Oxygen Delivery

Metabolizing tissues require an adequate supply of oxygen to efficiently produce the energy needed for cellular function. The quantity of oxygen loaded onto the arterial bloodstream per unit time (O₂ delivery) is the product of the cardiac output and the oxygen contained within each milliliter of blood. Therefore, a deficiency of either cofactor can be partially offset by a compensatory increase of the other. Conversely, sluggish blood flow, whether caused by low cardiac output or high resistance through the tissues, can limit the O₂ actually delivered to the cell. Increased blood viscosity impedes the transit of erythrocytes through the capillary bed, tending to limit oxygen consumption (VO₂).³ For this reason, paraproteinemia, extreme leukocytosis, and polycythemia can pose life-threatening challenges to O₂ consumption that are independent of any impact on cardiac output.³ Studies performed in animal models demonstrate that hematocrit (Hct), a primary determinant of viscosity, bears a nonlinear relationship to oxygen delivery that varies somewhat with circulating blood volume.⁴ At low values of Hct, a rising Hb concentration predictably adds to O₂ content and delivery. Above a Hct of 30% to 34%, however, it is difficult to demonstrate in critically ill patients an increase in oxygen consumption or an outcome benefit that derive from increases of O₂ content arising from further increments of Hb.⁵ At an Hct of around 55% to 57%, O₂ delivery reaches its maximum in normal subjects, falling sharply with each further rise in Hct (Figure 44-1). Above an Hct of approximately 65%, phlebotomy may be required to avert a hemodynamic crisis, as vital tissues may be deprived of delivered oxygen. Viscosity, and therefore tolerance for higher Hct, is partially determined by the circulating blood volume; the polycythemia associated with intravascular volume contraction is much less well tolerated than that of polycythemia vera, a condition in which circulating blood volume is expanded.³ As might be expected, patients with vascular disease are less tolerant to the adverse rheologic effects of high Hct.

Figure 44-1 Effect of hematocrit on viscosity and oxygen delivery.

Raising hematocrit simultaneously increases oxygen content and viscosity, which adversely affects blood rheology. Consequently, oxygen delivery reaches a maximum at hematocrit values in the upper midrange.

At the mitochondrial level, oxygen acts as the terminal acceptor in a chain of organic electron donors known as cytochromes. The PO₂ within the mitochondrion needed to sustain this process is very low—estimated to be much less than 1 mm Hg.⁶ To provide that needed level of mitochondrial oxygen tension, an appropriate oxygen diffusion gradient must be established from the arterial blood, across tissue and cellular boundaries, and into the cellular organelles. At sea level, normal levels of mitochondrial O₂ are achieved at a PaO₂ of about 95 mm Hg. The actual PO₂ within the mitochondrion, however, is affected by many factors other than arterial PO₂: tissue metabolic rate, microvascular control, tissue properties, and blood flow. Over time, varying degrees of accommodation to subnormal PaO₂ occur by adjustments of the cardiovascular system, Hb concentration, capillary system, and mitochondrial density.⁷ Although this adaptive phenomenon is commonly observed in patients with chronic lung diseases, the extent to which gradual accommodation to hypoxemia can occur and should be encouraged in patients who are critically ill is a provocative and largely unexplored question.

Oxygen Transfer Across the Lung

Oxygen is driven from the airspace to the pulmonary capillary by a diffusion gradient determined by the PO₂ difference between them and the resistance to diffusion presented by the intervening tissues and fluids. To keep the alveolar O₂ tension adequate, the oxygen supplied to the alveolus must be replaced at a rate equal to or greater than that at which the oxygen is removed by the passing capillary blood. Classically, six mechanisms can account for hypoxemia:

Low FIO₂ is an important mechanism of hypoxemia occurring at altitude and in fires that occur in confined spaces. Although the relationship is not a strictly linear function, as a rough estimate, inspired O₂ declines approximately 15 mm Hg for each 1000 meters of altitude above sea level.⁸ For practical purposes, however, a reduced concentration of inspired oxygen does not account for hypoxemia that occurs in the setting of critical illness. Hypoventilation alters the alveolar oxygen tension (PAO₂) in proportion to the rise of PaCO₂ (and PACO₂) and becomes an important factor when it occurs during breathing of room air (as in narcotic overdose) or of relatively low inspired concentrations of supplemental oxygen (e.g., via nasal cannulae). The importance of impaired diffusion as a hypoxemic mechanism is sometimes debated, since the transfer of O₂ from alveolus to Hb usually requires only a brief time for completion—somewhat less than the normal transit time of the erythrocyte through the capillary.⁹ Yet under many conditions that are commonly encountered, fewer capillaries are available to accept the cardiac output, so the rate at which blood flows through the lung is accelerated. Simultaneously, diffusion distances are lengthened, and the driving gradient is reduced by disease. For this reason, impaired diffusion is likely to contribute to hypoxemia occurring in the stressed patient with critical illness who receives near-normal FIO₂.

