Central composite design augmented quality-by-design-based systematic formulation of erlotinib hydrochloride-loaded chitosan-poly (lactic-co-glycolic acid) nanoparticles

Abstract

Aim: This study aimed to formulate erlotinib hydrochloride (ERT-HCL)-loaded chitosan (CS) and poly (lactic-co-glycolic acid) (PLGA) nanoparticles (NPs) using Quality-by-Design (QbD) to optimize critical quality attributes (CQAs). Materials & methods: Quality target product profile (QTPP) and CQAs were initially established. Based on L8-Taguchi screening and risk assessments, central composite design (CCD) design was used to optimize NPs. Results: ERT-HCL-loaded CS-PLGA NPs had a mean particle diameter, zeta potential and entrapment efficiency of 226.50 ± 1.62 d.nm, 27.66 ± 0.64 mV and 78.93 ± 1.94 %w/w, respectively. The NPs exhibited homogenous spherical morphology and sustained release for 72 h. Conclusion: Using systematic QbD approach, ERT-HCL was encapsulated in CS-PLGA NPs, optimizing CQAs. These findings propel future research for improved NSCLC treatment.

TWEETABLE ABSTRACT

Innovative erlotinib-loaded chitosan-PLGA nanoparticles, developed through a systematic QbD approach, promise enhanced drug delivery for NSCLC. Optimized for size, potential and entrapment efficiency, these particles demonstrate sustained release over 72 h. #DrugDelivery #QBD #NSCLC

GRAPHICAL ABSTRACT

Plain language summary

Summary points.

• The study presents a rational Quality-by-Design (QbD) methodology for the systematic development of ERT-HCL-loaded CS-PLGA NPs, improving the characteristics of drug delivery.

• The approach successfully identified the product-centric Quality Target Product Profile (QTPP), with the selected Critical Quality Attributes (CQAs) for this study being mean particle diameter, zeta potential and entrapment efficiency of ERT-HCL-loaded CS-PLGA NPs.

• Key independent variables (CMAs, FVs and PPs) affecting drug product CQAs were identified through rigorous initial risk assessments, comprising Ishikawa cause-effect diagram and methodologies like risk estimation matrix (REM), and failure mode effect analysis (FMEA) allowing for proactive risk mitigation.

• Quantitative FMEA analysis yielded seven independent variables with Risk Priority Score (RPN) of ≥25, which were used in factor screening studies.

• As per L8 orthogonal Taguchi design, out of seven factors, two factors namely drug: polymer ratio and CS concentration showed significant effect on the selected CQAs and would be later used in the DoE model.

• A 21 experimental run, 2-factor, and 5-level (-α, -1, 0, +1, +α), central composite design (CCD) was utilized where two independent variables namely drug: polymer ratio (X1) and CS concentration (X2) were optimized for their effects on the dependent variables; mean particle diameter (Y1), zeta potential (Y2) and entrapment efficiency (Y3) of the developed ERT-HCL-loaded CS-PLGA NPs.

• As per the model desirability factor, a final optimized batch was obtained with a mean particle diameter of 226.50 ± 1.62 nm, a zeta potential of 27.66 ± 0.64 mV, and an entrapment efficiency of 78.93 ± 1.94%.

• The NPs had a drug loading capacity of 4.33% w/w along with a satisfactory PDI value of less than 0.3 and exhibited a uniform spherical morphology.

• The in vitro release studies showed sustained ERT-HCL release up to 72 h, attributed to CS mitigating initial burst release, kinetic modeling affirmed a zero-order drug release and Fickian drug release behavior.

• The study lays the groundwork for further studies on ERT-HCL-loaded CS-PLGA NPs, which show promise in treating non-small-cell lung cancer and offer a solid foundation for therapeutic validation.

1. Introduction

Lung cancer along with breast and colorectal cancer is the leading cause of mortality in both males and females on a global scale [1]. As of 2023, there are an estimated 117,550 male and 120,790 female new cases with 127,070 fatalities attributed to lung cancer [1,2]. There are two types of lung cancer: non-small-cell lung cancer (NSCLC) and small-cell lung cancer (SCLC) [3]. NSCLC accounts for ∼87% of lung cancer cases [4]. Patients with NSCLC have a pessimistic prognosis, with just 15% of those 60 and elderly surviving for more than 5 years along with a significant recurrence rate [5,6]. Radiation treatment and chemotherapy represent the current strategies for treating NSCLC, yet, both possess drawbacks such as toxicity to healthy cells, substantial systemic side effects and poor selectivity in distinguishing cancerous from non-cancerous cells, leading to unintended harm to normal cells or tissues. Furthermore, these treatments lack targeted action, meaning they do not specifically target cancer cells or tumor sites, missing the opportunity for precise therapeutic intervention [7].

Amidst the therapeutic challenges, erlotinib hydrochloride (ERT-HCL), a specific and potent tyrosine kinase inhibitor (TKI), is an efficacious treatment drug for NSCLC. It binds toward the nonreceptor tyrosine kinase of EGFR at ATP binding sites with an affinity. It thereby inhibits its autophosphorylation, resulting in the induction of the apoptosis pathway and further leading to the inhibition of cellular growth and angiogenesis [8]. The aqueous solubility of ERT-HCL is pH-dependent, exhibiting very low solubility at neutral pH but increased solubility at pH levels below 5 [9]. To address this, ERT-HCL, marketed under the commercial brand name Tarceva^®, was developed. Its maximal solubility is approximately 0.4 mg/ml at a pH of around 2 [9]. Classified as a Biopharmaceutical Classification System (BCS) class II drug, ERT-HCL demonstrates low solubility and high permeability characteristics. Tarceva is typically administered as a once-a-day film-coated tablet, with strengths ranging from 25 to 125 mg, and should be taken continuously until disease progression or unacceptable toxicity occurs [10]. The oral absorption of Tarceva is approximately 60% [11] While food can enhance the bioavailability (BA) of Tarceva up to 100%, the extent of absorption improvement varies depending on dietary factors. As a general guideline, Tarceva is ideally taken on an empty stomach, either at least 1 h before or 2 h after a meal. The drug's system-wide distribution also diminishes the route of administration's capacity to specifically target cancer cells [12]. Further, ERT-HCL also exhibits several dose-related side effects including skin rashes, diarrhea, fever, chills and intestinal disorders [13]. Last, there also has been a rising concern with the use of ERT-HCL and other TKIs such as afatinib and gefitinib which is indeed the progression of acquired resistance toward the treatment, which typically occurs within a year of treatment initiation [14]. Because the overwhelming majority of treatments for NSCLC that have been authorized are plagued by off-target effects and acquired resistance, treating NSCLC with innovative and novel nano-delivery systems has gained the interest of researchers over the last few decades [2,15]. Owing to their highly efficient and versatile drug delivery characteristics, nanomaterial methods have received significant attention in recent years, as shown by the authorization of more than 100 products based on nanotechnology and nanomedicines [16,17]. Entrapping a chemotherapeutic drug in a nanoparticulate delivery system provides several benefits, including preservation from in-vivo degradation, improved solubilization for hydrophobic drugs, allowing interaction with cells or biomolecules, targeted therapy, reduced systemic side effects, increased drug length of exposure, decreased drug resistance, etc. [18–20].

