Development and optimization of a topical self-microemulsifying drug delivery system of Mammea americana L. using a quality by design and Box–Behnken approach

2.1. QbD framework applied to formulation development

The development of the topical self-microemulsifying formulation was structured under the QbD paradigm, emphasizing a process-oriented and predictive approach rather than empirical formulation. Over the last two decades, the pharmaceutical industry has become increasingly less reliant on models based solely on end-product testing, shifting toward proactive QbD approaches in which variability is anticipated and controlled rather than corrected [3].

Following the principles outlined in the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) Q8 (R2) guideline [4,5], this study applied QbD tools to identify and control the factors most critical to product quality. Accordingly, the experimental workflow involved:

• Establishing the quality target product profile (QTPP);
• Defining the critical material attributes (CMAs) related to the M. americana L. ethanolic extract and excipients;
• Identifying critical process parameters that influence emulsification and stability; and
• Determining critical quality attributes (CQAs) such as droplet size, transmittance, and physical stability.

Each step was based on preformulation data and previous reports describing QbD-guided topical formulation development [2,3]. This structured design enabled proactive control of variability and enhanced reproducibility of the final topical formulation.

2.2. Definition of the QTPP

To define the QTPP, the intended use and desired performance of the topical self-microemulsifying drug delivery system (SMEDDS) containing the ethanolic extract of M. americana L. were first specified. The process considered critical factors influencing patient safety, product stability, and application behavior.

The QTPP was established in accordance with ICH Q8 (R2) guidelines [4,5] and previous methodological reports on QTPP construction for semisolid and lipid-based formulations [6].

In accordance with the ICH Q9 guideline on Quality Risk Management [5], a preliminary risk assessment was conducted to support variable selection during product development. Performing a structured risk analysis facilitates the design of targeted studies and helps define critical variables, thereby strengthening the overall control strategy and deepening process understanding [7]. In this study, an Ishikawa (cause-and-effect) diagram and a simplified risk matrix were conceptually developed to evaluate how different factors could influence the development of a topical system derived from natural sources. These tools were used only as supportive elements to identify and prioritize formulation-related factors that may impact the final product’s quality, such as stability, droplet size, and emulsification behavior. Among these, physical stability and droplet size were considered the most CQAs, in agreement with previous reports describing formulation risk mapping for complex lipid-based systems [8].

This risk assessment was applied exclusively as a conceptual guide for selecting the experimental variables included in the Box–Behnken optimization design and was not treated as a quantitative result, in accordance with reviewer recommendations.

2.3. Characterization of the ethanolic phytotherapeutic extract

The ethanolic extract of M. americana L. was prepared from the endosperm of mature fruits collected in the Bolívar Department, Colombia (10°25′N, 75°30′W). The plant material was taxonomically authenticated by a certified botanist and deposited in the Herbarium of the Guillermo Piñeres Botanical Garden under voucher code JBC 467, ensuring full traceability and reproducibility of the biological source [9].

Extraction was carried out by dynamic maceration using extra-neutral ethanol (96%, USP grade) at 30°C for 24 hours under continuous agitation. The solvent was removed under reduced pressure at 40–45°C and 115 rpm, yielding a semi-solid ethanolic extract designated as EPE-01.

Phytochemical characterization was performed using Fourier Transform Infrared Spectroscopy (FTIR) (Jasco 4,700 spectrometer, 4,000–600 cm^-¹) and High-Performance Liquid Chromatography–Mass Spectrometry (UHPLC–MS) Platin Blue UHPLC system, KNAUER GmbH, coupled to an LTQ Orbitrap mass spectrometer, ThermoElectron Corporation [10–12]. The chromatographic and spectroscopic profiles revealed characteristic signals corresponding to coumarin-type and polyphenolic metabolites, consistent with those previously reported in the endosperm of M. americana. These findings support its suitability as an active phytopharmaceutical ingredient for topical formulation development.

