INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients with advanced, stable congestive heart failure (CHF) often have limited ability to participate in activities of daily living. Exercise limitation in CHF is multifactorial, but recognized contributing factors include impaired cardiac performance (1), abnormal ventilatory responses (2), peripheral muscle dysfunction (5, 6), or any combination thereof. Pharmacological interventions such as vasodilators or inotropes that improve central hemodynamics in patients with stable CHF have inconsistent acute effects on exercise performance (7, 8). These findings support the notion that exercise curtailment in CHF is not determined by cardiac factors alone, but likely depends on complex, integrated cardiopulmonary and locomotor muscle interactions.

It has become clear that patients with CHF may stop exercise because of intolerable exertional symptoms of leg fatigue, discomfort, and/or dyspnea at a point where there is apparent cardiopulmonary reserve (9). The mechanisms of exertional leg discomfort in stable CHF are unknown but are thought ultimately to reflect the net effect of a combination of peripheral muscle weakness and metabolic alterations in the active locomotor muscle. With respect to the latter, poor peripheral blood perfusion, reduced oxygen delivery, reduced capacity for oxidative metabolism, and reduced clearance of CO₂ and active metabolites are all likely to be instrumental (6). The presence of peripheral muscle weakness would dictate that increased neural activation or motor drive would be required to perform a given muscular task (10, 11). It is postulated that in the sensory domain, increased locomotor drive may be directly perceived (via central corollary discharge) as an increased sense of contractile muscle effort (10). In addition, local alterations in the metabolic milieu may activate muscle mechanoreceptors (12) whose afferent inputs may directly or indirectly give rise to unpleasant sensations perceived as leg discomfort (12).

The mechanisms of exertional dyspnea in CHF are also poorly understood, but likely contributing factors include excessive ventilatory requirements (9), restrictive ventilatory mechanics (14), inspiratory muscle weakness (15), or any combination of these factors. All of these conditions necessitate increased central motor command output to respiratory muscles, which may be perceived as increased sense of effort (10), as in the case of peripheral skeletal muscles.

It is reasonable to assume that any intervention that will favorably alter integrated cardiopulmonary function, so as to reduce leg discomfort and dyspnea, should improve exercise endurance. Therefore, this study was designed to determine the effects of ventilatory assistance on exertional symptoms and cardiopulmonary exercise responses in patients with stable CHF. Studies by Naughton and coworkers (16) have shown that the administration of continuous positive airway pressure (CPAP) to patients with stable CHF at rest acutely improved cardiac performance and also reduced the work of breathing. We reasoned that in patients with CHF who are exercising, ventilatory assistance, by similarly unloading the ventilatory muscles and improving cardiac performance, and possibly also by improving peripheral muscle blood flow, could have salutary effects on dyspnea and leg discomfort with consequent improvement in exercise endurance.

Therefore, we compared the effects of CPAP and inspiratory pressure support (PS) on exertional leg discomfort and dyspnea, exercise endurance time, dynamic ventilatory mechanics, breathing pattern, and cardiac responses in symptom-limited patients with stable CHF. A comparison of the effect of PS and CPAP of equal magnitude allowed us to examine the effects of the positive end-expiratory pressure (PEEP) component in isolation in these patients.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Design

This study was a randomized, blinded, controlled trial in which 12 patients with stable CHF participated. After providing informed consent, patients completed a screening visit that included a medical history and clinical assessment, chronic dyspnea evaluation, pulmonary function testing, symptom-limited maximal incremental exercise testing, and optimization of CPAP levels during exercise (see below). During the subsequent experimental visit, patients performed three constant-load exercise tests at approximately 75% of the previously determined maximum work rate, each separated by recovery periods of at least 1 h. In random order, these exercise tests were conducted while breathing under each of the following conditions: (1) optimized CPAP, (2) inspiratory pressure support (PS) of the same magnitude as CPAP, and (3) a control consisting of a minimal assist of 1 cm H₂O CPAP. Exertional symptoms of dyspnea and leg discomfort, and cardioventilatory responses to exercise, were compared during each mode of ventilatory assistance. Mechanisms for improvement in exercise endurance, if any, were also studied.

