Show Summary Details
Page of

Gas exchange assessment in the critically ill 

Gas exchange assessment in the critically ill
Chapter:
Gas exchange assessment in the critically ill
Author(s):

Peter D. Wagner

DOI:
10.1093/med/9780199600830.003.0076
Page of

PRINTED FROM OXFORD MEDICINE ONLINE (www.oxfordmedicine.com). © Oxford University Press, 2020. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in Oxford Medicine Online for personal use (for details see Privacy Policy and Legal Notice).

date: 29 November 2020

Key points

  • The cornerstone of insightful monitoring is directly sampled systemic arterial blood, but there are both more and less complex and invasive methods.

  • Pulse oximetry is the least invasive measure, but limitations include poor signal and inability to track substantial change in PO2 at higher SaO2 (flat O2-Hb curve).

  • The PaO2/FiO2 ratio is used in the ICU so measurements at different FiO2 can be compared, although it is not completely independent of FiO2 in the critically ill.

  • The alveolar-arterial PO2 difference corrects the alveolar PO2 for values of PaCO2 different than 40 mmHg. It becomes less useful as FiO2 is increased if V ˙ A / Q ˙ inequality is the main problem.

  • Using oximeter-tipped catheters placed in the pulmonary artery, mixed venous saturation is continuously available. A high arteriovenous difference implies high O2 demand in relation to supply, and vice versa but correct interpretation depends on V ˙ o 2 and cardiac output.

Introduction

Since most patients admitted to the intensive care unit (ICU) have damaged lungs, monitoring of gas exchange is critical to informed management. Monitoring can be descriptive (how abnormal is gas exchange?) and mechanistic (how may gas exchange reveal pathophysiological insights?)

The cornerstone of insightful monitoring is directly sampled systemic arterial blood, but there are both more and less complex and invasive methods. Usually, the more complex and invasive, the greater will be the insights gained, and vice versa. The key is to balance the complexities against the information content to select an approach that is sufficiently useful yet not overly complex.

Pulse oximetry

The simplest, least invasive approach is pulse oximetry, which measures arterial O2 saturation (SaO2). With immediate, continuous, low cost, non-invasive response requiring no specialty training, it is widely used to good advantage. The clinical range of use is mostly 85–100%. If SaO2 is <85–90%, direct arterial blood sampling and urgent therapeutic responses are usually required.

Oximetry does not measure arterial PCO2/pH and usually does not reveal mechanistic information. Moreover, as the O2-Hb dissociation curve shifts when pH, temperature, and PCO2 change, a given saturation may correspond to different PO2 values.

Two major limits to pulse oximetry are:

Inadequate signal: most oximeters are lightly clamped on a finger. They are easily dislodged and require adequate digital perfusion.

The shape of the O2-Hb dissociation curve: when arterial PO2 lies on the flat part of the O2-Hb curve (SaO2> ~95%), large changes in PO2 may occur from significant changes in lung function without measureable differences in SaO2.

Since the standard deviation of SaO2 is ~2%, a true SaO2 of 97% (normal) may read 95%, while a true value of 93% (abnormal) may also read 95%. It is thus problematic to know whether such a value represents poor lung function or random variation from normal.

Additionally, when saturation falls below ~70%, many oximeters will indicate an even lower (i.e. erroneous) value. This should not happen in the ICU, because saturations this low imply life-threateningly poor lung function, requiring placement of an arterial catheter to directly sample arterial blood.

Arterial blood gas sampling

Arterial blood sampling, via in-dwelling catheter or needle puncture, is the current backbone of ICU gas exchange assessment. Unlike oximetry, sampling is discrete, not continuous (continuously recording in-dwelling electrodes are under development). Sample volume can be less than 200 µL, enabling paediatric use. Analysis (blood gas analyser) takes only 1–2 minutes, and yields PO2, pH, PCO2, SaO2 [Hb], lactate, bicarbonate, and base excess, thus evaluating acid–base state and gas exchange.

Precautions in arterial blood sampling

Several important precautions are necessary to avoid errors:

  • The sample must be taken and kept without exposure to air until measured.

  • The sample syringe must be free of air (bubbles). These distort the results and should be immediately expelled from the syringe.

  • The syringe must be pre-heparinized to prevent clotting, which obstructs analysers and causes malfunction. Use just enough heparin to prevent coagulation without diluting the sample.

