Methodology

The evidence review methodology has changed from NICE methodology to the BTS NICE accredited guideline production process which is based on the Scottish Intercollegiate Guideline Network (SIGN) methodology and adheres to AGREE methodology (see section 1).

Evidence levels and grade of recommendation

These are now in SIGN format (see section 1 and tables 6 and 7).

Evidence base

The evidence base for the guideline has been updated to August 2013 (and extended to late-2016 for key references). None of the 2008 recommendations have been challenged by new evidence but many of the existing recommendations are supported by new information. There have been many observational studies but few randomised trials directly relevant to the guideline since 2008.

The remit of the guideline has been extended

The new guideline covers not just emergency oxygen use but most oxygen use in healthcare settings. It also covers short-term oxygen use by healthcare workers outside of healthcare settings but domiciliary oxygen use by patients is covered by the BTS guideline for home oxygen use in adults.13

The scope of the guideline has been widened

The present guideline includes the following new topics and settings which have been requested by guideline users:

• Postoperative and perioperative care including PCA,

• Endoscopy and other procedures requiring conscious sedation,

• Palliative care settings including hospices,

• Use of helium–oxygen mixtures (Heliox) and nitrous oxide/oxygen mixtures (Entonox),

• Use of CPAP,

• Use of oxygen by healthcare professionals in patients' homes,

• Use of oxygen by voluntary rescue organisations and other non-NHS first responders,

• High-flow nasal cannulae (HFNC).

The structure and format of the guideline has been changed since 2008:

The 2008 guideline was published as a supplement in Thorax.2 Additional educational materials and other resources including audit tools were made available on the BTS website. The new guideline exists in two complementary formats.

• A concise guideline which contains recommendations and good practice points is published in BMJ Open Respiratory Research.

• The full guideline including evidence review, physiology overview, illustrations and references is published in this edition of Thorax and is available on the BTS website http://www.brit-thoracic.org.uk.

Key question 1: physiology and pathophysiology of oxygen

• What are the dangers of hypoxia/hypoxaemia (ie, what happens to the human body)?

• What level of hypoxaemia is dangerous to all patients (even healthy adults)?

• What level of hypoxaemia is dangerous to vulnerable groups (eg, ischaemic heart disease, stroke, elderly)?

  • Repeat the above searches with additional key words: elderly, stroke, myocardial infarction, heart failure, COPD, trauma, renal failure.

• Same questions for hypercarbia/hypercapnia:

  • Search for ‘hypercapnia’ combined with terms implying a harmful outcome (death/tissue injury/brain damage/coma).

• What level of hypercapnia is dangerous to all patients?

• What level of hypercapnia is dangerous to vulnerable groups (as above)?

• Same questions for respiratory acidosis:

  • Search for ‘respiratory acidosis’ combined with terms implying a harmful outcome (death/tissue injury/brain damage/coma).

• What level of respiratory acidosis is dangerous to all patients?

• What level of respiratory acidosis is dangerous to vulnerable groups (as above)?

Key question 2: clinical aspects of hypoxaemia and oxygen therapy for common medical emergencies

• How to assess hypoxaemia (clinical, EWS systems, oximetry, arterial and capillary blood gases).

• How to assess hypercarbia/hypercapnia.

• Use of oxygen to relieve symptomatic breathlessness.

• Use of oxygen in acute COPD.

• Use of oxygen in acute asthma.

• Use of oxygen in pneumonia.

• Use of oxygen for pulmonary embolus.

• Use of oxygen in trauma.

• Use of oxygen in heart failure.

• Use of oxygen in myocardial infarction and unstable coronary artery syndrome.

• Use of oxygen in angina.

• Use of oxygen for other patients with less common conditions were searched individually (eg, CF, muscular dystrophy, motor neurone disease, severe kyphoscoliosis, anaphylaxis, hyperventilation).

Key question 3: oxygen prescription, oxygen delivery systems and oxygen transport

• Oxygen carriage in transport (practical issues; safety issues).

• Oxygen delivery systems in ambulances.

• Prescription of oxygen.

• Local hospital guidelines for oxygen use.

• Oxygen delivery systems in hospitals and other healthcare and emergency settings.

• Advantages/disadvantages of each delivery system (Venturi masks, simple face masks, nasal cannulae, high-flow masks such as non-rebreathing reservoir masks). Use of oxygen-driven nebulisers.

• Use of CPAP.

• Use of ‘alert cards’, alert bracelets or similar hazard warning systems for patients who are known to be at risk of hypercapnia.

3.1.1 Normal ranges for oxygen saturation (SaO₂ and SpO₂) and PO₂ (PaO₂) in the blood at sea level

As there is a fixed amount of haemoglobin circulating in the blood, the amount of oxygen carried in the blood is often expressed in terms of how saturated with oxygen the circulating haemoglobin is. This is what is meant by ‘oxygen saturation level’. If this is measured directly from an arterial blood sample, it is termed SaO₂. If it is measured from a pulse oximeter it is termed SpO₂. Alternatively, one can measure the PO₂ of the blood (PaO₂), known as the ‘partial pressure of oxygen’ in the blood. This measurement can be expressed in kilopascals (kPa; normal range 12.0–14.6 kPa) or in millimetres of mercury (normal range 90–110 mm Hg for young adults).17 The precise normal ranges for oxygen saturation and PO₂ are difficult to determine accurately due to a lack of data from the ‘normal’ population, that is, non-hospitalised healthy participants.

A US observational study of nearly 900 asymptomatic patients in the emergency department showed the median SpO₂ value to be 99% with SpO₂ values <97% occurring in 5.7% of the study group.18 However, the study group were young (median age 38) and predominantly African-American, so these data are not generalisable to the UK population.

Another North American study measured blood gases and SpO₂ at sea level in 96 healthy individuals aged between 18 and 81.17 They found that for adults 2SD range for SaO₂ is ∼94–98% at sea level but this may decline gradually with advancing age (table 8).

A much larger observational study of over 37 000 patients admitted to four acute medical admissions units across the UK19 found that median SpO₂ was 98% for young adults aged 18–64 years (IQR 97–99%). For adults aged ≥65 years the median saturation was 96% with IQR 95–98% (table 9). While the authors of this study recommend setting a normal range of 96–98%, their study excluded nearly 20% of patients who were receiving oxygen, many whom were likely to have a SpO₂ in the lower end of the normal range. In further considering these study results, it is of note that the PaO₂ is 0.8 kPa (6 mm Hg) lower in the supine position than in the upright position and most emergency measurements are made in the supine position.20

3.1.2 Oxygen saturation in older adults

The mean SaO₂ may be lower in older people than in young adults. However, it is difficult to dissociate the effects of advancing age from the effects of the diseases that become commoner in old age. Some papers have reported a fall in the blood PaO₂ in older participants. Indeed a fall in SpO₂ in patients >65 was demonstrated in Crapo and Smith's studies, and the former study shows a decreasing PaO₂ with age. However, others have failed to confirm this observation.21–23 The mean SaO₂ in seated adults aged >64 years in one published study was 95.5% compared with 96.9% in adults aged 18–24 years, and the SD was wider in the older age group with a 2SD range of 92.7–98.3% (tables 8 and 9).17 The mean (SD) SaO₂ for recumbent healthy men aged >70 years in another study was 95.3% (1.4%) giving a 2SD range of 92.5–98.1% for men of this age.21 The mean (SD) SaO₂ was 94.8% (1.7%) for recumbent healthy women aged >70 years with a 2SD range of 91.5–98.2%. The authors of this study did not observe any age-related decline in SaO₂ beyond the age of 70 years. The mean SaO₂ in this study of ∼95.0% for recumbent healthy men and women aged >70 years was below the normal range for seated healthy young adults. The mean PaO₂ in older participants in this study was 10.3 kPa for men and 9.8 kPa for women, which is lower than two other studies which reported mean PaO₂ values of 11.2 kPa and 11.1 kPa in healthy older participants.22 ,24 Some of these differences are probably due to participant selection, but there may also be variations in the results obtained by different blood gas analysers.25

Further data from an unpublished audit of 320 stable hospital patients in Salford and Southend with no history of lung disease found a mean (SD) SpO₂ of 96.7% (1.77%; 2SD range 95.2–100%) in patients aged >71 years (R O'Driscoll, A Davison, L Ward, personal communication). These values were measured by pulse oximetry in UK hospitals in 2008 and are more likely to represent the expected normal range of pulse oximetry measurements in the elderly UK population. The variation with age, sex and posture makes it difficult to give a precise target range that will apply to all adults who might require oxygen therapy, but the guideline development committee believe that a target range of 94–98% will achieve normal or near-normal SpO₂/SaO₂ for most adults in the UK and will avoid any risk of clinically significant hypoxaemia.

3.1.3 Oxygen saturation at altitude

While the percentage of oxygen in inspired air is constant at 21% and does not change with altitude, the fall in atmospheric pressure at higher altitudes decreases the partial pressure of inspired oxygen. The SaO₂ at a given altitude varies with age, sex, ethnic group and degree of acclimatisation to altitude. For example, a sample of 3812 people of all ages living in Tibet at an altitude of about 4000 m had a mean SaO₂ of only 88.2%, but people native to the Andes had an SaO₂ about 2.6% higher than Tibetans living at the same altitude.26 ,27 However, the physiology of hypoxaemia at altitude is very different to the physiology of hypoxaemia at sea level, for example, ventilator changes usually lead to a marked fall in arterial CO₂ level. Millions of people live at these altitudes with SaO₂ values that would cause concern at sea level. The city of La Paz in Bolivia has a mean altitude of 3600 m and a population of ∼1.5 million people. The SaO₂ of climbers on Mount Everest (8848 m) can fall below 55%.28 Sudden exposure to altitudes above about 4000 m can cause mountain sickness, high altitude pulmonary oedema and high altitude cerebral oedema in unacclimatised individuals. Long-term exposure to high altitude (or to hypoxaemia for any other reason) can lead to pulmonary hypertension.

3.1.4 Oxygen saturation in acute and chronic diseases

If the blood oxygen level falls to extremely low levels for even a few minutes (eg, during cardiac arrest), tissue hypoxia and cell death will occur, especially in the brain. The brain appears to be the most vulnerable organ during profound hypoxaemia; brain malfunction is the first symptom of hypoxia and brain injury is the most common long-term complication in survivors of cardiac arrests and other episodes of profound hypoxaemia. Sudden exposure to low SaO₂ below about 80% can cause altered consciousness even in healthy participants. It is likely that other organs in patients with critical illness or chronic organ damage are vulnerable to the risk of hypoxic tissue injury at oxygen levels above 80%.

Most experts emphasise the importance of keeping the SaO₂ above 90% for most acutely ill patients.29–32 However, the degree of hypoxia that will cause cellular damage is not well established and probably is not an absolute value. Healthy older adults, for instance, have lower SaO₂ values at rest than younger adults. Many patients with chronic lung disease, congenital cyanotic heart disease or chronic neuromuscular conditions have oxygen saturations substantially below the normal range, even when clinically stable.

However, although chronically hypoxaemic patients may tolerate an abnormally low SaO₂ at rest when in a clinically stable condition, these resting oxygen levels may not be adequate for tissue oxygenation during acute illness when the tissue oxygen demand may increase (eg, sepsis, trauma, pneumonia, head injury; see section 8).

Acute hypoxaemia with SaO₂<90% and sometimes <80% is seen in many acute illnesses such as pneumonia and heart failure and it is likely that the clinical manifestations of hypoxaemia in illness would be similar to those of experimental hypoxaemia in hypobaric chambers (impaired mental function followed by loss of consciousness). However, the clinical manifestations of the illness itself make it difficult to identify which symptoms and signs are due to hypoxaemia. Pure hypoxaemia, as seen in hypobaric chambers and at altitude, does not seem to cause breathlessness in resting participants. Patients with chronic diseases such as COPD, lung fibrosis, neuromuscular disorders or congenital heart disease may routinely attend outpatient clinics with SaO₂ levels well below 90% even at a time when their disease is stable. In an emergency, a clinician who was not familiar with such a patient (when stable) might interpret the low saturation as having occurred acutely and aim to achieve an oxygen saturation that was well above the patient's usual oxygen saturation level. Many such patients would qualify for long-term oxygen therapy. The UK COPD guideline recommends a threshold of 7.3 kPa (55 mm Hg) below which most patients with COPD will benefit from long-term oxygen therapy (equivalent to a SaO₂ of about 88–89%) and an PaO₂ threshold below 8.0 kPa (60 mm Hg) for patients with established cor pulmonale and some other subgroups.1

3.1.5 Variation in oxygen saturation during sleep

Healthy participants in all age groups have greater variation in SaO₂ when sleeping than while awake. A study of 330 people referred to a sleep laboratory with normal results of overnight polysomnography (patients with cranial facial or neurological abnormalities or previously diagnosed pulmonary disease were excluded) showed that desaturation routinely occurred with a mean (SD) minimum SaO₂ or ‘nadir’ of 90.4% (3.1%) during the night (2SD range 84.2–96.6%).33 The mean (SD) overnight SaO₂ ‘nadir’ was 89.3% (2.8%) for participants aged >60 years.33 In this study, participants aged 20–30 years spent 10% of the night with SaO₂ levels below 94.8% and half the night below 96.3%, and those aged 60 years spent 10% of the night below 92.8% and half the night below 95.1%. Furthermore, the authors of this study excluded obese patients with any features of sleep apnoea or hypopnoea because these patients are known to desaturate to very low levels during sleep (often below 70%). The variation in SaO₂ during sleep is exaggerated by alcohol and by sedative drugs. This makes it difficult to evaluate a ‘spot reading’ of SaO₂ on a sleeping participant. It is suggested that SaO₂ measurements of sleeping participants should be interpreted with caution and ideally observed for a few minutes to see if the participant has got sustained hypoxaemia or just a transient normal ‘nocturnal dip’.

Oxygen saturation in acute and chronic disease is discussed in section 3.1.4.

Evidence statements (see recommendations A1–A2):

• Normal daytime haemoglobin SaO₂ is 96–98% in young adults in the seated position at sea level but the lower limit falls slightly with age and is ∼94–98% in adults aged >70 years (evidence 2+).

• All participants have transient dips in oxygen saturation at night with a mean nadir of 90.4% (2SD range 84.2–96.6%) in healthy participants in all age groups (evidence level 3).

3.1.6 Normal range for arterial carbon dioxide tension

The reference range for arterial carbon dioxide tension (PaCO₂) is 4.6–6.1 kPa (34–46 mm Hg) for healthy adult men aged 18–38 years.34 Although this study was undertaken in 1948, it is consistent with the clinical experience of the guideline group members and with most modern reference values for PaCO₂. Although different laboratories and textbooks give slightly different reference values, all are within 0.2 kPa of the above reference range. Any value of PaCO₂ of >6.1 kPa (45 mm Hg) should be considered abnormal, but values up to 6.7 kPa (50 mm Hg) may be obtained by breath holding.35

Hypoxaemia

Hypoxaemia refers to low PO₂ or partial pressure of oxygen (PaO₂) in the blood. For practical reasons, hypoxaemia can also be measured in relation to oxyhaemoglobin saturation. In adults the normal range is influenced by age and comorbidity and the normal ranges for healthy adults are given in section 3.1.1. The precise level at which a patient becomes hypoxaemic is debatable. One could argue that any saturation below the lower limit of normal constitutes hypoxaemia. Various authors have defined hypoxaemia as SaO₂ of (1) <94%; (2) <92%; (3) <90%; or (4) PaO₂<60 mm Hg or 8 kPa.4 ,36–38 Most authors who have studied this area have defined hypoxaemia as PaO₂<60 mm Hg (8 kPa) or SaO₂<90%.39 There is no known risk of hypoxic tissue injury above this level and other guidelines on critical care set 90% as the minimum below which SaO₂ should not be allowed to fall.31 ,32

Hypoxia: Hypoxia occurs when oxygen supplies are insufficient to meet oxygen demands in a particular compartment (eg, alveolar or tissue hypoxia). Tissue hypoxia may be subdivided into four main causes: hypoxaemic, anaemic, stagnant or histotoxic. Oxygen therapy may only correct hypoxia due to hypoxaemia, other ways to improve oxygen delivery to the tissues need to be considered.

Hypoxaemic hypoxia: Hypoxaemic hypoxia (sometimes also referred to as hypoxic hypoxia) is present when the oxygen content in the blood is low due to reduced partial pressure of oxygen. This occurs naturally at altitude or occurs secondary to right-to-left shunts, ventilation–perfusion (V/Q) mismatch, alveolar hypoventilation or diffusion impairment.

Type 1 respiratory failure, which is defined as PaO₂<8 kPa or 60 mm Hg (equivalent to SaO₂ of ∼90%) with a normal or low PaCO₂ level is due to hypoxaemic hypoxia.

