Quality by design driven development of a Monteverdia ilicifolia herbal medicine: from plant to granules

1 Introduction

Monteverdia ilicifolia (Mart. Ex Reissek) Biral (syn. Maytenus ilicifolia Mart. Ex Reissek) is a medicinal plant of the Celastraceae family. The genus name has historical roots in the term “maytén,” used by Mapuches, the indigenous peoples of southern South America to designate medicinal plants traditionally employed for therapeutic purposes. The specific epithet ilicifolia refers to the morphological resemblance of its leaves to those of the genus Ilex, characterized by coriaceous texture and spiny margins. This species, commonly known as “espinheira-santa,” has a long-standing ethnobotanical relevance and is widely used in traditional medicine, particularly as infusions and decoctions for the treatment of gastrointestinal disorders (Silva et al., 2024).

Beyond its ethnobotanical relevance, M. ilicifolia is also associated with challenges related to botanical identification, as different species may share similar morphological features and the same popular name. Species such as Sorocea bonplandii and Zollernia ilicifolia, which belong to distinct botanical families, are often confused with M. ilicifolia. This overlap in vernacular nomenclature and morphological traits may lead to the substitution or adulteration of plant material, directly affecting chemical composition and, consequently, the therapeutic consistency and safety of herbal preparations (Di Stasi et al., 2000; Antunes et al., 2025).

These challenges go beyond species identification and reflect a broader issue inherent to herbal medicines due to their intrinsic chemical variability. Unlike synthetic drugs, herbal products are composed of multiple bioactive compounds whose concentration and composition may vary significantly depending on factors such as pre- and post-harvest conditions, processing methods, and extraction procedures. These sources of variability directly impact the quality, efficacy, and safety of the final product, posing significant challenges for standardization and industrial production. Furthermore, traditional preparation methods, such as infusions and decoctions, are often poorly controlled and lack reproducibility, which limits their translation into robust pharmaceutical formulations (Park et al., 2022; Kasera et al., 2024).

Over time, the concept of quality in pharmaceutical development has evolved from an approach primarily based on end-product testing to a more comprehensive paradigm focused on process understanding and control. In this context, Quality by Design (QbD), as proposed by the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH Q8), represents a significant advancement by promoting a systematic, science- and risk-based framework for product development (Juran et al., 1999; ICH, 2009; 2023).

QbD is structured around key elements, including the definition of the Quality Target Product Profile (QTPP), the identification of Critical Quality Attributes (CQAs), and the assessment of Critical Material Attributes (CMAs) and Critical Process Parameters (CPPs). These elements are supported by risk management tools and Design of Experiments (DoE), enabling a deeper understanding of the relationships between input variables and product performance. Within this framework, quality is built into the process rather than tested solely at the end, and this knowledge base supports the establishment of a Design Space (DS), defined as the multidimensional combination of material attributes and process parameters that ensures consistent product quality. The definition of the DS provides a scientific foundation for process control, contributing to the development of herbal medicines (ICH, 2009; Fukuda et al., 2018; Devangan et al., 2024).

In this context, the present study aimed to develop an herbal medicine based on a water-soluble dry extract of Monteverdia ilicifolia using a Quality by Design (QbD) approach. The study encompassed the main stages of the production chain, including harvesting, drying, milling, extraction, concentration, spray drying and granulation. Risk assessment tools and Design of Experiments (DoE) were employed to identify critical factors and evaluate their effects on responses associated with product quality. The integration of these strategies enabled a comprehensive understanding of process variability and supported the establishment of a robust and scalable manufacturing route for the proposed herbal product.

2 Materials and methods

2.1 Materials, standards and reagents

The plant material consisted of fresh leaves from Monteverdia ilicifolia. The samples were collected from the experimental field of the Multidisciplinary Center for Chemical, Biological, and Agricultural Research (CPQBA-UNICAMP) in Paulínia, Brazil, located at coordinates S 22° 47′ 52″ W 47° 6′ 49”. Botanical identity was confirmed by a voucher specimen deposited in the CPQBA herbarium under code CPMA 2429 (https://specieslink.net/col/CPMA/), and the activities were registered in the National System for the Management of Genetic Heritage (SISGEN) under code AFCE992.

Chemical reference standards included pyrogallol (Vetec, Rio de Janeiro, Brazil and Sigma-Aldrich, Steinheim, Alemanha), epicatechin, and catechin (Sigma-Aldrich, Cotia, Brazil). Processing reagents and materials included anhydrous sodium carbonate (Êxodo, Sumaré, Brazil), Folin-Ciocalteu reagent (Êxodo, Sumaré, Brazil), acetonitrile (Tedia, Rio de Janeiro, Brazil), and acetic acid (Êxodo, Sumaré, Brazil). Maltodextrin (Ingredion, Mogi Guaçu, Brazil) was utilized as a carrier agent for atomization, and hydroxypropyl methylcellulose, HPMC (Sigma-Aldrich, Cotia, Brazil) was used as a binder for granulation.

