# Pulse Oximetry, Plethysmography, Capnography, and Respiratory Monitoring

Rearranging the equation to employ the measured ratio of transmitted to incident light, we can define absorbance (A) as:

The absorption coefficient (a) is unique value for each wavelength and substance of interest. Oxygenated and reduced hemoglobins have different absorption coefficients across a spectrum of wavelengths, shown in Figure 23.1.

If we define Co as the relative concentration of oxygenated hemoglobin (compared to the total amount of hemoglobin) and Cr as the relative concentration of reduced hemoglobin, we can then define the total absorbance of light at a given wavelength as:

where W is the weight of hemoglobin per unit volume. Note that Co + C = 1.0.

In Figure 23.1, it is seen that at a wavelength of 805 nm, the coefficient for oxygenated hemoglobin (ao) and reduced hemoglobin (ar) are the same, a unique instance enabling determination of total hemoglobin (oxygenated and reduced) in the light path. A measurement at this wavelength will give us a value for WL:

If we now measure at any other wavelength where ar and ao are not equal, we can exploit the ratio to eliminate WL and determine the relative concentration of oxygenated hemoglobin (Co):

All the absorption coefficients a are constants that depend on the physical media and wavelength. We can thus group those together and rewrite this equation as:

This shows that we can measure the relative concentration of oxygenated hemoglobin by looking at the ratio of the absorbance (which we can measure) and adjusting with some known constants.

In a laboratory instrument we can have tight control of path length (a cuvette) and wavelength (a laser at 805 nm); these conditions are not possible for biologic or clinical application. For example, if we use a finger as our measurement compartment, Eq. [2] will generally hold, but we need to add the absorption of other finger tissues:

Since we are only interested in Aarterial, the other absorbance values need to be eliminated. This can be done by taking the time derivative of Atotal. The volume of the tissue and venous components can be treated as constants since they do not vary nearly as much as arterial pulsations, so the time derivative of the venous and tissue components will be about zero. The pulsatile part of Atotal is the arterial component with a nonzero time derivative. Hence the name “pulse” oximetry: It analyzes only the pulsatile portion of the absorbance.

When using a finger sensor, specific wavelengths are selected; LEDs emit light at specific wavelengths, 660 nm (red light) and 910 nm (infrared light), and have been employed for pulse oximetry. Other similar wavelengths can be chosen for particular designs (19). We can then define a ratio, R, as the ratio of the derivative of absorbances at two different wavelengths:

Applying similar logic as in the 805 nm case, we can derive an equation for Co:

where constants k are a combination of the (oxygenated and reduced) absorption coefficients a at the two wavelengths. While these coefficients are only dependent on the physical optical properties of hemoglobin, in practice, the constants k in Eq. [9] are determined empirically by pulse oximeter manufacturers, and are generally unpublished. Simpler (19) and more exact (20) derivations are available.

#### PERFUSION INDEX

Perfusion index (PI), or peripheral flow index (PFI), is a simple ratio of the varying to the nonvarying absorbance in the infrared, since it is minimally affected by changes in oxyhemoglobin saturation affecting the red absorption. Conceptually, it is analogous to alternating (AC) and direct (DC) electrical currents, represented as a ratio (or as percentage, × 100):

PI = (AC/DC)

PI changes with variation in peripheral perfusion. Relative change for an individual is useful, but variation between individuals can be considerable (19,21).

#### REASONS FOR ERRORS

Pulse oximetry fundamentally relies upon adequate perfusion of the vascular bed being monitored. Without a sufficient pulsatile signal, O2 content cannot be adequately analyzed. Decreased perfusion may be caused by a variety of factors including hypotension, medications, ambient temperature, poor circulation, and so forth. Clinicians will often search multiple fingers, toes, and earlobes for a site that can provide a saturation value. Decreased perfusion, leading to the inability of the pulse oximeter to provide a saturation value, is very common. While central site probes may never be ubiquitous, their utility in patients with poor peripheral perfusion cannot be overstated. Since these central sites reflect carotid artery flow, they will rarely experience errors due to poor perfusion of the vascular bed. Sometimes even rotation of a pulse oximeter probe by 45 or 90 degrees on a digit can generate a usable signal that was otherwise inadequate.

Pulse oximeters are calibrated using saturation curves of healthy adult volunteers. They are, therefore, most accurate at high saturation levels and less so at low saturation levels (22,23). Unfortunately, from the clinical standpoint, low saturations pose the greatest dangers. Nonetheless, this is rarely of major consequence, as the clinical difference between a saturation of 83% and 80% is probably minimal. The utility of oximetry at low saturation is principally for rapid assessment of the direction of change with therapeutic interventions.