2.4. Solubility of the extract in formulation auxiliaries

The water solubility analysis was in line with OECD Test No. 105 guideline for water solubility [13], in modification for applications with pharmaceutical excipients. Briefly, a fixed volume of each formulation adjuvant under conditions from Table 1 was pipetted into sealed test tubes containing 0.1 g of the extract. Orbital and reciprocal motion were applied for 10 minutes under 150 rpm with a 20°C ± 0.5°C temperature control via a ROTATERM model 10747ERM.

Table 1. Solubility in formulation auxiliaries.

Source: Adapted from the Organization for Economic Cooperation and Development [13].

^aSolubility categories were estimated based on the volume of auxiliary required to dissolve 0.1 g of extract, following the OECD water-solubility classification system.

The solution was stirred after each addition, and the onset and end of solubilization were followed. Aliquots were also taken for total dissolved solids determination using a Milwaukee MA871 refractometer that correlates solid content with light refraction [14]. A blank was taken first, followed by a mean of three reads for each sample, followed by subtraction of the blank to provide the final reading. The data facilitated the determination of each excipient’s ability to solubilize together with optimization of dissolution conditions for formulation optimization [15].

2.5. Box–Behnken experimental design

An appropriate understanding of product and process parameters is inevitable for the successful implementation of the QbD philosophy. Along with mechanistic modeling, design of experiments is a strong statistical tool that enables intentional control of significant formulation variables based on a pre-selected design [16,17]. Based on solubility screening data and typical component ranges for lipid-based systems published by Pouton [18], a BBD was adopted with Design-Expert^® software (version DX13). There were three independent variables in the experimental design: oily phase concentration (rosemary oil), surfactant (Tween 80), and co-surfactant (Tween 20), investigated under three levels each, within recommended formulation ranges:

• Rosemary oil: 10%, 15%, 20%
• Tween 80: 5%, 10%, 15%
• Tween 20: 5%, 10%, 15%.

The system’s performance was determined based on two principal responses: self-emulsification time and self-emulsification grade, according to Khoo et al. [19] and Goméz et al. [20]. The results allowed response surface plotting for the identification of optimum formulation parameters for the topical self-emulsifying system. Transmittance was additionally recorded as a descriptive parameter to support the emulsification performance, but it was not included in the statistical modeling or optimization.”

2.6. Emulsification time and self-emulsification grade

The emulsification time was assessed using a magnetic stirrer set at 200 rpm. Each formulation was added dropwise into 250ml of distilled water maintained at 37°C in a beaker. The formation of the emulsion was monitored visually using a stopwatch, and the time required to achieve complete emulsification was recorded [19,20].

Immediately after emulsification, the formulations were visually classified according to the self-emulsification grade, following the criteria established by Khoo et al. [19]. The emulsions were categorized based on appearance, clarity, and formation speed, as described in Table 2. This classification helped assess the dispersion quality and optical characteristics of each system.

Table 2. Degrees of self-emulsification.

Source: Adapted from Khoo [18].

2.7. Transmittance evaluation of the emulsions

Due to their nano- and microstructural characteristics, self-emulsifying systems can exhibit optical properties similar to those of true solutions, such as translucency. This phenomenon occurs when droplet sizes fall below 100nm, allowing light to pass through the dispersion without significant scattering [21].

To evaluate optical clarity, samples of each emulsion were analyzed using an Agilent Cary 60 UV-Vis spectrophotometer, and the percentage of transmittance was recorded. This test was included as a complementary quality attribute, intended to qualitatively support droplet dispersion assessment, rather than as a formal response variable in the BBD.

2.8. Characterization of optimized formulation

Formulation optimization was performed using a BBD, where the ratios of rosemary oil, Tween 80, and Tween 20 were evaluated to minimize emulsification time and emulsification grade. Statistical extrapolation identified an optimal composition, which was subsequently prepared.

The optimized replicates were prepared in deionized water and subjected to characterization. Droplet size distribution was determined by dynamic light scattering (DLS) at 25°C using quartz cuvettes, with five consecutive runs per sample. Results were expressed as mean diameter, median, and standard deviation, reporting multimodal distributions when present. Zeta potential was measured by laser Doppler electrophoresis under the Smoluchowski model at 25°C using disposable capillary cells, with three independent measurements per replicate. Mean values with standard deviations were calculated for each formulation.