2.2 Quality target product profile definition

The responses investigated were selected based on their relevance to achieving the predefined Quality Target Product Profile (QTPP), which was established as a water-soluble dry extract of M. ilicifolia exhibiting industrially suitable technological properties while preserving the chemical composition of the aqueous extract. These responses correspond to the Critical Quality Attributes (CQAs), which are properties that must be controlled within appropriate limits to guarantee product quality.

The responses were categorized into physicochemical (related to chemical composition), technological (related to pharmacotechnical performance of the granulated product), and operational (related to process feasibility). Table 1 relates these attributes to their targets and process stages.

TABLE 1. List of critical quality attributes for evaluating the respective stages.

Abbreviations: D = drying; M = milling; E = extraction; C = concentration; S = spray drying; G = granulation.

2.2.1 Total Polyphenols Content

Total Polyphenols Content (TPC) was evaluated by the Folin-Ciocalteu (FC) spectrophotometric method according to (Makkar et al. 2013) using an Agilent 8453 spectrophotometer (Santa Clara, United States of America).

A 500 µL aliquot of the diluted extract (0.3 mg/mL) was mixed with 200 µL of 1 M FC reagent, followed by the addition of 1.25 mL of sodium carbonate solution (7.5%, w/v). The reaction mixture developed a blue coloration, and absorbance was measured with a 1 cm quartz cuvette at 760 nm after a reaction time of 30 min at 20 °C ± 2 °C and protected from the light. Quantification (Equation 1) was performed using a calibration curve prepared with pyrogallol as the reference standard. The results were expressed as percentage and mass of total polyphenols extracted (Makkar, 2013).

T P = A b s − β 0 β 1 × C d w (1)

Abbreviations: T P C = Total Polyphenols Content, concentration of polyphenols in the sample (% m/m); A b s = absorbance of the sample; β 1 = angular coefficient of the pyrogallol calibration curve (Abs/μg/mL); β 0 = linear coefficient (Abs); C d w = equivalent concentration of the dry extract of the plant drug (µg/mL).

2.2.2 Total Tannins Content

Total Tannins Content (TTC) was determined by the difference between TPC and the fraction of polyphenols not absorbed by Polyvinylpolypyrrolidone, PVPP (Sigma-Aldrich, Cotia, Brazil). The sample was treated with 100 mg of PVPP to remove tannins, followed by centrifugation to separate the adsorbed fraction. The resulting supernatant, containing non tannin phenolic compounds, was diluted to 1.0 mg/mL and reanalyzed using the FC assay. (Makkar, 2013).

2.2.3 Epicatechin and catechin content

Analyses were performed by liquid chromatography with diode array detection (LC-DAD) using an Alliance system (Waters, Milford, United States of America). The chromatographic separation was achieved on ACE C₁₈ column (5 μm, 250 mm × 3.9 mm) coupled to a C₁₈ pre-column (10 mm × 4 mm) and using 1% (v/v) acetic acid and the mixture of acetonitrile: water: acetic acid (50:49:1, v/v/v) as eluent A and B, respectively. A gradient elution was carried, as shown in Supplementary Table S1 and the chromatograms were obtained at 280 nm.

2.2.4 Moisture content

Moisture content was analyzed using an infrared moisture analyzer (Ohaus MB25, Parsippany, United States of America) set to 120 °C until constant weight, calculating the percentage of mass loss.

2.2.5 Initial (ρ B m i n) and final (ρ B m a x) bulk density

The powder density was determined using a 25 mL graduated cylinder. The initial bulk density was calculated by the mass to volume ratio of the poured powder and the final bulk density was obtained after subjecting the cylinder to mechanical agitations until a constant volume was reached, and calculated as described by Equations 2, 3 (Aulton and Taylor, 2013). Abbreviations: m = mass (g); V = Volume (mL); o = original state; f = final state.

ρ B m i n = m V o (2)

ρ B m a x = m V f (3)

The Carr Index (%CI) indirectly assesses the compressibility properties of powders by comparing their initial and final bulk density (ρ B m i n and ρ B m a x). The %CI was calculated using the formula shown in Equation 4.

% I C = ρ B m a x − ρ B m i n ρ B m a x × 100 (4)

In order to meet technological requirements and make the production chain viable, a Carr index ≤25% was adopted as an acceptance criterion, as it provides minimally adequate characteristics, as recommended by (Aulton and Taylor, 2013).

2.2.6 Angle of repose (θ)

The angle of repose was measured by pouring the extract through an analytical funnel at 2.5 cm height (h) from the countertop to the funnel stem. The radius of the base of the formed cone was measured, and the angle calculated via trigonometry Equation 5, with results verified by image analysis using ImageJ software (version 1.54 g).

θ = arctan h r (5)

To achieve technological objectives and make the production chain viable, an angle of repose ≤40° was considered as an acceptance criterion, as it provides acceptable flow characteristics, as recommended by Aulton and Taylor, 2013.

2.2.7 Hygroscopicity (%H)

Samples were exposed in a climatic chamber (Tecnal TE-4003, Piracicaba, Brazil) set to 75% relative humidity and 50 °C. Mass gain was monitored until stabilization and expressed as a percentage of absorbed moisture, according to Equation 6.

