Artificial intelligence based software facilitates spirometry quality control in asthma and COPD clinical trials

Area of Science:

• Pulmonary Medicine
• Medical Informatics
• Clinical Trial Methodology

Background:

• High-quality spirometry data is crucial for clinical trials, especially for endpoints like forced expiratory volume in 1-second (FEV1) and forced vital capacity (FVC).
• Current American Thoracic Society (ATS)/European Respiratory Society (ERS) standards for spirometry quality include subjective evaluations, leading to inter-rater variability and potential errors.

Purpose of the Study:

• To evaluate the effectiveness of an artificial intelligence (AI)-based software (ArtiQ.QC) in assessing spirometry quality.
• To compare the AI's performance against traditional over-reading by experts.

Main Methods:

• A random sample of 2000 spirometry sessions (8258 curves) from COPD and asthma trials was analyzed.
• Spirometry acceptability was assessed using 2005 ATS/ERS standards by both expert over-readers and the ArtiQ.QC AI software.
• A subset of curves underwent joint review by three respiratory physicians to establish a consensus (gold standard).

Main Results:

• The AI software demonstrated high agreement (91%) with expert over-readers, achieving 97% sensitivity and 93% positive predictive value.
• While 88% of all curves were of good quality, the AI's performance was notably better in the asthma cohort.
• In a subset analysis, the AI achieved 73% agreement with the expert consensus, outperforming individual over-reader agreement (46%).

Conclusions:

• AI-based software offers comparable accuracy to human experts in measuring spirometry data quality.
• The subjective nature of spirometry assessment introduces variability, even with precise criteria.
• AI can enhance clinical trial conduct by providing consistent results and immediate feedback, thereby reducing variability.