A Non Invasive Estimate of Dead Space Ventilation from Exercise Measurements

Rationale During exercise, heart failure patients (HF) show an out-of-proportion ventilation increase, which in patients with COPD is blunted. When HF and COPD coexist, the ventilatory response to exercise is unpredictable. Objectives We evaluated a human model of respiratory impairment in 10 COPD-free HF patients and in 10 healthy subjects, tested with a progressive workload exercise with different added dead space. We hypothesized that increased serial dead space upshifts the VE vs. VCO2 relationship and that the VE-axis intercept might be an index of dead space ventilation. Measurements All participants performed a cardiopulmonary exercise test with 0, 250 and 500 mL of additional dead space. Since DS does not contribute to gas exchange, ventilation relative to dead space is ventilation at VCO2 = 0, i.e. VE-axis intercept. We compared dead space volume, estimated dividing VE-axis intercept by the intercept on respiratory rate axis of the respiratory rate vs. VCO2 relationship with standard method measured DS. Main results In HF, adding dead space increased VE-axis intercept (+0 mL = 4.98±1.63 L; +250 mL = 9.69±2.91 L; +500 mL = 13.26±3.18 L; p<0.001) and upshifted the VE vs.VCO2 relationship, with a minor slope rise (+0 mL = 27±4 L; +250 = 28±5; +500 = 29±4; p<0.05). In healthy, adding dead space increased VE-axis intercept (+0 mL = 4.9±1.4 L; +250 = 9.3±2.4; +500 = 13.1±3.04; p<0.001) without slope changes. Measured and estimated dead space volumes were similar both in HF and healthy subjects. Conclusions VE-axis intercept is related to dead space ventilation and dead space volume can be non-invasively estimated.


Introduction
The behaviour of ventilation during exercise in heart failure (HF) and in chronic obstructive pulmonary disease (COPD) patients may differ, being characterized in the former by an out-ofproportion increase of ventilation (VE), which is greater the greater the HF severity [1] and, in the latter, by a normal or excessive increase of ventilation in mild or moderate COPD and a blunted ventilation increase in severe COPD patients [2][3][4]. The elevated ventilatory response in HF patients seen before lactic acidosis ensues and the carbon dioxide (CO 2 ) [5] generated by the lactate is trivial relative to the rate of metabolic CO 2 production (VCO 2 ) [6,7]. The relationship between VE and VCO 2 is used to evaluate ventilatory efficiency [8]; in HF, as well as in pulmonary arterial hypertension, an increase of the slope of the VE vs. VCO 2 relationship is associated with a poor prognosis [9][10][11][12][13][14][15][16]. In COPD, ventilatory limitation to exercise is defined either as a reduction of ventilatory reserve or as a lowering of inspiratory capacity [17]. In case of severe COPD, the rise of ventilation during exercise is blunted, and consequently the slope of VE vs. VCO 2 relationship is normal or low, being the slope lower the more pronounced the emphysema profile [2].
HF and COPD often coexist with a reported prevalence of COPD in HF patients ranging between 23 and 30% [18] and with a relevant impact on mortality and hospitalization rates [19]. In patients with COPD and HF, the ventilatory response to exercise is poorly predictable. Indeed, HF hyperventilation can be counteracted by the incapacity of increasing tidal volume (VT) and alveolar ventilation, both being distinctive features of VE during exercise in COPD patients [17]. As a result, the slope of VE vs.VCO 2 relationship might be elevated, normal or even low in patients with COPD and HF, regardless of the presence and of the severity of ventilatory inefficiency. Up to now, only few studies have evaluated the ventilatory behaviour during exercise in patients with coexisting HF and COPD, being patients with comorbidities usually excluded from research trials dedicated to HF or COPD [20].
In the present study, we evaluated HF patients and healthy individuals through a progressive workload exercise with different added DS, hoping to mimic at least in part the effects of COPD on ventilation behaviour during exercise. We hypothesized that increased serial DS upshifts the VE vs. VCO 2 relationship and that the VE-axis intercept (VE Yinter ) might be an index of DS ventilation. Indeed, since DS does not contribute to gas exchange, VE relative to DS is VE at VCO 2 = 0, i.e., VE Yint on the VE vs. VCO 2 relationship.