Anaemic hypoxia: Anaemic hypoxia results from a reduced level of haemoglobin available for oxygen transport. Although the patient may not be hypoxaemic (with a normal PaO₂ and SpO₂), the reduced oxygen content of the blood may lead to tissue hypoxia. Carbon monoxide poisoning may also produce a form of anaemic hypoxia by impairing the ability of haemoglobin to bind oxygen, thereby reducing oxygen-carrying capacity.

Stagnant hypoxia: Stagnant hypoxia is a low level of oxygen in the tissues due to inadequate blood flow (either globally or regionally). This condition may occur in the extremities if a person is exposed to cold temperatures for prolonged periods of time and it is the cause of gangrene in tissue that is deprived of blood in severe peripheral vascular disease. Stagnant hypoxia may occur in low cardiac output states.

Histotoxic hypoxia: Histotoxic hypoxia is an inability of the tissues to use oxygen due to interruption of normal cellular metabolism. The best known example of this occurs during cyanide poisoning which impairs cytochrome function. It is increasingly thought that mitochondrial dysfunction may lead to decreased oxygen usage in sepsis despite adequate oxygen delivery. This has also been termed ‘cytopathic dysoxia’.40

Hyperoxia and hyperoxaemia: Hyperoxia and hyperoxaemia are the counterparts to the above terms and in this guideline refer to high oxygen content in a compartment and high PO₂ in the blood, respectively. As stated above, for practical purposes the PO₂ in the blood is often measured as oxyhaemoglobin saturation. Furthermore, this guideline is centred on providing target saturations for various conditions, but it should be noted that above a PaO₂ of ∼16 kPa (120 mm Hg) the oxyhaemoglobin saturation will obviously not change from 100%, yet the effects of further increases in PaO₂ may be important in certain conditions such as COPD. This is discussed in further detail in sections 5 and 6.

Respiratory acidosis

Carbon dioxide (CO₂) can combine with water (H₂O) to form carbonic acid (H₂CO₃) in the blood which, in turn, dissociates to bicarbonate (HCO₃⁻) and a hydrogen ion (H⁺). Acute respiratory acidosis occurs if the pH of the blood falls below 7.35 ([H⁺]>45 nmol/L) in the presence of a raised CO₂ level. If respiratory acidosis has been present for more than a few hours, the kidney retains bicarbonate to buffer the acidity of the blood and, over hours to days, this may be sufficient to produce a normal pH. This situation (high PaCO₂ with high bicarbonate and normal pH) is known as ‘compensated respiratory acidosis’. This situation is common in patients with chronic severe but stable COPD, but they may have an additional acute rise in PaCO₂ during an acute exacerbation giving rise to ‘acute on chronic’ respiratory acidosis despite their high bicarbonate level. This happens because the bicarbonate level was equilibrated with the previous CO₂ level and is insufficient to buffer the sudden further increase in CO₂ level that may occur during an exacerbation of COPD. Respiratory acidosis is common in clinical practice. Plant et al43 showed that about 20% of patients with AECOPD requiring hospital admission have respiratory acidosis.

Metabolic acidosis: This can be caused by failure to excrete acid produced by the body's normal metabolic processes (eg, during renal failure) or by increased production of acid from abnormal metabolic conditions such as diabetic ketoacidosis. Alternatively, it may result from direct loss of bicarbonate from the kidney or gut (eg, during chronic diarrhoea). In all forms of metabolic acidosis, there is a low blood bicarbonate level, either due to loss of bicarbonate or due to buffering of excess acid by bicarbonate which is excreted as CO₂. A common cause of metabolic acidosis is lactic acidosis caused by tissue hypoxia. This may result from decreased oxygen delivery such as occurs in hypoxaemia, or low cardiac output states or conditions such as sepsis where oxygen consumption is impaired in the face of adequate oxygen delivery. In health, metabolic acidosis will occur at peak exercise where oxygen delivery is insufficient to meet demand.

5.1.1 Arterial oxygen tension

Ventilatory drive

If PaO₂ falls, the peripheral chemoreceptors in the carotid body drive an increase in ventilation to increase PaO₂.54 This will not significantly increase PO₂ in blood leaving already well-ventilated units, but will increase PO₂ leaving less well-ventilated alveolar units by increasing PAO₂ in these units. Although the ventilatory response to SaO₂, and therefore CaO₂, is linear (figure 4), the carotid body senses PaO₂ and not CaO₂. This prevents excessive ventilation in response to anaemia which would be ineffective in increasing CaO₂. The peripheral chemoreceptors are able to do this because the very high ratio of DO₂ to oxygen consumption of the carotid body means that the tissue PO₂ in the carotid body continues to reflect PaO₂ and will not fall even in the presence of anaemic hypoxia.55 ,56

Example

• Before oxygen therapy assume 50% of pulmonary flow is passing through an area of low V/Q and that the pulmonary venous oxyhaemoglobin saturation (SpvO₂) from this compartment is 80% (ie, just above mixed venous SO₂). The other 50% is passing through an area of matched V/Q, resulting in a SpvO₂ of 97%. The final mixed SpvO₂ will be 88.5%.

• Following maximal oxygen therapy, assuming no change in flow as a result of release of HPV, SpvO₂ from the low V/Q compartment rises to 85% and SpvO₂ from the matched compartment rises to a maximum of 100%. The resulting admixed SpvO₂ will now only be 92.5%. This has occurred because fully saturated blood cannot increase its oxygen content beyond full saturation despite an increase in PO₂, apart from the minimal contribution from dissolved oxygen; that is, the relationship between PO₂ and oxyhaemoglobin saturation/blood oxygen content is not linear.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Oxygen dissociation curve with Bohr effect. 2,3-DPG, 2,3-diphosphoglycerate; PO₂, oxygen tension.

5.1.2 Haematocrit

Erythropoiesis is controlled by a negative feedback system involving erythropoietin. By contrast with the carotid bodies, the peritubular cells in the kidney are well suited to sensing oxygen delivery as oxygen extraction is relatively high compared with oxygen delivery.57 ,58 Although oxygen delivery to the kidneys as a whole organ is high due to high renal blood flow, DO₂ is reduced to the renal medulla as oxygen can pass from arterioles to the postcapillary venous system by shunt diffusion due to the parallel organisation of arterial and venous systems.42 Consequently, the peritubular cellular PO₂ is low. It falls to even lower levels following reductions in DO₂ either as a result of hypoxaemia or low haematocrit.

5.1.3 The oxyhaemoglobin dissociation curve and the Bohr effect

The oxygen–haemoglobin dissociation curve shows the relationship between haemoglobin SaO₂ and PaO₂ (figure 5 and table 10). The curvilinear shape of the curve has two particular features that serve to protect from tissue hypoxia. The upper portion of the curve is flat which means that a marked fall in PaO₂ is still compatible with a nearly complete oxygen saturation. Second, the steep portion of the curve means that despite rapidly falling oxyhaemoglobin saturation, the PO₂ remains relatively well preserved. This property facilitates the continued delivery of oxygen to the tissues despite progressively lower levels of saturation.59

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

Total carbon dioxide dissociation curve. HbCO₂, carbamino haemoglobin; PCO₂ carbon dioxide tension.

The oxygen-carrying capacity of haemoglobin is regulated in response to other metabolic factors to increase the efficiency of oxygen pick-up and delivery.60 The curve is shifted to the right because of an increase in temperature, PaCO₂ and hydrogen ion concentration (low pH) or an increase in 2,3-diphosphoglycerate (DPG). This shift to the right enhances release of oxygen to the tissues and improves oxygen availability, named the Bohr effect (figure 5).

The converse holds true for the lungs where lower carbon dioxide levels favour oxygen loading of haemoglobin. Consequently, SaO₂ cannot accurately predict PaO₂ and vice versa, but table 10 gives approximate equivalents.

Chronic hypoxaemia increases 2,3-DPG in erythrocytes, shifting the dissociation curve to the right and therefore increasing oxygen delivery to the tissues. Levels of 2,3-DPG are reduced in stored blood reducing to zero after about 2 weeks. This may reduce oxygen delivery to the tissues but the magnitude of this effect is not thought to be clinically significant. Once transfused, levels of 2,3-DPG increase to about 50% normal after about 6 hours and back to normal within 48 hours.

5.1.4 Regulation of DO₂

Acutely, the cardiovascular effects of hypoxaemia will tend to counter the impact of lower CaO₂ on DO₂ by increasing cardiac output through increased heart rate and myocardial contractility and by decreasing afterload by reducing systemic vascular resistance.61 ,62 Anaemic hypoxia is sensed in the aortic body, presumably owing to lower perfusion relative to oxygen consumption. Consequently, the aortic body can act as a sensor of reduced oxygen delivery as a result of either low PO₂ or low haematocrit (unlike the carotid body).55

At local tissue level, oxygen delivery can be adjusted to changes in local oxygen consumption. For example, exercising skeletal muscle receives a greater proportion of total cardiac output than resting skeletal muscle. This relates in part to hypoxaemia recruiting a larger proportion of the capillary bed by the relaxation of pericytes, and also through arteriolar vasodilation.63

5.2.1 Hypoxaemic hypoxia

Hypoxaemic hypoxia in blood leaving an alveolar capillary unit in the lung may be induced by alveolar hypoxia or incomplete gas exchange (see definition in section 3.2). This may be due to a reduction in inspired oxygen partial pressure, that is, high altitudes, intrapulmonary shunt, V/Q mismatching, alveolar hypoventilation or diffusion impairment.

The alveolar gas equation calculates the oxygen level in the alveolus using the following formula:

where PAO₂ and PACO₂ represent alveolar levels of oxygen and carbon dioxide, RER is the respiratory exchange ratio or the ratio of carbon dioxide production to oxygen consumption and inspired PO₂ (PIO₂)=FiO₂×(barometric pressure (100 kPa, 750 mm Hg)—water vapour pressure (∼6 kPa, 45 mm Hg)).

Considering this equation, alveolar hypoxia can be induced by decreased PIO₂ or increased PACO₂. If an alveolar capillary unit is relatively underventilated for its degree of perfusion (low V/Q ratio), PACO₂ will rise due to inadequate clearance and thus PAO₂ will fall. This may happen for a number of reasons such as increased dead space ventilation during the non-fatiguing pattern of shallow respiration in respiratory failure or abnormal lung mechanics in advanced COPD. In diseases that cause global hypoventilation such as respiratory muscle weakness, effectively all areas of lung have low V/Q ratios and this explains the hypercapnia and hypoxaemia associated with these conditions.

An extreme form of low V/Q pathophysiology occurs in intrapulmonary and extrapulmonary shunt where no gas exchange occurs at all. An example of intrapulmonary shunt is when the airway to a lung segment is obstructed by mucus creating an area of lung tissue that is perfused but not ventilated, thus acting as a right-to-left shunt. An example of extrapulmonary shunt is a ventricular septal defect with right-to-left shunting in Eisenmenger's syndrome.

In health and at rest, PO₂ has equilibrated with the tissue PO₂ across the alveolar capillary membrane one-third of the way along the length of the capillary. With increased thickness of this membrane, as in fibrotic lung disease, equilibration may take longer and an oxygen gradient may persist between the alveolus and blood at the end of the capillary. The overall effect of this when multiple alveolar capillary units are affected will lead to an increased alveolar-to-arterial (A–a) gradient. This is exacerbated during exercise, when capillary transit time decreases.

5.2.2 Other mechanisms of hypoxia

Anaemia and carbon monoxide poisoning may result in ‘anaemic hypoxia’ by reducing oxygen-carrying capacity (see definitions in section 3.2). A low cardiac output state will reduce oxygen delivery even in the absence of hypoxaemia. Tissue hypoxia may develop in these circumstances and this is often termed ‘stagnant hypoxia’. ‘Histotoxic hypoxia’ is due to an inability to metabolise oxygen at the mitochondrial level and may be due to severe sepsis or is seen in certain types of poisoning.

5.2.3 Hyperoxia

Hyperoxia can be caused by hyperoxaemia and polycythaemia. Considering again the alveolar gas equation in the previous section, hyperoxaemia can only exist in the presence of high PIO₂ or low PACO₂ (resulting from hyperventilation). The term ‘hyperoxia’ could technically be used to describe a patient with polycythaemia without hyperoxaemia, but most clinicians use the term only to describe situations in which the PaO₂ is raised.

5.3.1 Normal carbon dioxide homoeostasis

Carbon dioxide is principally carried in the blood in three forms: carbon dioxide, bicarbonate and as a carbamino compound.64 In the normal physiological range of 4.6–6.1 kPa (34–46 mm Hg) the relationship between PaCO₂ and carbon dioxide content can be considered linear (figure 6).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7

Effect of PaCO₂ on ventilation with interaction of acidosis and hypoxaemia. PaCO₂, arterial carbon dioxide tension; PCO₂ carbon dioxide tension; PO₂, oxygen tension.

5.3.2 Regulation of carbon dioxide

PaCO₂ is sensed at the peripheral54 and central chemoreceptors (in the medulla oblongata) by its effect on intracellular pH.65 Consequently, the regulation of PaCO₂ is intimately related to pH homoeostasis (figure 7).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8

Example of oxygen alert card.

It is often not appreciated how V/Q matching relates to PaCO₂. As discussed in section 5.2.1, alveolar capillary units with a low V/Q ratio have increased PACO₂. Because of the high solubility and diffusibility of carbon dioxide, there is little A–a gradient for carbon dioxide at the end of the capillary, so blood leaving low V/Q alveolar capillary units has a high PCO₂.

As described above, areas of low V/Q are usually minimised through HPV. It is also thought that a high PCO₂ can cause pulmonary vasoconstriction, adding to the homoeostatic mechanisms of the lung, matching perfusion to ventilation.66 ,67 As the relationship between PCO₂ and carbon dioxide dissolved in the blood is approximately linear over the physiological range (unlike oxygen), blood does not become saturated with carbon dioxide and therefore a high pulmonary venous PCO₂ from low V/Q areas can be partially balanced by a low pulmonary venous PCO₂ from high V/Q areas. Consequently, by increasing overall alveolar ventilation, the cardiopulmonary system is able to prevent hypercapnia despite significant V/Q mismatch or shunt, unless respiratory mechanics are limiting.

As with the carriage of oxygen (Bohr effect), there is a reciprocal relationship between PO₂ and carbon dioxide carriage. This is known as the Haldane effect.60 Deoxygenated haemoglobin has a higher carbon dioxide buffering capacity than oxygenated haemoglobin. This favours carbon dioxide pick-up in the systemic venous circulation and carbon dioxide offloading in the lungs.

Acutely, carbon dioxide acts as a sympathomimetic on the heart: it increases heart rate and stroke volume, increasing cardiac output. Peripherally it causes vasodilation, reducing systemic vascular resistance. Locally, carbon dioxide acts as a vasodilator, thus diverting blood flow to tissues with high metabolic demand. The resulting physical signs of hypercapnia are described in section 7.2.

5.4.1 Mechanisms of hypercapnia

The mechanisms of hypercapnia are simpler than hypoxaemia and there are four possible causes:68

1. Increased concentration of carbon dioxide in the inspired gas.

2. Increased carbon dioxide production.

3. Hypoventilation or ineffective ventilation.

4. Increased external dead space.

5.4.2 Hypoventilation and hyperventilation

Hypoventilation may be physiological—for example, in the face of a metabolic alkalosis. Pathological hypoventilation will occur either when the respiratory muscles are unable to ventilate the lungs sufficiently because they are pathologically weak or they are unable to overcome abnormal lung mechanics such as during an exacerbation of COPD. Reduced respiratory drive caused by drugs with sedative properties or by neurological injury will also produce hypoventilation.

Using the same physiological principles but in reverse, hyperventilation for any reason will produce hypocapnia. This may occur during pure hyperventilation during an anxiety attack or during physiological hyperventilation.

5.6.1 Optimising PaO₂

The physiology of oxygen therapy has already been discussed in the previous section. However, increasing FiO₂ is only one component in increasing oxygen uptake in the lungs. Other key manoeuvres to ensure oxygen delivery to the alveolar capillary bed include:

• Maintaining a satisfactory airway.

• Ensuring adequate alveolar ventilation.

• Reversing any respiratory depressants such as narcotics.

• Invasive ventilation or NIV where necessary.

• Treating airflow obstruction by bronchodilation or sputum clearance techniques.

• Optimising transfer factor (diffusion capacity).

• Treatment of pulmonary oedema.

