Introduction

Nanotechnology offers broader opportunities for the development of novel drug delivery systems to overcome the limitations of traditional pharmaceutical formulations. Conventional drugs suffer from drawbacks such as poor solubility and low permeability, which subsequently result in poor bioavailability and therapeutic efficiency.1 These issues are particularly pronounced in oral drug delivery, owing to environmental variability, enzymatic degradation, and first-pass metabolism.2 Nanotechnology-based drug delivery systems exploit their size and specific structure to circumvent these hurdles. Various nanocarriers, including micelles, dendrimers, polymer-based nanosystems, and lipid-based nanosystems, have been developed to enhance the effectiveness of drug delivery via various routes.3–5 Lipid-based nanosystems, which are often categorized as vesicular systems, such as liposomes, niosomes, ethosomes, transferosomes, and matrix systems (or lipid nanoparticles, LNPs), such as solid lipid nanoparticles (SLNs) and nanostructured lipid carriers (NLCs), have attracted considerable interest owing to their potential as drug delivery systems.6–9 In oral drug delivery, LNPs can protect the drug from degradation in the acidic environment of the stomach and enzymatic hydrolysis of the intestines.10,11 They also proved to be effective in enhancing permeability across epithelial membranes, facilitating lymphatic transport, and bypassing first-pass metabolism, thereby improving systemic bioavailability.12–14

Despite their potential as novel drug delivery systems, a comprehensive understanding in LNPs production remains lacking. The influences of input materials and process parameters on product properties and therapeutic performance have not been clearly explained, which can be attributed to inconsistencies such as batch-to-batch variations in particle size, zeta potential, encapsulation efficiency, and drug release profiles during formulation. Furthermore, the traditional experimental procedure of changing one-factor-at-a-time (OFAT) is deemed inefficient.15 For instance, modifying the lipid type without adjusting surfactant concentration may not improve drug loading, and important interactions between formulation variables may be overlooked. Correspondingly, it was realized that adding more tests did not enhance the product quality. The trial-and-error nature of this method is also resource-intensive, as researchers often need to restart the entire experiment when target quality attributes are not achieved. This approach poses challenges not only in terms of cost, but also in meeting predefined product quality requirements.16,17

The quality by design (QbD) approach offers a more systematic design for production, reducing exhaustive experimental procedures to more efficient ones. The QbD strategy for pharmaceutical formulations has been implemented since the publication of several guidelines by the International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH).18–23 The foundation of QbD is based on scientific design and manufacturing. Its core value is that the quality should be incorporated into the product. In addition to pharmaceutical knowledge, a mathematical-statistical understanding is important when applying QbD elements in pharmaceutical production. As part of QbD, optimization via design of experiment (DoE) is crucial for defining the mathematical model of the relationship between independent and dependent variables. DoE can also identify the statistical significance and optimal conditions to achieve the desired drug quality.24

In this review, the challenges of oral administration and the corresponding role of LNPs are discussed. Subsequently, we explored the implementation of QbD principles by identifying the quality profile in relation to the selection of materials and preparation methods based on risk assessment to understand the path for improvement in SLN and NLC production for oral drug delivery. Additionally, this article summarizes the findings from recent studies regarding the impact of input variables on output responses based on DoE results, where the optimized products were evaluated in vitro, in vivo, and ex vivo for various purposes.

Challenges of Oral Drug Delivery

Oral delivery is generally considered a more favorable drug administration route for patients and physicians owing to its effortlessness, noninvasive nature, and high patient acceptability.25 Additionally, orally administered drugs may specifically target certain regions, thereby localizing therapy in gastrointestinal (GI) diseases, such as gastroesophageal reflux disease, inflammatory bowel disease, GI cancer, and colorectal cancer.26–29 Despite these advantages, oral drug delivery still faces challenges associated with the inherent properties of drugs and the complexity of the GI system during their traversal.

The GI tract is particularly intriguing because of the variability in its environment. Greatly different pH values are encountered by the drugs, from acidic conditions in the stomach (pH 1 − 2.5), duodenum (pH 6.1), intestines (pH 7.1 − 7.5), and higher pH conditions in the colon (pH 7 − 8).30 Furthermore, the presence of various GI enzymes is particularly challenging for protein-based drugs. They may be hydrolyzed and degraded by pepsin in the stomach or by other proteolytic enzymes in the small intestine.31 In contrast, the enzymatic activity of pancreatic lipase may promote lipolysis in the GI tract and enhance the solubilization of lipophilic drugs or lipid-based formulations.32 GI enzymes also play pivotal roles in the first-pass metabolism. This presystemic metabolism is mediated mainly by cytochrome P450 enzymes such as CYP3A4, which are predominantly found in the liver and small intestine.33 The large surface area and low blood flow in the small intestine might prolong CYP3A4 exposure toward drugs, enabling more extensive metabolism and thus decreasing the oral bioavailability of the drugs.34 This intestinal first-pass effect is markedly enhanced by drugs that are substrates of CYP3A4, such as felodipine, nifedipine, atorvastatin, and simvastatin.35

The small intestine is considered the primary site of oral drug absorption, owing to its extensive surface area and various transport modes. The microvilli in the small intestine are lined with goblet cells, which facilitate glycoprotein secretion and form a mucosal layer.36 This layer is primarily composed of mucin, an oligosaccharide-rich glycosylated protein that provides an overall negative charge to mucus, thus facilitating electrostatic interactions with positively charged substances.37,38 Intestinal mucus is also involved in the formation of an unstirred water layer (UWL) between the intestinal bulk fluid phase and the epithelial brush border. The presence of UWL can be detrimental to lipophilic substances because of hindered access to the epithelium, limiting their passage to the systemic circulation.39

