Outcome-based Definition of the Lower Limit of Normal in Spirometry: A Study of 26,000 Young Adult Men.

Rationale The definition of the lower limit of normal (LLN) of spirometric variables is not well established. Objectives To investigate the relationship between spirometric abnormalities defined with different thresholds of the LLN and clinical outcomes, and to explore the possibility of using different LLN thresholds according to the pre-test probability of disease. Methods We studied the associations between pre-bronchodilator spirometric abnormalities (FEV1 < LLN, FVC < LLN, airflow obstruction, spirometric restriction) defined with different thresholds of the LLN (10th, 5th, 2.5th, 1st percentile) and multiple outcomes (prevalence of spirometric abnormalities, respiratory symptoms, all-cause and respiratory mortality) in 26,091 30-46 years old men who participated in a general population survey in Norway in 1988-1990 and were followed for 26 years. Analyses were performed with both local and GLI-2012 reference equations, stratified by pre-test risk (presence or absence of respiratory symptoms), and adjusted for age, BMI, smoking, and education. Results In the total population, the prevalence of airflow obstruction was 11.6% with GLI-LLN10, 11.0% with Local-LLN5, 6.1% with GLI-LLN5, 7.6% with Local-LLN2.5, and 3.5% with GLI-LLN2.5. The prevalence of spirometric restriction was 5.9% with GLI-LLN10, 5.2% with Local-LLN5, and 2.8% with GLI-LLN5. Increasingly lower thresholds of the LLN were associated with increasingly higher odds of respiratory symptoms and hazard of mortality for all spirometric abnormalities with both reference equations. Spirometric abnormalities defined with Local-LLN2.5 in asymptomatic subjects were associated with lower hazard of all-cause mortality (hazard ratio 1.50 (1.15, 1.95, 95% confidence intervals) for FEV1 < LLN) than those defined with Local-LLN5 in the general population (1.67 (1.50, 1.87) for FEV1 < LLN) and symptomatic subjects (1.67 (1.46, 1.91) for FEV1 < LLN). Overall, prevalence of spirometric abnormalities and associations with outcomes obtained with Local-LLN5 were comparable to those obtained with GLI-LLN10, and those obtained with Local-LLN2.5 to GLI-LLN5. Conclusions There is a relationship between statistically-based thresholds of the LLN of spirometric variables and clinical outcomes. Different thresholds of the LLN may be used in different risk subgroups of subjects, but the choice of the threshold needs to be evaluated together with the choice of reference equations.