Our retrospective study was performed based on the analysis of data obtained from the medical records of subjects with COPD who underwent incremental cardiopulmonary exercise testing in the pulmonary function laboratory of the Centro Hospitalar e Universitário de Coimbra (Portugal) between January 2016 and December 2019. The number of patients was determined by the availability of incremental exercise test and CT scan of the thorax that were performed within a 6-month period. The inclusion criteria were the diagnosis of COPD according to international guidelines1 and being clinically stable during the previous 3 months (without baseline oxygen desaturation, increased wheezing, cough, dyspnea or sputum). The exclusion criteria were prior diagnosis of other cardio-respiratory conditions (i.e., asthma, interstitial lung diseases, primary pulmonary hypertension, chronic heart failure, muscular and peripheral vascular diseases) that could affect exercise performance. None of the patients had undergone any pulmonary rehabilitation program and none were on long-term oxygen therapy. Written informed consent was obtained from all subjects. This study was approved by the Ethics Commission of Centro Hospitalar e Universitário de Coimbra (234/2021).

Pulmonary function tests were performed in all subjects before CPET including spirometry, diffusing capacity of the lung for carbon monoxide (DLCO) and body plethysmography (MasterScreen PFT, Jaeger, Wurzburg, Germany). All procedures were performed according to the international recommendations.8-10 Measurements included forced expiratory volume in one second (FEV1), forced vital capacity (FVC), residual volume (RV), total lung capacity (TLC) and the DLCO as measure of lung diffusion capacity. Spirometric, DLCO and lung volumes measurements were standardized as percentages of predicted values by using prediction equations.11-13

All patients underwent physician-supervised standard incremental cycloergometer (Blue Cherry, XScribeTM) using a protocol of incremental intensity according to international recommendations.14,15 Typical values were described by Wasserman et al.16 Oxygen uptake (VO2), carbon dioxide production (VCO2), ventilation (VE), tidal volume (VT) and end-tidal CO2 partial pressure (PETCO2) were measured by breath-by-breath and averaged every 10 s for subsequent analysis. The cardiac frequency was monitored via 12 leads ECG. Arterial oxygen saturation was monitored by the use of a pulse oxymeter (SpO2). Peak work rate (WR), peak VO2 and peak VE were defined, respectively, as the highest level of exercise and the highest VO2 and VE that could be sustained for at least 15 s during the last stage of incremental exercise. The oxygen pulse was calculated by dividing VO2 by heart rate. Breathing reserve (BR) measured at peak exercise was estimated as the ratio of VE over the maximal ventilatory ventilation (MVV) that was calculated by multiplying the FEV1 by 40. The anaerobic threshold was detected using the V-slope method using computerized regression analysis of the slopes of the VCO2 vs VO2 plot, which detected the beginning of the excess CO2 output generated from the buffering of [H+]. The physiological dead space-to-tidal volume ratio (VD/VT), an index of lung gas exchange efficiency, was calculated. The modified Borg scale was used to rate the intensity of breathing discomfort and of leg discomfort at peak exercise.17

Subjects with COPD were classified into four phenotypes based on CT scans using two cutoffs as previously described:%LAA-950 of 15% and the 75th percentile of the percentage of WA%. Patients with an %LAA-950 lower than 15% and WA% lower than the 75th percentile were categorized into normal phenotype group; patients with an %LAA-950 higher than 15% and %WA lower than the 75th percentile value were categorized into the emphysema dominant phenotype; patients with an %LAA-950 lower than 15% and WA% higher than the 75th percentile were categorized into the airway dominant phenotype; and patients with an %LAA-950 higher than 15% and %WA higher than the 75th percentile were categorized into the mixed phenotype.

Quantitative data with a normal distribution were expressed as mean and standard deviation (SD) and those variables without normal distribution were expressed as median and interquartile range (P25-P75). Differences of continuous variables with normal distribution were determined by ANOVA and those with non-normal distribution with nonparametric Kruskal-Wallis test. Categorical variables were compared using the χ2 test. Pearson's correlation coefficients were used to assess bivariate relationships between CT-based and CPET measures. Multivariate regression linear models (method: Enter) were performed to identify predicting variables of ventilatory response to exercise. A p value less than 0.05 was considered statistically significant. Statistical analysis was performed by using the SPSS 25.0 software.

A total of 80 subjects with COPD followed in the Centro Hospitalar e Universitário de Coimbra were included in this retrospective study and their characteristics are described in detail in Table 1. The mean age of participants was 65.0 ± 11.3 years and 78.8% were male subjects. According to the GOLD classification, 19% of the patients were categorized as GOLD 1, 49% as GOLD 2, 25% as GOLD 3 and 7% as GOLD 4. In overall, the subjects exhibited a moderate obstruction (FEV1 59.0 ± 20.0% of predicted), mild level of hyperinflation (RV/TLC 45.6 ± 11.0%), and a mildly impaired pulmonary diffusion (DLCO 65.3 ± 20.7).

