Five-year Progression of Emphysema and Air Trapping at CT in Smokers with and Those without Chronic Obstructive Pulmonary Disease: Results from the COPDGene Study

Abstract

Background

CT is used to quantify abnormal changes in the lung parenchyma of smokers that might overlap chronic obstructive pulmonary disease (COPD), but studies on the progression of expiratory air trapping in smokers are scarce.

Purpose

To evaluate the relationship between longitudinal changes in forced expiratory volume in 1 second (FEV₁) and CT-quantified emphysema and air trapping in smokers.

Materials and Methods

Cigarette smokers with and those without COPD participating in the multicenter observational COPDGene study were evaluated. Subjects underwent inspiratory and expiratory chest CT and spirometry at baseline and 5-year follow-up. Emphysema was quantified by using adjusted lung density (ALD). Air trapping was quantified by using mean lung density at expiratory CT and CT-measured functional residual capacity–to–total lung volume ratio. Linear models were used to regress quantitative CT measurements taken 5 years apart, and models were fit with and without adding FEV₁ as a predictor. Analyses were stratified by Global Initiative for Chronic Obstructive Lung Disease (GOLD) stage (GOLD 0, no COPD; GOLD 1, mild COPD; GOLD 2, moderate COPD; GOLD 3, severe COPD; GOLD 4, very severe COPD). Subjects with preserved FEV₁-to-forced vital capacity ratio and reduced FEV₁ percentage predicted were categorized as having preserved ratio impaired spirometry (PRISm).

Results

A total of 4211 subjects (503 with PRISm; 2034 with GOLD 0, 388 with GOLD 1, 816 with GOLD 2, 381 with GOLD 3, 89 with GOLD 4) were evaluated. ALD decreased by 1.7 g/L (95% confidence interval [CI]: −2.5, −0.9) in subjects with GOLD 0 at baseline and by 5.3 g/L (95% CI: −6.2, −4.4) in those with GOLD 1–4 (P < .001 for both). When adjusted for changes in FEV₁, corresponding numbers were −2.2 (95% CI: −3.0, −1.3) and −4.6 g/L (95% CI: −5.6, −3.4) (P < .001 for both). Progression in air trapping was identified only in GOLD stage 2–4. Approximately 33%–50% of changes in air trapping in GOLD stages 2–4 were accounted for by changes in FEV₁.

Conclusion

CT measures of emphysema and air trapping increased over 5 years in smokers. Forced expiratory volume in one second accounted for less than 10% of emphysema progression and less than 50% of air trapping progression detected at CT.

© RSNA, 2020

Online supplemental material is available for this article.

Summary

A large proportion of emphysema progression in smokers with and those without airflow limitation was not accounted for by changes in forced expiratory volume in 1 second, suggesting that CT-detected disease can progress before any measurable changes are detected with pulmonary function tests.

Key Results

• ■ CT measurements of adjusted lung density (ALD) declined by 1.7 g/L in subjects without chronic obstructive pulmonary disease symptoms (Global Initiative for Chronic Obstructive Lung Disease [GOLD] stage 0) and by 5.3 g/L in subjects with mild to severe chronic obstructive pulmonary disease (GOLD stages 1–4) after adjusting for age and height (P < .001 for both).

• ■ Estimates of progression of CT measures were approximately 10% less for inspiratory CT measurements (ALD and 15th percentile of the attenuation histogram) and approximately 33%–50% less for expiratory CT measurements (ratio of forced residual capacity to total lung capacity, mean lung density) after adjusting for time-varying forced expiratory volume in 1 second (FEV₁), suggesting that changes in FEV₁ can account for just a small portion of changes in inspiratory CT measurements but can account for a larger portion of changes in expiratory CT measurements.

• ■ The 6-minute walking distance test and St George’s Respiratory Questionnaire results accounted for very little of the progression in inspiratory and expiratory CT variables.

Introduction

Chronic obstructive pulmonary disease (COPD) is a chronic progressive lung disease characterized by nonreversible airflow limitation (1). This can be due to irreversible destruction of lung tissue (emphysema) and inflammation, remodeling of the small airways (2), or both. Although COPD is diagnosed and staged with pulmonary function testing (1), CT enables quantification of abnormal changes in the lung parenchyma that often, but not always, overlap with COPD.