Subjects
Ten HF patients and 10 healthy subjects were enrolled in the present study.
HF patients were regularly followed-up at our HF unit. Study inclusion criteria for HF patients were New York Heart Association functional classes (NYHA) I to III, echocardiographic evidence of reduced left ventricular systolic function (left ventricular ejection fraction #40%), optimized and individually tailored drug treatment, stable clinical conditions for at least 2 months, capability/willingness to perform a maximal or near maximal cardiopulmonary exercise test (CPET). Patients were excluded if they had obstructive and/or restrictive lung disease (forced expiratory volume in first second/forced vital capacity ratio (FEV 1 /FVC) ,0.70% and/or lung vital capacity (VC) ,80% of predicted value [21]), clinical history and/or documentation of pulmonary embolism, primary valvular heart disease, pulmonary artery hypertension, pericardial disease, exercise-induced angina, ST changes, severe arrhythmias and significant cerebrovascular, renal, hepatic and haematological disease.
A group of age matched healthy subjects was recruited among the hospital staff and from the local community through personal contacts. Inclusion criteria were absence of history and/or clinical evidence of any cardiovascular or pulmonary or systemic disease contraindicating the test or modifying the functional response to exercise, any condition requiring daily medications, and the inability to adequately perform the procedures required by the protocol. No subjects were involved in physical activities other than recreational.
The investigation was approved by the local ethics committee (''Ethics committee Centro Cardiologico Fondazione Monzino'', Institutional Review Board no. S186/311) and all participants signed a written informed consent before enrolling in the study.

Study protocol
At enrolment, demographical and clinical data were collected, lung function measurements and echocardiographic evaluation were performed to verify that the subjects screened met the study inclusion/exclusion criteria, and the informed consent was obtained.
Spirometry (Vmax 29C, SensorMedics, Yorba Linda, CA, US) was performed by all participants in accordance with the recommended technique [22], and measurements were standardized as percentages of predicted normal values [23].
To become familiar with the procedure, both HF patients and healthy subjects had been previously trained to perform an exercise test in our laboratory [24]. Thereafter, on different days, following a random order, exercise testing was done with additional DS equal to 0 mL, 250 mL and 500 mL.
All participants underwent incremental CPET on an electronically braked cycle-ergometer (Ergometrics-800, SensorMedics, Yorba Linda, CA, US) using a personalized ramp protocol that was chosen aiming at a test duration of 1062 minutes. The exercise was preceded by 5 minutes of rest gas exchange monitoring and by a 3-minute unloaded warm-up. A 12-lead ECG, blood pressure and heart rate were also recorded, and arterial oxygen saturation was monitored through a pulse oxymeter. The participants wore a nose clip and breathed through a mouthpiece connected to a mass flowmeter (Vmax 29C, SensorMedics, Yorba Linda, CA, US). Subjects were asked to cycle at a pedalling rate of 60-70 rpm, and CPET were selfterminated by the subjects when they claimed that maximal effort had been achieved. Oxygen consumption (VO 2 ), VCO 2 and VE were measured breath by breath with flowmeter and respiratory gas sampling lines at the end of the added DS. They were averaged every 20 seconds. Anaerobic threshold (AT) was calculated with the standard technique [25]. All tests were executed and evaluated by 2 expert readers.
In the absence of psychogenic hyperventilation, below the respiratory compensation point [26], the relation between VE and VCO 2 is characterized by a linear relationship (VE = aVCO 2 + b), with ''a'' as the slope and ''b'' as the intercept on the VE axis (VE Yint ) [8]. Since DS does not contribute to gas exchange, it is possible to hypothesize that the ventilation relative to DS is similar or related to the VE at VCO 2 = 0, which is the Y intercept of VE vs. VCO 2 relationship. To calculate DS volume (VD) from VE Yint (VD Yint ), we need to identify the corresponding respiratory rate (RR). This was obtained as the intercept of the RR vs. VCO 2 relationship on the RR axis (RR Yint ). Specifically, the RR vs. VCO 2 relationship was calculated through its linear portion that starts from the beginning of exercise and ends when RR increases more steeply, which corresponds to the tidal volume inflection/ plateau [27,28]. An example on how we calculate VE Yint and RR Yint is reported in figure 1.
We compared estimated VD values (VD Yint ) with resting and exercise values of VD, measured with standard method [8] (VD meas ), in the 3 experimental conditions, with 0 mL, 250 mL and 500 mL of added DS. The volume of mouthpiece and flowmeter (50 mL) was subtracted from VD. The standard calculation of VD [8] (VD meas ) is obtained by the following equation: with 863 as a constant and P a CO 2 as pressure for arterial CO 2 .
In healthy individuals [29], but not in HF patients [30], P a CO 2 can be reliably estimated from end-tidal expiratory pressure for CO 2 (P ET CO 2 ). Therefore, we measured P a CO 2 from arterial gas sampling in HF patients, and we estimated P a CO 2 from P ET CO 2 in healthy subjects. Thus, only in HF patients, a small catheter was introduced into a radial artery, blood samples were obtained at rest and every 2 minutes during exercise, and P a CO 2 was determined with a pH/blood gas analyzer (GEM 4000, Instrumentation Laboratory, Bedford, MA, US).
We calculated possible VD changes during exercise, and we evaluated whether an added DS modifies the slope of the VE vs. VCO 2 relationship and/or it simply upshifts it.