To reach the systemic circulation, drugs must be able to cross the intestinal epithelium, which primarily comprises enterocytes, along with a smaller number of other cells, such as goblet cells, microfold cells (M cells), and Paneth cells.40 Epithelium crossing involves several different transport mechanisms owing to its particular structure and constituents. Passive diffusion is the most common pathway, which relies on the movement of drugs along their concentration gradient, from a higher amount on the apical side to a lower amount on the basolateral side of the membrane. This mechanism may occur via the paracellular or transcellular routes. In the paracellular route, drugs traverse the enterocytes through intercellular tight junctions, allowing the movement of smaller hydrophilic molecules. In contrast, lipophilic molecules can naturally diffuse through the cell membrane transcellularly, owing to their similar affinities.41 Furthermore, transcellular absorption may also occur via endocytic mechanisms, such as phagocytosis by immune cells or M cells, macropinocytosis, clathrin-mediated or clathrin-independent endocytosis, and caveolae-mediated or caveolae-independent endocytosis.42

Unlike passive diffusion, active transport allows molecules to move against the concentration gradient. This transporter-mediated movement offers specific pathways during drug traversal and requires a certain amount of energy derived from adenosine triphosphate (ATP). Among the ATP-dependent transporters, the role of P-glycoprotein (P-gp) has been particularly highlighted in oral drug administration. P-gp is an efflux pump that acts as a defense mechanism by actively pushing xenobiotics out of enterocytes.36 However, this mechanism can be disadvantageous for orally administered drugs because it reduces the intracellular concentration of drugs, thereby limiting drug absorption into systemic circulation and reducing oral bioavailability. This limitation is particularly pronounced for the absorption of drugs that are P-gp substrates.43

Role of Lipid Nanoparticles in Oral Drug Delivery

Drug incorporation into LNPs, such as solid lipid nanoparticles (SLNs) and nanostructured lipid carriers (NLCs), is a feasible strategy for overcoming the variability of GI barriers and the complexity of transport modes during oral administration (Figure 1). As a nanoscale system, the structure of LNPs intrinsically generates a larger surface area. This results in increased interaction between the systems and biological membranes, facilitating enhanced absorption into blood circulation.44 The components of LNPs, which mainly comprise lipids and surfactants as well as additional excipients, provide specific advantages for oral drug traversal.

Figure 1 Drug transport mechanisms across intestine. Created in BioRender. Suliman, (K) (2025) https://BioRender.com/vmkadsf.

Protection From Degradation

The solid lipid components in LNPs, particularly long-chain fatty acids, can slow the degradation process by digestive enzymes, resulting in higher drug stability in the GI environment.40 This allows the drug to reach its target active form, thereby enhancing its therapeutic effectiveness.45 A study conducted by Veni and Gupta showed that the drug release of linagliptin formulated into SLN using stearic acid as a solid lipid at pH 6.8 was higher than that at pH 1.2. This signifies the ability of stearic acid to protect linagliptin from the gastric environment, thereby ensuring intestinal release of the drug.46

Using cetyl palmitate as a solid lipid component in the LNP formulation can also be advantageous for minimizing drug degradation owing to its low susceptibility to lipase hydrolysis, leading to prolonged drug retention in the formulation.47 El-Dakroury et al reported that the cumulative release of fexofenadine HCl in an acidic medium was lower when formulated in SLN using cetyl palmitate compared with the pure drug.48 Similarly, the release of doxorubicin from NLC containing the same solid lipid was lower in simulated gastric fluid than in simulated intestinal fluid.49

Inhibition of Cytochrome P450

First-pass metabolism, mainly mediated by cytochrome P450 enzymes, such as CYP3A4, is a major drawback of effective oral administration. Several LNP constituents may act as modulators of CYP3A4 in bypassing the first-pass metabolism, and subsequently increasing drug absorption.50 Unsaturated fatty acids, such as oleic acid, can inhibit CYP isoforms including CYP3A4. This may be due to the ability of fatty acids to disrupt the microsomal membrane and prevent drug binding to the enzyme active site.51 Several reports have shown that formulating NLC using oleic acid as a liquid lipid can improve the oral bioavailability of telmisartan and fexofenadine HCl.52,53 In a study by Sharma et al, the use of oleic acid in NLC markedly increased the plasma concentration and bioavailability of atorvastatin, which is a CYP3A4 substrate, compared to the marketed drug.54

Surfactants also play a pivotal role in the inhibition of CYP3A4. Non-ionic surfactants, such as Cremophor® EL and Cremophor® RH-40 have inhibitory effects on CYP3A4, with IC50 values ranging between 0.40 to 0.80 mM. Cremophores can alter drug absorption owing to agent-produced membrane fluidization, causing perturbations toward the CYP3A4 microenvironment, thus decreasing enzyme function.51 An in vivo pharmacokinetic study demonstrated increased bioavailability of silybin in SLN formulated using Cremophor® RH40 as a surfactant.55 In another study, it was observed that using Cremophor® EL in the SLN of fenofibrate, a CYP3A4 substrate, resulted in higher plasma concentrations than those of the pure drug.56

