Mechanisms of Exertional Dyspnea in Patients with Mild COPD and a Low Resting DL_CO

Abstract

Patients with mild chronic obstructive pulmonary disease (COPD) and lower resting diffusing capacity for carbon monoxide (DL_CO) often report troublesome dyspnea during exercise although the mechanisms are not clear. We postulated that in such individuals, exertional dyspnea is linked to relatively high inspiratory neural drive (IND) due, in part, to the effects of reduced ventilatory efficiency. This cross-sectional study included 28 patients with GOLD I COPD stratified into two groups with (n = 15) and without (n = 13) DL_CO less than the lower limit of normal (<LLN; Global Lung Function Initiative criteria) and 16 healthy controls. We compared dyspnea (Borg scale), IND (by diaphragm electromyography), ventilatory equivalent for CO₂ (V̇_E/V̇CO₂), and respiratory mechanics during incremental cycle exercise in the three groups. Spirometry and resting lung volumes were similar between COPD groups. During exercise, dyspnea, IND and V̇_E/V̇CO₂ were higher at equivalent work rates (WR) in the DL_CO<LLN group compared with the other two groups (all p < 0.05). In patients with DL_CO<LLN, severe respiratory mechanical constraints, indicated by end-inspiratory lung volume of approximately 90% of total lung capacity, occurred at a lower WR than the other two groups (p < 0.05). The dyspnea/IND relationship was similar across groups; therefore, the increased dyspnea at a standardized WR in the low DL_CO<LLN group reflected the higher corresponding IND. Higher dyspnea ratings in patients with mild COPD and DL_CO<LLN were associated with higher IND and V̇_E/V̇CO₂ at a given work rate. Higher ventilatory requirements in the DL_CO<LLN group accelerated dynamic mechanical abnormalities earlier in exercise, further increasing IND and dyspnea.

Keywords:

• COPD
• inspiratory neural drive
• DLCO
• dyspnea
• cardiopulmonary exercise testing

Introduction

Patients with mild (Global Initiative for Chronic Obstructive Lung Disease (GOLD) stage I) chronic obstructive pulmonary disease (COPD) can experience significant dyspnea during physical exertion [1,2], leading to reduced physical activity [2], poor quality of life [2], and increased mortality [3]. It is evident that heterogenous pathophysiological abnormalities of the small airways [1] and pulmonary microvasculature [4–6] exist in mild COPD and may be clinically important [6]. Broadly speaking, these abnormalities point to derangements of dynamic respiratory mechanics [7] and/or pulmonary gas exchange [8], which are amplified by exercise and require compensatory responses by the respiratory controller to maintain alveolar ventilation in pace with increasing metabolic demands [9]. While such physiologic adaptations (i.e. increased inspiratory neural drive, IND) are usually successful in maintaining arterial blood gas and acid-base homeostasis during exercise, the cost is often increased perceived dyspnea which can contribute to exercise intolerance [8,9]. In mild COPD, the interrelationships between these diverse physiological abnormalities of pulmonary gas exchange and dynamic mechanics, increased IND, and dyspnea, are poorly understood and are the main focus of this paper.

Small airway loss and dysfunction are the hallmark of mild COPD and can lead to variable expiratory flow limitation, dynamic hyperinflation and prematurely reduced inspiratory reserve volume (IRV) at relatively low minute ventilation (V̇_E) during exercise [1,10]. Thus, increased resistive and elastic loading of the inspiratory muscles can lead to compensatory increases in IND (cortical-motor command output) and perceived increased respiratory effort or discomfort [9].

Importantly, the impact of abnormal respiratory mechanics on dyspnea in mild COPD needs to be considered in the context of the prevailing ventilatory demand, which mainly reflects the adaptations of the central respiratory controller to altered chemical stimuli. Key to this is the influence of reduced ventilatory efficiency, which was present in ∼30% of individuals with spirometrically-defined COPD in a Canadian open population sample (n = 715) [11]. Ventilatory efficiency is measured as the ventilatory equivalent for carbon dioxide production (V̇_E/V̇CO₂) during exercise (specifically as the lowest value which occurs at the anaerobic threshold or soon thereafter), and is considerably higher (i.e. reduced efficiency) in mild COPD patients than in age- and sex-matched healthy controls [1,8,9,12]. High V̇_E/V̇CO₂ is believed to mainly reflect an increase in chemo-stimulation due to a high physiological dead space (V_D) [8], rather than increased central chemosensitivity, per se [8,13]. Thus, high V̇_E/V̇CO₂ during sub-maximal exercise suggests that regional pulmonary perfusion is diminished relative to alveolar ventilation [8,14,15].