Statistical analysis
Data are mean 6 standard deviation (SD). Cardiopulmonary measurements were collected breath by breath and reported as average over 20 s. Comparisons between the two groups were done through unpaired t-test. Both in HF and in healthy subjects, analysis of variance for repeated measures with Bonferroni post hoc test was performed to analyze the effect of the adding of different DS and to evaluate the changes of VD meas during exercise in the 3 experimental conditions. Bland and Altman relationship was calculated to compare VD Yint values and VD meas values in HF patients and in healthy individuals.
Statistical significance was set at p,0.05. All statistics were performed with IBM SPSS statistics 20.0 for windows.

Results
We enrolled 10 HF patients (9 males; mean age 61613 years) and 10 age-matched healthy subjects (8 males; mean age 59610 years). The main anthropometric data were not significantly different between the two groups. Patients with HF and healthy subjects were free from obstructive defects; although within the predicted normal limits, lung volumes tended to be smaller in HF patients than in normal subjects (table 1).

HF patients
Mean left ventricle ejection fraction was 3365%. The cause of HF was ischemic dilated cardiomyopathy in 4 cases and primary dilated cardiomyopathy in 6 cases. Three patients had an implantable cardioverter defibrillator; 9 were in sinus rhythm and 1 was in permanent atrial fibrillation. Four patients were in NYHA class I, 5 in NYHA class II and 1 in NYHA class III. All HF patients were on b-blockers, 9 with angiotensin-converting enzyme inhibitors, 4 with aldosterone receptor antagonists, 5 with diuretics and 3 with amiodarone.
All HF patients performed CPET without added DS and with 250 mL and 500 mL of additional DS without complications. In the HF group, peak VO 2 was slightly reduced compared to healthy subjects. With the exception of reduced peak workload and of an increased VT, the adding of different DS did not significantly impact on CPET data at peak of exercise and on VO 2 at AT (table 2). In table 3 VE, RR, VT, VD/VT, VCO 2 , P ET CO 2 and P a CO 2 during exercise are reported with 0, 250 and 500 mL of added DS.
Values of VE Yint , RR Yint , VD Yint , VD meas and the slope of VE vs VCO 2 relationship in HF patients with 0 mL, 250 mL and 500 mL of additional DS are reported in table 4.
With the adding of DS, the VE Yint increased significantly, whereas RR Yint showed a limited increase. Adding DS upshifted the VE vs. VCO 2 relationship with a minor slope increase (figure 2).
The calculated VD Yint rose as added DS increased; mean VD Yint increase with 250 and 500 mL of added space was 2266127 mL and 4466123 mL. VD meas increased during exercise in the 3 conditions albeit only as a trend when DS was not added (table 5). Figure 3 reports the Bland and Altman plot of VD Yint vs. VD meas at rest for HF patients in the 3 exercise conditions. As an average, a good agreement was observed when VD was calculated either by VE Yint , or VD meas , with or without additional DS.

