Residual Volume versus FRC Computed Tomography Assessment of Functional Small Airway Disease in Smokers with and without Chronic Obstructive Pulmonary Disease

To the Editor:

We write this letter to demonstrate the importance of using residual volume (RV) rather than FRC when assessing air trapping (functional small airway disease [fSAD]) via quantitative computed tomography (CT). A hallmark of obstructive lung disease is peripheral air trapping. Air trapping unrelated to emphysema has been defined as fSAD. Both emphysema and fSAD are responsible for the physiological feature of airflow obstruction. The main lung volume that increases with chronic obstructive pulmonary disease (COPD) severity is RV (1).

Using image-matching methods to tie CT voxels from images acquired at full inspiration (TLC) to voxels from images acquired at expiration, Galban and colleagues (2) adapted the parametric response map (PRM) to identify voxels classified on expiration as air trapped (less than −856 Hounsfield units [HU]) but not classified at TLC as emphysema (less than −950 HU). Such voxels were labeled PRM-fSAD. To move away from an HU threshold (less than −856 HU) on expiration (2), an alternative approach dubbed the “disease probability map” (DPM) (VIDA Diagnostics), uses a method based on the continuous distribution of HU values between inspiration and expiration and regional volume change (Jacobians) assessed from inspiratory-expiratory image matching to assign the probability of each voxel belonging to tissue subtypes: DPM-emphysema, DPM-fSAD, or DPM-normal (3, 4).

Although air trapping is traditionally evaluated at full exhalation (RV), some CT studies have acquired images at FRC. We hypothesized that fSAD assessed at RV will correlate more closely to spirometrically derived lung function indices, which are assessed by coaching a subject to a full expiratory effort. We further hypothesized that PRM- and DPM-fSAD measures will be affected differently when comparing assessments derived from RV efforts with those derived from FRC efforts. Some of the results of these studies have been reported previously in the form of an abstract (5).

Methods

We conducted this study using data from the University of Iowa cohort of the Genetic Epidemiology of COPD (COPDGene) database. Details of the study protocol have been published previously (6). The Iowa component is approved by the local institutional review board. Images of the Iowa cohort were acquired using a SOMATOM Force scanner (Siemens) at 3 lung volumes by coaching subjects to TLC, RV, and FRC. The coaching instructions are posted online (PDF found at www.spiromics.org/spiromics/node/4176). Post-bronchodilator spirometry was performed for measurement of FEV₁, FVC, and peak expiratory flow within 25–75% of the pulmonary FVC (FEF_25–75). CT-determined lung volumes were assessed by segmenting the lungs from the chest CT images.

CT image analysis

CT images were evaluated using the Apollo 2.0 software package (VIDA Diagnostics). CT DPM-fSAD was generated by TLC-to-RV registration (fSAD_RV) and TLC-to-FRC registration (fSAD_FRC) (LungPrint, VIDA Diagnostics), as previously described (3). CT PRM-fSAD measurements were generated via the Imbio software (Imbio), using the approach described by Galban and colleagues (2).

Statistical analysis

A Kruskal-Wallis test with a Dunn’s multiple comparisons test was performed for statistical comparison between control; preserved ratio impaired spirometry (PRISM); Global Initiative for Chronic Obstructive Lung Disease (GOLD) stages 1, 2, and 3/4; participant demographics, pulmonary function; and imaging measurements. For categorical variables, Fisher’s exact test was used. Spearman correlations were performed for DPM and PRM fSAD with pulmonary function measurements. A multivariable linear regression analysis (SAS 9.4 software, SAS Institute) was used to determine the associations for pulmonary function measurements with PRM and DPM fSAD at FRC and RV adjusted for age, sex, race, body mass index, and smoking pack-years.

Results

A total of 139 participants were evaluated (see Table 1 for demographics and CT-derived metrics by COPD group). There were no significant differences between the groups regarding age or sex. CT-TLC was significantly larger in GOLD 2 (5,453 ± 952 ml) compared with PRISM (4,522 ± 1,014 ml), whereas CT-RV was significantly larger in GOLD 2 (2,684 ± 603 ml) and GOLD 3/4 (3,154 ± 846 ml) than in GOLD 0 (1,978 ± 464 ml) and PRISM (2,060 ± 459 ml). There were no significant differences between COPD groups for CT-FRC (P < 0.05).