Not only is ventilation perfusion (V/Q) imbalance the most common contributor to clinical hypoxemia but it is also the mechanism least well understood among practitioners. It is the relative distribution of ventilation and perfusion that is critical to effective oxygenation. Ventilation must take place where perfusion does, or else the same levels of each that normally allow oxygenation and alveolar ventilation may produce both hypoxemia and wasted ventilation (ventilatory dead space). With respect to impaired oxygenation, this concept is perhaps best understood by considering the fall in alveolar oxygen tension that occurs as a result of regional alveolar hypoventilation. Owing to the sigmoidal shape of the oxyhemoglobin dissociation curve, excess ventilation of normal alveoli cannot fully compensate for regional desaturation elsewhere, so the net PaO₂ declines after blood from these two types of unit admix in the pulmonary veins. Like hypoxemia due to low FIO₂, hypoventilation, and diffusion impairment, hypoxemia resulting from V/Q imbalance responds to supplementation of inspired oxygen. Poor ventilation of a given lung unit can be compensated by raising the O₂ concentration of the inspired gas it receives.

Whereas the relationship of FIO₂ to PaO₂ is more or less linear for the first three oxygen-responsive mechanisms already covered, the response to oxygen supplementation for V/Q imbalance depends on the distribution of abnormal V/Q units contributing to the problem.¹⁰ Hypoxemia due to a relatively small number of lung units with very low V/Q characteristics may not respond noticeably to supplemental oxygen unless a very high FIO₂ is employed. Conversely, a lung comprised predominantly of lung units with mild V/Q impairment tends to respond in more linear fashion (Figure 44-2). It is also possible to convert very poorly ventilated lung units into airless, unventilated units with inspired gas having a very high FIO₂, owing to replacement of unabsorbable nitrogen with diffusible oxygen, leading to the unit’s contraction and eventual collapse as this process continues below the closing volume of the compromised region (absorption atelectasis).¹¹ Unless compensation by hypoxic vasoconstriction is complete, raising FIO₂ can paradoxically increase shunt even as it improves O₂ transfer in units that remain patent.

Figure 44-2 Influence of severity of ventilation/perfusion inequality on the FIO₂/arterial PO₂ relationship.

Arterial PO₂ rises linearly in normal and mildly affected lungs, whereas very high inspired oxygen fractions may be necessary to raise arterial PO₂ when the V/Q abnormality is severe.

Given the importance of matching blood flow to ventilation, it is not surprising that several mechanisms have developed to effect pulmonary microvascular regulation. Autonomic control, although less prominent and less precise than in the peripheral vasculature, is important nonetheless. Severe head injury, for example, can cause dysregulation and hypoxemia via this mechanism.¹² Local acidosis, such as that existing in poorly ventilated areas, tends to vasoconstrict the pulmonary arterial microvessels. The strength of this reflex, however, pales before that of hypoxic pulmonary vasoconstriction, which for most individuals is a well-developed protection against the consequences of perfusing underventilated areas.¹³ These mechanisms may be overpowered by pathologic processes or by pharmacologic interventions. For example, local release of inflammatory mediators or use of certain vasoactive drugs (e.g., nitroprusside) may counter these protective reflexes,¹⁴ and an abrupt rise of pulmonary artery pressure may overwhelm them.

Shunting occurs when systemic venous blood is not brought into close proximity with the inspired gas. Shunt can originate in the heart (e.g., through a patent communication at the atrial or ventricular level). Rarely, direct venous-to-arterial transfer occurs through micro- or macrovascular defects known as pulmonary arteriovenous fistulae. Such communications are encountered in relatively common diseases such as hepatic cirrhosis as well as in other settings, exemplified by the heritable Osler-Weber-Rendu abnormality. Diseases that affect the lung parenchyma are much more common causes of shunt than these cardiovascular disorders. Filling of the airspaces with fluid (e.g., edema) or cellular infiltrate (e.g., pneumonia) prevents effective gas-blood interchange. Inflammatory conditions may inhibit hypoxic vasoconstriction, worsening arterial hypoxemia, as does hypocapnic alkalosis.¹⁵ Collapse of lung units may occur on any anatomic scale, resulting in shunt through the affected regions. Causes for collapse vary from compression (e.g., by a pleural effusion), to disease-induced surfactant depletion or inactivation, to airway plugging, such as by retained secretions or a misplaced endotracheal tube. Sustained reversal of atelectasis requires attention to the inciting cause as well as recruitment of the problem area by deep lung expansion. Pure oxygen breathing will not improve hypoxemia due to shunting. Conversely, reduction of FIO₂ will not cause shunt-related hypoxemia to worsen and may spare ventilated areas the exposure to potentially toxic concentrations of inspired O₂ (Figure 44-3).

Figure 44-3 Effect of shunt percentage on arterial PO₂ for a range of FIO_2.

When shunt percentage exceeds 35% to 40%, variations of FIO₂ only modestly affect arterial PO₂. Moreover, because the risk of oxygen toxicity rises hyperbolically with inspired oxygen concentration, reductions of FIO₂ from 1.0 to 0.8 may yield benefit with only marginal impact on arterial oxygenation.

