FEV1

FEV1 predicts long-term prebronchodilator FEV1 trajectories and estimates individual risk of airflow limitation (FEV1/FVC below the lower limit of normal) using longitudinal spirometry data.

Key Features:

• Primary outcome: Prebronchodilator Forced Expiratory Volume in one second (FEV1) as the longitudinal outcome.
• Secondary outcome: Risk of airflow limitation defined as FEV1/FVC below the lower limit of normal.
• Development cohort: Derived from spirometry assessments in the Framingham Offspring Cohort comprising 4,167 participants aged ≥20 years with at least two valid spirometry measurements.
• Modeling approach: Individualized predictions generated using mixed effects regression models.
• Predictor selection: A machine learning algorithm was used to identify essential predictors, with the model utilizing 20 common predictors.
• Derivation performance: The model explained 79% of the variation in FEV1 decline in the derivation cohort.
• External validation cohorts: Validated across two independent multicenter cohorts of 2,075 and 12,913 participants.
• Validation performance: Low prediction error for FEV1 decline (root mean square error 0.18–0.22 L) and high discriminative power for airflow limitation risk (C-statistic 0.86–0.87).

Scientific Applications:

• COPD risk stratification: Predicts individual trajectories of FEV1 decline and probability of airflow limitation to stratify COPD risk.
• Epidemiological studies: Characterizes population-level lung function trajectories and determinants of FEV1 decline.
• Clinical research: Supports prognostic modeling, cohort characterization, and trial enrichment based on predicted lung function trajectories.
• Public health: Informs estimation of population burden of airflow limitation and prioritization of prevention strategies.

Methodology:

Computational methods explicitly reported include mixed effects regression models for individualized prediction, a machine learning algorithm for predictor selection using 20 common predictors, and external validation in two independent multicenter cohorts.