3.1. Determination and evaluation of the QTPP and CMA

The QbD approach, which emphasizes defining the QTPP and identifying the CMAs, was employed to ensure the topical formulation met desired performance and safety requirements. Accordingly, the QTPP for a lipid-based topical dosage form was designed to guide excipient selection and formulation strategy, focusing on attributes most relevant to quality (Fig. 1).

Figure 1. Ishikawa diagram of factors influencing the quality of a topical system based on natural products. [Click here to view]

To support CMA identification, a bibliographic review highlighted that secondary metabolites such as tannins, polyphenols, and coumarins abundant in the ethanolic extract of M. americana L. endosperm (EPE-01) exhibit antioxidant, antimicrobial, and anti-inflammatory activities. These bioactivities justify their integration into the formulation and underpin the therapeutic rationale of the developed system (Tables 3–5).

Table 3. Quality target product profile for a topic formulation.

Table 4. Critical material attribute for a topical product.

Table 5. Risk analysis matrix for topical formulation components based on Quality by Design principles.

Additionally, in accordance with the ICH Q9 guideline on Quality Risk Management [5], a conceptual risk assessment was performed to support variable selection during product development. An Ishikawa (cause-and-effect) diagram and qualitative risk matrix were developed to identify potential factors affecting product quality [7,8]. This analysis served only as a conceptual guide to prioritize the formulation variables (rosemary oil, Tween 80, and Tween 20) later optimized in the BBD. No quantitative results were generated from this analysis, in accordance with reviewer recommendations.

3.2. Characterization of the ethanolic plant extract

The ethanolic extract obtained from the endosperm of M. americana L. was characterized by FTIR and UHPLC–MS analyses to elucidate its main phytochemical constituents.

The FTIR spectrum (Fig. 2) displayed characteristic absorption bands indicative of diverse functional groups. The region between 1,000 and 1300 cm^-¹ showed strong C–O–C and C–O stretching vibrations, typically associated with coumarins, polyphenols, and tannins [12,22]. Absorption bands within 1,500–1,600 cm^-¹ corresponded to aromatic C=C stretching, confirming the presence of phenolic rings, while the distinct peak near 1,700–1,800 cm^-¹ was assigned to carbonyl (C=O) stretching vibrations characteristic of lactone-type coumarins. Finally, a broad absorption region at 3,200–3,400 cm^-¹, attributed to hydroxyl (–OH) stretching, further supported the presence of phenolic and polyphenolic compounds extracted with ethanol.

Figure 2. Fourier Transform Infrared (FTIR) spectrum of the EPE. Legend: FTIR spectrum of the EPE (EPE-01), with key regions highlighted to facilitate identification of characteristic functional groups. Shaded zones indicate specific vibrations: light blue for hydroxyl groups (–OH), dark blue for carbonyl groups (C=O), pink for aromatic rings, and turquoise for ether bonds (C–O and C–O–C). [Click here to view]

Taken together, the FTIR profile revealed a complex mixture of bioactive secondary metabolites, mainly phenols, polyphenols, tannins, and coumarins, consistent with previously reported data for M. americana endosperm extracts.

Complementary UHPLC–MS analysis (Fig. 3) confirmed the presence of coumarin-related compounds, with prominent peaks at retention times of 22.54, 23.08, 23.53, 24.18, and 24.47 minutes, corresponding to molecular ions of m/z 431.2056, 389.1952, 373.2002, and 495.2725. These signals are consistent with coumarin derivatives reported in the Mammea genus, supporting their role as phytochemical markers for both quality control and potential biological activity.

