Smaller Left Ventricle Size at Noncontrast CT Is Associated with Lower Mortality in COPDGene Participants

Abstract

Background

Smokers with chronic obstructive pulmonary disease (COPD) have smaller left ventricles (LVs) due to reduced preload. Skeletal muscle wasting is also common in COPD, but less is known about its contribution to LV size.

Purpose

To explore the relationships between CT metrics of emphysema, venous vascular volume, and sarcopenia with the LV epicardial volume (LV_EV) (myocardium and chamber) estimated from chest CT images in participants with COPD and then to determine the clinical relevance of the LV_EV in multivariable models, including sex and anthropomorphic metrics.

Materials and Methods

The COPDGene study (ClinicalTrials.gov identifier: NCT00608764) is an ongoing prospective longitudinal observational investigation that began in 2006. LV_EV, distal pulmonary venous blood volume for vessels smaller than 5 mm² in cross section (BV5), CT emphysema, and pectoralis muscle area were retrospectively extracted from 3318 nongated, unenhanced COPDGene CT scans. Multivariable linear and Cox regression models were used to explore the association between emphysema, venous BV5, pectoralis muscle area, and LV_EV as well as the association of LV_EV with health status using the St George’s Respiratory Questionnaire, 6-minute walk distance, and all-cause mortality.

Results

The median age of the cohort was 64 years (interquartile range, 57–70 years). Of the 2423 participants, 1806 were men and 617 were African American. The median LV_EV between Global Initiative for Chronic Obstructive Lung Disease (GOLD) 1 and GOLD 4 COPD was reduced by 13.9% in women and 17.7% in men (P < .001 for both). In fully adjusted models, higher emphysema percentage (β = –4.2; 95% confidence interval [CI]: –5.0, −3.4; P < .001), venous BV5 (β = 7.0; 95% CI: 5.7, 8.2; P < .001), and pectoralis muscle area (β = 2.7; 95% CI: 1.2, 4.1; P < .001) were independently associated with reduced LV_EV. Reductions in LV_EV were associated with improved health status (β = 0.3; 95% CI: 0.1, 0.4) and 6-minute walk distance (β = –12.2; 95% CI: –15.2, –9.3). These effects were greater in women than in men. The effect of reduced LV_EV on mortality (hazard ratio: 1.07; 95% CI: 1.05, 1.09) did not vary by sex.

Conclusion

In women more than men with chronic obstructive pulmonary disease, a reduction in the estimated left ventricle epicardial volume correlated with a loss of pulmonary venous vasculature, greater pectoralis muscle sarcopenia, and lower all-cause mortality.

© RSNA, 2020

Online supplemental material is available for this article.

Summary

Greater degrees of expiratory airflow obstruction were associated with smaller left ventricle size in smokers estimated from unenhanced chest CT; however, smaller left ventricle size was associated with lower mortality and better clinical outcomes.

Key Results

• ■ The estimated left ventricle epicardial volume (LV_EV), derived from postprocessing analysis of nongated, unenhanced chest CT, was reduced for participants with Global Initiative for Chronic Obstructive Lung Disease (GOLD) 4 chronic obstructive pulmonary disease (COPD) than for those with GOLD 1 COPD (13.9% in women and 17.7% in men; P < .0001 for both).

• ■ More emphysema, fewer pulmonary vessels less than 5 mm² in cross section, and lower pectoralis muscle area were independently associated with a smaller LV_EV.

• ■ A smaller LV_EV was associated with better health status, better 6-minute walk distance, and a reduced risk of death.

Introduction

Cardiac dysfunction is a major cause of morbidity and mortality in patients with chronic obstructive pulmonary disease (COPD). This dysfunction has been attributed to a combination of direct exposure to noxious substances such as tobacco smoke, inflammatory injury, and mechanical interdependence with a diseased lung (1–3). Several investigations that have explored these latter processes consistently demonstrated that higher amounts of emphysema and hyperinflation were associated with a smaller left ventricle (LV) (4–6). This phenomenon of the shrinking heart has been attributed to the compressive effects of the lung as well as to impaired LV filling because of emphysematous destruction of the intraparenchymal pulmonary vasculature (7). However, the clinical significance of this process has not been explored.

