Special Feature: In Search of the Holy Grail: The Ideal Index of Hypoxemia in ARDS

By Karen Johnson, PhD, RN

Over the past 50 years, our understanding of the acute respiratory distress syndrome (ARDS) has evolved, not only with respect to the pathophysiology of lung injury and hypoxemia, but also with the definition of this syndrome and its treatment. We still have a long way to go in our understanding about the variability in pulmonary gas exchange in patients with ARDS. ARDS is characterized by severe hypoxemia,1 which is an important element in the definition of the syndrome.2,3 Numerous indices have been devised to describe hypoxemia and measure its severity. The search for an ideal index of hypoxemia has been like the search for the Holy Grail.

What would be an ideal descriptor of hypoxemia? Gould and colleagues offer some insight.4 An ideal index of hypoxemia: 1) should be reliable—or, under certain conditions repeated measurements should yield similar values; 2) should measure what it’s supposed to measure; 3) should be responsive to changes in the true value of the measured physiologic parameter, and should reflect the degree of these changes with acceptable sensitivity; and 4) should provide some clinically useful diagnostic and prognostic information. In addition, another criterion of an ideal descriptor, particularly for hypoxemia, should be that the index should measure physiologic function directly, rather than the therapeutic intervention employed to support pulmonary function.5 This is particularly important when assessing hypoxemia because the index should not vary with alterations in FiO2.

Several indices have been proposed over the past fifty years to quantify hypoxemia including intrapulmonary shunt (Qs/Qt), the alveolar-arterial gradient (PAO2 - PaO2), the arterial/alveolar ratio (PaO2/PAO2), and the PaO2/FiO2 ratio. The purpose of this special feature is to review these indices and identify the index that is the most ideal descriptor of hypoxemia in ARDS.

Measurement of Intrapulmonary Shunt

Intrapulmonary shunt (Qs/Qt) is often identified as the gold standard for assessing impairment of pulmonary gas exchange in critically ill patients.6-10 The value of measuring Qs/Qt is to identify alterations in pulmonary gas exchange and to determine the degree to which the lung deviates from ideal as an oxygenator of blood. Measurements of Qs/Qt identify contributions to hypoxemia as the result of ventilation/perfusion mismatch, diffusion limitation, and true shunting.11

Qs/Qt can be measured by calculating the difference between the content of fully oxygenated pulmonary capillary (CCO2) and arterial blood (CaO2) divided by a difference between fully oxygenated pulmonary capillary blood (CCO2) and mixed venous blood (CvO2) according to the formula in Table 1.

Measurement of Qs/Qt has been favored by several investigators to assess hypoxemia in ARDS.9,10,12 Information needed to calculate oxygen content in the Qs/Qt formula is available only in patients who have pulmonary and peripheral artery catheters in place for the determination of CaO2 and CvO2. This has forced a situation where clinicians have had to estimate arterial oxygen content difference, which is often misleading.12 Estimation of Qs/Qt based on assumed arterial-venous difference can vary widely in ARDS.7,9 There appears to be a good correlation between measured and estimated Qs/Qt (r2 = .92) in patients with arterial-venous oxygen content differences greater than 4.5 mL/dL. However, the investigators reporting these results determined that the average estimated shunt (27.8%) was one and a half times greater than the measured shunt (18.7%).10

Qs/Qt is a complex formula to calculate, although most ICU bedside monitoring systems have the capability to calculate Qs/Qt once mixed venous and arterial blood gas data are available and entered into the system. It is time consuming to obtain, complex to calculate, and costly.

Oxygen Tension-Derived Indices

Complexity of the Qs/Qt, necessity for frequent sampling of arterial and mixed venous blood, and the need for a pulmonary artery catheter led to the development of a number of indices to estimate Qs/Qt.12,13 The indices are collectively referred to as "oxygen tension derived indices" and include: PAO2 - PaO2, PaO2/PAO2 and PaO2/FiO2. All relate the arterial oxygen tension to the driving force for oxygen transfer, expressed either as the FiO2 or PAO2. There has been some debate on the use of PAO2 vs FiO2 in a hypoxemia index. Some argue that PAO2 is more advantageous because it includes the effect of CO2.14 Others contend that this addition has little value, adds further assumptions, and may vary somewhat with permissive hypercapnia.15 In ARDS, PAO2 is usually elevated as a result of increased FiO2, and subtraction of a much smaller, relatively constant PaCO2 adds little discrimination.15

Alveolar-Arterial PO2 Difference ("A-a gradient")

The A-a gradient was developed based the relationship between alveolar-arterial oxygen tension: PAO2 equals PaO2 when ventilation and perfusion are perfectly matched. To determine this index, PAO2 is calculated from the following alveolar gas equation:

In healthy conditions there is generally a small difference between PAO2 and PaO2. With perfect matching of ventilation to perfusion, the expected PAO2 - PaO2 gradient would be zero.