Subjects

Patients with CHF satisfied the following inclusion criteria: (1) CHF documented for at least 6 mo, (2) echocardiography showing left ventricular ejection fraction (LVEF)  35% within the previous 6 mo, (3) New York Heart Association class II-IV disease of either ischemic or idiopathic etiology, (4) stable disease with no hospitalizations in the previous month. Patients were excluded from the study if they had significant obstructive lung disease (FEV₁/FVC < 80% predicted), unstable angina or significant cardiac arrhythmias, myocardial infarction within the previous 3 mo, primary valvular heart disease, inability to exercise owing to neuromuscular or musculoskeletal disease, or other potential causes of dyspnea or fatigue.

Pulmonary Function

Routine spirometry was performed at rest in the sitting position, in accordance with recognized standards (17), using an automated pulmonary function testing center (6200 Autobox DL; SensorMedics, Yorba Linda, CA). Baseline lung volumes (i.e., functional residual capacity [FRC]) and specific airway resistance (SRaw) were determined by constant-volume body plethysmography (6200 Autobox DL). Total lung capacity (TLC) was calculated as the sum of the FRC and spirometric inspiratory capacity (IC). Single-breath diffusing capacity for carbon monoxide (DL_CO-SB) was also measured (6200 Autobox DL). Maximum inspiratory and expiratory mouth pressures (MIP and MEP, respectively) measured at FRC and TLC, respectively, were assessed by the use of a standard mouthpiece and a direct-reading pressure manometer (Magnehelic; Dwyer Instruments, Michigan City, IN). Predicted normal values for spirometry, lung volumes, DL_CO-SB, and SRaw were those of Morris and associates (18), Goldman and Becklake (19), Gaensler and Wright (20), and Briscoe and Dubois (21), respectively.

Evaluation of Exertional Symptoms

Dyspnea, or breathlessness, was defined as an "unpleasant sensation of labored or difficult breathing." Chronic activity-related dyspnea was evaluated using the modified Baseline Dyspnea Index (22). Exertional dyspnea and leg discomfort were evaluated using the Borg Scale (23) during exercise. Before exercise testing, the Borg Scale was explained and its end points anchored such that "0" indicated no dyspnea or leg discomfort and "10" represented the maximal dyspnea or leg discomfort that the patient could imagine or had ever experienced. Immediately after exercise cessation and the completion of mechanical measurements, subjects were asked the reason(s) for exercise termination (i.e., dyspnea, leg discomfort/fatigue, both, or other).

Application of CPAP and PS

CPAP and PS were applied using a Respironics ventilator (Respironics, Murrysville, PA). To determine whether IC could be reliably used to track end-expiratory lung volume (EELV) during the application of ventilator support, we tested the effects of breathing with 5, 10, and 15 cm H₂O of CPAP on the TLC (FRC + IC) derived from body plethysmography performed while sitting at rest.

During the screening visit, CPAP was carefully adjusted upward in increments of 1 cm H₂O while cycling at the selected level of constant-load exercise to the point of maximum symptom benefit. The applied level of PS was identical to that of the selected CPAP. These "optimized" levels of CPAP and PS were applied during separate constant-load exercise tests on the day of the experimental visit.

Exercise Testing

Exercise tests were conducted on an electronically braked cycle ergometer (Ergometrics 800S; SensorMedics) to a symptom-limited end point. Steady-state resting measurements were taken over a 1-min interval after at least 3 min of quiet breathing through the breathing circuit. For the initial progressive exercise tests, subjects began cycling at a work rate of 10 W for 2 min and thereafter the load was increased in 2-min intervals by increments of 10 to 20 W until a maximum work rate was reached. Subsequent constant-load exercise tests were conducted at a work rate of approximately 75% of the predetermined maximal work rate. Pedalling rates were maintained between 50 and 70 revolutions per minute (rpm) for all tests.