Inhibition of P-Glycoprotein

P-glycoprotein (P-gp) is an ATP-binding cassette transporter predominantly found in the apical layer of epithelial cells. Its primary function as a xenobiotic efflux pump can be detrimental for oral delivery due to restricted drug transport across the basolateral layer, thereby limiting the amount of drug in the systemic circulation.57 Designing LNPs with appropriately selected constituents that act as P-gp inhibitors is an effective strategy to improve drug transport across the intestinal barrier. In general, the commonly proposed P-gp inhibitory mechanism involves obstruction of drug-binding sites, disruption of ATP hydrolysis, and alteration of cell membrane integrity.58 Lipid-based excipients, such as glyceryl monooleate and hard fat, as well as surfactants, such as Cremophor®, poloxamers, and polysorbates, are among the components that can be utilized in LNP formulations for these purposes.59 Tween® 80 (polysorbate 80) has been specifically recognized for its synergistic role as an inhibitor of P-gp and CYP3A4.60 Beloqui et al showed that the transport of saquinavir, a known P-gp substrate, across Caco-2 cells was enhanced when incorporated into an NLC system prepared with polysorbate 80.61 Furthermore, an in vitro permeability study demonstrated higher apical-basolateral transport of tilmicosin-loaded NLC prepared using poloxamer 188 and polysorbate 80 as surfactants.62 The use of Gelucire® 44/14 and polysorbate 80 as liquid lipids and surfactants in NLC also significantly increased the plasma concentration of iloperidone, confirming the potential of both excipients as P-gp inhibitors.63

Enhancement of Lymphatic Pathway

The enhancement of oral bioavailability offered by LNPs can also be facilitated through lymphatic transport, allowing lipophilic drugs to bypass first-pass metabolism. Lipids in LNPs may promote the formation of chylomicrons, which are large lipoproteins that are partly responsible for transcellular drug absorption.36 In the GI tract, triglycerides from lipids are broken down by lipase into monoglycerides and free fatty acids, which are then absorbed by enterocytes and incorporated into chylomicrons. These drug-loaded chylomicrons subsequently enter the lymphatic capillaries and avoid hepatic metabolism to finally reach the systemic circulation.45 Molecules with Log P > 5 are naturally transported via the lymphatic pathway. However, drugs with lower Log P values, such as atazanavir (Log P = 4.1), can also utilize this route when encapsulated within the NLC system. Using the chylomicron flow block model, Gurumukhi and Bari revealed the ability of NLC to circumvent first-pass metabolism, resulting in a higher plasma concentration of atazanavir.64

Lymphatic transport can also be facilitated transcellularly through uptake by M cells, which are specific epithelial cells essential for the intestinal immune system. These cells are primarily found in the gut-associated lymphoid tissue (GALT) or Peyer’s patches.40 Functionalization of LNPs with targeting ligands such as lectins enables specific binding to receptors on the M cell surface, thereby enhancing absorption and subsequent transport into the lymphatic system. A study by Hädrich et al demonstrated increased phagocytosis of quercetin NLC when its surface was functionalized with wheat germ agglutinin.65

Enhancement of Mucoadhesion

LNPs can also contribute to enhanced mucoadhesion in oral drug delivery via electrostatic, covalent, hydrogen, and van der Waals interactions.66 Intestinal epithelial cells are protected by a hydrophilic negatively charged mucus layer that serves as a barrier against foreign particles. The functionalization of LNP using positively charged polymers, such as chitosan, is considered a feasible strategy to enhance mucoadhesion via electrostatic interactions, subsequently prolonging the drug residence time in the GI tract, enabling controlled drug release, and improving oral bioavailability. A study by Pyo et al showed higher plasma concentrations of chitosan-coated fenofibrate NLC than of uncoated fenofibrate NLC. The authors suggested that chitosan contributes to enhanced mucoadhesion while also acting as a tight junction modulator in intestinal enterocytes, thus facilitating drug transport into the systemic circulation.67 In another study, a chitosan-functionalized SLN of thymoquinone showed a higher mucoadhesive efficiency than that of free thymoquinone. This may be due to electrostatic interactions between the cationic chitosan-functionalized SLN and anionic mucin molecules. In addition, the hydrophilic properties of chitosan further intensified mucoadhesion.68

Mucoadhesion can also be enhanced through covalent bonding between mucosal cysteine residues and thiomer (thiolated polymer) molecules. For instance, it was found that the use of thiolated polyoxyethylene oleyl ether surfactant in aprepitant-loaded NLC displayed prolonged adhesion to goat intestinal mucosa compared to unmodified surfactant-based aprepitant NLC and aprepitant suspensions. Furthermore, the modified NLC formulation also exhibited increased plasma concentration and relative oral bioavailability compared to the drug suspension and unmodified NLC.69

Quality by Design Framework

The development of LNPs involves complex interactions between formulation components and process parameters, which often result in challenges such as variability in particle size distribution, low encapsulation efficiency, poor drug loading, low zeta potential, and inconsistent drug release behavior. These attributes are critical to the performance of LNPs but are difficult to optimize using traditional OFAT approaches because of the absence of interaction analysis and the need for numerous trial-and-error iterations. In this context, the QbD framework serves as a powerful and structured approach for systematically exploring the formulation space. By employing design of experiments (DoE), QbD enables the identification and control of critical material attributes (CMAs) and critical process parameters (CPPs) that influence the critical quality attributes (CQAs). Beyond its conceptual strengths, QbD also offers practical advantages such as shortened development time, reduced experimental workload, and enhanced precision in targeting desired product characteristics. This is particularly valuable for complex systems like SLNs and NLCs, where small variations in formulation or processing can significantly impact particle size, encapsulation efficiency, and release profile, which subsequently translates into therapeutic performances. Consequently, QbD not only improves formulation robustness and scalability but also supports regulatory alignment and cost-effective LNP development.70,71