5.6.2 Optimising oxygen carriage

Oxygen is carried in blood mainly by haemoglobin with only a very small amount of oxygen dissolved in the blood itself. Adequate haemoglobin is therefore essential for CaO₂. The ideal haemoglobin level for optimal CaO₂ and therefore for optimal DO₂ has long been a subject of debate. Previous practices have favoured haemoglobin levels close to 100 g/L (10 g/dL), providing adequate CaO₂ as well as reducing viscosity of blood for better perfusion in critically ill patients. However, studies by Canadian researchers in the late 1990s have shown that haemoglobin levels of 70 g/L (7 g/dL) were as safe as higher levels and may produce fewer complications in the critically ill.72 However, this study was conducted using non-leucocyte depleted blood and it is possible that some of the infective complications in the group who were given more transfusions might have been avoided by the use of leucocyte-depleted blood. The optimal transfusion target for critically ill patients therefore remains the subject of ongoing discussion among experts in critical care medicine. Although the issue of optimal haemoglobin in patients with unstable or symptomatic coronary artery disease is not settled, haemoglobin levels of 100 g/L (10 g/dL) are recommended for adequate DO₂ (see box).

5.6.3 Optimising delivery

Besides adequate CaO₂ and PaO₂, delivery of oxygen depends on adequate flow of oxygenated blood. Cardiac output in turn depends on adequate blood (circulating) volume, adequate venous return and adequate and optimal myocardial function. To avoid tissue hypoxia, attention must therefore be paid to the volume status of the patient and the adequacy of cardiac function, as well as initiating oxygen therapy. In severely shocked patients (eg, cardiogenic shock, septic shock), invasive monitoring and inotropic/vasopressor therapy will usually be indicated in appropriate higher dependency environments. Some studies have shown that deliberately increasing oxygen delivery in critically ill patients as well as high-risk surgical patients reduces organ failure, reduces length of ICU stay and, most importantly, improves mortality.73–76 However, there is a substantial body of evidence showing no favourable effect on morbidity and mortality by achieving supranormal values for the cardiac index or normal values for mixed venous oxygen saturation.77–79

6.1.1 Desirable oxygen saturation ranges in acute illness

• Acute hypoxaemia is considered dangerous to healthy participants below a PaO₂ of about 6 kPa (45 mm Hg) or an SaO₂ of about 80% due to impaired mentation and risk of tissue hypoxia, but patients with acute illness or chronic organ disease or ischaemia are likely to be at risk above this level and critical care guidelines recommend aiming to achieve saturation >90%.

• Changes in physiological ‘track and trigger’ systems such as the NEWS96 may occur in acute illness with either no change or only a small change in SaO₂ levels.

• Critical illness may present initially with only a small fall in SaO₂ level because of compensating mechanisms.

• The upper end of the recommended range in this guideline (98%) is the upper limit of SaO₂ in healthy adults.

• The lower end of the suggested target saturation range (94%) is close to the lower end of the normal range and ensures that the SaO₂ remains above 90% even if it falls slightly below the target range.

6.2.1 Hyperoxaemia has been shown to be beneficial in the following clinical situations

• Carbon monoxide and cyanide poisoning (see section 8.10.7)

• Spontaneous pneumothorax (see section 8.11.6)

• Some postoperative complications (see section 8.15.3)

• Cluster headache (see section 8.13.7)

6.2.2 Other potential benefits and potential harms of oxygen therapy in non-hypoxaemic patients

• Most guidelines for CPR and the care of patients with critical illness recommend the use of the highest feasible oxygen concentration in the initial stages of resuscitation. Although these recommendations are not evidence-based, it is unlikely that controlled trials would ever be undertaken using different levels of oxygen therapy in these emergencies and it seems intuitive to maximise oxygen delivery for critically ill patients with circulatory collapse and respiratory failure. However, randomised trials have been undertaken of resuscitation of neonates breathing room air or oxygen and the unexpected outcome of a Cochrane review was that the survival was better when room air was used.97 This surprising finding cannot be extrapolated to adult patients, but it does emphasise the need for clinical trials even in areas where one might intuitively believe that oxygen would be beneficial. Furthermore, there is theoretical evidence that patients who have had a restoration of spontaneous circulation after cardiac arrest may be managed more safely with <100% inspired oxygen.98 ,99 Recent observational studies suggesting possible harmful outcomes from hyperoxaemia in survivors of cardiopulmonary resuscitation are discussed in section 8.10.1.

• It has been shown that early intervention to increase oxygen delivery to the tissues in critically ill patients as well as high-risk surgical patients reduces organ failure, reduces length of ICU stay and, most importantly, improves survival.73–76 Increased oxygen delivery in part involves oxygen therapy, but these studies did not show any benefit from aiming at supraphysiological oxygen delivery. There is some evidence that hyperoxaemia may be associated with increased mortality in patients on ICUs, this is discussed in section 8.10.2.

• Reported benefits of oxygen therapy in healing of established wounds and in treatment of wound sepsis are controversial. Hyperbaric oxygen reduced the risk of amputation in patients with chronic diabetic foot ulcers and may improve the chance of healing over 1 year, but the Cochrane reviewers had concerns about the size and quality of existing studies and recommended further trials.100 It is not known if conventional oxygen therapy has any effect on wound healing.101

• The potential benefit of oxygen for prevention of postoperative wound infection is discussed in section 8.15.3.

• Through relief of breathlessness and work of breathing, oxygen therapy may decrease carbon dioxide production and consequently offset some of the potential increase in PaO₂ that might otherwise occur due to the mechanisms described in section 6.3.1. However, there are no controlled trials supporting the use of oxygen for this indication.

6.3.1 Respiratory system

The most significant effect of high-concentration oxygen therapy on the respiratory system is hypercapnic respiratory failure in a population of vulnerable patients as described more fully in section 8.12. This does not occur in the absence of significant pulmonary disease or musculoskeletal disease affecting the thorax, and it can occur while the PaO₂ is still within the normal range or slightly below normal, especially in patients with exacerbated COPD for whom the baseline oxygen saturation level may be well below the lower limit of normal.

It has been stated for several decades that ‘hyperoxaemia causes hypercapnic respiratory failure by producing decreased respiratory drive in patients with intrinsic lung disease, such as COPD’, and many—if not most—medical textbooks from the 1960s to the present time refer to ‘loss of hypoxic drive’ as the main cause of hypercapnia and acidosis when high-concentration oxygen is given to patients with an AECOPD. This assertion is usually attributed to Campbell who championed the concept of controlled oxygen therapy in the 1960s. However, Campbell87 has been widely misquoted. What he actually said in 1967 was as follows: “It is usual to attribute the rise in PaCO₂ in these patients to removal of the hypoxic drive to ventilation but I share the doubts of Pain and co-workers108 that this is the whole story; changes in the pulmonary circulation may also be important.” Most but not all subsequent studies48 ,49 ,108–113 have shown that Campbell was correct in this assumption as discussed in section 4.2. Another curious feature of hypercapnia in AECOPD is that it is not universal.87 Some patients with COPD are prone to repeated episodes of hypercapnic respiratory failure and others may not ever suffer from this complication. Even among patients with COPD with chronic hypercapnia, not all will develop an increased carbon dioxide level (and acidosis) during acute exacerbations. The theory of ‘loss of hypoxic drive’ as the cause of hypercapnia is further confounded by the observation that PaCO₂ continues to rise as PaO₂ is increased above 13 kPa (100 mm Hg), which has little impact on decreasing ventilation and most patients with respiratory acidosis during an exacerbation of COPD have PaO₂ above 10 kPa, equivalent to saturation above about 93%.43 Therefore, while a small reduction in ventilation may be a contributing factor to the rise in carbon dioxide levels during oxygen therapy in COPD, the major factor is the worsening of V/Q matching. Additional effects of increasing FiO₂ will relate to atelectasis and perhaps worsening airflow obstruction due to increased viscosity. In conditions where there is little intrinsic lung disease but significant respiratory muscle weakness, loss of hypoxic respiratory drive will be a greater factor in the development of hypercapnia. However, HPV remains a significant regulator of V/Q matching even in non-diseased lung.

While the mechanism of oxygen-induced hypercapnia remains controversial it is likely there are at least six mechanisms responsible for this. These are ranked in likely order of importance below:

• V/Q mismatch,

• Ventilatory drive,

• Haldane effect,

• Absorption atelectasis,

• Higher density of oxygen compared with air,

• Rebreathing can occur if low oxygen flow rates are used through face masks.

6.3.2 Rebound hypoxaemia following sudden cessation of supplementary oxygen therapy

Patients who have developed decompensated hypercapnic respiratory failure following high-concentration oxygen therapy face a further significant danger of rebound hypoxaemia if oxygen is suddenly withdrawn in an attempt to correct the effects of excess oxygen therapy.116 ,117

Rebound hypoxaemia can be explained using the alveolar gas equation and, given its importance, is best illustrated with a working example (box). For the purposes of simplicity, this example makes several assumptions such as a constant respiratory exchange ratio and A–a gradient between stages 1 and 3. It also assumes that ventilation remains unchanged. Although when PaO₂ falls to 3.4 kPa on removal of oxygen, ventilation will rise, by definition it will not be able to rise sufficiently to meet the need to clear the carbon dioxide stores for the same reason that hypercapnic respiratory failure developed in the first instance. Rebound hypoxaemia is a major risk and could cause death. Consequently, this guideline will recommend that oxygen therapy be stepped down gradually while monitoring saturation continuously.

Good practice point

• Sudden cessation of supplementary oxygen therapy can cause rebound hypoxaemia with a rapid fall in PO₂ to below the tension that was present prior to the start of supplementary oxygen therapy.

6.3.3 Cardiovascular and cerebrovascular system

The effects and potential harms of hyperoxaemia on the circulation have been summarised by Thomson et al105 in an editorial which made a strong case for more trials. Hyperoxaemia causes coronary and cerebral vasoconstriction and, if the haematocrit is sufficiently low, this may theoretically cause paradoxical tissue hypoxia because of overall reduction in DO₂. There have been reports of possible harm from the administration of oxygen to non-hypoxaemic patients with strokes of mild-to-moderate severity or to non-hypoxaemic patients with myocardial infarction as discussed in sections 8.13.1 and 8.13.2. Thomson et al105 have suggested that oxygen should be ‘prescribed, administered and monitored with care’ in order ‘to achieve optimal tissue oxygenation’, not maximal oxygenation. This view was proposed by other authors such as Bryan and Jenkinson118 in the 1980s, but standard medical practice had not taken note of this advice prior to the publication of the first version of this guideline in 2008. Because there are no published data suggesting benefit from hyperoxaemia for most medical conditions and because of the theoretical risks, optimal management should aim for physiological oxygenation. Targets for oxygen therapy in specific circumstances, with evidence, are discussed in section 8.

6.3.4 Reactive oxygen species, tissue toxicity and reports of increased mortality

Aside from the potentially detrimental physiological effects of hyperoxaemia, the toxic effects mediated by reactive oxygen species (ROS) have potential risk.118 Excess ROS are generated in the presence of high tissue PO₂ in the form of hydrogen peroxide and superoxide, causing oxidative stress and free radical damage.119 At physiological levels ROS act as signalling molecules, but at higher levels they are cytotoxic, notably being released by primed neutrophils as a host defence mechanism. It is thought that ROS are responsible for the development of bronchopulmonary dysplasia in ventilated hyperoxygenated premature infants120 and reperfusion injury postmyocardial infarction.121 It has been known since the 19th century that prolonged exposure to high concentrations of oxygen in animal models leads to diffuse alveolar damage, haemorrhage, alveolar collapse, infiltration of inflammatory cells, necrosis, apoptosis and injury to the endothelium and epithelium in the lungs which causes death in the rat model after 4 days of exposure to 73% oxygen at one atmosphere.122 Griffith et al123 demonstrated in 1986 that normal human lungs are injured by breathing 30–50% oxygen for 45 hours with leakage of albumin into bronchoalveolar lavage fluid and other markers of lung injury.123

6.3.5 Delay in recognition of physiological deterioration

It was previously believed that a high FiO₂ is protective and gives patients a margin of safety. However, Downs102 and Beasley et al106 have argued that unstable patients may actually be placed at risk by the precautionary use of high-concentration oxygen therapy.107 ,124 During physiological deterioration, a patient given high-concentration oxygen therapy would have a normal or high pulse oximeter reading masking a progressive decline in the PaO₂/FiO₂ ratio and therefore not alerting staff to impending deterioration requiring mechanical support. A worked example of this is shown in reference.106 Furthermore, a patient who deteriorated physiologically while at a low FiO₂ would be detected early by pulse oximetry and could have the FiO₂ increased while being transferred to an ICU, whereas a patient who was already receiving a high FiO₂ would desaturate more slowly but, when the oximeter eventually detected desaturation, there would be fewer treatment options because increasing the FiO₂ further would have little effect.88 ,92 ,107

6.3.6 Acute lung injury in patients with acute paraquat poisoning, bleomycin lung injury and acid aspiration

Oxygen is known to be hazardous to patients with paraquat poisoning125 ,126 and oxygen potentiates bleomycin lung injury and may potentiate lung injury from aspiration of acids.127–129 Further details concerning these conditions are given in section 8.13.4.

6.3.7 Summary of risks of hyperoxaemia and supplemental oxygen therapy

Physiological risks

1. Worsened V/Q mismatch.

2. Absorption atelectasis.

3. Coronary and cerebral vasoconstriction.

4. Reduced cardiac output.

5. Damage from oxygen free radicals.

6. Increased systemic vascular resistance.

Clinical risks

1. Worsening of hypercapnic respiratory failure.

2. Delay in recognition of clinical deterioration.

3. Potentially worse outcomes in mild-to-moderate stroke.

4. Specific risk in patients with previous bleomycin lung damage or with paraquat poisoning or acid aspiration.

5. Unknown risk–benefit balance in acute coronary artery disease with normal oxygen saturation.

6. Association with increased risk of death in survivors of cardiac arrest and among patients on ICUs.

7. Uncontrolled supplemental oxygen therapy can be harmful to patients who are at risk of hypercapnic respiratory failure, especially if the PaO₂ is raised above 10 kPa.

8. High-concentration oxygen therapy to produce hyperoxaemia (above normal oxygen saturation) and has been associated with increased risk of death in some patient groups (eg, patients with mild and moderate strokes, survivors of cardiac arrests and ICU patients; see section 8).

6.4.1 Effects of a raised blood carbon dioxide level

Nervous system

Carbon dioxide exerts its effect either directly or as a consequence of acidosis. Hypercapnia increases cerebral blood flow and thereby may influence the cerebrospinal fluid pressure. It is the main factor influencing the intracellular pH which has an important effect on cellular metabolism. It exerts an inert gas narcotic effect similar to that of nitrous oxide. It influences the excitability of neurones particularly relevant in the reticular activating system. Carbon dioxide can induce narcosis when the PaCO₂ rises above 12–16 kPa (90–120 mm Hg).130

Pulmonary circulation

An elevated PACO₂ causes vasoconstriction in the pulmonary circulation although the effect is less marked than that of hypoxaemia.139 In healthy participants an end expiratory PCO₂ of 7 kPa (52 mm Hg) increases pulmonary vascular resistance by 32% which, along with raised cardiac output, increases mean pulmonary artery pressure by 60%.140 Changes in pH are thought to be the primary factor responsible for carbon dioxide-mediated changes in the pulmonary vasculature.141 ,142 Consequently, as with HPV, changes in PACO₂ help to match perfusion to ventilation.

Respiratory system

As explained in section 5.2.1, a raised carbon dioxide level may worsen hypoxaemia and its effects because the concentration of carbon dioxide in the alveolar gas reduces that of oxygen if the concentration of nitrogen remains constant. Also an increase in PaCO₂ shifts the oxygen dissociation curve to the right.

Cardiovascular system

In general, both hypercapnia and acidosis have direct depressant effects on cardiac myocytes and vascular smooth muscle cells.143 These effects are normally opposed by the increase in catecholamines caused by the raised PaCO₂. The overall effect of carbon dioxide on the cardiovascular system is therefore unpredictable. In artificially ventilated children a rise in carbon dioxide increases cardiac output and reduces total peripheral resistance and blood pressure tends to rise.144 Although an increase in carbon dioxide depresses heart rate, tachycardia is more common because of the effects of catecholamine stimulation over-riding the depressant effects on the heart. Arrhythmias have been reported but are seldom clinically significant in normal participants. Carbon dioxide is a systemic vasodilator.

Kidneys

Renal blood flow and glomerular filtration rate are reduced abruptly in the presence of levels of PaCO₂ above about 65 mm Hg (8.7 kPa).88 ,145

Blood electrolyte levels

The acidosis that accompanies hypercapnia may cause a rise in potassium if the acidosis is severe and sustained.

Endocrine system

Hypercapnia increases plasma levels of endogenous adrenaline and noradrenaline.

6.4.2 Clinical signs

The clinical signs of hypercapnia are produced by the physiological changes described above and are described in detail in section 7.2.1.

