Carbon dioxide is produced in the tissues by aerobic plus/minus anaerobic metabolism (Fig. 1a), transported in blood to the lung by venous return (essentially equal to cardiac output [QT]), and eliminated from the lung by minute ventilation (VE) [1]. In this model the lung is a simple mixing chamber and the alveolar fractional carbon dioxide (FACO₂) is given by

FACO₂ = ♵CO_2,ti/♵A + FICO₂   (1)

where ♵CO_2,ti is the tissue carbon dioxide production, ♵A is alveolar ventilation, and FICO₂ is the inspired FCO₂. If one assumes no diffusion defect for carbon dioxide, then the partial carbon dioxide tension (PCO₂) of arterial blood (PaCO₂) leaving the lung is the perfusion-weighted average alveolar PCO₂ (PACO₂). Note that pulmonary shunt will add mixed venous blood with high PCO₂ (PVCO₂) to arterial blood and slightly increase PaCO₂ [2]. ♵A is the product of respiratory frequency and expired tidal volume (VT). Expired VT is composed of alveolar VT and total physiologic dead space (VD_phy). The fraction VD_phy/VT is given by

VD_phy/VT = (PaCO₂ - P♾CO₂)/PaCO₂   (2)

where P♾CO₂ is the mixed expired PCO₂ [2]. In turn, VD_phy is partitioned into anatomic dead space (VD_ana; conducting airways that do not participate in gas exchange) and alveolar dead space (VD_alv; ventilated alveolar units that are devoid of perfusion; Fig. 2). VD_alv/VT_alv is given by

VD_alv/VT_alv = (PaCO₂ - PACO₂)/PaCO₂   (3)

where PACO₂ is the alveolar PCO₂, estimated either from PETCO₂ or PǢCO₂ [2] (see below). The PaCO₂-PETCO₂ gradient results from the presence of VD_alv or high alveolar ventilation-to-blood flow (VA/Q) lung regions (see also Capnography during weaning from mechanical ventilation, below).

The normal capnogram is the measurement of PCO₂ at the airway opening during the ventilatory cycle (Fig. 1b) [1]. Phase I (inspiratory baseline) reflects inspired gas, which is normally devoid of carbon dioxide. Phase II (expiratory upstroke) is the transition between VD_ana, which does not participate in gas exchange, and alveolar gas from the respiratory bronchioles and alveoli. Phase III is the alveolar plateau. Traditionally, PCO₂ of the last alveolar gas sampled at the airway opening is called the PETCO₂. Finally, phase IV is the inspiratory downstroke, the beginning of the next inspiration.

However, the capnogram contains no volume information. Accordingly, the PǢCO₂ [2,3], which is the volume-averaged alveolar PCO₂, is a better index of PACO₂ than is PETCO₂, which is just a single measurement of PCO₂ at the end of exhalation [2]. A more informative determination of pulmonary carbon dioxide elimination is VCO_2,br, which is starting to garner clinical acceptance. VCO_2,br is the multiplication and integration of airway flow and PCO₂ over an entire respiratory cycle [4,5,6]. See the section on Future directions of carbon dioxide kinetics monitoring, below, for an interpretation and contrast of the measurements of VCO_2,br and PETCO₂.

The disposition of carbon dioxide can also be represented in a hydraulic model (Fig. 3) [3,7]. The large peripheral tissue compartment drains through a conduit ( T) into the small central pulmonary compartment. The central compartment can be further divided into pulmonary shunt (♵A/ = 0), normal lung (♵A/ near unity), and high ♵A/Q lung units, including VD_alv. The tissues produce carbon dioxide (VCO_2,ti), which empties into the peripheral tissue compartment. Then, carbon dioxide flows by gravity (QT) from the higher level peripheral tissue compartment to the lower level pulmonary compartment. ♵A, which equals ♵E minus VD_ana and the effects of high ♵A/Q units, eliminates carbon dioxide from the lung. In this model, T affects the distribution and total amount of carbon dioxide in the body. For example, at low T, retention of carbon dioxide occurs in the peripheral tissue compartment, and higher PVCO₂ is required to restore carbon dioxide delivery to the lungs. This hydraulic model can help to understand the meaning of PETCO₂ during successful cardiopulmonary resuscitation (CPR), and to compare PETCO₂ with PaCO₂ in the assessment of ventilator parameters. See the section, Effect of positive end-expiratory pressure on carbon dioxide kinetics, below, which highlights the utility of the hydraulic model.

