Chronic obstructive pulmonary disease (COPD) is a common, preventable, and treatable disease characterized by persistent airflow limitation. The chronic airflow limitation progressively traps gas during expiration, resulting in hyperinflation, increased dyspnea, and limitations in exercise capacity. Acute exacerbation of COPD (AECOPD) is defined as an acute worsening of respiratory symptoms necessitating additional therapy. The severity of AECOPD and its increasing frequency is associated with a higher risk of death ^[1]. However, there is a lack of effective indicators to predict AECOPD. A deteriorating airflow limitation is associated with an increasing prevalence of exacerbations ^[2]. Forced expiratory volume in 1 sec (FEV₁) is an indicator of airflow limitation, but it lacks sufficient precision to be used clinically as a predictor of exacerbation or mortality in patients with COPD ^[3]. Only a weak correlation is observed between FEV₁, symptoms, and impaired health status of patients^[4]. Thus, it is important to explore other effective indicators for predicting AECOPD.

Exercise capacity is limited in COPD patients and is significantly correlated with overall survival ^[5]. The probable mechanisms include impaired oxygen delivery to skeletal muscle and ventilatory limitations ^[6]. Cardiopulmonary exercise testing (CPET) is used to assess exercise tolerance and evaluate the pathophysiological mechanism(s) of dyspnea and exercise limitation in patients with COPD ^[7]. Oxygen pulse (O₂P), an important CPET variable, is a noninvasive and reliable method used to estimate stroke volume and cardiac function ^[8] with the exclusion of coexisting diseases, but it has rarely been used to evaluate the severity of COPD. O₂P is defined as oxygen uptake (V̇O₂) divided by heart rate (HR). One study showed that O₂P is impaired during exercise in patients with COPD, and this is partly related to lung hyperinflation ^[9]. Miniati et al. ^[10] reported that peak O₂P is significantly lower in patients with moderate to severe emphysema than in those with no or mild emphysema. Thus, O₂P may be an important indicator for differentiating COPD severity. O₂P_max represents the O₂P at maximal exercise and reflects the maximal aerobic metabolic function. O₂P_max is extensively used for testing patients with cardiovascular diseases, but it has limited use in those with COPD. Furthermore, increased hyperinflation and gas trapping occur during an exacerbation, with reduced expiratory flow ^[11]. A worsening of gas exchange can also result in hypoxemia ^[12]. O₂P is directly related to hyperinflation and reflects gas exchange, so O₂P may be a better predictor of AECOPD. Accordingly, we hypothesized that O₂P_max has a role in COPD and may be related to AECOPD.

In this study, we examined the relationship between O₂P_max and exercise capacity, ventilatory conditions, gas exchange efficiency, and health-related quality of life (HRQL). We also estimated the predictive value of O₂P_max for AECOPD and determined the cut-off value.

This retrospective analysis involved patients with COPD who underwent incremental cardiopulmonary exercise in our exercise laboratory at Beijing Friendship Hospital, Capital Medical University (Beijing, China) from 1 January 2017 to 31 December 2020. This study was conducted in compliance with the Declaration of Helsinki. The Institutional Ethics Committee of Beijing Friendship Hospital of Capital Medical University approved this study (no. 2021-P2-334-01).

All participants were men and women aged ≥ 40 years with a body mass index of 18–32 kg/m² and no episodes of AECOPD within the previous 6 weeks. All of the participants had been diagnosed with COPD and a grade I, II, III, or IV airflow limitation according to the Global Initiative for Chronic Obstructive Lung Disease (GOLD) criteria and underwent CPET in our clinic. After enrollment, the participants’ smoking history, comorbidities (hypertension, coronary heart disease, type 2 diabetes, and hypercholesterolemia), medications used for COPD, and medications used for cardiovascular disease were recorded. The patients were followed up outside of the hospital by telephone. AECOPD was classified as mild (treated with short-acting bronchodilators only), moderate (treated with short-acting bronchodilators plus antibiotics and/or oral corticosteroids), or severe (required hospitalization or visits to the emergency room) ^[13]. The exclusion criteria were respiratory diseases other than COPD, heart failure, uncontrolled diabetes mellitus, uncontrolled hypertension, orthopedic problems, neurologic problems, and hemopathy.