Figure 3. UHPLC–MS chromatographic profile of the Phytotherapeutics ethanolic extract (EPE-01) recorded at 332 nm, showing peaks corresponding to metabolites of different polarities (more polar to less polar). [Click here to view]

3.3. Solubility of the extract in different formulation excipients

As shown in Figures 4 and 5, the ethanolic plant extract (EPE) exhibited greater solubility when mixed with rosemary oil and eucalyptus oil using a solvent volume of 1ml. This was attributed to their chemical composition and strong affinity with the extract’s lipophilic metabolites. Rosemary oil contains major terpenes such as 1,8-cineole (15%–55%) and α-pinene (9.0%–26%), while eucalyptus oil includes 1,8-cineole (63.1%), p-cymene (7.7%), α-pinene (7.3%), and α-limonene (6.9%) [23,24]. These components may favor the dissolution of coumarins and other phytoconstituents [25]. In contrast, cocamidopropyl betaine and Cremophor RH 40 showed poor solubility performance, likely due to a mismatch between their hydrophilic-lipophilic balance (HLB) and the extract’s chemical profile [23].

Figure 4. Total solids (Brix grades) content dissolved per 1l of solvent in selected formulation auxiliaries. [Click here to view]

Figure 5. Total solids (Brix grades) content dissolved per 10 ml of solvent in selected formulation auxiliaries. [Click here to view]

When higher volumes were used, Tween 80 and Tween 20 yielded the best results. Their high HLB values (16.7 and 15.0, respectively) reflect their hydrophilic nature. An increase in alkyl chain length appeared to expand micelle volume, thereby enhancing solubilization of lipophilic compounds [26,27].

3.4. Selection of excipients and BBD setup

According to the solubility study, excipients with higher affinity for the EPE were selected for formulation. In line with the philosophy of QbD, CMAs for these excipients were established so that they could be proactively monitored for any impact they could pose on formulation functionality and product quality, as shown in Tables 6 and 7. A design matrix (Table 8) was generated using Design-Expert^® DX13 software, with some variables and levels determined based on solubility behavior observed from previous experiments. Input factor levels were determined based on recommended ranges for lipid-based systems reported by Pouton [18]. The response variables adopted were time to self-emulsification and grade of self-emulsification based on the evaluation criterion proposed by Khoo et al. [19].

Table 6. Justification for the selection of formulation excipients.

Table 7. CMAs of selected excipients.

Table 8. BBD matrix.

3.5. Emulsification time results

The emulsification time for each formulation (Table 9) was evaluated to assess dispersion efficiency and system responsiveness. The experimental data, analyzed using Design-Expert^® DX13, produced a predictive model with an R² of 90.00% and an adjusted R² of 72.16%, indicating good model fit considering the complexity of the system. The maximum variance inflation factor among the variables (rosemary oil, Tween 80, and Tween 20) was 1.011, confirming the absence of multicollinearity and validating the reliability of the regression estimates [28]. Formulations 1, 4, 8, 9, 11, 12, and 13 exhibited emulsification times ranging from 87 to 156 seconds, which are considered suitable for SEDDSs. These values indicate rapid and efficient emulsification, crucial for ensuring uniformity and stability of the topical formulation. The 3D response surface plot (Fig. 6) illustrates that increasing the concentration of rosemary oil significantly reduced emulsification time. Conversely, Tween 80 showed minimal influence within the tested range, suggesting its effect may be non-linear or synergistic with other variables outside the evaluated levels.

Figure 6. 3D surface and contour plots showing the effect of rosemary oil and Tween 80 concentrations on emulsification time. [Click here to view]

Table 9. Emulsification time for each formulation.

ANOVA results confirmed the linear effect of rosemary oil (p = 0.0044) and the quadratic term AA (p = 0.042) as statistically significant, with no autocorrelation observed in the residuals. Although Tween 80 was not statistically significant within this design space, high-HLB surfactants like Tween 80 are known to support emulsification by reducing droplet size and enhancing dispersion kinetics [29].

3.6. Self-emulsification grade

In addition to emulsification time (Table 10), the self-emulsification grade provides qualitative insight into the clarity and visual characteristics of the emulsions formed. This evaluation is crucial for topical systems, where appearance and dispersion quality impact patient acceptability and formulation stability. The statistical model showed an excellent fit, explaining 96.04% of the variability (R²), with an adjusted R² of 88.92%, indicating high predictive capability. At a 95% confidence level, the significant variables were:

Table 10. Self-emulsification grade of each formulation.