  • The sample should be analysed immediately. If delay is unavoidable, keep it iced to reduce white blood cell metabolism, which will reduce PO2 and raise PCO2. This is especially important when PO2 is high, since PO2 may otherwise fall 10–20 mmHg/min.

  • Aggressively rotate the sample until the moment of analysis to avoid red cell settling causing errors in [Hb] and haematocrit.

  • Document patient temperature, to understand temperature-induced shifts in the O2-Hb dissociation curve, and to correct of PO2, PCO2 and pH back to body temperature (clinical analysers usually run at 37°C). Suppose a patient’s temperature is 39°C. As the sample cools to 37°C in the analyser, the number of O2 molecules per unit volume cannot change (no air contact). Thus, SaO2 is not affected by cooling. However, the O2-Hb curve shifts leftward, and so PO2 falls. We usually need PO2 at body temperature, and so temperature-correct the value accordingly, about 6% per degree Centigrade. Thus, a PO2 of 80 mmHg would become ~90 mmHg expressed when the patient’s temperature was 39°C. Directional changes are similar for PCO2 and blood [H+], yielding a higher PCO2 and lower pH than indicated by the analyser. The opposite is true when the patient is below 37°C. Most analysers temperature-correct automatically.

  • Document inspired [O2] at sampling. A PO2 of 90 mmHg, breathing room air, is interpreted very differently from the same PO2 measured during 100% O2 breathing.

Interpretation of blood gas variables

First, ask only how different PO2, PCO2, pH, SaO2, and [Hb]/haematocrit are from normal. Two things are required—expected normal values, and analyser measurement error, to establish the 95% confidence limits (roughly, mean ± 2 SD). If the values lie outside those limits, there is at least a 95% chance that the data are abnormal. When FiO2 is elevated, (common in the ICU), normal alveolar PO2 expected at any FiO2 can be estimated from the alveolar gas equation:

Expected     alveolar     PO 2   =   F I O 2 * ( PB   P H 2 O )     P A CO 2 /R
[eqn 1]

FiO2 is inspired O2 fraction, Pb is barometric pressure; Ph2o is saturated water vapour pressure (47 mmHg at 37°C); PaCO2 is arterial PCO2 and R is respiratory exchange ratio, usually assumed to be 0.8.

In sea level room air, normal alveolar PO2 is about 100 mmHg. On 100% O2, it rises to ~660 mmHg. The relationship is essentially linear in between. Normal arterial PO2 should be a few mmHg lower than alveolar.

Clinical analysers commonly have a coefficient of variation for PO2 of ~5% (thus if PO2 = 100 mmHg, 95% confidence limits are 90–110 mmHg). If PO2 is 500 mmHg, 95% confidence range is ~450–550 mmHg.

PCO2 usually shows less variance; if PCO2 = 40 mmHg, 95% confidence limits should be 38–42 mmHg, while when pH = 7.40, 95% confidence limits should be 7.38–7.42. However, ignoring the above sampling precautions will increase uncertainty.

Arterial blood gas variables can answer more than whether the results are normal or not. Well-established equations and relationships exist for quantifying gas exchange disturbances to estimate the extent of lung damage, and responses to interventions and therapies.

Pao2/Fio2 ratio

This simple and popular ratio [1]‌ is used in the ICU so measurements at different FiO2 can be compared. Since arterial PO2 in health rises linearly with FiO2, PaO2/ FiO2 is somewhat constant as FiO2 changes. This attractive concept has limitations: Fig. 76.1 shows the PaO2 (upper panel) and PaO2/ FiO2 ratio (lower panel) over the FiO2 range from room air to 100% O2 in several computational models [2] of varying gas exchange abnormality, labelled A through G.

Fig. 76.1 PaO2 and PaO2/FiO2 ratio change with inspired O2 concentration between room air and 100% O2 in normal lungs (A), in lungs with moderate to severe . 
  
 
    
     
      V
      ˙
     
  
 
    A
   /
    Q
    ˙
   

 inequality (B–D) and in lungs with varying amounts of shunt (E–G) (see text for description).

Fig. 76.1 PaO2 and PaO2/FiO2 ratio change with inspired O2 concentration between room air and 100% O2 in normal lungs (A), in lungs with moderate to severe . V ˙ A / Q ˙ inequality (B–D) and in lungs with varying amounts of shunt (E–G) (see text for description).