All participants underwent static spirometry (MasterScreen Body; CareFusion, San Diego, CA, USA), including measurements of FEV₁, forced vital capacity (FVC), vital capacity (VC), peak expiratory flow, and maximal mid-expiratory flow rate 75/25. Body plethysmography was performed to measure the residual volume (RV), total lung capacity (TLC), and the transfer factor of the lungs for carbon monoxide/alveolar volume. All pulmonary function tests were performed according to the American Thoracic Society/European Respiratory Society guidelines^[14].

CPET was performed according to the American Thoracic Society/European Respiratory Society guidelines ^[15]. All patients rested for 3 min, performed unloaded pedaling for 3 min, and then underwent an incremental, symptom-limited exercise test using an electronically braked cycle ergometer (ViaSprint, CareFusion, Hoechberg, Germany) at a pedaling rate of 40–70 rpm (approximately 60 rpm). The work rate (WR) was set to 5–20 W/min and was gradually increased by 5–15 W/min (5 W/min if FEV₁ < 1.0 L and 10 W/min if FEV₁ ≥ 1.0 L). The test ended when the cadence fell below 40 rpm and did not return with an exhortation, the patients requested termination, or the technician terminated the test for safety ^[16]. The limiting symptom was defined as the inability to maintain pedaling frequency or the development of intolerable shortness of breath. Chest pain (suggesting myocardial ischemia), ventricular tachycardia, and blood pressure (BP) ≥ 240/130 mmHg also prevented further exercise. The O₂P_max was defined as V̇O_2max/HR. According to the current guidelines ^[17], O₂P_max values ≥ 80% of the predicted value were considered normal, whereas O₂P_max < 80% of the predicted value was considered pathological. Accordingly, Group 1 was defined as patients with impaired O₂P_max (< 80% predicted), and Group N was defined as patients with normal O₂P_max (≥ 80% predicted). V̇O₂, carbon dioxide production (V̇CO₂), minute ventilation (V̇_E), breathing reserve (BR), and 12-lead electrocardiography were measured continuously. BP was measured at the end of every 2 min. Load was reported at maximum exertion. V̇O₂/kg, the ventilatory equivalent for carbon dioxide (EqCO₂), and HR recovery after 1 min of rest (HRR₁) were calculated automatically. The anaerobic threshold (AT) was identified using the modified V-slope method.

The COPD Assessment Test (CAT) and the St. George’s Respiratory Questionnaire (SGRQ) were used to measure the impaired health status of patients with COPD. The SGRQ includes four categories, such as symptoms, physical activities, psychosocial impacts, and the total score. The CAT total score ranges from 0 to 40, and the SGRQ total score ranges from 0 to 100. An SGRQ score ≥ 25 was used as the threshold for breathlessness, and the equivalent cut-off point for the CAT was 10^[13]. The modified British Medical Research Council Questionnaire (mMRC) was used to assess dyspnea severity. The mMRC score ranges from 0 to 5, and ≥ 2 was considered the threshold for separating “less breathlessness” from “more breathlessness.” Because our patients were using mouthpieces and could not conveniently answer the doctors’ questions during exercise, the dyspnea level and leg muscle fatigue were scored using Borg’s 10-point category-ratio scale (Borg CR10 scale) after the patients finished exercising. A Borg CR10 scale score of 4–6 indicated moderate to severe symptoms.

Fifty-seven patients (63%) completed the 3-year follow-up. AECOPD was tracked during the next 3 years. Telephone calls were made to the patients every 4 months to determine whether they had experienced any exacerbations, had taken any short-acting bronchodilators, antibiotics, or oral corticosteroids, or had been hospitalized. According to the GOLD ^[13], patients with AECOPD (AE2 group) developed more than one acute exacerbation episode within 1 year. Patients without AECOPD (AE1 group) had developed no or only one acute exacerbation episode within the year.

Continuous data were expressed as mean ± standard deviation, number, and percentage, or median (IQR), while categorical data were expressed as frequencies and percentages. Continuous data were compared using the t-test or Mann-Whitney test, and categorical data were compared using the chi-square test. A P-value < 0.05 was considered significant.