It is clinically important to be able to detect rapid hemoglobin desaturation with minimal delay. There is often, regrettably, a delay in numerical display of saturation due to the default moving average-calculated SpO2 set by device manufacturers. However, many pulse oximeters annunciate a tone associated with saturation of each pulsation which audibly decreases even with small changes in saturation. That tone can be the clinician’s earliest warning of deterioration of oxygenation (17,18). The delayed response time problem for pulse oximetry–detected desaturations can be partially overcome by reducing the user selectable averaging setting to the shortest duration, usually 2 seconds (24). However, because of an increased likelihood of false alarms and artifact, this is seldom done clinically. To overcome the problem of delayed response time, it is necessary to develop processing algorithms sensitive enough to detect changes quickly, while allowing for artifact rejection and avoiding false alarms, an area of active research.

Other sources of error have been explored by Trivedi and colleagues (25). The researchers looked at very common sources of error, including ambient light and motion artifact. Error rates with excess ambient light were as high as 63% for heart rate and 57% for saturation. For motion artifact, simulated with 2-Hz and 4-Hz tremors, all tested pulse oximeters showed clinically significant error rates in saturation with both movement artifact rates. Error rates were low in the 2-Hz motion for heart rate calculations; however, all devices failed at 4-Hz motion. Other investigators have also reported on the errors and false alarms associated with movement artifact (26–29). Additional sources of error include darkly pigmented skin (30,31), nail polish (32), thermal injuries to fingers and/or toes, and inaccessibility of the extremities. While unconventional, sometimes simply rotating application of a pulse oximeter probe on the axis of a finger or toe can improve signal and performance by changing the light path, though some report this maneuver is ineffective (32).

Another very common problem encountered is the lack of compatibility between probes and devices of the various manufacturers. van Oostrom and Melker (33) compared the accuracy of nonproprietary probes designed for use with a variety of pulse oximeters with that of their corresponding proprietary probes. A controlled signal was used on the Human Patient Simulator to simulate apnea. Statistical significance was not found in most of the comparisons, but in some instances the proprietary probes were closer to arterial oxygen than the nonproprietary probes. Thus, whether or not the manufacturer of the probe is the same as the manufacturer of the pulse oximeter may have importance. Table 23.1 summarizes the various reasons for errors (34).

 TABLE 23.1 Reasons for Artifactual Measurement in Pulse Oximetry

As can be seen from Eq. [2], the intensity of light measured at the detector varies by path-length changes caused by the arterial pulsations. All other parameters in the equation are constant (at least within several minutes). Path-length changes are caused by volume changes at the sensor site, and for this reason, volume artifacts are a frequent problem. Other errors can be caused by light interference from sources outside of the measurement system or electrical interference, and will typically show up in the plethysmogram as additional waveform fluctuations. The effects of this noise are a distorted plethysmogram and can cause incorrect heart rates and saturations to be calculated. One last source of errors is the patient himself or herself, especially for smokers: Carboxyhemoglobin present in the blood can cause an inaccurate determination of saturation.

#### ALTERNATE-SITE PROBES

The standard location to measure pulse oximetry is the finger. Such probes are common and work well in many cases, but become unreliable with poor perfusion, vasoconstriction, and hypothermia. The nasal septum was explored as a possible monitoring site in 1937 (41). Groveman and colleagues (42) also explored the nasal septum, believing that it represents a constant picture of the internal carotid circulation and reflected cerebral flow. Cucchiara and Messick (43) showed that plethysmography from the nasal septum failed to estimate cerebral blood flow during carotid occlusion. In 1991, the nasal septum was explored during hypothermia (44). Fourteen patients were monitored every 20 minutes during major abdominal procedures. The nasal septum probe was superior to the finger probe in detecting a pulse during hypothermia. The authors concluded that monitoring at the nasal septum was more reliable than monitoring at the finger in hypothermic patients. They acknowledged several limitations, including use during nasal intubation, in patients with extremely small nostrils, or in the presence of a nasogastric tube.

Buccal probes have been evaluated as an alternative probe site. They were prepared by taping a malleable metal bar securely over the back of a disposable Nellcor finger probe and bending the metal bar and probe around the corner of a patient’s mouth (45). It was determined that buccal SpO2 was greater than finger SpO2 and agreed more closely with SaO2. The authors determined that buccal pulse oximetry is a viable alternative to the finger. Limitations included longer preparation time, difficult placement, and possible dislodgement during airway maneuvers.

Awad and colleagues (46) demonstrated that the ear plethysmographic waveform is relatively immune to vasoconstriction. They also determined that the photoplethysmographic width has a good correlation to cardiac output. They concluded that the ear is more suitable for monitoring hemodynamic changes than the finger.

