Most practitioners try to improve tissue oxygenation by increasing cardiac index, arterial oxygen tension, or hemoglobin concentrations. It is unusual for hemoglobin-oxygen affinity to be considered at these times. 

This is because hemoglobin-oxygen affinity has complex effects on tissue oxygenation. In fact, changes in the oxyhemoglobin dissociation curve can have simultaneously opposing actions through oxygen uptake in the lungs versus oxygen unloading in the tissues.

The oxyhemoglobin dissociation curve

The oxyhemoglobin dissociation curve displays the relationship between the oxygen tension of blood and the oxygen saturation (Figure 1). Although the whole curve is the best representation of hemoglobin-oxygen affinity, p50 is often used as the sole descriptor. p50 is the oxygen tension when hemoglobin is 50 % saturated with oxygen. When hemoglobin-oxygen affinity increases, the oxyhemoglobin dissociation curve shifts to the left and decreases p50.

The standard p50 is the oxygen tension at which hemoglobin is 50 % saturated at pH = 7.4, pCO₂ = 40 mmHg (5.3 kPa), temperature = 37 °C with carboxyhemoglobin < 2 %, whereas the in vivo p50 is the oxygen tension at which hemoglobin is 50 % saturated at the pH, pCO₂, temperature, and carboxyhemoglobin concentration of the blood in the subject. The standard p50 thus depends primarily on red-cell 2,3-DPG concentrations and hemoglobin structure. The in vivo p50 reflects the total effect of 2,3-DPG, hemoglobin structure, acid-base balance, temperature, and the dyshemoglobins. From the perspective of oxygen loading and unloading, in vivo p50 is what matters. Unless otherwise specified, the term p50 in this paper means in vivo p50.

For highly accurate p50 determinations it is necessary to construct the full oxyhemoglobin dissociation curve in the laboratory. However, for clinical purposes, p50 values can be calculated much more simply from a single-point measurement of blood gases and hemoglobin-oxygen saturation. The Siggaard-Andersen Oxygen Status Algorithm is the most useful single-point method [1]. This is because it remains accurate up to a hemoglobin-oxygen saturation of 97 % provided the oxyhemoglobin dissociation curve maintains its shape.

The p50 of each animal species has evolved over the millennia through environmental selection pressures, and presumably is optimal for the tissue oxygen consumption, organ capillary density, and environmental oxygen tension of the animal. The value for humans is 26.7 kPa (3.5 kPa). In general, smaller animals have higher p50 settings (Figure 3).

In critical illness, many factors affecting p50 can be operating, at times simultaneously. Acidemia increases p50 via the Bohr effect, but at the same time reduces 2,3-DPG production, decreasing p50. Alkalemia does the opposite. Hypophosphatemia reduces 2,3-DPG production and hyperphosphatemia increases it. Prolonged hypoxemia increases 2,3-DPG concentrations and thus p50. Fever increases p50.

The final result is difficult to predict. Recently, a group of Australian critically ill patients were shown to have a normal mean in vivo p50 [2], despite reduced mean 2,3-DPG concentrations (and thus standard p50). This 2,3-DPG reduction was due almost solely to acidemia. Two other studies of critically ill patients also revealed a reduced standard p50, implying low 2,3-DPG concentrations [3,4]. The one exception comes from Belgium, where patients with acute respiratory distress syndrome and marked hypoxemia were found to have elevated 2,3-DPG concentrations [5].

Investigational drugs and hemoglobin substitutes which can reliably alter p50 are now in existence. As a result, much attention is being given to the best way to manipulate p50 (if at all) when tissue oxygen delivery is under threat. In simple terms, the p50 which best preserves mixed venous oxygen tensions is the appropriate defense of mitochondrial oxygenation.

Mathematical modeling predicts that a reduced p50 will defend mitochondria against severe environmental hypoxia, and animal and human data support this idea [6]. For example, animals adapted to hypoxic environments such as deep burrows or high altitude have a lower p50 than similar species breathing normal ambient oxygen tensions. Similarly, fetal blood (HbF; p50 = 19.4 mmHg (2.6 kPa)) has a low p50 as an adaptation to the hypoxic conditions in utero.

In contrast, when inadequate tissue oxygen delivery occurs in critical illness, it is virtually never due to ambient hypoxia. With the exception of severe hypoventilation, arterial hypoxemia is normally associated with a raised A-a gradient, usually from lung pathology or (rarely) from intra-cardiac shunting. Critically ill patients also suffer reduced tissue oxygen delivery from varying combinations of low output states and anemia. 

Major vascular obstruction can cause severe regional ischemia. In all these scenarios, increasing the p50 should improve venous oxygen tensions for a given oxygen extraction. However, as oxygen delivery continues to fall and the extraction fraction increases, the advantage afforded by increasing p50 will decrease progressively. In extreme hypoperfusion it virtually disappears.

It may soon be possible to achieve significant p50 elevations using artificial hemoglobin solutions or drugs which affect hemoglobin molecular shape. However, despite encouraging theoretical and experimental data, it remains to be established that manipulations of p50 in critical illness can improve gas exchange, tissue oxygenation, or outcome. When we have better evidence that this is true, the status of p50 will warrant routine quantification and consideration.