In addition to high exercise V̇_E/V̇CO₂, reduced single breath DL_CO is an indirect marker of microvascular dysfunction (reduced capillary blood volume) and/or capillary destruction in emphysema [16–18]. These related physiological markers have together been closely linked to higher dyspnea ratings in mild COPD [18–20]. However, the relative importance of the effects of microvascular dysfunction (low DL_CO, high V̇_E/V̇CO₂) and mechanical factors (reduced dynamic IRV) in contributing to increased IND and exertional dyspnea in mild COPD has never been determined.

Therefore, the main objective of this study was to compare dyspnea ratings, IND, V̇_E/V̇CO₂, and dynamic respiratory mechanics in two mild COPD groups with and without a DL_CO less than the lower limit of normal (LLN) and a reference control group during exercise. We postulated that: 1) patients with DL_CO<LLN would have higher V̇_E/V̇CO₂, IND and exertional dyspnea intensity than patients with similar resting mechanical abnormalities but with DL_CO≥LLN and 2) that the higher IND in the DL_CO<LLN group would be explained by higher ventilatory demand prompting an earlier onset of significant mechanical constraints.

Materials and methods

See online supplement for detailed methods.

Participants

This cross-sectional observational study included 28 participants with GOLD stage I COPD (forced expiratory volume in 1 s (FEV₁)/forced vital capacity (FVC) ratio < 0.7, and FEV₁≥80% of predicted value) [21] and 16 healthy controls who completed pulmonary function tests (PFTs) and a cardiopulmonary exercise test (CPET) with measurements of diaphragm electromyography (EMGdi) and respiratory pressure manometry. They represent a convenience sample recruited from previous Respiratory Investigation Unit and Canadian Respiratory Research Network Studies (Figure E1). Some participants’ data were included in previous studies, but there is no overlap in the analysis or results. Individuals with cardiovascular, metabolic, or other conditions that could contribute to reduced ventilatory efficiency and/or dyspnea were excluded. All protocols were approved by the Queen’s University and Affiliated Teaching Hospitals Research Ethics Board (DMED-1659-13).

Study design and procedures

Medical history and symptom and activity questionnaires [22–24] were completed. Spirometry, body plethysmography, and resting single-breath DL_CO were performed using automated equipment following established guidelines [25–27]. Chest computed tomography (CT) scans were qualitatively assessed by two experts (DEO,AFE) for the presence and severity of emphysema when available (19/28 COPD participants) [28]. COPD participants were stratified into two groups by DL_CO values ≥ or < the 5th percentile of the LLN (z score=–1.64) using Global Lung Function Initiative (GLI) equations [29]. Symptom-limited, incremental CPET was conducted on an electronically braked cycle ergometer, as previously described [8,9]. Select PFT and CPET measurements were expressed as percent of predicted normal values and/or z scores [29–31].

An esophageal catheter with multi-paired electrodes and dual pressure balloons was inserted nasally and positioned as previously described [9,32–34]. Raw EMGdi signal was sampled continuously and converted to a root mean square (RMS). The RMS EMGdi from tidal breaths was determined manually by selecting peak data between cardiac QRS complexes and averaging values of all breathing during exercise time periods. Maximal EMGdi (EMGdi, max) was measured during serial IC maneuvers (pre-exercise baseline, submaximal and peak exercise). Tidal EMGdi relative to maximum (EMGdi%max) was used as a surrogate measure for IND to the crural diaphragm based on a number of assumptions previously described [32–34].

Esophageal (Pes) and gastric (Pga) pressures were measured continuously via differential pressure transducers. Transdiaphragmatic (Pdi) pressure was calculated as Pga minus Pes. Maximal Pes (Pes,max) and Pdi (Pdi,max) were obtained from serial sniff (pre-exercise baseline and peak) and IC maneuvers, respectively. Tidal Pes relative to maximum Pes (Pes%max) was used as an index of respiratory muscle effort [9]. The breath selection for EMGdi and respiratory pressures occurred in the same 30 s window (first 30 s of second minute of exercise in each work rate period) as all other cardiorespiratory parameters.

