Resting Physiological Correlates of Reduced Exercise Capacity in Smokers with Mild Airway Obstruction

ABSTRACT

Smokers with minor spirometric abnormalities can experience persistent activity-related dyspnea and exercise intolerance. Additional resting tests can expose heterogeneous physiological abnormalities, but their relevance and association with clinical outcomes remain uncertain. Subjects included sixty-two smokers (≥20 pack-years), with cough and/or dyspnea and minor airway obstruction [forced expiratory volume in one-second (FEV₁) ≥80% predicted and >5th percentile lower limit of normal (LLN) (i.e., z-score >−1.64) using the 2012-Global Lung Function Initiative equations]. They underwent spirometry, plethysmography, oscillometry, single-breath nitrogen washout, and symptom-limited incremental cycle exercise tests. Thirty-two age-matched nonsmoking controls were also studied. Thirty-three (53%) of smokers had chronic obstructive pulmonary disease by LLN criteria. In smokers [n = 62; age 65 ± 11 years; smoking history 43 ± 19 pack-years; post-bronchodilator FEV₁ z-score −0.60 ± 0.72 and FEV₁/FVC z-score −1.56 ± 0.87 (mean ± SD)] versus controls, peak oxygen uptake (̇VO₂) was 21 ± 7 vs. 32 ± 9 ml/kg/min, and dyspnea/̇VO₂ slopes were elevated (both p < 0.0001). Smokers had evidence of peripheral airway dysfunction and maldistribution of ventilation when compared to controls. In smokers versus controls: lung diffusing capacity for carbon monoxide (D_LCO) was 85 ± 22 vs. 105 ± 17% predicted, and residual volume (RV)/total lung capacity (TLC) was 36 ± 8 vs. 31 ± 6% (both p < 0.01). The strongest correlates of peak ̇VO₂ were D_LCO% predicted (r = 0.487, p < 0.0005) and RV/TLC% (r = −0.389, p = 0.002). D_LCO% predicted was also the strongest correlate of dyspnea/̇VO₂ slope (r = −0.352, p = 0.005). In smokers with mild airway obstruction, associations between resting tests of mechanics and pulmonary gas exchange and exercise performance parameters were weak, albeit consistent. Among these, DLCO showed the strongest association with important outcomes such as dyspnea and exercise intolerance measured during standardized incremental exercise tests.

KEYWORDS:

• Diffusing capacity
• dyspnea
• exercise
• gas trapping
• smokers

Chronic obstructive pulmonary disease (COPD) is a common inflammatory disease of the airways, alveoli, and microvasculature that is underdiagnosed in smokers at risk for the disease. The majority of COPD patients have mild airway obstruction (1,2). Symptomatic smokers with mild COPD are at higher risk for earlier mortality and poorer perceived quality of life than nonsmokers (3,4). Moreover, dyspnea and activity restriction are common symptoms among smokers with minor spirometric abnormalities (5,6). In fact, recent large epidemiological studies confirmed that a sizable proportion of smokers with normal spirometry had reduced exercise capacity and increased requirement for inhaled medications (7,8).

Tobacco-related inflammatory injury of the lungs manifests as heterogeneous physiological impairment with highly variable clinical expression (9–12). Thus, simple spirometry provides only a crude assessment of disease pathophysiology, especially in the early stages of the disease. Moreover, spirometric assessments using different criteria for airflow limitation [Global Initiative for Chronic Obstructive Lung Disease (GOLD) fixed ratio versus lower limit of normal (LLN) of forced expiratory volume in one-second (FEV₁)/forced vital capacity (FVC)] can produce discrepant results in smokers being screened for COPD, leading in some instances to diagnostic confusion (13–15). Accordingly, the question arises whether additional pulmonary function tests are useful for the clinical evaluation of smokers with troublesome respiratory symptoms despite unremarkable spirometry.

Recent physiological studies conducted during exercise in symptomatic smokers with or without mild COPD have highlighted that exercise limitation is common and is multifactorial (9–12,16). Nevertheless, it has now become clear that respiratory factors such as increased dynamic mechanical constraints and reduced ventilatory efficiency are consistent contributory factors (9–12,16). However, since detailed exercise testing is not a realistic option for many clinicians, the current exploratory physiological study was designed to examine the relationship between a number of resting physiological tests that are widely available and exercise performance parameters such as dyspnea intensity and peak oxygen update (̇VO₂). We selected smokers with chronic respiratory symptoms but with mild airway obstruction, as arbitrarily defined by a post-bronchodilator FEV₁ greater than 80% predicted and the 5th percentile LLN (i.e., >−1.64 z-score) (15), regardless of whether they fitted COPD diagnostic criteria based on fixed FEV₁/FVC ratio. However, we planned a post-hoc comparison of resting and exercise tests (and their inter-relationships) in subgroups of healthy controls and smokers with and without COPD by FEV₁/FVC criteria.

The overarching aim of the study was to examine the potential clinical relevance of a panel of resting tests that interrogate respiratory mechanics and pulmonary gas exchange. These include spirometry, plethysmographic lung volumes, tests to measure respiratory impedance [e.g., impulse oscillometry (IOS)] and distribution of ventilation [e.g., single-breath nitrogen washout (SBNW)] (17,18); and lung diffusing capacity for carbon monoxide (D_LCO).