Generally, any site that has an arterial bed and is thin enough to safely transmit red and infrared light through can be used for pulse oximetry. Several monitoring locations have become standard of care, including the fingers and toes. A flexible earlobe probe is also used quite frequently when the fingers and toes are inaccessible. Since finger probes work quite well for the majority of patients, it is unlikely that alternate-site probes will ever become ubiquitous. As the nasal septum, nares, cheek, and ear measure oxygenation from central sites, reflecting blood flow of the carotid arteries, their potential for measuring other physiologic parameters is only beginning to be explored.

Reflectance pulse oximetry has become available as an alternative, commonly applied to the forehead. It is less prone to motion artifact, may be less prone to errors due to poor perfusion from vasoconstriction of peripheral digits, and assesses a peripheral site derived from carotid perfusion (47). Nasal ala have also been employed as another site derived from carotid circulation (48).

#### OVERCOMING LIMITATIONS

Errors and interference on pulse oximetry can largely be eliminated or prevented. Volume/movement artifact can be reduced by ensuring that the measurement site is kept in place or moved slowly. Masimo (35) developed probes that use a third light source; with this additional measurement, it is possible to estimate nonarterial volume changes due to movement artifact. This allows for compensation for those artifacts and creates a more stable signal that is not as susceptible to motion artifact. Light interference can largely be eliminated by covering the measurement site through the use of a properly sized probe, or by external means such as towels or other covers. Patient-related artifacts can be reduced by fully understanding the patient’s physiology, and by proper selection of the measurement site.

#### PHOTOPLETHYSMOGRAPHY

Photoplethysmography is the measurement of volume changes with light transmission. The photoplethysmograph (PPG) is displayed on most pulse oximeters; however, it is frequently ignored as oxygen saturation and pulse rate are the numbers of interest. There is an abundance of physiologic information that can potentially be extracted from this rarely used and noninvasively obtained signal.

##### Fundamentals

There are two main frequencies of variation in the value of light hitting the photodiode, and both are affected by absorption of the light by blood and various tissues. The low-frequency component (LFC)—DC or nonpulsatile component—represents the baseline amount of light hitting the detector. This value is affected by the total path traveled by the light; skin, bone, cartilage, adipose, blood, and so forth, all absorb light, and it is this relatively constant path that results in a baseline amount of light hitting the detector. This baseline amount fluctuates at a lower frequency than the heart rate. Since the biologic tissues in the path of the light are constant, with the exception of venous and arterial blood, the changes in the LFC correspond to changes in baseline blood volume in the path of the light. The majority of this baseline blood resides in the venous system.

The pulsatile cardiac component (PCC) corresponds to changes in the arterial blood volume with each heartbeat. The magnitude of change of the PCC with each heartbeat is related to stroke volume, and the area under the curve of each heartbeat is related to the volume of blood entering the vascular bed with each beat (49). The PCC is, therefore, a representation of flow into a vascular bed while the LFC is a representation of changes in venous volume (Figure 23.2A).

The typical pulse oximeter displays a processed waveform (Figure 23.2B). Since the raw data collected by the device corresponds to light hitting the photodiode, which is inversely related to blood volume, the waveform must be inverted to resemble an arterial pressure waveform. If the PPG was displayed as raw data and not inverted, point A would represent increasing light hitting the photodiode, corresponding to a decrease in blood volume, and point B would correspond to the point of maximum light hitting the photodiode, or the point of least blood in the vascular bed being monitored. Ideally, the plethysmograph would completely mimic the characteristics of an invasive arterial pressure tracing. Sometimes it comes very close. However, other factors influence the photoplethysmogram so that the waveform can be quite unlike an intra-arterial pressure. The steepness of the flow of the inflow phase A may be used as an indicator of ventricular contraction, and the amplitude of the phase may be used as an indicator of stroke volume (49). The vertical position of the dicrotic notch can be used as an indicator of vasomotor tone. Under most circumstances, the notch descends to the baseline during increasing vasodilation and climbs toward the apex with vasoconstriction (49). However, in many cases likely due to the variety of proprietary algorithms employed to process the PPG, unusual waveforms bearing little resemblance to the invasively measured arterial pressure waveform may be displayed by pulse oximeters, especially in conditions of poor peripheral perfusion (19,50,51).

##### Signal Processing

Prior to the advent of powerful personal computers, many researchers printed the PPG waveform and measured various parameters with a ruler; more recent efforts involve elaborate mathematical and signal processing models. Bhattacharya and colleagues (52) employed a novel concept aimed at detection of the dominant nonsinusoidal period and the extraction of the associated periodic component. This detection and extraction was performed with a moving window to accommodate the variations of the physiologic oscillations. They also characterized the system with a nonlinear dynamic system.