7.1.1 Clinical assessment of breathless patients and assessment of cyanosis

Clinicians examining a critically ill patient should remember the ‘ABC’ of emergency medicine. In the case of critically ill patients it may be necessary to secure the airway and resuscitate a patient before a detailed history can be obtained and before a full physical examination can be undertaken. Additionally, it is important to remember that supplemental oxygen is given to improve oxygenation but it does not influence the underlying cause(s) of hypoxaemia which must be diagnosed and treated as a matter of urgency.

In assessing an ill patient the SpO₂ level is only one of several physiological variables that should be monitored. Many patients with sudden acute illness such as postoperative pulmonary emboli will have a sudden alteration in physiological ‘track and trigger’ variables as assessed by the mEWS system94 ,95 ,163 and the NEWS developed by the Royal College of Physicians.96 Such patients may have only a small fall in SpO₂ due to physiological compensation mechanisms such as increased ventilation. Clinicians therefore need to be alert for falls in SpO₂ even within the recommended target ranges.

Although the use of ‘track and trigger’ systems are strongly recommended, the guideline group had some reservations of their use in patients at risk of hypercapnic respiratory failure, and in particular, the 2012 version Royal College of Physicians NEWS as there was no option to set a target range for oxygen saturations.164 ,165 Patients at risk of hypercapnic respiratory failure who are within the target range of 88–92% as recommended in this guideline would score 2–3 EWS points, and may prompt nursing staff to increase the inspired oxygen level to achieve ‘normal’ oxygenation (>96%) and put the patient at risk. The guideline group recommends use of the 2017 update of the NEWS chart which allows appropriate adjustments to the scoring for saturation for patients at risk of hypercapnic respiratory failure. Patients should score EWS points for saturations below the target range or if significantly above the target range while breathing oxygen. The 2017 NEWS chart will have a special oximetry section for patients at risk of hypercapnia.

Evidence statement

• There are no controlled trials of different methods of assessment of acutely unwell patients. Advice on clinical assessment is based on expert opinion and retrospective studies of outcomes (evidence level 4).

Recommendations

B1: Fully trained clinicians should assess all acutely ill patients by measuring respiratory rate, pulse rate, blood pressure and temperature, and assessing circulating blood volume and anaemia. Expert assistance from specialists in intensive care or from other disciplines should be sought at an early stage if patients are thought to have major life-threatening illnesses and clinicians should be prepared to call for assistance when necessary including a call for a 999 ambulance in prehospital care or a call for the resuscitation team or ICU outreach team in hospital care (grade D).

Traditional clinical assessment of hypoxaemia involves clinical inspection of the skin and buccal mucous membranes to decide whether central cyanosis is present or absent. This is a difficult clinical skill, especially in poor lighting conditions. Clinical assessment of hypoxaemia is made even more unreliable by the presence of anaemia or polycythaemia. Some patients may have peripheral cyanosis due to poor peripheral circulation in the presence of normal SaO₂. Several studies have shown that hypoxaemia is often not recognised by emergency medical service providers, especially if the patient does not complain of respiratory distress.166–168 A systematic review of the literature in 2005 reported that most hypoxaemic patients had at least one vital sign abnormality but skin colour was a poor indicator of hypoxaemia compared with pulse oximetry.39 For these reasons it is recommended that clinicians should not rely on visual assessments of ‘cyanosis’ but should instead use pulse oximetry to obtain an accurate assessment of a patient's oxygen saturation.

The nature of a patient's presenting illness may make hypoxaemia a likely outcome, thus prompting a careful clinical search for evidence of cyanosis complemented by urgent pulse oximetry. This situation applies to many common acute illnesses such as heart failure, COPD exacerbation, pneumonia and pulmonary embolism. A study of 2276 patients with pneumonia showed that hypoxaemia was independently associated with six risk factors: age >30 years (OR 3.2), COPD (OR 1.9), congestive heart failure (OR 1.5), respiratory rate >24/min (OR 2.3), altered mental status (OR 1.6) and chest radiographic infiltrate involving >1 lobe (OR 2.2).36 Acutely ill patients with significant hypoxaemia are likely to have an increased pulse rate or respiratory rate and, for this reason, usually score several NEWS points.94 ,95 ,163 ,165 ,169

The respiratory rate is the single best predictor of severe illness.95 However, many patients with marked hypoxaemia may present with non-specific findings such as restlessness and confusion rather than breathlessness, and oxygen saturation has been shown to be an independent predictor of mortality in multivariate analysis of the outcome of emergency medical admissions.19 ,170 Furthermore, the work of Thrush et al171 on normal volunteers has shown that heart rate, blood pressure and respiratory rate are not reliable indicators of hypoxaemia down to saturation levels as low as 70%. This would suggest that the changes in vital signs which are seen in most hypoxic patients are due to the underlying illness rather than hypoxaemia per se.

Hypoxaemia may be associated with increased or decreased ventilation. Although some hypoxaemic patients may have reduced levels of ventilation as a causative factor, the majority of hypoxaemic patients have increased minute ventilation in an attempt to increase the blood oxygen level. For example, a patient with an opioid overdose may have reduced ventilation causing hypoxaemia despite having structurally normal lungs, whereas a patient with pneumonia or major pulmonary embolism may have significant hypoxaemia due to V/Q mismatch despite an increased level of ventilation. The first patient in this example may appear peaceful and non-distressed despite significant hypoventilation and hypoxaemia, while the second patient is likely to have increased ventilation and tachycardia. The clinician therefore needs to make separate assessments of a patient's oxygen saturation and level of ventilation.

Having completed the history and rapid assessment of the patient, more detailed physical examination may reveal signs of an illness such as major pleural effusion, major pneumothorax or unexpected heart failure that may prompt the clinician to anticipate the presence of hypoxaemia.

Good practice points for clinical assessment of patients with suspected hypoxaemia

• The medical history should be taken when possible in an acutely breathless patient and may point to the diagnosis of a particular acute illness such as pneumonia or pulmonary embolism or an exacerbation of a chronic condition such as COPD, asthma or heart failure.

• Never discontinue oxygen therapy to obtain an oximetry measurement on room air in patients who clearly require oxygen therapy.

• Physical examination should be undertaken urgently. This may provide evidence of a specific diagnosis such as heart failure or a large pleural effusion, but it is common for the cause of breathlessness to remain undiagnosed until the results of tests such as chest radiographs are available.

• Record arterial oxygen saturation measured by pulse oximetry (SpO₂) and consider blood gas assessment in patients with unexplained confusion and agitation as this may be presenting feature of hypoxaemia and/or hypercapnia (cyanosis is a difficult physical sign to record confidently, especially in poor light or with an anaemic or plethoric patient).

• Carefully measure respiratory rate and heart rate because tachypnoea and tachycardia are more common than a physical finding of cyanosis in hypoxaemic patients.

• Appropriate changes should be made to any ‘track and trigger’ system used to allow for a lower target range in patients at risk of hypercapnic respiratory failure. These patients should score no EWS points for saturation if within their target range and they should score points if the oxygen saturation falls below the target range or if the saturation rises above the target range while breathing oxygen. The 2017 update of the NEWS chart has a special section for oximetry measurements for use with patients who have target range 88–92% and it is recommended that the 2017 NEWS chart should be used in all hospitals (see recommendation B4).

• The presence of a normal SpO₂ does not negate the need for blood gas measurements especially if the patient is on supplemental oxygen therapy. Pulse oximetry will be normal in a patient with a normal oxygen tension (PO₂) but abnormal carbon dioxide tension (PCO₂) or with a low blood oxygen content due to anaemia. For this reason, blood gases and full blood count tests are required as early as possible in all situations where these measurements may affect patient outcomes.

• All clinical staff who use oximeters must be trained in their use and made aware of the limitations of oximetry. (Oximetry is a valuable clinical tool but subject to artefact and errors of interpretation).

7.1.2 Value and limitations of pulse oximetry

Clinical assessment of hypoxaemia has been revolutionised by the advent of pulse oximetry in much the same manner as the clinical assessment of blood pressure was transformed by the invention of the sphygmomanometer. However, it was common in the past to see patients with acute respiratory illness who have had multiple measurements of their blood pressure but no record made of their oxygen saturation, peak expiratory flow or FEV₁. In addition to the clinical consequences of underassessment, Howes et al172 and Macnab et al173 have reported that the availability of a pulse oximeter was highly cost-effective because the finding of normal oximetry (>94%) in many patients allowed paramedics to use oxygen less frequently with a potential financial saving of up to $2324 (∼£1200) per ambulance per annum. The availability now of highly portable, low-cost pulse oximeters should make the measurement of oxygen saturations in many clinical scenarios commonplace and used as the ‘fifth vital sign’. We would encourage all clinical staff who are involved in the care of acutely ill patients to carry and use one so that measurement of oxygen saturations is as normal a part of assessment of a patient as listening to their chest.

Pulse oximetry measures haemoglobin oxygen saturation by detecting the absorption of light at two specific wavelengths that correspond to the absorption peaks of oxygenated and deoxygenated haemoglobin. Oximeters are less reliable at low saturation such as 80%, but modern oximeters reflect the SaO₂ accurately at saturation above about 88%.174–178 In almost all clinical circumstances covered by this guideline, patients with a saturation below 88% will be given intensive therapy to bring the saturation up to at least 90%, so the inaccuracy of the instruments at very low saturation levels should not affect patient management.

In one study of 123 adult patients who had simultaneous measurements of pulse oximetry and SaO₂ measured in ABGs, the 95% CI for the median difference ranged from −0.6 to +0.5%.176It has been estimated that an oxygen saturation of 92% or above measured by pulse oximetry has a sensitivity of 100% and specificity of 86% for excluding hypoxaemia defined as an arterial oxygen partial pressure below 60 mm Hg (8 kPa).179

Oximetry may be less accurate in acutely ill patients on ICUs, but there are no direct comparisons of the accuracy of pulse oximetry in critically ill patients compared with stable patients and healthy individuals. The study of Perkins et al180 showed a mean SpO₂ of 94.6% compared with a mean SaO₂ of 95.9% from 1132 simultaneous oximeter and ABG measurements on an ICU. Fortunately, this average difference of 1.3% was lower for pulse oximeter readings, thus allowing a margin of safety in most cases. This study also showed that fluctuations in oxygen saturation measured by oximetry tended to be greater than changes in SaO₂ measured with samples from an indwelling radial artery catheter.

Although oximetry is widely used, there are few clinical studies examining its utility. The Cochrane meta-analysis of the use of oximetry in perioperative monitoring of more than 20 000 patients failed to show any reduction in complications or deaths where oximetry was used, although oxygen was given more often to patients who were monitored with pulse oximetry.181 The authors suggested that the correction of modest hypoxaemia probably does not have much effect on clinical outcomes.

Pulse oximetry gives no information concerning pH, PCO₂ or haemoglobin level. Blood gases and full blood count tests are therefore required as early as possible in all situations where these measurements may affect patient outcomes.

The accuracy of pulse oximetry is diminished in patients with poor peripheral perfusion which may occur chronically in conditions such as systemic sclerosis or acutely in patients with hypotension or hypovolaemia. However, it has been suggested that many types of oximeter may remain accurate at arterial pressures as low as 20 mm Hg so long as the machine is able to obtain a reading despite the low pulse pressure.182 Most oximeters give an indication of the pulse signal strength. It is important to ensure that the oximeter has a good signal if technically possible, and the probe may need to be tried on different fingers or toes or on the earlobe to obtain the best available signal for the individual patient. There are some patients with poor perfusion for whom pulse oximetry measurements cannot be made. This includes patients with cold peripheries (eg, Raynaud's phenomenon), severe hypotension and peripheral ‘shut down’.

It must be remembered that oximetry gives a normal reading for oxygen saturation in most patients with anaemia because the oxygen saturation of the available haemoglobin is normal although the total amount of haemoglobin available for oxygen transport is reduced. These patients have normal oxygen saturation levels despite having ‘anaemic hypoxia’ which may cause considerable reduction in the total oxygen content of the blood. It is often not recognised that a patient with an SpO₂ of 98% but a haemoglobin of 7 g/dL ((7×0.98×1.34)=9.2 mL O₂/dL) will have a greatly reduced blood oxygen content compared with a patient with a haemoglobin of 15 g/dL and a saturation of 85% ((15×0.85×1.34)=17 mL O₂/dL; each g/dL haemoglobin when fully saturated carries 1.34 mL oxygen).

The accuracy of oximetry is unreliable in the presence of carbon monoxide or methaemoglobin. Both of these substances have similar light absorption characteristics to oxyhaemoglobin so an apparently normal SpO₂ in a patient with carbon monoxide poisoning or methaemoglobinaemia may be falsely reassuring. Carboxyhaemoglobin levels above 2% may cause falsely elevated SpO₂ measurements.183 Many smokers will have carboxyhaemoglobin levels above this level shortly after smoking a cigarette, and the carboxyhaemoglobin level may be elevated to 15% in some smokers and up to 50% or more in acute carbon monoxide poisoning. It is not known if the reduced blood oxygen content in smokers who develop sudden illness within a few hours of smoking cigarettes has any effect on clinical outcomes, or if heavy smokers might benefit from a slightly higher target saturation range than non-smokers during the first few hours of a serious illness in an effort to maintain a similar blood oxygen content.

Skin pigmentation may also influence the accuracy of pulse oximetry readings (usually overestimation but sometimes underestimation). In particular, the accuracy of pulse oximetry is impaired in dark skinned participants at saturation levels below 80–85%.184–186 However, this should rarely be a problem in clinical practice if the saturation is maintained in the range suggested in the present guideline (94–98% for most patients), although the work of Jubran and Tobin31 on ventilated participants suggested that an oxygen saturation of 92% was useful in predicting a PaO₂ above 60 mm Hg (8 kPa) in ventilated white participants but was less reliable in ventilated black participants who sometimes had an SpO₂ reading that was more than 4% above the directly measured PaO₂. In the case of sickle cell crisis, pulse oximetry may underestimate the level of oxygenation.187 In these circumstances, under-reading is safer than over-reading because no truly hypoxaemic patient would be denied oxygen therapy. In sickle cell patients, one study has found that pulse oximeters did not misdiagnose either hypoxaemia or normoxaemia during a sickle cell crisis provided a good wave signal was present,188 but discrepancy between SpO₂ and that measured by co-oximetry (SO₂) has been found by others during a sickle cell crisis189 and stable patients.190 In these studies, the SpO₂ was generally lower than the SaO₂, so pulse oximetry may misdiagnose hypoxaemia in sickle cell patients when none is present.

Oximeters can be affected by motion of the patient's hand, but this is less of a problem with modern oximeters than with older devices.191 Motion artefact is more of a problem if the patient also has reduced perfusion of the measuring site.192 A malpositioned oximeter sensor can cause artefact which can overestimate or underestimate the true oxygen saturation; this can be a particular problem during repositioning of ill patients.193

The site of oximetry is also important. Finger and earlobe measurements are more accurate than measurements from a probe applied to the toe, and finger probes may be more accurate than ear probes.194 ,195 Finally, clinical staff need to remember to remove nail varnish and false nails to avoid artefacts in oximetry measurements.

Evidence statements

• Pulse oximeters are accurate to within 1–2% of directly measured SaO₂ in most participants but the error (usually overestimation but sometimes underestimation) is greater in dark skinned participants, especially with very low saturation (below 80–85%) (evidence level 2+).

• The accuracy of oximeters in shock, sepsis and hypotension is largely unknown, but most errors are likely to result in falsely low readings which would result in additional oxygen being given. Most errors in oximetry are therefore not likely to place patients at risk, but it is important to ensure that the oximeter has a good signal and to avoid artefact due to motion, nail varnish or other potential sources of error (evidence level 3).

• It is advised that oximetry measurements on sleeping patients should be recorded over several minutes to avoid the possibility of being misled by a normal transient nocturnal dip in oxygen saturation and clinicians should look for evidence of snoring or sleep apnoea if desaturation is noted during sleep (evidence level 4).

• Pulse oximetry can be misleadingly normal in smokers because of raised blood carboxyhaemoglobin levels which will cause a reduced blood oxygen content despite an apparently normal oxygen saturation and a normal PO₂. Patients who have smoked cigarettes in the previous 10 hours may therefore be at increased risk from hypoxaemia (evidence level 3).

• There are no controlled trials regarding implementation of these observational studies and the following recommendations are based on expert opinion (evidence level 4).

Recommendations

B2: Oxygen saturation, ‘the fifth vital sign’, should be checked by trained staff using pulse oximetry in all breathless and acutely ill patients (supplemented by blood gases when necessary) and the inspired oxygen device and flow rate should be recorded on the observation chart with the oximetry result (grade D).

B3: Initial clinical assessment and subsequent monitoring of acutely unwell patients should include the use of a recognised physiological ‘track and trigger’ system, such as the NEWS which may trigger clinical review due to hypoxaemia, need for supplementary oxygen or for other reasons (grade D). BTS

B4: For patients who are at risk of hypercapnic respiratory failure, it is recommended that the relevant section of the 2017 NEWS chart should be used. Points are awarded if the oxygen saturation is below or above the target range (grade D).