In the clinical setting, shunting due to collapse of unstable alveolar units can be addressed by increasing transpulmonary pressure after they are reopened. Raising mean alveolar pressure by elevating end-expiratory airway pressure (positive end-expiratory pressure [PEEP]), extending the inspiratory time fraction or changing body position can be effective once the collapsible alveoli are reopened (recruited) to become part of the communicating airspace. Available evidence does not clearly indicate the best method to select PEEP in patients with hypoxemic respiratory failure. If recruitment potential is low, an increase in PEEP will have marginal effects on shunt and arterial oxygen tension. Simultaneously, higher PEEP may contribute to overdistention of open alveoli, increasing the risk of ventilator-induced lung injury (VILI) and dead-space formation as pulmonary blood flow is redirected to less well-ventilated regions. PEEP may adversely affect arterial oxygen tension in the presence of unilateral or asymmetric lung disease or when PEEP impairs venous return, limits oxygen delivery, and obligates oxygen extraction.

While the benefit of PEEP in patients with refractory hypoxemia depends on the potential for alveolar recruitment, not providing PEEP to a recumbent patient is usually inappropriate because of the associated positional loss of resting lung volume. In the early stage of hypoxemic respiratory failure, a PEEP setting of 8 to 15 cm H₂O is suitable for most patients. Higher levels of PEEP should be used when a greater potential for recruitment can be demonstrated to be effective in improving oxygen delivery and/or compliance of the respiratory system; however, PEEP above 24 cm H₂O is seldom required.

Modification of body position may dramatically affect shunting and blood oxygenation when the lungs are injured, especially when disease is asymmetrically distributed. Prone positioning routinely improves oxygen exchange by altering the distribution of transpulmonary pressure as it modifies chest wall compliance and allows the heart to sink to a dependent position that does not compress the lungs. The dorsal lung zones, which are generally the best perfused, tend to reopen when prone. Drainage of secretions and lymphatic efficiency may also improve (Figure 44-4).

Figure 44-4 Lung volumes in various body positions.

Compression by the abdominal viscera compresses the dependent (dorsal) lung in the supine position, decreasing the functional residual capacity (FRC). Changing body position modifies both chest wall compliance and resting lung volume. RV, residual volume; TLC, total lung capacity.

With healthy lungs, variations in mixed venous oxygen content do not influence PaO₂ perceptibly; recharging of desaturated Hb with oxygen takes place at the alveolar-capillary junction, even during exercise. In the presence of shunt or very low V/Q units, however, the influence of mixed venous oxygen content may be profound because of its admixture with well-oxygenated pulmonary venous blood. Because mixed venous O₂ content is influenced primarily by the ratio of oxygen consumption to oxygen delivery, hypoxemia may be at least partially alleviated by reducing O₂ demand or improving O₂ delivery. The equation relating these variables, which is derived by rearrangement of the Fick equation for oxygen, is:

On-line measurements of with a fiberoptic Swan-Ganz catheter enable venous desaturation to be detected and monitored.

Venous oxygen saturation is a clinical tool to evaluate the relationship between oxygen uptake and delivery for the whole body. In the absence of pulmonary artery catheter–derived mixed venous oxygen saturation (), the central venous oxygen saturation (ScvO₂) is increasingly being used as an imprecise but convenient surrogate measure. Central venous catheters are simpler to insert, less expensive, and associated with fewer complications than pulmonary artery catheters. Blood sampling through central venous catheters allows measurement of ScvO₂ or even continuous monitoring if a fiberoptic oximetric catheter is used. The normal range for is 68% to 77%, and ScvO₂ is considered to be approximately 5% above these values.^16,17

A decrease in Hb is associated with a decrease in oxygen delivery when cardiac output remains unchanged, since oxygen delivery is the product of cardiac output and arterial oxygen content. A decrease in Hb is one of four determinants responsible for a decrease in (or ScvO₂). Anemia can act alone or in combination with hypoxemia, increased oxygen consumption, or reduced cardiac output. When oxygen delivery decreases, oxygen consumption is maintained (at least initially) by an increase in oxygen extraction (O₂ER). O₂ER and are linked by a simple equation:

In human studies, dysoxia is usually present when falls below 45%. Tissue oxygen privation may occur at higher levels of when oxygen extraction is impaired. Ideally, efforts to boost cardiac output (by intravenous fluids or inotropes), Hb, and/or arterial oxygen saturation return to levels above 65% (or ScvO₂ to ≥70%).^16,18

