Spirometry was performed in seated position according to American Thoracic Society/European Respiratory Society (ATS/ERS) guidelines [19]. Reference values for FVC and FEV₁ were from Crapo et al. [20]. Chronic airflow limitation was defined as an FEV₁/FVC ratio of below 0.7 [19]. Restrictive respiratory impairment was defined as an FEV₁/FVC ratio of ≥0.7 and FVC of <80% predicted [19]. Lung volumes measured by plethysmography were not available.

The study incorporated the following details: (a) clinical and anthropometric characteristics of patients with cancer undergoing CPET; (b) adherence to international guidelines for methods of CPET [13]; and (c) the safety of CPET defined as the reported adverse events.

The exercise testing equipment consisted of a stationary cycle ergometer (Med Graphics CPX Ultima system, Medical Graphics Corporation, St. Paul, MN) that was calibrated before and after each test. The mechanical dead space volume, depending on the mouthpiece and connections used, ranged from 45 to 65 mL for this system. Calibration of gas concentrations using primary standard gases and flow was performed using a 3 L syringe prior to each test. All tests were performed by the same 3 certified exercise technologists.

Subjects were asked not to exercise on the day of the test and or to eat or drink caffeinated beverages 4 hours before the test. Following explanation each procedure, an informed consent was obtained. Prior to beginning the test subjects were familiarized with the stationary cycle ergometer and mouthpiece and cycled on the ergometer for approximately 10 minutes. They were seated and breathed through a mouthpiece with a nose clip in place. After a minimum of five minutes of resting measurements, they were exercised on the ergometer with increasing workloads at increments of 5 - 15 Watts, based on patients’ tolerability, using the Godfrey protocol [21]. Maximal effort was determined by the patient achieving a plateau in VO₂max (average of 5 highest consecutive V’O₂ values near peak exercise) and a respiratory exchange ratio of at least 1.1 at peak exercise. Data collection continued for several minutes post-exercise for gas collection and ECG monitoring purposes.

The following variables were measured every 15 seconds: minute ventilation (V’_E), inspired oxygen concentration, expired oxygen tension, inspired carbon dioxide output, oxygen uptake (V’O₂), expired carbon dioxide output. Anaerobic threshold (AT) was determined by the V-slope method and verified by the crossover of ventilatory equivalents for O₂ and CO₂ [22]. Heart rate and rhythm were monitored continuously throughout the study with the 12-lead ECG. In addition, the following variables were derived: maximum voluntary ventilation (MVV), ventilatory reserve (V’_E/MVV), respiratory exchange ratio (RER), O₂-pulse (V’O₂/HR), ventilatory equivalents for carbon dioxide and oxygen (V’_E/V’CO₂ and V’_E/V’O₂, respectively). Normal values for these variables were derived from Sun et al. [22]. Criteria for achieving maximal effort included: (a) a constant plateau in V’O₂ (average of 5 highest consecutive V’O₂ values near peak exercise with <150 mL/min variability) (b) achieving an RER of ≥1.05, and (c) heart rate <10 beats/min of the age-predicted maximum. Anaerobic threshold was achieved when there was a discernable increase in the V’_E vs V’CO₂ relationship and when V’_E/V’O₂ increased without simultaneous increase in V’_E/V’CO₂ during progressively increasing work rate [12]. Exercise testing was stopped when symptoms developed, including intolerable dyspnea, chest pain, significant ST-segment depression on electrocardiogram, drop in systolic blood pressure or arterial oxygen saturation ≤88%.

All patients with V’O₂max (expressed in mL/min) less than 84% predicted were considered as having functional limitation of cardiovascular, pulmonary vascular or ventilatory origin [12] [13]. Predicted values for V’_E/MVV, physiologic dead space (Vd/Vt) and O₂-pulse were derived from Sun et al. [22]. No patients experienced adverse events during testing.

The primary source of exercise limitation was determined based on CPET results with predicted values based on age, gender and BMI [12]. Patients with a peak V’O₂ below 84% predicted were considered as having a ventilatory and/or circulatory limitation [12] [22]. A normal V’_E/MVV (<70%), an O₂-pulse that remained low and failed to increase and elevated ventilatory equivalents that failed to decrease with exercise indicated a circulatory deficit [12]; if, in this group, there was no clinical, imaging or echocardiographic evidence of left heart failure, they were categorized as having pulmonary vascular limitation [12]. Finally, the slope and intercept for V’_E/V’CO₂ was computed according to the relationship y = a + bx, where y was the difference between V’_E at rest and peak exercise, x was the difference between V’CO₂ at rest and peak exercise, a was the intercept and b was the slope [18]. Musculoskeletal impairment was defined as exercise limitation in the absence of ventilatory or circulatory limitation and having achieved anaerobic threshold and with V’O₂max remaining within normal limits.