7.1.3 Arterial and arterialised blood gases

ABGs are the ‘gold standard test’ for assessing respiratory failure. However, published studies have shown that arterialised capillary gases from the earlobe (but not from the finger) can provide an assessment of pH and PCO₂ that is almost identical to that obtained from an arterial sample (indications for blood gas sampling are given in section 8.4 and recommendation 13).196–200 In acute and stable situations the earlobe specimen gives a PO₂ measurement which is 0.5–1 kPa (3.7–7.5 mm Hg) lower than the simultaneous arterial measurement with most of the divergence occurring at PO₂ above 8–10 kPa (60–75 mm Hg).197 ,200This means that most patients can be managed safely based on the pH and PCO₂ levels measured from earlobe blood gases supplemented by oxygen saturation measured by a pulse oximeter.197 ,200 In critically ill patients, the initial specimen should be an arterial specimen to guarantee an accurate initial assessment, but capillary gases are especially valuable for monitoring progress of the blood gases as a patient stabilises.

Patients who have had simultaneous arterial and earlobe samples rated the earlobe puncture procedure as being considerably less painful than arterial puncture.201However, the administration of local anaesthesia before ABG sampling produced a significant reduction in pain.202 ,203 A recent randomised trial found that arterial punctures using insulin needles cause less pain and fewer procedural complications compared with standard needles.204

The pain score measured on a 100 mm visual analogue scale mean (±SD) in punctures with the insulin needle was lower than the standard needle (23±22 vs 39±24 mm; mean difference=−15 mm; 95% CI −22 mm to −7 mm; p<0.001). However, due to the higher rate of haemolysis, the authors recommended that the use of insulin needles should be limited to conditions that do not require a concurrent potassium value in the same blood sample.

There is a very small risk of arterial damage from arterial puncture, especially if the radial site is used. Most reports of hand ischaemia have involved indwelling radial artery cannulae, but the vessel could also be injured by needle puncture.205 The guideline therefore recommends that arterialised earlobe specimens should be used more widely than at present as a safer and less painful alternative to ABG sampling and local anaesthetic should be used wherever possible for ABG sampling, but this is often not practical in medical emergencies and blood gas sampling should not be delayed in these circumstances. However, the accuracy of earlobe samples in shock or hypotension is not known and it is recommended that ABGs should be used in all cases of shock or hypotension (systolic blood pressure <90 mm Hg).

The technique of patient preparation, sample acquisition and sample processing for arterialised capillary gases is complex and should only be undertaken by fully trained staff. Capillary gases are very vulnerable to errors in technique and they should only be implemented for emergency use in units where staff have been fully trained in their use. See section 7.2.3 for advice concerning venous blood gases.

Evidence statements for use of arterialised capillary blood gas measurements

• Patients find earlobe specimens less painful than arterial puncture without local anaesthesia (evidence level 2++).

• ABG sampling causes pain which can be reduced by the use of local anaesthetic prior to the test (evidence level 1+).

• Arterialised earlobe blood gases will provide accurate information about PaCO₂ and pH but do not provide accurate information concerning PO₂ (evidence level 1−).

• The earlobe specimen gives a PO₂ measurement which is 0.5–1 kPa (4–7.5 mm Hg) lower than the simultaneous arterial measurement with greater divergence at oxygen levels above 8–10 kPa (60–75 mm Hg) (evidence level 1–).

• However, a combination of earlobe gases (to monitor pH and PCO₂) and oximetry (to measure oxygen levels) will allow safe management of most patients, even in emergency settings. (The only published evidence is for patients with COPD but this finding is likely to be generalisable to most patients other than those with shock or poor peripheral circulation) (evidence level 4).

• The technique of patient preparation, sample acquisition and sample processing for arterialised capillary gases is complex and should only be undertaken by fully trained staff (evidence level 4).

Recommendations for use of ABGs and arterialised capillary blood gases

C1: For critically ill patients or those with shock or hypotension (systolic blood pressure <90 mm Hg), the initial blood gas measurement should be obtained from an arterial sample. For most patients who require blood gas sampling, either ABGs or arterialised earlobe blood gases may be used to obtain an accurate measure of pH and PCO₂. However, the PO₂ is less accurate in earlobe blood gas samples (it underestimates the PO₂ by 0.5–1 kPa) so oximetry should be monitored carefully if earlobe blood gas specimens are used and a repeat arterial specimen should be taken if there is any concern about the accuracy of a capillary sample (grade D).

C2: Local anaesthesia should be used for all ABG specimens except in emergencies (grade A).

7.1.4 Transcutaneous oxygen assessments

Transcutaneous oxygen devices give different information from pulse oximetry. They are more sensitive to reduced perfusion and may be used to monitor tissue oxygenation in trauma patients but their use is beyond the scope of this guideline.206

7.2.1 Clinical assessment

In patients with lung disease, hypercapnia may be accompanied by visible respiratory distress, but this will be absent when hypercapnia is a consequence of a reduction in minute ventilation. Patients may have a flushed face, a full and bounding pulse and muscle twitching together with the characteristic flap of the outstretched hands. In severe cases, consciousness may be depressed and convulsions may occur.

Gross hypercapnia usually occurs with profound hypoxaemia and it is therefore difficult to disentangle the direct effect of hypercapnia per se. Coma will usually occur when the PaCO₂ is in the range 12–16 kPa (90–120 mm Hg). Survival has been seen following a PaCO₂ of 67 kPa (500 mm Hg).207

The presence of hypercapnic respiratory failure can be anticipated in patients with severe exacerbations of COPD or other diseases such as severe neuromuscular disorders. Carbon dioxide is a vasodilator so patients with hypercapnia may develop headache. Carbon dioxide in high concentrations has hypnotic effects and patients with hypercapnia may progress from drowsiness with flapping tremor to confusion to coma.84 ,130–134 A study of 127 episodes of acute respiratory acidosis showed that the best clinical predictors of respiratory acidosis were drowsiness (OR 7.1), flushing (OR 4.1), the presence of known COPD (OR 3.3) and the presence of intercostal retraction (OR 2.9).208

Clinical symptoms and signs of carbon dioxide retention include:

• Headache.

• Poor appetite.

• Vasodilation producing flushing and warm peripheries with dilated blood vessels (including retinal veins).

• Bounding pulse.

• Drowsiness.

• Flapping tremor.

• Confusion.

• Coma.

7.2.2 Blood arterial and arteriolar gases

Arterial or arterialised earlobe capillary blood gases will give an accurate estimation of pH and PaCO₂ (see section 7.1.3 for further details).196–198 The blood gases will need to be repeated in 30–60 min in patients with significant hypercapnia or acidosis to monitor the response to treatment. Patients with COPD who remain acidotic despite 30–60 min of standard treatment (including targeted low-concentration oxygen therapy) are likely to need NIV.209

7.2.3 Venous PCO₂ sampling

It has been suggested that the venous PCO₂ level can be used to screen for hypercapnia in patients with acute respiratory disease. A study of 196 paired samples of arterial and venous blood from patients with acute respiratory disease showed that the PCO₂ in the venous sample was an average of 0.77 kPa (5.8 mm Hg) higher than the simultaneous arterial sample.210 A venous PCO₂ below 6 kPa (45 mm Hg) had 100% sensitivity for eliminating the risk of hypercapnia (arterial PCO₂ above 6 kPa or 45 mm Hg), although the specificity was low at 57% and there was more variation in some studies.211–213 For patients who are not at risk of metabolic acidosis, the presence of a satisfactory oxygen saturation measured by pulse oximetry and a venous PCO₂ below 6 kPa (45 mm Hg) can exclude the possibility of significant arterial hypoxia or hypercapnia and may obviate the need for ABG measurements. This advice is strengthened by a recent prospective cohort study and a meta-analysis of five previous studies.214 ,215 Although the use of venous PCO₂ is becoming more common, it is still not widely enough used in clinical practice for the guideline committee to make a clear recommendation on its use.

7.2.4 Carbon dioxide monitors and non-invasive assessments of hypercapnia

Waveform capnography to assess end-tidal CO₂ is used primarily to confirm tracheal intubation during anaesthesia, intensive care and for any patients requiring endotracheal intubation. They are considered the ‘gold standard’ by the Royal College of Anaesthetists. The absence of any detectable carbon dioxide output indicates a failed intubation. The management of intubated patients is outside the remit of this guideline.

Waveform capnography monitors are also useful in the management of cardiac arrest and circulatory collapse. Waveform capnography during CPR allows: (1) confirmation of correct tracheal tube placement; (2) monitoring of the respiratory rate; (3) assessment of CPR quality—improved chest compression quality results in increased end-tidal CO2 level; (4) indication of return of spontaneous circulation (ROSC)—a large increase in end-tidal CO₂ during CPR can indicate ROSC; (5) a very low end-tidal CO₂ value during CPR despite high-quality chest compressions is associated with a low probability of ROSC.216–218 These devices are also useful in the care of intubated patients in the emergency department because, through visualising a typical ‘box wave form’, they can confirm that the tube is in the airway even in the absence of carbon dioxide production during a cardiac arrest. A sudden increase in carbon dioxide may be the first sign of spontaneous circulation.219

End-tidal carbon dioxide measurements correlate poorly with arterial carbon dioxide levels in patients with COPD, but they may be useful in some research studies of hyperventilation syndromes. However, these devices are inaccurate in patients with airways disease and those with a high respiratory rate, so they should not be used in the management of patients with respiratory failure and they will not be discussed further in this guideline.

The recent availability of portable devices to measure transcutaneous partial pressure of carbon dioxide (PtCO₂) at the bedside provide a promising alternative to an ABG. They function on the principle that CO₂ diffuses extremely well through tissues. A probe is attached to an area of skin (usually the earlobe) and warms to 42°C which ‘arterialises’ the underlying capillaries. The CO₂ diffusing through the skin changes the pH of a thin electrolyte membrane in the probe and the resulting signal is converted to an estimate of the PaCO₂. As well as measuring PtCO₂, the probe also measure SpO₂. Early studies indicate that such devices can be accurate in normal volunteers. Transcutaneous CO₂ devices have provided accurate estimates of PaCO₂ in a variety of clinical scenarios including AECOPD, invasive and NIV in ICUs and overnight studies of sleep-disordered breathing.220–224

Recently transcutaneous CO₂ monitoring has been used in the emergency department in patients presenting with acute severe asthma136 and community-acquired pneumonia225 The PtCO₂ accurately assessed PaCO₂ as measured by the ‘gold standard’ ABG.226 A rise in PtCO₂ was demonstrated in both groups of patients on administration of high-flow oxygen for 60 min, the significance of this will be discussed further in section 8.12. These new techniques are not yet recommended for routine clinical use.

8.7.1 Oxygen saturation target range for most patients

As discussed in sections 4–6 of this guideline, there is no evidence of benefit from above normal oxygen saturation in most clinical situations and there is evidence that excessive concentrations of oxygen can have adverse effects, even in some patients who are not at risk of hypercapnic respiratory failure. A target oxygen saturation range of 94–98% will achieve normal or near-normal oxygen saturation for most patients who are not at risk of hypercapnic respiratory failure. Furthermore, the suggested lower limit of 94% allows a wide margin of error in the oximeter measurement, thus minimising the risk of any patient being allowed to desaturate below 90% due to inaccurate oximetry.

8.7.2 Oxygen requirements for specific groups of patients

• Patients with critical illness requiring high-concentration oxygen therapy are discussed in section 8.10.

• Patients with medical emergencies which frequently cause breathlessness and hypoxaemia are discussed in section 8.11.

• Patients with COPD and other conditions that may predispose to hypercapnic respiratory failure are discussed in section 8.12.

• Medical emergencies for which oxygen was commonly given in the past but is not actually indicated unless the patient is hypoxaemic are discussed in section 8.13.

8.9.1 Devices used in emergency oxygen therapy in hospitals (see section 10 for further details)

• High-concentration oxygen from reservoir mask (15 L/min) or when using a bag mask or advanced airways (supraglottic airway or tracheal tube) for assisted ventilation during respiratory arrest or during CPR.

• Nasal cannulae (2–6 L/min) or simple face masks (5–10 L/min) for medium-concentration oxygen therapy (and HFNC using specialised equipment for high-concentration oxygen therapy, often as an alternative to a reservoir mask).

• Twenty-four per cent Venturi mask at 2 L/min or 28% Venturi masks at 4 L/min for patients at risk of hypercapnic respiratory failure (alternatively use nasal cannulae at 1–2 L/min, especially when the patient has stabilised).

• Tracheostomy masks for patients with a functioning tracheostomy (adjust flow to achieve desired saturation).

8.10.1 Cardiac arrest and other conditions requiring CPR

The 2015 guideline for Adult Advanced Life Support issued by Resuscitation Council UK recommends the use of the highest possible inspired oxygen level for ventilation during CPR; that guideline also makes recommendations regarding choice of airway and ventilation technique in these circumstances.231 The present guideline endorses these proposals during the period of resuscitation. Subsequent management will depend on the underlying condition and the patient's degree of recovery. There is theoretical evidence that patients with a ROSC may be managed more safely with a lower initial inspired oxygen than with 100% oxygen.98 ,99 Recently, in an observational study, Kilgannon et al151 ,152 reported increased mortality among survivors of cardiac arrest who had been exposed to very high levels of oxygen (PaO₂>300 mm Hg or 40 kPa) in the first 24 hours on the ICU. Bellomo et al154 also found an excess of deaths in this group of patients but argued that the risk was no longer statistically significant after multivariate analysis. Elmer et al232 reported a small single-centre study in 2015 in which severe hyperoxaemia (PaO₂>300 mm Hg or 40 kPa) was associated with increased hospital mortality but moderate hyperoxaemia (101–299 mm Hg) was not associated with decreased survival.

A systematic review and meta-analysis by Wang et al233 concluded that hyperoxaemia appears to be correlated with increased in-hospital mortality in survivors of adult cardiac arrest but the result should be interpreted cautiously because of the significant heterogeneity and the limited number of studies that were analysed.

Based on the above evidence, current UK and European Resuscitation guidelines recommend a target oxygen saturation of 94–98% once spontaneous circulation had been restored.231 This guideline endorses these recommendations.

Evidence statement

• The use of oxygen during and after CPR is based on expert opinion and extrapolation from observational data (evidence level 4).

Recommendation

E1: Use the highest feasible inspired oxygen for ventilation during CPR (see table 1 and section 8.10). Once spontaneous circulation has returned and arterial blood oxygen saturation can be monitored reliably, aim for a target saturation range of 94–98% and take an ABG sample to guide ongoing oxygen therapy. If the blood gas shows hypercapnic respiratory failure, reset the target range to 88–92% or consider mechanical ventilation (grade D).

8.10.2 Critically ill patients including major trauma, shock and major sepsis

There is evidence that early intervention to normalise oxygen delivery to the tissues using volume expansion and vasoactive agents is beneficial in the management of critically ill patients with shock or sepsis, but there is no evidence of benefit from attempts to achieve supranormal oxygen delivery.73–75 ,234–238 In fact, there is evidence that hyperoxia can cause a paradoxical decrease in whole body oxygen consumption in critically ill patients,239 and it has been demonstrated that hyperoxia can impair oxygen delivery in septic patients.240 A study using a rat model of progressive haemorrhage found that both hypoxaemia and hyperoxaemia compromised haemodynamics. There was a paradoxical fall in tissue PO₂ in hyperoxaemic animals despite the high PaO₂.241 These factors may account for the observation by deJonge (but disputed by Eastwood) that mortality was elevated among intensive care patients who were hyperoxaemic during the first day on the ICU.153 ,155 In de Jonge's study of 36 307 patients on ICUs in the Netherlands, the adjusted OR for death was lowest in a PO₂ range of 8.9–10.6 kPa (equivalent to saturation about 93–96% at normal pH). Two recent systematic reviews with meta-analysis have addressed this subject. Both studies concluded that hyperoxaemia was associated with increased hospital mortality but caution is required in interpreting the results because of heterogeneity in the included studies, including different definitions of hyperoxaemia.242 ,243 A randomised trial published in 2016 reported ICU mortality of 11.6% in 216 patients randomised to a target saturation range of 94–98% (median PaO₂ 11.6 kPa) compared with 20.2% mortality among 218 patients randomised to conventional therapy (median PaO₂ 13.6 kPa).244 The absolute risk reduction was 8.6% (risk reduction 0.086; 95% CI 0.017 to 0.150; p=0.01) and hospital mortality was 24.2% vs 33.9% (absolute risk reduction 0.099; 95% CI 0.013 to 0.182; p=0.03). However, the trial was prematurely terminated due to difficulty with recruitment, and the authors advise caution in interpreting the results for this reason.