Table 1.

Subject Demographics

	GOLD 0 (n = 65)	PRISM (n = 19)	GOLD 1 (n = 15)	GOLD 2 (n = 25)	GOLD 3/4 (n = 15)
Age, yr	68.5 (10.0)	68.8 (7.2)	68.8 (8.8)	70.4 (8.6)	74.6 (6.5)
Female sex, n (%)	39 (60)	8 (42)	6 (40)	6 (24)	8 (53)
FEV1, %pred	100 (11)	73 (6)*	89 (7)	66 (9)*†	38 (7)*†
FVC, %pred	95 (11)	72 (8)*	103 (8)‡	81 (9)*†	66 (11)*†
FEV1/FVC, %	79 (4)	76 (5)	64 (4)*‡	61 (7)*‡	43 (6)*‡
FEF25-75, L	2.58 (0.91)	1.78 (0.74)	1.23 (0.38)*	0.96 (0.37)*‡	0.32 (0.11)*‡
Imaging measurements
Air volume (TLC), ml	4,830 (1,033)	4,522 (1,014)	5,478 (1,303)	5,453 (952)‡	5,030 (967)
Air volume (FRC), ml	3,457 (1,094)	3,204 (1,412)	3,559 (1,267)	3,213 (1,234)	3,589 (1,190)
Air volume (RV), ml	1,978 (464)	2,060 (459)	2,403 (708)	2,684 (603)*‡	3,154 (846)*‡
DPM-fSADFRC, %	13.2 (11.4)	14.4 (11.2)	10.7 (10.9)	17.3 (13.0)	16.4 (6.9)
DPM-fSADRV, %	10.7 (5.8)	13.0 (4.8)	12.1 (5.1)	18.9 (8.5)*	34.6 (13.4)*†‡
PRM-fSADFRC, %	32.2 (20.0)	23.2 (18.5)	31.3 (20.0)	35.1 (20.5)	39.8 (12.1)‡
PRM-fSADRV, %	8.2 (6.7)	7.2 (5.5)	12.1 (7.4)	15.4 (9.3)*‡	33.2 (12.7)*†‡

Definition of abbreviations: DPM = disease probability map; FEF_25-75 = peak expiratory flow within 25–75% of the pulmonary FVC; fSAD = functional small airway disease; GOLD = Global Initiative for Chronic Obstructive Lung Disease; PRISM = preserved ratio impaired spirometry; PRM = parametric response map; RV = residual volume.

Data are presented as mean ± SD unless otherwise indicated.

^*Significantly different from GOLD 0.

^† Significantly different from GOLD I.

^‡ Significantly different from PRISM.

As shown in Table 2 and Figure 1, there were highly significant correlations for DPM-fSAD_RV but only modest to borderline correlations of DPM-fSAD_FRC with FEV₁, FVC, FEV₁/FVC, and FEF_25–75. Multivariable linear regression analysis showed that after adjusting for covariates, only the associations for DPM-fSAD_RV with lung function measurements remained significant. Similarly, there were significant correlations between PRM-fSAD_RV and all pulmonary function testing metrics, but only weak correlations between PRM-fSAD_FRC and FEV₁/FVC and FEF_25–75 and no significant correlations with FEV₁ percent predicted or FVC percent predicted. Again, only the associations for PRM-fSAD_RV with lung function measurements remained after adjusting for covariates in the multivariable linear regression model. PRM-fSAD_FRC measures were more than double the PRM-fSAD values at RV (Table 1) in all COPD groups except GOLD 3/4.

Table 2.

Correlations between Disease Probability Map and Parametric Response Map Functional Small Airway Disease Acquired at FRC and Residual Volume with Pulmonary Function Measurements