Curve A (open squares) represents the normal lung. PaO2 rises linearly (upper panel) and PaO2/FiO2 is very high and is essentially constant. However, even here, PaO2/FiO2 is not totally independent of FiO2—small changes must not be over-interpreted.

Curves B, C, and D represent lungs with increasing heterogeneity of ventilation with respect to blood flow ( V ˙ A / Q ˙ inequality), quantified by the term SDQ (in effect, the standard deviation of the V ˙ A / Q ˙ distribution). The normal range is 0.3–0.6 [3]‌. SDQ = 1.0 represents moderate inequality as in stable chronic lung disease; SDQ = 2.0 corresponds to a very ill ICU patient. In B, C, and D, there is zero shunt. PaO2/FiO2 increases considerably and non-linearly with increasing Fio2 despite a constant amount of V ˙ A / Q ˙ inequality. Both PaO2/FiO2 itself and its change with FiO2 depend on the amount of V ˙ A / Q ˙ inequality and the FiO2, making PaO2/FiO2 somewhat difficult to interpret.

Curves E, F, and G represent lungs without V ˙ A / Q ˙ inequality, but with shunting of 10, 20, and 30%, respectively. The behaviour of PaO2/FiO2 as FiO2 is different again—the ratio first falls as FiO2 increases, before modestly rising as FiO2 approaches 1. Especially important is the region encompassed by the dashed-line box: over a wide FiO2 range (room air to 0.6), PaO2/FiO2 is largely independent of FiO2, but lungs with moderate to severe V ˙ A / Q ˙ inequality and lungs with 20–30% shunt are, in practice, unable to be distinguished.

When FiO2 = 1, no matter the severity of V ˙ A / Q ˙ inequality, PaO2/ FiO2 is very high while with shunt, the ratio falls progressively as shunt increases. Fig. 76.1 also suggests that it may be useful to assess PaO2/FiO2 at two different values of FiO2. Its behaviour as FiO2 changes allows more insight into the gas exchange defect.

The alveolar-arterial Po2 difference: A-apo2

The alveolar gas equation allows calculation of alveolar PO2, and subtracting arterial PO2 from this yields the A-aPO2. The advantage of A-aPO2 is that it corrects the alveolar PO2 for values of PaCO2 different than 40 mmHg. Thus, if a patient with normal lungs, breathing room air, hypoventilates and PaCO2 rises to 80 mmHg, alveolar PO2 is 50 mmHg (PaO2 = 150 – 80/0.8 = 50). If arterial PO2 were 48 mmHg, A-aPO2 = 50 – 48 = 2 (normal). This shows that the hypoxaemia is fully explained by hypoventilation, which is not evident from looking at PaO2 alone. Had arterial PO2 been 40 mmHg, A-aPO2 would have been 10 mmHg, and this would have suggested the presence of significant additional gas exchange defect from shunt or V ˙ A / Q ˙ inequality (or both).

While very useful in patients breathing room air, A-aPO2 becomes less useful as FiO2 is increased if V ˙ A / Q ˙ inequality is the main problem. This is because at high FiO2, A-aPO2 falls, approaching zero on 100% O2, no matter how severe the V ˙ A / Q ˙ mismatch is. However, if A-aPO2 on 100% O2 is elevated, the reason has to be a shunt, and A-aPO2 is very useful.

In between room air and 100% O2, A-aPO2 will change with FiO2 in a manner specific to the nature and extent of the gas exchange disturbance, as shown in Fig. 76.2, which uses the same data as Fig. 76.1. With V ˙ A / Q ˙ inequality (open symbols), A-aPO2 first rises as FiO2 is increased, and then falls, to nearly zero when 100% O2 is breathed. The maximal A-aPO2 and the FiO2 at which it occurs depends on the severity of the inequality as shown. This behaviour is explained by the shape of the O2-Hb dissociation curve. However, when the problem is shunt, A-aPO2 rises steadily with FiO2 as shown.

Fig. 76.2 Alveolar-arterial PO2 difference in the same six lung models of 
  
 
    
     
      V
      ˙
     
  
 
    A
   /
    Q
    ˙
   

 inequality and shunt as shown in Fig. 76.1.

Fig. 76.2 Alveolar-arterial PO2 difference in the same six lung models of V ˙ A / Q ˙ inequality and shunt as shown in Fig. 76.1.