Synthetic polymers including polyethylene glycol (PEG), polylactide, poly (lactic-co-glycolic acid) (PLGA), poly (caprolactone) (PCL) and natural polymers, for instance, collagen, chitin, keratin, elastin, chitosan (CS), albumin, gelatin and alginate are the most utilized polymers for the synthesis of polymeric nanoparticles (PNPs) [21,22]. PLGA is a US FDA-approved and widely regarded as the most prominent polymer for the production of PNPs for cancer therapy [23,24]. This could be attributed to its high biocompatibility and well-known biodegradability in the human body [25]. Its adaptable structural and chemical characteristics make it an outstanding choice among other PNP polymers for entrapping active chemical moieties and for targeting NSCLC [26–28]. The benefits of PLGA nanoparticles (NPs) are undeniable, yet the material is not without significant drawbacks. This involves nonspecific cellular interaction with the targeted tumor tissue region, accumulation of these NPs in off-target neighboring regions and poor cellular adhesion and penetration via the cell surface as a result of a large negative surface charge [29,30]. Reports of unintentional burst release of PLGA NP formulations, which could reduce treatment efficacy or even cause harm to the host, have also arisen [31,32]. Last, PLGA NPs have limited functional groups available on the surface for conjugation [33,34]. It has been proven that modifying the PLGA surface with specific ligands will improve cellular drug delivery with modified release properties. These restrictions can be overcome by conjugating or coating the fundamental PLGA NPs with CS [35,36].

CS is a cationic naturally biodegradable and biocompatible polymer [37,38]. The CS's anti-cancer effect has been demonstrated in vitro and in vivo, suppressing cancer cell growth through complex apoptotic pathways [39]. Furthermore, CS's hydroxyl (-OH) and amino (-NH2) groups allow it to form hydrogen and covalent bonds with other moieties such as PLGA's carboxylic group (-COOH) or other targeting ligands [40,41]. Protonation of amine groups of CS with the help of dilute acid such as acetic acid helps in imparting a positive charge to the CS surface [42]. In our study, CS was combined with PLGA NPs that will prevent the initial unintentional burst release [36] and also allow for opportunistic conjugation of targeting ligands such as (folic acid, hyaluronic acid, peptides, etc.) through electrostatic interactions with protonated amine groups of CS [43].

Several formulation and process variables, such as the choice and concentration of polymer and stabilizer, preparation method, size reduction process, stirring speed and duration, etc., are involved in developing a nanoparticulate drug delivery system (DDS) [44]. The formulation procedures of such DDS can be optimized using traditional one-factor-at-a-time (OFAT) methods, but these methods are time consuming, costly and only yield partially optimal results [45,46]. Many people seem to have different interpretations of the famous quality statement by Joseph M. Juran: “Quality cannot be tested; it needs to be developed into a product”.

Quality-by-Design (QbD) is the panacea that the pharmaceutical sector has been striving for to produce the most desired product with desirable attributes while also easing legal burdens, thanks to the extensive input of numerous ICH guidelines [47]. Developing the ‘best’ preparation for a given circumstance with as little trial and error and input as possible, a QbD-based systemic approach has recently been practiced to provide a more in-depth insight into the formulation system by determining all the probable interaction(s) among product and/or process-related variables. QbD's component, the Design-of-Experiments (DoE), offers a statistical justification for the formulation development-stage parameters. Whenever you employ the DoE, you may use screening and risk assessment procedures to identify the critical material attributes (CMAs), critical formulation variables (CFVs) and critical process parameters (CPPs) that will help you achieve the specified critical quality attributes (CQAs) while staying within the design space [48].

To our knowledge, there exists no prior research specifically employing systematic QbD principles for developing a robust method and optimizing ERT-HCL-loaded CS-PLGA NPs, particularly utilizing preliminary risk assessment to identify potential risks, followed by a Failure mode effect analysis (FMEA) analysis to identify critical factors. The Taguchi-based L8-orthogonal screening model was further exploited to identify independent variables to be used in DoE which led to the establishment of a robust suitable model with observed variance for the development of the formulation with desired CQAs.

2. Materials & methods

2.1. Materials

Erlotinib HCl was kindly provided by MSN Laboratories Pvt. Ltd (Hyderabad, India) as a gift sample. Acid-terminated Poly (lactic-co-glycolic acid) (PLGA, 75:25, 4–15 kDa, Resomer 752H) was provided as a gift sample from Evonik Pvt Ltd (Mumbai, India). Low molecular weight chitosan (CS); (Mw: 50,000–190,000 Da, viscosity: 20–300 cP) and poloxamer 407 were purchased from Sigma Aldrich Pvt Ltd (Mumbai, India). Dichloro methane (DCM), dimethyl formamide (DMF) and methanol were analytical grade (AR) and purchased from Research Lab Fine Chemicals (Mumbai, India).

2.2. Preparation of ERT-HCL-loaded CS-PLGA NPs

ERT-HCL-loaded CS-PLGA NPs were prepared using the single-solvent evaporation methodology already reported in the literature with slight modifications [49,50]. An accurately weighed amount of ERT-HCL (10 mg) and PLGA (100 mg) was dissolved in dichloro methane (DCM) to form the organic polymeric solution. Separately, the required amount of poloxamer 407 (2% w/v) and CS (0.1% w/v) was dissolved in 1% glacial acetic acid solution to obtain the aqueous phase. The prepared organic phase was then emulsified with the aqueous phase by adding the organic phase dropwise under continuous magnetic stirring (2MLH, REMI, India). The resulting dispersion was then subjected to probe sonication (2-s on-off cycle at 40% amplitude, Sonics VCX-130, Sonics & Materials, Inc, USA) over an ice bath. The organic solvent was evaporated from the produced NPs by stirring them at ambient temperature overnight. ERT-HCL-loaded NPs were separated from the aqueous medium by ultracentrifugation (Optima MAX-XP ultracentrifuge with TLS-55 rotor, Beckman Coulter, USA) at 15,000 rpm for 25 minutes (min) at 4°C. The sediment was washed three to four-times with Milli-Q water to remove the bound drug and excess stabilizer. The concentrated NPs were redispersed in 1–2 ml of Milli-Q water, lyophilized and preserved in glass vials for further analysis.

2.3. Defining the QTPPs & CQAs

Identifying QTPP and CQA parameters is the crucial initial step in the QbD-based strategy toward a systematic formulation development of CS-PLGA NPs of ERT-HCL. A product-centric QTPP was initially established. QTPP can be defined as an “overview describing every one of the quality attributes of a pharmaceutical product that must be controlled to ensure the targeted product quality, safety and efficacy”. The framework for QTPPs would be developed by reviewing the literature, comprehending the pragmatic and scientific aspects of formulation, the intended results of the formulation and taking legal and risk considerations into account [51,52]. Various quality characteristics (QAs) were identified to fulfill the QTPP such as mean particle diameter, zeta potential, entrapment efficiency and in vitro drug release, from which the major key CQAs would be selected [53].

2.4. Initial risk assessment

For the CS-PLGA NPs of ERT-HCL, an initial risk assessment was conducted to ascertain the CPPs, CFVs and CMAs. The effect of process variables and quality characteristics on identified CQAs was examined [54]. Using Minitab 17 software (Minitab Inc, PA, USA), an Ishikawa cause-effect fishbone diagram was generated that illustrates a schematic cause-and-effect connection that ties the various MAs, FVs and PPs to the detected CQAs. Further, methodologies comprising risk estimation matrix (REM) and FMEA were employed for the qualitative and quantitative analysis to prioritize the high-risk CMAs, CFVs and CPPs, respectively. A REM was constructed in which the effects of different MAs, FVs and PPs were ranked as high, medium, or low risk (s) and wherein significantly higher risk(s) were considered CMAs, CFVs and CPPs. FMEA scoring was used to investigate and determine the risk of the numerous elements indicated in the Ishikawa fishbone diagram and REM using RPN scoring [55]. Every factor was scored on severity, occurrence and detectability. The range of each factor's scale was 1 to 5. On this scale, 1 indicates a low risk, while 5 indicates a substantial risk. The risk priority number (RPN) was obtained by multiplying the three parameters, occurrence (O), severity (S) and detectability (D) [54]. A threshold value of the RPN (i.e., ≥25) was decided for differentiating the high-risk factors against the low-risk factors. The factors linked with high RPN (≥25) would be further analyzed using the Taguchi model in the factor screening studies.

