Noninvasive Cardiac Output Measurement Using Partial Carbon Dioxide Rebreathing Is Less Accurate at Settings of Reduced Minute Ventilation and when Spontaneous Breathing Is Present

The study was approved by the ethics committee of the National Cardiovascular Center (Osaka, Japan), and written informed consent was obtained from each patient.

Twenty-five adult patients aged 19–75 yr (median age, 63 yr) who had undergone cardiac surgery were enrolled in this study (table 1). They were consecutively admitted patients whose cases matched the following criteria: insertion of a pulmonary artery catheter, stable hemodynamics in the intensive care unit, and no leakage around the endotracheal tube. We excluded candidates who had central nervous system disorders, might be adversely affected by induced hypercapnia, or demonstrated severe tricuspid regurgitation. 4Arterial blood pressure, heart rate, pulmonary artery pressure, central venous pressure, and pulse oximeter signal (PM-1000; Nellcor Inc., Hayward, CA) were continuously monitored in all patients. After waiting 1–2 h for hemodynamics to stabilize after surgery, we started the measurements. First, using an inspiratory-hold technique, 5we measured the effective static compliance and resistance of the respiratory system.

We measured CO using two methods: thermodilution (CO^TD) and noninvasive partial carbon dioxide rebreathing technique (CO^NI). CO^TDwas obtained using a 7.5-French pulmonary artery catheter (Abbott Laboratories, North Chicago, IL). Injection of 10 ml cold saline (0°C) was performed in triplicate, and the values were averaged. We standardized the timing of bolus injection after the first half of the expiratory phase. 6CO^NImeasurement was performed with the partial carbon dioxide rebreathing technique (NICO²software version 3.1, fast mode; Novametrix Medical Systems Inc., Wallingford, CT). This procedure has been described in detail elsewhere. 1,2Briefly, V̇co²is calculated on a breath-by-breath basis, and the differential Fick equation is applied to establish the relation between V̇co²and CO as follows:

where Cv̄co²represents the carbon dioxide content in mixed venous blood, and Caco²represents the carbon dioxide content in arterial blood. In the NICO²system of the current version, carbon dioxide rebreathing is performed for 50 s every 3 min. Assuming that CO remains constant during the carbon dioxide rebreathing procedure, the following equation is substituted for the previous one:

where ΔV̇co²is the change in V̇co²between normal breathing and carbon dioxide rebreathing, ΔCv̄co²is the change in mixed venous carbon dioxide content, and ΔCaco²is the change in arterial carbon dioxide content. Then, assuming that Cv̄co²also remains constant during the carbon dioxide rebreathing procedure, the following equation is introduced:

When end-capillary content (Ccco²) is used in place of Caco², pulmonary capillary blood flow (PCBF), the blood flow that participates in alveolar gas exchange, is measured rather than CO, and the following equation is plotted:

Assuming here that ΔCcco²is proportional to changes in Petco², the following equation can be plotted:

where ΔPetco²is the change in Petco²between normal breathing and carbon dioxide rebreathing, and S is the slope of the carbon dioxide dissociation curve from hemoglobin. CO is the sum of PCBF and intrapulmonary shunt flow (Q̇^S); then, CO is expressed in the following equation:

where Q̇^S/Q̇^Tis the intrapulmonary shunt fraction. The noninvasive method for estimating shunt fraction in the NICO²system is adapted from Nunn's iso-shunt plots, which are a series of continuous curves indicating the relation between arterial oxygen pressure (Pao²) and Fio²for different levels of shunt. 3Pao²is a function of arterial blood oxygen saturation (Sao²), which is noninvasively determined using the pulse oximeter signal. Before the start of the study protocol, the NICO²system was calibrated for zero CO². We entered the results of Pao², arterial carbon dioxide pressure (Paco²), Fio², and hemoglobin concentration into the machine when each patient was undergoing the baseline ventilation.

After admission to the intensive care unit, each patient was ventilated with an 8400 STi  ventilator (Bird Corp., Palm Springs, CA). Initial ventilatory settings were as follows: SIMV, volume-controlled ventilation, inspired V^Tof 10 ml/kg, respiratory rate of 10 breaths/min, inspiratory time of 1.0 s, PEEP of 4 cm H²O, and pressure support of 10 cm H²O. The Fio²settings were adjusted by attending physicians to maintain Pao²greater than 100 mmHg. With the patients maintained in the supine position, sedated with continuous intravenous injection of propofol (2 to 3 mg · kg⁻¹· h⁻¹), we started the measurements.