Emphysema can be measured from inspiratory chest CT scans as low-attenuation areas in the lung (3,4). Quantitative CT analysis of emphysema correlates well with histologic findings (5), is associated with symptoms among smokers (6) and the general population (7), and is used to independently predict mortality in subjects with and those without COPD (8). To evaluate emphysema on CT images, the percentage of lung voxels with attenuation below a certain threshold (commonly −950 HU) can be assessed, or the lung attenuation at a certain histogram percentile (eg, 15th percentile) can be evaluated. The density of the lung can be adjusted for inspiratory lung volume by using a sponge model that adjusts CT-derived lung volume for the physiologic predicted total lung capacity (TLC) (9). This model assumes that lung density increases proportionally as lung volume decreases.

Abnormalities in the small airways cannot be directly quantified on inspiratory chest CT images, but they can be inferred from air trapping on expiratory CT images and have been shown to correlate well with physiologic evidence of airflow obstruction (10). Several quantitative CT measures of air trapping have been suggested, including the percentage of lung voxels with attenuation less than a given threshold (commonly, −856 HU) and the ratio of functional residual capacity (FRC) to TLC.

Progression of emphysema and air trapping can be monitored by using serial CT scanning. For instance, in a trial of α-1-antitrypsin augmentation therapy, CT-quantified emphysema has been used as an outcome measure (11). Studies evaluating the progression of emphysema and air trapping in subjects who smoke cigarettes are scarce (12,13), especially studies that evaluate progression of expiratory air trapping. We hypothesized that progression in CT density measurements would be correlated with changes in forced expiratory volume in 1 second (FEV₁). The purpose of this study was to investigate the relationship between the 5-year progression in CT-quantified emphysema and air trapping and the changes in other indexes of disease severity, including FEV₁, 6-minute walking distance, and the St George’s Respiratory Questionnaire (SGRQ) in the COPDGene study cohort.

Materials and Methods

Study Participants

The COPDGene study is a multicenter prospective observational investigation of the genetic and epidemiologic factors associated with COPD (14) that recruited over 10 000 subjects with COPD, as defined by standard spirometric criteria (1), and control subjects who had smoked for at least 10 pack-years. The COPDGene study was approved by the institutional review boards at each of the 21 clinical sites. We obtained written informed consent from all subjects, and the study was compliant with the Health Insurance Portability and Accountability Act. Our study included subjects who were enrolled in phase 1 from 2008 to 2011 and who returned for 5-year follow-up in phase 2. Among the subjects who returned for an on-site visit in phase 2, subjects were excluded if they satisfied one or more of the following conditions: a CT scan was not performed, there was an issue with scan quality or processing failure, there was a change in smoking status between examinations (smoking cessation is associated with a decrease in lung attenuation, which could be misinterpreted as an increase in emphysema [15,16]), there were clinically important changes in the lungs that would affect quantification (eg, lobectomy, change in lung parenchymal opacities), there was a lack of spirometry data, or the interval between CT and spirometry was more than 90 days.

Thoracic CT Scans and Spirometry

Inspiratory and expiratory volumetric thoracic nonenhanced CT scans were obtained at baseline and 5-year follow-up by using a standardized protocol for multidetector CT scanners (14). All spirometry data were collected by using an ndd EasyOne spirometer (ndd, Andover, Mass).

Subjects were categorized according to Global Initiative for Chronic Obstructive Lung Disease (GOLD) status (GOLD 0, no COPD; GOLD 1, mild COPD [FEV₁ ≥ 80% of that predicted]; GOLD 2, moderate COPD [FEV₁ = 50%–79% of that predicted); GOLD 3, severe COPD [FEV₁ = 30%–49% of that predicted]; GOLD 4, very severe COPD [FEV₁ < 30% of that predicted]) (1). Subjects with preserved FEV₁-to-forced vital capacity ratio and reduced percentage predicted FEV₁ were categorized as having preserved ratio impaired spirometry (PRISm) (17).

Quantitative CT Analysis

Quantitative CT analysis was performed centrally by using LungQ, version 1.0.0 (Thirona, Nijmegen, the Netherlands), which is commercially available. The lungs and lobes were automatically segmented (18,19), visually checked by data analysts for accuracy, and corrected, as needed (20).