The specific objective of our study was, therefore, to answer the novel question: which of the available candidate resting physiological markers are most closely linked to increased activity-related dyspnea and reduced exercise tolerance in smokers with respiratory symptoms being screened for COPD?

Materials and methods

Subjects and study design

This study analyzed data compiled from 62 symptomatic smokers (n = 21 current, n = 41 former) who were screened for COPD for the Canadian Respiratory Research Network database at the Respiratory Investigation Unit (RIU), Queen's University and Kingston General Hospital between 2011 and 2014. Selection criteria included: age ≥40 years; smoking history ≥20 pack-years; body mass index (BMI) 18–35 kg/m²; and a preserved FEV₁ (≥80% predicted and greater than the 5th percentile LLN (z-score >−1.64) using Global Lung Function Initiative (GLI) 2012 reference equations) (15). Symptoms included cough/sputum and/or activity-related dyspnea [modified Medical Research Council (MRC) dyspnea scale ≥2 or COPD Assessment Test (CAT) ≥10 or Baseline Dyspnea Index ≤9] (19,20). Thirty-two age-matched nonsmoking healthy subjects from the RIU data registry were also included if they had an FEV₁ ≥80% predicted and FEV₁/FVC >0.7 and both with z-scores >−1.64 using the GLI-2012 equations (15). Subjects with bronchial asthma and/or other medical conditions that could contribute to dyspnea/exercise limitation or contraindicate exercise testing were excluded. Some subjects were included in previously published studies (10,11,16); however, the current data set is not the same, and there is no overlap with any of these previous analyses.

This study received ethical approval from the Queen's University and Affiliated Teaching Hospitals Research Ethics Board (DMED-1674-14). After signing written informed consents, participants completed two visits to the RIU. Visit 1 included screening for eligibility, medical history, symptom, and activity assessment questionnaires (19–22), Charlson Comorbidity Index (23), pre- and post-bronchodilator (400 μg salbutamol) pulmonary function tests, small airway function tests, and an incremental cardiopulmonary exercise test (CPET) for familiarization. Visit 2 included spirometry followed by an incremental CPET. Subjects refrained from smoking and avoided caffeine, heavy meals, alcohol, and vigorous exercise ≥4 hours before visits. Subjects were not tested if they had respiratory tract infections within 4 weeks. A high-resolution computed tomography (HRCT) scan of the thorax was available for smokers with COPD, and was used for the quantitative assessment of the extent of emphysema (24).

Procedures

Pulmonary function and small airway function tests (SBNW and IOS) (17,18) were performed using automated equipment (Vmax229d and Autobox V62J; and MasterScreen IOS; SensorMedics, Yorba Linda, CA); measurements were expressed relative to predicted normal values (15,25–27). We used predictive equations for D_LCO of Burrows et al. (26); these equations worked well in a similar population of 128 healthy men and women previously tested in our laboratory (28). CPETs were conducted on an electronically braked cycle ergometer (Ergoline 800 s; SensorMedics) using a SensorMedics Vmax229d system according to guidelines, as previously described (9). CPETs included steady-state rest followed by stepwise increments (20-watt/2-min) until symptom limitation. Exercise measurements included: breath-by-breath cardiorespiratory and breathing pattern parameters; oxygen saturation (pulse oximetry); heart rate (HR, 12-leads ECG); dynamic lung volumes calculated from inspiratory capacity (IC) with rest-to-peak decrease in IC, reflecting the magnitude of dynamic hyperinflation (DH) (29). Dyspnea and leg discomfort intensity were measured by the modified 10-point Borg scale (30). Exercise measurements were expressed as % predicted normal values when applicable (31–33). The anaerobic threshold (AT) was assessed using the V-slope method, and we also employed the ventilatory equivalent method (34). Furthermore, two different investigators independently verified the identification of the AT.

Statistical analysis

Primary end points included peak ̇VO₂, the slope of dyspnea intensity over ̇VO₂ and rest-to-peak change (∆) in IC. A sample size of 30 per group should provide sufficient power (>80%) to detect a significant difference in each of these primary end points based on the assumptions: α = 0.05, β = 0.80, a two-tailed test of significance, a 20% between-group difference in peak ̇VO₂ and dyspnea/̇VO₂ slopes or a 0.2 L difference in ∆IC, and mean and SD values obtained in similar populations tested at our laboratory (9).

Unpaired t-tests were used to compare between-group differences in pulmonary function tests and dyspnea intensity, cardiorespiratory parameters at rest, iso-work rates, and peak exercise. Results are reported as means±SD unless otherwise specified. Statistical significance was set at p < 0.05.

Within the smokers, associations between each of the primary end points and relevant independent variables were tested using Pearson's correlations. Independent resting variables included: post-bronchodilator FEV₁/FVC and FEV₁, IC, functional residual capacity (FRC), residual volume (RV), total lung capacity (TLC), RV/TLC, D_LCO, resistance at 5 Hz (R5), change in airway resistance from 5 to 20 Hz (R5-20), distal capacitive reactance (X5), and phase-III N₂ slope. Spirometry and lung volume measurements were expressed as % predicted, and ratios were expressed as % in this analysis. Based on descriptive statistics such as means, Pearson's correlation, and literature review, variables were entered as an important predictor in a linear regression model for each of the primary outcome. The final regression model was based on the diagnostic test such as F-test, adjusted r-square, collinearity test, and normal probability plot. Relationships between important outcomes were tested across smoker subgroups (former versus current smokers, COPD versus non-COPD) using a regression model incorporating an interaction term (group × independent variable). p-values less than 0.05 were used for statistical significance, and all the analyses were conducted using statistical software SAS version 9.3.