Most signal processing algorithms employed in pulse oximeters are proprietary and unpublished. The reasons and implications have been vigorously debated (51,53,54). Goldman and colleagues (35) from the Masimo Corporation published a detailed description of their signal extraction for error reduction. Masimo Signal Extraction Technology (SET) uses a new conceptual model of light absorption for pulse oximetry and employs discrete saturation transformed to isolate individual saturation components in the optical pathway. Johansson (55) processed the PPG signal using a 16th-order bandpass Bessel filter and a 5th-order bandpass Butterworth filter; a neural network analysis was then performed to determine respiratory rate. Nilsson and colleagues (56) employed three separate methods for the evaluation of the PPG (called the blood volume pulse) for changes caused by exercise. First, they derived a single parameter from the distribution found in the average histogram of the time-aligned beats. Their second approach analyzed the ratio observed between the first harmonic and higher harmonics in the signal. The third approach evaluated the dicrotic notch depth directly from the PPG waveform. Alian and Shelley further discussed the detrimental effects of excessive filtering and autogain on the utility and interpretation of the PPG, potential use of heart rate variability (HRV) as an index of autonomic nervous system activity, pulse transit time (PTT) as a correlate of arterial system rigidity, respiratory modulation as a guide to goal-directed fluid therapy, and artifacts due to venous engorgement from over-resuscitation or steep Trendelenburg positioning (19,50,51). Further significance of these findings will be elaborated below.

##### Uses of the Photoplethysmograph

The PPG is a noninvasively obtained window into many physiologic parameters. Since a pulse oximeter probe can be placed by those with minimal or no training and are found in virtually every aspect of medical care, there are many active research projects exploring the potential uses of the PPG (50,51).

Several researchers have attempted to construct mathematical relationships between the PPG and various indices of arterial mechanics. Kato and colleagues (57) measured the PPG from a finger pulse oximeter probe and pressure at the ipsilateral radial artery, simultaneously. The authors concluded that a four-element, two-compartment model can be applied to the PPG to determine peripheral vascular wall mechanics. Chowienczyk and colleagues (58) determined that PPG assessment may provide a useful method to examine vascular reactivity. Millasseau and colleagues, in 2002 (59), concluded that contour analysis of the digital volume pulse (DVP) provides a simple, reproducible, noninvasive measure of large artery stiffness, and, in 2003 (60), determined that indices of pressure wave reflection and large artery stiffness can be used as an index of vascular aging. Bortolotto and colleagues (61) concluded that the second derivative of the PPG and the pulse wave velocity can both be used to evaluate vascular aging in hypertensives.

Other researchers are exploring the use of the PPG for noninvasively determining respiratory rate. Changes in intrathoracic pressure during the respiratory cycle displace venous blood, affecting the LFC. These changes also affect cardiac return, changing the amplitude of the PCC. During spontaneous breathing, subatmospheric pressure during inspiration draws air and blood together into the lungs; blood is drawn from the vena cava into the right heart and pulmonary vascular bed; a minor decrease in peripheral venous pressure (PVP) ensues. Soon thereafter, the expiratory pressure normalizes the system. During positive pressure ventilation, the inspiration is driven by positive pressure, which raises intrathoracic pressure and reduces venous return to the right heart. Simultaneously, and very briefly, blood forced from the low-pressure pulmonary vascular bed increases return to the left heart as well as stroke volume (62). This is followed by a decrease in cardiac output as venous return into the central circulation drops off. The extent of the fluctuations caused by positive pressure ventilation depends on the state of filling of the peripheral vascular bed, the intrathoracic pressure changes, peripheral vasoconstrictor activity, and central blood volume (49). Since positive pressure ventilation often accompanies general anesthesia, which causes vasodilation and damped vasomotor response, respiratory fluctuations are emphasized. It was also discovered that early hypovolemia may be reflected in an exaggerated respiratory wave before other more classic signs of decreased urine output, tachycardia, or hypotension (49).

Nilsson and colleagues (63) extracted the cardiac- and respiratory-related components, applied a mathematic algorithm, and developed a new PPG device for monitoring heart rate and respiratory rate simultaneously; their study determined that the PPG has the potential for respiratory rate monitoring. Nilsson and colleagues (56) also hypothesized that the filling of peripheral veins is a major mechanism behind the LFC signal, and found that a correlation exists in the amplitudes of the LFC in the PPG and the respiratory variations in PVP (p < 0.01). Leonard and colleagues (64,65) concluded that baseline respiratory rate was easily identified from a pulse oximeter PPG using wavelet transforms. The study of Foo and Wilson (66) determined that the respiratory rate obtained from the PPG was significantly related to that estimated by a calibrated air pressure transducer during tidal breathing in the absence of motion artifact (p < 0.05). Nilsson and colleagues (56) concluded that respiration can be monitored by the PPG with high sensitivity and specificity regardless of anesthesia and ventilatory mode. Leonard and colleagues (67) continued their work by developing a fully automated algorithm for the determination of respiratory rate from the PPG.