Descriptive data were shown as mean and standard deviation. Comparisons amongst subcohorts were conducted by multifactorial analysis of variance (ANOVA) with adjustments for age, gender and BMI [23]. Associations between physiologic variables were determined by Pearson’s correlation, expressed as r². The relationships between V’O₂max and FVC and FEV₁ were adjusted for anthropometric characteristics. Linear regression was used to assess the slope and intercept of V’_E/V’CO₂ based on y = a + bx, where y was V’_E and x was V’CO₂. Data for the entire exercise from rest to peak exercise was included to compute the slope and intercept [22]. A p-value of <0.05 was considered statistically significant for intergroup comparisons and for inter-variable associations.

The records of 43 patients referred for evaluation of dyspnea at the cancer center were reviewed; 7 did not have evidence for cancer. Complete clinical and physiological data were available for the remaining 36 patients (males 20, females 16), of which 31 (86%) were identified as having a ventilatory or circulatory limitation, and one was classified as having musculoskeletal limitation, based on clinical and physiologic data. Ten patients exhibited normal lung function and CPET findings and were categorized as having normal exercise capacities. Smoking history was available in 22 patients; ten were former smokers (including one individual who was classified as normal), ranging from 10 to 66 pack-years; the remaining 12 denied exposure to tobacco products. One patient (a former smoker) also gave a history of occupational chemical exposure. Three patients had hemoglobin levels <10 gm/dL.

Thirty-two patients (89%) had solid cancers, the most common being lung (11), prostate (6) and breast (5). Fourteen had a history of more than one tumor, 12 with solid type, the most common being lung (3) and prostate (3). Eight patients were diagnosed with hematopoietic malignancies: non-Hodgkin lymphoma (5), Hodgkin disease (2) and leukemia (1). Two of the lymphomas occurred in patients with solid tumors (lung and breast, one each).

Thirty-two patients (89%) had received treatment for their cancer(s) prior to undergoing CPET: Eleven with chemotherapy and 10 hormonal; 10 patients were treated with biologics (including 1 with Bacille-Calmette-Guerin [BCG] vaccine), 4 in combination with hormonal or chemotherapy. Three of five breast cancer patients received anthracycline derivatives. Four patients also received radiotherapy, 2 in combination with chemotherapy, and all directed at chest or breast fields. At the time CPETs were done, no patients were receiving biologic agents (other than BCG).

Table 1 lists anthropometric and lung function characteristics for the healthy

Values represent mean ± SD. FVC, forced vital capacity; FEV₁, forced expiratory volume in 1 sec; MVV, maximum voluntary ventilation. ^†Two-tailed Student t-test.

control subjects (n = 10) and all patients combined (n = 36). Patients were older by a mean of 11.7 years (p = 0.01). All lung function variables in patients corrected for age, gender and BMI were statistically significantly higher than in the control group.

Table 2 lists anthropometric and lung function characteristics for five separate cohorts including the control subjects. The largest cohort comprised the pulmonary vascular group (n = 18, 50% of patients). The oldest patients were in the ventilatory and (one) musculoskeletal patients. The ventilatory group exhibited the highest mean BMI (29.8 kg/m²). The lowest group mean FVC, FEV₁, FEV₁/FVC and MVV was in the ventilatory cohort (75% predicted, 72% predicted, 0.67 and 68% predicted, respectively). Twenty-six (72%) patients had expired by 2018, one year or more before the end of the survey period; all succumbed to progression of their malignancies.

Table 3 shows cardiopulmonary exercise data recorded at peak exercise for control subjects and all 36 patients combined. Amongst patients, mean exercise duration was 9 minutes, ranging between 6.5 minutes (the musculoskeletal patient, n = 1) and 11.4 minutes (ventilatory, n = 10). The mean V’O₂max for all patients was 15 mL/kg/min (64.6% ± 18.9% predicted). Table 4 lists cardiopulmonary exercise variables at peak exercise in the subjects subdivided according to circulatory or ventilatory etiology of dyspnea. The lowest mean ventilatory reserve was in the ventilatory cohort (26% of the maximum voluntary ventilation,

Values represent mean ± SD. “Normal”, patients with cardiorespiratory variables within normal limits; FVC, forced vital capacity; FEV₁, forced expiratory volume in 1 sec; MVV, maximum voluntary ventilation. ^†ANOVA.