Most critically ill patients are at risk of multiorgan failure and therefore require intensive care assessment as a matter of urgency. Critical care consensus guidelines set 90% saturation as the minimum level below which oxygen saturation should not be allowed to fall.31 ,32 ,245 The 2012 Surviving Sepsis Campaign guideline is mainly based on mixed venous oxygen saturation or central venous saturation but also recommends a target SaO₂ of 88–95% for patients with sepsis.31 ,32 ,245 However, these recommendations are based on directly measured SaO₂ in critical care settings with intensive levels of nursing and monitoring. The present guideline recommends a slightly higher target saturation range prior to the transfer of these seriously ill patients to critical care facilities.

For most critically ill or severely hypoxaemic patients, initial oxygen therapy should involve the use of a reservoir mask, aiming at an oxygen saturation of 94–98%. If the patient has concomitant COPD or other risk factors for hypercapnic respiratory failure, the initial saturation target should also be 94–98% pending the results of blood gas estimations and assessment by intensive care specialists. If critically ill patients with COPD have hypercapnia and acidosis, the correction of hypoxaemia must be balanced against the risks of respiratory acidosis and ventilatory support using NIV or invasive ventilation should be considered.

It is also recognised that many patients with long bone fractures may develop hypoxaemia even in the absence of injury to the airway or chest (possibly due to opioid treatment and fat embolism) and they should be monitored with oximetry and given oxygen if necessary.38 ,246–248 These patients, if not critically ill, should have a target oxygen saturation of 94–98% or 88–92% if they have coexisting COPD or other risk factors for hypercapnic respiratory failure.

Evidence statement

• The use of oxygen in critical illness is based on expert opinion guided by physiology and observational data (evidence level 4).

Recommendation

E2: In critical illness, including major trauma, sepsis, shock and anaphylaxis, initiate treatment with a reservoir mask at 15 L/min and aim at a saturation range of 94–98%. This advice also applies to patients with critical illness who have risk factors for hypercapnia pending the results of blood gas measurements and expert assessment. In patients with spontaneous circulation and a reliable oximetry reading, it may be possible to maintain a saturation of 94–98% using lower concentrations of oxygen (grade D).

8.10.3 Drowning

Survivors of drowning may have suffered inhalation of fresh or sea water into the lungs and may become hypoxaemic. Supplemental oxygen should be given to all patients with saturation below 94%, aiming at a target saturation of 94–98%.249

Evidence statement

• The use of oxygen in drowning is based on expert opinion (evidence level 4).

Recommendation

E3: In cases of drowning, aim at an oxygen saturation of 94–98% once adequate circulation is restored (grade D).

8.10.4 Anaphylaxis (see recommendation E2)

Patients with anaphylaxis are likely to suffer from tissue hypoxia due to a combination of upper and/or lower airway obstruction together with hypotension. In addition to specific treatment of these problems, the Resuscitation Council UK recommends initial high-concentration oxygen (10–15 L/min, by reservoir mask) for patients with anaphylaxis.231 ,250 The present guideline would endorse this practice in the immediate management of anaphylaxis followed by a target saturation of 94–98% once the patient's oxygen saturation can be measured reliably.

Evidence statement

• The use of oxygen in anaphylaxis is based on expert opinion (evidence level 4).

8.10.5 Major pulmonary haemorrhage or massive haemoptysis

Major pulmonary haemorrhage and massive haemoptysis can occur for a large number of reasons ranging from acute pulmonary vasculitis to erosion of a blood vessel by a lung tumour. In addition to specific treatment of the causative condition, most such patients require supplemental oxygen treatment (also see table 1). A target saturation range of 94–98% is recommended. Treatment should be initiated with high-concentration oxygen via a reservoir mask and subsequently adjusted according to chart 2 to maintain a saturation of 94–98% pending the results of blood gas measurements.

Evidence statement

• The use of oxygen in pulmonary haemorrhage is based on expert opinion (evidence level 4).

8.10.6 Epileptic fits

As this is a life-threatening condition where a patient may suffer from cerebral hypoxia (and oximetry may not be possible), patients with status epilepticus should be treated in accordance with table 1 (reservoir mask until clinically stabilised).

Evidence statement

• There have been no clinical trials of oxygen use during epileptic fits and it may be impossible to obtain an oximetry signal in these circumstances. Expert opinion advises the administration of high-concentration oxygen until oximetry measurements can be obtained (evidence level 4).

Recommendation

E4: In patients with acute seizures due to epilepsy or other causes, high-concentration oxygen should be administered until a satisfactory oximetry measurement can be obtained and clinicians should then aim for an oxygen saturation of 94–98% or 88–92% if the patient is at risk of hypercapnic respiratory failure (grade D).

8.10.7 Major head injury

Patients with major head injury are at risk of hypoxaemia and hypercapnia. They require urgent assessment and maintenance of airway patency, either through positioning, simple adjuncts or early tracheal intubation and ventilation to avoid further brain injury due to brain oedema which may be aggravated by hypoxaemia and/or hypercapnia. These patients should be triaged directly where possible to a major trauma centre. Initial treatment should include high-concentration oxygen via a reservoir mask pending availability of satisfactory blood gas measurements or until the airway is secured by intubation which will also protect against the risk of aspiration due to reduced consciousness. Although hypoxaemia is common in patients with head injury (and associated with poorer outcomes), the relative contribution of hypoxaemia to outcome is not yet established.37 ,251–255 All authors agree that hypoxaemia should be corrected, but a review of the literature concluded that there is no evidence of clinical benefit from hyperoxia in brain injured patients and a subsequent clinical study showed that normobaric hyperoxia did not improve brain metabolism in five patients with acute severe brain injury.256 ,257 There are no UK guidelines for oxygen therapy in the immediate phase after head injury, but US guidelines recommend maintaining an oxygen saturation above 90% for patients with acute brain injury.30 An American study published in 2009 describes a cohort of 3420 patients and the authors reported worse outcomes in patients with severe traumatic brain injury associated with hypoxaemia and extreme hyperoxaemia. Outcomes were optimal in the PaO₂ range of 110–487 mm Hg (14.6–65 kPa).258 Other observational studies have suggested that hyperoxaemia may be associated with improved outcomes. Asher et al259 recently reported the outcomes for 193 adult patients with severe traumatic brain injury who had overall survival of 57%. They reported that a PaO₂ threshold between 250 and 486 mm Hg (33–65 kPa) during the first 72 hours after injury was associated with improved all-cause survival, independent of hypocarbia or hypercarbia. Two recent observational studies have provided further evidence that hyperoxaemia may be associated with increased mortality in traumatic brain injury.260 ,261 The present guideline advises giving supplementary oxygen if required to maintain an oxygen saturation in the range of 94–98% pending the publication of randomised studies comparing normoxaemia with normobaric hyperoxaemia.

Evidence statement

• The use of oxygen in cases of major head injury is based on expert opinion and extrapolation from observational studies (evidence level 4).

Recommendation

E5: In cases of major head injury, aim at an oxygen saturation of 94–98%. Initial treatment should involve high-concentration oxygen from a reservoir mask at 15 L/min pending availability of satisfactory blood gas measurements or until the airway is secured by intubation (grade D).

8.10.8 Carbon monoxide poisoning

Patients with carbon monoxide poisoning may have a normal level of PaO₂ but a greatly reduced level of oxygen bound to haemoglobin because this has been displaced by carbon monoxide.262 Standard pulse oximetry cannot screen for carbon monoxide exposure as it does not differentiate carboxyhaemoglobin from oxyhaemoglobin and blood gas measurements will show a normal PaO₂ in these patients. Some newer co-oximeters can measure carboxyhaemoglobin levels as well as SpO₂.The blood carboxyhaemoglobin level must be measured to assess the degree of carbon monoxide poisoning. The half-life of carboxyhaemoglobin in a patient breathing air is ∼300 min; this decreases to 90 min with high-concentration oxygen via a reservoir mask. The most important treatment for a patient with carbon monoxide poisoning is therefore to give high-concentration oxygen via a reservoir mask. Comatose patients or those with severe mental impairment should be intubated and ventilated with 100% oxygen. The role of hyperbaric oxygen remains controversial. A 2011 Cochrane review concluded that existing randomised trials did not establish whether the administration of hyperbaric oxygen to patients with carbon monoxide poisoning reduced the incidence of adverse neurological outcomes.263 However, a randomised trial published in 2007 has suggested that patients with loss of consciousness or high carboxyhaemoglobin levels may have less cognitive sequelae if given hyperbaric oxygen.264

Evidence statement

• The use of oxygen in carbon monoxide poisoning is based on expert opinion and extrapolation from observational studies (evidence level 4).

Recommendation

E6: In cases of carbon monoxide poisoning, an apparently ‘normal’ oximetry reading may be produced by carboxyhaemoglobin, so aim at an oxygen saturation of 100% and use a reservoir mask at 15 L/min irrespective of the oximeter reading and PaO₂ (grade D).

8.11.1 Patients with acute onset of hypoxaemia of unknown cause with no pre-existing respiratory disorders or risk factors

It is common for breathless and hypoxaemic patients to have no firm diagnosis at the time of presentation. For most acutely hypoxaemic patients whose medical problem is not yet diagnosed, an oxygen saturation range of 94–98% will avoid the potential hazards associated with hypoxaemia or hyperoxaemia (see sections 4–6 and tables 1–2). Aiming for an oxygen saturation in the normal range (rather than an abnormally high oxygen level) will also have the effect of allowing the lowest effective FiO₂ to be used, thus avoiding risks such as absorption atelectasis and V/Q mismatch that may be associated with the use of very high fractions of inspired oxygen (see sections 5 and 6). The priority for such patients is to make a specific diagnosis as early as possible and to institute specific treatment for the underlying condition. Early blood gas measurement is mandatory in the management of patients with sudden unexplained hypoxaemia.

Evidence statement

• The use of oxygen in most instances of acute hypoxaemia is guided by expert opinion based on physiology and observational studies except where indicated otherwise in the sections that follow (evidence level 4).

Recommendations

D1: For acutely breathless patients not at risk of hypercapnic respiratory failure who have saturations below 85%, treatment should be started with a reservoir mask at 15 L/min in the first instance (see figures 1–2 (charts 1–2) and table 2 and sections 8.10 and 10).* The oxygen concentration can be adjusted downwards (using nasal cannulae at 1–6 L/min or a simple face mask at 5–10 L/min) to maintain a target saturation of 94–98% once the patient has stabilised (grade D).

D2: In other cases of acute hypoxaemia without critical illness or risk factors for hypercapnic respiratory failure, treatment should be started with nasal cannulae (or a simple face mask if cannulae are not tolerated or not effective) with the flow rate adjusted to achieve a saturation of 94–98% (grade D).

D3: If medium-concentration therapy with nasal cannulae or a simple face mask does not achieve the desired saturation, change to a reservoir mask and seek senior or specialist advice (grade D).

Good practice point

• High-flow nasal oxygen using specialised equipment should be considered as an alternative to reservoir mask treatment in patients with acute respiratory failure without hypercapnia.

*For initial management of patients at risk of hypercapnic respiratory failure, see recommendations G1 and G2.

8.11.2 Acute asthma

The BTS/SIGN guideline for the management of acute asthma recommends that the oxygen saturation should be maintained between 94% and 98%.229 Although there is no danger of tissue hypoxia at any saturation above 90%, a drop of oxygen saturation below 94% may indicate deterioration and should prompt a further assessment. Supplemental oxygen should be started using nasal cannulae at 2–4 L/min or a simple face mask at 5 L/min or 35–40% Venturi mask and adjusted as necessary to maintain a saturation of 94–98%.265

A study which was published in 2003 showed that the administration of 100% oxygen to patients with acute severe asthma produced an increased PaCO₂ and a decreased peak expiratory flow compared with patients treated with 28% oxygen.266 The authors of that study recommended the use of targeted oxygen therapy rather than giving high-concentration oxygen to all patients with acute severe asthma. This advice has been reinforced by the randomised study of Perrin et al136 who found a significant increase in transcutaneous PCO₂ when oxygen was given at 8 L per minute via simple face mask. Of 106 patients randomised to high-concentration oxygen or targeted oxygen therapy (target 93–95%), all 10 instances of hypercapnia occurred in patients given high-concentration oxygen therapy. Hypercapnia in acute asthma indicates a near-fatal attack and indicates the need for consideration of intensive care admission and ventilation.229 The trial of Perrin et al suggests that many cases of hypercapnic respiratory failure in acute asthma may be due to excessive oxygen therapy. It remains appropriate to give oxygen to patients with acute severe asthma in the absence of oximetry or blood gas results, but there is no evidence of benefit from giving oxygen to patients who are not hypoxaemic and the above study suggests that unnecessary oxygen administration may cause harm. Oxygen should not be withheld from hypoxaemic patients with severe asthma because of concerns about possible hypercapnia, but oxygen therapy should be confined to what is necessary to achieve a saturation range of 94–98%.266 ,267

Evidence statements

▸ Patients with acute asthma should not be given high-concentration oxygen therapy in the absence of severe hypoxaemia (evidence level 4).

▸ The suggested target saturation range of 94–98% in acute asthma is based on expert opinion in the absence of randomised trials (evidence level 4).

Recommendation

F1: In acute asthma, aim at an oxygen saturation of 94–98% (see tables 2 and 3 and sections 8.11 and 8.13) (grade D).

8.11.3 Pneumonia

The BTS guideline for pneumonia recommends aiming at an oxygen saturation target range of 94–98% in uncomplicated pneumonia with appropriate adjustments for patients with COPD, guided by blood gas measurements.268 The present guideline endorses these principles. It has recently been demonstrated that high-concentration oxygen therapy can increase transcutaneous CO₂ level in patients presenting with suspected community-acquired pneumonia.137

Evidence statement

• The use of oxygen in pneumonia is based on expert opinion (evidence level 4).

Recommendation

F2: In cases of pneumonia who are not at risk of hypercapnic respiratory failure, aim at an oxygen saturation of 94–98% (grade D).

8.11.4 Lung cancer and other cancers with pulmonary involvement

Most patients with cancer who present with acute breathlessness have a specific causative factor such as a pleural effusion, pneumonia, COPD, anaemia or collapse of a lobe or collapse of a whole lung.269 ,270 Relief of breathlessness may be achieved by treating these causative factors or by the use of opioid medicines but there is increasing evidence that supplemental oxygen does not relieve breathlessness for non-hypoxaemic patients with cancer, as discussed in section 8.17. However, it is likely that patients with cancer with significant hypoxaemia may have some relief from breathlessness if given oxygen.

Evidence statement

• The use of oxygen in the acute management of lung cancer is based on expert opinion and extrapolation from observational studies (see separate advice on palliative care in section 8.17; evidence level 4).

Recommendation

F3: In acute breathlessness due to lung cancer, aim at an oxygen saturation of 94–98% unless there is coexisting COPD. See also ‘Oxygen use in palliative care’ section 8.17 (grade D).

8.11.5 Deterioration of fibrotic lung conditions and other conditions involving parenchymal lung disease or alveolitis

It is recognised that patients with fibrosing lung conditions such as idiopathic pulmonary fibrosis may have acute deteriorations or exacerbations, often during intercurrent chest infections. Other patients may present acutely with breathlessness due to extrinsic allergic alveolitis, sarcoidosis or other types of parenchymal lung disorders. These patients often have a high degree of V/Q mismatch and a requirement for high oxygen concentrations to achieve satisfactory blood gases and they are not at risk of hypercapnia. The oxygen level should be adjusted to maintain an oxygen saturation in the range of 94–98%, but this level may not be achievable or only achievable with a reservoir mask. Patients with severe idiopathic pulmonary fibrosis are rarely suitable for invasive ventilation or NIV because of the progressive nature of the condition.

Evidence statement

• The use of oxygen in deteriorating pulmonary fibrosis is based on expert opinion (evidence level 4).

  Recommendation

F4: In acute deterioration of pulmonary fibrosis or other interstitial lung diseases, aim at an oxygen saturation of 94–98% or the highest possible if these targets cannot be achieved (grade D).

8.11.6 Pneumothorax

As with pleural effusions, patients with a large pneumothorax may be breathless and hypoxaemic and may require supplemental oxygen for symptom relief pending definitive treatment by aspiration or drainage. However, high-concentration inhaled oxygen can also increase the rate of reabsorption of air from a pneumothorax up to fourfold.271 A 2012 study using rabbits with traumatic pneumothorax confirmed that resolution time was significantly faster with high FiO₂ but there was evidence of lung toxicity above 60% oxygen.272 The BTS guideline on the management of pneumothorax recommends the use of high-concentration oxygen (reservoir mask) in all patients without COPD who require hospital admission for observation due to a moderate-sized pneumothorax that does not require drainage.273 Once a pneumothorax is drained or aspirated successfully, the patient should not require oxygen therapy unless there is additional pathology such as pneumonia, asthma or COPD requiring specific treatment.