Relationship of PO₂ to Blood O₂ Content

Even though the oxyhemoglobin dissociation relationship is implicitly used for clinical decision making, there are important nuances (Figure 44-5). Over the clinically relevant range, the oxyhemoglobin dissociation curve is highly nonlinear, so that a drop of a few percentage points in SaO₂ over the 95% to 100% interval reflects a much larger change in PaO₂ than does a similar decrement that occurs over the 80% to 85% interval. Pulse oximeters record the relative absorption of light by oxyhemoglobin and deoxyhemoglobin. For a fixed value of Hb, the O₂ saturation parallels its relative O₂ content, but a high saturation guarantees neither its total O₂ content nor the adequacy of tissue O₂ delivery. For example, a patient may have a “full” SaO₂ after inhaling a high concentration of carbon monoxide, and yet directly measuring arterial oxygen content per deciliter of blood (e.g., using a co-oximeter) may demonstrate profound arterial O₂ depletion (see Figure 44-5). Moreover, a patient in circulatory shock may maintain a perfectly normal SaO₂ despite serious O₂ privation. Because cyanide blocks the uptake of oxygen by the tissues, O₂ consumption is low in cyanide poisoning, even though arterial and mixed venous saturations are normal or increased. It is occasionally forgotten that arterial oxygen saturation bears no direct relationship to the adequacy of ventilation; a patient breathing a high inspired concentration of oxygen will maintain a nearly normal SaO₂ for a brief period in the face of a full respiratory arrest.

Figure 44-5 Relationship of PaO₂ to blood oxygen content (CaO₂).

The oxyhemoglobin dissociation curve normally plateaus at a PO₂ of approximately 100 mm Hg (upper solid line). Alkalosis and hyperthermia (A) shift the relationship up and to the left, whereas acidosis and hyperthermia (B) shift it downward to the right. Carbon monoxide causes tighter binding of oxygen to hemoglobin (Hb) but reduces the capacity of Hb to bind oxygen.

Controversy has surrounded the concept of supply dependency of oxygen consumption for patients who have sustained trauma, massive surgery, or sepsis. Prognosis in these conditions is somewhat better for critically ill patients in whom higher oxygen delivery is manifest. By inference, it has been suggested that in these settings, supranormal oxygen delivery is needed to satisfy the oxygen demands of key vital organs. There is little doubt that prompt and vigorous resuscitation must be carried out, or that patients who do not spontaneously generate sufficient oxygen delivery or who cannot extract oxygen effectively have a worse prognosis than other patients undergoing the same stress who do. But it is inappropriate to sustain oxygen delivery at supranormal values in critically ill patients. Some data even suggest potential harm.¹⁸

A multicenter Italian trial demonstrated that aggressive fluid administration toward supranormal values for oxygen delivery conferred no routine benefit for patients in the medical/critical care unit.¹⁹ For nonmoribund patients with sepsis and/or acute respiratory distress syndrome (ARDS), supply dependency may not, in fact, exist. Therefore, without better evidence, maximizing oxygen delivery cannot be accepted as a goal for circulatory support in patients admitted to the ICU. An often cited study of patients in septic shock provides strong evidence in support of aggressive, early resuscitation in improvement of outcome. This trial was directed at early normalization of central venous oxygen saturation rather than a supranormal physiologic response.²⁰

Two studies emanating from a large National Institutes of Health (NIH)-sponsored multicenter trial provide additional data in regard to fluid resuscitation and end-organ function. In the first, fluid management protocols compared the impact of central venous versus pulmonary artery catheter monitoring in patients sustaining acute lung injury.¹⁷ Mortality in the first 60 days was similar in patients whose fluid management was guided by data from central venous and pulmonary arterial catheters. Pulmonary artery catheter–guided therapy did not improve outcomes for patients in shock at the time of enrollment in the study. There were no differences between groups in renal or pulmonary function or the use of other end-organ support. Patients receiving pulmonary artery catheterization experienced approximately twice as many catheter-related complications (mainly arrhythmias). In the second article from this seminal study, conservative and liberal strategies for fluid management were compared in patients with acute lung injury. In this trial, the difference in fluid administration was approximately 7 liters over 7 days. The rate of death at 60 days was comparable between patients receiving a conservative fluid administration strategy and a more liberal fluid administration strategy. Conservative fluid management was associated with improved pulmonary function, reduced duration of mechanical ventilation, and shorter ICU stay. These outcomes were achieved without increasing the incidence or severity of nonpulmonary organ failure.²¹

Assessing the Efficiency of Oxygen Exchange

Mean alveolar oxygen tension (PAO₂) must first be computed to judge the efficiency of gas exchange across the lung. The ideal PAO₂ is obtained from the modified alveolar gas equation:

where R is the respiratory exchange ratio, and PIO₂ is the inspired oxygen tension adjusted for FIO₂ and water vapor pressure at body temperature (47 mm Hg at 37°C). Therefore,

Under steady-state conditions, R normally varies from around 0.7 to 1.0, depending on the mix of metabolic fuels (see later discussion). When the same patient is monitored over time, R generally is assumed to be 0.8 or neglected entirely. Under most clinical conditions, the alveolar gas equation can be simplified to:

For example, at sea level with a normally ventilated patient breathing room air:

Simplified Measures of Oxygen Exchange

Several pragmatic approaches have been taken to simplify bedside assessment of O₂ exchange efficiency. The first is to quantitate P(A-a)O₂ during the administration of pure O₂. After a suitable wash-in time (5-15 minutes depending on the severity of the disease), shunt (uncontaminated by V/Q mismatch) accounts for the entire P(A-a)O₂. Furthermore, if Hb is fully saturated with O_2, dividing the P(A-a)O₂ by 20 approximates shunt percentage (at FIO₂ = 1). As pure O₂ replaces alveolar nitrogen, some patent but poorly ventilated units may collapse—the process of absorption atelectasis.¹¹ Moreover, because shunt percentage is affected by changes in cardiac output and mixed venous O₂ saturation, these simplified measures may give a misleading impression of changes within the lung itself.