Several studies have compared the A-a gradient and Qs/Qt, and the 2 measures correlate moderately well in critically ill patients (r = 0.58-0.68).9,10,16,17 However one study found that 51% of the A-a gradients did not accurately reflect Qs/Qt17 and another actually found that in 25% of the measurements, Qs/Qt and A-a gradient varied in opposite directions.16 The use of the A-a gradient in critically ill patients is limited because it cannot differentiate the severity of different clinical situations.9,10,16,17 For example, a patient receiving FiO2 70% with a PaCO2 40 mm Hg and a PaO2 of 70 mm Hg has an A-a gradient of 379 mm Hg. Another patient receiving an FiO2 90% with PaCO2 40 mm Hg and PaO2 200 mm Hg has an A-a gradient of 392 mm Hg. Another limitation of the A-a gradient is that varying FiO2 concentrations affect the measurement.18 At increasing FiO2, A-a gradient increases markedly, causing this index to lose clinical utility in critically ill patients. Its clinical usefulness is limited to patients with a Qs/Qt < 15% and FiO2 < 0.50.12

Arterial-Alveolar Oxygen Tension Ratio ("a/A ratio")

The a/A ratio was developed based on the relationship between alveolar-arterial oxygen tension. As PaO2 decreases relative to PAO2, the ratio decreases and intrapulmonary shunt increases. A normal value is > 0.75 and a ratio < 0.75 indicates pulmonary dysfunction due to ventilation-perfusion abnormality, shunt, or a diffusion limitation.19

There are conflicting data on the accuracy with which this index reflects Qs/Qt. Cane and colleagues examined the relationship of Qs/Qt and a-A ratio in a heterogeneous group of 75 critically ill patients (50 medical and 25 surgical ICU patients) and reported a high correlation between Qs/Qt and a/A ratio (r = -0.78).10 Rasanen and colleagues examined the relationship between Qs/Qt and the a-A ratio in 17 critically ill patients with respiratory failure, but reported a low correlation (r = - 0.47).20 Like A-a gradient, a-A ratio appears to be vulnerable to changes in peripheral circulation and oxygen therapy.

PaO2/FiO2 ("P/F ratio")

The PaO2/FiO2 ratio was introduced in an attempt to overcome the limitations of A-a gradient and a/A ratio and permit the evaluation of PaO2 at varying FiO2.21 A normal ratio is 300-500 mm Hg, and a value less than 250 mm Hg reflects a clinically significant impairment of pulmonary gas exchange.21

Several studies have compared PaO2/FiO2 and Qs/Qt and the 2 measures have been found to be moderately to highly correlated in hemodynamically stable patients.6, 9,10,20 Covelli and colleagues found that a PaO2/FiO2 < 200 mm Hg correlated with a Qs/Qt > 20% in critically ill patients with ARDS.9 In a retrospective analysis of previously published data of 16 patients with ARDS, PaO2/FiO2 in patients with moderate shunts (< 30%) varied considerably with alteration in FiO2.15 In all patients in this review, when the use of the ratio was restricted to FiO2 values ³ .50 and PaO2 values £ 100 mm Hg, there was little variation in the ratio for a given patient. With low values of true shunt or with substantial perfusion of alveolar units with low ventilation/perfusion ratios, PaO2 increased to > 100 mm Hg and diminished the value of PaO2/FiO2 as an index of hypoxemia.

The American-European consensus conference definition of acute lung injury and ARDS is partially based on the PaO2/FiO2: PaO2/FiO2 200-300 mm Hg defines acute lung injury and PaO2/FiO2 < 200 mm Hg defines ARDS.2 PaO2/FiO2 has been shown to be higher in ARDS patients who survive22 and significantly lower in non-survivors.5 However, several studies22,23 and a meta-analysis24 suggest that it is an inconsistent predictor of outcome in patients with ARDS. Doyle and colleagues found no difference in mortality between patients with a PaO2/FiO2 ratio of 150-300 mm Hg and those whose ratio was < 150 mm Hg.23 The Prostaglandin E1 Study Group found that an improvement in the PaO2/FiO2 ratio after day one of conventional therapy predicted a favorable prognosis in their control patients.22 This improvement in oxygenation among survivors was maintained over a 7-day period. Thus, monitoring trends over time may provide more useful information than any single measurement.