Pearson’s correlation coefficient (r) was used to assess the correlations between the static lung function variables (FEV₁/FVC%, FEV₁%, FVC%, and VC) and the CPET variables (AT%, WR_AT, WR_max, WR%, V̇O_2AT, V̇O_2max, V̇O₂%, V̇O₂/kg_max, V̇O₂/kg%, V̇_Emax, V̇_E%, BR_AT, BR_max, EqCO_2AT, EqCO_2max, and HRR₁) and O₂P_max.

Univariate logistic regression analysis was performed to assess the efficiency of O₂P_max for predicting AECOPD. Static lung function and CPET variables that were significantly different between the AE2 and AE1 groups were assessed to identify risk factors for AECOPD using multivariate logistic regression analysis. Wald’s test was used to obtain the logistic regression parameters.

The discriminative ability of O₂P_max in predicting AECOPD was evaluated by receiver operating characteristic (ROC) analysis. The area under the ROC curve (AUC) and the 95% confidence interval (CI) were calculated. The best cut-off value was determined.

Table 1 shows the patient demographics and baseline characteristics. Ninety-one patients with COPD were divided into Group 1 (n = 27) and Group N (n = 64) according to their O₂P_max value. No significant differences in age, sex, body height, body mass index, smoking history, comorbidities, or medications were observed between the two groups. However, significant differences in body weight and classification of the airflow limitation were detected between the two groups (P < 0.05). The airflow limitation was more severe in Group 1 than in Group N.

Table 2 shows the static lung function parameters. Patients with a lower O₂P_max had lower FVC%, VC%, FEV₁%, and FEV₁/FVC% values (P < 0.05).

Table 3 shows the CPET variables. AT, WR, V̇O₂, V̇O₂/kg, HR, systolic BP, diastolic BP, V̇_E, BR, and EqCO₂ were measured at the AT and maximum exercise during CPET. HRR₁ was calculated after the exercise. Significant differences in AT, WR%, V̇O₂%, V̇O₂/kg%, V̇_E%, EqCO_2AT, and EqCO_2max were observed between the two groups (P < 0.05). Among these variables, EqCO_2AT and EqCO_2max were higher in patients with impaired O₂P_max, whereas the other variables were lower. Significant differences were also detected in HRR₁ (P < 0.05), which was regarded as a circulatory parameter.

Table 4 shows the CAT, mMRC, Borg CR10 scale, and SGRQ scores. The Borg CR10 scale, SGRQ total, SGRQ symptom, and SGRQ activity scores were significantly higher in Group 1 than those in Group N (P < 0.05).

Figure 1 shows the correlation coefficients between O₂P_max and FEV₁/FVC%, FEV₁%, and FVC%, which were regarded as lung ventilation function variables. O₂P_max was slightly but significantly correlated with FEV₁/FVC% (r = 0.207, P < 0.05), FEV₁% (r = 0.267, P < 0.05), and FVC% (r = 0.288, P < 0.01).

Figure 1.  Correlations between O₂P_max and FEV₁/FVC%, FEV₁%, and FVC%. Abbreviations: O₂P, oxygen pulse; FEV₁, forced expiratory volume in 1 sec; FVC, forced vital capacity.

Figure 2 shows the correlation coefficients between O₂P_max and the CPET variables. O₂P_max was positively correlated with AT (r = 0.623, P < 0.001), WR% (r = 0.613, P < 0.001), V̇O_2max% (r = 0.681, P < 0.001), V̇O₂/kg_max% (r = 0.707, P < 0.001), and V̇_Emax% (r = 0.399, P < 0.001), but negatively correlated with EqCO_2AT (r = −0.302, P < 0.01) and EqCO_2max (r = −0.214, P < 0.05).

Figure 2.  Correlations between O₂P_max and the CPET parameters Abbreviations: O₂P, oxygen pulse; AT, anaerobic threshold; WR, work rate; V̇O₂, oxygen uptake; V̇O₂/kg, oxygen uptake per kilogram; V̇E, minute ventilation; EqCO₂, ventilatory equivalent for carbon dioxide.

Table 5 shows that FVC%, FEV₁%, O₂P_max%, WR_max, and V̇O_2max were significantly lower in the AE2 group than in the AE1 group (P < 0.05).