Results

The smoker and control groups had similar age, height, and BMI (Table 1). Thirty-three (53%) smokers had COPD based on GLI classification (post-bronchodilator FEV₁/FVC z-score <−1.64): those with COPD had similar resting (FRC, IC) and dynamic lung volumes, D_LCO, IOS, peak ̇VO₂ and dyspnea/̇VO₂ slopes compared with those without COPD. Both smoker groups (with and without COPD by LLN criteria) were different from controls with regard to respiratory symptoms and resting lung functions (online supplement). Forty-four (71%) smokers met GOLD grade 1 criteria, and eighteen (29%) met the formerly named GOLD 0 (at risk): both groups had similar resting (FRC, IC) and dynamic lung volumes, IOS, and peak ̇VO₂. Smokers had a significant smoking history (43 ± 19 pack-years), and controls were never-smokers except for five subjects who had an insignificant smoking history and had stopped smoking >20 years before the study. Smokers had worse activity-related dyspnea, higher CAT scores, lower physical activity (CHAMPS), poorer health status (SGRQ), and higher Charlson Index compared with controls (Table 1). 36 of 44 smokers with COPD had chest HRCT scans, and the results revealed 15 ± 9% (range: 2–41%) of their lungs as low-attenuation areas (LAA, less than −950 Hounsfield Units). Comorbidities in the smoker and control groups, respectively, included stable: hypertension (43.5 and 9.3%), gastroesophageal reflux disease (35.5 and 9.3%), hypercholesterolemia (30.6 and 12.5%), osteoarthritis (21.0 and 21.9%), hypothyroidism (14.5 and 9.3%), diabetes mellitus (9.7 and 6.3%), stable angina (8.1 and 0%), and previous myocardial infarction (1.6 and 0%). Inhaled and noninhaled medications are shown in Table 1.

Table 1. Subject characteristics and medications.

	Smokers	Healthy controls	p
	(n = 62)	(n = 32)	values
Male: Female, n	38:24	17:15	0.510
Age, years	65 ± 11	64 ± 9	0.596
Height, cm	168 ± 9	168 ± 9	0.987
Body mass, kg	78 ± 14	75 ± 12	0.249
BMI, kg/m^2	28 ± 5	27 ± 3	0.129
Smoking history, pack-years	42.5 ± 18.9	0.6 ± 1.9	<0.0001
Smoking status, n:
• Current	21	0	—
• Never	0	27	—
• Ex–smoker	42	5	—
BDI focal score (0–12)	8.7 ± 1.6	11.5 ± 0.8	<0.0001
Modified MRC dyspnea scale (0–4)	1.13 ± 0.8	0.06 ± 0.2	<0.0001
CAT score (0–40)	12.6 ± 7.3	3.2 ± 3.3	<0.0001
Oxygen cost diagram (0–100), mm	71 ± 15	91 ± 10	<0.0001
CHAMPS, kcal/week for all activities	3903 ± 3618	6529 ± 4469	<0.0001
SGRQ total score	25.5 ± 16.2	3.7 ± 3.4	<0.0001
Charlson Comorbidity Index	3.4 ± 1.7	2.1 ± 1.1	<0.0001
Noninhaled medications, n (%)
• β-Blocker	8(13)	0(0)	—
• Diuretic	6(10)	1(3)	—
• ACE inhibitor	12(19)	2(6)	—
• Ca channel blocker	7(11)	0(0)	—
• Statin	18(29)	4(13)	—
• Anti–diabetic medications	5(8)	1(3)	—
• Proton pump inhibitor	16(26)	2(6)	—
• Acetyl salicylic acid	6(10)	4(13)	—
• L–thyroxin	9(15)	3(9)	—
Inhaled medications, n (%)
• SABA	19(31)	0(0)	—
• SAMA	3(5)	0(0)	—
• LABA	3(5)	0(0)	—
• LAMA	15(24)	0(0)	—
• ICS	1(2)	0(0)	—
• ICS–LABA	18(29)	0(0)	—

Values are means ± SD.

Abbreviations: BDI = Baseline Dyspnea Index; BMI = body mass index; CAT = COPD Assessment Test; CHAMPS = Community Healthy Activities Model Program for Seniors questionnaire; COPD = chronic obstructive pulmonary disease; MRC = Medical Research Council; SGRQ = St. George's Respiratory Questionnaire. ACE = angiotensin converting enzyme; Ca = calcium; ICS = inhaled corticosteroid; LABA = long-acting β₂-agonist; LAMA = long-acting antimuscarinic; SABA = short-acting β₂-agonist; SAMA = short-acting antimuscarinic; oxygen cost diagram is a visual analog scale (100 mm in length) and that lists activities of daily living according to their levels of oxygen consumption. It is used for quantifying a patient's evaluation of his tolerance to exercise, as a component of health-related quality of life. The BDI score ranges from 0 (very severe impairment) to 4 (no impairment) for each of the following 3 domains (functional impairment, magnitude of task and magnitude of effort); they are summed to determine the BDI focal score (0–12).