^†Values represent mean ± SD, Vt, tidal volume; V’_E, minute ventilation; V’O₂, oxygen consumption; RER, respiratory exchange ratio; V’CO₂, carbon dioxide output; HR, heart rate, ^†Two-tailed Student t-test, **Slope and intercept represent the entire curve from resting to peak exercise.

Values represent mean ± SD, ^†ANOVA with adjustment for age, gender and BMI, Vt, tidal volume; V’_E, minute ventilation; V’O₂, oxygen consumption; RER, respiratory exchange ratio; V’CO₂, carbon dioxide output; HR, heart rate, ^†multiple group ANOVA.

n = 10). The mean peak V’_E/V’O₂ and V’_E/V’CO₂ were highest in the cardiovascular cohort (182% and 189% predicted, respectively), but was not statistically significant because only 3 patients comprised this group. The lowest mean O₂-pulse was in the cardiovascular group (47% predicted), 47% of the mean value for the control group. End-tidal PCO₂ (PetCO₂) was lowest in the pulmonary vascular group (compared to other subcohorts, p < 0.05, ANOVA). Figure 1 and Figure 2 highlight differences between V’O₂ and V’_E/V’CO₂ at peak exercise between control subjects and all patients combined.

Six of 10 patients with ventilatory limitation (V’_E/MVV > 70% and normal peak O₂-pulse) exhibited an obstructive pattern on spirometry, with the remainder showing a restrictive deficit. Patients in the pulmonary vascular cohort

(n = 18) typically exhibited normal spirometry or mild restrictive changes. Their mean (±SD) V’O₂max was 61% ± 17% predicted. They exhibited a mean pulse-O₂ intermediate to that of the controls and cardiovascular cohort (74% predicted). Their V’_E/V’CO₂ and V’_E/V’O₂ were 134% and 133% predicted, respectively. Six of these patients had evidence for pulmonary hypertension by echocardiography; 2 had pulmonary thromboembolism confirmed by CT angiography.

Of 31 patients (86%) who exhibited increases in the ventilatory equivalent for CO₂ and O₂ at peak exercise, 16 had adequate ventilatory reserve and no clinical or echocardiographic evidence for left ventricular impairment, indirectly indicating presence of circulatory limitation. The slope of V’_E/V’CO₂ was highest in the ventilatory and both circulatory cohorts; their combined slope was 42% higher than that of the 10 control subjects (p < 0.02). The intercept of V’_E/V’CO₂ tended to be the highest in the control subjects and the single musculoskeletal

patient, and lowest in the cardiovascular group, although not statistically significant because of variability.

A close association between V’O₂max and FVC and/or FEV₁ indicates that impaired respiratory function contributes strongly to exercise limitation. Reduction in spirometric values can also be seen in patients with circulatory limitation [2] [15] [16] [17] [18]. As can be expected, the strongest associations between spirometric and CPET variables occurred amongst the ventilatory and pulmonary vascular groups. For the entire cohort of 36 patients, V’O₂max (ml/min) was positively correlated with FVC (in L) and FEV₁ (in L) [r² = 0.33, p = 0.00024, and r² = 0.38, p = 0.00006, respectively (Figure 3(a) and Figure 3(b))].

There were weaker negative associations of FVC (% pred) with V’_E/V’CO₂ (% pred) and V’_E/V’O₂ (% pred) (r² = 0.16, p = 0.017, and r² = 0.12, p = 0.039,

respectively) at peak exercise. Negative associations of FEV₁ (% pred) with peak V’_E/V’CO₂ (% pred) and peak V’_E/O₂ (% pred), however, were stronger (r² = 0.42, p = 0.00002, and r² = 0.37, p = 0.0001, respectively) at peak exercise, respectively (Figure 4(a) and Figure 4(b)).

In the ventilatory group (n = 10), there was a weak positive correlation between and peak V’_E/V’O₂ and FVC (L) (r² = 0.47, p < 0.05, respectively) at peak exercise. The association between FVC (L) and peak V’_E/V’CO₂ did not quite reach statistical significance (r² = 0.39, p = 0.055). There also were positive associations between FEV₁ (L) and peak V’_E/V’CO₂, and between FEV₁ (L) and peak V’_E/V’O₂ (%) (r² = 0.43, p= 0.039 and r² = 0.68, p = 0.003, respectively) at peak exercise.

