Concerning quality demands of arterial partial pressure of oxygen

To the Editor,

Searching the literature for quality demands or biological variation for arterial partial pressure of oxygen (pO₂), we found no criteria that could be used for defining the necessary limits for neither bias nor Coefficient of Variation (CV%) of pO₂.

Usually, according to the Stockholm criteria of 1999 and further revised in Milan 2014 [1, 2], the criteria should preferably be based on medical needs according to application, or if that cannot be done, according to intra- and interindividual variation.

When we constructed the Tempus600 system [3] and later documented the effect on biochemical parameters by transporting sample tubes through the system, we experienced an unacceptable effect on true values on especially O₂-related parameters [4].

As no data is yet available through the Westgard: (Ricos et al.) or EFLMs databases on Coefficient of Variations and biological variation of pO₂, we discussed with our clinicians how they use the pO₂ in clinical application. Based on these clinical judgements and through the known information on the normal ranges, we defined the precondition for our considerations for pO₂ quality specifications in arterial blood as follows:

1. The normal reference range is 11–14 kPa

2. An acceptable value in an emergency patient is down to 10 kPa

3. The definition of respiratory failure is a value below 8 kPa

4. A life-threatening value is 5 kPa

According to the clinician’s decision limits for acting on changes in pO₂, the following could be deduced:

1. It is clinically relevant to detect a decrease from 10 to 9 and from 9 to 8 kPa

2. It is clinically necessary to detect a decrease of 0.5 kPa at the level of 5 kPa

3. An acceptable total error at a level of 5 kPa is required

We assumed that in such situations the biological variation is unimportant, as the actual value is what determines the patient’s situation and the clinician’s corrective medical interventions. Therefore, at these critical values it is important to establish as accurately as possible the value, but also to detect an improvement or deterioration over time.

Consequently, the considerations on acceptable variation comprise both sampling, other pre-analytical and transport sources of variation, and the analytical variation itself. The measurement should thus be capable of detecting a change in pO₂ from 10 to 9 kPa with at least a 95 % certainty.

In evaluating this decrease, the bias is of no importance if measured on the same instrument and thus, the critical difference mandatory to be able to detect with 95 % certainty is 1 kPa. Using analysis of variance, the difference should be within 2.77 times the total analytical and pre-analytical variation. This means that pre- and analytical variation of 0.36 kPa (1 SD) will fulfill this. At the pO₂ level of 9 kPa, this is 0.36 × 100/9=a CV% corresponding to 4 %.

The maximal total analytical and pre-analytical variation are thus defined to be less than 4 %.

This defined minimum quality criteria for the pO₂ measurement will consequently also enable a measurement of a decrease from 8 to 5 kPa with a certainty of above 99 %, fulfilling the ability to detect this life-threatening value.

For the absolute measurement of the value 5 kPa with a total error of <0.5 kPa giving the clinician’s limit for acting (Bullet 6: It is clinical necessary to detect a decrease of 0.5 kPa at the level of 5 kPa), the bias plus 1.65 times the total analytical variation should not exceed 0.5 kPa.

(Bullet 7: An acceptable total error at a level of 5 kPa is required).

This can be achieved with a maximal bias of 0.2 kPa and a total coefficient of variation of 4 % (0.2 + 1.65 × (4 % of 5.0=0.2) ∼ 0.5). For the absolute measurement of a value, this means that the total coefficient of variation of sampling, transport, other pre-analytical sources, and the analytical uncertainty should be below 0.36 kPa at level 9 kPa and below 4 % through the range from 1 to 14 kPa.

This means that at 5 kPa, a significant change of 2.77 × 4 % of 5 kPa=2.77 × 0.2=0.55 kPa is detectable, largely fulfilling the clinical needs for the application.

If these criteria of maximum bias of 0.2 kPa and a total coefficient of variation below 4 % are fulfilled, the quality specifications are satisfactory for all applications for arterial blood measurements of oxygen tension.

For the measurement of venous values, at least the same specifications are required.

In conclusion, we suggest that the quality demands for arterial partial pressure of oxygen are:

Bias<0.2 kPa

CV<4 %

TE<0.5 kPa at a level below 5 kPa and TE<0.6 kPa at a level above 5 kPa

TE<0.2 + (1.65 × 4 % of the actual level).

We have implemented these quality demands in our laboratories until more detailed scientifically based data are available.

The calculations and resulting specifications have been distributed to members of the Nordic lab Network CCL (Laboratory medicine in Nordic countries) for criticism, and remarks from Anders Kallner were subsequently incorporated.