A44-30 Simplified Assessment of Inhaler Capability: A Predictive Model for Peak Inspiratory Flow and Suboptimal Inhalation Risk Across Device Resistances

Abstract Introduction Inhaled therapy is essential for chronic pulmonary disease management, but is often compromised due to insufficient peak inspiratory flow (PIF). Current multi-measurement PIF assessment methods are clinically laborious. This study developed a mathematical model to predict PIF value and suboptimal PIF risk from minimal measurements, thereby simplifying inhalation capability evaluation. Methods This study included data from 281 participants, comprising In-Check DIAL measurements of PIF at resistance (R) 1-3, spirometry parameters, and demographic characteristics. The cohort was randomly divided into training (70%) and validation (30%) sets. Two models were developed: (1) multiple linear regressions to predict PIF values at R1 and R3; (2) a logistic regression to identify suboptimal PIF probability (PIF 60 L/min at R3). Models were developed using stepwise regression and validated using R² for linear regression and area under the curve (AUC) for logistic regression. Results In the general cohort, the multiple linear regression model for R1 PIF was: R1PIF = 13. 108 + 0. 888 * R2PIF + 0. 630 * PEF + 0. 088 * Height, with an R² of 0. 774 upon validation. The model for R3 PIF was: R3PIF = 3. 941 + 0. 825 * R2PIF + (-0. 079) * FEV1/FVC, with an R² of 0. 771 upon validation. Better R2 were observed in patients with chronic obstructive pulmonary disease (COPD) (R2=0. 872 for R1 and 0. 833 for R3) than other cases (R2=0. 721 for R1 and 0. 740 for R3). The subgroup of patients with COPD (n = 67 in training set and 29 in validation set) was further analyzed. The model for R1 PIF was: R1PIF = 28. 700 + 0. 873 * R2PIF + 2. 681 * FEF50, with an R² of 0. 903 upon validation. The model for R3 PIF was: R3PIF = 4. 375 + 0. 768 * R2PIF, with an R² of 0. 865 upon validation. In the general cohort, 28% participants exhibited suboptimal PIF at R3. The logistic model for suboptimal PIF risk at R3 incorporated the value of PIF at R2 and FEV1/FVC and achieved an AUC of 0. 911 on the validation set. Conclusions These models enabled accurate PIF and risk prediction from a single measurement, especially in patients with COPD, overcoming multi-step limitations to guide personalized inhaler selection and optimize disease management. This abstract is funded by: Shanghai Municipal Science and Technology Major Project (ZD2021CY001), Science and Technology Commission of Shanghai Municipality (20Z11901000, Shanghai Municipal Key Clinical Specialty (shslczdzk02201)