Pulmonary function and small airway function

Table 2^{Figure 1} summarizes pulmonary function tests. Smokers [post-bronchodilator FEV₁/FVC z-score −1.56 ± 0.87 (65 ± 8%); FEV₁ z-score −0.60 ± 0.72 (90 ± 11% predicted)] had evidence of pulmonary gas trapping (higher RV/TLC), peripheral airway dysfunction (lower maximal mid-expiratory flow, higher R5-20 and resonant frequency), and maldistribution of ventilation (higher phase-III N₂ slope) compared with controls (Figure 1). Average D_LCO was within normal limits in smokers, but significantly reduced compared with controls (85 ± 22 versus 105 ± 17% predicted). In addition, 40% of smokers had D_LCO values <80% predicted compared with only 3% of controls. V_A/TLC ratio was normal (>0.8) in both groups. A weak correlation was found between D_LCO% predicted and degree of emphysema (LAA), (r = −0.386, p = 0.02). In support of successful asthma exclusion, none of the smokers demonstrated significant bronchodilator reversibility.

Table 2. Resting pulmonary function.

	Smokers	Healthy controls	p values
Post-bronchodilator
FEV1, L	2.54 ± 0.6(90 ± 11)	3.07 ± 0.6(108 ± 12)	0.000(<0.0001)
FEV1 z-score	−0.60 ± 0.72	0.55 ± 0.78	<0.0001
FEV1/FVC, %	65 ± 8	76 ± 5	<0.0001
FEV1/FVC z-score	−1.56 ± 0.87	−0.24 ± 0.68	<0.0001
Pre-bronchodilator
FEV1, L	2.41 ± 0.67(85 ± 12)	2.95 ± 0.67(105 ± 13)	0.000(<0.0001)
FEV1/FVC, %	63 ± 9(80 ± 10)	73 ± 6(94 ± 7)	<0.0001(<0.0001)
FEF25–75%, L/s	1.25 ± 0.78(50 ± 22)	2.23 ± 0.90(94 ± 32)	<0.0001(<0.0001)
SVC, L	3.94 ± 0.8(109 ± 14)	4.23 ± 0.9(118 ± 12)	0.138(0.005)
IC, L	2.83 ± 0.7(104 ± 18)	2.97 ± 0.8(111 ± 18)	0.367(0.098)
FRC, L	3.27 ± 0.8(103 ± 23)	3.11 ± 0.8(98 ± 23)	0.348(0.367)
TLC, L	6.12 ± 1.1(103 ± 13)	6.08 ± 1.1(104 ± 13)	0.876(0.816)
RV, L	2.17 ± 0.6(100 ± 24)	1.86 ± 0.5(86 ± 21)	0.012(0.008)
RV/TLC, %	36 ± 8	31 ± 6	0.003
DLCO, ml/min/mmHg	17.5 ± 5.1(85 ± 22)	21.6 ± 4.9(105 ± 17)	0.000(<0.0001)
DL/VA, ml/min/mmHg/L	3.3 ± 0.7(90 ± 18)	4.1 ± 0.6(110 ± 15)	<0.0001(<0.0001)
VA/TLC	0.85 ± 0.09	0.87 ± 0.06	0.357
sRaw, cmH2O•s	9.2 ± 3.9(217 ± 93)	5.6 ± 1.9(136 ± 50)	<0.0001(<0.0001)
Closing volume, L	0.51 ± 0.3	0.43 ± 0.2	0.151
N2 slope, %/L	4.9 ± 2.1	3.2 ± 1.6	0.000
R5, cmH2O/L/s	4.9 ± 1.6	4.2 ± 1.6	0.051
R5-20, % R5	22.8 ± 10.9	12.9 ± 6.9	<0.0001
AX, cmH2O/L	8.9 ± 8.9	3.4 ± 3.0	0.001
X5, cmH2O/L/s	−1.6 ± 0.9	−1.1 ± 0.5	0.006
Fres, Hz	15.8 ± 5.3	11.9 ± 3.7	0.000

Values are means ± SD, with values in parentheses representing % of predicted normal values.

Abbreviations: AX = reactance area; D_LCO = diffusing capacity of the lung for carbon monoxide; D_L/V_A = D_LCO relative to alveolar volume; FEF_25–75% = forced expiratory flow between 25 and 75% of forced vital capacity; FEV₁ = forced expired volume in 1 second; FVC = forced vital capacity; FEV₁/FVC = ratio between FEV₁ and FVC; FRC = functional residual capacity; Fres = resonant frequency; IC = inspiratory capacity; N₂ = nitrogen; R5 = resistance at 5 Hz during impulse oscillometry; R5-20 = differential change in resistance between 5 and 20 Hz; RV = residual volume; sRaw = specific airway resistance; SVC = slow vital capacity; TLC = total lung capacity; X5 = distal capacitive reactance at 5 Hz.