Because of larger number of patients (n = 18), the pulmonary vascular group exhibited more robust correlations. In this group, V’O₂ (ml/min) correlated with

FVC (L) (r² = 0.44, p = 0.0003) (Figure 5). There were significant negative associations of FVC (% pred) with peak V’_E/CO₂ (% pred) and peak V’_E/O₂ (% pred) [(r² = 0.44, p = 0.003 and r² = 0.5, p = 0.0002, respectively (Figure 6(a) and Figure 6(b))], and of FEV₁ with both peak ventilatory equivalents [r² = 0.49, p = 0.002 and r² = 0.56, p = 0.0003, respectively (Figure 7(a) and Figure 7(a))].

For all patients combined, we found strong associations between the slope for V’_E/CO₂ and FEV₁, and between the slope and V’O₂max, particularly when expressed as percent predicted values (Table 5). Correlations were less strong between V’_E/V’CO₂ slope and absolute values of variables. Results were similar for

Same abbreviations as in previous tables. *p < 0.05, ANOVA, **p < 0.01, ***p < 0.001.

the pulmonary vascular group (n = 18). Fewer and weaker associations were found in the ventilatory-limited group because of their small numbers (n = 10).

Patients with cancer overall exhibited a mean V’O₂max of 15 mL/kg/min, 23% less than in the control group, similar to the findings of Wernhart and Halle [7] who found a difference of 24% between cancer survivors and healthy control subjects. Beaudry et al. [24] [25] also found a difference of 22% and 29%, respectively, between patients with early stage breast cancer treated with anthracyclines and age-matched healthy women. Cancer patients have impairment of cardiorespiratory fitness due to several factors: weight loss with decrease in muscle mass [25], respiratory impairment from effects of chemo-radiotherapy and tumor infiltration of the lung, cardiovascular limitation from drug effects [24], pericardial effusion and pulmonary vascular occlusion from thrombotic and tumor emboli [26] [27].

Many cancer therapies exhibit adverse cardiovascular effects predispose patients to heart failure both with reduced and preserved ejection fraction. Comorbid cardiovascular risk factors or cancer-related cardiometabolic effects add to the risk for heart failure. Additional research is needed to understand the incidence of heart failure, particularly that with preserved ejection fraction. This becomes even more important with the increasing use of biologic agents, including checkpoint inhibitors, which were not administered to any of our patients.

Within the cohort as a whole, peak V’O₂ was strongly associated with FVC and FEV₁, and amongst the subcohorts, this correlation was most prominent in the pulmonary vascular group (who exhibited a normal or mild restrictive pattern on spirometry) despite their ventilatory reserve being within normal limits. Airflow limitation is often associated with heart failure [16] [17] [18] because of smoking history and perivascular cuffing with edema. The association was less strong in patients with ventilatory impairment, but they constituted a smaller group. The primary cardiovascular group likely would also have shown a similar correlation had they been larger [28]. Considering the Fick equation [where V’O₂ = cardiac output × (arterial − venous oxygen content difference)], patients with impaired cardiac output and/or increased pulmonary vascular resistance at peak exercise have decreased oxygen delivery, while patients with chronic lung disease exhibit hypoxemia due to ventilation-perfusion mismatching and decreased gas transfer, which in turn, increases pulmonary vascular resistance. In addition, anemia contributes to impaired oxygen transport, as oxygen content is dependent on hemoglobin (1.34 gm carried in 100 mL blood).

Ventilatory equivalents for O₂ and CO₂ were increased in ventilatory and both circulatory cohorts. There were also negative correlations between spirometric volumes and ventilatory equivalents, particularly in the pulmonary vascular group. This association was less pronounced in the ventilatory group. Such a relationship has been described in patients with chronic heart failure [29] [30], pulmonary hypertension [31] and most lung diseases [32] [33] exhibit a decrease in ventilatory efficiency in proportion to the reduction in exercise capacity. Gas exchange in the majority of our patients was likely further impaired by the effects of drug, hormonal, biologic and radiation treatments. Impaired cardiac output may be associated with augmented chemoreceptor or peripheral ergoreceptor drive to ventilation leading to overactive efferent muscle nerve activity and exercise hyperventilation [34] [35]. Within the group as a whole we found negative associations of FEV₁ and FVC with V’_E/V’CO₂ and V’_E/O₂ at peak exercise. Patients with COPD and restrictive respiratory disorders exhibit increase in dead space breathing, respectively because of air trapping and increase in rapid shallow breathing, or because of concomitant pulmonary vascular changes [29] [36].

