Table 3

Associations between per cent emphysema and mortality due to lung disease in the Multi-Ethnic Study of Atherosclerosis, stratified by airflow limitation, 2004–2013

Per cent emphysema as continuous exposure in persons with valid spirometry measuresp Interaction for airflow limitation
Underlying cause of deathAirflow limitation (N=863)p ValueNo airflow limitation (N=2965)p Value
Respiratory diseases
 Deaths8120.597
 Person-years641822 843
 Mortality rate per 10 000 person-years12.55.3
 HR (95% CI)
  HRcrude (95% CI)1.38 (0.87 to 2.19)0.1711.41 (0.79 to 2.50)0.244
  HRadjusted (95% CI)1.32 (0.98 to 1.78)0.0691.30 (0.58 to 2.94)0.525
  HRextended (95% CI)1.40 (1.02 to 1.93)0.0382.38 (1.01 to 5.60)0.047
Lung cancer
 Deaths18180.593
 Person-years641822 843
 Mortality rate per 10 000 person-years28.17.9
 HR (95% CI)
  HRcrude (95% CI)1.39 (1.03 to 1.88)0.0311.03 (0.58 to 1.85)0.920
  HRadjusted (95% CI)1.31 (0.89 to 1.93)0.1701.14 (0.58 to 2.24)0.709
  HRextended (95% CI)1.52 (1.01 to 2.28)0.0461.56 (0.82 to 2.96)0.172
All lung diseases
 Deaths26300.930
 Person-years641822 843
 Mortality rate per 10 000 person-years40.513.1
 HR (95% CI)
  HRcrude (95% CI)1.39 (1.08 to 1.78)0.0111.18 (0.79 to 1.78)0.413
  HRadjusted (95% CI)1.32 (0.98 to 1.78)0.0711.16 (0.71 to 1.90)0.542
  HRextended (95% CI)1.38 (1.01 to 1.90)0.0441.63 (1.02 to 2.61)0.043
  • The end points were defined by an underlying cause of death of respiratory disease (J00–J99), lung cancer (C33–C34) and, combining these, all lung disease. We do not report mortality due to chronic lower respiratory disease (J40–47) due to very low event rates in this group (N=6 and N=2 in persons with and without airflow limitation, respectively).

  • HRs reported per IQR (4.5%), which is equivalent to the difference between the third quartile (5.7%) and the first quartile (1.2%) of per cent emphysema. Models adjusted for baseline age, sex, race/ethnicity, body mass index, site, smoking status, pack-years of smoking, coronary artery calcium score and educational attainment. Interaction terms are reported for this adjusted model. The extend model is additionally adjusted for the FEV1, absence/presence of restriction on spirometry and log-transformed volume of high attenuation areas.