The PaO₂/FIO₂ (or “P/F”) ratio is a convenient and widely used bedside index of oxygen exchange that attempts to adjust for fluctuating FIO₂. However, although simple to calculate, this ratio is affected by changes in PEEP and variations in and does not remain equally sensitive across the entire range of FIO₂—especially when shunt is the major cause for admixture. Another easily calculated index of oxygen exchange properties, the PaO₂/PAO₂ (or “a/A”) ratio, offers similar advantages and disadvantages as FIO₂ is varied. Like the P/F ratio, it is a useful bedside index that does not require blood sampling from the central circulation but loses reliability in proportion to the degree of shunting. Furthermore, in common with all measures that calculate an “ideal” PAO_2, even the a/A ratio can be misleading when fluctuations occur in the primary determinants of (Hb and the balance between oxygen consumption and delivery).

None of the indices discussed thus far account for changes in the functional status of the lung that result from alterations in PEEP, auto-PEEP, or other techniques for adjusting average lung volume (e.g., inverse ratio ventilation, lateral or prone positioning). If the objective is to categorize the severity of disease or to track the true O₂ exchange status of the lung in the face of such interventions, the P/F ratio falls short. The oxygenation index (OI):

takes mean airway pressure (mean Paw) resulting from PEEP and inspiratory time fraction into account. This calculation has gained popularity in neonatal and pediatric practice but has yet to be widely used in adult critical care. Although preferable to the unadjusted P/F ratio, this index too is imperfect; mean airway pressure and FIO₂ bear complex and nonlinear relationships to PaO₂ when considered across their entire ranges.

Physiologic Effects of CO₂

Carbon dioxide, the major waste product of oxidative metabolism, is a relatively well-tolerated gas. Apart from its key role in regulation of ventilation, the clinically important effects of CO₂ relate to changes in cerebral blood flow, pH, and adrenergic tone. Hypercapnia dilates the cerebral vessels and hypocapnia constricts them—a point of importance for patients with raised intracranial pressure. Acute increases in CO₂ depress consciousness, probably as the result of intraneuronal acidosis. Slowly developing increases in CO₂ can be easily endured, presumably because buffering has time to occur. Nonetheless, a higher PaCO₂ signifies alveolar hypoventilation which tends to cause associated decreases in alveolar and arterial PO₂. With hypoxemia and acidosis compensated by supplemental oxygen and compensatory retention of bicarbonate, some outpatients with PaCO₂ levels that chronically exceed 90 mm Hg continue to lead active lives. Conversely, patients with renal insufficiency lack the ability to buffer carbonic acid and tolerate hypercapnia poorly.

The adrenergic stimulation that accompanies acute hypercapnia causes cardiac output to rise and peripheral vascular resistance to increase. During acute respiratory acidosis, these effects may partially offset those of hydrogen ion on cardiovascular function, allowing better tolerance of lower pH than with metabolic acidosis of a similar degree. Constriction of glomerular arterioles also occurs by adrenergic stimulation, producing oliguria in some patients. Plethora, diaphoresis, muscular twitching, asterixis, and seizures may be observed at extreme levels of hypercapnia in patients made susceptible by electrolyte or neural disorders. Prompted by a favorable experience with “permissive hypercapnia” on important clinical outcomes of life-threatening asthma²² and ARDS,²³ considerable attention has been directed toward the beneficial actions of CO₂ as an antioxidant and antiinflammatory agent.^24,25 It is conceivable that in selected circumstances, hypercapnia may not only be acceptable but desirable.

The major cardiovascular effects of acute hypocapnia relate to alkalosis and diminished cerebral blood flow.²⁶ Abrupt lowering of PaCO₂ reduces cerebral blood flow and raises neuronal pH, altering cortical and peripheral nerve function. Lightheadedness, circumoral and fingertip paresthesia, and muscular tetany can result in this setting. Rarely, sudden major reductions of PaCO₂ (e.g., shortly after initiating mechanical ventilation) produce life-threatening arrhythmias and seizures, especially in those patients with elevated levels of serum bicarbonate. Because of the importance of adrenergic compensation for the vasodilatory effects of hypercapnic acidosis, hemodynamic manifestations of acute hypercapnia are more profound in the presence of β- and/or α-adrenergic blockade.