The breathing circuit for experimental exercise tests was designed to minimize pressure fluctuations during the application of CPAP, to avoid the occurrence of leaks during ventilatory assistance, and to minimize the possibility of rebreathing. With nasal passages occluded by a nose clip, all subjects breathed through a mouthpiece attached to a low-resistance two-way nonrebreathing valve (2700 series; Hans Rudolph, Kansas City, MO). The valve setup was adjusted and fixed at a comfortable height for each subject. Inspiratory and expiratory flow rates were measured by pneumotachographs (No. 3; Fleisch, Lausanne, Switzerland) and differential pressure transducers (MP45, ± 2 cm H₂O; Validyne, Northbridge, CA) and tidal volume (VT) was integrated from inspiratory flow. Expiratory air was circulated through a 2.5-L mixing chamber and analyzed for oxygen (FE_O2) (S-3A oxygen analyzer; Applied Electrochemistry, Pasadena, CA) and carbon dioxide content (FE_CO2) (LB-2 gas analyzer; SensorMedics). End-tidal CO₂ (FET_CO2) was analyzed continuously by the LB-2 gas analyzer. Oxygen saturation (Sa_O2) by pulse oximetry (503 pulse oximeter; Criticare Systems, Wakesha, WI) and electrocardiographic monitoring (Cardiovit CS-6/12Z; Schiller, Baar, Switzerland of heart rate (HR) were carried out continuously throughout testing.

All equipment was calibrated immediately before and after each test. Signals were collected using an eight-channel recorder (RS3800; Gould Instruments, Cleveland, OH). Flow, volume, mouth pressure (Pm), esophageal pressure (Pes), FET_CO2, FE_CO2 and FE_O2 were sampled by a computer at a rate of 100 Hz using computer data acquisition software (CODAS; Dataq Instruments, Akron, OH) and stored for later analysis. For each breath, computer software calculated timing (TI, TE, TI/Ttot), flow (VT/TI, VT/TE), and volume (VT) parameters. Minute ventilation (E), oxygen consumption (O₂), and carbon dioxide output (CO₂) were calculated using standard formulas (24). Anaerobic thresholds were assessed using the v-slope method (25). Exercise parameters were compared with the predicted normal values according to Jones (24).

Operational lung volumes. With the assumption that TLC did not change with exercise, measurements of end-expiratory lung volume (EELV) were derived from IC maneuvers performed at rest, every 2- 3 min during exercise, and at peak exercise; thus, changes in IC reflected changes in EELV during exercise. Confirmation of satisfactory technique and reproducibility of IC measurements for each subject was first established during an initial practice session at rest. This method has been previously tested and found to be both repeatable and responsive to change (26).

Breathing pattern analysis. Tidal flow-volume curves at rest, at a standardized time near end-exercise, and at peak exercise were constructed for each patient and placed within their respective maximal flow-volume envelopes according to coinciding IC measurements. Maximal flow-volume loops were performed at rest and immediately after exercise for this analysis. The presence or absence of flow limitation was then determined by comparing tidal expiratory flows with those of the maximal envelope at isovolume: we looked at the shape and limits of the maximal flow-volume curve in the tidal operating range (FEF₇₅, FEF₅₀, FEF₂₅, PEFR), as well as the extent of expiratory flow limitation (percentage of VT that encroached on the maximal flow envelope, the extent of encroachment of flow at the midrange of VT on the maximal envelope at isovolume).

Measurement and analysis of lung mechanics. Esophageal pressure (Pes) was recorded continuously using a balloon-tipped catheter system during exercise testing. A 10-cm latex balloon, containing 0.5 ml of air and connected by a polyethylene catheter to a Validyne differential pressure transducer, was positioned according to an accepted technique (27). At rest and immediately after exercise (i.e., peak), maximum inspiratory maneuvers against an occluded airway at EELV were performed to obtain maximum values for Pes (PI_max). The work of breathing (WOB) was calculated using Campbell diagrams constructed from tidal Pes-volume loops (28). The tension- time index of the inspiratory muscles (TTI) was calculated as the product of mean inspiratory Pes/PI_max and TI/Ttot (29). Finally, an index of neuromechanical coupling was calculated as the ratio of inspiratory Pes/PI_max to standardized tidal volume (VT/IC) (30).