% H = m e x p o s e d − m i n p u t m i n p u t × 100 (6)

2.2.8 Wetting time

The wetting time was evaluated by the time (in seconds) required for a unit dose of the extract (powder equivalent to 3 g of herbal drug) spread over water at 50 °C to sink or become completely wet (Hogekamp and Schubert, 2003).

2.2.9 Color evaluation

Color measurements was performed using a MiniScan colorimeter (HunterLab, Reston, United States), operating in the CIELAB system (L, a, and b), where L represents lightness, ranging from 0 (black) to 100 (white). The evaluation of color parameters was included as a complementary analysis to assess variations among experimental conditions, allowing the detection of differences between experimental conditions and supporting the assessment of process reproducibility. (Zhang and Wandell, 1996).

2.3 Process mapping

The process mapping was conducted to provide a detailed understanding of the activity flow and to identify hazards that influence the final product’s quality. The production process for the M. ilicifolia water-soluble dry extract was mapped using Bizagi Modeler software (version 4.1.0.010) (Bizagi, 2025).

Figure 1 illustrates the integrated workflow of the production process, highlighting the interdependence between agricultural stages and technological processing steps. The process begins with cultivation, carried out in the experimental field of CPQBA, followed by harvesting of the medicinal plant using pruning shears, drying in a forced-air circulation oven (Fanem 520-C, Guarulhos, Brazil), and milling in a knife mill (Primotécnica, Mauá, Brazil). The workflow then proceeds with extraction by infusion or decoction, according to the recommendations of the Brazilian Pharmacopoeia, followed by extract concentration using a rotary evaporator (Büchi R-215, Flawil, Switzerland), atomization by spray drying (Mini Spray Dryer Büchi B-290, Flawil, Switzerland), and manual granulation using a porcelain mortar (Brasil, 2021).

2.4 Risk assessment

Risk assessment was conducted in accordance with the recommendations of the International Organization for Standardization (ISO) 14,971 and used the Failure Mode and Effects Analysis (FMEA) tool. This systematic approach allowed for the evaluation of potential failures throughout the process stages by considering their effects on product quality. The analysis established a relationship between the probability of occurrence, detection capability, and the severity associated with each identified hazard, enabling the quantification of risk.

A comprehensive description of the risk assessment, including the identified hazards, scoring criteria, and calculated Risk Priority Numbers (RPN), is provided in Supplementary Table S2. This evaluation supported the identification of critical process steps and guided the selection of factors for subsequent experimental investigation.

2.5 Design of experiments (DoE)

Based on the risks assessment, the critical process steps were further investigated using Design of Experiments (DoE). Process stages with well-established conditions in the literature, such as drying, were considered to be under control, and their parameters were defined based on previous studies, thus not being subjected to experimental design investigation (Negri et al., 2009).

For the lab-scale extraction step, a screening study was conducted using a design including nine factors to estimate main effects. Subsequently, the process was scaled up to pilot scale, where the main effects and interactions between two factors previously identified as critical were evaluated. Finally, in the spray drying step, three factors were included in the experimental design, allowing the estimation of main, quadratic, and interaction effects. This approach enabled a deeper understanding of process behavior and supported the identification of optimal operating conditions (Box et al., 1978; Rodrigues and Iemma, 2015; Montgomery, 2017).

2.5.1 Lab-scale extraction using Plackett–Burman design (PBD)

The selection of this design was motivated by the large number of factors initially considered in the study, for which screening designs are more suitable to efficiently identify those with the main effects on the responses (Rodrigues and Iemma, 2015).

In addition, the set of evaluated factors included both qualitative and quantitative variables, a characteristic that further supports the application of the PBD for preliminary assessment of main effects. The investigated levels for each factor are presented in Table 2. A PB matrix of 12 trials was applied for each method of extraction (decoction and infusion), resulting in 24 trials.

TABLE 2. Factors and corresponding levels evaluated according to the Plackett–Burman design (PBD).

Applicable only to the decoction process.

2.5.2 Pilot-scale extraction using a central composite design (CCD)

Following the screening of key factors, the process was scaled up to pilot scale using equipment with a capacity 30 times greater than that employed at laboratory scale. It is important to note that scaling up may lead to significant changes in process behavior, as factors previously identified as critical may become non-critical, and vice versa.

Therefore, in order to evaluate and better understand process performance at a larger scale, a Central Composite Design (CCD) was conducted based on the factor levels presented in Table 3. The experimental assays were performed according to a design matrix consisting of six experiments, including factorial points and two central points. The reduced number of central points was justified by limitations in sample availability, since pilot-scale experiments require substantially higher consumption of plant material.

TABLE 3. Factors and corresponding levels evaluated according to the Central Composite Design (CCD).

2.5.3 Spray drying using a Central Composite Rotatable Design (CCRD)

Risk assessment also identified spray drying as a critical step in the process. Considering that a limited number of critical factors were identified for this stage, and that these correspond to continuous quantitative variables, a Central Composite Rotatable Design (CCRD) was selected as an optimization strategy.