Despite these reports, PaO2/FiO2 remains the most important clinical physiologic variable used in the diagnosis and assessment of ARDS.25-27 In a survey of 448 ICU Medical Directors in the United States, respondents considered the PaO2/FiO2 ratio to be the physiologic variable most important to determine the respiratory status of a patient with ARDS.26

The Ideal Index of Hypoxemia in ARDS

Which of the indices is the ideal index of hypoxemia in ARDS? Unfortunately, none of them are ideal, but the PaO2/FiO2 ratio appears to be the most ideal descriptor available at the present time. The Qs/Qt is considered to be the gold standard, but its measurement is limited to patients with a pulmonary artery catheter. Because it is the gold standard, it has been used as the measure to which other indices are compared. Of the 3 oxygen tension derived indices, PaO2/FiO2 is the most highly correlated with Qs/Qt. However, correlation is not the appropriate measure to use in these comparisons. Correlation is a poor estimate of the predictive value of an association between 2 measurements.28

Using criteria recommended by Gould and colleagues, PaO2 /FiO2 does appear to be the most ideal index of hypoxemia available at the present time: 1) The PaO2/FiO2 ratio appears to be the most reliable, but only under conditions when the FiO2 is ³ 50% and PaO2 is £ 100 mm Hg. However, these conditions are frequently present in ARDS. 2) Theoretically, the ratio appears to measure hypoxemia. It relates the arterial oxygen pressure to the driving force necessary for oxygen transfer. 3) It appears to be responsive to changes in the true value of hypoxemia and it reflects the degree of the changes with some sensitivity because the ratio gets smaller with worsening pulmonary pathology. It does appear to be useful in monitoring trends in hypoxemia over time. But it does not reflect overall severity of lung injury. 4) The ratio appears to provide some clinically useful diagnostic data, but several studies suggest that it is an inconsistent predictor of outcome in patients with ARDS. And finally, while it may be an index that measures pulmonary physiologic function, it does appear to be influenced by therapeutic interventions used to support pulmonary function. As with all the oxygen tension based indices, it is affected by changes in FiO2. At the present time, it is the most commonly used index of hypoxemia in ARDS because of its simplicity and ease of calculation of data that are readily available. The search for the Holy Grail continues.


1. Ashbaugh DG, et al. Lancet. 1967;11:319-323.
2. Bernard GR, et al. Am Rev Respir Crit Care Med. 1994;149:818-824.
3. Moss M, et al. Crit Care Med. 1995;23:1629-1637.
4. Gould MK, et al. Crit Care Med. 1997;25:6-8.
5. Marshall JC, et al. Crit Care Med. 1995;23:1638-1652.
6. Zetterstrom H. Acta Anaesthesiol Scand. 1988;32: 579-584.
7. Martyn JJ, et al. Crit Care Med. 1979:7:492-496.
8. Oliven A, et al. Crit Care Med. 1980:8:99-101.
9. Covelli HD, et al. Crit Care Med. 1983;11:646-649.
10. Cane RD, et al. Crit Care Med. 1988;16:1243-1245.
11. Farhi LE. Ventilation perfusion relationship and its role in alveolar gas exchange. In: Advances in Respiratory Physiology. Caro CG (Ed). London: Arnold; 1966, 148-197.
12. Rasanen J, et al. Crit Care Med. 1987;15:1058-1061.
13. Granger WM. Respir Care. 1996;41:607-610.
14. Gilbert R, et al. Crit Care Med. 1979;7:267-272.
15. Gowda MS, Klocke RA. Crit Care Med. 1997;25:41-45.
16. Keefer JR, et al. Anesth. 1981;55:A82.
17. Wallfisch HK, et al. Crit Care Med. 1981;11:147.
18. Hess D, Maxwell C. Respir Care. 1985;30:961-963.
19. Gilbert R, Keighley JF. Am Rev Respir Dis. 1974;109: 142-145.
20. Rasanen J, et al. Intens Care Med. 1988;14:118-122.
21. Horowitz JH, et al. Arch Surg. 1974;108:349-355
22. Bone RC, et al. Chest. 1989;96:849-851.
23. Doyle RL, et al. Am J Respir Crit Care Med. 1995;152: 1818-1824.
24. Krafft P, et al. Intens Care Med. 1996;22:519-529.
25. Carmicheal LC, et al. J Crit Care. 1996;11:9-18.
26. Zaccardelli DS, Pattishall EN. Crit Care Med. 1996;24: 247-251.
27. Offner PJ, Moore EE. J Trauma. 2003;55:285-289.
28. Bland JM, Altman DG. Lancet. 1986;1(8476):307-310.

Karen Johnson, PhD, RN, School of Nursing, University of Maryland, is Associate Editor of Critical Care Alert.