Blood lactate. A venous catheter was inserted into the forearm of each patient and a standard set-up was implemented to maintain patency of the line. Lactate samples were obtained after at least 5 min of steady-state rest, every 2-3 min of exercise, at peak exercise, and 3 min after exercise cessation. Samples were drawn into dry sterile tubes containing 6.0 mg of potassium oxalate and 7.5 mg of sodium fluoride for glycolytic inhibition, were stored on ice during testing, and taken to the hospital chemistry laboratory for analysis immediately after the completion of each exercise session. The plasma was separated by centrifugation and analyzed for lactate using a colorimetric analysis involving the reduction of NAD⁺ to NADH and the oxidation of lactate to pyruvate.

Statistical Analysis

Results are expressed as means ± SEM and a p < 0.05 level of significance was used for all analyses. Relevant exercise response slopes and their intercepts were determined using linear regression analysis of individual data sets. Although summarizing a curve into one data point (such as a slope) usually means that the curvature of the relationship is missed, this method is a simple and reliable way to represent changes in measurements during exercise and is generally appropriate when not trying to "fit" data. We have previously shown that slopes derived by linear regression are reproducible and responsive to change (26). Measurements were also compared at standardized times during constant-load exercise, specifically at the highest equivalent time achieved in all exercise tests by each subject (i.e., isotime). Comparisons were made between measurements taken during each pressure application (CPAP, PS) and control using statistical methods for repeated measures and longitudinal data. Analyses were designed to answer the following questions:

Does ventilatory support improve exercise endurance in CHF? By unloading the respiratory muscles, was dyspnea (perceived leg discomfort/fatigue) relieved? Symptom-limited exercise times were compared during control, CPAP, and PS exercise using an ANOVA for repeated measures. Comparisons were made between measurements of Borg-time slopes and standardized Borg ratings during control, CPAP, and PS tests using ANOVA for repeated measures.

Were improvements in exercise endurance, if any, related to reductions in exertional symptom intensity? To establish an association between the change in exercise endurance with ventilatory assistance and the concurrent change in exertional symptom intensity, simple regression analysis using Pearson correlations was performed with the change in symptom-limited exercise endurance time (T_lim) as the dependent variable and the change in the Borg-time slope as the independent variable. This was carried out for Borg-time slopes of both dyspnea and leg discomfort.

What were the possible mechanisms of improvement in response to ventilatory assistance? To establish associations between the change in exercise endurance during ventilatory assistance and possible contributing factors, simple regression analysis was performed using Tlim as the dependent variable and changes in operational lung volumes (IC, inspiratory reserve volume [IRV], EELV, end-inspiratory lung volume [EILV], VT/IC), breathing pattern (VT, f, VT/TI, VT/TE, TI/Ttot), ventilation (E), ventilatory mechanics (Pes/PI_max, other Pes-derived measurements such as WOB, TTI), an index of neuromechanical coupling (Pes/PI_max:VT/IC ratio), lactate concentrations, indices of expiratory flow limitation (extent of VT overlap), and cardiovascular parameters (HR, oxygen pulse, blood pressure, Sa_O2) as independent variables. Baseline symptomatology, lung function, pulmonary mechanics, breathing pattern, and other relevant factors were analyzed as possible covariates. Stepwise multiple regression analysis was carried out to establish the best predictive equation for T_lim, and the resultant model was reestimated with significant predictors only. Mechanisms of improvement in exertional symptom intensity were evaluated in a similar manner.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Subject characteristics are shown in Table 1. The mean left ventricular ejection fraction (LVEF) was significantly reduced, ranging from 10 to 35%. Medications being taken included angiotensin-converting enzyme inhibitors (n = 10), diuretics (n = 10), digoxin (n = 5), nitrates (n = 3), and -blockers (n = 3). Although all but one subject had a smoking history, only one had smoked within the previous 5 yr, and seven had quit smoking at least 10 yr before the study. Baseline pulmonary function parameters were within normal limits except for a mild reduction in lung volume (FVC), and a reduced expiratory reserve volume (ERV). In addition, maximal flow-volume curves showed progressive scooping of expiratory flows (i.e., reduced FEF₅₀, FEF₇₅).