In patients who exhibited increases in ventilatory equivalents at peak exercise (n = 31), at least half had normal ventilatory reserve and no clinical or echocardiographic evidence of left ventricular impairment, indicating presence of pulmonary vascular limitation. Increase in production of reactive oxygen species (ROS) can decrease vasodilatory properties of nitric oxide (NO), resulting in pulmonary vasoconstriction during exercise. Chemotherapy and radiation increase ROS generation with associated endothelial injury, vascular remodeling and increased arterial stiffness [9]. Patients with breast and lung cancer who received radiation therapy exhibit an increase in V’_E/V’CO₂ [37] [38] [39], suggesting effects on the pulmonary vascular circulation, as was the case in many of our patients.

We computed the slope for the V’_E/V’CO₂ ratio to further define the severity of circulatory impairment and to assess its ability to distinguish ventilatory from circulatory limitation [22], using resting and peak exercise values for V’_E and V’CO₂ in the relationship y = a + bx to generate the plots. We found that the slope was approximately 40% higher in the ventilatory and both circulatory groups than in the 10 control subjects. The mean slope at peak exercise for the latter group (28.2) is similar to those reported by Sun et al. [22] in their study of 474 healthy subjects (slope 25) and by Wernhart and Halle in 60 healthy subjects (slope 31.3) [7]. The ratio has been used to indicate the severity of heart failure [36], but as seen in this study, by itself could not differentiate between circulatory and ventilatory impairment, in contrast to others [18] [39]. More recently, the intercept of the V’_E/V’CO₂ slope has been used to further distinguish ventilatory from circulatory limitation [17] [18] [39]. In our patients, the mean intercept tended to be lower for patients with cardiovascular limitation, as reported by others [17] [18] [39] [40]. The intercept was highest in the lone patient with musculoskeletal weakness, likely because of rapid, shallow dead space breathing (respiratory rate 42 breaths/min).

We found that the slope of V’_E/V’CO₂ in all patients combined was variably associated with indices of airflow limitation and peak oxygen uptake and are similar to findings of others [3] [14] [41]. Again, the main mechanism for these findings is the occurrence of dead space breathing in such patients and the increase in oxygen uptake required to overcome the respiratory and pulmonary vascular constraints imposed during exercise. Such changes have been documented in patients with both chronic lung disease and heart failure [7] [24] [28] [34] [39]. The unique aspect of this hypothesis-generating investigation is the first time use of ventilatory equivalents for V’O₂ and V’CO₂ and the V’_E/V’CO₂ slope and its intercept in an attempt to distinguish between a respiratory and circulatory origin for dyspnea in patients with cancer. Larger studies in dyspneic individuals with cancer may help determine if the slope and intercept can further distinguish between primary cardiac and pulmonary vascular sources of impairment.

A study by Yu et al. [42] reinforces our understanding of exercise intolerance as a persistent, drug-related cardiorespiratory impairment in patients treated for breast cancer. The UPBEAT (Understanding and Predicting Breast Cancer Events After Treatment) trial] [43] is designed to evaluate the effect of various chemotherapies on peak oxygen uptake and 6-minute walk exercise capacity in women with breast cancer over time. Its estimated completion date is July 2034. Such trials are additionally likely to support the benefit of exercise programs as part of a rehabilitation program for such patients [44].

There were limitations to this study, most of which relate to its retrospective nature. Most important, the finding of increased dead space ventilation during exercise in most patients might suggest a selection bias against those who might not have exhibited such a finding. Yet we report all patients who underwent CPET for evaluation of dyspnea rendering selection bias less likely. Second, other than CPET and spirometry which were conducted in all patients, additional diagnostic studies that would have been helpful to confirm the presence of structural circulatory changes were not uniformly available. While all patients underwent chest CT scanning (for tumor staging purposes) within a few months of CPET and spirometry, only a few had CT angiography and/or ventilation-perfusion scanning which, respectively, would have demonstrated underlying pulmonary emboli (thrombotic or tumor) or chronic thromboembolic disease. Third, smoking history was available in only 59% of patients. Fourth, we acknowledge the small sample sizes of our cohorts; larger numbers of patients may unmask differences in variables that were not detected in this study. The small number of patients within cohorts also limits the generalizability to all cancer patients; however, the predominant finding of a circulatory source of exercise limitation in most patients is striking. It provides further evidence for the potential adverse effects of therapy on the cardiovascular system. Further systematic studies should provide additional information with regard to specific etiologies for the increase in dead space breathing in such patients.