This type of experimental design is suitable for response surface modeling, allowing the evaluation of linear, quadratic, and interaction effects between factors, as well as the identification of optimal operating conditions. The evaluated factors and their corresponding levels are presented in Table 4. The experiments were conducted according to a design matrix consisting of 17 assays, including factorial, axial, and central points.

TABLE 4. Factors and corresponding levels evaluated according to the Central Composite Rotatable Design (CCRD).

The value of ±α corresponds to ±1.682, according to the rotational rate assumed by the planning.

2.5.4 Statistical analysis

Data treatment and model validation were conducted using PROTIMIZA software. Prior to the statistical analysis, the assumptions required for parametric methods were evaluated by assessing the normality and homoscedasticity of the adjusted models. Subsequently, an analysis of variance (ANOVA) was performed to determine the significance of the factors and model terms. Effects were considered statistically significant when the p-value was ≤0.1. This significance threshold was deliberately adopted during the process development stage to reduce the risk of Type II errors (false negatives), which could lead to the exclusion of potentially influential factors. Although this approach may increase the likelihood of Type I errors, it allows a more comprehensive identification of variables that may affect process performance and therefore warrant further investigation and monitoring.

Mathematical models were adjusted through a sequential term elimination procedure, removing non-significant terms while monitoring the coefficient of determination (R ²) to ensure the model’s explanatory capacity remained high.

2.5.5 Desirability and design space

To identify the ideal operating conditions that satisfied multiple quality objectives simultaneously, the global desirability function was applied using Stat-Ease 360 software (version 25.0.1). In this approach, each individual response was transformed into a dimensionless desirability value ranging from 0 (completely undesirable) to 1 (completely desirable). The global desirability was then calculated to find the point within the experimental region where the combined performance of all CQAs reached its maximum value. This optimized condition served as the basis for the subsequent process validation trials.

2.6 Optimization and process validation

The factor levels applied in the validation experiments were defined based on previously established conditions aimed at meeting the target specifications for the set of response variables. This selection was guided by the parameter combinations identified during the modeling and optimization stages of the process, in order to experimentally verify the ability of the process to reproduce the proposed conditions.

Each validation condition was performed in triplicate to assess the reproducibility of the obtained results. When more than one validation condition was evaluated, the results were compared using Student’s t-test, performed in RStudio (version 4.4.0), adopting a significance level of p ≤ 0.05 (Posit team, 2025).

3 Results and discussion

3.1 Process overview

The overall process resulted in a cumulative yield of 17.6%, indicating that approximately 570 g of fresh leaves (dry basis) are required to obtain 100 g of granulated dry extract. This behavior reflects the cumulative nature of the unit operations involved, in which successive mass losses are expected throughout the different transformation stages, as shown in Table 5.

TABLE 5. Mass balance and yield distribution across process stages.

Abbreviations: : not applicable; PY: partial yield of each stage; CY: cumulative yield; HD: herbal drug.

Carrier solution: 20% (w/w) maltodextrin.

Binder solution: hydroalcoholic solution containing 2% (w/w) hydroxypropyl methylcellulose.

The progressive reduction in overall yield is mainly attributed to the extraction and spray drying steps, which account for the highest relative losses. Despite these reductions, the final yield is consistent with those typically reported for herbal medicine production processes, where selective extraction and phase transformations inherently limit mass recovery.

The extraction step was identified as the most critical stage of the process, with an average partial yield of 30.5% and a pronounced impact on cumulative yield. This behavior is associated with the intrinsic selectivity of the extraction process, in which only the water-soluble fraction of the plant matrix is recovered. As illustrated in Figure 2, these results in the removal of insoluble solids, representing an inherent and unavoidable mass loss characteristic of aqueous extraction processes (Siqueira et al., 2025).

From an operational perspective, drying emerged as the limiting step of the process due to its extended time (72 h), which is significantly longer than that required for the other stages. Although drying exhibited a high partial yield (93.5%), indicating efficient moisture removal with minimal solid loss, the prolonged processing time imposes constraints on production capacity, energy consumption, and overall process scalability.

3.2 Lab-scale extraction using Plackett–Burman design (PBD)

The screening phase was carried out through two separate 12 run Plackett–Burman designs. The physicochemical responses presented in Table 6 showed a wide variation in the extracted polyphenol content, ranging from 40.03 mg (2.84%, assay I24) to 145.33 mg (5.18%, assay D6). These results indicate that changes in individual factors or their interactions can lead to significant differences in the extracted dose, potentially impacting the daily dosage administered to the patient. A similar amplitude of variation was observed for total tannins and flavanol content.

TABLE 6. Plackett-Burman matrix with responses studied in lab-scale extraction.

Abbreviations: x₁: Drug-to-extract ratio (%); x₂: Form; x₃: Extraction vessel; x₄: Water quality; x₅: Exposure; x₆: Boiling time (min); x₇: Agitation; x₈: Extraction time (min). Y₁: Total Polyphenols (TP); Y₂: Total Tannins (TT); Y₃: epicatechin content; Y₄: catechin content; Y₅: process step yield; Y₆: process recovery; Y₇: Process time ; I: infusion; D: decoction.