CO₂ Production and Storage

The quantity of CO₂ produced is a function of oxygen consumption and any CO₂ that is liberated in the buffering of hydrogen ion. The metabolic exchange ratio, R, varies with the mix of metabolic fuels, with carbohydrate, protein, and fat associated with ratios of 1.0, 0.7, and 0.6 respectively. CO₂ is both more diffusible and more soluble than O₂, and most CO₂ carried in the blood is in dissolved form. A smaller but very significant proportion of CO₂ is bound within the erythrocyte as bicarbonate through the action of carbonic anhydrase.

As reflected by the relatively small difference between systemic venous (45 mm Hg) and arterial (40 mm Hg) concentrations, only about one-eighth of circulating CO₂ is discharged as blood passes through the lungs. Yet, because the amount dissolved CO₂ is a linear function of its partial pressure (PCO₂), large quantities of CO₂ can be efficiently extracted from relatively small quantities of blood through a gas-permeable membrane purged on its opposite side by fresh gas. This is the principle behind passive and active extrapulmonary CO₂ removal devices now commonly deployed in critical care. The ability of these devices to extract CO₂ is comparatively great relative to their capacity for oxygen loading, as the latter can only work with a potential Hb saturation difference between 25% and 50%. (As already noted, usually ranges between 50 and 75%).

Body stores of carbon dioxide are far greater than those of oxygen. When breathing room air, only about 1.5 L of O₂ are stored (much of it in the lungs), and some of this stored O₂ remains unavailable for release until life-threatening hypoxemia is underway. Although breathing pure O₂ can fill the alveolar compartment with an additional 2 to 3 L of oxygen (a safety factor during apnea or asphyxia), these O₂ reserves are still much less than the 120 L or so of CO₂ normally stored in body tissues. Because of limited oxygen reserves, PaO₂ and tissue PO₂ change rapidly during apnea at a rate that is highly dependent on FIO₂.

Carbon dioxide stores are held in several forms (dissolved, bound to protein, fixed as bicarbonate) and distributed in compartments that differ in their volumetric capacity and ability to exchange CO₂ rapidly with the blood.²⁷ Well-perfused organs constitute a small reservoir for CO₂ that is capable of quick turnover; skeletal muscle is a larger compartment with sluggish exchange, and bone and fat are high capacity chambers with very slow filling and release. From a practical point of view, the existence of large CO₂ reservoirs with different capacities and time constants of filling and emptying means that equilibration to a new steady-state PaCO₂ after a step change in ventilation (assuming a constant rate of CO₂ production, VCO₂) takes longer than generally appreciated—especially for step reductions in alveolar ventilation (Figure 44-6). With such a large capacity and only a modest rate of metabolic CO₂ production, the CO₂ reservoir fills rather slowly, so that PaCO₂ rises only 6 to 9 mm Hg during the first minute of apnea and 3 to 6 mm Hg each minute thereafter. Depletion of this reservoir can occur at a considerably faster rate.

Figure 44-6 Effect of step changes of ventilation on PaCO₂.

A stepped increase in ventilation will cause PaCO₂ to fall in approximately exponential fashion. A stepped decrease in ventilation will cause PaCO₂ to approach equilibrium exponentially at a slower rate that is influenced by the magnitude of CO₂ storage capacity and CO₂ production.

Measurement of CO₂ excretion is valuable for metabolic studies, computations of dead-space ventilation, and evaluation of hyperpnea. Estimates of CO₂ production are representative when the sample is collected carefully in the steady state over adequate time. The rate of CO₂ elimination is a product of minute ventilation () and the expired fraction of CO₂ in the expelled gas. If gas collection is timed accurately and the sample is adequately mixed and analyzed, an accurate value for excreted CO₂ can be obtained. However, whether this value faithfully represents metabolic CO₂ production depends on the stability of the patient during the period of gas collection—not only with regard to VO₂, but also in terms of acid-base fluctuations, perfusion constancy, and ventilation status with respect to metabolic needs. During acute hyperventilation or rapidly developing metabolic acidosis, for example, the rate of CO₂ excretion overestimates metabolic rate until surplus body stores of CO₂ are washed out or bicarbonate stores reach equilibrium. The opposite situation occurs during abrupt hypoventilation or transient reduction in cardiac output.

Efficiency of CO₂ Exchange

The volume of CO₂ produced by the body tissues varies with metabolic rate (and is affected by conditions such as fever, pain, agitation, and sepsis). In the mechanically ventilated patient, many vagaries of CO₂ flux can be eliminated by controlling ventilation and quieting muscle activity with deep sedation with or without paralysis. PaCO₂ must be interpreted in conjunction with the . For example, the gas exchanging ability of the lung may be unimpaired even though PaCO₂ rises when reduced alveolar ventilation is the result of diminished respiratory drive or marked neuromuscular weakness. As already noted, alveolar and arterial CO₂ concentrations respond quasi-exponentially after step changes in ventilation, with a half-time of about 3 minutes during hyperventilation, but a slower half-time (16 minutes) during hypoventilation.²⁸ These differing time courses should be taken into account when sampling blood gases after making ventilator adjustments.

