To the Editors:
I read with interest the article by Rees et al. 1. Their data on 40 patients with chronic lung disease supports the use of peripheral venous blood for an estimate of the arterial oxygen tension (Pa,O2) in a vast majority of patients. They also address this issue in a similar study 2. They found that mean Pa,O2 was not clinically larger than the peripheral venous oxygen tension (0.83±4 mmHg or 0.11±0.53 kPa). They also showed that peripheral venous carbon dioxide tension and pH correlate very well with arterial values, with the arterial values being slightly lower and clinically negligible compared with venous values for carbon dioxide tension (-0.7±2.1 mmHg or -0.09±0.28 kPa) and pH (-0.001±0.013) 1.
I would like to offer two comments on their findings. First, the authors describe accuracy (bias) as the mean difference between the two samples. However, true mean bias is best taken as the mean absolute difference between the two samples. Suppose, for example, that one subject has a Pa,O2 that is 5 mmHg more than the venous sample, and another subject has a Pa,O2 that is 5 mmHg less. In such a case, if one uses mean difference instead of absolute mean difference, one would erroneously conclude that the accuracy of peripheral venous blood to predict Pa,O2 is perfect (mean bias would be 0 mmHg), when in fact it is not. The authors calculate that the bias from their samples is 0.83 mmHg (0.11 kPa). However, as the above example indicates, the absolute bias may well be much larger.
Secondly, the findings are of no surprise. A recent meta-analysis of 29 studies indicates that the accuracy of predicting Pa,O2 from a venous sample depends on the arterial sample 3. While the meta-analysis compared earlobe and fingertip capillary samples to arterial samples, the findings of the meta-analysis still apply, since venous sampling is evaluated in both cases. The meta-analysis clearly shows that as the Pa,O2 drops, as in the case of altitude or other pathophysiology (i.e. lung disease), the accuracy in predicting Pa,O2 improves 3. The meta-analysis thus clarifies why the Mount Everest study 4, and any study using diseased patients with low Pa,O2 (e.g. Dar et al. 5), supports the use of capillary blood to represent Pa,O2, and any other study that uses subjects with normal Pa,O2 values does not. The data of Rees et al. 1 support the notion that peripheral venous blood reflects Pa,O2 because patients had low a mean Pa,O2 (67 mmHg or 9 kPa) to begin with, and a low Pa,O2 minimises the arterio-venous difference for oxygen tension, improving its accuracy in predicting Pa,O2. In table 7 of the meta-analysis the regression equation indicates that when arterial Pa,O2 is 67 mmHg, the mean difference between arterial and earlobe oxygen tension is small (0.79 mmHg; r2 = 0.86, mean square error 4.8 mmHg) 3. This difference is nearly identical to the 0.83 mmHg mean difference from Rees et al. 1 when using the same mean Pa,O2.
Consistent with the recent work of Rees and co-workers 1, 2, the meta-analysis demonstrated that venous sampling from an earlobe or fingertip could accurately reflect arterial pH or arterial carbon dioxide tension (Pa,CO2) over a wide range of values 3. Therefore, sampling either peripheral venous blood, arterialised capillary blood from the earlobe, or fingertip capillary blood for predicting Pa,CO2 or arterial pH is very robust over a wide range of clinical values. I would recommend substituting arterial sampling for venous sampling if one only wants to predict arterial pH or Pa,CO2.
For predicting Pa,O2, venous blood sampling from a peripheral vein or from an earlobe sample (which is better than a fingertip sample) is only appropriate for conditions where the Pa,O2 is low. The conundrum is, of course, that one cannot know if Pa,O2 is low (i.e. <70 mmHg) unless it is actually measured from an arterial sample. As the residual standard error in predicting Pa,O2 from venous samples is large, from 6 mmHg when predicting from earlobe sampling (which is similar to their peripheral venous sampling data) to 15 mmHg when predicting from fingertip sampling 3, I would recommend against using venous blood to predict Pa,O2 or the alveolar to arterial oxygen tension difference in patients or in research subjects.
Statement of interest
None declared.
- © ERS Journals Ltd