2.5. Factor screening studies by L8-Orthogonal Taguchi model

An L8 orthogonal Taguchi design was created using Design Expert (Design Expert 14, Stat-Ease, MN, USA) for the factor screening studies to identify the effects of different factors on the identified CQAs of CS-PLGA NPs of ERT-HCL. The factors deduced from RPN scoring were used to design the Taguchi model matrix. The factors were tested at low and high levels for a total of eight experimental runs and were evaluated for mean particle diameter (d.nm), zeta potential (mV) and entrapment efficiency (% w/w) (Supplementary Table S1). The model was adjusted using linear polynomial equations. Subsequently, the coefficients associated with each element pertaining to the CQAs of the product were assessed, following the elimination of any interaction term(s). Last, the Pareto chart constructed by the design expert software was analyzed for the three responses to understand the contribution and effect of each MAs, FVs and PPs.

2.6. Systematic formulation optimization of ERT-HCL-loaded CS-PLGA NPs by DoE

After the successful identification of independent variables, the DoE-based approach that further involves response surface methodology (RSM) was employed for the systematic formulation of CS-PLGA NPs of ERT-HCL with the expected responses within the required period [56]. A 2-factor and 5-level CCD design was employed to investigate the optimal values for the variables and explore the design space. The factors deduced from the Taguchi screening model would be selected as the main independent variables (X) in the CCD model and would be optimized for their effects on the mean particle diameter (Y1), zeta potential (Y2) and entrapment efficiency (Y3) of the developed ERT-HCL-loaded CS-PLGA NPs. A total of 21 experimental runs were performed with 4 factorial points, 8 axial points and 9 replication center points for a better estimation of curvature and to provide a more robust estimate of model parameters (Supplementary Table S2). The polynomial equation was generated using the software, which permitted the appropriate evaluation of linear, quadratic and interactive effects of the independent variables (X) on the responses (Y) [51,57]. ANOVA analysis was used to determine the effect of independent variables on the responses, where a p-value less than 0.05 was considered statistically significant.

2.7. Characterization & evaluation of ERT-HCL-loaded CS-PLGA NPs

2.7.1. ATR-FTIR analysis

The Fourier Transform Infrared (FTIR) spectra were obtained using a UATR Two Perkin Elmer (Perkin-Elmer, Inc, USA) equipped with attenuated transmittance reflectance (ATR). The Individual FTIR spectra were obtained for pure ERT-HCL, PLGA, CS and ERT-HCL-loaded CS-PLGA NPs and compared respectively. After cleaning the diamond surface of the ATR-FTIR instrument with carbon tetra chloride (CCl₄), 5 mg of the sample was placed and scanned by increasing the torque up to 70%. The automatic baseline corrections were made by the software spectrum [58]. The examined IR spectral wavelength ranged from 4000 to 400 cm^-1 with a 4 cm^-1 resolution [58,59].

2.7.2. DSC analysis

Differential scanning calorimetry (DSC) tests were performed on ERT-HCL, PLGA, CS and ERT-HCL-loaded CS-PLGA NPs using a DSC (Mettler Toledo, DSC 1 STAR, Switzerland). Individual samples were filled and crimped in 2 to 5 mg aluminum crucibles. The temperature ranged from 30 to 350°C, with a rising heating rate of 10°C/min [59].

2.7.3. Surface morphology

The blank PLGA NPs, ERT-HCL-loaded PLGA NPs and ERT-HCL-loaded CS-PLGA NPs were imaged using a Transmission electron microscope (TEM) (TEC-12, TECNAI G2 SPIRIT BIOTWIN) operating at 100 kV; utilizing protocols and methodology already available in the literature with some slight modifications [60]. Briefly, 50 μl of diluted NP samples were affixed on the carbon-coated copper grids obtained from Electron Microscopy Sciences, USA, and were allowed to adsorb to the copper grids for 2 min followed by drying of the sample with an infrared lamp for 4–5 min. The samples were next negatively stained for 20 s using 5 μl of phosphotungstic acid. The excessive phosphotungstic acid stain was patted down with filter paper. The staining procedure was repeated one more time and finally, the samples were air-dried for 2 min. The samples on the grids were then carefully placed in the TEM instrument for imaging and analysis. The samples were imaged using Olympus soft imaging solutions VELETA CCD Camera and analyzed using Tecnai imaging & Analysis software.

2.7.4. Mean particle diameter & Zeta potential

To determine the mean particle diameter and zeta potential of the CS-PLGA NPs of ERT-HCL, an aqueous dispersion of accurately weighed NPs was prepared, poured into an appropriate cuvette, and scanned using a Nano-ZS zeta sizer (Malvern Instruments, UK). The d value (90th percentile particle diameter) was used to describe the mean particle diameter (d.nm). The homogeneity and spread of the NPs in the samples were analyzed using the polydispersity index (PDI) value. Mean values from three independent samples (n = 3) obtained under the same experimental circumstances were reported.

2.7.5. Entrapment efficiency & drug loading capacity

The proportion of ERT-HCL content in the CS-PLGA NPs was measured by first preparing an aqueous dispersion of NP formulation and centrifuging at 40,000 rpm for 45 min at 4°C (Optima MAX-XP ultracentrifuge with TLS-55 rotor, Beckman Coulter, USA) [61]. The resulting supernatants were carefully withdrawn and diluted suitably with the mobile phase. The diluted supernatants were filtered through a 0.45 μm nylon filter. The quantification of free ERT-HCL in the supernatant was carried out using a UV-visible spectrophotometer (Shimadzu-1800) configured to operate at a wavelength of 247.67 nm. The drug loading capacity of the NPs was calculated by subtracting the amount of free ERT-HCL in the supernatant from the total amount of ERT-HCL used in the formulation and expressing it as a percentage relative to the total mass of the NPs.

2.7.6. In vitro drug-release studies & kinetic modeling

The in vitro release of ERT-HCL from CS-PLGA NPs was assessed using the dialysis bag method. Freeze-dried NPs equivalent to 10 mg of ERT-HCL were dispersed in 1 ml of Milli Q water. Subsequently, the NP dispersion was carefully transferred into a dialysis bag with a molecular weight cutoff (MWCO) of 12000, which had been previously soaked in Milli Q water for approximately 24 h. The dialysis bag was meticulously sealed at both ends using thread to ensure a secure enclosure. The sealed dialysis bag was then submerged in 50 ml of release media consisting of phosphate buffer saline (PBS) with a pH of 7.4. The setup was maintained at a temperature of 37 ± 0.5°C with continuous magnetic stirring at 100 rpm. At specific, predetermined time intervals (0.25, 0.5, 1, 2, 4, 6, 8, 12, 24, 36, 48 and 72 h), 1 ml of the sample was withdrawn from the dialysis bag. To ensure that sink conditions were maintained, an equal volume of fresh PBS was added to replace the withdrawn sample. The withdrawn samples were subjected to analysis to determine the content of ERT-HCL. This analysis was conducted using UV-Spectroscopy, with measurements taken at a fixed wavelength of 247.67 d.nm. The entire drug release study was conducted in triplicate (n = 3) to ensure the reliability and consistency of the results. The data obtained from these experiments were reported in terms of percent cumulative drug release (% CDR).