Statistical analysis

The sample size was calculated a priori to detect a one Borg-unit difference at 80% power. One-way analysis of variance (ANOVA) with post-hoc Bonferroni corrected t-tests were used for between-group comparisons of demographics, PFT, CPET, neurophysiologic, and sensory variables. Sex as a covariate was included in ANOVA for key outcomes. Prevalence of comorbidities, medication use, and smoking status was compared between groups using Fisher’s exact test. Emphysema severity comparisons (between COPD groups) were analyzed using Fisher’s exact test. Slope of the relationship between dyspnea-to-oxygen uptake (V̇O₂), and IND-to-work rate (WR) were determined using linear regression. Hey plots of tidal volume (V_T) versus V̇_E were used to determine individual V_T/V̇_E plateaus. Pearson’s correlation tested the associations between variables of interest. Data are presented as mean ± standard deviation (SD) unless otherwise indicated. Statistical significance was set a priori at p < 0.05. Data were analyzed using Statistical Package for Social Sciences (SPPS v. 26)

Results

Study participants

Participant characteristics are summarized in Table 1. Fifteen COPD participants had DL_CO<LLN and 13 had DL_CO≥LLN. The majority of COPD participants (24/28) were symptomatic by respiratory questionnaires (MRC and/or CAT, i.e. GOLD 1B), and COPD participants were more symptomatic than controls. Respiratory medication use and frequency of comorbidities were similar between COPD groups. Emphysema of trace-mild severity was evident on CT scans, and was not different between DL_CO groups (p = 0.245).

Table 1. Participant characteristics.

Parameter	Mild COPD DLCO<LLN (n = 15)	Mild COPD DLCO≥LLN (n = 13)	Control (n = 16)	p value
Age (years)	69 ± 8*	69 ± 7^†	61 ± 8	p = 0.0083
Males/females (n)	6/9	9/4	7/9	p = 0.2730
Height (cm)	163.2 ± 8.0	169.9 ± 8.6	166.4 ± 9.4	p = 0.1644
Body mass (kg)	70.10 ± 13.8	82.5 ± 15.2	77.5 ± 11.2	p = 0.0563
Body mass index (kg · m^-2)	26.3 ± 5.1	28.8 ± 6.0	27.9 ± 2.4	p = 0.3626
Smoking history (pack-years)	44.2 ± 23.9*	45.7 ± 25.7^†	0.2 ± 0.6	p < 0.0001
Smoking status, n (%)
Current	5 (33)	1 (8)	0 (0)	p < 0.0001
Ex-smoker	10 (67)	12 (92)	3 (19)	p < 0.0001
Never	0 (0)	0 (0)	13 (81)	p < 0.0001
MRC dyspnea scale (1–5)	2.1 ± 0.7*	1.9 ± 0.6^†	1.1 ± 0.3	p < 0.0001
BDI focal score (0–12)	8.6 ± 1.6*	8.6 ± 2.1^†	11.3 ± 0.9	p = 0.0001
CAT (0–40)	15.7 ± 8.6*	12.5 ± 7.4^†	3.7 ± 4.1	p = 0.0001
Inhaled medications, n (%)
SABA	5 (33)	5 (33)	0	p = 1.0000
SAMA	1 (7)	1 (7)	0	p = 0.5436
LAMA	5 (33)	5 (33)	0	p = 0.7222
ICS	2 (13)	0 (0)	0	p = 0.2777
ICS-LABA	2 (13)	4 (27)	0	p = 0.4564
Comorbidities, n (%)
Hypertension	8 (53)	8 (62)	3 (20)	p = 0.0410
Pulmonary hypertension	0	0	0	N/A
Other cardiac	5 (33)	7 (54)	2 (13)	p = 0.0568

Values are mean ± SD or number and (%) of participants. MRC: Medical Research Council; BDI: Baseline dyspnea index; CAT: COPD assessment test; SABA: short acting β₂-agonist; SAMA: short-acting muscarinic antagonist; LAMA: long-acting muscarinic antagonist; ICS: inhaled corticosteroid; LABA: long-acting β₂-agonist. p values are from one-way ANOVA or Fisher’s exact test. Significance markers indicate post hoc between group comparisons.

* p < 0.05 DL_CO<LLN vs. control.

ǂ p < 0.05 DL_CO<LLN vs. DL_CO≥LLN.

† p < 0.05 DL_CO≥LLN vs. control.

Both COPD groups had a relatively preserved post-bronchodilator FEV₁, a similar reduction in forced expiratory flow (25–75% of FVC) compared with control (Table 2). Both COPD groups had lower resting IC, and increased FRC and residual volume/total lung capacity (RV/TLC) compared with controls. The TLC-alveolar volume (V_A) difference was similar between COPD groups. The carbon monoxide transfer coefficient (K_CO) was lower in DL_CO<LLN compared with DL_CO≥LLN. All controls had a DL_CO≥LLN. During CPET, V̇O_2peak and WR_peak were lower in both COPD groups compared with controls (Table 2). The anaerobic threshold (AT) occurred at a lower V̇O₂ in DLCO < LLN (1.04 ± 0.23 L/min) compared with control (1.71 ± 0.70 L/min, p = 0.001), but was not different between DLCO < LLN and DLCO > LLN (1.21 ± 0.31 L/min, p = 0.505), or DLCO > LLN and control (p = 0.073).

