Validation of lung function prediction equations from patient survival data.

Area of Science:

• Pulmonary Medicine
• Medical Statistics
• Epidemiology

Background:

• Accurate lung function prediction equations are crucial for diagnosing and managing pulmonary diseases.
• Existing prediction equations may not be optimal for all patient populations.
• Patient survival data offers a novel approach to validate these equations.

Purpose of the Study:

• To identify the most accurate lung function prediction equations for a Caucasian patient cohort using survival data.
• To compare the predictive performance of different spirometry and gas transfer equations for all-cause mortality.

Main Methods:

• Retrospective analysis of spirometry, lung volumes, and gas transfer data from 8,139 Caucasian patients.
• Cox regression analysis using standardized residuals from various prediction equations to determine hazard ratios for death.
• Evaluation of prediction models for diffusing capacity of the lung for carbon monoxide (D(L,CO)) and forced vital capacity (FVC).

Main Results:

• Diffusing capacity of the lung for carbon monoxide (D(L,CO)) was the strongest predictor of all-cause mortality, followed by forced vital capacity (FVC).
• Miller's equations, derived from a US population, demonstrated the best predictive performance for both D(L,CO) and FVC in this cohort.
• The Lambda-Mu-Sigma (LMS) and Miller spirometry equations showed comparable, better performance than other models, with Miller's D(L,CO) equations exhibiting less sex bias.

Conclusions:

• Patient survival data is a valuable tool for selecting appropriate lung function prediction equations.
• Miller's equations for D(L,CO) and FVC, along with LMS or Miller spirometry equations, are recommended for this Caucasian patient population.
• Careful selection of prediction equations is essential to avoid biases, particularly sex bias, in pulmonary function assessment.