To maximize the therapeutic efficacy of inhaled medications and minimize adverse effects, it is essential to assess not only the total administered dose but also the regional distribution of drug deposition within the lungs (Beinart et al., 2024; Lazarinis et al., 2024; Cottini et al., 2025). In patients with asthma and chronic obstructive pulmonary disease (COPD), adequate drug distribution to the peripheral airways has been reported to be significantly associated with improved symptom control, reduced exacerbation frequency, and long-term maintenance of lung function (Mastalerz and Kasperkiewicz, 2011; Usmani et al., 2021; Clarà et al., 2023; Darquenne et al., 2024). Conversely, excessive oropharyngeal deposition of inhaled corticosteroids (ICS) is known to cause local adverse events, such as oral candidiasis and dysphonia, and systemic side effects through gastrointestinal absorption following swallowing (Dubus et al., 2001; Kaur and Singh, 2018; Arevalo and Buban, 2023; Gümrü and Akkitap, 2022). These adverse effects can lead to reduced patient adherence, which is a critical issue in clinical practice (Engelkes et al., 2015; Asamoah-Boaheng et al., 2021). Therefore, a method for quantifying and optimizing site-specific deposition on an individual basis is highly desirable.

Inhalation therapy is the initial treatment for asthma and COPD (Global Initiative for Asthma - GINA, 2025; Global Initiative for Chronic Obstructive Lung Disease, 2025). By enabling the direct delivery of medications to target sites, this approach facilitates a rapid onset of action while reducing systemic exposure and associated adverse effects (Chandel et al., 2019; Wang et al., 2024). Dry powder inhalers (DPIs) are widely used in clinical settings owing to their superior physicochemical stability and ability to achieve efficient pulmonary deposition through patient-generated inspiratory flow (Dhoble et al., 2024; Islam et al., 2024). However, the actual dose and distribution of drugs delivered to the lungs can vary substantially owing to multiple factors, including respiratory function, inhalation flow profiles, device design, formulation properties, and operational errors (Borghardt et al., 2018; Haidl et al., 2016; Nováková et al., 2022).

Previous attempts to predict therapeutic efficacy have used plasma drug concentration monitoring (Alahmadi et al., 2021). Although correlations with clinical outcomes have been observed, such approaches are limited by considerable interpatient and inter-formulation variability, making it difficult to selectively quantify pulmonary drug deposition (Getz and Carroll, 2017). In vivo imaging techniques using radiolabeled drugs have enabled the quantification of regional deposition; however, their invasive nature and procedural complexity restrict their utility to regulatory studies rather than routine clinical practice (Meseguer et al., 1994; Newman et al., 2000). Therefore, non-invasive methods such as in vitro and in silico testing are being explored. These studies have substantially advanced our understanding of how inhalation maneuver and device–powder interactions affect aerosol generation and regional deposition. However, these approaches typically depend on predefined inhalation patterns and airway geometries (Rebello et al., 2022). There is limited visibility into day-to-day, patient specific inhalation behavior in real-world clinical practice, where technique errors and self-adapted maneuvers may occur. Although the development of digital health integrated smart inhalers is underway, there is as yet no established method to directly evaluate drug deposition by accounting for individual inhalation patterns and airway anatomy (Bosnic-Anticevich et al., 2023). Therefore, practical methods are required to assess site-specific drug delivery based on real-time and non-invasive indices obtained during actual patient inhalation.

The results of our previous study demonstrated that a particle emission signal derived from the photo reflection method (PRM) can serve as non-invasive, and real-time indices for total pulmonary drug deposition via Symbicort^® Turbuhaler^® (Hatazoe et al., 2024). In PRM, the reflected light is measured as the aerosol released from an inhaler passes through the laser sensing region. Symbicort^® Turbuhaler^® is widely used as a combination therapy of an inhaled corticosteroid (budesonide) and a long-acting beta2-agonist (formoterol fumarate) for the treatment of asthma and COPD. Symbicort^® is characterized as a formulation where micronized particles (1–5 μm) of the two active ingredients and lactose form soft agglomerates (Buttini et al., 2016). It has been reported that these budesonide and formoterol exhibit comparable aerodynamic behavior in terms of the fine particle fraction (FPF). However, because the therapeutic and adverse effects of inhaled drugs originate from drug deposition in different anatomical regions, a detailed analysis of drug deposition sites in the respiratory system is necessary. Therefore, the aim of the present study was to develop a predictive model capable of simultaneously estimating drug delivery to the peripheral airways, central airways, and oropharyngeal regions using an established PRM-based monitoring system.