The in vitro drug-release data was fitted into various kinetic models such as zero order, first order, Higuchi and Korsmeyer Peppas to understand the release mechanism of ERT-HCL from the polymer system. The maximum correlation coefficient value (r²) signifies the mechanism of drug release. The Korsmeyer Peppas model was further utilized to distinguish between different release mechanisms, including Fickian release (diffusion-controlled release), non-Fickian release (anomalous transport) and case-II transport (relaxation-controlled release) [61].

2.8. Statistical analysis

All investigations were conducted in triplicates (n = 3), and results were provided as mean ± standard deviation (SD) unless stated otherwise. To statistically assess the data, an analysis of variance (ANOVA) using a one-way design was performed. For p-values <0.05, differences between two related parameters were deemed statistically significant.

3. Results

3.1. Preliminary batch trials

Preliminary lab experiments were carried out to refine the formulation conditions and assess the physicochemical characteristics of the NPs to adjust and modify parameters to get the desired quality attributes. Different formulation methods include emulsification–solvent evaporation, emulsification–solvent diffusion and nanoprecipitation [62–64]. The emulsion solvent evaporation method was used as it yielded maximum drug entrapment with control over the mean particle diameter. The choice of the organic solvent was deemed less critical thus, DCM was used as it yielded good results in the preliminary batch formulation. The changes in probe sonication amplitude were found to have a negligible impact on mean particle diameter and zeta potential, therefore this parameter was held constant at 30% amplitude. The probe sonication pulse was adjusted to 2 s on and off over an ice bath to prevent overheating of the samples.

3.2. Defining the QTPPs & CQAs

The QTPPs elements defined for ERT-HCL loaded CS-PLGA NPs were CS-PLGA NPs were identified as desired dosage/delivery type system as the unique combination of these two biocompatible and biodegradable polymers exhibit an array of favorable drug delivery characteristics. The pulmonary route was selected as the route of administration since it increases drug accumulation at the local site, extends residence time, protects against degradation and reduces off-target systemic toxicities. Low mean-particle diameter, zeta potential in the range of -20 to -30 mV, and high entrapment efficiency were desired. Long-term stability over several months to years at room temperature was also desired from the formulation so that they can sustain for a suitable period. The next step in the QBD approach involves identifying the key CQAs to ensure the attainment of the desired QTPP. Mean particle diameter (100–300 d.nm), zeta potential (-20 to -30 mV) and entrapment efficiency (more than >65% w/w) were identified as the main CQAs.

3.3. Initial risk assessment

The Ishikawa fishbone diagram illustrated in Figure 1 highlights the main cause-and-effect connection among the many parameters influencing the identified CQAs of the drug product. The selection of raw materials, material properties, process parameters, formulation variables and milieu factors all have a significant impact on the CQAs. In the next step, REM was constructed to prioritize the risk associated with the different MAs, FVs and PPs. Table 1 highlights the major high-risk factors in red which were taken as CMAs, CFVs and CPPs. The high-risk factors were the drug-polymer ratio, the concentration of CS and stabilizer, the organic: aqueous phase ratio, the rate of organic phase addition, probe sonication time and solvent evaporation rate. The organic solvent type and formulation stirring speed, etc. were medium/low risk. After the REM analysis had been carried out, a more structured and quantitative risk assessment was performed. Table 2 summarizes the FMEA analysis and shows the RPN score deduced for all the factors. High RPN scores (≥25) were observed for a drug: polymer ratio, CS concentration, stabilizer concentration, the ratio of organic to aqueous phase, rate of addition of organic to aqueous phase and solvent evaporation rate respectively.

Figure 1.

Ishikawa-Fish bone diagram depicting the different process conditions and material attributes. This thorough depiction directs the creation of successful nanoparticle (NP) optimization techniques and assists in the identification of possible factors that may influence critical quality attributes (CQAs).

Table 1.

Initial risk assessment of ERT-HCL-loaded CS-PLGA NPs, highlighting potential risks associated with the material attributes, formulation variables and process parameters.

CS: Chitosan; ERT-HCL: Erlotinib hydrochloride; FV: Formulation variables; MA: Material attribute; NP: Nanoparticle; PLGA: Poly (lactic-co-glycolic acid); PP: Process parameter.

Table 2.

The FMEA analysis representing the RPN scores, calculated based on occurrence, severity and detectability, providing a comprehensive evaluation of potential risks associated with the formulation of ERT-HCL-loaded CS-PLGA NPs.

Risk factor category		Occurrence (O)	Severity (S)	Detection (D)	Risk priority number (RPN = O × S × D)
Milieu
	Temperature	2	1	3	6
	Humidity	1	1	2	2
	Oxidation	1	2	1	2
	Personnel	3	1	1	3
Material attributes
Drug	Molecular weight	1	3	3	9
	Log p	1	2	3	6
	Pka	1	2	2	4
	Solubility in the organic phase	2	2	3	12
Polymer (PLGA)	Grade	1	2	3	6
	Mol. weight	1	3	3	9
	LA/GA ratio	1	2	2	4
	Viscosity	2	2	2	8
	Functional groups	2	2	2	8
	Solubility in the organic phase	2	3	2	12
Chitosan	Mol. weight	1	3	3	9
	Viscosity	2	2	4	16
	Deacetylation value	2	2	2	8
Stabilizer (Poloxamer 407)	Type	1	1	5	5
	Viscosity	2	1	4	8
	HLB value	2	1	4	8
	Solubility in the outer aqueous phase	4	1	3	12
Formulation variables
	Drug: polymer ratio	3	4	5	60
	Chitosan conc.	3	3	5	45
	Stabilizer conc.	3	2	5	30
	The ratio of organic to aqueous phase	2	3	5	30
Process parameters
	Preparation method	1	1	5	5
	Stirring temperature	2	2	4	16
	Stirring speed	2	2	5	20
	Probe sonication time	2	4	5	40
	Rate of addition of organic to aqueous phase	2	4	4	32
	Solvent evaporation rate	2	4	3	36
	Solvent evaporation temperature	3	2	3	18

CS: Chitosan; ERT-HCL: Erlotinib hydrochloride; FMEA: Failure mode effect analysis; NP: Nanoparticle; PLGA: Poly (lactic-co-glycolic acid); RPN: Risk priority number.

3.4. Factor screening studies by the L8-Orthogonal Taguchi model

Figure 2A showed a significant influence (i.e., a value higher than the t-value set of the Pareto chart) of the drug: polymer ratio factor for the mean particle diameter of the CS-PLGA NPs of ERT-HCL. The ratio of organic to aqueous phase although showed a positive result or contribution, was not considered significant for the selected CQA, and thus was set at a constant value. A significant effect of the CS concentration was observed on the zeta potential of the CS-PLGA NPs of ERT-HCL as it crossed the t-value limit of the Pareto chart (Figure 2B). Last, for the % entrapment efficiency response, a significant effect of the drug: polymer ratio was observed (Figure 2C). In conclusion, two factors, in other words, drug: polymer ratio and the CS concentration were selected as the main independent variables (X) that would be later used in the DoE model for the systematic formulation of the NPs.

Figure 2.

Pareto chart for screening of factors for the selected critical quality attributes (CQAs). Mean particle diameter (d.nm), zeta potential (mV) and entrapment efficiency (% w/w).

3.5. Systematic formulation optimization of CS-PLGA NPs of ERT-HCL by DoE

A design space was created and the two independent variables (X) were evaluated at five different levels and optimized for their effects on the mean particle diameter (Y1), zeta potential (Y2) and entrapment efficiency (Y3) of the developed ERT-HCL-loaded CS-PLGA NPs (Table 3).

Table 3.

The CCD design space illustrates the obtained responses Y1, Y2 and Y3 (mean ± SD, n = 3). This design space demonstrates the optimized conditions for ERT-HCL-loaded CS-PLGA NP formulation, leading to the desired NP characteristics.