Researchers have also been exploring the relationship between the PPG and volume status. Perel and colleagues (68) found that the difference between systolic pressure at end-expiration and the lowest value during the respiratory cycle (d-Down) correlated to the degree of hemorrhage. It also correlated with the cardiac output and the pulmonary capillary wedge pressure. Thus, the changes in systolic pressure with respiration, as demonstrated by arterial pressure waveforms (systolic pressure variation [SPV]) and its d-Down component, are accurate indicators of hypovolemia in ventilated dogs subjected to hemorrhage. Rooke and colleagues (69) also concluded that SPV and the d-Down appear to follow shifts in intravascular volume in relatively healthy, mechanically ventilated humans under isoflurane anesthesia. Building on this principle, Partridge (70) attempted to use pulse oximetry as a noninvasive method to assess intravascular volume status. The study showed that the PPG correlated with the systolic pressure variation (r = 0.61), which was previously shown to be a sensitive indicator of hypovolemia. Shamir and colleagues (71) investigated ventilation-induced changes in the PPG after removing and reinfusing 10% of the estimated blood volume in 12 anesthetized patients. The plethysmographic SPV was measured as the vertical distance between maximal and minimal peaks of waveforms during the ventilatory cycle and expressed as a percentage of the amplitude of the PPG signal during apnea. This was measured during five consecutive mechanical breaths before apnea and the mean value was obtained for analysis. The 10% loss of estimated blood volume resulted in increased heart rate without changes in mean arterial pressure. Both the PPG waveform changes and the SPV from the arterial blood pressure tracing increased significantly after blood withdrawal (p < 0.01). The changes in the PPG correlated with the changes in the SPV. After volume replacement, heart rate decreased while arterial pressure remained unchanged. There were no significant changes in the PPG waveform or the SPV with volume replacement. Fuehrlein and colleagues (72) investigated the use of PPG for volume status changes during blood donation and hemodialysis. They concluded that the PPG could be used to detect changes in volume status and vascular instability during hemodialysis and blood donation.

#### MULTIWAVELENGTH PULSE OXIMETRY

There is now an extensive literature concerning multiwavelength pulse oximetry, sometimes also called co-oximetry. Improvement of pulse oximetry performance, and new capabilities such as noninvasive estimation of concentrations of hemoglobin, carboxyhemoglobin, and methemoglobin are possible by using multiple wavelengths and new processing algorithms (73). Controlled studies in normal patients have reported generally good, but variable, correlations with laboratory determinations and trends in circulating hemoglobin concentrations, and responsiveness to blood and fluid administration (74). Because noninvasive results have not been reliably correlated with laboratory results, a debate continues concerning what is good enough or useful (54,75–77). The data have also demonstrated that there can be differences of 1 gm/dL in hemoglobin even between laboratory methods, as well as by oximetry (78). Correlation with laboratory values in low perfusion states, trauma resuscitation, and acute hemorrhage and circulatory changes have been less satisfactory than under more stable cardiovascular conditions (79). Assessment of hemoglobin concentration or trends in settings where laboratory determinations are not available or greatly delayed would have clear utility (80). The issues raised have contributed to the debate about which reference and transfusion triggers are appropriate, and how hemoglobin determinations could bias evidence or its application (53,81).

Rapid detection of high concentration of carboxyhemoglobin may be particularly significant for initial assessment of emergency patients. Detection of methemoglobins may be principally useful to avoid adverse effects of titratable drugs like nitroprusside (82).

A very interesting new development is Oxygen Reserve Index (ORI). ORI provides a graded, proportional change from 0 to 1 related to PaO2 from about 98 to over 200 torr, above the 100% SpO2 plateau with standard pulse oximetry (83). While this gradually increasing index is qualitatively proportional to changes in PaO2, it gives an indication of change over a range of partial pressures where there was previously no indication of change PaO2 after oxyhemoglobin reached full saturation, a useful feature for warning for impending desaturation or changes in lung function (84,85).

#### FUTURE DIRECTIONS OF OXIMETRY

While the foundational concepts of pulse oximetry are employed in many ways, innovation continues to provide improved performance and to break barriers and resolve limitations of previous generations of the technology. There is still room for improvement particularly at low saturations, for poor tissue perfusion, responsiveness to deterioration, alarm, and artifact management.

Clinically applicable assessment of mitochondrial oxidative state may probe previously inscrutable pathophysiology, for instance, in sepsis, hypoxia, and poor perfusion states (5).

Hemodynamic monitoring employing characteristics of PPG is beginning to realize objectives for cardiopulmonary optimization by noninvasive monitoring. Further investigation and pulse oximeter design optimization for PPG analysis will be needed (19).

