Oxygen Uptake Efficiency Slope and Breathing Reserve, Not Anaerobic Threshold, Discriminate Between Patients With Cardiovascular Disease Over Chronic Obstructive Pulmonary Disease

Objectives The study sought to compare the relative discrimination of various cardiopulmonary exercise testing (CPX) variables between cardiac and respiratory disease. Background CPX testing is used in many cardiorespiratory diseases. However, discrimination of cardiac and respiratory dysfunction can be problematic. Anaerobic threshold (AT) and oxygen-uptake to work-rate relationship (VO2/WR slope) have been proposed as diagnostic of cardiac dysfunction, but multiple variables have not been compared. Methods A total of 73 patients with chronic obstructive pulmonary disease (COPD) (n = 25), heart failure with reduced ejection fraction (HFrEF) (n = 40), or combined COPD and HFrEF (n = 8) were recruited and underwent CPX testing on a bicycle ergometer. Following a familiarization test, each patient underwent a personalized second test aiming for maximal exercise after ∼10 min. Measurements from this test were used to calculate area under the receiver-operator characteristic curve (AUC). Results Peak VO2 was similar between the 2 principal groups (COPD 17.1 ± 4.6 ml/min/kg; HFrEF 16.4 ± 3.6 ml/min/kg). Breathing reserve (AUC: 0.91) and percent predicted oxygen uptake efficiency slope (OUES) (AUC: 0.87) had the greatest ability to discriminate between COPD and HFrEF. VO2/WR slope performed significantly worse (AUC: 0.68). VO2 at the AT did not discriminate (AUC for AT as percent predicted peak VO2: 0.56). OUES and breathing reserve remained strong discriminators when compared with an external cohort of healthy matched controls, and were comparable to B-type natriuretic peptide. Conclusions Breathing reserve and OUES discriminate heart failure from COPD. Despite it being considered an important determinant of cardiac dysfunction, the AT could not discriminate these typical clinical populations while the VO2/WR slope showed poor to moderate discriminant ability. (Identifying an Ideal Cardiopulmonary Exercise Test Parameter [PVA]; NCT01162083)

C ardiopulmonary exercise testing (CPX) is recommended for the identification of the key-limiting organ in a patient presenting with exercise intolerance or dyspnea (1). Most diagnostic algorithms are similar (2)(3)(4)(5): peak oxygen uptake (VO 2 ) is used to determine the extent of limitation and the combination of anaerobic threshold (AT) and breathing reserve (BR) is used to determine cause. A BR cutoff of 30% and an AT cutoff of 40% of predicted peak VO 2 have typically been used to discriminate between respiratory and cardiac limitation respectively. However, it may be difficult to establish etiology if abnormalities in cardiac and respiratory function coexist (6). BR, measured at peak (7) or AT (8), discriminates patients with known respiratory disease from healthy adults and those with heart disease, in small, selective studies. These results have not been replicated in independent samples, these studies employed small sample sizes and participants were highly selected.
It is also unclear how best to assess change in status using serial measurements of a single patient when pulmonary and cardiac pathologies coexist, which is not uncommon. In 1 study of chronic heart failure (CHF) patients 40% had spirometry suggestive of chronic obstructive pulmonary disease (COPD) (9).
Another reported that CHF was present in w20% of people with COPD (10).
This study aimed to establish which CPX variables showed the best ability to discriminate between respiratory and cardiac limitation in a prospective cohort of patients with COPD, heart failure with reduced ejection fraction (HFrEF), and coexisting COPD and HFrEF.