The profiles of the inhalation flow parameters and particle emission signal indices for each inhalation pattern are summarized in Table 1. As PFR and FIR increased, both Peak PE and AUC_FR×PE showed a corresponding increase, with the highest values observed under the Quick60 condition, which featured the most rapid increase in flow rate.

The inhalation performance parameters for all inhalation patterns are shown in Table 2. The OE of the Quick30 pattern was significantly lower than that of the other PFR60 patterns. Both the FPF_WL and FPF_PA increased with higher PFR and FIR, ranging from 15.2% to 57.2% and from 3.1% to 11.9%, respectively. Drug deposition in the oropharynx ranged from 10.5% to 36.3%, with higher values observed under Quick60 (p < 0.05 vs. Mild60 and Quick30).

The aerodynamic particle distributions obtained from ACI are shown in Figure 2. In the Quick30 condition, deposition at Stage 0—which corresponds to the pharyngeal region—was higher than that in the other patterns. Conversely, deposition in the induction port representing the oral cavity increased markedly with higher PFR and FIR. Three patterns with a defined PFR of 60 L/min exhibited peak deposition in Stage 3, corresponding to the central airways. By contrast, Quick30 showed a shift in peak deposition to a deeper region at Stage 4. These trends are reflected in Figure 3, which compares deposition indices across inhalation patterns, indicating that both pulmonary delivery and oropharyngeal deposition increased concurrently under high PFR and FIR conditions.

SEM images of Symbicort^® Turbuhaler^® particles obtained before and after inhalation are shown in Figure 4. The undispersed aggregates were observed inside the device before inhalation (Figure 4A). Panels B–G show particles deposited on the stages 0 (pharynx region), Stage 3 (central airway) and Stage 5 (peripheral airway) of ACI, respectively, after a single inhalation. The upper images correspond to Quick60, and the lower images correspond to Quick30. Both inhalation patterns, aggregated drug particles larger than 10 μm were observed at stage 0. Partial disaggregation was evident in Stage 3, whereas Stage 5 showed well-dispersed fine particles. However, in Quick30, coarser agglomerates were observed in Stage 0. In addition, Figure 5 shows the geometric particle size distribution of particles deposited on Stage 3 using the AI-based particle size analysis. The image analysis appropriately identified the deagglomerated primary particles (Figures 5A,B). Furthermore, a sharp particle size distribution was confirmed, with a median particle diameter (D₅₀) of 2.82 μm.

The results of the univariate regression analyses for model fitting are shown in Figure 6. In Table 3, we list the best-fit model types and associated equations for each inhalation performance. Furthermore, model evaluation metrics, including the AIC, RMSE, and MAE for all univariate regression models, are summarized in Table 4. In the case of FPF_WL and FPF_PA, linear models with PFR exhibited strong correlations (R ² = 0.81–0.82), whereas AUC_FR×PE showed a better fit with the Hill model, achieving the highest R ² of 0.86 for FPF_WL. In addition, a high correlation was confirmed between FR at PE peak end and the Hill model. By contrast, oropharyngeal indices were significantly predicted by the linear model with FR at PE peak top/end, and the Hill model with FIR. Whereas the logX model using AUC_FR×PE and linear model using PFR also showed moderate predictive power.

We evaluated the impact of inhalation pattern on site-specific drug deposition and identified optimal predictors for each site in the respiratory tract. High PFR and FIR enhanced pulmonary drug delivery while also increasing oropharyngeal deposition. Univariate regression revealed that the optimal predictors differed by deposition site, indicating the need for site-specific evaluation in modeling inhalation performance.