Subjects had a Baseline Dyspnea Index focal score of 6.8 ± 0.6, indicating a moderate level of chronic activity-related dyspnea. Incremental maximal exercise tests were stopped owing to severe leg discomfort/fatigue (n = 4), dyspnea (n = 2), or a combination of both (n = 6). Exercise capacity was severely curtailed with significantly reduced maximum oxygen consumptions (O₂max) and peak work rates, and modestly increased E/CO₂ ratios (Table 1). Anaerobic thresholds were in the low to normal range in all subjects at a mean O₂ of 39 ± 2% of the predicted O₂max, where normal is between 40 and 60%. All subjects showed expiratory flow limitation at rest, i.e., tidal expiratory flows overlapped the maximum expiratory envelope at isovolume over the majority of VT (see example in Figure 1A). At end-exercise, EELV increased by a mean of 0.26 ± 0.06 L from rest, and EILV reached 92 ± 2% of TLC.

View larger version (17K): [in this window] [in a new window]   Figure 1.   Typical example of a CHF patient (A) Tidal flow-volume loops at rest and at the peak of symptom-limited exercise are shown relative to the maximal envelope. These indicate resting expiratory flow limitation, and dynamic hyperinflation with persisting flow limitation during exercise. (B) Flow-volume loops at isotime during control, CPAP, and PS exercise tests show no differences in EELV or in the extent of expiratory flow limitation.

Optimization of CPAP and PS

CPAP was optimized during cycle exercise at 54 ± 8 W, i.e., 75% of the maximum work rate achieved in incremental testing. A mean "optimal" CPAP of 4.8 ± 0.2 cm H₂O (range, 4-6 cm H₂O) was selected as the level providing the maximum symptom benefit. For each patient, PS of this same level was used during experimental testing.

CPAP between 5 and 15 cm H₂O did not alter measured lung volumes at rest (i.e., TLC, RV, FRC, IC). Therefore, with the assumption that TLC at rest did not change during exercise, we justified using the change in IC to derive the change in EELV as an indicator of dynamic hyperinflation during exercise with ventilator assistance.

Exercise Responses during Ventilatory Assistance

There were no significant carryover effects resulting from performing three constant-load tests in 1 d with respect to (1) exercise slopes of dyspnea, leg discomfort, O₂, CO₂, E, IC, tidal Pes, and lactate expressed over time, or (2) isotime measurements of O₂, CO₂, E, IC, tidal Pes, and lactate, therefore allowing intertest comparisons to be made.

Endurance during cycle exercise using the constant-load protocol was limited primarily by leg discomfort (n = 7), a combination of leg discomfort and dyspnea (n = 4), or by dyspnea (n = 1). Total exercise time increased significantly during exercise with PS (2.8 ± 0.8 min or 43 ± 14%; p = 0.004), but only modestly with CPAP (1.4 ± 0.7 min or 28 ± 15%; p = 0.079) compared with control (Table 2). The majority (n = 8) of subjects increased their exercise endurance in response to PS, whereas only 5 of 12 subjects improved in response to CPAP.

Despite differences in exercise time, peak Borg ratings of both dyspnea and leg discomfort did not change in response to PS or CPAP (Table 2). Slopes of Borg ratings of perceived leg discomfort over time were significantly reduced in response to PS compared with control (p < 0.05), but did not change significantly with CPAP compared with control (Figure 1). Slopes of Borg dyspnea ratings over time were not different between control, PS, or CPAP tests (Figure 2).

View larger version (17K): [in this window] [in a new window]   Figure 2.   Borg ratings of leg discomfort and dyspnea are shown relative to time during constant-load exercise (values are means ± SEM). *p < 0.05, significant difference in slope compared with control.

Cardiopulmonary measurements obtained at peak or end-exercise did not change significantly, other than a reduction in Pes-derived measurements during PS and CPAP indicative of respiratory muscle unloading (Table 2). Comparisons made at isotime for each exercise test are shown in Table 3. At isotime, significant differences were found for O₂ and various Pes- derived measurements: tidal Pes, peak inspiratory Pes, Pes/ PI_max, TTIdyn, and WOB. At this standardized time, WOB per minute was decreased from control by 39 ± 8 and 25 ± 4% during PS and CPAP, respectively. There were no differences between tests with respect to cardiovascular parameters, ventilation, peak expiratory Pes, operational lung volumes, gas exchanges, or venous lactate concentrations. Finally, the extent of expiratory flow limitation was not different between tests at isotime: tidal flow-volume loops overlapped the maximal expiratory envelope over a mean of 20 ± 8, 31 ± 8, and 37 ± 10% of VT during control, PS and CPAP, respectively; with no differences in EELV (see example in Figure 1B).