Use of SpO2 rather than directly measured SaO2 in proportion to FIO2 as an index of severity of lung injury and risk stratification has the advantage of being noninvasive and easily trended, though not perfectly correlated (86–90).

NIRS assessment of various organs and particularly cerebral oxygenation is promising but not uncontroversial (91,92), having suffered from competing claims of manufacturers, lingering concerns about what is measured, and limited outcomes data (93,94). Use in evaluation of resuscitation from cardiac arrest is particularly intriguing (95). An unconventional approach to measuring jugular venous oxygenation is a promising method that may have utility for assessment of cerebral oxygen supply and demand matching (96).

#### CAPNOGRAPHY AND RESPIRATORY MONITORING

Capnometry refers to the measurement of carbon dioxide, regardless of the method used. Capnography describes the method of obtaining a capnogram, that is, a tracing of carbon dioxide concentration as a function of time or volume. A capnograph is the instrument used to generate a capnogram. Capnometry has become the minimal standard of practice for the American Society of Anesthesiologists since 1986—most recently amended in 2015—whenever a patient’s airway is breached with an endotracheal tube, a laryngeal mask airway, or an esophageal tracheal airway (97). If CO2 can be detected with each breath after placing the artificial airway, the clinician has the first and best—if not the only reliable—indication that the artificial airway is ventilating the patient’s lungs rather than the esophagus and stomach. However, merely detecting carbon dioxide on one or two breaths is not sufficient. Persistence of capnograms with continuous ventilation is necessary, because there can be enough carbon dioxide insufflated in a stomach that even gas returned from a distended stomach can resemble a capnogram. With esophageal intubation, the concentration of carbon dioxide returned with each breath will decrease rapidly with dilution, and the waveform of the capnogram will not have the typical shape (see below). With impaired circulation, such as during CPR, capnograms will be diminished in amplitude, or may even be absent if the lungs are not perfused with blood. In this instance, an endotracheal tube should be evaluated for correct placement, but left in place even if there is no capnogram. Continuous capnograms should be expected with effective CPR that generates circulation of blood.

The delivery of CO2 to gas exhaled from the lungs is the final step in a complex system. Cellular metabolism generates CO2, which diffuses through local tissue and is absorbed by perfusing capillary blood. In blood, CO2 is dissolved in plasma and buffered with proteins, principally hemoglobin. Venous blood flow delivers the CO2 to the heart, which powers pulmonary perfusion, and delivers it to the lungs and, via ventilation, gathers it from the alveoli when a breath finally pushes it to the outside. Thus, the discovery of CO2 after intubation of the patient’s airway, while essential, is but the tip of the proverbial iceberg. In this section, we provide a clinician’s overview of capnometry and its applications in the perioperative period and the intensive care unit.

#### HISTORY

Carbon dioxide and its measurement have enjoyed a colorful history (98–100). Jan Baptista van Helmont (1579–1644) recognized a spirit escaping from burning wood and called it gas sylvester (from Latin silva = wood and silvester = woody). The recognition that a gas could reside in something solid found expression in the term “fixed air” introduced by J. Black in 1755. He discovered the gas to be a constituent of carbonated alkali. Later, Antoine-Laurent de Lavoisier (1743–1794) showed the gas to be an oxide of carbon. What an extraordinary discovery: carbon (of coal and diamonds) connected with oxygen was a gas! John Tyndall (1820–1893) spoke of “perfectly colorless and invisible gases and vapors” such as carbonic acid (now called carbon dioxide) that could well absorb radiant energy. This insight enabled him to detect carbon dioxide in the exhaled gas. However, before capnography based on physical methods could gain a foothold in clinical practice, a chemical method described by John Scott Haldane (1860–1936) became the “gold standard.” He caused a precisely measured volume of gas to be drawn into a closed system that made it possible to expose the gas to absorbents such as sodium or potassium hydroxide. These agents removed the carbon dioxide from the sample; the vanished volume was attributed to the absorbed CO2. Refinements of this method were widely used, yet the methods were time consuming. Chemical methods, in general, destroy the gas to be measured and allow only snapshots of respiratory CO2.

A number of methods exploited the physics of energy absorption. August Hermann Pfund (1879–1949), Professor of Optics in Baltimore, measured the effects of interposing more or less CO2 between a heat source and a temperature sensor. Karl Friedrich Luft (1900–1999) employed infrared energy beamed through cells with and without CO2, thus enabling the measurements of the energy absorbed by the gas in question. Later generations of this same principle gave rise to the currently most widely used infrared spectroscopic method of capnography.