Table 2. Resting pulmonary function and select peak exercise measurements.

Parameter	Mild COPD DLCO<LLN (n = 15)	Mild COPD DLCO≥LLN (n = 13)	Control (n = 16)	p value
Post-bronchodilator
FEV1 (%predicted)	90 ± 10*	92 ± 11^†	111 ± 10	p < 0.0001
FEV1 z score	–1.13 ± 0.64*	–0.82 ± 0.63^†	–0.76 ± 0.69	p < 0.0001
FVC (%predicted)	109 ± 19	109 ± 15	113 ± 10	p = 0.7317
FVC z score	0.28 ± 1.11	0.35 ± 0.72	0.86 ± 0.70	p = 0.1532
FEV1/FVC	0.61 ± 0.06*	0.63 ± 0.03^†	0.77 ± 0.05	p < 0.0001
FEV1/FVC z score	–2.03 ± 0.61*	–1.74 ± 0.20^†	–0.22 ± 0.80	p < 0.0001
FEV1/FVC < LLN (%)	67	69	0
Pre-bronchodilator
FEV1 (%predicted)	81 ± 12*	82 ± 13^†	107 ± 11	p < 0.0001
FEV1/FVC	0.58 ± 0.07*	0.59 ± 0.04^†	0.75 ± 0.06	p < 0.0001
FEF25 –75% (%predicted)	41 ± 21*	41 ± 12^†	110 ± 58	p < 0.0001
TLC (%predicted)	99 ± 16	99 ± 10	99 ± 13	p = 0.9965
SVC (%predicted)	110 ± 16	109 ± 14	116 ± 110	p = 0.3879
IC (%predicted)	88 ± 14*	88 ± 18^†	107 ± 13	p = 0.0011
FRC (%predicted)	109 ± 23*	112 ± 23^†	91 ± 21	p = 0.0251
RV (%predicted)	101 ± 29*	106 ± 23^†	81 ± 22	p = 0.0201
RV/TLC	0.39 ± 0.08*	0.38 ± 0.05^†	0.28 ± 0.06	p = 0.0001
DLCO (%predicted)	57 ± 12*^,ǂ	82 ± 7^†	98 ± 11	p < 0.0001
DLCO z-score	–3.42 ± 1.38*^,ǂ	–1.15 ± 0.43^†	–0.13 ± 0.67	p < 0.0001
V A (%predicted)	88 ± 8*	93 ± 9	98 ± 10	p = 0.0194
KCO (%predicted)	65 ± 13*^,ǂ	87 ± 8^†	100 ± 16	p < 0.0001
TLC–V A (L)	1.08 ± 0.58*	1.10 ± 0.42^†	0.57 ± 0.31	p = 0.0030
Peak Exercise
V̇O2, L/min	1.38 ± 0.34*	1.76 ± 0.39^†	2.59 ± 0.87	p < 0.0001
V̇O2 %predicted	86 ± 14*	90 ± 14^†	136 ± 30	p < 0.0001
Work rate, W	84 ± 29*	102 ± 25^†	169 ± 62	p < 0.0001
Work rate %predicted	80 ± 28*	81 ± 19^†	128 ± 25	p < 0.0001

Values are mean ± SD. FEV₁: forced expiratory volume in 1 s; FVC: forced vital capacity; FEF_25–75: forced expiratory flow between 25 and 75% of FVC; TLC: total lung capacity; SVC: slow vital capacity; IC: inspiratory capacity; FRC: functional residual capacity; RV: residual volume; DL_CO: diffusing capacity of the lung for carbon monoxide; V_A: alveolar volume; K_CO: carbon monoxide transfer coefficient; V̇O₂: oxygen uptake; W: watts. p values are from one-way ANOVA. Significance markers indicate post hoc between group comparisons.

* p < 0.05 DL_CO<LLN vs. control.

ǂ p < 0.05 DL_CO<LLN vs. DL_CO≥LLN.

† p < 0.05 DL_CO≥LLN vs. control.

Neurophysiologic response to exercise

IND for a given WR was consistently elevated in both COPD groups compared with controls and was significantly higher in DL_CO<LLN compared with DL_CO≥LLN at 40 and 60 W (Figure 1A). The DL_CO<LLN group had a greater degree of neuromuscular dissociation (EMGdi%max:Pdi%max, Figure 1B) at 60 W compared with DL_CO≥LLN and control. There was also a greater degree of neuromechanical dissociation (EMGdi%max:V_T/VC%predicted, Figure 1C) in both COPD groups compared with controls at 60 W. Throughout exercise, respiratory muscle effort (Pes%max) was similar in both COPD groups, but elevated compared to controls (Figure 1D).