Correlates of Improvement in Endurance

Increased endurance (T_lim) with both modes of ventilator assistance correlated best with the reduction in the slope of exertional leg discomfort over time (r = 0.71, p < 0.01) (Figure 3). In turn, the reduction in exertional leg discomfort correlated primarily with the reduction in O₂, both expressed as slopes over time (r = 0.61, p < 0.01); while the decrease in O₂/time correlated with the decrease in Pes/time (r = 0.55, p < 0.01). In other words, the reduction in total body O₂ was likely related to the fall in the oxygen cost of breathing.

View larger version (15K): [in this window] [in a new window]   Figure 3.   Changes in exercise endurance (Tlim) with CPAP (squares) and PS (circles) correlated inversely with changes in exertional leg discomfort expressed as Borg-time slopes (r = 0.71, p < 0.01).

Changes in exercise endurance during ventilator assistance were not related to changes in ventilation, breathing pattern, operational lung volumes, the level of respiratory muscle unloading, expiratory pleural pressures, or the extent of expiratory flow limitation.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The novel findings of this study in patients with stable advanced CHF are as follows: (1) PS significantly improved exercise endurance; (2) increased exercise endurance during PS correlated strongly with reduced intensity of exertional leg discomfort; (3) changes in Borg ratings of leg discomfort during PS correlated with changes in both O₂ and tidal esophageal pressure excursions; (4) CPAP of equal magnitude to PS did not consistently improve either perceived leg discomfort or exercise endurance; (5) both PS and CPAP had only modest effects on exertional dyspnea intensity during constant-load submaximal exercise; and (6) patients showed significant increases in end-expiratory lung volumes during exercise.

Mechanisms of Improved Exercise Endurance

The magnitude of improvement in exercise endurance (i.e., > 40%) during PS substantially exceeded previously reported acute effects of pharmacological agents in this population (7, 8). Improvement was explained primarily by altered symptoms and did not correlate directly with alterations in any of the measured cardiopulmonary physiological variables. Severe leg discomfort was the primary exercise-limiting symptom in the majority of study subjects and it was not surprising, therefore, that alleviation of this symptom during PS contributed to improved exercise endurance. We believe that the significant reduction in Borg-time slopes for leg discomfort in our patients was real and not a spurious finding. We have shown that Borg ratings of leg discomfort (and dyspnea) during constant load exercise are reliable, being both highly reproducible and responsive to change when tested in a population of patients with pulmonary disease (26).

The mechanisms of reduced leg discomfort during PS remain speculative but it is reasonable to assume that such improvements ultimately reflect acute metabolic changes within the exercising leg muscles. Several studies have shown profound metabolic and structural abnormalities in the peripheral skeletal muscles of patients with advanced CHF (6, 12). Our patients with dilated cardiomyopathies had significant cardiac impairment as evidenced by the average ejection fraction of only 24% (range, 10-35%), the significantly reduced maximal oxygen consumption (66 ± 5% of predicted maximum), and the reduced anaerobic threshold (39 ± 2% of predicted maximum O₂). One would therefore anticipate that, in these subjects, there were significant peripheral muscle metabolic abnormalities due to poor peripheral perfusion and its deleterious chronic effects on muscle structure and function.

The significant correlations between reduced intensity of exertional leg discomfort, reduced O₂-time slopes, and reduced Pes-time slopes during exercise with ventilator assistance support the contention that alleviation of leg discomfort is ultimately linked to the extent of inspiratory muscle unloading achieved. The small, but significant, reduction in O₂ during pressure support ventilation could be explained by small air leaks at the mouth. However, because we did not observe any leaks (i.e., puffing out of the cheeks, or escape of air at the mouth of any of our study subjects) while breathing during the low levels of pressure applications in this study, because the reduction in O₂ occurred in conjunction with reduction in esophageal pressure measurements including the tension time index (a measurement that reflects the oxygen cost of breathing), and because the changes in O₂ were greater with PS than CPAP, we believe that such changes reflect the effect of inspiratory muscle unloading rather than measurement error as a result of an undetected air leak.