Capnometry is the measurement of FCO₂ in tidal gas at the airway opening [1,8]. Capnography is the graphic display of measured FCO₂ versus time. Capnometry most commonly utilizes infrared light absorption or mass spectrometry [9]. Both methods are reliable and relatively accurate. Capnometers that are used in clinical practice use two different sampling techniques: sidestream or mainstream sampling. A mainstream capnometer has an airway adaptor cuvette attached in-line and close to the endotracheal tube (ETT). The cuvette incorporates an infrared light source and sensor that senses carbon dioxide absorption to measure PCO₂. A sidestream capnometer uses a sampling line that attaches to a T-piece adapter at the airway opening, through which the instrument continually aspirates tidal airway gas for analysis of carbon dioxide.

Because the lung is the only body compartment in which carbon dioxide normally and continuously accumulates, the presence of cyclic exhaled carbon dioxide can be used to confirm airway patency and pulmonary ventilation. Although initially adopted for anesthesia monitoring in the operating room, the use of capnography to confirm airway patency and lung ventilation has expanded over the past 8 years to include critical care, emergency medicine, field resuscitation, and conscious sedation settings [1,8,15,18,19,20,21,22].

However, there are pitfalls in the use of capnography to confirm endotracheal intubation. Potential problems with technology are described above. In addition, several scenarios have been described that impact on the ability of capnography to assess the airway and ventilation.

First, during circulatory arrest, pulmonary ventilation will result in low and decreasing values of exhaled carbon dioxide because QT and carbon dioxide transport from the tissues to the lung are decreased or absent in the presence of continuing VA [23,24,25]. In the clinical setting, however, Vukmir et al [19] demonstrated that infrared capnography was 100% specific and sensitive in the detection of endotracheal versus esophageal intubation in 100 critical care cases of airway management, 17 of which were cardiac arrests.

Second, positive-pressure ventilation by face mask can force pharyngeal gas, containing exhaled carbon dioxide from the previous breath, into the esophagus and stomach [1]. Likewise, ingestion of carbonated beverages can also generate carbon dioxide in the stomach [26]. Subsequent esophageal intubation and gastric ventilation can result in initial cyclic `exhaled' carbon dioxide. However, esophageal intubation usually causes an initial `PETCO₂' that is less than 10 mmHg and that decreases with each `exhaled breath' as inspiration dilutes carbon dioxide in the stomach [27]. In the case of suspected esophageal intubation, consider interpreting the value of exhaled carbon dioxide after the sixth breath [15].

Third, in a case report in a neonate weighing under 700 g [28], although the ETT tube was correctly positioned in the trachea, displacement of the ETT against the lateral wall of the trachea resulted in a flat capnogram and an erroneous diagnosis of esophageal intubation.

Fourth, pathology that causes absence of ventilation, including severe bronchospasm, patient apnea, or plugged ETT will result in absence of expired carbon dioxide and a falsely negative diagnosis that the ETT is not in the trachea.

Finally, it is prudent to remember that a normal capnogram confirms ventilation of the lungs through a patent airway, but not necessarily a secure airway. In a case report [29], a normal capnogram resulted during ventilation through an ETT positioned at the glottic opening, but not securely placed in the trachea.

Despite these potential drawbacks, capnography remains the most reliable monitor of airway patency in a variety of experimental and clinical settings. Mickelson et al [30] demonstrated that exhaled carbon dioxide was the most reliable indicator of esophageal intubation in canine model. Likewise, Knapp et al [31] studied current methods of verifying tracheal tube placement in the critical care setting, and found that capnography was superior to auscultation or other devices such as the lighted stylet. Capnography can also recognize esophageal intubation in neonates [32]. In the field, compared with other devices carbon dioxide monitoring best detects esophageal intubation by limiting the number of false negatives and false positives [15].