*x₆ (Boiling time) is not applicable to the infusion process. The experiments were carried out strictly following the PBD, matrix, ensuring appropriate randomization and correct allocation of factor levels. Proper implementation of this experimental design is essential to guarantee reliable estimation of main effects, preserving model resolution and avoiding bias in the interpretation of results.

The main effects analysis for extraction by decoction and infusion are presented in Table 7. These results indicate that process performance is primarily influenced by a limited set of factors, which consistently affect both extraction yield and the chemical composition of the extracts.

TABLE 7. Analysis of the effects on the physicochemical responses of lab-scale extraction.

Values in bold indicate statistical significance at p < 0.1.

Abbreviations: Avg: Average; x₁: Drug-to-extract ratio (%); x₂: degree of fragmentation; x₃: Extraction vessel; x₄: Water quality; x₅: Exposure; x₆: Boiling time (min); x₇: Agitation; x₈: Extraction time (min). Y₁: Total Polyphenols Content (TPC); Y₂: Total Tannins Content (TTC); Y₃: epicatechin content; Y₄: catechin content.

*x₆ (Boiling time) is not applicable to the infusion process.

Considering that the experiments were conducted in accordance with the recommendations of the Brazilian Pharmacopoeia Herbal Medicine Formulary (Brasil, 2021), the observed variation in extracted polyphenol content directly reflects the unit dose administered to the user. Therefore, the results suggest that variations in preparation conditions may lead to substantially different doses, potentially compromising treatment efficacy.

The operational responses, also showed variability in solid yield (Y₆) and processing time, providing further insight into the extraction step. To establish relationships between the observed responses and specific process factors, the data were subsequently subjected to effect analysis.

Overall, the form/particle size of the herbal drug (x₂) was the factor that most consistently influenced the evaluated responses. The powdered form resulted in positive effects on extraction yield and on the content of polyphenols, tannins, and flavanols (epicatechin and catechin). This behavior can be attributed to the increased contact surface area between the plant material and the solvent, which enhances mass transfer and facilitates the release of metabolites (Wijesekera, 1991; Pereira et al., 2023).

The drug-to-extract ratio (DER, x₁) showed a significant effect on the values (mg) of the extracted compounds, indicating that the more concentrated proportions lead to higher recovery metabolites results. However, the relative extraction (%) was not significantly affected. This finding highlights the importance of jointly evaluating absolute and relative responses, since the former is directly proportional to the amount of herbal drug used.

The type of vessel (x₃) showed a significant effect in decoction for total tannins (%) and epicatechin (mg and %), suggesting a possible interaction between the extraction vessel material and phenolic compounds. Similarly, water quality (x₄) exhibited specific effects on epicatechin and catechin, indicating that differences in ionic composition may selectively influence the extraction or stability of these flavanols. In addition, the observed variation in catechin and epicatechin profiles may also be associated with possible epimerization processes, since flavanols can undergo stereochemical interconversion under thermal conditions, leading to shifts in the relative abundance of these isomers (Huang et al., 2025).

The remaining operational factors such as exposure (x₅), boiling time (x₆), agitation (x₇), and extraction time (x₈) generally showed minor and non-significant effects for most responses. These results suggest that, within the studied ranges, such variables play a secondary role in determining the final product characteristics.

Taken together, the results indicate that particle size of the herbal drug and DER are the critical elements of the process (CMA and CPP), particularly with respect to the recovery of phenolic compounds, while the remaining factors contribute to a lesser extent.

3.3 Pilot-scale extraction using a central composite design

Following the screening of critical factors at laboratory scale, the extraction process was scaled up using a central composite design (CCD), comprising six experimental assays and evaluating two selected factors.

Based on the physicochemical responses, total polyphenol content ranged from 3.12% (assay E2) to 4.26% (assay E3), indicating that variations in individual factors or their interactions may lead to more efficient extraction conditions for routine production. A similar range of variation was observed for total tannins and flavanol content, as shown in Table 8. The extraction kinetics are represented in Supplementary Figure S1.

TABLE 8. Central Composite Design with responses studied in pilot-scale extraction.

Abbreviations: x₁: Drug-to-extract ratio; x₂: Agitation. Y₁: Total Polyphenols Content (TPC); Y₂: Total Tannins Content (TTC); Y₃: epicatechin content; Y₄: catechin content; Y₅: process step yield; Y₆: process recovery; Y₇: process time.

These results also indicated that all extraction conditions resulted in products meeting the specifications for total tannins and epicatechin, as established by the Brazilian Pharmacopoeia (>2.0% and >0.28%, respectively). It was consistent with those reported by Siqueira et al., who obtained 3.27% ± 0.02% of total tannins and 0.683% ± 0.03% of epicatechin in aqueous extracts of M. ilicifolia (Brazil, 2024; Siqueira et al., 2025).