Quality Target Product Profile

Quality target product profile (QTPP) is a foundational component of the QbD framework, outlining the desired profile of a drug product to provide optimal safety and efficacy.70 In pharmaceutical product development, the QTPP provides a prospective summary of the final drug product, including dosage form, delivery system, route of administration, dosage strength, container closure system, drug release, pharmacokinetic properties, purity, sterility, and stability.18 In the context of oral drug delivery, the identification of QTPP related to the enhancement of systemic and/or lymphatic absorption is particularly crucial.72 For example, in a QbD-based study of atazanavir-loaded NLC, the QTPP stated that the defined pharmacokinetic parameters should be higher than the reference to provide higher drug concentrations, ensuring higher lymphatic uptake of the drug.64

In the identification of the QTPP, researchers should consider the regulatory requirements for bioequivalence and patient adherence, ensuring that the product matches the therapeutic performance of the reference products.73 The QTPP not only guides formulation and manufacturing strategies but also provides a benchmark for assessing critical quality attributes (CQAs) throughout the development process, aiming for a high-quality pharmaceutical product that meets both safety standards and therapeutic goals.

Critical Quality Attributes

As defined in the ICH Q8 (R2) document, CQA is another important element of QbD that represents the physical, chemical, biological, or microbiological properties of a drug product, which must remain within a defined range, limit, or distribution to ensure the desired quality.18 Quality attributes can be either critical or noncritical. When failure to achieve a specified range results in no efficacy or potential harm to the patient, an attribute should be considered critical.73 Along with the QTPP, CQAs play a vital role in guiding product and process development, ensuring that the safety and efficacy standards are met. However, CQAs differ from QTPP in scope and function within the QbD framework. QTPP outlines the overall desired characteristics and quality of the final drug product, such as the expected release profile and therapeutic effect, whereas CQAs may include specific parameters, such as particle size and encapsulation efficiency, which need to be tightly controlled to meet the target profile.74 Furthermore, unlike the more fixed profile of the QTPP, CQAs serve as adjustable responses to changes in the formulation attributes or process parameters. Thus, CQAs play a critical role in bridging the quality objectives outlined in the QTPP with the practical aspects of formulation and process development.71

In the production of lipid-based nanocarriers, common CQAs include particle size, polydispersity index, encapsulation efficiency, drug-loading capacity, cumulative drug release, and zeta potential.75,76 For orally administered drugs, the control of uniformly small particles is substantial because nanosized particles (<1000 nm) provide a greater surface contact area, subsequently increasing the intestinal absorption.77 Furthermore, nano-sized particles can be transported both paracellularly and transcellularly (via endocytosis by enterocytes or via M cell uptake).25,78 However, a larger particle size may be useful for extended drug release. LNPs larger than 150 nm are more likely to be taken up by phagocytes, which act as reservoirs and accumulate inside the liver or spleen over an extended period before being gradually released into systemic circulation.79,80

The selection of zeta potential as a CQA also affects the performance of the final product. Zeta potential describes the surface charge of a colloidal particle, which is measured as the electrical potential at a layer relative to a certain point in the bulk medium. A higher absolute value of the zeta potential (≥ ±30 mV) indicates stronger repulsive electrostatic interactions between particles, thus preventing aggregation and ensuring the stability of the dispersion system.81 Moreover, from the perspective of drug delivery, the surface charge of nanoparticles is partially responsible for stronger membrane binding and cellular uptake enhancement.82,83 Nisini et al found that positively charged liposomes could interact with the negatively charged mucosal surfaces of tumor cells, facilitating liposome endocytosis by antigen-presenting cells, thus enhancing cell-mediated immune responses.84 The dependence of cytotoxicity on the zeta potential was also demonstrated in a study by Shao et al, where positively charged nanoparticles resulted in higher cytotoxicity toward L929 cells.85

Another property that may be considered a CQA in orally administered LNPs is the percentage of unpleasant taste, which describes the palatability of the drug product. In a study of diacerein-loaded SLN, it was found that an optimum amount of lipid was suitable for producing a palatable preparation.86 Several studies have also revealed the ability of lipids to control the release of bitter drugs in saliva, effectively sustaining the concentration of drugs that reach bitter taste receptors.87,88

Critical Material Attributes

Critical material attributes (CMAs) mainly identify the state of the input materials, such as the drug substances and excipients employed during production. These attributes encompass a wide range of material properties within an acceptable range, which can influence the quality profile of the final product.89 Determining the interrelationship between CMAs and CQAs is fundamental during the QbD process, where material attributes are systematically identified, screened, and controlled based on their impact on quality.90 In the development of SLNs and NLCs for the oral route, determination of CMAs is particularly crucial in relation to their objectives, both to protect the drugs from the GI environment and to deliver them into the systemic circulation.

