where the predicted in vivo behavior of LBFs based on in vitro data has failed to align with observed pharmacokinetic outcomes. On fenofibrate, Do et al. used in vitro dispersion data to examine the performance of four LBFs in comparison with in vivo data after administration in rats. However, the results failed to distinguish between LBFs administered in the fasted or fed state, and no correlation could be identified [16]. As examined by Feeney et al., it was found that of eight drugs studied using the pH-stat lipolysis device, only half correlated well with in vivo data [17]. Furthermore, in a paper published in 2014, Thomas and colleagues highlighted the lack of predictability of lipolysis through a study of fenofibrate and mini-pigs [18]. Another publication on an indirubin derivative failed to predict the in vivo performance of a lipid formulation with lipolysis data [19]. In the 2010s, attempts were made to establish IVIVCs for the BCS 2 molecule cinnarizine. Formulations were often similar between publications, and in vivo data were based on dog studies. Larsen et al. were only able to obtain a level D correlation and observed precipitation on one of the formulations during in vitro lipolysis, whereas in vivo performance was the same for all formulations [20]. Christophersen et al. were also able to obtain a qualitative correlation for the fasted state, but no quantitative correlation could be established for the fed state [21].

In terms of in vitro methods, dissolution tests are often used to perform in vitro–in vivo correlations, in particular the USP 2 apparatus. Lipolysis is useful for qualitative studies comparing formulations but is less suited to the development of IVIVC. Several articles demonstrate that the precipitation observed in vitro during lipolysis is not related to in vivo performance, and that studying the crystalline form of the precipitate provides a better understanding of the phenomena involved [22]. In most cases, lipolysis tests are designed to represent intestinal conditions only; therefore, lipolysis with a gastric step may have an impact on the predictivity of the method and may allow a closer mimicking of the biopharmaceutical process after administration [23].

Given the complexity of mimicking the in vivo behavior of lipid-based formulations using standard dissolution or lipolysis models, advanced approaches such as combined lipolysis–permeation models have been developed to provide deeper insights into their mechanisms. Indeed, when in vitro studies have failed to predict or reproduce the in vivo performance of LBFs, the lack of an in vitro absorption mechanism is often cited among the possible explanations for the lack of IVIVC [24]. These hybrid systems allow for a more comprehensive assessment by integrating key processes like lipid digestion and drug permeation, bridging gaps left by traditional methods. The following sections of this review will delve into these advanced tools and their applications in LBFs characterization.

To improve the success of IVIVC for lipid-based formulations, it is essential to understand and replicate their in vivo behavior. LBFs undergo digestion, forming mixed micelles that impact drug solubilization and absorption. Therefore, generating detailed in vitro data such as lipolysis profiles, dissolution, or permeability assays is a crucial first step. These data must then be linked to relevant in vivo studies to ensure translational value. When appropriately implemented, this approach could enhance predictive performance, making IVIVC a potentially powerful tool in LBF’s development.

3. Specific Considerations for LBFs

3.1. Dispersion/Digestion

When developing a new drug, the choice of formulation strategy is often guided by decision trees based on the physico-chemical properties of the active substance. In the case of LBFs, the behavior of lipids in the gastrointestinal tract, and most precisely their dispersion/digestion, must also be taken into account. One of the key characteristics of lipid-based formulations is their ability to solubilize active pharmaceutical ingredients through micellization. Surfactants can significantly enhance drug solubility by forming micelles, enabling solubilized concentrations that exceed the equilibrium solubility of the API alone [25]. Most of the time, above the CMC (Critical Micelle Concentration), the active molecule is directly dependent on the surfactant concentration. It has been observed that increasing surfactant concentrations can enhance the solubility of a molecule, a phenomenon reported for several different surfactants [26]. SEDDS are, therefore, mixtures of lipids and surfactants designed to enhance API solubilization. When they are dispersed in water or gastrointestinal fluids, especially those with high amounts of water-miscible cosolvents or hydrophilic surfactants, there may be a significant modification in solubilization capacity. If drug concentrations exceed the equilibrium solubilization capacity or precipitation does not occur immediately, the system is considered supersaturated [25]. This phenomenon is quantitatively described by the maximum supersaturation ratio (SRM), which reflects the supersaturation generated upon formulation dispersion and digestion [27]. SRM is calculated as the ratio between the maximum solubilized drug concentration achievable before digestion without precipitation and the equilibrium solubility of the drug in the digested aqueous phase.

After oral administration of lipid formulations, components are dispersed to form lipid droplets as explained above, followed by lipolysis and solubilization of digestion products by bile acids, forming a colloidal solution of mixed micelles. During lipolysis, triglycerides are digested into diglycerides, monoglycerides, and fatty acids by lipases and co-lipases. Lipase is a digestive enzyme produced by the pancreas. It becomes active only when it encounters the surface of emulsified fat droplets, working in conjunction with bile salts and co-lipase, a co-factor found in pancreatic juice. The presence of colipase and bile salts helps the enzyme bind effectively to its substrate and promotes the emulsification of fats. Once bound, lipase breaks the ester bonds in triglycerides, resulting in the formation of free fatty acids and 2-monoacylglycerols [28]. A well-dispersed lipid formulation is, therefore, necessary to ensure a contact surface with the gastric or intestinal environment and enable homogeneous lipolysis [29]. The intercalation of these digestion products with bile secretions generates lipid reservoirs for the active substance, ranging from liquid crystalline phases at the oil/water interface and smaller multilamellar and unilamellar vesicles to mixed micellar species in the bile-rich areas of gastrointestinal fluids (Figure 3) [30]. Surfactants can adsorb to undigested solid particles, partition into and emulsify food lipids, and form colloidal structures. They can also bind to species dissolved in luminal fluid, such as proteins. These interactions reduce the amount of monomers available to interact with the membranes of the intestinal epithelium. However, in the presence of a sufficiently high concentration of surfactant, monomers can interact with the epithelium at such a concentration that membrane alterations are observed [30].