	DPM-fSADFRC		DPM-fSADRV		PRM-fSADFRC		PRM-fSADRV
Unadjusted	r	P Value	r	P Value	r	P Value	r	P Value
FEV1 (%pred)	−0.19	0.03	−0.58	<0.0001	−0.03	0.72	−0.44	<0.0001
FVC (%pred)	−0.15	0.08	−0.44	<0.0001	0.09	0.29	−0.22	0.01
FEV1/FVC (%)	−0.18	0.03	−0.60	<0.0001	−0.26	0.002	−0.66	<0.0001
FEF25-75 (L)	−0.18	0.04	−0.66	<0.0001	−0.21	0.01	−0.65	<0.0001
Adjusted*	β-Value	P Value	β-Value	P Value	β-Value	P Value	β-Value	P Value
FEV1 (%pred)	−0.18	0.28	−1.41	<0.0001	−0.05	0.59	−1.16	<0.0001
FVC (%pred)	−0.11	0.36	−0.75	<0.0001	0.04	0.60	−0.51	<0.0001
FEV1/FVC (%)	−0.000	0.46	−0.009	<0.0001	−0.000	0.06	−0.008	<0.0001
FEF25-75 (L)	−0.002	0.78	−0.05	<0.0001	−0.004	0.41	−0.05	<0.0001

Definition of abbreviations: DPM = disease probability map; FEF_25-75 = peak expiratory flow within 25–75% of the pulmonary FVC; fSAD = functional small airway disease; PRM = parametric response map; RV = residual volume.

^*Adjusted analysis included age, sex, race, body mass index, and smoking pack-years; models with percent predicted values excluded age, sex, and race.

Figure 1.

Pearson correlations between disease probability map (DPM) and parametric response map (PRM) functional small airway disease (fSAD) measurements acquired at residual volume (RV) and FRC. Equations of the lines: (A) y = 0.21 × +5.62; (B) y = 0.24 × +11.70; (C) y = 1.44 × +11.67; (D) y = 0.98 × −2.56. Images to the right demonstrate DPM fSAD-RV (left column) versus DPM fSAD-FRC distributions (yellow) for representative participants across all Global Initiative for Chronic Obstructive Lung Disease categories (top to bottom). Green = DPM-normal; red = DPM-emphysema.

Figure 1 demonstrates significant differences between fSAD measurements generated from FRC and RV-lung CT images for both PRM (Figure 1A) and DPM (Figure 1B) (P < 0.001; R = 0.37; and P = 0.002; R = 0.26, respectively). At FRC, the relationship between fSAD assessed by PRM versus DPM (Figure 1C) was strong (r = 0.84; P < 0.0001; y = 1.44 × +11.67). However, as seen in Figure 1C and Tables 1 and 2, PRM-fSAD_FRC was approximately double that at RV for both PRM and DPM-fSAD_RV. The relationship between fSAD at RV by PRM and DPM (Figure 1D) is strong (r = 0.92; P < 0.0001) and close to the line of identity (y = 0.98 × −2.56). DPM fSAD distributions for representative participants across GOLD groups are demonstrated on the left of Figure 1.

Discussion

Our findings demonstrate that PRM-fSAD and DPM-fSAD RV measurements (2, 3, 6) provide a stronger differentiation between participants with and without COPD and greater associations with pulmonary function measurements than measurements acquired at FRC. These findings are not surprising, because spirometric measures incorporate an assessment using an RV effort. What is important to note is that a voxel label at FRC does not equate to the voxel label at RV, and voxels that represent fSAD are lost in analyses, which use FRC as the expiratory volume. The more than doubling of fSAD when assessed via PRM at FRC compared with RV is attributable to the use of a threshold criterion (voxels less than −856 HU). At FRC (more air than RV), more voxels fall below the −856 HU threshold. Thus, more are classified as fSAD. The DPM approach, which uses the probability of each tissue type versus a threshold, is more likely to detect emptying of some voxels less than −856HU at FRC and thus is less likely to classify these voxels as fSAD. A previous study comparing the spatial differences of PRM and DPM fSAD classified voxels also suggested that DPM may have increased sensitivity for classifying voxels close to the −856 HU threshold (3). However, further investigations into the pathophysiological mechanisms associated with the PRM and DPM classification differences are needed.

This study emphasizes that the measurement of fSAD by CT is critically dependent on breathing instructions. A precise assessment of fSAD is crucial for determining early disease etiology (7) and the efficacy of treatments in patients with early stages of COPD as well as in symptomatic smokers without spirometric evidence of obstructive lung disease (8). Therapies that target fSAD have the potential not only to reduce air trapping but also to slow and even reverse the development of COPD.

It is recognized that the measurement of fSAD by CT is critically dependent on breathing instructions. Coaching to RV in a clinical setting will require careful training of CT technologists to deliver instructions similar to those delivered in a pulmonary function laboratory.