Validation of lung function prediction equations from patient survival data

Diagnosing and managing pulmonary disease usually requires judging lung function against predicted values. We explored patient survival data to help identify the best equations for our population.The earliest spirometry, lung volumes and gas transfer data for all Caucasian patients was extracted from our database. Survival status was available for 8139 patients. Lung function as standardised residuals (SR) from various prediction equations was used in Cox regression to predict hazard ratio (HR) for death.The best lung function predictor of all-cause mortality was DL,co and then FVC, being best with the US equations of Miller, χ(2)=1468 and 1043 respectively having taken age, sex, smoking status and BMI into account. HR for SR