Lung function decline in former smokers and low-intensity current smokers: a secondary data analysis of the NHLBI Pooled Cohorts Study

Abstract

Background: Former smokers now outnumber current smokers in many developed countries, and current smokers are smoking fewer cigarettes per day. Some data suggest that lung function decline normalises with smoking cessation; however, mechanistic studies suggest that lung function decline could continue. We hypothesised that former smokers and low-intensity current smokers have accelerated lung function decline compared with never-smokers, including among those without prevalent lung disease.

Methods: We used data on six US population-based cohorts included in the NHLBI Pooled Cohort Study. We restricted the sample to participants with valid spirometry at two or more exams. Two cohorts recruited younger adults (≥17 years), two recruited middle-aged and older adults (≥45 years), and two recruited only elderly adults (≥65 years) with examinations done between 1983 and 2014. FEV₁ decline in sustained former smokers and current smokers was compared to that of never-smokers by use of mixed models adjusted for sociodemographic and anthropometric factors. Differential FEV₁ decline was also evaluated according to duration of smoking cessation and cumulative (number of pack-years) and current (number of cigarettes per day) cigarette consumption.

Findings: 25 352 participants (ages 17-93 years) completed 70 228 valid spirometry exams. Over a median follow-up of 7 years (IQR 3-20), FEV₁ decline at the median age (57 years) was 31·01 mL per year (95% CI 30·66-31·37) in sustained never-smokers, 34·97 mL per year (34·36-35·57) in former smokers, and 39·92 mL per year (38·92-40·92) in current smokers. With adjustment, former smokers showed an accelerated FEV₁ decline of 1·82 mL per year (95% CI 1·24-2·40) compared to never-smokers, which was approximately 20% of the effect estimate for current smokers (9·21 mL per year; 95% CI 8·35-10·08). Compared to never-smokers, accelerated FEV₁ decline was observed in former smokers for decades after smoking cessation and in current smokers with low cumulative cigarette consumption (<10 pack-years). With respect to current cigarette consumption, the effect estimate for FEV₁ decline in current smokers consuming less than five cigarettes per day (7·65 mL per year; 95% CI 6·21-9·09) was 68% of that in current smokers consuming 30 or more cigarettes per day (11·24 mL per year; 9·86-12·62), and around five times greater than in former smokers (1·57 mL per year; 1·00-2·14). Among participants without prevalent lung disease, associations were attenuated but were consistent with the main results.

Interpretation: Former smokers and low-intensity current smokers have accelerated lung function decline compared with never-smokers. These results suggest that all levels of smoking exposure are likely to be associated with lasting and progressive lung damage.

Funding: National Institutes of Health, National Heart Lung and Blood Institute, and US Environmental Protection Agency.

Figure 1.. CONSORT diagram.

HCHS/SOL = Hispanic Community Health Study/Study of Latinos; JHS = Jackson Heart Study; MESA = Multi-Ethnic Study of Atherosclerosis; NHLBI = National Heart, Lung, Blood Institute; SHS = Strong Heart Study.

Figure 2.. Predicted FEV1 curves according to (A) smoking status, (B) duration of smoking cessation, (C) cumulative cigarette consumption, and (D) current cigarette consumption.

Predicted FEV1 curves were generated from linear mixed models with cohort-specific unstructured covariance matrices adjusted for age, age², height, height², sex, race/ethnicity, weight, birth year, site, study, educational attainment, and the smoking parameter of interest. Multiplicative interactions with age were modeled for all covariates. 95% confidence intervals are indicated by shading. To reflect differences in age distributions across strata of smoking exposures, predictions at the extremes of each stratum-specific age distribution (<5^th or >95^th percentile) were not shown. Participants with variable smoking status were excluded from analyses of duration of smoking cessation and of cumulative and current cigarette consumption. With respect to categorizations of duration of smoking cessation and cumulative and current cigarette consumption, all ranges were inclusive of the lower boundary point and exclusive of the upper boundary point.

Figure 2.. Predicted FEV1 curves according to (A) smoking status, (B) duration of smoking cessation, (C) cumulative cigarette consumption, and (D) current cigarette consumption.

Predicted FEV1 curves were generated from linear mixed models with cohort-specific unstructured covariance matrices adjusted for age, age², height, height², sex, race/ethnicity, weight, birth year, site, study, educational attainment, and the smoking parameter of interest. Multiplicative interactions with age were modeled for all covariates. 95% confidence intervals are indicated by shading. To reflect differences in age distributions across strata of smoking exposures, predictions at the extremes of each stratum-specific age distribution (<5^th or >95^th percentile) were not shown. Participants with variable smoking status were excluded from analyses of duration of smoking cessation and of cumulative and current cigarette consumption. With respect to categorizations of duration of smoking cessation and cumulative and current cigarette consumption, all ranges were inclusive of the lower boundary point and exclusive of the upper boundary point.

Figure 2.. Predicted FEV1 curves according to (A) smoking status, (B) duration of smoking cessation, (C) cumulative cigarette consumption, and (D) current cigarette consumption.

Predicted FEV1 curves were generated from linear mixed models with cohort-specific unstructured covariance matrices adjusted for age, age², height, height², sex, race/ethnicity, weight, birth year, site, study, educational attainment, and the smoking parameter of interest. Multiplicative interactions with age were modeled for all covariates. 95% confidence intervals are indicated by shading. To reflect differences in age distributions across strata of smoking exposures, predictions at the extremes of each stratum-specific age distribution (<5^th or >95^th percentile) were not shown. Participants with variable smoking status were excluded from analyses of duration of smoking cessation and of cumulative and current cigarette consumption. With respect to categorizations of duration of smoking cessation and cumulative and current cigarette consumption, all ranges were inclusive of the lower boundary point and exclusive of the upper boundary point.

Figure 2.. Predicted FEV1 curves according to (A) smoking status, (B) duration of smoking cessation, (C) cumulative cigarette consumption, and (D) current cigarette consumption.

Predicted FEV1 curves were generated from linear mixed models with cohort-specific unstructured covariance matrices adjusted for age, age², height, height², sex, race/ethnicity, weight, birth year, site, study, educational attainment, and the smoking parameter of interest. Multiplicative interactions with age were modeled for all covariates. 95% confidence intervals are indicated by shading. To reflect differences in age distributions across strata of smoking exposures, predictions at the extremes of each stratum-specific age distribution (<5^th or >95^th percentile) were not shown. Participants with variable smoking status were excluded from analyses of duration of smoking cessation and of cumulative and current cigarette consumption. With respect to categorizations of duration of smoking cessation and cumulative and current cigarette consumption, all ranges were inclusive of the lower boundary point and exclusive of the upper boundary point.