Dead Space

The physiologic dead space (VD) refers to the “wasted” portion of the tidal breath that fails to participate in CO₂ exchange. A breath can fail to accomplish CO₂ elimination either because fresh (CO₂-free) gas is not brought to the alveoli or because fresh gas fails to contact systemic venous blood. Thus, tidal ventilation is wasted whenever CO₂-laden gas is recycled to the alveoli with the next tidal breath. Alternatively, a portion of the tidal volume is wasted if fresh gas distributes to inadequately perfused alveoli so that CO₂-poor gas is exhausted during exhalation (Figure 44-7). If this concept is understood, it becomes clear why VD should not be considered as a composite of physical volumes. Nonetheless, wasted ventilation traditionally is characterized as the sum of the “anatomic” (or “series”) dead space, and the “alveolar” dead space. Because the airways fill with CO₂-containing alveolar gas at the end of the tidal breath, the physical volume of the airways corresponds rather closely to their contribution to wasted ventilation (the series or “anatomic” dead space) provided that mixed alveolar gas is similar in composition to the gas within a well-perfused alveolus. This is almost true for a quietly breathing normal subject in whom the alveolar dead space (poorly perfused alveolar volume) is very small. When the lung parenchyma is well aerated and well perfused, the anatomic dead space is relatively fixed at approximately 1 mL per pound (0.4 kg) of body weight.²⁹ Of note, patients with endotracheal tubes and tracheostomies have less series dead space, while those with attached breathing apparatus may have more.

Figure 44-7 Concept of ventilatory dead space.

A, Wasted ventilation (dead space) develops as the result of inadequate perfusion (alveolar compartment A, upper panel) or B, from the failure of ventilation to eliminate carbon dioxide from the conducting airways (lower panel). In both instances efforts expended in ventilation do not result in effective CO₂ elimination from the affected lung units.

Anatomic dead space becomes an important concern at very low tidal volumes. For patients with lung disease that affects the lung parenchyma, and those ventilated at pressures that overinflate some lung units, alveolar dead space predominates. In these settings, the lung is composed of well-perfused and poorly perfused units, so the mixed alveolar gas within the airways at end exhalation has a CO₂ concentration lower than that of pulmonary arterial blood.

For normal subjects, dead space increases with advancing age and body size and is reduced modestly by recumbency, extended breath holding, and decelerating inspiratory gas flow patterns. External apparatus attached to the airway that remains unflushed by fresh gas may add to the series dead space, whereas tracheostomy reduces it. The supine position reduces dead space by decreasing the average size of the lung and by increasing the number of well-perfused lung units.

Numerous diseases increase VD. Destruction of alveolar septae, low-output circulatory failure, pulmonary embolism, pulmonary vasoconstriction or vascular compression, and mechanical ventilation with high tidal volumes or PEEP are common mechanisms that often act in combination to increase VD.

Dead-Space Fraction

In the setting of parenchymal lung disease, dead space varies in proportion to tidal volume over a remarkably wide range. Series dead space tends to remain fixed but generally constitutes a small percentage of the total physiologic VD, and is overwhelmed by the alveolar dead space component. Therefore, except at very small tidal volumes or when extensive tidal recruitment of collapsed units occurs, the fraction of wasted ventilation (VD/VT) tends to remain relatively constant as the depth of the breath varies. The dead-space fraction can be estimated from analyzed specimens of arterial blood and mixed expired (PECO₂) gas:

where PECO₂ is the CO₂ concentration in mixed expired gas. (This expression is known as the Enghoff-modified Bohr equation.) As already noted, PECO₂ can be determined on a breath-by-breath basis if exhaled volume is measured simultaneously. Alternatively, exhaled gas can be collected over a defined period. The PCO₂ of gas exiting a mixing chamber attached to the expiratory line provides a continuous “rolling average” value. In collecting the expired gas sample during pressurized ventilator cycles, an adjustment should be made for the volume of any sampled gas stored in the compressible portions of the ventilator circuit.

In healthy persons, the normal VD/VT during spontaneous breathing varies from roughly 0.35 to 0.15, depending on the factors noted earlier (e.g., position, exercise, age, tidal volume, pulmonary capillary distention, breath holding). In the setting of critical illness, however, it is not uncommon for VD/VT to rise to values that exceed 0.7. Indeed, increased dead-space ventilation usually accounts for most of the increase in the requirement and CO₂ retention that occur in the terminal phase of acute hypoxemic respiratory failure. High and increasing dead-space values may portend an adverse outcome in ARDS.³⁰ Conversely, improving dead space has been reported as a propitious sign in prone positioning.³¹ In addition to pathologic processes that increase dead space, changes in VD/VT occur during periods of hypovolemia or overdistention by high airway pressures. This phenomenon often is apparent when progressive levels of PEEP are applied to support oxygenation. Conversely, recruitment of functioning lung tissue tends to reduce the dead-space fraction. Examination of the airway pressure tracing under conditions of controlled, constant inspiratory flow ventilation may demonstrate concavity or a clear point of upward inflection, indicating overdistention, accelerated dead-space formation, and escalating risk of barotrauma. Small reductions in PEEP or tidal volume may then dramatically reduce peak cycling pressure and VD/VT.