Two other historical methods that were employed clinically for capnography, as well as other medical gas analyses, deserve mention. Both have been displaced by more economical—and less capable—infrared gas analyzers. One of the first methods used for clinical gas analysis, including capnography, was mass spectrometry, in which charged particles are separated by their mass when an ionized gas beam is passed through a magnetic field. The other was Raman scattering spectroscopy (101). Both of these spectrometers have other medical applications, including specialized gas analyses beyond needs for basic capnography. Gas mixtures add analytic considerations; nitrous oxide can interfere with PCO2 measurement. Infrared gas analyzers cannot directly quantify nonpolar molecules, such as nitrogen and inert gases like helium and argon; together they represent the residual difference from atmospheric pressure after accounting for the reported gases and water vapor. Raman and mass spectrometry can directly quantitate mixtures even of nonpolar gases.

#### SITES OF MEASUREMENT

Capnography, herein examined, focuses on the detection and monitoring of CO2 in the exhaled gases. There will always be small differences in PCO2 between blood and different respired gas sampling sites. There must be at least a small difference in partial pressure of CO2 to have a gradient for diffusion, even in health. With pathophysiology, increased gradients and mixing of gases from areas of lung with different ventilation–perfusion ratios will confound assumptions of uniform gradients. While direct measurement of alveolar CO2 (PACO2) is not clinically possible, the assumption that PetCO2 will provide a valid proxy measurement is frequently not valid due to dilution of alveolar gas by admixture, either within the lung or the breathing apparatus before the measurement sampling point (102). This point is one of the most unappreciated errors for interpretation of capnograms and incorrect assertion of a-A gradient (103). Thus, capnograms from breath sampling sites without a closed connection to the trachea will be substantially diluted and distorted by the free diffusion of surrounding atmospheric air (100,104).

CO2 can also be lost and consequently collected and measured from the skin (105), from the stomach (106), sublingually (107,108), and even the rectum (109), and other sites; these will not be discussed in this chapter.

With modern methods, continuous readings of exhaled CO2 tensions and volumes can be obtained. We speak of steady state when the partial pressure (tension, PCO2) of carbon dioxide in different tissue and organ compartments and blood and alveolar gas have reached equilibrium, and when the input of CO2 from metabolism equals the output of CO2 via ventilation and, to a small extent, via skin, flatus, feces, and urine. A steady state can exist with high, normal, or low arterial, alveolar, or end-tidal CO2 tension (PaCO2, PACO2, or PetCO2); however, all too often, we do not have a steady state: Tissue depleted of CO2 can absorb liters of the gas from blood until tissue PCO2 and blood PCO2 reach equilibrium; conversely, such tissue stores can contribute CO2 to that generated by metabolism. For example, prolonged hyperventilation can exhaust tissue stores of carbon dioxide and bicarbonate. Under these conditions, the maintenance of a normal PaCO2 requires less than normal ventilation because some of the metabolic CO2 filters back into the tissues instead of being exhaled. Conversely, high tissue stores of CO2 (e.g., after a cardiac and respiratory arrest) will call for greater than normal ventilation until steady state is once again reached. A depressed respiratory center (e.g., under the influence of an opiate) will lead to an imbalance of input and output as the tissues take up some of the CO2 while the rest leaves the body in the exhaled gas. Once the tissues and blood reach equilibrium, steady state will once again supervene in the presence of elevated levels of PaCO2, PACO2, and PetCO2. Renal compensation of metabolic acidosis or alkalosis, though much slower than respiratory changes, will also contribute to an unsteady state for many hours or days. The point is that capnograms can hide as much as they reveal, and thus the interpretation of capnograms calls for discerning clinicians.

#### CONVENTIONS OF MEASUREMENTS

Before describing the different methods of capnography, the conventions of reporting the tension or concentration of CO2 present in the exhaled gas must be discussed. The tension of the gas can be reported in mmHg (torr), with normal end-tidal values between 35 and 45 mmHg, which translates into 4.67 and 6.0 kPa, respectively (1 kPa = 7.5 torr). The amount of CO2 present in a gas sample can also be reported in percent or as a fraction of the volume of gas. Here caution needs to be exercised: 5.0% end-tidal PCO2 (equal to the end-tidal fraction, Fet, 0.05) at a barometric pressure at sea level of 760 mmHg (101 kPa) would amount to 38 mmHg (5.05 kPa). However, many millions of people live in cities at altitude. Assume, for example, Mexico City with an ambient pressure of about 550 mmHg (73 kPa). Here, 5.0% end-tidal CO2 would represent only 27.5 mmHg (3.65 kPa). The influence of the pressure exerted by water vapor (47 mmHg or 6.26 kPa at 37°C), which is not affected by barometric pressure but rises and falls as a function of temperature, also plays an important role at altitude (Table 23.2). As we wish to correlate alveolar gas tension with the tension of gases in blood (reported in mmHg or kPa), we prefer to report gaseous CO2 not in percent, but as PCO2 in mmHg or kPa.