METHODS
RECRUITMENT. Patients with a diagnosis of HFrEF or COPD were eligible for the study. HFrEF patients were prospectively recruited from a heart failure clinic. They must have been symptomatic at some point in the past. Sequential symptomatic COPD patients were recruited from an outpatient clinic.
Four patients with COPD found to have ventricular dysfunction, and 4 patients with HFrEF with obstructive spirometry and COPD features (smoking history, typical computed tomography findings, sputum production) were subsequently reclassified into a mixed group.  Discriminatory Capacity of CPX Variables highest 20-s average. The AT was identified using unaveraged breath-by-breath data using the V-slope method (12), and corroborated using other plots. VO 2 at AT, minute ventilation:carbon dioxide (VE/VCO 2 ) ratio at AT, and end-tidal CO 2 (P ET CO 2 ) (mm Hg) were taken at this time point. The oxygen uptake efficiency slope (OUES) was calculated as the slope of the regression line between log 10 minute ventilation (x-axis) and VO 2 (y-axis). The VE/VCO 2 slope was calculated using data until the ventilatory compensation point À slope 1, and using all exercise data, including exercise after ventilatory compensation point À slope 2. Maximal voluntary ventilation (MVV) was calculated as: 40 $ forced expiratory volume in 1 s (FEV 1 ). Predicted values for peak VO 2 and OUES were generated from the SHIP study (13,14). All other calculations were performed using standard methods.
The AT was described as a percentage of predicted peak VO 2 .
A number of CPX variables were never calculated in the SHIP cohort, and so these variables were only analyzed for the 2 disease groups.
OTHER MEASURES. Echocardiography was performed using an IE33 ultrasound system (Philips, Amsterdam, the Netherlands) and B-type natriuretic peptide (BNP) was measured.  Values are mean AE SD, n (%), or median (interquartile range). The p values are between disease groups (excluding healthy controls) by analysis of variance for continuous variables and chi-square analysis for categorical variables.
ACEI ¼ angotensin-converting enzyme inhibitor; ATII-R ¼ angiotensin II receptor blocker; BNP ¼ B-type natriuretic peptide; COPD ¼ chronic obstructive pulmonary disease; eGFR ¼ estimated glomerular filtration rate;  Table 1. Disease groups did not significantly differ by age, but did by gender (p ¼ 0.01) and weight (p ¼ 0.001), and further characteristics were corrected for age, gender, and weight. Most HFrEF patients were symptomatic (5 NYHA functional class I, 30 functional class II, 9 functional class III).
Within the COPD category, 2 patients were categorized as mild, 12 were moderate, 10 were severe, and 3 were very severe.
The SHIP cohort matched 134 healthy controls.
CPX RESULTS AND ASSOCIATIONS WITH EXERCISE CAPACITY. Unadjusted peak VO 2 was similar between the 2 principal groups (COPD 17.1 AE 4.6 ml/min/kg and HFrEF 16.4 AE 3.6 ml/min/kg, p ¼ 0.48). Table 2 gives the adjusted mean values for all CPX variables within each group. 8 patients in the COPD group, 1 in the HFrEF group, 1 in the mixed group, and 2 healthy controls, did not achieve AT and were excluded from analyses on AT dependent variables only.
Among the patients, peak VO 2 correlated with estimated glomerular filtration rate (r ¼ 0.30, p ¼ 0.003) and Log 10 BNP (r ¼ -0.35, p ¼ 0.001) but not with hemoglobin or sodium. Fourteen individuals had atrial fibrillation or flutter but there was no significant difference in peak VO 2 compared to sinus rhythm (p ¼ 0.24).

ASSOCIATIONS BETWEEN SPIROMETRY AND CPX IN
HFrEF. On multivariate regression analysis peak VO 2 did not relate to FEV 1 in patients with HFrEF Only breathing reserve at AT (p ¼ 0.03) and peak (p ¼ 0.01), and peak minute ventilation Table 2, and between the 2 disease groups in Figure 1.

DIFFERENCES IN VARIABLES BETWEEN GROUPS. A comparison of results across groups is shown in
All measures of peak VO 2 , the AT, the VE/VCO 2 -slope and ratio at AT, end-tidal CO 2 , and circulatory power all showed significant differences between healthy controls and each disease group, but not between the disease groups. OUES, OUES/kg, and percent predicted OUES differed significantly between COPD and HFrEF, and between HFrEF and healthy controls.
OUES also differed between COPD and healthy controls but not when corrected for weight or percent predicted. The unadjusted O 2 -pulse differed between all groups, and was significantly higher in the HFrEF than the COPD group but the 2 disease states did not differ as percent predicted. Double product differed between all groups, with lowest values in the HFrEF group, and highest in healthy controls. Breathing reserve at the AT was significantly lower in the COPD compared to the HFrEF group and at peak was significantly lower in the COPD compared with both other groups. VO 2 to work-rate relationship was significantly lower in patients with HFrEF compared to the other groups. OUEP was significantly lower in the COPD compared to the HFrEF group.  Table 3 shows comparisons of the discriminant abilities of the variables, quantified as AUCs. Variables with good discrimination between COPD and HFrEF were breathing AT ¼ anaerobic threshold; BR ¼ breathing reserve; CPX ¼ cardiopulmonary exercise testing; DP ¼ double product; HR ¼ heart rate; OUES ¼ oxygen uptake efficiency slope; PETCO2 ¼ end-tidal CO2; RER ¼ respiratory exchange ratio; VE ¼ minute ventilation; VO2 ¼ oxygen uptake; VO2/WR slope ¼ oxygen-uptake to work-rate relationship; other abbreviations as in Table 1. Other variables were either significantly worse discriminators or did not discriminate at all.
Variables with good discrimination between COPD and healthy controls were breathing reserve, peak VO 2 , VE/VCO 2 at AT, O 2 -pulse, and circulatory power.
Variables with good discrimination between HFrEF and healthy controls were OUES, double product, peak VO 2 , circulatory power, VE/VCO 2 slope, and VO 2 at AT.
Including patients with mixed disease under their primary diagnosis worsened discrimination marginally (Online Table 1).
To ensure that patients not achieving AT were not influencing its power to detect a difference in groups, peak VO 2 was substituted for the AT in these patients.
The AUC for the VO 2 at AT in ml/min was 0.60 and 0.57 as percent predicted peak VO 2 , both similar to the values seen when those not achieving the AT were excluded.  for BR showed sensitivity of 80% and specificity of 100% to predict respiratory disease, correctly classifying 56 of 65 patients without mixed disease.
Using a previously determined algorithm (5) that used the cutoffs for AT of 40% predicted peak VO 2 , and BR of 30%, 26 of 65 patients were correctly classified. Net reclassification improvement for percent predicted OUES over AT showed an improvement of 74.0% (p < 0.001), with an integrated discrimination improvement of 29.6% (p < 0.001).