Emphysema was quantified on inspiratory scans using the adjusted lung density (ALD) measurement, and Multi-Ethnic Study of Atherosclerosis (or MESA) equations were used to adjust for total lung volume (TLV) on CT scans (21). This measurement is based on the 15th percentile of the attenuation histogram (Perc15) method (attenuation [in Hounsfield units] below which 15% of voxels are situated + 1000 HU) that is adjusted for the difference in TLV between baseline and follow-up scans. The adjustment uses the sponge model principle, which assumes mass (volume times density) remains constant between two time periods. Equations can be found in Appendix E2 (online). We include results for both ALD and Perc15.

Air trapping was quantified on expiratory scans by using the mean lung density (MLD) at expiratory CT and the ratio of FRC measured on expiratory CT scans to TLV measured on inspiratory CT scans (hereafter, FRC-to-TLV ratio). Figure 1 shows how CT measurements were calculated.

Figure 1:

Left, Graph shows histograms for inspiratory and expiratory CT based on which emphysema and air trapping measurements are calculated. Emphysema is calculated by using the 15th percentile of the attenuation histogram, which represents the Hounsfield unit value below which 15% of the lowest voxels are distributed. This is converted to adjusted lung density by adding 1000 HU and adjusting for the difference in lung volume between visits 1 and 2. Air trapping was defined by using mean lung density (MLD) at expiratory CT (dotted line), as well as the ratio of CT-derived functional residual reserve to total lung volume. Total lung capacity (TLC) can be calculated by using the area under the inspiratory CT curve. Functional residual capacity (FRC) can be calculated by using the area under the expiratory CT curve. Middle and right, Images show three-dimensional lung volumes of inspiratory and expiratory CT scans, respectively, representing TLC and FRC.

Statistical Analyses

Each quantitative CT variable was modeled as an outcome in full and reduced linear mixed models. The full models contained visit number (visit 1 or 2, a class variable), mean FEV₁ over the two visits, visit-specific (or time-varying) FEV₁ and other covariates (subject race, age, sex, and height; scanner make) as predictors. The inclusion of mean FEV₁ allows the coefficient of visit-specific FEV₁ to represent the estimated change in the CT variable per unit increase in FEV₁ within a subject (the between- and within-subject FEV₁ effects are separated by including both mean FEV₁ and time-varying FEV₁; only subjects with FEV₁ at both visits were included in model fits). The reduced model simply removed the visit-specific FEV₁ variable. In our model, the expected progression of a CT measure is the time effect (average change between visits based on the visit predictor) plus average changes in time-varying covariates (ie, the time effect indicates CT progression not accounted for by the time-varying covariates). The impact of within-subject changes in FEV₁ on expected progression was assessed by comparing the difference in time effects between full and reduced models, as it was the only predictor that differed between the models. For example, if time effects are similar between full and reduced models, then little or no change in the CT variable is accounted for by changes in FEV₁. On the other hand, if the time effect in the full model is a small portion of what is in the reduced model without time-varying FEV₁, then changes in FEV₁ account for much of the CT variable progression.

To account for repeated measures, a random intercept was included in each model for each subject; random intercept terms were also included for scanner model and study center. Analyses were performed separately for each GOLD stage and then by grouping subjects into GOLD stage 0, GOLD stage 1–4, or PRISm. Similar methods were then applied to determine how much of CT progression was accounted for by 6-minute walking distance and SGRQ. P < .05 was considered indicative of a significant difference. Statistical software (SAS, version 9.4; SAS Institute, Cary, NC) was used for analyses.