Figure 1. Measures of small airway dysfunction, pulmonary gas trapping, and lung diffusing capacity in smokers and age-matched healthy controls. (A) R5-20, (B) X5, (C) phase-III N₂ slope, (D) FEF_25–75% (% predicted), (E) RV/TLC (%), and (F) D_LCO (% predicted). Values in box plots represent the median, 10th, 25th, 75th, and 90th percentile. *p < 0.05 smokers vs. healthy controls. Abbreviations: D_LCO = diffusing capacity of the lung for carbon monoxide; FEF_25–75% = maximal mid-expiratory flow; N₂ = nitrogen; R5-20 = change in airway resistance from 5 to 20 Hz during impulse oscillometry (IOS); RV = residual volume; TLC = total lung capacity; X5 = distal capacitive reactance during IOS.

Exercise responses

Table 3 summarizes peak exercise measurements. Peak work rate, ̇VO₂, ventilation (̇V_E), and HR were lower in smokers than in controls (all p < 0.0001). HR responses were higher throughout exercise in smokers compared with controls. None of our subjects showed any ischemic electrocardiographic changes during exercise. Smokers had significantly higher ̇V_E, ventilatory equivalent for carbon dioxide (̇V_E/̇VCO₂), and lower end-tidal CO₂ at submaximal exercise compared with controls (Figure 2). Resting ̇V_E/̇VCO₂ was higher in smokers than controls (47 ± 8 versus 41 ± 5, p < 0.001). The ̇V_E to ̇VCO₂ relationships expressed as nadir, slope, and intercept were 31 ± 4 versus 27 ± 3 (p < 0.0001), 28 ± 4 versus 27 ± 3 (p = 0.055), and 4.6 ± 1.5 versus 3.3 ± 1.8 (p = 0.000) in smokers versus controls, respectively. Within smokers, resting ̇V_E/̇VCO₂ correlated well with ̇V_E/̇VCO₂ nadir (r = 0.611, p < 0005). In turn, ̇V_E/̇VCO₂ nadir correlated with D_LCO% predicted (r = −0.353, p = 0.005). Smokers had lower IC and inspiratory reserve volume (IRV) during submaximal exercise than controls (Figure 2). Rest-to-peak ΔIC was −0.19 ± 0.47 and 0.09 ± 0.52 L (p = 0.006) in smokers versus controls, respectively. Smokers (with and without COPD) had higher dyspnea intensity and leg discomfort during submaximal exercise than controls (Figure 2, E1). Dyspnea/̇V_E, dyspnea/̇VO₂ and leg discomfort/̇VO₂ slopes were higher in smokers versus controls (0.12 ± 0.06 versus 0.07 ± 0.04 units/L/min, 0.47 ± 0.38 versus 0.22 ± 0.14 and 0.53 ± 0.32 versus 0.27 ± 0.14 units/ml/kg/min, respectively, all p < 0.0001); differences were also present when ̇VO₂ was expressed as % predicted [0.13 ± 0.10 versus 0.07 ± 0.05 (p = 0.001) and 0.16 ± 0.14 versus 0.08 ± 0.05 (p = 0.003), Borg scale units/% predicted]. AT occurred at a lower ̇VO₂ in smokers versus controls (1.30 ± 0.40 versus 1.63 ± 0.49 L/min, p = 0.001).

Table 3. Measurements at the peak of symptom-limited incremental cycle exercise.

	Smokers	Healthy controls	p values
Work rate, watts	102 ± 41	156 ± 52	<0.0001
(% predicted)	(69 ± 23)	(107 ± 25)	<0.0001
̇VO2, L/min	1.68 ± 0.60	2.41 ± 0.78	<0.0001
(% predicted)	(79 ± 20)	(116 ± 25)	<0.0001
̇VO2, ml/kg/min	21.5 ± 6.7	31.8 ± 9.1	<0.0001
̇VE, L/min	59.7 ± 21.5	90.7 ± 29.4	<0.0001
(% predicted MVC)	(71 ± 23)	(88 ± 24)	0.001
VT, L	1.73 ± 0.54	2.21 ± 0.58	0.000
Fb, breaths/min	34 ± 6	39.9 ± 6.5	0.000
TI/TTOT	0.46 ± 0.04	0.48 ± 0.02	0.003
IC, L	2.63 ± 0.66	2.98 ± 0.70	0.022
(% predicted)	(96 ± 20)	(112 ± 18)	0.000
Δ IC from rest, L	−0.19 ± 0.47	0.09 ± 0.52	0.006
IRV, L	0.90 ± 0.77	0.77 ± 0.37	0.165
EILV, % TLC	85 ± 8	87 ± 6	0.145
PETCO2, mmHg	33.9 ± 4.3	32.8 ± 4.6	0.269
̇VE/̇VCO2	33.9 ± 4.8	33.9 ± 5.4	0.864
SpO2, %	95.6 ± 1.8	96.0 ± 2.0	0.286
R	1.05 ± 0.1	1.12 ± 0.1	0.003
HR, beats/min	133 ± 24	152 ± 16	0.000
(% predicted)	(79 ± 14)	(91 ± 9)	<0.0001
O2 pulse, mL/beat	12.7 ± 3.8	15.9 ± 4.9	0.001
Dyspnea, Borg units	5.3 ± 2.4	5.1 ± 2.9	0.676
Leg discomfort, Borg units	6.2 ± 2.4	5.8 ± 2.9	0.514

Values are means ± SD.