#### METHODOLOGIES

##### Chemical

CO2 goes into solution, combines with water to form carbonic acid, and establishes an equilibrium between dissolved CO2 and bicarbonate. This reaction, which changes the pH, gives rise to chemical methods using pH indicators for the estimation of CO2 concentration in moist gas, or colorimetric etCO2 (Fig. 23.3).

Changes in temperature and the addition or removal of H+ ions, in turn, affect the ratio of bicarbonate to carbonic acid, and dissolved CO2, therefore, plays a crucial role in the acid-based equilibrium of blood, which is assessed by measuring pH and PCO2 and calculating bicarbonate in blood. This is discussed in detail elsewhere in this book, as is the importance of buffers (primarily hemoglobin) and carbonic anhydrase, which accelerates the reaction:

 TABLE 23.2 Carbon Dioxide and Barometric Pressure at 37°C

##### Sidestream and Mainstream (On-Airway) Capnography: Time-Based Methods

The gas to be analyzed has to be collected from the patient under conditions that prevent contamination of the gas with ambient air. Ideally, we would like to sample tracheal gas; that is rarely possible. A cuffed endotracheal tube with a port close to the mouth offers the next best opportunity to collect exhaled gas from the patient before it mixes with gas from the outside or the breathing circuit. First, we will take a look at the sidestream method.

###### Sidestream

On its way to the analyzer through a capillary, the gas cools (from body to room temperature) and the water vapor condenses, forming droplets. Two methods minimize the potential problem of having water obstruct the flow of gas or confound the spectroscopic analysis: Collecting capillaries made out of Nafion tubing enables the water vapor in the tube to diffuse through the tubing wall to equilibrate with the water vapor in air surrounding the tube. This leads to a reduction of water vapor burden in the capillary. The second, a more common method employs a water trap situated close to the analyzer.

Figure 23.4 shows a typical time-based capnogram obtained under ideal conditions. Here we plot the tension of carbon dioxide in the ordinate and time in the abscissa. Observe the angles α and β, both of which in a healthy individual approach 90 degrees. A respiratory pause at the end of exhalation will cause the plateau phase to become horizontal.

With the sidestream method, the costly analyzer can be kept out of the way; a thin, long aspirating capillary presents no encumbrance to the clinical team; and aspirated gas can be analyzed for CO2 as well as other gases. Furthermore, gas can be aspirated from nasal prongs with minimal annoyance to a conscious patient. Commercial configurations are available that enable the simultaneous aspiration of gas from one nostril while delivering oxygen to the other nostril or to the mouth. However, in that application, contamination of the aspirated gas with room air or oxygen is possible, although several studies have shown clinically satisfactory results with these arrangements. At a minimum, such a system can provide evidence of ventilation and enable the recording of respiratory rates.

The sidestream method has two well-recognized drawbacks. The longer the capillary is, the longer the travel time for the gas from the patient to the analyzer. That causes the capnogram to be out of phase with simultaneously recorded flow or pressure tracings. The long travel also gives the leading edge of a gas a chance to mix with the gas it is replacing in the tubing, which produces slurring of the capnogram, particularly noticeable with rapid respiratory rates (Fig. 23.5).

Many sidestream capnometers aspirate up to 200 mL gas per minute into the analyzer. Premature newborns, with their small tidal volumes and high respiratory rates, will then develop capnograms that show false low end-tidal and false high inspiratory PCO2 values. Two mechanisms contribute to these conditions: On the one hand, the tiny patient’s tidal volume might be so low as to cause the capnograph to aspirate gas from the breathing tube, thus diluting the exhaled gas from the patient. On the other hand, the time constant of the capnograph may be too long to respond adequately to rapid breaths. Modern capnographs offer low rates of aspiration (30–200 mL/min) and short time constants (110).

###### Mainstream

Instead of aspirating a sample of respired gas with a sidestream system, it is possible to determine the concentration of carbon dioxide close to the patient’s mouth in a “mainstream” or “on-airway” method (Fig. 23.6). The method eliminates two weaknesses of the sidestream system: Without a need to transport the gas through a capillary to the analyzer, mainstream capnograms show no slurring of the capnographic tracings. Of course, these advantages come with a (tolerable) cost: The carbon dioxide sensor must be brought close to the patient’s mouth, which adds weight to the breathing circuit/endotracheal tube; in this position, the sensor is exposed to potential damage, and obtaining gas samples from a spontaneously breathing, nonintubated patient is more difficult than would be true from a sidestream system. Nevertheless, mainstream systems have been adapted for capnography even in nonintubated infants.