• Factor A (rosemary oil, p = 0.0166)
• Factor C (Tween 20, p = 0.0081)
• Quadratic terms: AA (p = 0.0022), CC (p = 0.0049)
• Interaction terms: AB (p = 0.0103), AC (p = 0.0301)

As shown in Figure 7, the lowest self-emulsification grades, which correspond to clearer and finer emulsions, were obtained in the central region of the design space, particularly when rosemary oil and Tween 80 were used at intermediate concentrations. In contrast, high concentrations of rosemary oil or Tween 20, when used individually, tended to increase the emulsification grade, producing more turbid and slower-forming emulsions. This may be attributed to saturation effects or phase behavior that interferes with optimal droplet dispersion. These results emphasize the importance of a balanced ratio between the oily phase and surfactants to achieve optimal emulsion quality, especially in self-emulsifying topical systems where physical appearance and optical clarity are key formulation attributes.

Figure 7. 3D surface and contour plots showing the effect of rosemary oil and Tween 20 concentrations on self-emulsification grade. [Click here to view]

3.7. Transmittance of the emulsions

Self-emulsifying systems can have high optical clarity and frequently resemble real solutions because of their nano- or microemulsion nature. This characteristic is usually associated with droplet sizes smaller than 100 nm, which improves transparency and lessens light scattering [21]. A translucent appearance consistent with fine droplet dispersion was suggested by some formulations in the current study that displayed transmittance values near 92%. Transmittance was not regarded as a primary response variable in this experimental design; rather, it was used only as a supporting quality parameter that supported the interpretation of emulsification efficiency and assisted in qualitatively comparing samples (Fig. 8).

Figure 8. Transmittance (%) of each formulation measured by UV-Vis spectrophotometry. [Click here to view]

3.8. Characterization of optimized formulation

A predictive framework was made available by the BBD to determine the ideal ratio of Tween 80 (6.7%), Tween 20 (15.0%), and rosemary oil (13.5%) in order to achieve high optical clarity and quick emulsification. Although the model proposed a perfect formulation with a desirability index of 1.0, these predictions needed to be verified experimentally. Consequently, the optimized formulation was made and put through DLS and zeta potential characterization.

Trimodal distributions were found using DLS, with nanometric fractions (~21–26 nm), intermediate populations in the submicrometric range (344–684 nm), and a third peak in the micrometric domain (2.5–3.8 µm). A significant portion of the droplets in this multimodal profile reached the nanometric scale. Larger populations, however, persisted. This indicates incomplete homogenization, a phenomenon already reported with natural extracts. Due to their complex composition, certain extract components may not fully solubilize in the lipid phase or may interfere with the formation of nano-droplets. As a result, aggregates or separate phases can appear [30]. The nanometric fraction is expected to have functional relevance for SEDDSs, as it contributes to high surface area and rapid dissolution/dispersion behavior.

Recent research on SNEDDS indicates that multimodal size distributions can be compatible with sufficient stability and enhanced bioavailability, provided the nanometric fraction remains constant over time [31]. Therefore, the relatively high global PDI values should not be interpreted as instability per se but as an intrinsic feature of the complex droplet population in SEDDS formulations.

Zeta potential analysis provided additional insights into the stabilization mechanisms of the optimized formulation. The prototype exhibited a value of −18.5 mV, respectively. Corresponding to moderate electrostatic repulsion. Although these magnitudes are below the conventional ±30 mV often associated with high colloidal stability, it is important to highlight that non-ionic surfactants were used in the formulations. In this context, steric stabilization plays a key role. The observed multimodal droplet size distributions remained stable during short-term storage, supporting the system’s functional stability [32]. Overall, these findings validate the predictive accuracy of the BBD model and highlight critical physicochemical limitations. The optimized formulation generated nanometric droplets. However, its multimodal size distribution and moderate zeta potential indicate stability constraints, which must be addressed by adding stabilizing agents.