We have previously described the application of image-based analytic tools that can quantitatively assess the epicardial volume (myocardium and chamber) of the right ventricle and/or LV from noncardiac gated, unenhanced chest CT scans (8–10). Using these features, we wanted to replicate and expand on previously described associations among emphysema, the pulmonary vasculature (11), and LV volume. Patients with COPD also often have sarcopenia, but whether it may relate to a reduced LV epicardial volume (LV_EV) is unknown. Likewise, the potential relationship between smaller LV_EV and important clinical outcomes such as health status, exercise capacity, and/or risk of death, as well as how these relationships may vary by sex, are also unknown. We hypothesized that loss of the venous vasculature and lower pectoralis muscle area would be independent predictors of smaller LV_EV and that a reduced LV_EV would be predictive of poorer clinical outcomes such as worse health status, lower exercise capacity, and a higher risk of death. Here, we investigate these questions in COPDGene participants with expiratory airflow obstruction.

Materials and Methods

The COPDGene study (ClinicalTrials.gov identifier: NCT00608764) is a prospective longitudinal observational investigation (12). It was approved by the institutional review board of all participating centers, and all participants provided written informed consent. Health Insurance Portability and Accountability Act approval for this substudy was obtained as part of the institutional review board review at Brigham and Women’s Hospital. Participant characterization in COPDGene included volumetric inspiratory and expiratory nongated, unenhanced CT chest scans, questionnaires, and spirometry. Participants with previously diagnosed chronic lung diseases other than asthma, emphysema, or COPD were excluded.

Study Participants

This investigation of the clinical significance of the LV_EV represents a secondary analysis of the epidemiologic findings, imaging data, and outcomes collected as part of the round 1 COPDGene study visit performed between October 2006 and January 2011. Ever-smokers with expiratory airflow obstruction (ratio of forced expiratory lung volume in 1 second [FEV₁] to forced vital capacity <0.7 after the administration of a short-acting inhaled bronchodilator) were included in our analyses. Exclusion criteria for this substudy consisted of a lack of CT-based measures of emphysema, heart segmentation failure, and vascular segmentation failure.

CT Scan Acquisition and Postprocessing

The parameters used for image acquisition and reconstruction have been described in detail previously (12) and are briefly outlined in Appendix E1 (online). Objective characterization of emphysema-like tissue was performed using a local histogram-based technique and was reported as the percentage of the total lung volume occupied by emphysematous lung tissue (percentage of emphysema) for each participant (9,13–16). Morphologic assessments of the distal pulmonary vasculature were performed as described previously (17,18), and the arterial and venous vascular components were separated using a deep learning–based approach (10,19). Pulmonary blood volume for vessels smaller than 5 mm² in cross section (BV5) was calculated for each participant. Finally, cardiac segmentation was performed using an atlas-based approach as described previously (8,20). The estimated LV_EV reported herein included both the wall and chamber volume. The epicardial surface rendering of the heart was visually inspected for quality control for each participant. Postprocessing segmentation errors, such as the inclusion of chest wall in the cardiac segmentation, were excluded. Vascular and ventricular volumes were expressed in milliliters for all modeling. The ventricular volumes normalized to body surface area were also calculated for visual display of the mortality data (Appendix E1 [online]). Additional CT analyses included the quantification of pectoralis muscle area (expressed in square centimeters) on a single axial image above the level of the aortic arch (21–23). Additional details describing image postprocessing are available in Appendix E1 (online). Coronary artery calcium scoring was performed as described previously and reported as the nongated Agatston score (24). All CT analytics were performed on the inspiratory scan.

Lung Function, Health Status, 6-Minute Walk Distance, Mortality, and B-type Natriuretic Peptide

Spirometric measures of lung function, including FEV₁, forced vital capacity, and their ratio (FEV₁ to forced vital capacity) were performed using a spirometer (EasyOne; ndd Medical Technologies, Andover, Mass). Testing was performed before and after the administration of a short-acting inhaled bronchodilating medication (albuterol) per American Thoracic Society recommendations, and results were expressed as a percentage of predicted values (ie, FEV₁%) (25,26). Those with expiratory airflow obstruction were then classified as having Global Initiative for Chronic Obstructive Lung Disease (GOLD) stages 1–4 according to decrements in their FEV₁ (27). Health status was measured using the total score from the St George’s Respiratory Questionnaire (SGRQ) (28). Measures of the distance walked in 6 minutes (6-minute walk distance [6MWD]) were expressed in feet, and mortality was determined as described previously using data from the National Death Index and from the COPDGene long-term follow-up program (29). The B-type natriuretic peptide level was measured in a subset of COPDGene participants and is expressed as picograms per milliliter.