Generally, the type and amount of lipids are the primary considerations in LNPs production. Changes in the type of lipid, drug-to-lipid ratio, solid-to-liquid lipid ratio, or total lipid concentration can affect the particle size, drug encapsulation efficiency, and release profile of nanoparticles.62,91 For example, the use of solid lipids in different polymorphs can influence the phase transition temperature, which affects lipid crystallinity and the likelihood of drug expulsion from the nanoparticle matrix.92 In NLCs, the optimum amount of liquid lipids may provide greater encapsulation efficiency because of the lower melting point of the system, which subsequently enhances the dissolving capacity of the matrix.93 Furthermore, solid and liquid lipid compositions also affect the type of NLCs, which are classified as imperfect, amorphous, or multiple NLCs. In imperfect NLCs, a lower amount of liquid lipid is blended with solid crystalline lipids, such as glycerides, reducing crystallinity and promoting the formation of an unstructured matrix during cooling. This facilitated a higher drug loading inside the matrix voids.94 In the amorphous type, specific non-crystalline solid lipids, such as medium-chain triglycerides, hydroxystearate, or isopropyl myristate, form an amorphous core along with liquid lipids. The disordered nature of the matrix can minimize drug leakage during storage.95 In multiple NLCs, liquid lipids such as medium-chain and long-chain triglycerides or oleic acid are employed in higher amounts, which enables phase separation and oil compartment formation within the solid lipid. This compartment provides a suitable environment for solubilizing lipophilic drugs, and subsequently promotes sustained or controlled drug release during oral administration.96

Surfactants also play an important role as CMAs in LNPs formulation. The type and concentration of the surfactant can modify the surface properties of the nanoparticles, thereby influencing drug loading, stability, particle size distribution, and pharmacokinetic profile.97 For instance, cationic surfactants such as hexadecyltrimethylammonium bromide (CTAB) can improve mucoadhesion by forming electrostatic interactions with negatively charged endocytosis-inducing biological membranes, thus enhancing the cellular uptake of nanoparticles.98,99 Both the lipids and surfactants selected in the SLN and NLC formulations should be generally recognized as safe based on their biocompatibility, biodegradability, and non-toxicity (Table 1).

Table 1 Common Excipients in LNPs for Oral Drug Delivery

In surface-modified LNPs, the coating materials can also be considered CMAs. A study by Veni et al indicated that increasing the amount of eudragit, a pH-sensitive polymer, delayed the drug release from the SLN matrix. Eudragit-coated SLN remain stable in acidic gastric environments, whereas the drug is gradually released upon arrival in alkaline intestinal environments.46 The addition of surface charge modifiers to LNPs may also be beneficial. El-say et al reported that a higher zeta potential was obtained by increasing the concentration of stearylamine, a positive charge-inducing agent, thus improving the stability of LNP.123

Critical Process Parameters

Critical process parameters (CPPs) refer to specific input operating or process state variables that are controlled and monitored during production.89 These parameters can impact CQAs, and highly impactful factors should be prioritized. The criticality of a process parameter lies in its ability to satisfy the desired product quality. In LNPs production, the identification of potential CPPs is directly related to the selection of the preparation method (Figure 2). Based on preliminary information, the operating range of a specific parameter can be established to obtain the optimal conditions for the preferred CQA results.70 Several preparation methods, such as melt emulsification-ultrasonication, high-pressure homogenization (HPH), and high-shear homogenization (HSH), have been widely employed for the QbD-driven development of LNPs.100,110,111 Other methods such as hot-melt extrusion, phase inversion temperature, solvent diffusion, solvent evaporation, and spray drying have also been reported.91,130,134,135 However, there have been fewer QbD studies conducted using these methods.

Figure 2 Critical process parameters in various LNP preparation methods.

Each preparation technique offers a different mechanism along with its advantages and disadvantages. For example, in HPH, input materials are accelerated by high pressure (100 − 2000 bar) through a micron-size gap, where the resulting cavitation force and shear stress can break down the particle to the nanometer size.136 However, the high-energy nature of HPH potentially leads to a suboptimal polydispersity index (PDI) owing to uneven particle disruption.64 It is worth noting that process variations in HPH involving hot and cold conditions facilitate distinctive LNP characteristics. In hot HPH, the drug-lipid melt is combined with a surfactant above the lipid melting point. The mixture was emulsified. The hot emulsion was subsequently homogenized for several cycles at a specific pressure and cooled to room temperature to obtain the SLNs or NLCs. Conversely, in cold HPH, the drug-lipid melt is first solidified using dry ice or liquid nitrogen. The solid was then crushed to obtain micron-sized particles, which were dispersed in a cold surfactant solution, followed by high-pressure homogenization.137 Relatively smaller particles can be obtained using the hot process because of the decreased viscosity at higher temperatures. Nevertheless, high temperatures may also promote rapid degradation of the LNP system;136 thus, certain conditions should be optimized to control the quality of the final product. During the HPH process, the number of homogenization cycles also contributes to the particle size and PDI and is commonly considered a CPP in the HPH technique.138

High-shear homogenization and ultrasonication-based methods are other approaches involving high energy. Both techniques offer a similar mechanism. Prior to nanosizing, the molten lipid was prepared at 5 − 10 °C above its melting point and dispersed in a surfactant solution under stirring at the same temperature. The resulting emulsion was homogenized or sonicated to reduce the droplet size to the nanometer scale.139 The homogenization speed and cavitation-generating ultrasound amplitude are responsible for particle breakdown and thus may be considered as CPPs in their respective methods.140 Concurrently, the duration of sonication and homogenization should be optimized to obtain the desired LNP characteristics without overheating the sample.141