We performed the two protocols serially. In the first protocol, to prevent spontaneous breathing, if needed, we administered bolus vecuronium bromide (4–8 mg). We applied three settings of volume-controlled ventilation in random order as follows: (1) inspired V^Tof 12 ml/kg and respiratory rate of 10 breaths/min; (2) V^Tof 6 ml/kg and respiratory rate of 20 breaths/min; and (3) V^Tof 6 ml/kg and respiratory rate of 10 breaths/min. The first and second settings should result in identical V̇^E, and the last setting should result in half the V̇^Evalue. At each setting, the rebreathing loop was size adjusted according to the manufacturer's instructions recommended for a V^Tsetting of 12 ml/kg. PEEP and Fio²identical to baseline were used throughout the measurement period. After establishing steady state conditions (approximately 15 min) and confirming stable values of CO^NI(< 5% change in the successive readings), we measured both CO^TDand CO^NI. The values of expired V^Tand V̇^Ewere recorded from the digital display of the ventilator. Arterial blood samples were analyzed with a calibrated blood gas analyzer (ABL 505; Radiometer, Copenhagen, Denmark). Hemodynamic data were also recorded. Dead space fraction (V^D/V^T) and venous admixture fraction (Q̇^S/Q̇^T) were calculated as described elsewhere. 4 

In the second protocol, we examined the measurement of CO when there was spontaneous breathing effort. We stopped the infusion of vecuronium and decreased the propofol infusion rate to 0.5 mg · kg⁻¹· h⁻¹. When the patient recovered spontaneous breathing and satisfied our extubation criteria (recovery of cough reflex; V^T≥ 8 ml/kg and respiratory rate ≤ 20 breaths/min under pressure support of 10 cm H²O; arterial blood gas of pH, 7.35–7.45; Paco², 35–45 mmHg; and Pao²≥ 100 mmHg at Fio²≤ 0.5), we started the measurements. In random order, we applied three settings of partial ventilatory support: (1) SIMV plus PSV, mandatory breath rate of 5 breaths/min, mandatory V^Tof 12 ml/kg, and 6 cm H²O of pressure support; (2) continuous positive airway pressure plus PSV, 10 cm H²O of pressure support; and (3) the same setting of PSV with the shortest length of rebreathing loop. In the first and second settings, the rebreathing loop was sized according to the manufacturer's instructions recommended for a V^Tsetting of 12 ml/kg; at the other setting, the loop was fully retracted (150 ml). After establishing steady state, we measured both CO^TDand CO^NI. Because V^T, respiratory rate, and V̇^Eincreased according to the stimulus of carbon dioxide rebreathing, we recorded V^T, respiratory rate, and V̇^Eat the end of the normal breathing period and at the end of the carbon dioxide rebreathing period. We limited ourselves to performing a single measurement for each ventilatory setting per patient.

Data are presented as mean ± SD. Using analysis of variance with repeated measures, mean values were compared across different settings. When significance was observed, the mean values were tested by multiple comparison with the Bonferroni correction. We evaluated the correlation between CO^NIand CO^TDwith linear regression and Bland-Altman analysis. 7Statistical significance was set at P < 0.05.

Patients had respiratory system compliance of 48.4 ± 8.7 ml/cm H²O and resistance of 9.6 ± 2.8 cm H²O · s · l⁻¹. All patients safely underwent all the measurements and were extubated within 1 h after the measurement protocol.

Table 2shows respiratory and hemodynamic results during controlled mechanical ventilation. Although V̇^Ewas identical, there was higher Paco²and V^D/V^Tat the ventilatory setting of V^Tof 6 ml/kg and respiratory rate of 20 breaths/min than at V^Tof 12 ml/kg and respiratory rate of 10 breaths/min. When V̇^Ewas set to a smaller value (V^Tof 6 ml/kg and respiratory rate of 10 breaths/min), Paco²and Petco²were significantly higher, and CO^NIand V̇co²were significantly lower, compared with the other two settings that provided twice as much V̇^E. The values of CO^TD, a ratio of Pao²to Fio², and Q̇^S/Q̇^Tdid not differ significantly at any of ventilatory settings.

The results of Bland-Altman analysis and regression analysis are summarized in table 3and figures 1 and 2. CO^NIcorrelated moderately with CO^TDwhen V̇^Ewas high, regardless of the V^T(figs. 1 and 2). However, when V̇^Ewas set at half, CO^NIunderestimated CO (y = 0.70x; bias, −1.73 l/min) and the correlation coefficient (R) was small (0.34;table 3). Analysis of the results obtained at the setting of 6 ml/kg V^Tand 20 breaths/min respiratory rate showed bias values and slope of linear regression between those of the other two ventilatory settings.