Abbreviations: EILV = end-inspiratory lung volume; Fb = breathing frequency; HR = heart rate; IC = inspiratory capacity; IRV = inspiratory reserve volume; MVC = maximal ventilatory capacity; O₂ pulse = oxygen pulse; P_ETCO₂ = partial pressure of end-tidal carbon dioxide; R = respiratory exchange ratio; SpO₂ = arterial oxygen saturation measured by pulse oximetry; T_I/T_TOT = inspiratory duty cycle; TLC = total lung capacity; ̇V_E = minute ventilation; ̇V_E/̇VCO₂ = ventilatory equivalent for carbon dioxide; ̇VO₂ = oxygen uptake; V_T = tidal volume.

Figure 2. Selected parameters of responses to incremental cycle exercise plotted against work rate in smokers and age-matched healthy controls. (A) Exertional dyspnea intensity using Borg scale, (B) ̇V_E, (C) ̇V_E/̇VCO₂, (D) PetCO₂, (E) IC, (F) IRV. Values are mean ± SEM. Square symbols represent the points at the highest equivalent work arte of 60 watts. *p < 0.05 smokers vs. healthy controls at rest, standardized work rates or at peak exercise. Abbreviations: IC = inspiratory capacity; IRV = inspiratory reserve volume; PetCO₂ = partial pressure of end-tidal carbon dioxide; TLC = total lung capacity; ̇V_E = minute ventilation; ̇V_E/̇VCO₂ = ventilatory equivalent for carbon dioxide.

Physiological correlations

Resting physiological correlates of each of the primary end points within smokers are summarized in Table 4^{Figure 3}. D_LCO% predicted and RV/TLC% were the strongest correlates of peak ̇VO₂ (ml/kg/min); their combination explained 35% of the variance in peak ̇VO₂. In the secondary analysis conducted across smoker subgroups, significant relationships were found to be similar in former versus current smokers as well as in those with COPD versus without COPD. Although the cutoffs are arbitrary, it is interesting that 80% of smokers with a combination of high RV/TLC (>35%) (35) and low D_LCO (<80% predicted) had reduced peak ̇VO₂ (≤20 ml/kg/min) (36).

Table 4. Correlates of each of the primary outcomes within smokers.

Independent variables	DLCO % predicted	RV/TLC %	X5 cmH2O/L/s	R5-20 % R5	FEV1/FVC %	N2 slope %/L	IC % predicted
Dependent variable: peak ̇VO2 ml/kg/min
r	0.487	−0.389	0.377	−0.341	0.231	−0.182	−0.136
p	<0.0005	0.002	0.003	0.007	0.071	0.161	0.290
^†Model summary: DLCO (% predicted) and RV/TLC (%), R = 0.591, r^2 = 0.350, adjusted r = 0.328, f = 15.85, p < 0.001
Dependent variable: dyspnea/̇VO2 slope
r	−0.352	0.225	−0.290	0.161	−0.210	.330	0.0838
p	0.005	0.079	0.022	0.211	0.102	0.009	0.517
^†Model summary: DLCO (% predicted) and N2 slope, r = 0.447, r^2 = 0.200, adjusted r = 0.172, f = 7.25, p = 0.002
Dependent variable: rest-to-peak change in IC
r	0.298	−0.241	0.515	−0.520	0.352	−0.149	−0.350
p	0.0193	0.0595	<0.0005	<0.0005	0.005	0.253	0.005
^†Model summary: IC (% predicted), R5-20 and X5, r = 0.622, r^2 = 0.388, adjusted r = 0.356, f = 12.23, p < 0.001

Correlation analyses between each of the primary end points and relevant independent variables using Pearson's correlations.

^† Stepwise forward regression analysis showing the best combination of independent variables for each of the primary endpoints.

Abbreviations: D_LCO = diffusing capacity of the lung for carbon monoxide; FEV₁ = forced expired volume in 1 s; FVC = forced vital capacity; IC = inspiratory capacity; N₂ = nitrogen; R5-20 = differential change in resistance between 5 and 20 Hz; RV = residual volume; TLC = total lung capacity; X5 = distal capacitive reactance at 5 Hz; ̇VO₂ = oxygen uptake.

Peak ̇VO₂ (ml/kg/min) also correlated with exercise slopes of the intensity of dyspnea/̇VO₂ (r = −0.381, p = 0.002) and leg discomfort/̇VO₂ (r = −0.452, p < 0.0005). In turn, dyspnea/̇VO₂ correlated best with D_LCO% predicted (r = −0.352, p = 0.005) (Table 4). Slopes of leg discomfort/̇VO₂ were not associated with any resting respiratory measurements but did correlate inversely with ̇VO₂ at the AT (r = −0.324, p = 0.012). Figure 3 summarizes inter-relationships between D_LCO% predicted, peak ̇VO₂, dyspnea/̇VO₂ slope and ̇V_E/̇VCO₂ nadir within the smoker group.

Figure 3. Inter-relationships between D_LCO% predicted, peak ̇VO₂, dyspnea/̇VO₂ slope, and ̇V_E/̇VCO₂ nadir within the smoker group. D_LCO correlated with peak ̇VO₂ (r = 0.487, p < 0.0005), dyspnea/̇VO₂ slope (r = −0.352, p = 0.005), and ̇V_E/̇VCO₂ nadir (r = −0.353, p = 0.005). Abbreviations: D_LCO = diffusing capacity of the lung for carbon monoxide; ̇V_E/̇VCO₂ = ventilatory equivalent for carbon dioxide; ̇VO₂ = oxygen uptake.