Micelles are isolated from the rest of the intestinal contents in a layer of unstirred water at the level of the intestinal mucosa and dissociated by the pH effect. The mechanism of dissociation of lipolysis products from mixed micelles is not yet fully elucidated, but it has been suggested that the process is associated with an acidic microenvironment present in the layer of unstirred water overlying the intestinal membrane. According to the microenvironment theory, ionized fatty acids are converted to their non-ionized forms and dissociate from the micelles before being rapidly absorbed across the intestinal membrane [31]. After enzymatic digestion, these lipids are taken up by enterocytes at the apical brush border by diffusion. Once absorbed, some fatty acids can move directly into the blood via diffusion, while others are recombined in the endoplasmic reticulum to form triglycerides. These triglycerides are then subsequently secreted into the circulation in the form of chylomicrons [32,33]. The particle size of the dispersion also plays an important role, with smaller colloidal structures able to diffuse more easily through the unstirred water layer and bring the active substance to the site of intestinal absorption. Given the complex interactions in the gastrointestinal tract, it seems that the properties of the dispersion formed after interaction with bile and pancreatic secretions, including particle size, condition the performance of LBFs rather than the particle size of the dispersion at the initial stage [17].

3.2. Absorption

Absorption is a major factor influencing bioavailability. Absorption phenomena are described according to several modes: passive diffusion across membranes, transporter-mediated transport, endocytosis, or paracellular transport [34]. Lipid formulations have an impact on this phenomenon by promoting the passage of molecules from the intestinal lumen through the intestinal wall, with an increase in paracellular or intracellular permeability. To this end, certain lipid excipients can cause membrane fluidization, modify the type of intracellular transport, or impact the opening of tight junctions [35]. Indeed, medium-chain fatty acid esters can act as safe permeation enhancers, allowing transient and reversible opening of tight junctions between epithelial cells to promote the paracellular pathway or by interacting with the membrane to promote the transcellular pathway to the bloodstream (Figure 4) [36]. For example, the medium-chain fatty acid Labrasol® ALF (caprylocaproyl polyoxyl-8 glyceride), manufactured and commercialized by Gattefossé (Saint Priest, France), was studied by McCartney et al. in 2019 [37]. Using ex vivo Ussing chamber experiments, followed by in vivo confirmation, the authors demonstrated that this excipient transiently induces mild intestinal epithelial perturbation and reversibly opens tight junctions, allowing the transport of molecules with molecular weights comparable to that of insulin. In 2024, the same research group further confirmed the absorption-enhancing properties of medium-chain fatty acids by investigating a second excipient, Labrafac™ MC60 (glycerol monocaprylocaprate) [36]. This improvement in permeability could be beneficial for BCS class III and IV active substances.

On the other hand, long-chain fatty acid esters can promote transport to the lymphatic circulation for highly lipophilic actives, thus bypassing the liver and reducing the first-pass effect and the metabolism for actives highly sensitive to this degradation pathway (Figure 4) [38,39]. This is one of the major advantages of lipid-based formulations. To achieve this, drugs are co-absorbed with long-chain fatty acids that associate with lipoproteins in chylomicrons, and the contents of these chylomicrons are then released directly into the lymphatic system [2]. Halofantrine has been used as a model drug to study lymphatic transport in the presence of lipids. By comparing short-, medium-, and long-chain fatty acid-based vehicles, the authors demonstrated that lymphatic absorption of halofantrine is significantly enhanced in formulations containing long-chain glycerides in cannulated rats [40]. Similar conclusions were observed in cannulated dogs, where 1 g of lipid-based formulations containing long-chain fatty acids also resulted in increased lymphatic uptake [41]. Although molecular weight does not appear to significantly influence lymphatic transport, the lipophilicity of a drug (expressed as log p) seems to impact its distribution between the blood and lymphatic system. According to several studies, lymphatic uptake from simple lipid formulations or oil solutions is minimal or absent for compounds with a log p < 5 [42]. Lymphatic transport is primarily evaluated using in vivo models, which remain the most established approaches. These include animals with cannulated lymphatic ducts or the administration of chylomicron secretion inhibitors such as cycloheximide [43]. However, such in vivo models are difficult to establish, and consequently, data on lymphatic transport remain limited. In contrast, in vitro methods to assess this absorption pathway are far less common. The lymphatic system is a complex biological network involving transport, resynthesis, assembly, and dissociation processes, which is difficult to replicate in vitro [32]. However, recent developments have introduced in vitro approaches based on artificial chylomicrons derived from Intralipid® emulsions combined with absorption assays. These emerging models offer a preliminary evaluation of both the lymphatic transport potential of active pharmaceutical ingredients and the influence of lipid-based excipients on this process [39].