Variables	X1	X2	Y1	Y2	Y3
	Drug: polymer ratio	Chitosan concentration	Mean particle diameter	Zeta potential	Entrapment efficiency
Units	ratio	% w/v	d.nm	mV	% w/w
1	0	0	221.42 ± 1.32	26.5 ± 1.19	76.39 ± 2.07
2	0	0	223.42 ± 1.98	25.6 ± 0.72	75.19 ± 2.33
3	0	-1.41421	209.54 ± 0.74	14.2 ± 2.24	65.72 ± 3.46
4	1.41421	0	254.38 ± 1.19	28.2 ± 0.65	86.72 ± 2.54
5	1	1	242.63 ± 2.26	27.6 ± 1.32	83.81 ± 1.68
6	-1.41421	0	207.34 ± 0.92	20.6 ± 1.80	48.83 ± 3.59
7	1	1	240.41 ± 2.81	28.4 ± 1.31	82.71 ± 1.40
8	-1	1	210.44 ± 2.07	25.9 ± 1.10	68.91 ± 2.92
9	0	0	220.11 ± 3.40	25.7 ± 1.06	76.92 ± 1.74
10	-1	-1	210.43 ± 3.92	19.5 ± 2.02	66.48 ± 3.60
11	0	1.41421	222.51 ± 1.23	30.4 ± 2.16	78.99 ± 1.61
12	1.41421	0	257.64 ± 2.45	29.7 ± 1.48	87.31 ± 1.25
13	-1	1	215.76 ± 1.38	26.3 ± 1.73	68.79 ± 4.01
14	0	0	224.24 ± 3.51	25.2 ± 0.50	77.71 ± 3.74
15	1	-1	237.42 ± 3.08	20.3 ± 2.24	66.29 ± 2.88
16	-1.41421	0	209.63 ± 1.17	19.2 ± 1.38	54.82 ± 4.47
17	-1	-1	217.52 ± 1.38	19.7 ± 1.27	66.64 ± 2.63
18	0	-1.41421	207.43 ± 2.70	17.1 ± 2.11	64.14 ± 3.14
19	0	1.41421	224.71 ± 1.62	29.8 ± 0.61	80.93 ± 1.36
20	1	-1	239.53 ± 1.44	18.2 ± 1.98	65.72 ± 1.40
21	0	0	220.32 ± 1.60	25.3 ± 1.26	78.72 ± 1.92

CCD: Central composite design; CS: Chitosan; ERT-HCL: Erlotinib hydrochloride; NP: Nanoparticle; PLGA: Poly (lactic-co-glycolic acid).

3.5.1. Response (Y1): mean particle diameter

Following the experimental runs using the CCD design, it was observed that the lowest particle diameter of the CS-PLGA NPs was achieved in ranges 207–220 d.nm (Table 3). The ANOVA analysis and regression analysis of responses for mean particle diameter (Y1) are summarized in Supplementary Table S3. The model showed an F-value of 53.84, and a p-value of <0.0001 suggesting that the model was significant. The model's significance was further demonstrated by the nonsignificant ‘lack of fit’ p-value of 0.0016 and ‘lack of fit’ F-value of 9.73. The adjusted and predicted R2 values differed by 0.0398 with a good correlation of determination (R2) value of 0.9472 [65,66]. Factor X1 (drug: polymer ratio) had a high F-value of 224.43 and a p-value of <0.0001. Factor X2 (CS concentration) was also a significant factor having a p-value of 0.0102. The model demonstrated the significance of the two independent pure variables, X1 and X2, as well as one quadratic variable, X1². As a result, the polynomial Equation (1) for the (Y1) response, or mean particle diameter, can incorporate these variables.

$$Y1=221.902+15.0163 X1+2.945 X2+5.82775 X{1}^2$$ (1)

In the equation, the positive coefficients of variables X1 and X2 indicate that particle diameter significantly increases with p-value < 0.0001, and 0.0102 respectively with increasing levels of X1 and X2 variables. Similarly, the positive coefficient of the quadratic variable X12 indicates that the mean particle diameter increases (p of 0.0003) with an increase in the drug: polymer ratio. The 3-dimensional (3D) surface plot considering X1 and X2 variables for the mean particle diameter response is depicted in Figure 3A.

Figure 3.

3D surface plot for response. (A) Y1: Mean particle diameter (d.nm), (B) Y2: Zeta potential (mV), (C) Y3: Entrapment efficiency (% w/w).

3.5.2. Response (Y2): zeta potential

The ANOVA analysis and regression analysis of responses for zeta potential (Y2) are summarized in Supplementary Table S4. The model's significance was demonstrated by its F-value of 24.14 and p-value of less than 0.0001. The nonsignificant ‘lack of fit’ p-value was 0.0002 and the ‘lack of fit’ F-value was 13.01. The difference between the adjusted and predicted R2 values was 0.0852 with a good correlation of determination (R2) value of 0.8895 [65]. Based on the generated model, p-values of 0.0013, <0.0001 and 0.0181 were found for the pure variables X1 and X2, and one of the quadratic variables, X2², respectively, indicating their significance. The quadratic variable X2² was significant, suggesting that the zeta potential response is influenced by the CS concentration and this relationship may not be linear, but instead quadratic. The final polynomial equation for the (Y2) response – zeta potential is shown in Equation (2).

$$Y2=25.66+1.79358 X1+4.46067 X2+-1.49562 X{2}^2$$ (2)

The positive coefficients of variables X1 and X2 indicate that the zeta potential of the NPs significantly increases with increasing levels of X1 and X2 variables. The quadratic variable has a negative coefficient which is due to the relationship between CS concentration and zeta potential being concave downwards which was in accordance with 3D surface plot analysis. In Lehman's words, this indicates that the zeta potential initially increases as the CS concentration rises, but reaches an optimum at a certain CS concentration level. After this threshold, the zeta potential begins to decrease as CS concentration increases further. The 3D surface plot for the (Y2) response – zeta potential is depicted in Figure 3B.

3.5.3. Response (Y3): entrapment efficiency

The ANOVA and regression analysis of responses for entrapment efficiency (Y3) are summarized in Supplementary Table S5. Regression analysis of the model revealed an R2 value of 0.8276 and less than 0.2 difference between the adjusted and predicted R2 value indicating less deviation from the mean value, indicating that the model was significant with an F-value of 14.40 and a p-value of less than 0.0001. As per the model, both the pure variables X1 and X2 along with interaction variable X1X2, and one quadratic variable X1² were found significant with p-values of <0.0001, 0.0008 and 0.0261 respectively. Thus, the equation for the % entrapment efficiency response can be written as per Equation (3).

$$Y3=79.986+7.95265 X1+5.10008 X2+3.74125 X1X2+-3.72394X{1}^2$$ (3)

The positive coefficients of the two pure variables X1 and X2 signify that entrapment efficiency increases with increasing levels of both variables respectively. As per the model, the interaction variable was also significant, suggesting that the effect of changing one variable on the entrapment efficiency of NPs depends on the level of the other variable. Also, the combination of the drug: polymer ratio and CS concentration can have a synergistic effect on the entrapment efficiency of the NPs. The model revealed that the entrapment efficiency response is not linear, but rather quadratic with respect to the drug-polymer ratio. Potential contributors to this correlation include similarities in the molecular structures of the drug and polymer, the drug's solubility in the polymer, and the preparation technique employed to make the drug-polymer combination. The 3D surface plot for the entrapment efficiency response is depicted in Figure 3C.