Figure 1. Diaphragm electromyography (EMGdi) and respiratory pressure manometry responses to incremental exercise. A) ratio of tidal EMGdi to maximum EMGdi (EMGdi%max), B) ratio of EMGdi%max to transdiaphragmatic pressure (Pdi) as a percentage of maximum Pdi (Pdi%max), C) ratio of EMGdi%max to tidal volume (V_T) as a percentage of predicted vital capacity (VCpr), and D) esophageal pressure (Pes) as a percentage of maximum Pes (Pes%max) all relative to work rate. Data are presented as mean ± SEM. DL_CO: diffusing capacity for carbon monoxide; LLN: Lower limit of normal. *p < 0.05 DL_CO<LLN vs. control, ǂp < 0.05 DL_CO<LLN vs. DL_CO≥LLN, ^†p < 0.05 DL_CO≥LLN vs. control.

Ventilatory response to exercise

During exercise, V̇_E was consistently elevated in both COPD groups compared with control (Figure 2A). Additionally, V̇_E/V̇CO₂ was consistently elevated in the DL_CO<LLN compared to control throughout exercise (Figure 2B). At 60 W, V̇_E/V̇CO₂ was higher in DL_CO<LLN compared with DL_CO≥LLN (p = 0.0362) and DL_CO≥LLN compared with controls (p = 0.0456). V̇_E/V̇CO₂ nadir (lowest 30 s average) was higher in DL_CO<LLN (34 ± 4) compared with DL_CO≥LLN (29 ± 4, p = 0.007) and controls (27 ± 3, p = 0.001). The slope of V̇_E-to-V̇CO₂ was greater in DL_CO<LLN (32 ± 4) compared with the DL_CO>LLN (28 ± 4, p = 0.0209) and controls (27 ± 3, p < 0.0001). At the AT, V̇_E/V̇CO₂ was higher in DL_CO<LLN (35 ± 4) compared with DL_CO>LLN (31 ± 4, p = 0.0348) and control (29 ± 3, p < 0.0001), and was higher in DL_CO>LLN compared with control (p < 0.0001).

Figure 2. Ventilatory and gas exchange response to incremental exercise. A) minute ventilation (V̇_E), B) ventilatory equivalent for carbon dioxide (V̇_E/V̇CO₂), C) partial pressure of end-tidal CO₂ (P_ETCO₂), and D) oxygen saturation estimated by pulse oximetry (SpO₂). Data are presented as mean ± SEM. DL_CO: diffusing capacity for carbon monoxide; LLN: Lower limit of normal. *p < 0.05 DL_CO<LLN vs. control, ^ǂp < 0.05 DL_CO<LLN vs. DL_CO≥LLN, ^†p < 0.05 DL_CO≥LLN vs. control.

Partial pressure of end-tidal CO₂ (P_ETCO₂) was lower in DL_CO<LLN compared with controls throughout exercise but there were no differences between COPD groups (Figure 2C). Estimated arterial oxygen saturation was not different between groups throughout exercise (Figure 2D).

Tidal volume relative to vital capacity (V_T/VC) was elevated in DL_CO<LLN compared with controls. At 60 W, V_T/VC was elevated in DL_CO<LLN compared with DL_CO≥LLN at 60 W (Figure 3A). 87% of the DL_CO<LLN participants reached a V_T/V̇_E plateau compared with 69% of DL_CO≥LLN participants (Figure E2). The plateau occurred at a lower average WR in DL_CO<LLN than DL_CO≥LLN (66 ± 31 W vs. 90 ± 23 W, p = 0.049). During exercise, both COPD groups breathed at higher operating lung volumes (lower IC and IRV) compared to controls (Figure 3C,D). The DL_CO<LLN group encroached on critical mechanical limits earlier in exercise compared with DL_CO≥LLN, reaching an end-inspiratory lung volume (EILV)>90% of TLC at a lower WR (76 ± 25 vs. 114 ± 35 W, p = 0.025).

Figure 3. Breathing pattern and operating lung volume responses to incremental exercise. A) tidal volume (V_T) as a percentage of vital capacity (VC), B) breathing frequency (Fb), C) inspiratory capacity (IC) as a percentage of total lung capacity (TLC), and D) inspiratory reserve volume (IRV) as a percentage of TLC. Data are presented as mean ± SEM. DL_CO: diffusing capacity for carbon monoxide; LLN: Lower limit of normal. *p < 0.05 DL_CO<LLN vs. control, ^ǂp < 0.05 DL_CO<LLN vs. DL_CO≥LLN, ^†p < 0.05 DL_CO≥LLN vs. control.