Statistical Analysis

Data are presented as medians and interquartile ranges for continuous covariates and as percentages for categoric variables. Assessment of LV morphologic characteristics across sex-stratified GOLD groups was performed using the Jonckheere-Terpstra test (30). Multivariable linear regression, as well as multivariable Cox regression, was used to predict the LV_EV as well as to assess the association between the LV_EV and the SGRQ, 6MWD, and mortality. All models used continuous measures of the percentage of emphysema, venous BV5, and pectoralis muscle area as well as additional adjustments for age at enrollment (in years), sex, race, height (in centimeters), weight (in kilograms), FEV₁%, smoking status (at the time of enrollment), CT-measured total lung capacity expressed as a percentage of predicted values, pack-years tobacco history, self-reported history of physician-diagnosed hypertension and congestive heart failure, resting heart rate, resting diastolic blood pressure, and nongated Agatston score unless otherwise specified. Additional models using the LV_EV normalized to the body surface area are presented in Appendix E1 (online). The reported P values are two sided, and P < .05 was considered to indicate a statistically significant difference. Analyses were performed with software (SAS, version 9.4; SAS Institute, Cary, NC).

Results

Participant Characteristics

The study included 4510 COPDGene participants who had expiratory airflow obstruction at their baseline study visit. Of these, 260 were excluded because of a lack of emphysema measures, 378 were excluded because of cardiac segmentation failure (Fig E1 [online]), and 554 were excluded for vascular segmentation failure, leaving 3318 participants (Fig 1). The median age of this group was 64 years; 1512 participants (46%) were women and 617 (19%) were African American (Table 1). Figure 2 presents the automated CT reconstructions obtained in one participant to compute the quantitative traits used to assess ventricular volume, pectoralis muscle area, and small vessel venous vascular volume. Figure E2 (online) presents the aggregated and sex-stratified distributions of the LV_EV according to spirometrically defined GOLD stages. The median LV_EV between GOLD 1 and GOLD 4 COPD was reduced by 13.9% in women and 17.7% in men (P < .001 for both).

Figure 1:

Consort diagram for ever-smokers with chronic obstructive pulmonary disease (COPD). GOLD = Global Initiative for Chronic Obstructive Lung Disease.

Table 1:

Clinical, Epidemiologic, and CT-based Features of Study Participants with Complete Data on Vascular and Left Ventricle Morphologic Characteristics

Figure 2:

Images in a 57-year-old man with body mass index of 28.3 kg/m² and forced expiratory lung volume in 1 second expressed as a percentage of predicted values of 58.5% from the COPDGene cohort with the different reconstructions for the left ventricle, pulmonary veins, and pectoralis muscles. A, Coronal view of the inspiratory CT scan at the level of the trachea. Upper lobe shows predominant advanced centrilobular emphysema (emphysema percentage = 29.0%). B, Segmentation of the pectoralis major (blue) and minor (green) at the level of the aortic arch used to compute the pectoralis muscle area. The pectoralis muscle area was 25.9 cm². C, Lateral view of the left ventricle epicardium model used to calculate the estimated left ventricle epicardial volume (159.6 mL). D, Display of the venous phase of the pulmonary vascular tree extracted with the scale-space particles approach and segmented by an artery-vein deep learning method used to compute the venous blood volume of vessels smaller than 5 mm² in cross section (62.9 mL). Color shading of the vessels represents vessel caliber.

Univariable Associations

We began by examining the univariable associations between the LV_EV, percentage of emphysema, venous BV5, and pectoralis muscle area (Table E1 [online]). The percentage of emphysema was inversely related to the LV_EV and pectoralis muscle area (P < .001 for both), whereas the venous BV5 and pectoralis muscle area were directly related to the LV_EV (P < .001 for both). No relationship existed between the venous BV5 and percentage of emphysema (P = .1).

Multivariable Models

Multivariable models were then created to predict the LV_EV (Tables 2, E2 [online]). The initial model included the percentage of emphysema (model 1). In this model, emphysema was inversely related to the LV_EV, where a 10% increase in the percentage of emphysema was associated with a 4.2 mL smaller LV_EV (P < .001). The venous BV5 was then added (model 2), and the previously described relationship between emphysema and the LV_EV remained. In this second model, a 10-mL increase in distal venous vascular volume was associated with a 6.9-mL increase in the LV_EV (P < .001). Our final multivariable model to predict the LV_EV (model 3) included the pectoralis muscle area. Again, the direction and magnitude of the associations between the LV_EV and previously described metrics of emphysema and venous vascular volume were consistent with the first two models, and, in this fully adjusted model, a 10-cm² increase in the pectoralis muscle area was associated with a 2.7-mL increase in the LV_EV (P < .001).