In the spray-drying method, LNPs can be converted into dry powders that offer better physicochemical stability than the dispersion form. Atomization of the aqueous LNP dispersion may occur because of the high inlet temperature of the spray dryer.142 The feed flow rate also promotes physical quality alteration of the LNPs. Mozaffar et al reported that, at a higher temperature (180 °C) and flow rate (15 mL/min), the particles were more evenly distributed and free of aggregates.135

The hot-melt extrusion (HME) technique is a relatively less explored method for LNPs preparation, despite its scalability and environmentally friendly characteristics.143 During HME, the main ingredients are pumped and mixed inside an extruder barrel at 10 − 15 °C above the solid lipid melting temperature. Excipients can be subsequently added to certain feeding zones during extrusion. The resultant hot pre-emulsion was then subjected to either sonication or high-pressure homogenization to reduce the particle size.56,134,144 In this method, varying the extrusion barrel temperature and screw speed evidently influenced the particle size and encapsulation efficiency of the LNPs, and thus could be selected as CPPs during production.

In the phase-inversion temperature (PIT) method, a certain type of emulsion is transformed into its reversed type by continuously changing the mixture temperature.136 This technique primarily exploits the temperature-dependent properties of the hydrophilic-lipophilic balance (HLB) of surfactants. A mixture of drugs, lipids, and surfactants was prepared prior to the phase inversion. The emulsion was then subjected to several cycles of heating and cooling, followed by rapid cooling using cold water (0 °C). This treatment breaks down the emulsion system, resulting in the formation of stable LNPs.130 The temperatures for phase inversion and the heating/cooling rate may influence the resulting nanoparticles; thus, they can be considered as CPPs in the PIT method.145

To select CPPs for each preparation technique, it is necessary to recognize the tunable operating parameters and process state variables that influence the CQAs of the product. Subsequently, the established potential operating space could be further employed during continuous manufacturing of LNPs.

Risk Assessment

In the QbD framework, risk assessment is a vital yet distinct component that supports decision-making by identifying and prioritizing material attributes and process parameters that may impact CQAs.18 While QbD focuses on defining and achieving the QTPP through the systematic control of formulation and process variables, risk assessment serves to anticipate potential sources of variability or failure that could compromise product quality, safety, or cost efficiency.

Several risk assessment tools can be used to guide this process. For example, the Ishikawa (fishbone) diagram provides a broad overview of the contributing factors from each category, describing the cause-and-effect relationships between the variables (Figure 3).112,124 The risk estimation matrix (REM) ranks variables based on their qualitative impact on CQAs.101 In contrast, quantitative tools like failure mode and effects analysis (FMEA) assign numerical values to severity, occurrence, and detectability of potential failures to generate a risk priority number (RPN).90 This allows researchers to prioritize experimental efforts based on criticality. Thus, while QbD is primarily concerned with achieving product quality through robust design, risk assessment complements it by systematically evaluating failure points and hazards, ultimately enhancing the efficiency and reliability of the development process.

Figure 3 Example of an Ishikawa diagram in LNP optimization.

Design of Experiments

During the QbD course of action, the interactions between independent variables (CMAs and/or CPPs) and dependent variables (CQAs) were formalized through the design of experiments (DoE). In the optimization step, these interactions are typically represented by a polynomial equation that captures single-factor effects, two-factor interactions, and quadratic relationships.146 The generated polynomial equation describes the interplay between factors, providing recommended paths to follow upscaled production. Various aspects are considered when selecting DoE models, such as the purpose of the investigation, factors and responses to be investigated (number, levels, qualitative or quantitative), resources (materials, time, and budget), prior knowledge, and historical data.147 Based on the objective of this study, DoE is generally categorized into screening and optimization designs.148 To concise this article, the DoE models that are commonly used in the production of LNPs are briefly reviewed. More in-depth information regarding the technical details of DoE models has been discussed extensively in several other studies.149–151

Screening Designs

A screening design is considered an initial approach to isolate potentially more significant factors from the numerous possible factors influencing responses. Although methodically different, the function of the screening design is similar to that of risk assessment.152 In general, a screening step is used only to determine the important variables experimentally observed in subsequent optimization designs. Several models, such as two-level full factorial, fractional factorial, Plackett-Burman, and Taguchi, are the most frequently employed screening designs in pharmaceutical formulations.153

Factorial design is one of the most comprehensive DoE models which allows multiple factors to be screened simultaneously. The full factorial design (FFD) examines all possible combinations of factor levels, ensuring a comprehensive analysis of the main effects and interactions. A two-level FFD is commonly employed during the screening step, with levels denoted as high (+1) and low (−1) (Figure 4a). The total number of experimental runs is n = 2k, where 2 and k represent the number of levels and factors, respectively.70 This design allows the evaluation of both the main effects and the interaction effects between the variables. However, as the number of factors increases, the required number of experiments increases exponentially, making it impractical for a large number of factors. To address this, fractional factorial design (FrFD) offers a more efficient alternative by selecting only a subset (fraction) of the full factorial runs while still capturing the significant effects (Figure 4b). Instead of performing all the 2k experiments, a fraction, such as half (2k−1) or a quarter (2k−2) of the total runs, was conducted.151 For example, in a four-factor scenario, instead of performing all 16 runs for a full 24 design, an FrFD with 24–1 = 8 runs may be sufficient for the initial screening.

