Variability of Arterial PaO2 in Critically Ill Patients

Abstract & Commentary

In this study from taiwan, the spontaneous variability of arterial oxygen tension (PaO2) was determined in 23 critically ill but "stable" medical ICU patients. The patients all had pulmonary artery catheters and were being ventilated in pressure control mode for a variety of diagnoses, including acute respiratory distress syndrome (ARDS) in nine, severe sepsis in eight, and acute pulmonary edema in three. At a time when the patients were hemodynamically stable and judged not to be changing clinically, they were switched for one hour each among three combinations of inspiration-to-expiration (I:E) ratio and positive end-expiratory pressure (PEEP). Under each condition, after airway suctioning, with the patients deeply sedated and with no changes in ventilator settings or other therapy, arterial blood was drawn every 15 minutes for blood gas analysis.

There was substantial variability in PaO2 from measurement to measurement. Although mean PaO2 varied under the different conditions of I:E ratio and PEEP (from a low of 97 mmHg with I:E 1:2 and PEEP 5 cm H2O to a high of 127 mmHg on the same I:E ratio and 15 cm H2O PEEP), there was also substantial variation from sample to sample under each set of ventilation conditions. The coefficient of variation among the samples drawn every 15 minutes under the different conditions ranged from 5.9% to 7.2%, a difference that was not statistically significant. The intrapatient range of individual PaO2 values varied from 4-65 mmHg; this range of PaO2 values within a single patient was more than 20 mmHg during at least one of the one-hour observation periods in 15 of the 23 patients, and more than 30 mmHg at some time in seven of them.

Mixed venous oxygenation was also studied, and also demonstrated substantial measurement-to-measurement variation. The coefficients of variation for the alveolar-arterial oxygen tension difference [P(A-a)O2)] and ratio of arterial to alveolar oxygen tension (PaO2/PAO2) were about the same as for PaO2 alone, while that for P(A-a)O2/PaO2 was significantly greater (mean 12.5%). Tsai and colleagues conclude that, in critically ill medical ICU patients, despite sedation, the spontaneous variability in PaO2 over time is substantial, but that this variability is similar under different conditions of I:E ratio and PEEP.


This study confirms the earlier findings of Thorson et al (Chest 1983;84:14-18), Hess and Agarwal (J Clin Monit Comput 1992;8:111-115), and Sasse et al (Chest 1994;106:187-193), that "stable" patients who are ill enough to require invasive monitoring and other interventions in an ICU demonstrate more spontaneous variability in arterial blood gases—particularly in PaO2—than is commonly appreciated.

The fact that PaO2 varies on spot samples every 15 minutes by an average of about 7%—and by more than 20 mmHg in most patients at least some of the time—creates a real problem for management. In a patient with ARDS, a drop in PaO2 from 80 to 60 mmHg in two successive arterial specimens would likely trigger a ventilator change—an increase in inspired oxygen fraction, PEEP, or I:E ratio, for example—and yet such a change may well just reflect the baseline instability of oxygenation in such a patient. This problem is compounded several-fold by continuous monitoring with pulse oximetry, with which even momentary dips in arterial saturation are detected and often lead to management changes.

Should we ignore a 20 mmHg drop in PaO2, or the pulse oximeter’s alarm that saturation has dipped below 90%? Of course not. However, it is important to know that such changes may just reflect the instability of gas exchange at a time of critical illness rather than a change in the patient’s condition that mandates a change in ventilator settings or other intervention.

It is difficult to construct a management algorithm that eliminates the need for experience and clinical judgment. In this era of increasing comfort with "permissive hypercapnia," clinicians in the ICU need to learn when to allow "permissive hypoxia" as well. A pulse oximeter alarm indicating a saturation below 90%, or a "routine" arterial blood gas result with a PaO2 lower than on the last specimen, should trigger a close look at the patient rather than an automatic change in management. If the vital signs have not changed and the patient is clinically the same as before, there is probably time to watch and wait, at least for the next 15 minutes, and to repeat the measurements if there are no signs of deterioration at the bedside.