The univariate logistic regression confirmed that O₂P_max was an independent predictor of AECOPD [odds ratio (OR), 1.068; 95% CI, 1.023–1.116; P = 0.003]. Five variables (FVC%, FEV₁%, O₂P_max, WR_max, and V̇O_2max) were used to build a logistic regression model to predict AECOPD in a multivariate analysis (Tables 6-7). Only O₂P_max was a predictor of AECOPD (OR = 1.062, 95% CI = 1.012–1.114, P = 0.015). ROC curve analysis was applied to categorize the optimal cut-off value of O₂P_max for exacerbation (Figure 3). The AUC was 0.739 (95% CI = 0.609–0.869, P = 0.002). The cut-off value was 89.5%, with a sensitivity of 63.30% and a specificity of 77.80%.

Figure 3.  ROC curve of O₂P_max as an overall predictor of AECOPD. Abbreviations: AUC, area under the ROC curve; CI, confidence interval.

The main finding of the present study was that impaired O₂P_max in patients with COPD indicated an impaired ventilatory condition, gas exchange efficiency, exercise capacity, and HRQL. The correlation analysis showed that O₂P_max was correlated with FEV₁/FVC%, FEV₁%, FVC%, WR, AT, V̇O₂, V̇O₂/kg, V̇_E, and EqCO₂. Moreover, O₂P_max was significantly different between the AE2 and AE1 groups, which was a new finding. The univariate and multivariate logistic regression analyses demonstrated for the first time that O₂P_max predicted AECOPD (OR 1.068, 95% CI = 1.023–1.116). The optimal O₂P_max cut-off value was 89.5%. The AUC was 0.739 (0.609–0.869) with a sensitivity of 0.633 and a specificity of 0.778.

V̇O₂ represents metabolic efficiency, and WR represents exercise capacity. The AT represents a transition point from aerobic to anaerobic metabolism, and a lower AT indicates reduced oxygen delivery to muscle cells. Declines in AT%, WR, and V̇O₂ contribute to exercise intolerance ^[18]. The present study indicated that AT, WR%, V̇O₂%, V̇O₂/kg%, and V̇_E% were lower in Group 2 than those in Group 1 (impaired O₂P_max group). This finding suggests that O₂P_max might be an effective indicator for estimating exercise capacity. Patients with COPD develop dynamic hyperinflation during exercise. Dynamic hyperinflation can reduce left ventricular stroke volume secondary to the increased intrathoracic pressure, which, in turn, decreases preload by reducing venous return and the volume of the left ventricle. This explains why O₂P_max is impaired in patients with COPD. Hyperinflation leads to limited cardiac performance, which occurs when there is a mechanical limitation to an increase in V̇_E. Therefore, dynamic hyperinflation in patients with COPD may adversely affect exercise tolerance by reducing O₂P_max. Moreover, lung hyperinflation impairs cardiopulmonary interactions and leads to impaired muscle oxygen availability during exercise. Research has also shown that impaired circulation is not usually a limiting factor for exercise intolerance in patients with COPD ^[6]. Thus, O₂P_max may be a predictor of exercise capacity independent of cardiovascular function. The present study also demonstrated that O₂P_max was highly positively correlated with AT, WR%, V̇O_2max%, and V̇O₂/kg%. Montes de Oca et al.^[16] showed that inspiratory intrathoracic pressure has a direct relationship with O₂P at peak exercise in patients with severe COPD. Torres-Castro et al. ^[19] reported that V̇O_2peak and W_max decrease in patients with COPD and pulmonary hypertension, in line with the decrease in O₂P. An inadequate increase in stroke volume may contribute to the inability of V̇O₂ to increase. These results explain why V̇O₂ was positively correlated with O₂P_max. WR and O₂P_max represent exercise capacity; thus, it is easy to understand why O₂P_max was positive correlated with WR. In addition, the circulatory system is responsible for oxygen delivery. When the metabolic demands begin to exceed oxygen delivery to contracting muscles, anaerobic metabolism begins. The AT is the exercise level that reflects the metabolic condition of anaerobic glycolysis ^[20]. A lower AT is observed if a patient’s oxygen consumption efficiency is poor. The AT reflects oxygen consumption efficiency, and circulatory function can be used to determine the AT. This explains why O₂P_max was associated with the AT in our study.