Results

Subject Characteristics

Of the 10 198 subjects included in phase 1, 5697 returned for their 5-year follow-up visit. Among those who returned, 173 were excluded because of missing CT scans, 518 were excluded because of CT quality failures, 20 were excluded because of clinically important changes in the lung, 718 were excluded because smoking status changed, 20 were excluded because of missing baseline spirometry data, 37 were excluded because CT and spirometry were performed more than 90 days apart, and 68 did not have complete spirometry data. Thus, 4143 subjects were eligible for analyses involving FEV₁. The 68 subjects without complete spirometry data were eligible for analyses using other clinical variables (n = 4211). These data are shown in the flowchart (Fig 2). Of the 4211 subjects, 2034 (48.3%) had GOLD 0, 503 (11.9%) had PRISm, 388 (9.2%) had GOLD 1, 816 (19.4%) had GOLD 2, 381 (9.0%) had GOLD 3, and 89 (2.1%) had GOLD 4; 50% (n = 2121) of the subjects were men. Mean age of subjects in the PRISm and GOLD 0 groups was 58 years ± 9 (standard deviation). Mean age of subjects in the GOLD 1 and GOLD 2 groups was 63 years ± 8 and 63 years ± 9, respectively. Mean age of subjects in the GOLD 3 and GOLD 4 groups was 64 years ± 8 and 64 years ± 7 years, respectively. Table 1 shows baseline characteristics of the included subjects. Table 2 shows unadjusted longitudinal changes in physiologic measures. All GOLD groups exhibited decreases in mean FEV₁ over 5 years, with GOLD 0, 1, and 2 groups exhibiting the largest changes (−234.0 mL ± 259.3, −272.2 mL ± 263.2, and −231.6 mL ± 313.8, respectively). The number of subjects with a change in GOLD stage from visit 1 to 2 is shown in Table E1 (online).

Figure 2:

Flowchart shows 4143 subjects were eligible for analyses, including measurement of forced expiratory volume in one second (FEV₁). There were 4211 subjects in whom FEV₁ was missing who were eligible for analyses including other clinical variables (ie, 6-minute walking distance). BD = bronchodilation, COPD = chronic obstructive pulmonary disease, LVRS = lung volume reduction surgery, PFT = pulmonary function test.

Table 1:

Descriptive Statistics for Clinical, Demographic, and CT Variables at Visit 1 for All 4211 Smokers

Table 2:

Unadjusted Estimates of Change in Key Variables

Changes in Inspiratory CT Measures of Emphysema

Table 3 shows the 5-year change in all inspiratory quantitative CT measurements for subjects in the PRISm, GOLD 0, and GOLD 1–4 groups. After adjustment for covariates, mean ALD showed changes over 5 years for subjects in all groups (−2.26 g/L, −1.67 g/L, and −5.26 g/L in the PRISm, GOLD 0, and GOLD 1–4 groups, respectively; P = .01 in the PRISm group, P < .001 in the GOLD 0 and GOLD 1–4 groups). Mean Perc15 decreased in the GOLD 1–4 group (5.07 HU, P < .001) but did not decrease significantly in other groups (P > .05). When FEV₁ was included in the models for subjects in the GOLD 1–4 group, estimates of change over time were reduced (in magnitude) by approximately 10% for ALD (from −5.26 g/L to −4.60 g/L) and Perc15 (from −5.07 HU to −4.62 HU), showing that most of the changes in these measurements were not accounted for by FEV₁. Inclusion of FEV₁ in the models for subjects in the GOLD 0 or PRISm groups did not reduce estimates for ALD (GOLD 0, from −1.67 to −2.18 g/L; P < .001) (PRISm, −2.26 g/L to −2.64 g/L; P < .05). Results stratified by individual GOLD stage for ALD are shown in Figure 3, A, and Table E2 (online). Results stratified by GOLD stage for Perc15 are shown in Figure 3, B, and Table E3 (online). An example of an increase in emphysema at CT in a subject with GOLD 0 disease and a subject with GOLD 4 disease is shown in Figure 4.

Table 3:

Estimated 5-year Mean Change in Inspiratory CT Measurements from Linear Mixed Models for Subjects Classified by GOLD Stage after Adjusting for Time-varying Covariates and FEV₁

Figure 3:

Bar graphs show estimated 5-year mean change (visit 2 − visit 1) in, A, adjusted lung density and, B, 15th percentile of the attenuation histogram from the linear mixed models for smokers by Global Initiative for Chronic Obstructive Lung Disease (GOLD) group. Dark gray bars show results without forced expiratory volume in 1 second (FEV₁) included in the model; light gray bars, results with FEV₁ included in the model; error bars are black. GOLD 0 = no chronic obstructive pulmonary disease (COPD), GOLD 1 = mild COPD, GOLD 2 = moderate COPD, GOLD 3 = severe COPD, GOLD 4 = very severe COPD, PRISm = preserved ratio impaired spirometry.