Healthy subjects
Healthy subjects performed all CPET without complications. Peak exercise data and VO 2 at AT were not significantly affected by the adding of DS (table 2).
When DS was added, the value of the slope of VE vs. VCO 2 relationship and RR Yint did not change, whereas only the VE Yint increased significantly (table 4) with an upshift of the relationship (figure 4). Similarly to HF patients, VD Yint increased with added DS in the three experimental conditions, specifically by 3006150 mL and by 5706160 mL with 250 and 500 mL, respectively.
During exercise, VD meas remained constant without additional DS, whereas it significantly decreased during exercise with added DS, but this finding is likely due to the underestimation of P a CO 2 by P ET CO 2 with added DS (table 5). Figure 5 reports the Bland and Altman plot of VD Yint vs. VD meas at rest for healthy subjects and showed a good correlation between the two methods both with and without additional DS.

Discussion
In the present study, we evaluated a human model of increased dead space in HF patients and in healthy subjects, applying a progressive workload exercise with different added DS. We documented that a rise in serial DS, mimicking a rise in anatomical DS, was parallel to the VE Yint increase both in healthy individuals and in HF patients. Therefore, VE Yint is related to DS ventilation. Moreover, we showed that the value of DS can be non-invasively estimated as the ratio of VE Yint /RR Yint .
Few study limitations should be discussed at first. Firstly, our research was undertaken to analyze the role on ventilation behaviour during exercise of a respiratory comorbidity, COPD, in HF patients. We built a COPD model by adding an external dead space. We recognize that our model is only a partial COPD  primitive chemoreceptor abnormalities as drivers of the alveolar hypoventilation observed in COPD patients. Thirdly, with the Yintercept we analyze an index of overall DS. However, in the present setting, we were able to change DS only by adding an external (anatomical equivalent) DS, so that we do not know if changes in physiological DS similarly influence the VE Yint . Fourthly, VE changes during exercise are due to VCO 2 , VD/ VT and P a CO 2 changes, and all may influence the VE vs. VCO 2 relationship. In the present study, we added external DS, which at each step of exercise, was associated to an increase of VD/VT and P a CO 2 (the latter in 2 steps only as a trend) resembling what happens during exercise in COPD patients (table 3). Therefore both P a CO 2 and VD/VT changes have likely a role in the VE vs. VCO 2 relationship changes we observed after adding DS. It is recognized that P a CO 2 measurements were done only in HF patients and not in healthy subjects, but a different behaviour in healthy subjects is unlikely. Fifthly, the condition of VE at CO 2 production equal 0, as such at the VE Yint of the VE vs. VCO 2 relationship, is a mathematical extrapolation with no physiological meaning. Moreover, absolute DS changes during exercise, so that also the VE Yint value is likely close but different from the rest value. Indeed, we showed that VD tended to increase in HF patients and to reduce in healthy subjects during exercise without added DS. However, we suggest using VE Yint as a tool to evaluate the presence of an increased DS, regardless of its physiological meaning with respect to rest and exercise. The adding of DS significantly reduced the external work produced in HF patients, while a not significant reduction was observed in normal subjects. Peak VO 2 remained unchanged in both groups after adding DS; this finding suggests that added DS was associated to an increased work of breathing which, as a percentage of total work, seems to be greater in HF patients than in normal subjects.    We measured DS during exercise using a standard formula [8] in HF patients. To avoid systemic artery catheterization, we estimated P a CO 2 from P ET CO 2 in healthy subjects, which is a accepted method in the absence of lung disease [29]. It is recognized, however, that albeit largely used in the clinical setting, extrapolation of P a CO 2 from P ET CO 2 even in normal individual is approximate and likely to cause some of the variability observed ( figure 5). Moreover, the values obtained in normal subjects with added DS showed a progressive and unrealistic DS reduction. This is due to a P a CO 2 underestimation by P ET CO 2 when adding DS, confirming the need to directly measure P a CO 2 during exercise for DS evaluation [30]. The low P ET CO 2 compared to P a CO 2 observed during exercise with added dead space (Table 3) is likely due to the rapid rise of PCO 2 during exhalation, which does not reach a plateau.