The regression models and their statistical significance are summarized in Table 9, which presents the corresponding equations, coefficients of determination (R ²), and calculated and tabulated F-values. Models in which the calculated F-values exceeded the tabulated values at a significance level of 10% were considered statistically significant and were used to establish correlations with the studied factors. Additional model validation metrics, including complete ANOVA, lack-of-fit tests, adjusted R ², and predicted R ² values, are presented in Supplementary Table S3, providing further evidence of model adequacy and predictive performance.

TABLE 9. Models and summaries of analysis of variance for pilot-scale extraction.

Abbreviations: x₁: Drug-to-extract ratio (%); x₂: Agitation. Y₁: Total Polyphenols Content (TPC); Y₂: Total Tannins Content (TTC); Y₃: epicatechin content; Y₄: catechin content; Y₅: process step yield; Y₆: process recovery; Y₇: process time.

The contour maps shown in Figure 3a, generated from the statistical models, allowed the integrated evaluation of the effects of DER and agitation on the responses. Both parameters exhibited positive effects on the absolute extraction of chemical markers; however, DER showed a negative effect on the relative (%) content of extracted compounds, suggesting a solvent saturation phenomenon.

The plots presented in Figure 3b illustrate the correlation between experimental and predicted values for the evaluated responses, allowing the assessment of the predictive performance of the fitted models. A good agreement between values was observed, with data points distributed close to the identity line (y = x), indicating adequate model fitting and low data dispersion. Minor deviations around the line suggest the presence of residual errors inherent to the process, without compromising the representativeness of the models.

3.4 Spray drying and granulation using a Central Composite Rotatable Design (CCRD)

The spray drying and granulation processes, were conducted according to a Central Composite Rotatable Design (CCRD) comprising 17 experimental assays and evaluating three process factors. After obtaining the spray-dried extract, each assay was subjected to granulation, which improved the technological properties of the material, including flowability, compressibility, hygroscopicity, and wetting time. The average yield of the spray drying stage was 73.9%, while granulation showed a higher yield of 92.6%, as demonstrated in Table 10.

TABLE 10. Central Composite Rotatable Design with responses studied in spray drying and granulation.

Abbreviations: x₁: Maltodextrin concentration (%); x₂: Feed rate (mL/min); x₃: Input temperature (°C). Y₁: Total Polyphenols Content (TPC); Y₂: Total Tannins Content (TTC); Y₃: epicatechin content; Y₄: catechin content; Y₅: process step yield; Y₆: process recovery; Y₇: process time; Y₈: residual moisture; Y₉: Initial bulk density (ρ B m i n); Y₁₀: Final bulk density (ρ B m a x); Y₁₁: Carr index (%CI); Y₁₂: Angle of repose (θ); Y₁₃: Hygroscopicity (%H); Y₁₄: Wetting time; Y₁₅: Color evaluation; SG: spray drying and granulation.

Granulation was performed using hydroxypropyl methylcellulose (HPMC) as a binder due to its compatibility with tannin-rich plant matrices. Unlike commonly used binders such as polyvinylpyrrolidone (PVP), which may form insoluble complexes with tannins and compromise the availability of chemical markers, HPMC provides improved stability and functionality for such systems (Rowe et al., 2006).

The integrated evaluation of these responses, shown in allowed the identification of experimental conditions capable of meeting predefined requirements, supporting the subsequent stages of modeling, optimization, and process validation.

Based on the physicochemical responses of the granulated extracts, total polyphenol content ranged from 6.07% (assay SG10) to 14.24% (assay SG9). This variation indicates that the addition of maltodextrin leads to dilution of the dry extract, thereby increasing the unit dose required for the final product. A similar trend was observed for total tannins and flavanol content.

Although the Brazilian Pharmacopoeia, seventh edition (FB7), does not specify the total content of tannins and epicatechin in granulated extracts, the results obtained in this study were in agreement with the data from the literature. Siqueira et al. reported epicatechin contents of 2.91% ± 0.71% in spray-dried extracts of M. ilicifolia, supporting the validity of the obtained values (Brazil, 2024; Siqueira et al., 2025).

For epicatechin content, total polyphenols, and total tannins, it was observed that their contents followed the overall solid yield, indicating that losses are predominantly associated with physical and not to chemical process. This behavior suggests that the spray drying conditions employed were suitable for preserving the herbal matrix, likely due to the short residence time and rapid solvent evaporation. In contrast, catechin showed higher concentrations in the product, suggesting a possible conversion from its precursor compounds.

Regarding technological responses, the results show that processes conducted with lower proportions of maltodextrin resulted in denser powders. This finding has relevant logistical implications, as higher bulk density reduces volume per unit mass, improving storage and transportation efficiency. Consequently, obtaining a denser powder may represent an operational and economic advantage, particularly in large-scale production scenarios.

Figure 4 presents the powder cones formed in each assay for angle of repose determination. The top-view images allow visualization of differences in color among the granulated extracts and provide qualitative insight into their flow properties. Lighter-colored samples, associated with higher maltodextrin content, formed larger cones, indicating improved flowability.