Sensory response to exercise

Dyspnea intensity during exercise was greater in DL_CO<LLN compared with DL_CO≥LLN and controls at 40 and 60 W (Figures 4A and 5A). The relationship between dyspnea intensity as a function of increasing IND and V̇_E as a percentage of maximum voluntary ventilation (MVV) was similar between groups (Figure 4B,C).

Figure 4. Dyspnea intensity during incremental exercise expressed relative to A) work rate, B) the ratio of diaphragmatic electromyography to maximal EMGdi (EMGdi%max) C) ventilation (V̇_E) relative to maximum voluntary ventilation (MVV). Data are presented as mean ± SEM. Arrow indicates comparison at highest-equivalent work rate of 60 W between DL_CO<LLN and the DL_CO≥LLN and control groups. DL_CO: diffusing capacity for carbon monoxide; LLN: Lower limit of normal. *p < 0.05 DL_CO<LLN vs. control, ^ǂp < 0.05 DL_CO<LLN vs. DL_CO≥LLN.

Figure 5. Individual exercise responses at 60 W for A) perceived dyspnea intensity (Borg scale), B) ventilation relative to carbon dioxide clearance (V̇_E/V̇CO₂), C) diaphragm electromyography (EMGdi) as a percentage of maximal EMGdi during inspiratory capacity maneuvers (EMGdi%max), and D) end-expiratory lung volume (EELV) as a percentage of total lung capacity (TLC). Data points are individual participant values with mean crossbars and standard deviation error bars. DL_CO: Diffusing capacity for carbon monoxide; LLN: Lower limit of normal.

Correlative analysis

IND-WR slopes correlated with V̇_E/V̇CO₂ nadir p (r = 0.51, p < 0.001), and DL_CO%predicted (r=–0.67, p < 0.001). DL_CO%predicted in turn correlated with V̇_E/V̇CO₂ nadir (r = 0.43, p = 0.021) and the dyspnea-V̇O₂ slope (r=–0.69, p < 0.001).

Discussion

Our results showed that: 1) mild COPD patients with DL_CO<LLN had higher V̇_E/V̇CO₂, IND and dyspnea ratings during exercise compared with those with a DL_CO≥LLN, despite similar resting respiratory mechanics; 2) The higher ventilatory requirements in the DL_CO<LLN accelerated the development of significant respiratory mechanical constraints earlier in exercise, which further increased IND and dyspnea; 3) Dyspnea-IND slopes were similar in all three groups, therefore, the higher dyspnea ratings at a standardized WR in the DL_CO<LLN group reflected simultaneously higher IND. This study provides new insights into the nature of the relationship between low DL_CO and exertional dyspnea in patients with mild COPD.

Mild COPD is characterized by heterogeneous physiological derangements of pulmonary gas exchange and dynamic respiratory mechanics that are best uncovered by comparisons with healthy, age-matched controls during exercise (Figures 1–3, 5). The gradual evolution of these complex pathophysiological impairments in susceptible smokers ultimately results in adaptive compensatory increases of IND to maintain adequate alveolar ventilation (V̇_A) during physical activity [9,35]. In this context, our results showed that the rise in IND, as a function of increasing power output, is higher in the two COPD groups than in healthy controls, being most pronounced in DL_CO<LLN (Figure 1A). Higher IND for a given WR, V̇O₂ or V̇_E in COPD is multifactorial. Previous studies in mild COPD have focused on the role of increased airway resistance and mechanical loading of the inspiratory muscles [9]. Under these conditions, increased cortical motor command output and electrical activation of the diaphragm (and other inspiratory muscles) is required to achieve adequate force generation to sustain V̇_E commensurate with metabolic demand. Such factors are undoubtedly at play in the current COPD sample, particularly near peak exercise, when EILV is ∼89% of TLC [7,9]. The current study extends previous work by considering additional contributors to increased IND, such as increased chemo-stimulation and its negative impact on dynamic respiratory mechanics, particularly at lower exercise intensity.