There is an important cautionary note—Figs 76.1 and 76.2 reflect constancy of the main determining variables as FiO2 is altered. Thus, the extent of V ˙ A / Q ˙ heterogeneity and/or shunt, together with tidal volume and respiratory frequency, cardiac output, metabolic rate, [Hb], acid–base state, and temperature are all presumed to remain fixed. If any of these change over time, or from changing FiO2, the data shown in these figures must change. Absorption atelectasis, release of hypoxic pulmonary vasoconstriction, and changes in cardiac output may all occur as FiO2 is raised. Any of these may change shunt and V ˙ A / Q ˙ inequality.

Physiological shunt: Q ˙ S / Q ˙ T

Since the 1950s [4,5,6], a further, simple, analysis of blood gas data has proven useful. The calculation of physiological shunt, or ‘ Q ˙ S / Q ˙ T ’. Q ˙ denotes blood flow, s denotes shunt and T denotes total. It reflects the shunt equation, based on mass conservation, applied to a lung imagined as two compartments, one having zero ventilation. The second compartment performs all gas exchange. The shunt equation is:

Q˙ T × CaO 2   =   Q˙ S   ×   CvO 2   +   ( Q˙ T     Q˙ S )   ×   Cc O 2
[eqn 2]

Q ˙ T is total pulmonary blood flow and Q ˙ S is blood flow through the shunt compartment, which transmits blood unchanged (with mixed venous O2 concentration) to the pulmonary veins. CaO2 is systemic arterial O2 concentration, CvO2 is O2 concentration in mixed venous (pulmonary arterial) blood, and Cc’O2 is end-capillary O2 concentration of the second, ventilated compartment. Cc'O2 is computed from the O2-Hb dissociation curve based on the alveolar PO2 from the alveolar gas equation. The equation states that arterial O2 concentration is the perfusion-weighted average of O2 concentrations in the two compartments. After rearrangement:

Q ˙ s / Q ˙ T   =   [ C c O 2     C a O 2 ] / [ C c O 2     C v O 2 ]
[eqn 3]

The equation can be used at any FiO2. However at FiO2 < 1.0, Q s / Q ˙ T is termed ‘venous admixture’ or ‘physiological shunt’, while when FiO2 = 1.0, it is regarded as ‘true shunt’. This is because at FiO2 < 1.0, V ˙ A / Q ˙ inequality will contribute to Q s / Q ˙ T , while at FiO2 = 1.0, V ˙ A / Q ˙ inequality ceases to contribute and Q s / Q ˙ T does, in fact, represent only shunt (see Figs 76.1 and 76.2).

Fig. 76.3 shows Q ˙ s / Q ˙ T for the same data as in Figs 76.1 and 76.2. When the lung really does consist of two functional compartments—normal and shunt, as may happen in localized consolidation or atelectasis—the Q ˙ S / Q ˙ T model accurately depicts the extent of shunt across the whole FiO2 range. However, when there is V ˙ A / Q ˙ inequality, rather than shunt, Q ˙ s / Q ˙ T can be very high at low FiO2, but eventually falls to zero on 100% O2. Thus, the shunt equation will underestimate the true extent of V ˙ A / Q ˙ inequality at FiO2 above room air and may over-estimate true shunt at FiO2 < 1.0.

Fig. 76.3 Physiological shunt (
  
   (
    Q
    ˙
   
   s/
    Q
    ⋅
   
   T)
  
 in the same six disease models as shown in Figs 76.1 and 76.2.

Fig. 76.3 Physiological shunt ( ( Q ˙ s/ Q T) in the same six disease models as shown in Figs 76.1 and 76.2.

To use the equation, CaO2 is calculated from arterial PO2, saturation, and [Hb], and Cc'O2 is correspondingly computed from alveolar PO2. The weakness is in CvO2. To measure this directly requires a pulmonary artery catheter, which is often unavailable. Alternatively, CvO2 can be calculated from metabolic rate ( V ˙ o 2 ) and cardiac output ( Q ˙ T) using the Fick principle:

V ˙ O 2   =   Q ˙ T   ×   [ C a O 2     C v O 2 ]
[eqn 4]

Rearranging,

CvO 2 = CaO 2 V ˙ O 2 / Q ˙ T,
[eqn 5]

which can be combined with eqn 3 to yield:

Q ˙ s / Q ˙ T   =   [ C c O 2     C a O 2 ] / [ C c O 2     C a O 2   +   V ˙ O 2 / Q ˙ T ]
[eqn 6]

However, if Q ˙ T and / or V ˙ o 2 are not known, and must be assumed, application of the shunt equation can be considerably in error as Fig. 76.4 shows. The top panel shows how arterial PO2 would actually vary as cardiac output (and thus CvO2) varies for a lung with a fixed shunt that was 20% of the cardiac output, the patient always breathing 100% O2. The fall in PaO2 when cardiac output is low is due to the associated fall in mixed venous PO2 required to meet the unchanged metabolic needs of the tissues, and vice versa when cardiac output is elevated. The lower panel shows the shunt calculated from the arterial PO2 values in the upper panel if the venous O2 concentration was assumed constant at the value actually seen when cardiac output was 6 L/min. Only at that cardiac output will the correct shunt value (20%) be obtained (dashed lines). As cardiac output drops, the calculated shunt is erroneously high; when cardiac output is elevated, the shunt is underestimated. The errors may be considerable.

Fig. 76.4 (Upper panel) Arterial PO2 (PaO2) change with a fixed shunt (using an example of 20%) as cardiac output varies between 3 and 9 L/min, all other factors constant. As cardiac output falls, so too does mixed venous [O2], which in turn explains why PaO2 is lower. PaO2 may be as low as 100 mmHg or as high as 400 mmHg, depending on cardiac output. (Lower panel) Figure shows how assuming a standard arteriovenous [O2] difference (equal to that when cardiac output is 6 L/min using the same example) can cause large errors in shunt computed from the shunt equation. Only when cardiac output is 6 L/min will the equation return the correct value. There is two-fold range in apparent shunt.

Fig. 76.4 (Upper panel) Arterial PO2 (PaO2) change with a fixed shunt (using an example of 20%) as cardiac output varies between 3 and 9 L/min, all other factors constant. As cardiac output falls, so too does mixed venous [O2], which in turn explains why PaO2 is lower. PaO2 may be as low as 100 mmHg or as high as 400 mmHg, depending on cardiac output. (Lower panel) Figure shows how assuming a standard arteriovenous [O2] difference (equal to that when cardiac output is 6 L/min using the same example) can cause large errors in shunt computed from the shunt equation. Only when cardiac output is 6 L/min will the equation return the correct value. There is two-fold range in apparent shunt.

Mixed venous O2 saturation/Po2

The preceding illustrates the risk of assuming mixed venous oxygenation when computing the shunt and a rationale for knowing the correct value. Another reason for estimating mixed venous PO2 and/or saturation is to better understand the relationship between O2 supply and demand. Using oximeter-tipped catheters placed in the pulmonary artery, mixed venous saturation is continuously available, and arterial saturation is also usually available. Thus, with [Hb], the arteriovenous O2 concentration difference, CaO2—CvO2, becomes available. From the Fick principle in eqn 5, the arteriovenous difference is numerically equal to the ratio V ˙ o 2 / Q ˙ T , which is the ratio of whole-body O2 demand ( V ˙ o 2 ) to supply ( Q ˙ T) . A high arteriovenous difference implies high O2 demand in relation to supply, and vice versa, which may be clinically useful information. This concept is, however, complicated by a major uncertainty. Suppose V ˙ o 2 / Q ˙ T is high. Does that mean that demand is elevated or that supply is reduced? Measuring only the ratio cannot tell. We need the absolute value of either V ˙ o 2 or Q ˙ T to resolve this. This is important because inappropriate therapy might otherwise be applied.

Physiological dead space, Vd/Vt

Just as for shunt, a two-compartment approach can estimate how much ‘dead space’ would have to exist to explain the mixed expired PCO2 value. The concept is again to imagine the lungs as one normal compartment performing all gas exchange, and a second that is ventilated, but completely unperfused, and not participating in gas exchange (hence, the name ‘dead space’ for the ventilation associated with it). Again, we use mass conservation, applied to CO2, as follows:

V˙ E   ×   P E CO 2   =   V˙ A   ×   P A CO 2   +   V D   ×   P I CO 2
[eqn 7]