Figure 4 Schematic illustration of Full Factorial Design, FFD (A); Fractional Factorial Design, FrFD (B); Box-Behnken Design, BBD (C); Central Composite Circumscribed Design, CCCD (D); Central Composite Inscribed Design, CCID (E); Central Composite Face-centered Design, CCFD (F). Created in BioRender. Suliman, (K) (2025) https://BioRender.com/10a19ff.

The Plackett-Burman design (PBD) is a highly efficient screening design that focuses solely on identifying the most critical factors among a large number of factors while maintaining a minimum number of experiments. Unlike factorial designs, PBD focuses exclusively on estimating the main effects and assumes that interaction effects are negligible.150 The total number of runs in a PBD follows a multiplication of four greater than the number of factors. For instance, the design may require only twelve runs when there are seven or eight factors to screen.154,155 This makes PBD particularly useful for preliminary screening, where the goal is to quickly eliminate insignificant factors before moving to a more detailed optimization phase. However, because interaction effects are not accounted for, PBD is best suited for cases in which factor interactions are either minimal or of no primary interest.71

The Taguchi design is a specialized factorial design that incorporates orthogonal arrays to systematically reduce the number of experiments while ensuring robust results.156 However, in contrast to traditional factorial designs that focus on the main effects and interactions, the Taguchi design emphasizes the improvement of process stability using a signal-to-noise (S/N) ratio to quantify the stability and performance of a system under varying conditions. In the experimental setting, the signal (S) represents the desired output quality, whereas the noise (N) represents undesired response variability due to uncontrolled factors or external disturbances.157 Depending on the desired outcome of the experiment, the S/N ratio can be classified as larger-the-better (LTB), smaller-the-better (STB), or nominal-the-best (NTB).158 For example, in a study on erythrocyte-coated NLC, coating factors affecting particle size and PDI were screened based on the STB S/N ratio, successfully yielding ultrasmall NLC with potential for glioblastoma therapy.159 The orthogonal arrays in the Taguchi design allowed for a wide range of experimental possibilities. A two-level design is commonly used to screen for multiple factors in fewer runs. For instance, Pant et al conducted a 7-factor, 2-level experiment with only 8 runs (L8) using the Taguchi design, instead of 128 (27), to identify the critical factors in the production of raloxifene-loaded NLC.129 To effectively perform screening using the Taguchi design, it is essential to carefully select experiments that maintain a statistical balance and provide an unbiased estimation of the effect of each factor on the responses.

Optimization Designs

While FFD are often used for screening, they can also be applied in optimization by incorporating three or more levels per factor. A three-level FFD (3k) introduces an additional intermediate level (coded as 0), allowing for a more precise estimation of the quadratic effects.147 This is particularly useful when researchers suspect that the relationships between factors and responses are nonlinear. However, similar to two-level FFD, the total number of runs increases exponentially, making this approach efficient only when working with a small number of factors. Notably, a two-level FFD may be applied in an optimization that employs fewer factors.160 Mendes et al constructed two types of designs to develop NLC containing atorvastatin calcium: 32 (nine runs) and 22 (four runs) FFD. An observation of the influence of surfactant concentration on the particle size revealed similar interactions in both designs.159

Central composite design (CCD) is one of the most extensively used response surface methodology (RSM) designs for optimization because it is effective for modeling curvature and optimizing nonlinear processes. The CCD consists of three main components: a full factorial or fractional factorial cube (2k), axial (star) points (2k), and center points. The total number of runs required is determined by the formula 2k + 2k + cp, where k is the number of factors, and cp is the number of center points. Thus, for a 3-factor optimization, a minimum of 15 experiments were required. The axial points (coded as +α and − α) may extend the design space beyond the factorial region, allowing observations at extreme values.141 Based on the selection of α values, CCD can be categorized as: a) circumscribed (CCCD): the axial points extend beyond the factorial space (Figure 4d); b) inscribed (CCID): the axial points remain within the factorial region (Figure 4e); and c) face-centered (CCFD): the axial points are positioned on the faces of the factorial cube (Figure 4f).161 The α value in the CCD varies between 1 and 2k/4, with the latter usually selected to maintain the design rotatability. For example, Ayed et al used a rotatable CCD to optimize two factors, the lipid and surfactant amounts, over 13 runs (including five center points) in the production of quetiapine fumarate-loaded NLC, with an α value of 1.414 (22/4).131 A similar design was employed in a 3-factor, 20-run optimization of NLC containing ifosfamide (including six center points), with an α value of 1.682 (23/4).91 It is also noteworthy that the α value in CCD, particularly in face-centered design, can be set at the same level as the low (−1) and high (+1) values of the factorial space, omitting observations at extreme points, as demonstrated in the study of diacerein-loaded SLN performed by Al-Remawi et al.86