3.5.4. Overlay & desirability of response surface plot of Y1, Y2, & Y3 for optimization

Surface diagram plots derived for responses Y1, Y2 and Y3 were superimposed to produce a graphical optimization overlay plot to identify the optimal region and prediction of the optimized batch [51,67]. For obtaining the desired region, constraints were selected that were to minimize mean particle diameter (range 200–250 d.nm), maximize zeta potential (-20 to 30 mV), and maximize entrapment efficiency (75–90% w/w). The values of all the independent variables were adjusted to lie within the whole range (-1 to +1) [51,68]. The entrapment efficiency response (Y3) was given more importance (Importance +++++) than the other two responses (Importance +++). Figure 4 depicts the overlay plot of the responses with the set constraints. After defining the desired region and recognizing the limits that can be used, desirability-based numerical predictions were made to obtain the optimized batch. Narrower constraints were further added to refine the optimization process and the uncoded values of variables were obtained as summarized in Supplementary Table S6. The software predicted a batch with a desirability of 0.767 with predicted responses of the mean particle diameter of 225.01 d.nm, a zeta potential of 28.95 mV, and an entrapment efficiency of 81.49% w/w (Supplementary Figure S1). As per the software predictions, batches were formulated in triplicates, and the obtained results were mentioned as mean ± standard deviation. A final optimized batch was hence obtained having a mean particle diameter of 226.50 ± 1.62 d.nm, zeta potential of 27.66 ± 0.64 mV, and entrapment efficiency of 78.93 ± 1.94% w/w. The results obtained concord with the predicted results as per the software. Hence the model developed is robust and the variability has been predicted to a large extent. The drug loading capacity observed was 4.33% w/w. All formulation batches achieved a PDI of less than 0.3. The optimized batch as per DoE was further characterized and evaluated.

Figure 4.

Overlay plot for the three responses. Y1: Mean particle diameter (d.nm), Y2: Zeta potential (mV), Y3: Entrapment efficiency (% w/w) with the set constraints. The overlay plot displays the region of interest, which is marked in yellow and can be used in the prediction of the optimal batch.

3.6. Characterization & evaluation of ERT-HCL-loaded CS-PLGA NPs

3.6.1. ATR-FTIR analysis

The ATR-FTIR spectra of ERT-HCL, PLGA, CS and ERT-HCL-loaded CS-PLGA NPs are depicted in Figure 5A. The spectra of ERT-HCL showed characteristics peak at about 3277 cm^-1 (C-H stretching of the carboxylic group), 1629 cm^-1 (N-H bend of the secondary amine), 1508 cm^-1 (C=C stretch; aromatic ring) and at 1285 cm^-1 (C=O stretch) [59]. The PLGA showed characteristics peak at 1743 cm^-1 (C=0 stretching) and 1184 cm^-1 (C-0-C stretching vibrations) [69]. CS showed characteristic bands at 3281 cm^-1 (N-H, O-H stretching vibrations), and 1606 cm^-1 due to the presence of residual N-acetyl groups (C=O stretch of amide) [70].

Figure 5.

Characterization of ERT-HCL-Loaded CS-PLGA nanoparticles. (A) FTIR spectra, (B) DSC thermogram of ERT-HCL, PLGA, CS and ERT HCL-loaded CS-PLGA NPs, (C) TEM images of ERT-HCL-loaded CS-PLGA NPs, (D) in vitro drug-release profiles of pure ERT-HCL, ERT-HCL-PLGA NPs and ERT-HCL-CS-PLGA NPs (mean ± SD, n = 3).

CS: Chitosan; DSC: Differential scanning calorimetry; ERT-HCL: Erlotinib hydrochloride; FTIR: Fourier transform infrared spectroscopy; NP: Nanoparticle; PLGA: Poly (lactic-co-glycolic acid); TEM: Transmission electron microscope.

3.6.2. DSC analysis

The DSC thermograms of ERT-HCL, PLGA, CS and ERT-HCL-loaded CS-PLGA NPs are shown in Figure 5B. The drug ERT-HCL showed a sharp endothermic peak at around 234°C, corresponding to its melting point [59]. The DSC thermogram of PLGA revealed a thermal transition endothermic peak associated with the polymer's glass transition (Tg) between 30 and 40°C, corresponding to the energy absorption required for the transition from a glassy to a rubbery state [30,65]. The DSC thermogram of CS showed an endothermic at around 94°C, which signifies loss of water molecules, where the energy has been absorbed by the sample particles, which is consistent with the water evaporation process. An exothermic peak was observed at 306°C, which corresponds to the thermal degradation of the CS; and decomposition of the amine units (-RNH₂) [43,71].

3.6.3. Surface morphology

The ERT-HCL-loaded CS-PLGA NPs were spherical, and uniform in structure with mean particle diameter in agreement with optimized batch mean particle diameter analysis as depicted in Figure 5C. The zoomed-out figure at the scale of ∼1 μm, shows homogeneous structure and colloidal stability of the NPs. The CS coating appears as an evenly distributed thin layer encircling the PLGA core, providing the NPs with an extra layer of protection [72].

3.6.4. In vitro drug-release studies & kinetic modeling

The cumulative amounts of ERT-HCL released from the two NP formulations compared with the pure drug solution are depicted in Figure 5D. The drug release profiles of PLGA NP and CS-PLGA NP were significantly different (p < 0.05). A biphasic release pattern of the ERT-HCL from both the NP formulation can be observed [59]. Where there is an initial burst release of the drug followed by a much slower, sustained release of up to 72 h. The initial burst release can be attributed to the free excess drug absorbed on the NP's surface, which gets released much faster in the dissolution medium [73]. Conversely, the sustained release can be attributed to the slow gradual diffusion of the hydrophobic drug ERT-HCL which is packed in the dense matrix of PLGA and CS polymer system [74,75]. The initial burst release in the in vitro drug release of the NPs cannot be completely prevented but can be reduced to some extent, which was achieved by the addition of CS in the PLGA NP formulation. The initial release rate of ERT-HCL in CS-PLGA NPs was much slower than that in PLGA NPs, due to the charge attraction and adsorption of ERT-HCL on CS [35]. At the 12-h time, the PLGA NPs showed a cumulative drug release of 56.07% compared with CS-PLGA NPs, which showed a 35.83% release.

Release kinetics of ERT-HCL from two NP formulations according to different kinetic models is summarized in Supplementary Table S7. The ERT-HCL-loaded PLGA NPs were best fitted into first-order drug release (r² value of 0.977). On the other hand, when CS was incorporated into the formulation, in other words, ERT-HCL-loaded CS-PLGA NPs, it followed a zero-order drug release behavior (r² value of 0.993). The Korsmeyer Peppas model fitting indicated a Fickian diffusion-controlled drug release behavior (n value ≤ 0.43) for both the PLGA and CS-PLGA NPs formulation of ERT-HCL with n values of 0.34 and 0.41 respectively [76].

4. Discussion

The systematic approach to formulation development and optimization outlined in this study demonstrates the importance of understanding and controlling critical parameters in CS-PLGA NPs production. For the formulation of ERT-HCL loaded CS-PLGA NPs, preliminary lab trials were conducted, allowing optimization of various MAs, FVs and PPs. Those that were observed to be affecting the identified CQAs were screened and utilized in the DoE model for the preparation of CS-PLGA NPs.

In the application of a systematic and step-by-step QBD approach, the initial step involved defining and finalizing the QTPPs and CQAs. In our study, CS-PLGA NPs were chosen as the preferred drug delivery system due to their unique combination of biocompatible and biodegradable properties, which offer favorable characteristics for drug delivery. Pulmonary administration was identified as the primary route of delivery, recognizing its benefits such as increased drug accumulation at the local site, prolonged residence time, degradation protection and reduced systemic toxicities. Our formulation aimed for specific attributes to ensure effective drug delivery: low mean particle diameter, a zeta potential within the range of -20 mV to -30 mV, and high entrapment efficiency. Additionally, we sought long-term stability at room temperature over several months to years, ensuring sustained efficacy of the formulation.