PS could improve peripheral perfusion by improving cardiac output and/or by altering regional vascular distribution. Cardiac performance is probably curtailed during exercise in advanced CHF, even further than at rest, because of greater inspiratory pleural pressure swings required to maintain ventilation in pace with metabolic demands. Such increases in negative inspiratory pleural pressure would increase the left ventricular (LV) systolic transmural pressure gradient of the failing heart, and increase LV afterload. The addition of PS (5 cm H₂O) would predictably reduce this transmural pressure gradient and reduce LV afterload with resultant improvement in peripheral blood flow to the legs. Improved perfusion would result in improved nutrient supply to the exercising muscle and enhanced clearance of CO₂ and waste metabolites, with consequent improved acid-base status. There were no significant changes in any of the measured cardiac parameters (blood pressure, heart rate, oxygen pulse, CO₂) or in blood lactate measurements during PS. However, these findings alone do not preclude improvements in peripheral blood flow in our group, which would require invasive, central, and peripheral hemodynamic measurements to confirm. Similarly, lack of reduction of nonarterialized venous blood lactate concentrations does not accurately reflect local metabolic alterations in the active muscles.

Another possible mechanism of improved leg blood flow involves altered vascular distribution during ventilatory muscle unloading. In this respect, Harms and coworkers (31) have shown that in healthy subjects exercising at supramaximal O₂, when their capacity to increase cardiac output was limited and ventilatory demands were excessive, ventilatory muscle unloading with proportional assist ventilation (PAV) decreased leg vascular resistance (presumable by altering sympathetic tone), increased leg blood flow, and increased leg O₂ (despite reducing total O₂). The authors postulated that in these subjects, where the ability to increase cardiac output was already maximal and where there are two competing groups of active working muscles (i.e., leg locomotor muscles and the diaphragm), blood flow to the peripheral muscles may be compromised because of a diaphragmatic vascular "steal" effect (31). Ventilatory muscle unloading may similarly improve peripheral blood flow and fractional O₂ to the leg muscles of exercising patients with CHF, whose ability to increase cardiac output is greatly diminished at low work rates and whose work of breathing is increased compared with unaffected individuals. However, invasive thermodilution studies would be required to confirm this hypothesis.

The contention that improvement in leg discomfort ultimately reflects an improved metabolic milieu at the active muscle is supported by the findings of a randomized, controlled study on the acute effects of hyperoxia (60%) on exertional leg discomfort in mildly hypoxemic patients with advanced COPD (32). In this study, 60% oxygen, not room air, significantly relieved exertional leg discomfort, and oxygen therapy resulted in delayed lactate accumulation compared with control (32). From a neurophysiologic perspective, improving the ionic and acid-base status of the active muscle could alter afferent sensory input (12), or reduce requirements for neural activation and the attendant sense of contractile muscular effort (10, 11).

Comparison of CPAP and PS Effects

The negative effects of CPAP of equal magnitude to PS on leg discomfort and exercise endurance in CHF are possibly due to the failure to improve cardiac performance to a similar degree, or to less effective ventilatory muscle unloading. While the addition of CPAP (5 to 10 cm H₂O) to healthy individuals at rest is uniformly negative (i.e., reduced venous return and LV preload, and reduced cardiac output) (16), the effect of CPAP in patients with acute or chronic CHF are often positive (16, 33). Naughton and colleagues (16), have shown that brief administrations of 5 to 10 cm H₂O of CPAP to patients with stable chronic CHF resulted in significant respiratory muscle unloading, reduced LV systolic transmural pressure gradients, and reduced LV afterload without compromising the cardiac index. However, responses to CPAP therapy in CHF are not easily predicted and depend on such multiple factors as baseline LV and right ventricular preload and afterload, ventricular contractility, intravascular volume, and LV end-diastolic pressure. The factors that predict the acute effects of CPAP during exercise have not been ascertained. Some of our study subjects (5 of 12) did improve exercise performance and reduced exertional leg discomfort during CPAP, whereas the majority did not benefit. Bradley and co-workers (33), showed that CPAP was beneficial in the acute setting at rest only in patients with high left ventricular end-diastolic pressures. Central hemodynamic variables were probably quite variable within our patient group; we anticipate that improvements in cardiac performance would not occur in patients whose cardiac output was predominantly preload dependent during exercise, or in patients who did not have high capillary wedge pressures (33).