DPIs are designed to contain agglomerated drug particles or drug-carrier complexes to avoid adhesion to internal surfaces. Turbulent energy from inhalation facilitates deagglomeration and generates fine particles with aerodynamic diameters suitable for peripheral airway delivery (Islam et al., 2024; Zheng et al., 2021). After drug release, these particles deposit through inertial impaction or gravitational sedimentation, exerting pharmacological effects (Demoly et al., 2014). This mechanism is particularly relevant for Turbuhaler^®, which reportedly requires a PFR of at least 60 L/min to achieve adequate dispersion (Haidl et al., 2016). Consistent with this device feature, the overall emitted dose from Turbuhaler^® has been reported to increase with inhalation flow rate, typically reaching approximately from 40% to 80% (Farkas et al., 2023). In the present study, our OE showed a comparable flow rate dependency trend (Table 2). Moreover, FPF_WL and FPF_PA were significantly higher at 60 L/min than at 30 L/min, with the maximum FPF_WL observed being 57.2% under Quick60. SEM images (Figure 4) and particle size distribution analysis (Figure 5) showed dispersed particles at ACI Stages 3 and 5 under Quick60, whereas large agglomerates remained at Stage 0 under Quick30. Additionally, drug deposition at ACI Stage 0 was markedly higher under Quick30 than under the other inhalation patterns (Figure 2). These findings from SEM and ACI data collectively indicate that lower flow rates result in insufficient deagglomeration and the emission of coarse agglomerates, whereas high PFR facilitates progressive deagglomeration along the impactor stages and promotes distal lung delivery.

We found that the value of FPF_WL ranged from 15.2% to 57.2%, depending on the inhalation pattern. In previous in vivo studies with Turbuhaler^®, investigators have reported pulmonary deposition values of 26.1% ± 10.5% in adult patients with asthma (Thorsson et al., 1998) and a range of 15.6%–47.2% in pediatric patients (Wildhaber et al., 1998). The lower end of the range of values in our study is consistent with these previous reports; however, the upper end exceeded what was previously reported. In clinical settings, 19%–78% of patients with asthma or COPD fail to achieve a target PFR of 60 L/min during DPI use (Broeders et al., 2001; Haikarainen et al., 2020; Bagherisadeghi et al., 2017; Weers, 2022). Therefore, ideal conditions, as in the Quick60 inhalation pattern, may not be replicable in real-world practice, contributing to differences between in vitro and in vivo findings.

The FPF_PA in this study ranged from 3.1% to 11.9%, which is lower than the reported 8.6%–24.8% range from in silico analyses of ICS delivery with Turbuhaler^® (Watz et al., 2021). This underestimation likely results from limitations of the ACI device. The ACI is designed for standard flow rates (28.3 L/min), and its aerodynamic cut-off diameters shift depending on the applied flow. At higher flow rates, enhanced inertial forces reduce the cut-off diameters of each stage, leading to premature capture of fine particles in upstream stages (Yoshida et al., 2017; Kamiya et al., 2009; Almeziny et al., 2025; van Holsbeke et al., 2018). In the present study, Oral-Pharynx increased under the higher PFR and FIR conditions (Figure 3A), and peak deposition shifted from Stage 3 to Stage 4 under Quick30 (Figure 2). These results suggest the occurrence of this inertial force impaction phenomenon. Conversely, because the deagglomeration of drug aggregates increases under high flow rate conditions, the Turbuhaler^® yielded higher FPF_PA results at higher flow rates (Figure 3C). As site-specific deposition of DPI formulations is determined by the balance between deagglomeration efficacy and inertial impaction, it is crucial to predict site-specific deposition across a wide range of inhalation patterns. Although this may lead to underestimation of peripheral delivery in absolute terms, the evaluation by ACI grouping remains valuable for assessing the relative influence of inhalation patterns on site-specific drug deposition.

Additionally, increased FIR was associated with elevated drug deposition in the oropharyngeal region, particularly under Quick60 conditions. Oropharyngeal deposition was higher at steeper flow acceleration, indicating that excessive FIR may enhance drug impaction in the upper airway. This phenomenon is consistent with the results of previous in silico studies that showed that particles in the 5–10 μm range are prone to inertial impaction at elevated flow rates (Sosnowski et al., 2006; Kadota et al., 2022). The oropharyngeal deposition we observed in the present study (10.5%–36.3%) was lower than that reported in previous in vivo studies with Turbuhaler^®, where deposition to the oropharyngeal region ranged from 12.8% to 61.8% (Wildhaber et al., 1998; Hirst et al., 2001; Devadason et al., 1997). Such variability among studies is likely influenced by inter-individual differences in inhalation techniques, device handling, tongue positioning, and pharyngeal anatomy (Sosnowski et al., 2006; Xi and Yang, 2019). In the present study, we used the USP induction port (USP-IP), ACI cone and Stage 0 to Stage 2 as the surrogates for oropharyngeal regions. Although the USP-IP lacks anatomical fidelity, combining it with Stage 0 to Stage2 and applying cooling to reduce re-scattering (Zhou et al., 2011) enabled detection of flow-dependent differences in oropharyngeal deposition.