Figure 4:

Baseline and follow-up inspiratory CT scans in two smokers show an increase in emphysema (blue areas). A threshold of −950 HU was used to define emphysema. A, Baseline and, B, 5-year follow-up scans in a smoker with Global Initiative for Chronic Obstructive Lung Disease (GOLD) stage 1 disease. Adjusted lung density (ALD) decreased from 66.5 g/L to 49.2 g/L. Forced expiratory volume in 1 second (FEV₁) was 3.2 L, and ratio of FEV₁ to forced vital capacity (FVC) was 74% at both visits. C, Baseline and, D, 5-year follow-up scans in a smoker with GOLD stage 4 disease. ALD decreased from 74.5 g/L to 59.0 g/L. FEV₁ increased from 0.5 L to 0.6 L, but FEV₁-to-FVC ratio decreased from 40.3% to 25.7%.

Results of the 6-minute walking distance test and SGRQ were similar to FEV₁ and are provided in Tables E4 and E5 (online) and are stratified by PRISm, GOLD 0, and GOLD 1–4 groups.

Changes in Expiratory CT Measures of Air Trapping

The 5-year change estimates for the FRC-to-TLV ratio and MLD at expiratory CT for subjects in the GOLD 0, PRISm, and GOLD 1–4 groups are provided in Table 4. For subjects in the GOLD 1–4 group, progression was identified for both expiratory measurements. When FEV₁ was included in the models, estimates for subjects in the GOLD 1–4 group were reduced by about 50% for the FRC-to-TLV ratio (from 2.67% to 1.34%, P < .01 for both) and by about 33% for MLD at expiratory CT (from −13.73 HU to −9.31 HU, P < .01 for both), suggesting change in expiratory CT measurements was not fully accounted for by changes in FEV₁ but more than for the inspiratory measures. These reductions were also consistent with individual GOLD 1–4 groups (Fig 5; Tables E6, E7 [online]). These tables show that the largest changes occurred for GOLD 2 and GOLD 3 groups; after including time-varying FEV₁, estimates of progression over 5 years decreased by 20%–50% but remained significant (P < .01). In the PRISm and GOLD 0 groups, mean changes over time were not significant (P > .05) whether or not time-varying FEV₁ was included. An example of an increase in air trapping on CT scans in a subject with GOLD 0 disease and a subject with GOLD 4 disease is shown in Figure 6.

Table 4:

Estimated 5-year Mean Change in Expiratory CT Measurements from Linear Mixed Models for Subjects Classified by GOLD Stage after Adjusting for Time-varying Covariates and FEV₁

Figure 5:

Bar graphs show estimated 5-year mean change (visit 2 − visit 1) in, A, ratio of forced residual capacity to total lung volume measured with CT, and, B, mean lung density at expiratory CT from the linear mixed models for smokers by Global Initiative for Chronic Obstructive Lung Disease (GOLD) group. Dark gray bars show results without forced expiratory volume in 1 second (FEV₁) included in the model; light gray bars, results with FEV₁ included in the model; error bars are black. GOLD 0 = no chronic obstructive pulmonary disease (COPD), GOLD 1 = mild COPD, GOLD 2 = moderate COPD, GOLD 3 = severe COPD, GOLD 4 = very severe COPD, PRISm = preserved ratio impaired spirometry.

Figure 6:

Baseline and follow-up expiratory CT scans in two smokers show an increase in the amount of air trapping (yellow areas). A threshold of −856 HU was used to define air trapping. A, Baseline and, B, 5-year follow-up scans in a smoker with Global Initiative for Chronic Obstructive Lung Disease (GOLD) 1 disease. Mean lung density at expiratory CT (MLD_exp) decreased from −686 HU to −762 HU. Forced expiratory volume in 1 second (FEV₁) was 2.4 L and did not change over time. Ratio of FEV₁ to forced vital capacity (FVC) changed from 79.2% to 79.3%. C, Baseline and, D, 5-year follow-up CT scans in a smoker with GOLD 4 disease. MLD_exp decreased from −769 HU to −816 HU. FEV₁ changed from 0.7 L to 0.6 L, and FEV₁-to-FVC ratio changed from 27.3% to 30.3%.

Results of 6-minute walking distance and SGRQ were similar to FEV₁ results (Tables E4, E5 [online]) and are stratified by PRISm, GOLD 0, and GOLD 1–4 groups.