Evidence statement

• The use of oxygen in pneumothorax is based on expert opinion and extrapolation from observational studies (evidence level 4).

Recommendations

F5: In most cases of pneumothorax, aim at an oxygen saturation of 94–98% if the patient is not at risk of hypercapnic respiratory failure (grade D).

F6: In patients with pneumothorax having hospital observation without drainage, the use of high-concentration oxygen (15 L/min flow rate via reservoir mask) is recommended unless the patient is at risk of hypercapnic respiratory failure (grade D).

8.11.7 Pleural effusion

If a pleural effusion is causing significant breathlessness, the most effective treatment is to drain the effusion. Hypoxaemic patients with pleural effusions are likely to benefit from supplementary oxygen therapy. The BTS guidelines for management of pleural effusions do not give any specific advice concerning oxygen therapy, but it seems reasonable to give supplementary oxygen to hypoxaemic patients to maintain a saturation of 94–98%.

Evidence statement

• The use of oxygen in pleural effusion is based on expert opinion (evidence level 4).

Recommendation

F7: In pleural effusion, aim at an oxygen saturation of 94–98% (or 88–92% if the patient is at risk of hypercapnic respiratory failure) (grade D).

8.11.8 Pulmonary embolism

Most patients with suspected pulmonary embolism have normal oxygen saturation and the main focus of treatment is to reach a specific diagnosis and to start anticoagulant treatment. These patients do not require oxygen therapy unless there is hypoxaemia. In these cases, the lowest concentration of oxygen that will achieve a target saturation of 94–98% is recommended. However, patients with massive or multiple pulmonary embolism may be profoundly hypoxaemic and should initially be given high-concentration oxygen via a reservoir mask to achieve an oxygen saturation of 94–98% pending definitive treatment such as thrombolysis. It has been suggested that the blood oxygen saturation may underestimate the severity of pulmonary artery obstruction in acute pulmonary embolism if shock is present.274

Evidence statement

• The use of oxygen in pulmonary embolism is based on expert opinion (evidence level 4).

Recommendation

F8: In pulmonary embolism, aim at an oxygen saturation of 94–98% (or 88–92% if the patient is at risk of hypercapnic respiratory failure) (grade D).

8.11.9 Acute heart failure

Most patients with acute heart failure are breathless, usually due to pulmonary oedema or low cardiac output, especially if cardiogenic shock is present. The pathophysiology of oxygen transport in cardiogenic shock has been discussed in detail by Creamer et al.275 It has been shown in an animal model that the ventilatory failure of cardiogenic shock may be due to an impairment of the contractile process of the respiratory muscles.276 Miñana et al277 did not find increased mortality in Spanish patients with acute heart failure who had hypoxaemia defined as PaO₂<60 mm Hg (8 kPa) which would correspond to saturation below about 91% at normal pH. In addition to specific treatment for heart failure, patients should be given supplemental oxygen to maintain a saturation of 94–98%. This is consistent with the European Society of Cardiology Task Force and European Society of Intensive Care recommendation that patients with acute heart failure should receive oxygen to maintain SpO₂ of >90%.278 Patients with marked hypoxaemia (saturation <85%) should be treated with a reservoir mask initially and patients with coexisting COPD will require a lower target saturation of 88–92% pending the availability of blood gas results.

Evidence statements

• The use of oxygen in acute heart failure is based on expert and extrapolation from observational studies (evidence level 4).

• In hospital settings, patients with acute pulmonary oedema may benefit from CPAP and from non-invasive ventilatory support (evidence level 1+, see evidence statement in section 8.19.2).

Recommendations

F9: In acute heart failure, aim at an oxygen saturation of 94–98% (or 88–92% if the patient is at risk of hypercapnic respiratory failure) (grade D).

F10: CPAP with entrained oxygen or high-flow humidified nasal oxygen to maintain saturation 94–98% (or 88–92% if at risk of hypercapnia) should be considered as an adjunctive treatment to improve gas exchange in patients with cardiogenic pulmonary oedema who are not responding to standard treatment (or NIV if there is coexistent hypercapnia and acidosis) (grade B).

8.11.10 Breathlessness due to severe anaemia

• If breathlessness is due to severe anaemia, the specific treatment is blood transfusion. Studies by Canadian researchers in the late 1990s have shown that haemoglobin levels of 70 g/L (7 g/dL) were as safe as transfusion to higher levels and may produce fewer complications in the critically ill.72 However, this study was conducted using non-leucocyte-depleted blood and it is possible that some of the complications in the group who were given more transfusions might have been avoided by the use of leucocyte-depleted blood. The optimal transfusion target for critically ill patients therefore remains the subject of ongoing discussion among experts in critical care medicine and readers are referred to national guidelines for blood transfusion.279 Giving oxygen to increase an already normal oxygen saturation will have very little effect on the oxygen-carrying power of the blood, but it is reasonable to administer supplemental oxygen to maintain a saturation of 94–98% if the saturation is below these levels breathing air or if breathlessness is a very prominent symptom.

Evidence statement

• The use of oxygen in anaemia is based on expert opinion based on blood gas physiology (evidence level 4).

Recommendations

F11: In anaemia, aim at an oxygen saturation of 94–98% or 88–92% if the patient is at risk of hypercapnic respiratory failure (grade D).

Good practice point

✓ Correction of anaemia by blood transfusion should be based on national guidelines.

8.11.11 Sickle cell crisis

Patients with sickle cell disease frequently present with an acute painful crisis and less frequently with an ‘acute chest syndrome’ comprising breathlessness, chest pain and fever with pulmonary infiltrates on the chest radiograph. The exact causes and mechanisms are not well understood, but oxygen should be given to all hypoxaemic patients with sickle cell crisis to avoid further intravascular sickling. There are no randomised studies of oxygen therapy in acute chest syndrome and no randomised studies of acute painful crisis in adults, but two small randomised trials showed no clinical benefit in non-hypoxaemic children with acute painful crisis.280 ,281 Patients with sickle cell disease may have a reduced oxygen saturation even when clinically stable. Homi et al282 reported a mean saturation of only 92.5% (95% CI 92.0% to 93.0%) in a group of children and young adults (age 9–18 years) with stable sickle cell disease compared with an average saturation of 97.1% (95% CI 98.8% to 97.3%) in a local control group. The British Committee for Standards in Haematology recommended in 2003 that oxygen should be given if the oxygen saturation falls below what is normal for the individual patient or a default target of 95% if the usual saturation is unknown.283 The 2012 NICE guideline reinforced this advice.284 This is consistent with the advice in the present guideline to aim at a normal or near-normal oxygen saturation for non-hypoxaemic patients with a target saturation of 94–98%. Readers are referred to the guideline on sickle cell disease for disease-specific management of this condition.283

Evidence statement

• During an acute sickle cell crisis, low SaO₂ as assessed by pulse oximetry is common. Research suggests that this usually indicates a low PaO₂ although during an acute painful crisis pulse oximetry may underestimate the true SaO₂. Desaturations during an acute sickle cell crisis should be treated with oxygen to reduce further sickling and progression of the crisis with further vaso-occlusive events (evidence level 1−).

Recommendation

F12: In sickle cell crisis and acute chest syndrome, aim for an oxygen saturation of 94–98% or aim at the saturation level that is usual for the individual patient (grade D).

Good practice point regarding sickle cell crisis

✓ Arterial or arterialised capillary blood gases should be sampled if there is any doubt about the reliability of oximetry during a sickle cell crisis.

8.12.1 COPD exacerbations

There is an extensive literature documenting the effects of high-concentration oxygen therapy in acute COPD.43–45 ,48 ,85 ,87 ,103 ,114 ,117 ,130 ,285–298 These reports show that the administration of supplemental oxygen to patients with exacerbated COPD (or with severe but stable COPD) often causes a rise in PaCO₂ with subsequent respiratory acidosis for reasons summarised in section 5.3. The literature up to 2001 is summarised in detail in the review by Murphy et al.84 Some patients with COPD are prone to repeated episodes of hypercapnic respiratory failure in response to oxygen therapy and others may not ever suffer from this complication. Apart from patients with recurrent hypercapnic respiratory failure, it is not possible to predict if individual patients with COPD will develop hypercapnia during an acute exacerbation, so all patients with moderate or severe COPD should be considered to be at risk of this complication until the results of blood gas measurements are available. It is therefore important that patients who are at risk of having COPD should be diagnosed accurately, and this can only be done by measurement of spirometry.1 The advice given for patients with COPD also applies to other forms of fixed airflow obstruction such as patients with chronic irreversible asthma or fixed airflow obstruction associated with bronchiectasis.

Some patients with acute severe exacerbations of COPD may be too breathless to undertake spirometry on arrival in hospital, but NICE guidance is that all patients should have the test performed before discharge from hospital to confirm the diagnosis of COPD and to assess the severity of the condition.1 There is very little literature describing the effects of oxygen therapy in the other conditions listed above, but patients with these conditions are recognised to be at risk of hypercapnic respiratory failure and should be treated in a manner analogous to patients with COPD.

It has been shown that patients with COPD with a pH reading <7.35 ([H⁺]>45 nmol/L) despite controlled oxygen therapy are more likely to die and more likely to meet criteria for intubation and ventilation.43 ,285 ,289 One of these reports also showed that patients with a high PaO₂ on arrival in hospital (>10.0 kPa or 75 mm Hg) were more likely to meet criteria for ventilation and the severity of acidosis was related to high PaO₂ values.43 Based on these results, Plant et al43 recommended an upper limit of about 92% saturation for patients with exacerbations of COPD to prevent the PaO₂ rising above 10 kPa. This report was supported by the work of Joosten et al45 which showed that a PaO₂ of >74.5 mm Hg (10 kPa) in acute COPD was associated with an increased likelihood of admission to a HDU, increased need for NIV and a longer stay in hospital.

Roberts et al46 reported outcomes among 5052 patients with exacerbated COPD where oxygen therapy in the ambulance was documented. They found that the mortality was 7.2% in patients who were managed with <35% oxygen but mortality was 11.1% in those given ≥35% oxygen and the need for NIV increased from 9% to 22% when high-flow oxygen was given.

Based on the above literature, this guideline recommends a maximum saturation of 92% while awaiting blood gas results in AECOPD and other conditions that may predispose to hypercapnic respiratory failure. Although the rise in PaCO₂ (and fall in pH) is greatest in patients who are given sufficient oxygen therapy to elevate the PaO₂ above 10 kPa, it is important to note that hypercapnia can occur in acute COPD even if the oxygen saturation is <88%.296 The best management strategy for patients with persistently acidotic COPD is a trial of NIV with supplemental oxygen therapy, if required to maintain oxygen saturations 88–92%.43 ,209 ,299

Some patients with previous hypercapnic respiratory failure will have alert cards or an entry in their electronic record to alert the emergency team to the optimal concentration of oxygen required during the patient's previous hospital admissions (see section 9.7). In the absence of such information, it is suggested that a target of 88–92% should be set initially for patients with a history of previous NIV or invasive ventilation and, if necessary, modified later based on blood gas results. These patients should be categorised as very urgent by ambulance teams and emergency services, requiring immediate blood gas measurement and senior assessment on arrival at the hospital emergency department. Against this background, the study of Austin et al51 has strengthened the advice given in this guideline. These authors randomised 405 patients with presumed exacerbation of COPD to high-concentration oxygen (8–10 L/min via reservoir mask) or to titrated oxygen treatment with target range 88–92% using nasal cannulae (with nebulised bronchodilator therapy driven by compressed air). Mortality was 9% among those given high-concentration oxygen and 4% among those randomised to targeted oxygen therapy. For patients with spirometrically confirmed COPD, mortality was 9% vs 2%, respectively. Further support for targeted oxygen therapy was supplied in the 2012 study of Cameron et al.135 They undertook retrospective review of the records of 254 COPD admissions to one hospital and they found that both hypoxaemia (PaO₂<60 mm Hg or 6 kPa) and hyperoxaemia (PaO₂>100 mm Hg or 13.3 kPa) were associated with increased risk of serious adverse outcome, defined as a composite of hypercapnic respiratory failure, assisted ventilation or inpatient death. Twenty-four per cent of patients were hyperoxaemic and these patients had OR of 9.2 for serious adverse outcome compared with normoxaemia. Hypoxaemia occurred in 33% of patients and was associated with an OR of 2.2 for serious adverse outcome. Expressed in terms of the BTS guideline target range, the OR for serious adverse outcome was 2.0 for those with saturation below 88% compared with 88–92%. There was no increase in risk in the range of 93–96% but the OR for serious adverse outcome was 2.4 if the saturation was above 96%. The evidence for targeted oxygen therapy in exacerbated COPD is now so strong that it is unlikely that ethical approval would be granted for any further trials that randomised patients to high-concentration oxygen. However, it is not known if 88–92% is necessarily the ideal saturation range and it would be ethical to randomise patients with COPD to targeted oxygen with a target range slightly above or below or slightly wider or narrower than 88–92%.

Unfortunately, clinical studies have shown that patients with COPD are frequently given very high concentrations of oxygen, either because of misdiagnosis or because the risks of hyperoxia in patients with COPD have been overlooked.43–45 ,285 Many patients with COPD are unaware of the diagnosis or are mislabelled as having asthma (see section 9.5). The diagnosis of COPD should be communicated clearly to the patient because ambulance teams have difficulty differentiating asthma and COPD in the prehospital setting. In a recent Australian study, only 57% of 1048 patients recorded as COPD by ambulance teams had an emergency department discharge diagnosis of COPD.300

The consensus from the literature in the past was that patients with AECOPD should be treated with Venturi masks to minimise the risks of hypercapnic respiratory failure.84 However, with the availability of continuous oximetry during medical emergencies, it is now possible to titrate the flow of oxygen from nasal cannulae to achieve the desired target range and nasal cannulae were used in the only randomised trial of targeted oxygen therapy for patients with COPD.51 If Venturi masks are used, it is not yet known if it is better to start with a 28% Venturi mask or a 24% Venturi mask. Initial management with a 28% Venturi mask appears to be safe.301 The current guideline recommends starting with either a 28% Venturi mask of 1–2 L of nasal oxygen in cases of COPD with no known history of hypercapnic respiratory failure, with downward adjustment to a 24% mask (in hospital) if the saturation rises above 92%. In cases of prior hypercapnic failure who do not have an oxygen alert card, it is recommended that prehospital treatment should be started using a 28% Venturi mask at 4 L/min or a 24% Venturi mask in hospitals (and prehospital if available) with a target saturation of 88–92%. Observational studies in the 1960s suggested that a PaO₂ of 50 mm Hg or 6.7 kPa (saturation about 84%) will prevent death from hypoxaemia in acute COPD exacerbations.290 ,291 If the saturation should fall below 88% despite treatment with a 24% or 28% Venturi mask, the patient should be treated with nasal cannulae or a simple face mask with the flow adjusted to maintain a saturation of 88–92% pending the availability of blood gas results. This small subgroup of patients is at very high risk of death and should be treated as a high priority on arrival in emergency departments, requiring immediate senior assessment and ABG measurements.

Evidence statements

▸ The target saturation in patients with COPD who are at risk of hypercapnia is 88–92% (evidence level 1+).

▸ The target saturation for patients with other risk factors for hypercapnia (eg, morbid obesity, chest wall deformities or neuromuscular disorders or fixed airflow obstruction associated with bronchiectasis) is 88–92% based on expert opinion which is extrapolated from observational studies (evidence level 4).

Recommendations

G1 (also A3): For most patients with known COPD or other known risk factors for hypercapnic respiratory failure (eg, morbid obesity, CF, chest wall deformities or neuromuscular disorders or fixed airflow obstruction associated with bronchiectasis), a target saturation range of 88–92% is suggested pending the availability of blood gas results (grade A for COPD, grade D for other conditions).

G2: Some patients with COPD and other conditions are vulnerable to repeated episodes of hypercapnic respiratory failure. In these cases, it is recommended that treatment should be based on the results of previous blood gas estimations during acute exacerbations. For patients with prior hypercapnic failure (requiring NIV or intermittent positive pressure ventilation) who do not have an alert card, it is recommended that low-concentration oxygen treatment should be started using a 24% Venturi mask at 2–3 L/min (or a 28% Venturi mask at 4 L/min or nasal cannulae at 1–2 L/min if a 24% mask is not available) with an initial target saturation of 88–92% pending urgent blood gas results. these patients should be treated as a high priority by emergency services and the oxygen concentration should be reduced if the saturation exceeds 92% but increased if it falls below 88% (grade D).

Good practice points for COPD and other conditions that may cause hypercapnic respiratory failure

Diagnosis of COPD or suspected exacerbation of COPD:

• If the diagnosis is unknown, patients over 50 years of age who are long-term smokers with a history of chronic breathlessness on minor exertion such as walking on level ground and no other known cause of breathlessness should be treated as having suspected COPD for the purposes of this guideline.