Among CPX variables OUES and breathing reserve
displayed the greatest ability to discriminate between HFrEF and COPD, and were the only CPX discriminators with AUC >0.8. This discriminant ability was similar to that seen with BNP. OUES also strongly discriminated HFrEF from healthy adults, while BR discriminated COPD from healthy adults.
A potential algorithm to help distinguish patients based on these variables is shown in Figure 3.
Peak VO 2 , the most widely known CPX variable, had no capacity to discriminate between cardiac and pulmonary causes of exercise limitation (AUC: w0.50); both diseases depressed peak VO 2 . Importantly this similarity of peak VO 2 between our 2 principal groups allowed us to compare the ability of other measures to discriminate between cardiac and lung disease without concerns that observed differences merely related to differences in peak VO 2 .  (16). In our current study individualized protocols resulted in similar, and recommended, exercise times (1), suggesting that cardiovascular limitation may lower the slope even with optimal exercise duration.
The O 2 -pulse, a surrogate for stroke volume, was unexpectedly higher in HFrEF, compared to COPD, although this difference was not seen when corrected using predictive equations. We believe that high beta-blockade use within the HFrEF group led to lower heart rates, greater filling times, and therefore higher stroke volumes. Second, the O 2 -pulse is dependent on arteriovenous oxygen content difference, often reduced in COPD patients with lower arterial saturations and higher peak venous saturations. The predicted O 2 -pulse was unsurprisingly significantly higher in healthy controls compared to both disease groups.
All measurements of the VE/VCO 2 relationship failed to discriminate the disease groups, but were significantly higher than healthy controls. In patients with HFrEF they are abnormal due to hyperventilation and perfusion to ventilation mismatching (17). In COPD a number of causes lead to an abnormal VE/ VCO 2 relationship including mismatching of ventilation to perfusion.
OXYGEN UPTAKE EFFICIENCY SLOPE. The OUES, largely effort independent, is calculated as the slope of the semilog relationship between O 2 and minute ventilation (18). OUES appears to be unaffected by COPD. Our group has previously found, within a large retrospective heart failure cohort, that patients with  Abbreviations as in Tables 1 and 2. low percent predicted FEV 1 have lower peak VO 2 but not OUES (19). In the current study OUES did not relate to FEV 1 or K CO (Hb).
OUES was significantly lower in HFrEF than COPD, despite similar exercise capacities, and healthy con- Therefore, the VO 2 /log 10 VE curve may be "shifted" rightward (higher log 10 VE for any given VO 2 ) but the curve's gradient itself is unchanged. This hypothesis may also explain why our mixed cohort had OUES values close to predicted.
BREATHING RESERVE. Breathing reserve has long been suggested as a discriminator of respiratory limitation (3)(4)(5)8). BR at AT has been proposed to reduce the influence of voluntarily cessation of exercise (8).
Both BR at AT and peak showed good discriminatory power. However 32% of our COPD patients did not achieve AT, similar to the study advocating the BR at AT (40%), limiting its widespread applicability. In contrast BR at peak and the OUES are measurable in all. ANAEROBIC THRESHOLD. In previous CPX algorithms a reduced AT would identify heart failure (4,5), yet evidence supporting its role is scarce. In HFrEF VO 2 at AT is reduced (23) and superior to peak VO 2 at predicting prognosis (24). Nery et al. (7), showed VO 2 at AT in patients with mitral valve disease was lower than patients with COPD and healthy controls; however the numbers were small with significant differences in gender and age between groups. These studies are the foundation of what has become a firmly held belief-namely that VO 2 at AT reflects cardiac function. Very few CPX studies performed on patients with COPD report the VO 2 at AT however Medoff et al. (8) found no difference between COPD and heart failure patients with similar exercise capacities, consistent with our findings. We showed that VO 2 at AT (as a percent predicted peak VO 2 ) has poor discriminant ability between the disease groups, and only showed moderate discrimination between healthy adults and the 2 groups. We suggest that this variable is critically determined by muscle function and any chronic process that impairs muscular function will reduce the AT.
There may be a concern that the number of patients without a measured AT influenced our results. Our results are similar to other studies in terms of numbers of COPD patients failing to achieve AT (8)