The present study also illustrated that O₂P_max can be used to assess a ventilation obstruction. FVC%, VC%, FEV₁%, FEV₁/FVC%, V̇_E%, and BR_AT were lower, and EqCO_2AT and EqCO_2max were higher in patients with impaired O₂P_max than in those without impaired O₂P_max. Additionally, O₂P_max was negatively correlated with EqCO_2AT, and EqCO_2max was positively correlated with FEV₁/FVC%, FEV₁%, FVC%, and V̇_Emax%. Our data corroborate the findings of Frazão et al. ^[21], who reported that the ability to increase V̇_E during aerobic exercise is impaired in patients with COPD because exhalation may not be completed prior to the onset of the next breath, causing an increase in operational lung volume and progressive air retention. EqCO₂ is a measure of ventilatory efficiency and is calculated as V̇_E divided by V̇CO₂. EqCO₂ symbolizes the capacity to remove carbon dioxide. In patients with COPD who develop impaired O₂P_max, these higher EqCO₂ values may be explained by disproportional reductions in V̇_E and perfusion. The decrease in ventilation and perfusion are different ^[21], and perfusion may have a higher degree of reduction in patients with impaired O₂P_max. The correlations between V̇_Emax%, EqCO_2AT, EqCO_2max, and O₂P_max can be explained by the fact that airflow limitation impairs cardiac function ^[21]. The increase in inspiratory intrathoracic pressure decreases lung elastic recoil pressure. The combined effects of decreased lung elastic recoil pressure and increased airway resistance result in reduced V̇_E and the development of hyperinflation ^[21]. Improvement in ventilatory mechanics results in improved cardiac function, which manifests as a decrease in HR with improved O₂P ^[22]. Lung hyperinflation also causes mechanical constraints on the heart and a reduction in preload ^[23]. These interactions between hyperinflation, ventilation and cardiac function may explain why O₂P_max was positively correlated with V̇_Emax%. EqCO₂ was negatively correlated with V̇_E, and O₂P_max was negatively correlated with EqCO₂.

The positive correlations between FEV₁/FVC%, FEV₁%, FVC%, and O₂P_max can be explained by the fact that the decreases in FEV₁/FVC%, FEV₁%, and FVC% contribute to lung hyperinflation, which, in turn, reduces left ventricular stroke volume. Furthermore, O₂P is negatively correlated with lung hyperinflation ^[24], and lung hyperinflation is negatively correlated with FEV₁ ^[25]. In addition, Silvestre et al. ^[26] suggested that the association between lung function and exercise capacity may be related to specific abnormalities in cardiopulmonary performance. All of these findings support the relationships between FEV₁/FVC%, FEV₁%, FVC%, and O₂P_max. BR reflects the respiratory reserve capacity during extreme exercise, and a lower BR is the main feature of ventilation limitation in patients with pulmonary disease. It is calculated as the maximum ventilatory volume minus V̇_E. BR decreases in patients with COPD but increases in those with cardiovascular disease. The present study showed that BR decreased in patients with impaired O₂P_max, whereas no correlation was detected between BR and O₂P_max. This finding could indicate that the decreased O₂P_max in patients with COPD is not only caused by reduced filling of the ventricles but may also be attributable to other mechanisms.

HRR is a marker of cardiac autonomic function independent of the workload and the change in HR during exercise. A low HRR is a common finding in patients with impaired lung function and is associated with decreased survival in patients with COPD ^[27]. In these patients, autonomic nervous dysfunction contributes to the increased work of breathing, which promotes a wider airway caliber ^[27]. HRR₁ is defined as the decrease from peak HR to 1 min of recovery and is an independent predictor of morbidity and all-cause mortality ^[28]. To our knowledge, no study has explored the correlation between HRR₁ and O₂P in patients with COPD until now. Our present results indicate that HRR₁ decreased significantly in patients with impaired O₂P_max. This may be attributed to the fact that both parameters are regulated by the autonomic nervous system, and an increase in sympathetic tone and/or a decrease in parasympathetic tone may reduce HRR₁ and contribute to tachyarrhythmia.