The combined evaluation of angle of repose and Carr index enabled the identification of conditions capable of simultaneously meeting desirable flowability and compressibility criteria. Acceptance ranges were defined as Carr index values below 25% and angle of repose below 40°. Although no exact overlap between optimal conditions was observed, assays SG1 and SG9 showed performance close to the target ranges, indicating satisfactory technological behavior.

The hygroscopicity behavior of the granulated dry extracts over time exhibited a sharp increase in hygroscopicity during the initial hours, particularly within the first 24 h, followed by a stabilization trend, indicating rapid initial moisture uptake and subsequent equilibrium. Some assay presented hygroscopicity values exceeding the upper limit of 10.0%, suggesting greater sensitivity to environmental moisture.

Moisture content (%) of the spray-dried extracts remained below the upper limit of 5% for all experimental conditions, demonstrating that the process parameters were adequate to produce a dry product within the established criteria. Moisture values ranged from 0.65% (assay SG5) to 4.60% (assay SG3), reflecting differences between experimental conditions due to variations in process parameters.

Maintaining moisture content below 5% is critical for the stability of dry extracts, as higher water levels may promote hydrolysis, chemical degradation, microbial growth, and deterioration of technological properties such as flowability. Therefore, the results indicate that, regardless of the specific parameter configuration, the spray drying process was effective in producing a stable and technologically suitable final product (Mutavski et al., 2025).

The regression models and their statistical significance are summarized in Table 11, which presents the fitted equations, coefficients of determination (R ²), and calculated and tabulated F-values. In cases where the calculated F-values exceeded the tabulated values at a significance level of 10% (p < 0.1), the models were considered statistically significant, allowing the establishment of relationships between the response variables and the investigated factors. Additional model validation metrics, including complete ANOVA, lack-of-fit tests, adjusted R ², and predicted R ² values, are presented in Supplementary Table S4, providing further evidence of model adequacy and predictive performance.

TABLE 11. Models and summaries of analysis of variance for spray drying and granulation.

Abbreviations: x₁: Maltodextrin concentration (%); x₂: Feed rate (mL/min); x₃: Input temperature (°C). Y₁: Total Polyphenols Content (TPC); Y₂: Total Tannins Content (TTC); Y₃: epicatechin content; Y₄: catechin content; Y₅: process step yield; Y₆: process recovery; Y₇: process time; Y₈: residual moisture; Y₉: Initial bulk density (ρ B m i n); Y₁₀: Final bulk density (ρ B m a x); Y₁₁: Carr index (%CI); Y₁₂: Angle of repose (θ); Y₁₃: Hygroscopicity (%H); Y₁₄: Wetting time; Y₁₅: Color evaluation.

Models with Fcalc < Ftab were considered not statistically significant and therefore not validated.

The contour plots shown in Figure 5a illustrate the combined effects of maltodextrin concentration, inlet air temperature, and feed rate on the technological properties of the granulated extract. In parallel, the plots presented in Figure 5b show the correlation between experimental and predicted values for each response, enabling the evaluation of the predictive performance of the fitted models. The good agreement observed between these values confirms the reliability of the regression equations in describing the behavior of the responses within the studied experimental domain.

Maltodextrin concentration and inlet air temperature were found to significantly influence final bulk density, with direct implications for downstream operations such as storage, transportation, and packaging. In particular, an increase in tapped density indicates a product that occupies less volume for a given mass, contributing to improved space utilization during storage. This characteristic directly supports logistical optimization, reducing transportation costs and facilitating handling throughout subsequent stages of the production chain.

The Carr index was also influenced by maltodextrin concentration and inlet air temperature. Increases in these factors resulted in higher Carr index values, indicating reduced compressibility and poorer flow properties. This behavior may be associated with the formation of more cohesive particles or an increased tendency toward particle compaction. The observed variations in these parameters may be partially explained by the concentration of maltodextrin, which is known to influence the glass transition temperature and surface characteristics of spray-dried particles. Higher maltodextrin contents can modify particle morphology, reduce stickiness, and alter interparticle interactions, thereby affecting powder cohesiveness and flowability (Dupas et al., 2024). These changes may consequently influence bulk and tapped densities, resulting in variations in Carr Index values.

For the angle of repose, maltodextrin concentration was identified as a significant factor. Lower angle values are associated with improved flowability, whereas higher values indicate increased particle cohesiveness. This parameter is particularly relevant for industrial operations, as it directly affects powder flow through processing equipment. Therefore, proper control of process conditions influencing the angle of repose is essential to ensure process reliability and efficiency, as well as to obtain a product with suitable technological properties.

Regarding color attributes (CIELAB system), the contour plots indicate a pronounced effect of maltodextrin concentration on increasing lightness (L*), as well as an interaction between feed rate and inlet air temperature. These findings highlight the importance of a multivariate approach to process understanding, as both sensory and technological attributes can be simultaneously influenced by multiple process parameters.