Table 2:

Multivariable Models to Predict Left Ventricle Epicardial Volume in Ever-Smoking COPDGene Participants with Chronic Obstructive Pulmonary Disease

Clinical Associations

Following examination of pulmonary and extrapulmonary radiologic features that predict epicardial morphologic characteristics of the LV, we wanted to explore its clinical associations. In fully adjusted models, a 10 mL higher LV_EV was associated with a 0.3 point higher (worse) SGRQ (P < .001) (Tables 3; E3 [online]). We also examined the relationship between the LV_EV and both the 6MWD and mortality. In both multivariable models (which included adjustment for the FEV₁% and percentage of emphysema), larger LV_EV volumes were associated with poorer clinical outcomes. A 10-mL increase in the LV_EV was associated with a 12-ft decrease in 6MWD (P < .001) (Tables 3; E4 [online]) and a 6% higher risk of death (hazard ratio: 1.06; 95% confidence interval: 1.04, 1.08 per 10-mL increase; P < .001) (Table 4; Fig E3 [online]).

Table 3:

Multivariable Models to Predict the SGRQ Total Score, 6MWD, and Brain Natriuretic Peptide in Ever-Smokers with COPD in the COPDGene Study

Table 4:

Predictors of All-Cause Mortality

Sex-based Differences in Clinical Outcomes

Finally, we sought to determine if some of the sex-based differences in clinical outcomes previously observed in COPD could be ascribed to the cardiovascular system as assessed by our measures of the LV epicardial volume. To do so, we performed additional stratified analyses and testing for sex interactions in our models. We found that an increased LV_EV was associated with greater decrements in both the health status as assessed according to the SGRQ and 6MWD in women than men (Table 3). However, there was no effect of sex in the association between LV_EV size and mortality (P = .97 for interaction) (Table 4).

B-type Natriuretic Peptide

To further understand the observed effect of sex on the association of LV_EV and clinical outcomes, we examined B-type natriuretic peptide data available from a subset of 236 COPDGene participants in our study cohort. The LV_EV was associated with the B-type natriuretic peptide (β = 6.6; 95% confidence interval: 4.8, 8.4), and this association was modified by sex. In stratified models, the increase in B-type natriuretic peptide per milliliter increase in LV_EV was more than twice as high in women than in men (Table 3).

Discussion

Using imaging, clinical, and epidemiologic data from the COPDGene study, we examined the relationship of CT-based metrics in left ventricle (LV) morphologic findings, emphysema, venous vascular morphologic characteristics, and pectoralis muscle area. We found that while increasing amounts of emphysema were associated with a smaller LV epicardial volume (LV_EV), greater distal venous vascular volume and pectoralis muscle area were associated with a larger LV_EV. Subsequent clinical modeling in analyses adjusted for lung function and percentage of emphysema revealed that a larger LV_EV was associated with poorer health status (increased St George’s Respiratory Questionnaire score), a reduced 6-minute walk distance (6MWD), and a higher risk of death. These clinical associations, at least for health status and 6MWD as well as the effect of LV enlargement on B-type natriuretic peptide level, varied significantly by sex.

As demonstrated by prior investigations, we found that the LV was smaller in those with more severe expiratory airflow obstruction and emphysematous destruction of the lung parenchyma (5,7). This phenomenon was thought to reflect a pathologic process such as underfilling of the ventricle. However, our findings showed that a smaller LV volume was associated with better and not worse clinical outcomes in COPD (5). A second finding of our analyses is that the association between the LV volume and clinical outcomes appears to reflect a sexually dimorphic process and that increases in LV size had a stronger effect on clinical outcomes and plasma-based measures of B-type natriuretic peptide in women than in men.

The association between emphysema-like tissue on CT scans and reduced LV volume has been demonstrated previously. The largest of these investigations was presented by Barr and colleagues (7) using population-based data from the Multi-Ethnic Study of Atherosclerosis. The authors found that a 10-point increase in the percentage of emphysema was associated with a 4.1-mL reduction in the LV end-diastolic volume. Our data are highly consistent with these results, as we demonstrated that a 10% increase in emphysema was associated with a 4.2-mL reduction in the LV_EV of those with COPD and a 6.6-mL reduction in the LV_EV of ever-smokers with normal lung function.

Several proposed mechanisms exist for the association between emphysema and LV volume. These include lung hyperinflation leading to both extrinsic compressive effects on the myocardial surface, impaired venous return to the thoracic cavity, and the secondary presence of pulmonary vascular remodeling resulting in reduced LV filling (5,7). Although our CT-based measures of the LV_EV cannot discern chamber and wall volumes, we sought to explore the latter hypothesis of impaired ventricular filling using our metrics of distal venous vascular volume.