The Box-Behnken design (BBD) is another widely used optimization design, particularly when the relationship between the factors and responses is expected to be nonlinear. Unlike CCD, which includes axial points that may extend beyond the factorial space, BBD distributes the experimental runs evenly at the midpoint of the factor pairs, eliminating the need to test for extremely high and low values (Figure 4c).146 CCD, particularly the circumscribed design, includes axial points that extend beyond the factorial region, which can be valuable for exploring a broader response surface but may introduce impractical conditions. On the other hand, face-centered CCD maintains experimental points within the factorial space but requires more runs because of more levels on account of axial points (+α and −α).153 The number of runs in a BBD is given by the formula 2k (k−1) + cp, where k represents the number of factors, and cp represents the number of center points. This means that to optimize three factors, a minimum of 13 experiments are required, whereas CCD requires at least 15 runs. This makes BBD a more efficient choice for optimization in fewer runs, while still maintaining a robust quadratic model. In the development of LNPs, BBD can be used as a follow-up to the screening stage. For instance, in a study of lurasidone HCl-loaded SLN fabricated by high-pressure homogenization, Patel et al screened seven factors using the Plackett-Burman design. Highly critical factors, namely lipid concentration, homogenization pressure, and homogenization cycle, were subsequently optimized to obtain the preferred particle size and entrapment efficiency using a 15-run BBD (including three center points).112 Nevertheless, numerous studies have reported the direct implementation of BBD for SLN or NLC optimization using 1 − 5 center points, resulting in 13 − 17 experiments.102,110,113,125

Design Space

In developing LNPs, identifying an optimal design space is essential to ensure a well-balanced formulation that meets predefined quality attributes because each factor can influence a response differently, sometimes even in a contradictory manner. For example, increasing the sonication time may reduce the particle size, but it can also lead to an undesirably low entrapment efficiency.68,103,104,124 Therefore, it is important to describe an optimal space that simultaneously achieves a balanced response for multiple factors. The selection of the most optimized formula is guided by the established DoE results, which help map the relationships between the critical factors and their respective responses. This process involves defining a design space that serves as a multidimensional region where the combination of input variables ensures the achievement of optimal responses.74 One of the most effective tools in this process is contour plot overlay, which enables the simultaneous evaluation of multiple responses by superimposing their individual contour plots.162 This graphical approach helps to identify an intersection where all quality attributes meet predefined requirements, thereby defining the most suitable formulation space. In addition, a desirability function is commonly used to simultaneously optimize multiple responses. This method transforms each interaction into a desirability scale ranging from 0 (least desirable) to 1 (most desirable), allowing the calculation of a composite desirability index that reflects the overall optimization outcome.163 By assigning specific weights to different responses based on their importance, this function aids in resolving conflicting optimization criteria, ensuring that the selected formulation maintains a balance among all critical parameters.

Optimization of Lipid Nanoparticles for Oral Drug Delivery

In QbD-based studies on the development of LNPs for oral administration, various independent variables (CMAs and/or CPPs) have been examined to elucidate their effects on the dependent variables (CQAs). The formation of LNPs relies on physicochemical principles such as lipid melting and recrystallization, emulsification, and colloidal stabilization. Upon cooling, the dispersed lipid phase solidifies into nanoparticles, while surfactants reduce the interfacial tension and provide steric or electrostatic stabilization to maintain colloidal stability.137 The nature and concentration of lipids, their crystallinity, and the compatibility with surfactants play key roles in controlling particle characteristics and drug incorporation. Process parameters, such as homogenization speed, pressure, and sonication time, directly influence the nucleation and growth of particles. Thus, the relationship between the independent variables and LNP formation mechanisms underpins their influence on CQAs. The most commonly evaluated CQAs during the optimization stage included particle size, polydispersity index, entrapment efficiency, drug loading, zeta potential, and drug release, as presented in Table 2.

Table 2 Design of Experiments in the Optimization of LNPs for Oral Drug Delivery

Influences of Independent Variables on Particle Size

Particle size is considered to be one of the principal CQAs in LNP development. SLNs and NLCs with smaller particle sizes inherently have larger surface areas, facilitating higher drug dissolution and absorption. A smaller particle size is also preferable because of the varied transport mechanisms during oral administration.25 Several variables significantly affect the size of the LNPs. In most cases, a higher amount of solid lipids leads to an increase in particle size. At higher solid lipid concentrations, the increased viscosity may resist oil droplet breakdown, resulting in a larger particle size.114 For example, Diwan et al demonstrated that, with other variables held constant, increasing the solid lipid amount from 50 mg to 300 mg led to a particle size increase from 279.2 nm to 837.6 nm during production of SLN containing cilnidipine.105 A similar trend is also observed in NLC optimization when the total lipid content or solid-to-liquid lipid (S/L) ratio is considered a CMA, particularly when the solid lipid concentration exceeds that of the liquid lipid, as evident in various reports.53,115,116,128 However, contradictory interactions have been observed in several studies. For instance, in the study of paliperidone-loaded NLC by Rehman et al, a three-fold increase in total lipid concentration (at an S/L ratio of 70:30) resulted in a particle size reduction from 487.9 nm to 332.6 nm.117 This finding aligns with the results of Pant et al in the development of raloxifene-loaded NLC, in which smaller particles were obtained at higher solid lipid levels.129

The use of higher liquid lipid concentrations has only been investigated in a few studies. Predominantly, an increased amount of liquid lipids is associated with a reduction in the particle size. For instance, in the optimization of eplerenone-loaded NLC, Abd-Elhakeem et al demonstrated that increasing the liquid-to-solid lipid (L/S) ratio from 1:1 to 2:1 markedly reduced the particle size.118 This reduction can be attributed to the low viscosity of the liquid lipid, which facilitates the rapid movement of surfactant molecules, effectively preventing aggregation, and promotin