An updated comparison of standard and novel FEV1 indices’ association with all-cause mortality

Abstract

Rationale Current recommendations for defining FEV1 abnormalities are based on Z-score cutoffs. Alternative approaches may better correlate with patient-related outcomes, including mortality. Objective This study evaluates the association between FEV1 value and mortality in six FEV1 indices in a large, U.S based cohort. Methods This is a cohort of 2007–2012 National Health and Nutrition Examination Study (NHANES) participants with spirometry and linked mortality data through 2019. We transformed FEV1 values to the following indices: raw FEV1, FEV1-Z scores (FEV1-Z), FEV1-Percent Predicted (FEV1-PP), FEV1/Height2 (FEV1/Ht2), FEV1/Height3 (FEV1/Ht3), and FEV1-Q. We compared association with all-cause mortality between lowest and highest FEV1 deciles of each index using Cox Proportional Hazards models. Two sensitivity analyses were performed, one after applying NHANES sample weighting and one including lower quality spirometry. A subgroup analysis of participants with airflow obstruction, defined as FEV1/FVC ≤ LLN, was performed. Results Of the 12,994 included participants, 971 (7.5%) had died. The majority (56.2%) were nonsmokers and 13.2% had an FEV1/FVC < LLN. When comparing the most to least severe deciles of FEV1, the indices with the largest magnitude unadjusted hazard ratios were FEV1/ht3 (HR 26.4, 95%CI 16.0–43.6), FEV1/ht2 (HR 21.8, 95%CI 13.8–34.7), FEV1-Q (HR 17.5, 95%CI 11.6–26.5), and FEV1 (HR 14.4, 95%CI 9.7–21.5). After adjusting for age, gender, and tobacco pack-years, FEV1/Ht3 (HR 4.9, 95%CI 2.6–9.3) and FEV1/Ht2 (HR 4.8, 95%CI 2.7–8.7) had the highest hazard ratios, however the confidence intervals had significant overlap with other indices. In adjusted analyses, the C-statistic (0.81) was the same across indices. Sensitivity and subgroup analyses yielded a similar pattern. Conclusions Reference-range-independent indices based on absolute FEV1 (raw FEV1, FEV1-Q, FEV1/Ht2, FEV1/Ht3) are equally or more strongly associated with mortality than reference-range- dependent FEV1 indices (FEV1-Z scores, FEV1-Percent Predicted) in a large U.S. cohort.