Through a comprehensive QBD approach, mean particle diameter, zeta potential and entrapment efficiency were identified as the main CQAs essential for achieving the desired QTPP. Notably, we did not consider in vitro drug release as a primary CQA. This decision was made based on the inherent properties of our CS-PLGA NP system, which facilitates the gradual and sustained release of the hydrophobic drug ERT-HCL over an extended period, typically up to 72 h or more.

The Risk assessment is a methodical approach to determining and evaluating probable occurrences that may have a negative impact on the outcome and deciding regarding the tolerability of the risk-based risk analysis. Risk identification, analysis, evaluation, control, reduction and acceptance are all components of risk assessment under the ICH Q9 [77,78]. In the Ishikawa fishbone diagram for CS-PLGA NPs, the factors were divided into the choice of raw materials, MAs, PPs, FVs and milieu factors that would affect the CQAs. This depiction is similar to the form of fish bones, with branches representing factors that might affect the outcome shown on the fish's head.

The REM and FMEA methods were employed for the qualitative and quantitative analysis to prioritize the high-risk factors. The REM analysis conducted in our study unveiled several factors influencing the formulation process, categorizing them based on their risk levels. High-risk factors included the drug-polymer ratio, concentration of CS and stabilizer, organic-to-aqueous phase ratio, rate of organic phase addition, probe sonication time and solvent evaporation rate. These factors pose significant challenges and require careful consideration during formulation development to mitigate potential risks and ensure product quality. Conversely, certain factors were identified as medium to low risk, such as the choice of organic solvent and formulation stirring speed. While these factors may have less immediate impact on formulation outcomes, they still warrant attention to maintain consistency and optimize process efficiency. Understanding the risk profile of each factor is instrumental in prioritizing efforts during formulation optimization. High-risk factors demand optimization strategies, including robust DoE and systematic evaluation, to mitigate their impact on product quality. Conversely, medium to low-risk factors can be managed through standard operating procedures and established protocols, with periodic reassessment to ensure continued process robustness.

Quantitative FMEA analysis was carried out to identify and mitigate the plausible risk associated with the MAs and PPs. This was done by assigning a numerical value or a score to each MAs, FVs and PPs illustrated in an Ishikawa fishbone diagram. The factors having an RPN score ≥25 were used in the Taguchi factor screening study.

Dr. Genichi Taguchi's L8 orthogonal Taguchi design is an experimental method for determining the most effective combinations of variables to improve the quality of a final product or process [79]. In the L8 design, each column represents a control factor and each row indicates a combination of levels for each component; hence, the design is termed “orthogonal” since it uses a matrix with eight columns and eight rows. Typically, low, and high levels are used so that the impacts of the variables on the response variable may be tested [80]. The L8 design is especially convenient for small-scale experiments because it permits the evaluation of multiple factors with a relatively small number of experimental runs. Additionally, the design is balanced, which means that each level of each factor is examined an equal number of times. Each quality attribute used in the factor screening studies contained distinct first-order polynomial equations and polynomial coefficients, indicating no interactions between the factors. Furthermore, each factor was analyzed using Pareto charts that ultimately affected the selected responses. The factors on the left side of the chart have the highest impact on the response variable, while the factors on the right side have less impact. The two factors namely drug: polymer ratio and the CS concentration showed a significant effect (crossed the t-value limit of the Pareto chart) toward the selected CQAs and thus were selected as the main independent variables (X) to be used in the DoE model. The other five factors were set at a constant value as per preliminary batch trials, prerequisite knowledge and data from the Taguchi model to yield the desired responses of the selected CQAs.

With statistical implements, RSM is a tried-and-true approach for developing and optimizing pharmaceutical formulations, enabling researchers to acquire as much information as possible from a small number of experimental runs [81]. One approach in RSM that is well suited to pharmaceutical challenges is central composite design (CCD), which permits exploration using a minimal number of experimental runs with the identification of the ideal composition for reaching the pre-setting objective. Orthogonality and rotatability are crucial CCD properties that influence the quality and interpretability of the resultant models. In a CCD, orthogonality refers to the independence of the design variables, which means that the effect of one variable is not confounded with the effects of the other variables [82]. In other words, each variable is altered separately and does not influence the others. Orthogonality is desirable because it permits independent estimation of the effects of each variable on the response. Rotatability refers to a CCD's capacity to generate a comprehensive range of experimental conditions that can be rotated in multidimensional space without influencing the estimation of the individual effects of the variables [83]. A rotatable design permits a comprehensive design space exploration, resulting in greater comprehension of the response surface and the relationship between input variables and output responses. A CCD achieves orthogonality and rotatability by incorporating a set of evenly distributed ‘design points’ throughout the design space. Typically, these design points are arranged in a factorial design, with additional ‘star points’ and ‘center points’ inserted to estimate the response surface's curvature. The design elements are meticulously chosen to assure orthogonality and rotatability. By utilizing Design Expert software; version 14 (Stat-Ease, MN, USA), CCD design space was created where the independent variables (X) were evaluated at five different desired levels (-α, -1, 0, +1, +α). The α value of 1.414 was chosen to maintain the rotatability and orthogonality of the CCD design.

For the optimization of mean particle diameter (Y1), small particles are desirable as they can facilitate the permeability and penetration of physiological membranes/barriers, thereby facilitating passive administration and tumor cell targeting. The F-value of 53.84, (with only a 0.01% chance that an f-value this large could occur due to noise) [84], a p-value of <0.0001, and a good R2 value of 0.9472 with adjusted and predicted R2 values differing by 0.0398 (<0.2, indicating a lower deviation from the mean value, and a good correlation of determination) showed that the model can be used for the prediction of mean particle diameter [65,66]. The nonsignificant ‘lack of fit’ p-value of 0.0016 and ‘lack of fit’ F-value of 9.73 (indicating that there is a 0.16 percent possibility that such a large value could be the result of noise) [47,85] further demonstrated the model significance.

Both Factors X1 (drug: polymer ratio) and X2 (CS concentration) with F-values and p-values of 224.43, 8.63, <0.0001 and 0.0102 respectively were significant. The mean particle diameter had a significant dependency on the drug: polymer ratio factor in comparison to others. Factor X2 was also significant but dependence on this factor was comparably less. Software analysis showed that the quadratic model should be implemented over the linear model. Further, only factors having a p-value of less than 0.05 are significant and must be used for generating the polynomial equation [51]. As observed in the 3D surface plot, there is an increase in the mean particle diameter of the NPs with increasing levels of both variables. Although both the variables are significant, the X1 variable has a lot more effect as compared with the X2 variable on the mean particle diameter. The drug–polymer ratio influences the concentration of the drug and polymer in the NP formulation, which can impact the intermolecular forces within the components and their ability to self-assemble into small NPs. Similarly, CS; a polycationic polymer can form electrostatic interactions with the negatively charged surface of the PLGA. As the concentration of CS increases, the thickness of the CS coating on the NPs also increases, leading to an increase in the mean particle diameter of the NPs.

The second response (Y2) investigated was the zeta potential of the CS-PLGA NPs. The optimal zeta potential value is critical for particulate distribution stability and homogeneity. The F-value of 24.14, a p-value of <0.0001, and a good R2 value of 0.8895 with adjusted and predicted R2 values differing by 0.0852 showed that the model can be used for the prediction of zeta potential. The model significance was further supported by the nonsignificant ‘lack of fit’ p-value of 0.0002 and ‘lack of fit’ F-value of 13.01. Similarly, to the mean particle diameter response, the software suggested a quadratic model to be chosen over a linear one. The pure variables X1 and X2, and one of the quadratic variables, X2² were incorporated in the polynomial equation for zeta potential. As observed in the 3D surface plot, increasing levels of the drug: polymer ratio and the CS concentration increase the electrostatic repulsion between particles, leading to a higher zeta potential [86]. Although both the variables were significant, the (X1) variable; drug: polymer ratio has more effect as compared with the (X2) variable; CS concentration on the zeta potential of the NPs.