Table 4shows respiratory and hemodynamic results when patients had spontaneous breaths. At the phase of normal breathing, all of the three ventilatory settings showed similar V̇^E(mean value of 0.11 l · min⁻¹· kg⁻¹) and respiratory rate (12.3–12.4 breaths/min). When the ventilatory mode was PSV, V^Tat the phase of normal breathing was similar for different sizes of rebreathing loop. When PSV was applied with the rebreathing loop adjusted for 12 ml/kg V^T, at the phase of carbon dioxide rebreathing, V̇^Eincreased by 46%, V^Tincreased by 20%, and respiratory rate increased by 30%. Similarly, when SIMV–PSV was applied with the same size of rebreathing loop, at the phase of carbon dioxide rebreathing, V̇^Eincreased by 28%, V^Tof spontaneous breaths increased by 23%, and total respiratory rate increased by 28%. By contrast, when the shortest rebreathing loop was used with PSV, carbon dioxide rebreathing caused smaller increases in V̇^E(+17%), V^T(+10%), and respiratory rate (+10%). There were no significant differences at the three ventilatory settings in blood gas analysis data, V̇co², Petco², V^D/V^T, and Q̇^S/Q̇^T.

The results of Bland-Altman analysis and regression analysis when spontaneous breaths were present are summarized in table 5and figures 3 and 4. During SIMV–PSV mode, the correlation between the CO^NIand CO^TDwas poor (precision, 1.41 l/min and R = 0.23). When PSV was applied with the rebreathing loop adjusted for 12 ml/kg V^T, the correlation was moderate (precision, 1.26 l/min and R = 0.75). When the shortest rebreathing loop was used during PSV, CO^NIoverestimated CO^TD(bias, 1.2 l/min and slope = 1.19) with large precision (1.80 l/min).

The main findings of this study are as follows. (1) Rather than small V^T, low V̇^Eled to less accuracy of CO^NI. (2) When spontaneous breathing effort was present, CO^NIwas less accurate than during controlled mechanical ventilation. (3) During carbon dioxide rebreathing, spontaneous breathing V^Tand respiratory rate increased. (4) Shortening the rebreathing loop reduced the accuracy of CO^NI, although causing less increase in V^Tand respiratory rate during carbon dioxide rebreathing.

We had previously shown that, when V^Tis constant during controlled mechanical ventilation, CO^NIcorrelates well with CO^TD, regardless of inspired oxygen fraction, PEEP, or whether ventilation was pressure or volume controlled. 4At constant respiratory rate, however, reduced V^Tresults in an underestimation of CO^NI, and the reason for this discrepancy remained to be clarified. We designed the first part of this study to investigate whether V^Tor V̇^Eis the dominant factor for the accuracy of CO^NI. For V^T, when V̇^Eof volume-controlled ventilation was set to maintain normocapnia, a correlation between CO^NIand CO^TDwas clinically acceptable (bias < 1 l/min, precision ≤ 1 l/min), whether V^Twas large or small (table 3). The percentage error, which was calculated from the precision divided by the mean CO value, was also acceptable (18% for large V^Tsetting, 13% for small V^Tsetting) because acceptable range is reported to be less than 20%. 8By contrast, when V̇^Ewas reduced to half, CO^NIunderestimated CO^TDwith worse precision (1.27 l/min) and percentage error (29%). These findings clearly indicate that V̇^Eis more important than V^Tfor CO^NIaccuracy. The NICO²system, by using the following equation, assumes that Cv̄co²is constant during the measurement period:

Cv̄co²may increase during carbon dioxide rebreathing, however, when V^D/V^Tis large and alveolar ventilation is low. When using the above equation, the neglecting of ΔCv̄co²could lead to an underestimation of CO. At the end of the 50-s rebreathing period in the NICO²system, mixed venous Pco²was reported to increase by 0.53 mmHg (median) or by 2.5% (average) from the initial value. 9,10Even at identical V̇^E, we observed that CO^NIunderestimated CO at a high respiratory rate (20 breaths/min) and small V^T, compared to ventilation at a low respiratory rate (10 breaths/min) and large V^T(table 3). We speculate that increased V^D/V^Tand decreased alveolar V̇^Eat high respiratory rate leads to this inaccuracy and that a change in V^Tcan also affect CO^NIaccuracy by this mechanism. It is clear that controlled mechanical ventilation with constant V̇^Eand constant V^Tprovides more reliable CO^NImeasurement.

