Area under the expiratory flow-volume curve: predicted values by regression and deep learning methods and recommendations for clinical practice

• Type of Material: Article
• Authors: Octavian Ioachimescu, Emory University; José A Ramos, Cleveland Clinic; Michael Hoffman, Cleveland Clinic; James K Stoller, Cleveland Clinic
• Language: English
• Date: 2021-01-01
• Publisher: BMJ PUBLISHING GROUP
• Publication Version: Final Published Version
• Copyright Statement: © Author(s) (or their employer(s)) 2021. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.
• License: Creative Commons Attribution-NonCommercial 4.0 International
• Final Published Version (URL): http://dx.doi.org/10.1136/bmjresp-2021-000925
• Title of Journal or Parent Work: BMJ OPEN RESPIRATORY RESEARCH
• Volume: 8
• Issue: 1
• Grant/Funding Information: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
• Keywords: SPIROMETRY; PARAMETER; Science & Technology; Respiratory System; Life Sciences & Biomedicine; lung physiology; LOOP; respiratory measurement
• Research Categories: Health Sciences, Medicine and Surgery

Abstract

In spirometry, the area under expiratory flow-volume curve (AEX-FV) was found to perform well in diagnosing and stratifying physiologic impairments, potentially lessening the need for complex lung volume testing. Expanding on prior work, this study assesses the accuracy and the utility of several models of estimating AEX-FV based on forced vital capacity (FVC) and several instantaneous flows. These models could be incorporated in regular spirometry reports, especially when actual AEX-FV measurements are not available. We analysed 4845 normal spirometry tests, performed on 3634 non-smoking subjects without known respiratory disease or complaints. Estimated AEX-FV was computed based on FVC and several flows: peak expiratory flow, isovolumic forced expiratory flow at 25%, 50% and 75% of FVC (FEF25,FEF50and FEF75, respectively). The estimations were based on simple regression with and without interactions, by optimised regression models and by a deep learning algorithm that predicted the response surface of AEX-FV without interference from any predictor collinearities or normality assumption violations. Median/IQR of actual square root of AEX-FV was 3.8/3.1–4.5 L2/s. The per cent of variance (R2) explained by the models selected was very high (>0.990), the effect of collinearities was negligible and the use of deep learning algorithms likely unnecessary for regular or routine pulmonary function testing laboratory usage. In the absence of actual AEX-FV, a simple regression model without interactions between predictors or use of optimisation techniques can provide a reasonable estimation for clinical practice, thus making AEX-FV an easily available additional tool for interpreting spirometry.