The distal venous vascular volume was directly related to the LV_EV in univariable analyses and after multivariable adjustment, which included the CT-measured total lung capacity expressed as a percentage of predicted values and percentage of emphysema. Our data are complementary to and extend those reported by Aaron and colleagues (11), who found that the nonarterial and/or venous segmented total intraparenchymal blood vessel volume (from CT) and pulmonary microvascular blood volumes (from MRI) were directly related to LV end-diastolic volume. The aggregate of these efforts confirms that an aspect of reduced LV size in those with emphysema is due to reduced chamber blood volume because of impaired ventricular filling.

COPD is, however, a complex condition with a multitude of pulmonary and extrapulmonary manifestations. Acknowledging that, we expanded our analyses beyond a focused assessment of the interdependence of heart and lung and demonstrated that the loss of skeletal muscle (reduction in pectoralis muscle area) is associated with reduced epicardial volume of the LV. Further work is needed to explore this observation, but our findings suggest that reductions in pectoralis muscle area and LV_EV reflect a systemic process with a shared loss of skeletal muscle and myocardium.

Our multivariable modeling also demonstrated that a larger LV_EV was associated with modest reductions in both health status and 6MWD as well as a higher risk of death. These observations are somewhat paradoxical given the consistent observation of reduced LV size in smokers with more advanced COPD. In light of the well-documented association between higher LV volume and adverse clinical outcomes in patients with heart failure and other cardiac conditions (31–34), our results are not surprising. LV enlargement in the COPDGene cohort could simply represent the superimposition of additional pathologic findings, relative dilation, or enlargement of the ventricle on an already shrinking heart.

Finally, we found that the effects of the LV_EV on both health status and 6MWD were significantly modified by sex. Prior investigation has demonstrated that women may be at a higher risk for more severe dyspnea and poorer health status than men with the same degree of chronic tobacco smoke exposure (35), and part of this differential susceptibility for clinical impairment may be found in the cardiovascular system. Previous echocardiographic studies also suggest this may be the case. For example, Ky et al (36) found that among patients with chronic angina and enlarged ventricles, women had a greater risk for developing heart failure than men.

The sex-based differences in the associations of the LV volume extend beyond clinical outcomes to include plasma-based measures of B-type natriuretic peptide. Although previous work reported that B-type natriuretic peptide levels are higher in women than men (37), those findings were not related to differences in cardiac structure in a population-based study of largely healthy individuals. Our data expand on those observations by suggesting that in COPD, there is greater production of B-type natriuretic peptide (which may be due to relatively greater increases in transmural wall stress [38]) per milliliter increase in LV volume in women than in men. Although these findings do not directly explain the sex-based differences in health status or 6MWD reported herein, it is possible that excess myocardial wall stress per unit volume change of the LV_EV in women may be responsible for the differences we observe in the SGRQ score and 6MWD.

Our study has limitations. Our CT-based analytics of the epicardial surface of the LV did not discern ventricular chamber from myocardium, and we therefore could not differentiate the pattern of cardiac remodeling, including the relationship between ventricular wall thickness and chamber volume, nor could we directly assess cardiac function (31). These data were not cardiac gated, and some of the data had both systolic and diastolic phases on the same slice, causing blurring of the epicardial boundaries or cardiac motion between the slices. This will give rise to blurred epicardial boundaries, which are most evident on sagittal and coronal reformations. Those participants with pericardial fluid collections may have a spuriously high estimated LV_EV index. We also lacked a large cohort of age-, sex-, and race-matched controls from which we could generate normative standards for our measures of ventricular volume, pectoralis muscle area, and venous vascular volume. These data would allow us to better understand the prevalence of pathologic findings represented by these features rather than by our provided associations with these features. Finally, we used cross-sectional radiologic data, which may or may not correctly infer heart size with advancing age and disease progression. Additional observational studies (prior to the onset of disease) would be needed to establish the dynamic nature of heart volumes in health and disease.

In summary, multiple pulmonary and extrapulmonary manifestations of smoking-related injury are found to be significantly associated with the estimated left ventricular (LV) volumes derived from noncontrast chest CT in those participants with chronic obstructive pulmonary disease. Reductions in the estimated LV volume correlated with the severity of expiratory airflow obstruction and may reflect an adaptive process to the pulmonary and systemic manifestations of this chronic respiratory disease. Deviations from this tendency (ie, LV enlargement or lack of LV shrinkage from preexisting cardiac disease) were associated with poorer clinical outcomes and associations, which are most evident in women.