The best independent correlates of rest-to-peak ΔIC were, in order of significance: IOS measurements (R5-20, X5), FEV₁/FVC, and resting IC% predicted (Table 4). The combination of R5-20, X5, and IC% predicted explained 39% of the variance in DH.

Discussion

The main findings of this study were that symptomatic smokers with a preserved FEV₁ had heterogeneous respiratory impairment as exposed by a number of sensitive tests. Among these, decreased D_LCO and increased RV/TLC provided the most consistent associations with reduced exercise tolerance. A low D_LCO was also the best correlate of increased exertional dyspnea. Beyond the FEV₁, these additional resting tests helped expose the nature and extent of the respiratory impairment in individual smokers and provided insights into the source of their symptoms.

Our group of smokers who had persistent respiratory symptoms (cough, sputum, and/or activity-related dyspnea) was recruited as a result of targeted screening for COPD. As anticipated (37), different diagnostic spirometric criteria for COPD yielded discrepant results: 44 of 62 met GOLD grade 1 criteria, but 33 of 62 had COPD using the more stringent GLI criteria, which accounts for effects of aging. Smokers with/without COPD were similar in baseline lung function and peak exercise measures, but both the groups were significantly different from controls (online supplement). An a priori selection of smokers with FEV₁ ≥80% predicted and > the 5th percentile LLN was made to increase the likelihood of recruiting smokers without significant airway obstruction. Among such individuals, those with persistent symptoms have been shown in several population studies to have increased morbidity and mortality compared with nonsmokers (38,39). Indeed, the smoker group of the current study had poorer perceived health status, lower self-reported daily physical activity, and greater chronic activity-related dyspnea (measured by various questionnaires) compared with controls (Table 1). This was corroborated by the reduced peak ̇VO₂ (by over 30% on average) and higher dyspnea intensity ratings in smokers. In accordance with previous studies (9–12,16), reduced exercise tolerance in smokers was multifactorial and included severe dyspnea and leg effort. Our results confirmed that contributing respiratory factors included reduced ventilatory efficiency, higher ̇V_E, lower IC, and critical dynamic respiratory mechanical constraints and associated respiratory discomfort at relatively lower exercise intensities compared with controls (Figure 2). Notably, in smokers, IRV decreased toward its minimal value at a peak work rate that was 38% lower than that of healthy controls.

There has recently been renewed interest in IOS and N₂ washout tests and their potential utility in early detection of respiratory impairment in smokers. In the current study, these tests, on average, showed greater respiratory impedance and maldistribution of ventilation in smokers than in healthy controls (Figure 1). Interestingly, R5-20 correlated with the extent of DH during exercise, both measurements indicating an exaggerated frequency dependence of dynamic resistance in smokers. Abnormal configuration of the mid-volume maximal expiratory flow-volume loop and increased pulmonary gas trapping were more likely in smokers compared with healthy controls (Figure 1): this pointed to the presence of extensive small airway dysfunction despite a mean FEV₁ of 90% predicted and plethysmographic lung volumes and resting IC within the normal range.

Lung microvascular inflammation, disruption of the alveolar-capillary interface and reduced pulmonary perfusion are increasingly identified in smokers with minor spirometric abnormalities although precise quantification of these derangements is challenging (40,41). In this context, a recent imaging study has provided evidence of reduced pulmonary blood flow (by 30%) even in mild COPD patients (41). We have shown that reduced ventilatory efficiency during exercise (high ̇V_E/̇VCO₂ nadir) is common in mild COPD and that this mainly reflects high physiological dead space (rather than alveolar hyperventilation or relatively reduced tidal volume) and thus a preponderance of lung units with high ventilation–perfusion ratios (11).

The determinants of D_LCO include: ventilation distribution, ventilation/perfusion matching, resistance at the alveolar-capillary membrane, and combination rates with hemoglobin. It is noteworthy that in the current study, D_LCO and KCO (D_LCO/V_A), while within the low normal range by the American Thoracic Society criteria, were consistently lower in smokers than in controls by an average of 20%, with no difference in V_A/TLC (a crude measure of ventilatory inhomogeneity) (42). D_LCO provides a window into the microvasculature, and low values have previously been linked with activity-related dyspnea (Borg scale and Baseline Dyspnea Index) and low peak ̇VO₂ in moderate-to-severe COPD (29,43,44). A plausible mechanism is increased chemostimulation of ventilation secondary to effects of high physiological dead space as a result of presumed emphysema and reduced surface area for gas exchange