In addition to confirmation of ETT placement in the trachea, capnography may aid in cases of difficult intubation. During awake, blind, nasotracheal intubation, the end of a sidestream capnometer sampling probe can be placed through and positioned at the distal end of the ETT [33]. Then, increasing values of cyclic exhaled PCO₂ can help guide the ETT to the glottic opening. During a difficult intubation, effective ventilation can be maintained through a tube at the tip of the pharynx (guided by the expiratory carbon dioxide waveform), until other adjuncts to intubation are available [34].

Sidestream capnography adapts well to the nonintubated, sedated patient. Croswell et al [35] compared monitoring by capnography, pulse oximetry and clinical observation in sedated, pediatric, dental patients. Capnography provided a minimum 15 s warning of potential arterial desaturation, and was the most sensitive method for detecting airway compromise, especially during deeper levels of sedation. With oral/nasal capnometry in pediatric patients after active seizures, Abramo et al [36] demonstrated that PETCO₂ is a useful predictor of hypercapnia and is more sensitive than pulse oximetry in predicting impending respiratory failure. Other studies [8,37] have supported the assertion that capnography provides the earliest warning of airway obstruction and respiratory compromise.

Finally, capnography is a useful monitor during transport of intubated, critically ill patients [38,39]. Beside the obvious advantage of early warning against ETT dislodgment and/or compromise of ventilation, monitoring of PETCO₂ (as an estimate of PaCO₂) may aid the management of patients in whom hypercapnia is detrimental, such as patients with head injury with raised intracranial pressure and pediatric patients with pulmonary hypertension [38].

Capnography has been considered a potentially useful noninvasive monitor to assess the weaning of patients from mechanical ventilation in critical care settings [40]. However, studies have shown variable results in the ability of PETCO₂ to predict PaCO₂. Whether the use of PETCO₂ can limit the need for invasive arterial blood gas monitoring has yet to be established.

In a 1985-1991 literature review of the efficacy of noninvasive blood gas monitoring in the adult critical care unit [41], the Technology Subcommittee of the Working Group on Critical Care (Ontario Ministry of Health) concluded that changes in PETCO₂ need to be interpreted with extreme caution. Healey et al [42] compared the correlation of PETCO₂ with PaCO₂ before and after withdrawal of assist control mechanical ventilation. PETCO₂ paralleled changes in PaCO₂ (r = 0.82). Saura et al [43], in a prospective study to evaluate the relationship between PaCO₂ and PETCO₂ before and during weaning with continuous positive airway pressure ventilation, also found that PETCO₂ could detect clinically relevant hypercapnic episodes. However, there was a high incidence of false positives that led to arterial blood gas sampling. Withington et al [44] found that, after a gradient between PaCO₂ and PETCO₂ was established, PETCO₂ was a useful parameter in the weaning of postcardiac surgery patients.

The assessment of PETCO₂ may be misleading if not considered in the context of changing hemodynamics and ventilatory pattern. Although there can be significant correlation of PETCO₂ with PaCO₂, clinically acceptable sensitivity and specificity may only occur in the absence of significant changes in T or ♵A/Q relationships. In evaluating the use of capnography as a noninvasive monitor of PaCO₂ in critical care patients, Morley et al [45] observed that PETCO₂ was useful as a predictor only in patients without significant parenchymal lung disease. Prause [46] found that PETCO₂ was useful for the adjustment of ventilatory parameters in prehospital emergency care patients only if they had no major cardiopulmonary damage. As depicted in Fig. 2, the gradient between PETCO₂ and PaCO₂ depends on VD_alv (ie the amount of lung regions with high or infinite ♵A/ ratios) [2,25]. Lung regions with high ♵A/ ratios can result from high alveolar pressures (eg large VT, positive end-expiratory pressure [PEEP]), low pulmonary perfusion pressures (eg low T, upright position), and obstruction of pulmonary blood flow (eg thrombus, gas, or fat embolism). Thus, in the critically ill patient, VD_alv often changes and affects the ability of PETCO₂ to predict PaCO₂ and be a substitute for arterial blood gas sampling.

In our opinion, better understanding of pathophysiology of carbon dioxide kinetics during steady and nonsteady state should improve clinical care during intensive care treatment. Capnography and the measurement of PETCO₂ will gradually be augmented by relatively new measurement methodology (including VCO_2,br and PǢCO₂). Future directions include the study of oxygen kinetics [1].