There have been few clinical reports on the accuracy of the NICO²system when spontaneous breathing is allowed. It is now common for patients receiving intensive care to be ventilated with modes that allow some spontaneous breathing. Consequently, we need to confirm whether the NICO²technique provides effective monitoring when spontaneous breathing is present, such as during mixed ventilation consisting of spontaneous breaths and mandatory ventilation. Although several reports have compared CO^NIwith continuous CO^TDmeasurement under mixed ventilation, actual ventilatory settings were not specified, 11–13and modified algorithms were used. 11,12Meanwhile, using a system different from the NICO², Gama de Abreu et al.  14have reported that the intraindividual variability of CO measured by partial carbon dioxide rebreathing technique was significantly larger during irregular spontaneous breathing than when respiratory rate and V^Twere fixed. In this study, when spontaneous breathing effort was present, precision (1.26–1.80 l/min) and percentage error (20–30%) were large, indicating less accuracy of the NICO²system. 8The exact reason for this inaccuracy remains unknown, but there are several plausible reasons.

First, under the influence of spontaneous breathing, V̇^Emay both drift by time and increase in response to carbon dioxide rebreathing. These changes in V̇^Emay foul the assumption of constant Cv̄co²and affect accuracy of the NICO²system. 9,10Second, it may be possible that the stimulus of carbon dioxide rebreathing increases CO in spontaneously breathing patients. Because only minimal dosage of propofol (0.5 mg · kg⁻¹· h⁻¹) was used in our experiment, it is likely that the sympathetic nerve of the patient, as well as the respiratory center, is stimulated during carbon dioxide rebreathing. Third, when there is spontaneous breathing effort, V^Tchanges breath by breath, which may affect accurate measurement of V̇co²or Petco².

Figure 5shows representative V̇co²and Petco²traces from a patient during controlled mechanical ventilation, SIMV–PSV, and PSV. During controlled mechanical ventilation, both V̇co²per breath and Petco²produced a stable plateau during normal breathing and carbon dioxide rebreathing. On the other hand, during SIMV–PSV, the V̇co²per breath changed drastically on a breath-by-breath basis because of the variation in V^Tbetween mandatory breaths (830 ml) and spontaneous breaths (540–620 ml). In the NICO²system, values for V̇co²and Petco²during 60-s baseline normal ventilation were calculated as the average of samples of 33–60 s, and those during the 50-s rebreathing period were calculated for intervals of 25–50 s. 3Because CO^NIis derived from changes in V̇co²and Petco², during SIMV–PSV, the presence of marked breath-by-breath changes in V̇co²and Petco²may affect CO^NIaccuracy. During PSV, breath-by-breath changes in V̇co²or Petco²are smaller than during SIMV–PSV. Even so, neither V̇co²nor Petco²produced a steady plateau during normal breathing or rebreathing periods, probably because V̇^Echanged under the influence of carbon dioxide rebreathing. Worse precision during SIMV–PSV and PSV supports the speculation that irregular spontaneous breathing affects the accuracy of the current version of the NICO²system. In fact, a correlation between CO^NIand CO^TDwas better even during PSV in the previous study 4than in the current one, probably because the patients had been sedated more deeply with 2 to 3 mg · kg⁻¹· h⁻¹propofol, resulting in more regular spontaneous breathing. 4 

Once the rebreathing loop is adjusted for a given V^T, it is possible that an awake patient may increase V^T, making the rebreathing loop relatively too short. In light of this, it may be rational to adjust the rebreathing loop for the largest anticipated V^T. In this case, however, when the rebreathing loop is adjusted for large V^T, carbon dioxide rebreathing stimulates the respiratory center of the patient, resulting in increased V^T, respiratory rate, and V̇^E. The increase in V̇^Ewas 28% during SIMV–PSV and 46% during PSV (table 4), suggesting that respiratory efforts increase during carbon dioxide rebreathing. Although fully retracting the rebreathing loop minimized respiratory efforts due to carbon dioxide rebreathing, minimal loops resulted in overestimation and poor correlation (table 5).

The current study has several limitations. First, we waited for approximately 15 min to establish steady state after each change of ventilatory condition. However, when spontaneous breathing effort is present and V̇^Eis changed, more time may be required to attain stable conditions and more accurate CO^NI, which impairs clinical usefulness of NICO²monitoring. Second, all patients in this study were sedated, but awake patients may respond differently to carbon dioxide rebreathing. Third, effects of gas compression on the V̇^Eor V^Tmeasurement should be considered in patients with low compliance or high resistance, although the patients enrolled in this study showed normal lung mechanics. Finally, the range of CO measured in this study was relatively small (3.26–9.6 l/min) and only one steady state CO was studied in each patient. Further study is needed to evaluate the accuracy and reproducibility of the NICO²system under various hemodynamic conditions.

In conclusion, the change in V̇^Eor alveolar ventilation rather than V^Taffects the accuracy of CO measurement using noninvasive partial carbon dioxide rebreathing. When spontaneous breathing efforts are present, the CO measurement using the partial carbon dioxide rebreathing method becomes less accurate. Moreover, during the carbon dioxide rebreathing phase, the respiratory efforts during SIMV–PSV or PSV mode increase.