AI Over-Reading Based On ATS/ERS 2019 Criteria Is A Reliable Option For Instant Spirometry Quality Control In Clinical Trials


Acquiring high-quality spirometry data in clinical trials is important, particularly when using FEV1 or FVC as primary endpoints and clinical decision-making. In addition to quantitative criteria, the ATS/ERS quality control standards include subjective evaluation which introduces inter-rater variability.  


This study explores the value and reliability of artificial intelligence (AI) based quality control software (ArtiQ.QC) to assess spirometry quality in clinical trials using the 2019 ATS/ERS guidelines. 


A total of 3103 spirometry sessions (16743 curves) were randomly selected from 2 Chiesi COPD and chronic bronchitis clinical trials. Acceptability was determined by over-readers using a protocol based on the 2019 ATS/ERS guidelines and compared with acceptability defined by ArtiQ.QC (Das et al. ERJ 2020). The comparison was performed at curve level assigning  2 categories: ‘acceptable’ and ‘usable or unacceptable’, based on the over-reading protocol.  


Concordance between the AI algorithm and human over-reading was observed in 95% of the cases for FEV1, and in 91% of the cases for FVC.  


The results reconfirmed a high level of concordance between the AI quality control software and the human over-readers also with the use of the 2019 ATS/ERS guidelines. By providing immediate and consistent results, the use of the AI algorithm may optimize clinical trial conduct and reduce the variability of spirometry variables often used as primary endpoints in respiratory trials. 


P. Desbordes, B. Cuyvers, E. Topole, S. Biondaro, I. Montagna, S. Corre, M. Topalovic 


[1] ArtiQ NV, Leuven, Belgium 

[2] Global Clinical Development, Chiesi Farmaceutici S.p.A; Parma, Italy