To identify the predictors of drug deposition, univariate regression analysis was performed using the inhalation flow rate and particle emission signal. Generally, for Turbuhaler^®, FPF<5 µm increases with inspiratory flow rate, although several studies have reported that the incremental gain in FPF<5 µm diminishes at excessively high flow rates (Farkas et al., 2023). This behavior is considered to reflect a shift from disaggregation by shear stress within the device to enhanced inertial impaction in the upper airways. On this basis, we assumed that particle emission signal indices representing drug release may also exhibit nonlinear behavior. Accordingly, we adopted four different mathematical models (linear, logarithmic, Hill, and Emax) to capture such potential nonlinearities. As summarized in Table 3, pulmonary deposition indices (FPF_WL and FPF_PA) correlated strongly with actual PFR, AUC_FR×PE, and FR at PE peak end. These predictors reflect the intensity and duration of inhalation and drug release, suggesting that sustained inhalation efforts may be related to lung deposition. By contrast, oropharyngeal deposition showed strong correlation with the parameters related to the early phase of inhalation, namely, FR at PE peak top/end and FIR. Unlike time-averaged indices such as AUC_FR×PE, FIR and FR at PE peak top/end are reflective of initial flow acceleration and the onset of turbulent airflow, both of which promote inertial impaction within proximal-airway structures.

Notably, AUC_FR×PE was a significant predictor for lung deposition. Symbicort^® Turbuhaler^® contains micronized drug particles blended with fine lactose particles of size <10 μm, which are loaded as large and soft agglomerates (∼0.5 mm in size) (Leung et al., 2015). Accordingly, the particle emission signals are likely to reflect light from both drug and lactose particles. As a result, even if lactose contributes to the particle emission signals, drug particle release involves a stepwise deagglomeration process that is dependent on the level of shear stress, making it difficult to describe the relationship using a simple linear model. Whether similar relationships between site-specific drug deposition and particle emission signals can be observed for DPIs with different formulation designs and deagglomeration mechanisms remains to be investigated.

By contrast, oropharyngeal deposition was best described by a linear model with FR at PE peak top. This likely reflects a process dominated by inertial impaction of particles on the upper airway walls, which increases proportionally with the magnitude of initial inhalation flow. In contrast to pulmonary deposition, which involves nonlinear processes such as stepwise deagglomeration, oropharyngeal deposition followed a more direct relationship in which the strength of inhalation was linearly associated with the amount of deposition. These findings highlight the need for site-specific model selection that captures the mechanisms underlying aerosol deposition, even when using the same predictors.

Traditional lung deposition predictions rely primarily on aerodynamic particle size and inhalation flow rates (Horváth et al., 2020; Longest and Holbrook, 2012; Kim and Hu, 2006). Although these emphasize airflow characteristics, they do not consider real-time drug release kinetics. For instance, profiles with comparable PFR can yield different aerosol release depending on the early phase inspiratory dynamics. In contrast, particle emission indices, such as AUC_FR×PE, capture the aerosolization process in real time, improving the ability to potentially predict variability in site-specific deposition across individual inhalation patterns.

This study had several limitations. First, the present analysis was conducted using only Symbicort^® Turbuhaler^®. Given the flow resistance, the amount and size of drug and lactose differences across inhalers, these models need to be validated with other inhalers. Second, we restricted our analysis to univariate regression models. The class of PFR and FIR was limited, and the intercorrelations among flow rate and PRM indices could lead to overfitting and unstable parameter estimates. Future studies with larger and more diverse inhalation patterns and devices will be required to develop multivariate prediction models. Finally, the clinical relevance of the particle emission signal remains unvalidated, and further human studies are warranted to establish external validity.

In this study, we demonstrated that regional drug deposition could be predicted based on inhalation pattern–dependent flow characteristics and particle emission signal measurements. Indicators reflecting inhalation intensity were effective in predicting pulmonary delivery, whereas parameters representing early-phase flow acceleration were more predictive of oropharyngeal deposition. These findings suggest that distinct factors govern the drug distribution across anatomical sites. Thus, compared to conventional assessments based on individual parameters such as flow rate, real-time flow and emission signals may serve as useful tools for assessing inhalation performance.