The resting DL_CO and V̇_E/V̇CO₂ are arguably indirect markers of pulmonary microvascular dysfunction/destruction and are linked to important clinical outcomes [18–20,36]. The single-breath DL_CO consists of three components: pulmonary capillary blood volume (Vc), diffusing membrane capacity (Dm) and distribution of V̇_A. Importantly, the TLC-V_A differences were similar in the two COPD groups suggesting that lower V̇_A as a fraction of TLC (i.e. poor distribution of inhaled gas mixture) did not explain differences in DL_CO, thus confirming that the other factors are important. Tedjasaputra et al. [16] recently demonstrated that both DL_CO and Vc were reduced during exercise (while Dm and SpO₂ were similar) in mild COPD, compared with healthy controls. Thus, low Vc may partly explain the link between low DL_CO and high V̇_E/V̇CO₂.

Participants with lower DL_CO had higher V̇_E/V̇CO₂ (nadir and slope) than the other two groups, indicating reduced ventilatory efficiency. In fact, V̇_E/V̇CO₂ nadir averaged ∼34 in DL_CO<LLN– a cutoff value that has previously been linked to poorer survival in COPD [36]. It is known that V̇_E/V̇CO₂ is influenced by: 1) the regulated resting arterial CO₂ (PaCO₂) (a variable known to be influenced by increased chemo-sensitivity) [8,13] and 2) the magnitude of the physiological dead space (number of alveolar lung units with high V̇_A/Q̇ ratios) [8]. Previous studies in mild COPD have established that enhanced central (and peripheral) chemo-sensitivity are unlikely contributors to increased V̇_E/V̇CO₂ nadir during exercise [8,13,37,38]. Indeed, in one such study high V̇_E/V̇CO₂ was more closely correlated with measured dead space than resting PaCO₂ in mild COPD [8]. Further although peripheral chemoreceptors are known to influence the ventilatory response to hypoxia during exercise in health [39,40], previous work in non-hypoxemic mild/moderate COPD has shown that enhanced peripheral chemosensitivity is unlikely to explain the increased ventilatory response (i.e. high V̇_E/V̇CO₂ nadir) during exercise [37,38].

Interestingly, in the current study, P_ETCO₂ as a function of increasing WR was lowest in the DL_CO<LLN group even though breathing pattern (which can influence P_ETCO₂ at a given PaCO₂) was similar between the COPD groups (Figure 2C). The low P_ETCO₂ at rest and throughout exercise in DL_CO<LLN could reflect the effects of reduced pulmonary blood perfusion (and V_c) and consequent inefficiencies in CO₂ clearance [5,8]. In addition, we cannot exclude possible between-group differences in alveolar ventilation, though alveolar hyperventilation is not typical in mild COPD. Collectively, therefore, the consistently lower P_ETCO₂ in patients with DL_CO<LLN were more likely to represent the consequences of increased V_D/V_T. Given the well-known disparities between measured P_ETCO₂ and arterial PaCO₂ (even in mild COPD), direct measurements of arterial blood gases (unavailable in the current study) would be required to measure dead space and V̇_A (calculation using the modified Bohr equation) and rule out any contribution of alveolar hyperventilation to the low P_ETCO₂, by considering it in the context of the measured PaCO₂ [8]. This would allow for more definitive assessment of the contributions to reduced ventilatory efficiency in this population.

The relative importance of reduced ventilatory efficiency as a contributor to high IND and dyspnea in the DL_CO<LLN group is best appreciated in the early phase of exercise and at the standardized 60 W comparison (Figure 5). This point generally occurred before the anaerobic threshold (and the respiratory compensation point), where operating lung volumes, breathing pattern and arterial O₂ saturation were similar in the two COPD groups. As anticipated, at higher exercise intensities, mechanical events became important. The higher IND (by 18%) and V̇_E (by 11%) at 60 W in the DL_CO<LLN group forced earlier dynamic mechanical constraints: the V_T plateau (minimal IRV) and attendant tachypnea occurred at a lower WR, V̇_E, and V̇O₂ than the other two groups (Figure E2). Thus, the results bolster the contention that significant dynamic respiratory mechanical constraints and high peak IND (62%) and severe dyspnea (Borg 6) can occur in patients with apparent, adequate breathing reserve, as traditionally estimated from FEV₁ [9,41].

The association between activity-related dyspnea, low DL_CO, high physiological dead space and high ventilatory demand is long established in advanced COPD and is often attributed to emphysema [42]. Our current COPD participants, regardless of DL_CO grouping, had similar minimal degrees of centrilobular emphysema on CT, which supports previous reports of a poor correlation of resting DL_CO with emphysema scores in COPD [18]. In this context, reduced pulmonary blood perfusion is known to occur in non-emphysematous lung regions in mild COPD [15]. There is evidence that “vasculocentric” abnormalities (inflammation, endothelial dysfunction) can occur in mild COPD, independent of parenchymal destruction [4,5,43,44]. Notably, Barbera and colleagues [4] originally showed intimal thickening and luminal narrowing of the pulmonary arteries in lung resection specimens in mild to moderate COPD, leading to high V̇_A/Q̇ lung units and creating a skewed ventilation dispersion [5]. Thus, it is conceivable that impaired capillary perfusion played an important role in decreasing DL_CO and increasing V_D in this sample with only minor or trace apical emphysema.