In all subjects, the values of CT emphysema (%LAA) an airway wall thickening (%WA) were 9.4 ± 9.0% and 41.4 ± 4.8%, respectively. The 75th percentile for %WA, used for defining the presence of significant airway disease, was 44.6%.

Based on the CT findings of emphysema (%LAA) and air wall thickening (%WA), the subjects were classified into four phenotypes. Of the 80 subjects with COPD, 32 (40%) were defined as normal, 21 (26%) as airway dominant, 20 (25%) as emphysema dominant and 7 (9%) as mixed phenotype.

There was no difference in age and smoking history (measured by pack-years) among the phenotypes (p>0.05) (Table 1). Body mass index was lower in the emphysema dominant and mixed phenotypes than in the normal and air dominant subgroups.

Subjects with COPD in the emphysema and mixed phenotype had more severe airflow limitation as assessed by FEV1. Moreover, we found a significant difference in DLCO, with lower values for emphysematous and mixed type subjects than for the remaining two CT-based phenotypes.

CPET was performed in all subjects. In overall, the exercise tolerance was reduced in the study group (peak VO2 16.0 ± 4.2 ml/min/Kg [67.4% predicted]). CPET data obtained from of the whole sample of patients are summarized in Table 2.

According to the classification in phenotypes, we found significant differences in distribution in peak VO2 (%pred) and peak workload, both with a lower value in the emphysema and mixed phenotype. The other cardiovascular and metabolic variables distribution was not significant diverse among phenotypes, as well as the dyspnea intensity and leg fatigue at peak exercise (p>0.05).

Regarding the ventilatory response, we analyzed the distribution of VE/VCO2, VE/VCO2 slope and PETCO2 at peak exercise according to CT phenotypes (Fig. 1). VE/VCO2 was significantly higher in the mixed, emphysema and air dominant phenotypes than in the normal phenotype (p = 0.031, p = 0.002, p = 0.038, respectively). Concerning VE/VCO2 slope, values were significantly higher in mixed and emphysema phenotypes than in the normal one (p = 0.006 and p = 0.033, respectively). The variable PETCO2 was significantly lower in the emphysema phenotype and air dominant phenotypes than in the normal phenotype (p = 0.006 and p = 0.026, respectively). No significant differences were found in the distribution of peak-to-rest change in PaCO2 (ΔPaCO2) or VD/VT.

Figure 1.

Distribution of VE/VCO2, VE/VCO2 slope and PETCO2 at peak according to phenotypes of COPD.

(0.26MB).

CT indices of both lung emphysema severity and airway wall thickening were negatively associated with exercise capacity and ventilatory response (Table 3). The results demonstrated that, in all four phenotypes, LAA% was inversely correlated with peak VO2, O2 pulse and peak workload (r=−0.376, p = 0.01; r=−0.297, p = 0.008; r=−0.333, p = 0.003; respectively). In contrast, LAA% showed a positive correlation with peak VE/VCO2 and VE/VCO2 slope (r = 0.437, p = 0.006 and r = 0.503, p<0.001). The extent of emphysema was related with increased dead space as LAA% was positive correlated with VD/VT (r = 0,270, p = 0,023). With regards to WA%, it was inversely correlated with O2 pulse and peak workload (r=−0.575, p<0,001 and r=−0.419, p = 0.001), but not with peak VO2. On the other hand, airway wall thickening was positive correlated with peak VE/VCO2, VE/VCO2 slope and VD/VT (r = 0.541, p<0.001; r = 0.299, p = 0.033; r = 0.292, p = 0.037; respectively).

A multiple linear regression was run to predict exercise capacity and ventilatory response in unadjusted and adjusted models for age, gender, BMI and FEV1 (Table 4). Emphysema severity (%LAA) was the only independent predictive variable for VE/VCO2 in unadjusted (β 0.345; SE 0.151; 95% CI 0.004/0.618; p = 0.047) and adjusted models (β 0.343; SE 0.147; 95% CI 0.009/0.610; p = 0.044), as well for VE/VCO2 slope in unadjusted (β 0.496; SE 0.085; 95% CI 0.152/0.492; p<0.001) and adjusted models (β 0.496; SE 0.081; 95% CI 0.130/0.455; p = 0.001). The evaluation of the variable peak VO2 demonstrated that %LAA was also the only independent predictive variable in unadjusted model (β −0.441; SE 0.203; 95% CI −1.062/−0.247; p = 0.002), but not in adjusted model.