COPD is largely characterized by breathlessness. The mMRC is considered adequate for assessing breathlessness symptoms. However, COPD affects patients beyond dyspnea. The SGRQ is the most comprehensive disease-specific health status questionnaire. The CAT is simpler. The Borg CR10 scale is mainly used to evaluate dyspnea and leg muscle fatigue during maximum exercise ^[29]. Because our patients had instruments in their mouths and could not conveniently answer questions during exercise, we evaluated the patients using the Borg CR10 scale at the end of the exercise. As a result, the Borg CR10 scale score and the SGRQ symptom, activity, and total scores were higher in the O₂P_max impaired group than those in the other groups. Souza et al. ^[30] showed that patients with COPD who had lower activities of daily living exhibit lower O₂P, which may reflect the presence of a limiting cardiac component. Lan et al. ^[31] indicated that the increased O₂P_max is accompanied by improvements in the SGRQ. Therefore, O₂P_max is a valid indicator of HRQL and should be measured in patients with COPD. This may be attributed to the increase in right ventricular afterload during exercise, impairing stroke volume, which was augmented by increased intrathoracic pressure swings and is related to O₂P_max ^[32].

AECOPD is associated with an increased risk of death ^[33]. It is defined as an acute worsening of respiratory symptoms that results in additional therapy ^[34]. AECOPD is classified as mild (treated with short-acting bronchodilators only), moderate (treated with short-acting bronchodilators plus antibiotics and/or oral corticosteroids), or severe (requires hospitalization or an emergency room visit). No effective indicator has been established to predict AECOPD. History of earlier AECOPD may be a predictor of frequent exacerbations ^[35]. Deteriorating airflow limitation is associated with an increasing prevalence of exacerbations ^[2], but FEV₁ lacks sufficient precision to be used clinically as a predictor of exacerbations or mortality in patients with COPD^[36]. A novelty of our study is that O₂P_max was an effective predictor of AECOPD, and, to the best of our knowledge, this is the first study to identify the O₂P_max cut-off value for predicting AECOPD. Wu et al. ^[37] reported that patients with impaired peak O₂P were hospitalized more frequently during a 1-year follow-up. In the present study, the O₂P_max decreased in the AE2 group. Multivariate and univariate logistic regression analyses showed that O₂P_max predicted AECOPD. We used a ROC curve to determine the best cut-off value of O₂P_max for AECOPD (89.5%). The AUC was 0.739 (0.609–0.869) with a sensitivity of 0.633 and a specificity of 0.778. Our findings indicate moderate accuracy of O₂P_max for predicting AECOPD. Further research is needed to validate whether O₂P_max, in combination with other parameters, would increase the accuracy of prediction. Nevertheless, it is important to inform patients of the high risk of acute exacerbations once the CPET reveals an impaired O₂P_max value (< 89.5%). Doctors should choose higher-level treatment strategies for these patients.

Why lower O₂P_max is associated with more frequent AECOPD is not entirely clear. To our knowledge, progressive respiratory failure, cardiovascular disease, and other diseases are the primary cause of death in patients with COPD hospitalized for an exacerbation ^[36]. In addition, deteriorating airflow limitation is associated with an increasing prevalence of exacerbations ^[2]. Hyperinflation and gas trapping increase with the reduced expiratory flow during exacerbations ^[11]. Therefore, the association between impaired O₂P_max and AECOPD may be related to hyperinflation, which can be caused by airflow limitation and is associated with poor cardiac function. A larger scale study is needed to validate the role of O₂P_max in the risk of AECOPD.

This study had several limitations. First, this was a retrospective study, and not all of the patients finished the 3-year follow-up for AECOPD. Additionally, the sample size was small. Second, O₂P_max is an effective indicator of cardiac function and muscle function, but the echocardiographic parameters and stroke volume were not tested. Third, we did not measure dynamic lung hyperinflation parameters, such as the end-expiratory lung volume, inspiratory capacity, or RV using CPET. Thus, the evidence indicating that O₂P_max is related to dynamic lung hyperinflation is insufficient. Fourth, because not all of the enrolled patients underwent hemoglobin concentration and blood gas analyses, we did not consider these factors in the analysis. Hemoglobin is very important for oxygen carrying capacity, and blood gas values reflect disease severity to some extent. Nevertheless, our current study found that an indicator can play a role in the assessment of COPD severity and prognosis.

This present study illustrated that impaired O₂P_max indicates impairments in exercise capacity, ventilatory condition, gas exchange efficiency, and HRQL. The most important finding was that O₂P_max was an independent predictor of AECOPD. This is the first study to identify the O₂P_max cut-off value.