Participants with DL_CO<LLN reported higher dyspnea intensity ratings at relatively low WRs compared with DL_CO≥LLN and control (Figure 4A). Furthermore, dyspnea/IND and dyspnea/V̇_E/MVV slopes were similar in all three groups (Figure 4B,C). It is clear that at 60 W, the DL_CO<LLN group had higher IND and ventilatory demand/capacity imbalance, which helps explain their consistently higher exertional dyspnea ratings compared with the other two groups (Figure 5). This association between reduced ventilatory efficiency/high demand and exertional dyspnea has been corroborated by a recent study by Phillips et al. [45] which showed that in mild COPD, selective pulmonary vasodilation by inhaled nitric oxide (iNO, compared with placebo) was associated with decreased V̇_E/V̇CO₂ and dyspnea intensity. Importantly, this occurred in the absence of any change in dynamic lung mechanics or O₂ saturation during iNO, highlighting the dyspneogenic role of reduced ventilatory efficiency in mild COPD. Taken in the context of current neurophysiological constructs of dyspnea, we posit that increased efferent electrical activation (increased IND) from both medullary (chemical) and cortical (mechanical) centers in the DL_CO<LLN group, is relayed to the somatosensory cortex (increased central corollary discharge), where unpleasant respiratory sensations are consciously perceived [46].

Limitations

Despite the best efforts to match controls on the basis of habitual physical activity in this retrospective analysis, the control group had greater than expected exercise capacity which may exaggerate group differences. However, the differences are unlikely to explain the key physiological differences at submaximal WR comparisons. Our study design did not allow us to explore multiple other potentially important afferent stimuli (e.g. from lungs, respiratory and locomotor muscles) that could potentially influence IND and the intensity and quality of dyspnea [45]. Further, our study design did not include measurement of arterial blood gas, which limits our ability to calculate dead space, or consider the effects of carotid chemoreceptor response to arterial PO₂. While our patients had no evidence of significant hypoxemia from pulse oximetry, or clinical diagnosis pulmonary hypertension, we did not determine if pulmonary artery pressures during exercise were higher in the DL_CO<LLN than the other two groups, which would have contributed to a higher V̇_E/V̇CO₂. Future studies that included measurements of arterial blood gas and pulmonary artery pressures could add further mechanistic insight into genesis of dyspnea in mild COPD.

Conclusions

The results support the contention that IND during exercise is relatively high in COPD patients with only mild airway obstruction and lower DL_CO. Moreover, we present new evidence that in those with mild COPD and low DL_CO, abnormalities of pulmonary gas exchange likely contribute to higher IND from the early stages of exercise. Finally, the higher ventilatory demand in this group likely accelerated development of significant mechanical constraints with further amplification of IND and neuro-mechanical dissociation, which conflated to provoke greater respiratory discomfort as exercise proceeded.

Implications

DL_CO measurement should be considered as part of a more comprehensive physiological evaluation of smokers with mild airway obstruction who present with disproportionate persistent dyspnea during physical activity. Combined abnormalities of DL_CO and exercise V̇_E/V̇CO₂ suggest clinically important impairment of pulmonary gas exchange. The results set the stage for a deeper interrogation of the nature of microvascular dysfunction in smokers, including innovative biological studies to identify novel therapeutic targets.

Author contributions

All authors played a role in the content and writing of the manuscript. DEO had the original idea for the study. MDJ performed the majority of the data analysis. All authors contributed to data interpretation. MDJ and DEO wrote the first draft of the manuscript. All authors approved the final version of the manuscript.

Disclosure statement

M.D. James was supported by a Queen Elizabeth II Graduate Scholarship in Science and Technology. Dr. D.B. Phillips was supported by a postdoctoral fellowship from the Natural Sciences and Engineering Research Council of Canada. Dr. A.F. Elbehairy acknowledges the support of the European Respiratory Society, Fellowship LTRF 2019. Dr. D.E. O’Donnell has received research funding via Queen’s University from Canadian Institutes of Health Research, Canadian Respiratory Research Network, AstraZeneca, and Boehringer Ingelheim. The funders had no role in the study design, data collection and analysis, or preparation of the manuscript. All other authors have no conflicts of interest to declare.