In an attempt to standardise the degree of impairment of gas exchange in the lungs of ICU patients a number of indices of oxygen exchange have been developed. These include: the arterial-alveolar oxygen tension ratio (a/APO2), the arterial oxygen tension-inspired oxygen concentration ratio (PaO2/FIO2), the respiratory index (RI), [A-a)DO2/PaO2) and the alveolar-arterial oxygen tension difference [A-a)DO2). These are used clinically despite the fact that they do not accurately predict gas exchange particularly with changing the amount of inspired oxygen or when there is mixing of oxygenated blood with unoxygenated blood in the patient (called shunt). None of the indices reliably reflect the behaviour of the physiological shunt. In a pilot study as inspired oxygen (FIO2) was increased incrementally from 0.30 to 1.00, up to 55 per cent of the measured changes in these indices were opposite in direction to the corresponding changes in the physiological shunt. The investigators have developed a computer algorithm and a slide rule to try and improve on these clinical measures. The investigators wish to test their slide rule on ventilated patients by altering the inspired oxygen to see if it will predict what the actual change in oxygen will be.](streamdown:incomplete-link)

30 adult patients admitted to Aberdeen Royal Infirmary ICU, requiring invasive mechanical ventilation and with impaired gas exchange will be recruited and the following baseline demographic data will be recorded: cause of respiratory failure, APACHE II score, cardiovascular shock (presence / absence), duration of mechanical ventilation prior to inclusion, duration of respiratory failure prior to inclusion. Also the following baseline physiological data will also be recorded: FIO2, SaO2, PEEP, arterial blood pressure, central venous pressure, haemoglobin, PaO2, PaCO2, arterial pH and inotrope / vasopressor infusion rates.

The FIO2 will then be varied in random steps to 0.35, 0.4, 0.45, 0.5, 0.55, 0.6 and 0.65. After changing the FIO2 a 3-4 minute equilibration period will be allowed prior to taking an arterial blood gas sample for analysis. In the event of a PaO2 < 8kPa being recorded, further reductions in FIO2 will not be undertaken.

The data collected will then be utilised using a slide rule computer programme to generate FIO2 vs. SpO2 graphs. Other indices of gas exchange, e.g. PaO2/FIO2 ratios will also be generated from the data. Predicted versus actual PaO2 will be plotted and a correlation coefficient obtained.

Patients will then be followed for up to 5 days and depending on progression the same procedures will be repeated between days 3 and 5.

Inclusion Criteria:

• FIO2 on enrollment = 0.4-0.6, PEEP 5cm H2O, "Stable" respiratory failure - same or less FIO2 for 2 hours prior to enrollment.

Exclusion Criteria:

• Age < 16 years
• Pregnancy
• Known intracranial disease
• Prone position
• Nitric oxide therapy
• Pneumothorax/bronchopleural fistula/intercostal drain in situ
• Severe respiratory failure (FIO2 0.7), OR
• Severe haemodynamic instability (escalating inotrope requirements, fluid resuscitation > 1 litre/hour in 2 hours prior to commencing intervention, mean arterial blood pressure < 65 mmHg).

30 adult patients admitted to Aberdeen Royal Infirmary ICU, requiring invasive mechanical ventilation and with impaired gas exchange will be recruited.