Adding DS increased the slope of VE vs. VCO 2 relationship in HF patients but not in control subjects. This is different from what happens in patients with severe COPD who show a high VE/ VCO 2 ratio at the beginning of exercise but a blunted VE increase during exercise, so that the slope of VE vs. VCO 2 relationship is normal or low [2]. In our model, the DS increase was too modest to generate a ventilatory limitation to exercise, being the ventilatory reserve at peak exercise always preserved. Accordingly, in HF patients, but not in healthy subjects, we observed a minor exercise performance reduction with the adding of DS.
The VE vs. VCO 2 relationship is frequently used as a prognostic tool in HF patients [9][10][11][12][13]. Some laboratories prefer Table 5. Values of volume of dead space at rest and during exercise in heart failure patients and healthy subjects with no additional dead space and with 250 mL and 500 mL of additional dead space.  to analyze the ratio of the relationship [31], others the slope [32]. However, the ratio varies during exercise, so that which exercise VE/VCO 2 ratio value should be considered is still a matter of debate [31]. Moreover, while the behaviour of VE/VCO 2 ratio during exercise is well described in normal and HF individuals [31], its behaviour in COPD or in patients with HF and COPD is less characteristic and not used as a diagnostics/prognostic tool. To avoid the above-mentioned uncertainties, many authors prefer to study the VE vs. VCO 2 relationship throughout the exercise [33] or up to the respiratory compensation point [8]. To do so, the slope of the VE vs. VCO 2 relationship is calculated, but no attention is dedicated to the intercept of this relationship on the VE axis. However, the increase of the slope of VE vs. VCO 2 relationship may be blunted when COPD is associated to HF [2]. Notably, the presence of COPD in HF may be difficult to be defined because some lung impairment is typical of HF and particularly in more advanced cases regardless of COPD [5]. In the present study, we showed that a DS increase is parallel to the VE Yint increase, so that its value should be taken into account when analyzing the VE vs. VCO 2 relationship. Indeed, VE Yint differences were observed even by adding a relatively small DS (250 mL), which corresponded to 1/10 of peak VT in healthy subjects. It is recognized, however, that whilst the means of estimated and measured VD are similar, the individual values differ up to 60% in case of no added DS and up to ,20% when 500 mL DS were added. This suggests caution when analyzing specific individual data, particularly in the presence of no or modest lung disease.
In the present study, we added 250 mL and 500 mL of DS during exercise. To confirm that VE Yint increase was related to DS increase, we calculated VD Yint . To do so, we need to divide VE by RR, but the value of RR to be chosen is an open question. We used the intercept of the RR vs. VCO 2 relationship on the RR axis because this is the RR value corresponding to VE Yint . Interestingly, the changes of VD Yint values with added DS were very similar to the amount of added DS.
In conclusion, we provide the rational basis for the assessment of VE Yint during exercise as a tool to evaluate DS. Further studies are needed to confirm and to analyze the clinical meaning of the present observation.  volume calculated as VE Yint /RR Yint (VD Yint ) and measured DS volume (VD meas ) at rest, calculated as (1-863/P a CO 2 (VE/ VCO 2 )*VT) with P a CO 2 for healthy subjects with 0 mL (diamonds), 250 mL (circles) and 500 mL (crosses) of additional DS. The grey line identifies the mean difference of VD meas -VD Yint; the black lines identify the mean difference of VD meas -and VD Yint 61.96*standard deviation. P a CO 2 was estimated from P ET CO 2 . P a CO 2 = carbon dioxide pressure; P ET CO 2 = tele-expiratory carbon dioxide pressure; VE = ventilation; VT = tidal volume. doi:10.1371/journal.pone.0087395.g005