To the extent that reduced leg discomfort and improved endurance depend on the degree of inspiratory muscle unloading, differences in responses between CPAP and PS could reflect deleterious effects of the PEEP component of CPAP. It is noteworthy that the tension-time index (a measure of the oxygen cost of breathing) (34) and the O₂, both measured at a standardized time during exercise, were both significantly reduced with PS compared with CPAP. The inferior responses of CPAP relative to PS could not be explained by differences in ventilation, breathing pattern, or operational lung volumes.

Effect of Ventilatory Assistance on Dyspnea in CHF

Although leg discomfort was the predominant symptom during exercise, our patients also experienced severe dyspnea. During PS, dyspnea did not increase further despite sustained high levels of ventilation (> 50 L/min) for a period of approximately 3 min longer than unassisted control. However, PS did not appreciably affect dyspnea in the earlier phases of the constant-load exercise test. This lack of response may reflect the fact that patients were more focused on the salutary effects of PS on their most prominent exertional symptom of leg discomfort, or alternatively, that the mechanisms of dyspnea in CHF were not manipulated by PS. The fact that effective inspiratory muscle unloading had little effect on dyspnea during the earlier stage of exercise suggests that factors other than mechanical loading (i.e., elastic and resistive loads) contributed predominantly to dyspnea.

In contrast to normal individuals, in whom CPAP (5 cm H₂O) during exercise has been shown to increase perceived breathing effort (35), CPAP did not worsen exertional dyspnea in CHF and likely had a modest effect toward the end of the endurance run. The response of normal subjects to CPAP has previously been studied; subjects had the option of allowing EELV to increase or to defend mean expiratory flows and the EELV by recruiting expiratory muscles (35). Both strategies are disadvantageous and likely give rise to unpleasant respiratory sensations: lung hyperinflation compromises inspiratory muscle function and expiratory muscle recruitment may increase the sense of expiratory effort. In contrast, a CPAP of 5 cm H₂O in our patients with CHF had no significant effects on EELV, EILV, VT/TE, or peak expiratory Pes, therefore suggesting the presence of significant expiratory limitation. In this circumstance, expiratory flow rate is independent of downstream pressure (36). The contention that our patients with CHF were flow limited is supported by the finding that tidal expiratory flow-volume loops met or exceeded the maximal expiratory flow envelope in all patients at rest and during exercise. Maximal expiratory flow curves were concave to the volume axis at low lung volumes, and ERV was uniformly diminished. During exercise, EELV increased significantly by a mean of approximately 0.3 L as a result of expiratory flow limitation. It is possible that the neutral effects of added CPAP on respiratory sensation, particularly expiratory effort, in CHF are a consequence of expiratory flow limitation. We have previously shown that the addition of small increases in positive pressure on expiration during exercise in patients with chronic air flow limitation does not effect sensation compared with healthy normal subjects who experienced increased expiratory effort (35). The mechanisms of expiratory flow limitation in CHF were not determined, but may relate to airway dysfunction as a result of previous smoking history or advanced age, to the effects of obesity (body mass index was elevated at 29 kg/m²), or to airway dysfunction as a result of CHF (bronchial edema or increased reactivity).

In summary, added PS resulted in marked improvement in exercise endurance in CHF. Alleviation of exertional leg discomfort was the main contributor and correlated with the extent of inspiratory muscle unloading. We postulated that PS improved cardiac performance or vascular distribution, thus altering the metabolic milieu of the active muscle. The clinical implication of our findings is that noninvasive PS could potentially be used as an adjunct to cardiac rehabilitation: patients could train at higher work loads, and for longer periods of time, which should translate into enhanced physiological training effects and enhanced quality of life.