The third response (Y3) investigated was the entrapment efficiency of the ERT-HCL-loaded CS-PLGA NPs. The entrapment efficiency was obtained in the range of 48.83–87.31% w/w. A high entrapment efficacy is desired so that a greater quantity of drug can be encapsulated within the NPs, leading to a greater drug loading capacity and a more sustained drug release profile [87]. This, in turn, can result in improved therapeutic outcomes by allowing for a more controlled and prolonged release of the drug, thereby reducing the frequency of dosing and the likelihood of adverse effects. The F-value of 14.40, a p-value of <0.0001, and a good R2 value of 0.8276 with adjusted and predicted R2 values differing by 0.133 showed that this model is significant. The model significance was further supported by the nonsignificant ‘lack of fit’ p-value of <0.0001 and ‘lack of fit’ F-value of 44.38. the polynomial equation was created by incorporating significant variables X1 and X2, along with interaction variable X1X2, and one quadratic variable X1². In the 3D surface plot, the X1 variable has a higher effect as compared with X2 on the entrapment efficiency. When the concentration of PLGA is increased, more polymer molecules are available to interact with the drug or other substance, which can result in increased encapsulation [88]. In addition, the increased concentration of PLGA can result in a denser clustering of the drug and polymer molecules making it harder for the drug to escape the matrix [50].

The three responses were overlayed to obtain a desirable region. This overlay plot displays the region of interest, which is marked in yellow, and where values are expected to fall within the specified range. The gray zone indicates the area where the response deviated from specified criteria for the desired product characteristics. After adding narrower constraints to refine the optimization process, the software predicted a batch with a desirability of 0.767. The moderate to low desirability value observed may be attributed to several factors. These include potential trade-offs between different responses, narrower constraints for mean particle diameter (207.34 to 257.64 d.nm), zeta potential (20 to 30 mV), and entrapment efficiency (48.83 to 87.31% w/w), as well as the weighting assigned to each response in the optimization process. Specifically, the higher importance given to entrapment efficiency (importance of 5) compared with particle size and zeta potential (importance of 3 each), and model robustness could have influenced the desirability value. A similar desirability value was achieved by Kim et al., utilizing the Box-Behnken design for optimizing mesoporous silica nanoparticles [89].

As per the values suggested for the X1 and X2 factors, batches were formulated in triplicates, and a final optimized batch was obtained with a mean particle diameter of 226.50 ± 1.62 d.nm, a zeta potential of 27.66 ± 0.64 mV, and entrapment efficiency of 78.93 ± 1.94% w/w. The drug loading capacity for the NPs was 4.33% w/w. A PDI of less than 0.3, was observed for the NPs batches. which is satisfactory for polymeric nanocarriers. It is preferred to have a low PDI value, which indicates prepared NPs with minimal variance in mean particle diameter.

The FTIR analyses were conducted to investigate potential intermolecular interactions between the drug and the polymer. Additionally, the objective was to elucidate any observed peak shifts that occurred during this interaction. The characteristic bands of PLGA and CS were observed in the ERT-HCL-loaded CS-PLGA NPs, which confirms the successful adsorption of CS on the PLGA surface. The slight shift and decrease in the characteristic peak of CS can be attributed to the complex bonding between the amino and carboxylic groups of CS and PLGA respectively.

The DSC curve of ERT-HCL-loaded CS-PLGA NPs revealed no characteristic melting peak of the drug, indicating the lack of crystalline drug. The polymers may have prevented the drug from crystallizing when NPs were being formed. After encapsulation, ERT-HCL in the NPs likely assumed a molecular dispersion or solid solution amorphous or unstructured crystalline phase in the polymer matrix system [13,59].

The TEM analysis was performed to investigate the nanoscale surface morphology of NPs, which can be used to optimize their biological behavior and drug delivery efficiency [90]. The zoomed-out image demonstrated the homogeneity and colloidal stability of the uniform spherical NPs.

The in vitro drug release studies of ERT-HCL loaded PLGA and CS-PLGA NPs showed a biphasic release pattern, where there was an initial burst release followed by a gradual sustained release. When CS was incorporated into the PLGA NPs, the burst release was reduced, and at the 12 h time, CS-PLGA NPs showed a 35.83% release, and the PLGA NPs showed a release of 56.07%. As per fittings into different kinetic models, ERT-HCL-loaded PLGA NPs were best fitted in first-order release, while CS-PLGA NPs followed zero-order release, as the incorporation of CS reduced the initial burst release of the drug and resulted in the constant release of ERT-HCL from the NP formulation. Several investigations from other research groups have previously revealed comparable drug-loaded CS/PLGA NP release patterns [91]. As per the Korsmeyer Peppas model, both formulations showed Fickian drug release behavior release (n value ≤ 0.43), suggesting the release of ERT-HCL from these NPs is primarily governed by diffusion through the polymer matrix

5. Conclusion

The systematic formulation development and optimization discussed herein highlight the crucial role of understanding and controlling critical parameters in producing ERT-HCL-loaded CS-PLGA NPs. Preliminary lab trials informed the optimization of multiple MAs, FVs and PPs identified through their impact on the CQAs. Establishing product-centric QTPP and CQAs was pivotal in our formulation approach, selecting CS-PLGA NPs due to their biocompatibility and biodegradability for pulmonary drug delivery. Through risk assessment using Ishikawa fishbone diagrams, REM and FMEA, high-risk factors affecting formulation were prioritized for optimization. Taguchi's L8 orthogonal design facilitated factor screening, with the drug:polymer ratio and CS concentration identified as the main variables influencing CQAs. An effective and robust CCD model was employed to explore the design space. Based on the model's desirability factor, the optimal desired batch was obtained with a mean particle diameter of 226.50 ± 1.62 d.nm, a zeta potential of 27.66 ± 0.64 mV, and entrapment efficiency of 78.93 ± 1.94% w/w. The drug loading capacity of CS-PLGA NPs was determined to be 4.33% w/w, with a satisfactory PDI value of less than 0.3, indicating minimal variance in the mean particle diameter of NPs. Characterization studies further confirmed successful CS adsorption on PLGA, absence of drug crystallinity and homogeneity of NP morphology. In vitro release studies revealed sustained drug release from CS-PLGA NPs up to 72 h, exhibiting zero-order kinetics, indicative of Fickian diffusion-controlled release mechanisms. To our knowledge, this study is one of a kind representing a comprehensive formulation strategy integrating systematic QbD principles, risk assessment, experimental design and characterization techniques to develop and optimize CS-PLGA NPs of ERT-HCL for efficient drug delivery applications. For future studies, we will carry out surface modification of the optimized formulated ERT-HCL-loaded CS-PLGA NPs for better targeting and providing therapeutic efficacy.

Supplementary material

Supplemental data for this article can be accessed at https://doi.org/10.1080/20415990.2024.2342771

Author contributions

HP Nijhawan: conceptualization, investigation, data curation, visualization writing-original draft. B Prabhakar: supervision, visualization, writing – review & editing of the draft. KS Yadav: conceptualization, supervision, writing – review, editing and proofreading of the draft.

Financial disclosure

The authors have no financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Competing interests disclosure

The authors have no competing interests or relevant affiliations with any organization or entity with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, stock ownership or options and expert testimony.

Writing disclosure

No writing assistance was utilized in the production of this manuscript.