Where V ˙ E is total minute ventilation (tidal volume ( V ˙ T ) times breathing frequency), V ˙ A is ventilation of the normal compartment and Vd is ventilation of the unperfused compartment. Note that V ˙ E= V ˙ A + V D . PeCO2 is the PCO2 in the mixed exhaled gas. PaCO2 is the alveolar PCO2 of the normal compartment, and PiCO2 is the inspired PCO2, since that must be the alveolar PCO2 of the unperfused compartment. Replacing V ˙ A by V ˙ A - V D , and rearranging eqn 7 yields:

V D / V T   =   [ P A CO 2     P E CO 2 ] / P A CO 2
[eqn 8]

If, as is commonly done, we substitute PaCO2 (arterial PCO2) for PACO2, because the latter is actually very difficult to measure, we end up with:

V D / V T   =   [ P A CO 2     P E CO 2 ]   / P A CO 2
[eqn 9]

Vd/Vt is the fraction of the total ventilation passing to the unperfused compartment (without gas exchange), and thus ‘wasted’. It is a composite of three kinds of dead space: one is normal, and is the contribution from the conducting airway volume (from mouth to the last terminal bronchiole), called the anatomic dead space. The second is the tubing volume, if any, between a ventilated patient and the valve of the ventilator or fork in the ventilator tubing that separates inspired from expired gas. This includes the endotracheal or tracheostomy tube volume. The third is the ‘alveolar dead space’, which is a virtual volume of gas ventilating hypothetical alveoli that are completely unperfused. It is normally zero. Application of eqn 9 captures all three kinds inseparably as a single value, and what then has to be done is to subtract the first two kinds of dead space to arrive at the alveolar dead space, which is usually the number of interest. Tubing volume is calculated from its length and diameter. Anatomic dead space, rarely measured, is often taken as 1ml per pound (2 mL/kg), of lean body weight. It will be a little less if the pharynx is bypassed by an endotracheal or tracheostomy tube.

It is important to distinguish absolute dead space from percentage dead space. The latter is most frequently reported and, while useful, is affected directly by the tidal volume. A dead space of 200 mL with 500 mL tidal volume, yields 40% Vd/Vt, while the same dead space in a 1-L tidal volume reduces Vd/Vt to 20%. Therefore, reporting of alveolar dead space should also use absolute volume. In a normal lung, alveolar dead space (i.e. total, less anatomic, ventilator tubing volume) is essentially zero, but in lung disease with V ˙ A / Q ˙ inequality it can be substantial. In the cases presented in Figs 76.176.3, alveolar dead space is 17% when SDQ = 1.0, 35% when SDQ = 1.5, and 52% when SDQ = 2.0.

The popular perception is that only areas of above-normal V ˙ A / Q ˙ ratio contribute to alveolar dead space. However, areas of low V ˙ A / Q ˙ ratio and even shunt contribute to Vd/Vt when arterial PCO2 is used as in eqn 9 because such areas not only depress arterial PO2, they elevate arterial PCO2.

Conclusion

In conclusion, several longstanding approaches to gas exchange allow quantification of gas exchange disturbances in the ICU. Perhaps the key point is that, for all, assumptions are made that may critically affect their interpretation and lead to erroneous therapeutic decisions. This discussion is intended to illustrate those important assumptions and thereby mitigate their consequences.

References

1. Villar J, Perez-Mendez L, Blanco J, et al. (2013). A universal definition of ARDS: the PaO2/FiO2 ratio under a standard ventilatory setting—a prospective, multicenter validation study. Intensive Care Medicine, 39, 583–92.Find this resource:

2. West JB. (1969). Ventilation/perfusion inequality and overall gas exchange in computer models of the lung. Respiration Physiology, 7, 88–110.Find this resource:

3. Wagner PD, Hedenstierna G, and Bylin G. (1987). Ventilation-perfusion inequality in chronic asthma. American Reviews of Respiratory Diseases, 136, 605–12.Find this resource:

4. Riley RL and Cournand A. (1951). Analysis of factors affecting partial pressures of oxygen and carbon dioxide in gas and blood of lungs: theory. Journal of Applied Physiology, 4, 77–101.Find this resource:

5. Riley RL, Cournand A, and Donald KW. (1951). Analysis of factors affecting partial pressures of oxygen and carbon dioxide in gas and blood of lungs: methods. Journal of Applied Physiology, 4, 102–20.Find this resource:

6. Riley RL and Houston CS. (1951). Composition of alveolar air and volume of pulmonary ventilation during long exposure to high altitude. Journal of Applied Physiology, 3, 526–34.Find this resource: