Statistical analysis

Statistical analyses were performed with Statistical Package for the Social Sciences, Version 22 (IBM Corporation, Armonk, NY, USA). P-values <0.05 were considered statistically significant. A low frequency of the subjects in GOLD 3 and 4 prompted a merge into GOLD 3–4. The Student’s t-test was utilized to compare mean HGS values between groups. To analyze HGS in relation to reference values, the proportion below the mean reference value and the proportion below the 95% CI of the mean were compared between those with and without COPD using chi-squared test. Differences in mean HGS between non-COPD and COPD and non-COPD and GOLD grades were analyzed in multiple linear regression models adjusted for sex, age, height, heart disease, and smoking habits. Separate multiple linear regression analyses of the association between HGS and heart disease and FEV₁% of predicted value were performed among the subjects with and without COPD, respectively. All regression analyses were performed with HGS as dependent variable in models, including all subjects (adjusted for sex) as well as stratified by sex.