Modest but consistent reduction in D_LCO has been previously reported in smokers with or without mild COPD in the absence of structural changes of emphysema on computed tomography (CT) (37,45,46). In one study by Kirby et al. (40), low D_LCO correlated with reduced 6-minute walking distance in ex-smokers with normal chest CT scans and spirometry. The observation in the current study that D_LCO was associated with high ̇V_E/̇VCO₂ nadir, low peak ̇VO₂, and increased exertional dyspnea ratings within smokers (regardless of the presence of COPD by GLI criteria) supports the notion that similar physiological mechanisms of dyspnea may exist as have been postulated in more advanced COPD (Figure 3) (29,43,44). It is noteworthy that smokers who met GOLD grade 1 criteria had minor upper lobe structural changes consistent with centrilobular emphysema (15% LAA <−950 HU). In this group, we found a weak correlation between % LAA and D_LCO consistent with the results of other studies in more advanced COPD (47–49). Thus, in the smoker group, the lower D_LCO likely reflects the presence of greater ventilation/perfusion inequalities, reduced efficiency of CO₂ elimination, and consequent higher ventilatory stimulation. We postulate, based on the results of previous physiological studies in mild COPD that (9–12), that this higher demand in the presence of increased peripheral airway dysfunction and expiratory flow limitation forces earlier dynamic mechanical constraints at lower exercise intensities than in nonsmoker controls. In some individuals, ventilatory demand was likely further amplified by lower ventilatory thresholds reflecting earlier metabolic acidosis due to deconditioning (Table 1). This basic physiological construct is in line with recent exercise studies in smokers at risk for COPD or in those with mild COPD where increased exertional dyspnea (compared with healthy controls) at standardized ventilation was explained by the relatively higher fractional inspiratory neural drive to the diaphragm (12,16).

This is the first study to examine the relative strength of the association of a wide range of resting pulmonary function tests to peak symptom-limited ̇VO₂ and exertional dyspnea intensity in smokers with a preserved FEV₁. We found that although exercise limitation was undoubtedly multifactorial and included both respiratory and nonrespiratory factors, the combination of reduced D_LCO and respiratory mechanical factors (increased RV/TLC) explained 35% of the variance in peak ̇VO₂. This relation was not affected by differences in FEV₁/FVC within the smokers and was similar in active and former smokers. The practical implication is that mechanical factors such as increased RV are amenable to inhaled bronchodilators and have the potential to improve exertional dyspnea, at least based on the results of a previous study in patients with GOLD grade 1 COPD (50). By contrast, a reduced D_LCO is currently not responsive to pharmacological treatment but improves with smoking cessation (51).

Clearly, nonrespiratory factors also contributed to the lower peak symptom-limited ̇VO₂ in the smoker group who reported higher intensity of perceived leg discomfort and had lower AT compared with controls. This suggests a higher prevalence of skeletal muscle deconditioning in smokers, which is potentially amenable to multimodality exercise training.

Limitations

Our study sample was confined to symptomatic Canadian Caucasian smokers at risk for COPD who were targeted for screening according to recent guidelines. Our results may not therefore be generalizable to asymptomatic smokers. However, we and others have observed that so-called “asymptomatic” smokers (by modified MRC and CAT criteria) often have greater dyspnea and exercise intolerance than healthy controls when challenged with formal exercise testing (52). D_LCO was not corrected for between-subject differences in circulating hemoglobin, and although smokers were instructed to refrain from smoking prior to the test, baseline carboxyhemoglobin was not collected. Factors such as reduced physical activity and increased leg discomfort contributed to the lower peak ̇VO₂ in smokers; however, this did not detract from our analysis of the associations between the panel of tests of respiratory impairment and activity-related dyspnea.

Conclusions/implications

This study adds to the evidence that cardiopulmonary exercise testing is useful in uncovering significant and often unsuspected physiological derangements in smokers with preserved FEV₁. The overarching message is that the presence of persistent respiratory symptoms in smokers—regardless of whether they meet COPD LLN criteria—should not be ignored. Such individuals who were formerly designated “smokers at risk” (or GOLD 0) likely have combinations of peripheral airway dysfunction and some degree of pulmonary gas exchange inefficiency to explain their relatively reduced exercise tolerance. While the value of screening of symptomatic smokers using spirometry is undisputed, the results show that mild reduction in FEV₁ may mask physiological impairment that can contribute to negative clinical outcomes in some smokers. In this context, it is evident that among the available resting physiological tests, those that measure lung diffusing capacity and pulmonary gas trapping show the strongest associations with the common symptoms of exercise intolerance and dyspnea.

Funding

Prof. Denis O'Donnell has received funding for this research through Ontario Thoracic Society; William Spear/Richard Start Endowment Fund, Queen's University; Canadian Respiratory Research Network (CRRN). The CRRN is supported by grants from the Canadian Institutes of Health Research (CIHR)—Institute of Circulatory and Respiratory Health; Canadian Lung Association (CLA)/Canadian Thoracic Society (CTS); British Columbia Lung Association; and Industry Partners Boehringer Ingelheim Canada Ltd, AstraZeneca Canada Inc., and Novartis Canada Ltd. Amany F. Elbehairy has received financial support through the John Alexander Stewart scholar award from Queen's University.

Declaration of interest

The funders had no role in the study design, data collection and analysis, or preparation of the manuscript. Amany F. Elbehairy, Azmy Faisal, Jordan A. Guenette, Dennis Jensen, Katherine A. Webb, Rashid Ahmed and J. Alberto Neder have no conflicts of interest that are directly relevant to the content of this article.

Author contributions

All the authors played a role in the content and writing of the manuscript. In addition, D.E.O. was the principal investigator and provided the original idea for the study; D.E.O. and K.A.W. had input into the study design and conduct of study; K.A.W., J.A.G., and A.F.E. collected the data; A.F.E., A.F., J.A.G., K.A.W., and R.A. performed data analysis and prepared it for presentation.