Discussion

Emphysema and air trapping in smokers can be evaluated with CT, but data on disease progression monitoring with CT are needed. We found that CT-quantified emphysema progressed over a 5-year period in smokers with and those without chronic obstructive pulmonary disease (COPD) (preserved ratio impaired spirometry [PRISm] or Global Initiative for Chronic Obstructive Lung Disease [GOLD] 0–4 disease) and that air trapping significantly progressed in the PRISm and GOLD 1–4 groups. Inclusion of forced expiratory volume in 1 second (FEV₁) in the models resulted in a small to moderate decrease in the magnitude of change (around 10% for emphysema, up to 50% for air trapping), indicating that changes in emphysema and air trapping are only partly explained by changes in FEV₁. The relative amount of change in the quantitative CT measurements explained by FEV₁ was higher for expiratory measurements than for inspiratory measurements, indicating that expiratory measurements are more strongly related to changes in FEV₁ (22).

Two other studies (22,23) have investigated changes over time for quantitative CT in large cohorts of smokers. In the Evaluation of COPD Longitudinally to Identify Predictive Surrogate Endpoints (or ECLIPSE) study, which included 1928 current and former smokers with GOLD 2–4 symptoms, Perc15 was 1.13 g/L/y, which was quite similar to the decline of 5.48 g/L over 5 years found in our study for smokers with GOLD 1–4 disease. However, whereas the ECLIPSE study found no difference in the rate of change of emphysema in higher GOLD stages, our study showed a clear pattern of greater change in this group. The ECLIPSE investigators also showed a weak correlation between decline in FEV₁ and change in Perc15 in 1953 subjects with COPD. Our study confirms and extends these findings by showing that the change in ALD persists when FEV₁ is included in the model for a cohort with the full range of lung function. The lack of a strong association between FEV₁ decline and change in lung density is not surprising since the correlation between Perc15 and FEV₁ at baseline in previously published COPDGene data was relatively weak, similar to results from the ECLIPSE study (23,24).

Mohamed Hoesein et al (22) investigated the rate of progression of emphysema in 3670 smokers included in the Dutch-Belgian Randomized Lung Cancer Screening Trial (or NELSON) lung cancer screening study. The authors showed that there was a significant annual increase in emphysema for all smokers over 3 years (irrespective of smoking status), but emphysema progression was not significantly correlated with decline in FEV₁ over the observation time. These results support our finding that the decline in FEV₁ only explains a small portion of the progression of CT-quantified emphysema, and together they suggest that ALD may have value in longitudinal studies of COPD as an independent marker of lung destruction.

One of the primary sources of variation in longitudinal quantification of emphysema is variation in lung volume. Previous studies have addressed this issue by using various methods to correct lung volume. Dirksen et al (24) proposed using a sponge model for correction, where the lung density was adjusted based on the ratio of TLV achieved at CT to the predicted TLC (from plethysmographic equations). The ECLIPSE investigators used a similar approach (23). However, plethysmographic TLC is measured in the upright position; CT performed in the supine position results in lower lung volume. For this reason, we used CT-based predicted TLV values calculated with equations from the MESA study (21).

The multicenter design of our study could have introduced possible sources of variability, such as differences in scanner model. We addressed this by including scanner model and study center in our analyses, but measurements could still be vulnerable to other sources of noise, such as differences in lung volume. As longitudinal quantification of emphysema and air trapping can change with variation in lung volume, spirometric control of inspiratory and expiratory acquisitions would be desirable. Because spirometric control would be very difficult to implement in a multicenter trial, we used CT-based predicted TLV values, as stated previously. Because we only included smokers who returned for an on-site 5-year follow-up visit, a selection bias might have been introduced by the absence of smokers who were ill or deceased.

Our findings, together with the already compelling amount of evidence from other groups, add to the value of quantitative CT in evaluating smokers with and those without COPD. CT could play an important role in future drug or other interventional trials, as well as in trials to evaluate specific subgroups of smokers, such as those in the PRISm group and rapid progressors. Despite this, routine use of quantitative CT to evaluate smokers in clinics or trials remains largely underutilized.

In conclusion, our results showed significant progression in emphysema and air trapping during 5-year follow-up in smokers with and those without chronic obstructive pulmonary disease. Progression of emphysema and air trapping correlate in part with progression in airflow obstruction, as measured with forced expiratory volume in 1 second (FEV₁) but can also progress without a significant change in FEV₁.