3.5 Desirability and design space

The desirability analysis simultaneously evaluated the behavior of technological responses as a function of the combined levels of material attributes and process parameters (CMAs and CPPs). Based on this approach, an operational region was defined in which the process meets the target objectives, as described in Table 1.

Figure 6 presents the individual desirability profiles for the evaluated factors, as well as the contour plot of global desirability. The simultaneous optimization of the responses indicated a region with relatively low global desirability (<0.7), centered approximately at A = 0.0% (maltodextrin), B = 4.30 mL/min (feed rate), and C = 140.9 °C (inlet air). This result suggests limitations in achieving a Design Space for all evaluated responses simultaneously.

3.6 Optimization and process validation

Due to the limitations observed in the desirability analysis, two validation conditions were evaluated. Condition 1 was defined based on the factor levels corresponding to assay SG9 (without maltodextrin addition, with a feed rate of 20% and an inlet air temperature of 160 °C), as it represents a combination close to the region indicated by the desirability results obtained from the models. In contrast, condition 2 (45% maltodextrin, a feed rate of 5.4 mL/min, and an inlet air temperature of 150 °C) was established with the aim of reducing product hygroscopicity and improving flowability.

Table 12 demonstrates that no significant differences were observed in process yield between the evaluated conditions, indicating that the addition of maltodextrin and adjustments to the operational parameters did not affect the overall efficiency of material recovery.

TABLE 12. Comparison of Spray Drying and Granulation process validation results between Condition 1 and Condition 2 using Student’s t-test.

Values in bold indicate statistical significance at p < 0.05.

Abbreviations: SD: standard deviation.

No significant differences were observed in process yield between the evaluated conditions, indicating that maltodextrin addition and adjustments in operational parameters did not affect the overall efficiency of material recovery. Similarly, the retention rates of chemical markers across process stages did not differ significantly between conditions, demonstrating that, despite differences in formulation and processing parameters, the relative stability of the chemical markers was maintained throughout the process. These findings indicate that the modifications introduced in condition 2 did not compromise the retention of bioactive compounds during spray drying and granulation, reinforcing the robustness of the process.

In contrast, statistically significant differences were observed for total polyphenols, total tannins, epicatechin, and catechin contents, with higher values obtained under condition 1. This behavior is directly associated with the dilution effect promoted by the addition of maltodextrin in condition 2, which reduces the relative concentration of chemical markers in the final product without necessarily affecting their retention during processing.

The physical properties of the powders also showed significant differences between the evaluated conditions. Initial and final bulk densities were higher in condition 1, resulting in a significantly lower Carr index, indicative of better compressibility. However, condition 2 exhibited higher Carr index values, consistent with the presence of maltodextrin and its influence on particle formation.

Regarding color attributes, significant differences were observed for lightness (L*). These results indicate that maltodextrin addition increased product lightness and altered its color profile.

4 Conclusion

Within a QbD framework, the integration of process mapping, risk assessment, and Design of Experiments enabled a systematic understanding of the relationships between CMAs, CPPs, and CQAs across the production chain. This approach provided a scientific basis for defining operational conditions that ensure consistent product quality and offered evidences for further establishment of a Design Space in accordance with QbD principles.

Extraction was identified as a critical step, with particle size and Drug to Extract Ratio significantly influencing the recovery of chemical markers. The results also demonstrated that process optimization should consider not only physicochemical responses but also operational and economic aspects due to the absolute and relative extraction.

Desirability analysis of the spray drying stage indicated limitations in achieving optimal conditions for all responses simultaneously, highlighting trade-offs between technological attributes. Process validation demonstrated that different operating conditions can be used to adjust properties such as compressibility, flowability, and hygroscopicity.

Overall, the results support the application of QbD to improve process understanding and guide the development of scalable manufacturing routes for herbal medicines. In this context, the incorporation of relevant granulation technologies, such as high-shear granulation and fluidized bed granulation, should be considered in future scale-up studies to enhance process reproducibility and industrial applicability. Future research should also investigate the variability of raw materials from different suppliers and focus on long-term and accelerated stability assessments. Additionally, incorporating biological activity evaluations will be crucial to further confirm the therapeutic relevance of the developed product.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fntpr.2026.1887810/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Hector Henrique Ferreira Koolen, University of the State of Amazonas, Brazil

Reviewed by: Askur Rahman, University of Trunodjoyo Madura, Indonesia

Jyoti Dahiya, Central Council for Research in Ayurvedic Science, India

Footnotes

ANOVA, One-way analysis of variance; CCD, Central Composite Design; CCRD, Central Composite Rotatable Design; CMA, Critical Material Attributes; CPP, Critical Process Parameters; CPQBA, Multidisciplinary Center of Chemical, Biological, and Agricultural Research; CQA, Critical Quality Attributes; DER, Drug to Extract Ratio; DoE, Design of Experiments; DS, Design Space; FC, Folin-Ciocalteu; PBD, Plackett–Burman Design; QbD, Quality by Design; QTPP, Quality Target Product Profile; SD, Standard deviation; TPC, Total Polyphenols Content; TTC, Total tannins content.