PaCO₂ is influenced by CO₂ production, minute ventilation, and the ventilatory dead space according to the following equation:

In a different form:

Here PACO₂, VA and PB refer to alveolar PCO₂, alveolar ventilation, and barometric pressure, respectively. In view of the hyperbolic relationship of PaCO₂ to alveolar ventilation (Figure 44-8), it can be understood that relatively small changes of effective ventilation can profoundly influence PaCO₂ and pH when alveolar ventilation is low and PaCO₂ is high. Once PaCO₂ has climbed to approximately double its normal value, fluctuations of pH and PaCO₂, with their attendant adverse effects on hemodynamics and pulmonary artery pressure, place the critically ill patient at increased risk. Moreover, ventilatory drive is blunted when PaCO₂ values are increased, while small changes in ventilation may cause PaCO₂ to plummet. In the context of increased PaCO₂, it is interesting to consider tracheal gas insufflation (TGI), a novel technique in which fresh gas is injected near the carina so as to wash the proximal airway free of carbon dioxide during exhalation and thereby improve ventilation efficiency with little effect in inspiratory airway pressure.³² In the setting of extreme hypercapnia, the ordinarily small improvement in alveolar ventilation that TGI affords proves valuable in reducing PaCO₂ and its attendant consequences.

Figure 44-8 Relationship of alveolar ventilation and alveolar CO_2.

Despite the varying conditions depicted, the hyperbolic function relating alveolar ventilation implies that small changes of effective ventilation translate into marked changes of alveolar PCO₂ and consequently of PaCO₂ and pH.

Monitoring of Exhaled Gas

Capnography analyzes the CO₂ concentration of the expiratory air stream, plotting CO₂ concentration against time or, more usefully, against exhaled volume. Although most capnometers in clinical use currently display PCO₂ as a function of time, much of the attention will focus on the CO₂ versus volume plot because it provides more information of clinical value. After anatomic dead space has been cleared, the CO₂ tension rises progressively to its maximal value at end exhalation, a number that reflects the CO₂ tension of mixed alveolar gas. For normal subjects, the transition between phases of the capnogram is sharp, and once achieved, the alveolar plateau rises only gently. Furthermore, when ventilation and perfusion are evenly distributed, as they are in healthy subjects, end-tidal PCO₂ (PETCO₂) closely approximates PaCO₂, with PETCO₂ normally underestimating PaCO₂ by 1 to 3 mm Hg. The difference between PETCO₂ and PaCO₂ widens when ventilation and perfusion are matched suboptimally, so that alveolar dead-space gas admixes with CO₂-rich gas from well-perfused alveoli.

When plotted against a volume axis, as opposed to the more commonly encountered time axis, the capnogram offers data of considerable clinical value. Inspection of such tracings can yield estimates for the “anatomic” (Fowler) dead space, as well as for the end-tidal and mixed expired CO₂ concentrations (Figure 44-9). Knowing the barometric pressure, the mixed expired value can be expressed as a percentage of the exhaled volume, which is also immediately available from the tracing. If the VT remains constant, the product of the PECO₂ : PB ratio and is the VCO_2, and the mixed expired CO₂ concentration can be used in the Enghoff-modified Bohr equation to estimate the physiologic dead-space fraction.

Figure 44-9 Capnogram with expired PCO₂ plotted against exhaled volume.

Important data can be derived from the expired capnogram obtained under steady-state passive conditions: anatomic or “Fowler” dead-space; physiologic dead space (the difference [B] between PaCO₂ and mixed expired CO₂, PECO₂, expressed as a fraction of PaCO₂); slope of the “alveolar plateau” (A), an indicator of the heterogeneity of ventilation; and an estimate of CO₂ production (obtained from the product of mixed expired CO₂ referenced to total barometric pressure and exhaled volume).

As with other monitoring techniques, exhaled CO₂ values must be interpreted cautiously. The normal capnogram comprises an ascending portion, a plateau, a descending portion, and a baseline. In disease, the sharp distinctions between phases of the capnogram as well as the slopes of the composite segments are blurred. Moreover, failure of the airway gas to equilibrate with gas from well-perfused alveoli invalidates PETCO₂ as a reflection of PaCO_2, especially as respiratory frequency fluctuates; the PECO₂ per cycle, however, remains valid under these conditions. End-tidal PCO₂ gives a low range estimate of PaCO₂ in virtually all clinical circumstances, so that a high PETCO₂ strongly suggests hypoventilation. Abrupt changes in PETCO₂ may reflect such acute processes as aspiration or pulmonary embolism if the and breathing pattern (f, VT, and I : E ratio) remain unchanged. Although breath-to-breath fluctuations in PETCO₂ can be extreme, the trend of PETCO₂ over time helps identify underlying changes in CO₂ exchange.

Key Points