• Spirometry should be measured at least once during hospital admissions for suspected COPD (as per NICE COPD guideline1). Measurement of spirometry may confirm (or exclude) a diagnosis of airflow obstruction and the FEV₁ level is a useful indicator of disease severity in COPD.

Immediate management of patients with known or suspected COPD:

• If the saturation remains below 88% in prehospital care despite a 28% Venturi mask, change to nasal cannulae at 2–6 L/min or a simple face mask at 5 L/min with target saturation of 88–92% and alert the emergency department that the patient is to be treated as a high priority.

• Patients with a respiratory rate >30 breaths/min should have the flow rate from Venturi masks set above the minimum flow rate specified for the Venturi mask packaging to compensate for the patient's increased inspiratory flow (see figure 11B). Increasing the oxygen flow rate into a Venturi mask does not increase the concentration of oxygen which is delivered.

• Patients with a significant likelihood of severe COPD or other illness that may cause hypercapnic respiratory failure should be triaged as very urgent on arrival in hospital emergency departments and blood gases should be measured on arrival in hospital.

• Prior to availability of blood gas measurements, use a 24% Venturi mask at 2–3 L/min or nasal cannulae at 1–2 L/min or 28% Venturi mask at 4 L/min and aim for an oxygen saturation of 88–92%

Initial hospital management of patients with exacerbation of COPD:

• Patients with exacerbations of COPD need careful monitoring for hypercapnic (type 2) respiratory failure with respiratory acidosis which may develop in the course of a hospital admission even if the initial blood gases were satisfactory.

• Avoid excessive oxygen use in patients with COPD. The risk of respiratory acidosis in patients with hypercapnic respiratory failure is increased if the PaO₂ is above 10.0 kPa due to previous excessive oxygen use.

• If following blood gas measurements the pH and PCO₂ are normal, aim for an oxygen saturation of 94–98% unless there is a history of previous hypercapnic respiratory failure requiring NIV or intermittent positive pressure ventilation or if the patient's usual oxygen saturation when clinically stable is below 94% (these patients should have a target range of 88–92%). Blood gases should be repeated at 30–60 min to check for rising PCO₂ or falling pH.

• Recheck blood gases after 30–60 min (or if there is evidence of clinical deterioration) for all patients with COPD or other risk factors for hypercapnic respiratory failure even if the initial PCO₂ measurement was normal.

• If the PCO₂ is raised but pH is ≥7.35 ([H⁺]≤45 nmol/L) and/or a high bicarbonate level (>28 mmol/L), the patient has probably got long-standing hypercapnia; maintain target range of 88–92% for these patients. Blood gases should be repeated at 30–60 min to check for rising PCO₂ or falling pH.

• If the patient is hypercapnic (PCO₂>6 kPa or 45 mm Hg) and acidotic (pH<7.35 or [H⁺]>45 nmol/L), start NIV with targeted oxygen therapy if respiratory acidosis persists for more than 30 min after initiation of standard medical management.

• For patients using Venturi masks, consider changing from Venturi mask to nasal cannulae once the patient has stabilised.

• For patients who use LTOT for severe COPD, a senior clinician should consider setting a patient-specific target range if the standard range of 88–92% would require inappropriate adjustment of the patient's usual oxygen therapy while the patient is in hospital.

Management of hypercapnia or respiratory acidosis due to excessive oxygen therapy—avoidance of life-threatening rebound hypoxaemia (see section 6.3.2)

• If a patient is suspected to have hypercapnic respiratory failure due to excessive oxygen therapy, the oxygen therapy must be stepped down to the lowest level required to maintain a saturation range of 88–92%. This may be achieved using 28% or 24% oxygen from a Venturi mask or 1–2 L/min via nasal cannulae depending on oxygen saturation and subsequent blood gas measurements.

• Sudden cessation of supplementary oxygen therapy can cause life-threatening rebound hypoxaemia with a rapid fall in oxygen saturations below the starting oxygen saturation prior to the start of supplementary oxygen therapy.

8.12.2 Exacerbation of CF

Patients with breathlessness due to CF should be managed in a Cystic Fibrosis Centre unless this is not possible for geographical reasons. If not possible, all cases should be discussed with the Cystic Fibrosis Centre or managed according to a protocol that has been agreed with the regional centre. Patients with advanced CF may suffer from exacerbations which are similar to exacerbations of advanced COPD with associated hypoxaemia and hypercapnia. The principles of management are similar to those in AECOPD, including a need to maintain adequate oxygen saturation and avoiding excessive hypercapnia and acidosis. As in COPD, NIV may be of value in severe cases.302 NIV in CF may also be helpful to reduce symptoms (eg, work of breathing and breathlessness) and assist in airway clearance.

It is recommended that patients with acute exacerbations of CF should be managed on similar lines to patients with AECOPD with a target oxygen saturation of 88–92% for most patients, but recognition that individual patients may need to be managed differently on the basis of previous and current blood gas measurements. One study has shown that patients with a respiratory rate above 30 breaths/min often have an inspiratory flow rate above the minimum flow rate specified on the mask packaging and may therefore benefit from a doubling of this flow rate.303 However, there is no direct experimental evidence of the clinical effectiveness of increased flow rates from Venturi devices. It is possible that patients with very high inspiratory flow rates might benefit from a 28% Venturi mask with the flow rate set at 6–8 L/min to minimise the risk of the inspiratory flow rate exceeding the gas flow rate (see section 10.2.3). Patients with CF who have had previous episodes of hypercapnic respiratory failure should be issued with an oxygen alert card with recommendations based on previous blood gas measurements.

Evidence statement

▸ The optimal oxygen saturation level for patients with acute exacerbations of CF has not yet been investigated in clinical trials. Expert opinion advises treating these patients in a similar manner to patients with AECOPD who are at risk of hypercapnic respiratory failure (evidence level 4).

Recommendation

G3: Initial oxygen treatment of CF exacerbations should be similar to the initial oxygen treatment of COPD exacerbations with target saturation 88–92% (see sections 8.12.1–8.12.2) (grade D).

8.12.3 Chronic musculoskeletal and neurological disorders

Hypoxaemia due to musculoskeletal and neurological disorders may be associated with acute illness (such as a chest infection) superimposed on a chronic neuromuscular condition. However, muscle weakness can be acute, subacute (eg, Guillain-Barre’ syndrome) or chronic (see section 8.13.6). For most patients with inadequate ventilation due to neuromuscular weakness, non-invasive or invasive ventilatory support is more useful than supplementary oxygen and these patients are at risk of hypercapnic respiratory failure which may be aggravated by high concentrations of oxygen. For this reason, it is recommended that spirometry should be monitored carefully and blood gases should be obtained as early as possible in all such cases. Pending the availability of blood gas results, a saturation target of 88–92% will avoid the risks of severe hypoxaemia or severe hypercapnia. The NICE guidance on NIV in motor neurone disease is also applicable to other neuromuscular conditions causing respiratory failure.304

▸ Evidence statement

The optimal oxygen saturation level for patients with musculoskeletal and neurological disorders has not yet been investigated in clinical trials. Expert opinion advises urgent blood gas assessment and early consideration of NIV with a target saturation of 88–92% while blood gas results are awaited (evidence level 4).

Recommendation

G4: In the initial management of musculoskeletal and neurological disorders with acute respiratory failure or acute-on-chronic respiratory failure, aim at an oxygen saturation of 88–92% and measure blood gases to determine if NIV will be required (grade D).

Good practice point regarding patients with neurological disorders

• Patients with respiratory failure due to neurological disorders or muscle disease are at high risk of dying and require urgent assessment to determine if they are likely to require non-invasive or invasive ventilator support rather than oxygen therapy. Monitor these patients with blood gases and regular spirometry (forced vital capacity). Patient's wishes regarding this form of treatment should be established as early as possible in the course of the illness, ideally before an acute episode has developed.

8.12.4 Obesity-hypoventilation syndrome

Patients with the OHS are chronically hypercapnic unless treated with NIV and may decompensate acutely resulting in hypercapnic respiratory failure with acidosis.305 One study has found the incidence of hypoventilation to be as high as 31% in a cohort of obese hospitalised patients.306 Wijesinghe et al47 have demonstrated in 2011 that breathing 100% oxygen causes worsening hypercapnia in stable patients with obesity-associated hypoventilation. For purposes of oxygen therapy, patients with morbid obesity should be treated in a similar manner to patients with an AECOPD. The target saturation will usually be 88–92% but, as with COPD, a lower target range may be appropriate for individual patients based on blood gas measurements during a previous exacerbation or due to acute acidosis. Assessment of patients with increasing shortness of breath or worsening oxygen saturation must include blood gases. As in COPD, patients with respiratory acidosis may benefit from NIV.

Evidence statement

• Patients with OHS are at risk of worsening hypercapnia when administered high-flow oxygen (evidence level 1+).

• However, the optimal oxygen saturation level for patients with OHS has not yet been investigated in clinical trials. Expert opinion advises treating these patients in a similar manner to patients with AECOPD who are at risk of hypercapnic respiratory failure (evidence level 4).

Recommendations

G5: Morbidly obese patients (BMI>40 kg/m²), even without evidence of coexistent OSA are at risk of hypoventilation and should be given titrated oxygen to maintain a target saturation of 88–92% (grade D).

G6: NIV should be considered for hypercapnic patients with COPD, CF, neuromuscular disorders or morbid obesity who are at risk of hypercapnic respiratory failure if the pH is <7.35 or [H⁺]>45 nmol/L (grade D). See BTS/ICS Guideline for the ventilatory management of acute hypercapnic respiratory failure (ref 299).

8.13.1 Acute myocardial infarction, suspected myocardial infarction and acute coronary syndromes

Some patients with acute myocardial infarction have heart failure and should be treated accordingly (see section 8.11.9). Most patients with suspected or confirmed myocardial infarction are not hypoxaemic and most are not breathless. In the case of non-hypoxaemic patients, it is not known if supplementary oxygen may be beneficial by increasing the amount of oxygen delivered to the hypoxaemic area of myocardium or whether it may actually cause vasoconstriction with increased systemic vascular resistance and reduced myocardial oxygen supply with worsened systolic myocardial performance.307–315 A study of patients having coronary arteriography found that breathing 100% oxygen reduced coronary blood flow velocity by 20% and increased coronary resistance by 23%.316 A study of 16 healthy individuals using MRI found that coronary artery oxygen delivery fell by 11% due to decreased cardiac output and decreased left ventricular perfusion while breathing oxygen at 8 L/min via a mask despite the increased blood oxygen content.317 However, Ranchord et al318 found that high-concentration oxygen increased the time to onset of exercise-induced myocardial ischaemia in patients with stable ischaemic heart disease, suggesting that the physiology of oxygen delivery to the heart may be different during exercise in ischaemic heart disease. Reduced myocardial oxygen consumption might also be associated with reduced contractility.

There is a theoretical possibility that high oxygen levels might exacerbate reperfusion injury to the heart.314 Despite a multitude of large studies of intervention in myocardial infarction, there had been only three published randomised studies of oxygen therapy prior to 2015. A randomised trial in 1976 did not identify any benefit from such therapy but found some evidence of potential harm.319 This trial reported a significantly greater rise in aspartate aminotransferase and heart rate which could suggest increased infarct size in the oxygen treated group. There was a difference in mortality between the two groups that did not reach statistical significance (3 deaths in 77 patients treated with air vs 9 deaths in 80 patients given oxygen at 6 L/min via a simple face mask for 24 hours). This study was undertaken at a time before angiographic interventions were possible for myocardial infarction.

Ranchord et al320 randomised 136 patients with ST-elevation myocardial infarction to 6 L of oxygen via simple face mask or titrated oxygen to achieve saturation of 93–96% for 6 hours after randomisation in a hospital setting. The trial was conducted between 2007 and 2009 and most patients had angiographic interventions. There was no difference in troponin T level between the groups and the number of deaths was too low (3) to permit meaningful analysis. Eighty-five per cent of patients in both groups had received oxygen in the ambulance and/or emergency department for a mean of 62 min prior to randomisation and almost all patients received thrombolysis or percutaneous coronary intervention within 4 hours of onset of chest pain. Therefore, this study really evaluated the effect of stopping oxygen therapy prior to therapeutic intervention in a group of patients of whom most had already received oxygen for more than 1 hour.

The third study was an open-label randomised study conducted in Russia which was reported with limited methodological details.321 Twenty-eight patients received oxygen in addition to standard therapy for 30 min prior to myocardial revascularisation and for 3 hours after the intervention, 30 patients received oxygen for 3 hours after the intervention and 79 patients received ‘standard therapy’. Details of oxygen therapy prior to randomisation were not given. This study reported one death among 58 patients randomised to oxygen therapy and no deaths among 79 patients randomised to standard therapy but the authors reported fewer complications among those randomised to oxygen.

A systematic review and a historical review of oxygen therapy in acute myocardial ischaemia have both concluded that there was no evidence to support this practice in non-hypoxaemic patients and some evidence of possible harm.322 ,323 Further systematic reviews and meta-analysis in 2011 and 2012 concluded that the existing studies lacked power, no clear conclusions can be drawn and further studies are needed.324 ,325 One study from 1969 showed that hypoxaemia did not affect the availability of oxygen for myocardial metabolism in normal participants until the oxygen saturation fell to about 50%, but evidence of myocardial ischaemia was seen at saturations of 70–85% in participants with coronary artery disease.326 In these circumstances, it is advised that patients with myocardial infarction or chest pain suspicious of myocardial infarction should be given supplementary oxygen if required to maintain a saturation of 94–98%.

The study by Wilson and Channer4 in 1997 showed that desaturation below 90% was common in patients with myocardial infarction within the first 24 hours of admission to a coronary care unit, but these authors may not have been aware that nocturnal desaturation to this level is very common in healthy individuals.33 Wilson and Channer4 did not demonstrate any correlation between hypoxaemic events and adverse cardiac events. They did, however, show that monitoring by oximetry was inadequate in UK coronary care units in the mid-1990s.

The 2010 NICE Guideline for Chest Pain of recent onset recommends setting an oxygen target saturation range of 94–98% based on the arguments discussed above and the advice given in the 2008 BTS Emergency Oxygen Guideline.327 The 2012 guidelines for acute myocardial infarction of the European Society for Cardiology recommends the following: “Oxygen (by mask or nasal prongs) should be administered to those who are breathless, hypoxic (SaO₂<95%), or who have heart failure.”328 The 2011 American College of Cardiology/American Heart Association guideline for Management of Patients With Unstable Angina/Non-ST-Elevation Myocardial Infarction gives the following advice:329 “Supplemental oxygen should be administered to patients with UA/NSTEMI with an arterial saturation <90%, respiratory distress, or other high-risk features for hypoxemia.” This is followed by the following unreferenced advice: “It is reasonable to administer supplemental oxygen to all patients with UA/NSTEMI during the first 6 h after presentation.” The limited available evidence supports the suggestion that clinicians should aim at normal or near-normal oxygen saturation in patients with myocardial infarction, acute coronary syndrome and chest pain suspicious of coronary artery disease. A target saturation range of 94–98% will meet all of these goals, and further research of this topic should be prioritised because this is such a common medical problem and there is so little existing evidence.

Fortunately, there is one recent published study and other forthcoming studies of oxygen use for myocardial infarction in ambulances. The AVOID study in Australia was study was designed to determine whether withholding routine supplemental oxygen therapy in patients with acute ST-elevation myocardial infarction but without hypoxaemia prior to reperfusion decreases myocardial infarct size.330–332 The results have suggested that cardiac enzyme levels were higher and infarct size at 6 months was 55% bigger as determined by cardiac MR scan when oxygen was administered. Mean peak troponin was similar in the oxygen and no oxygen groups (57.4 vs 48.0 μg/L; ratio 1.20; 95% CI 0.92 to 1.56; p=0.18). There was a significant increase in mean peak creatine kinase in the oxygen group compared with the no oxygen group (1948 vs 1543 U/L; means ratio 1.27; 95% CI 1.04 to 1.52; p=0.01). There was an increase in the rate of recurrent myocardial infarction in the oxygen group compared with the no oxygen group (5.5% vs 0.9%; p=0.006) and an increase in frequency of cardiac arrhythmia (40.4% vs 31.4%; p=0.05). At 6 months, the oxygen group had an increase in myocardial infarct size on cardiac MR (n=139; 20.3 vs 13.1 g; p=0.04).331 A subsequent publication by these authors showed that there was a significant increase in troponin and creatine kinase levels which was dependent on the total dose of supplemental oxygen (litres of supplemental oxygen received) and a typical patient receiving supplemental oxygen in the first 12 hours after ST-elevation myocardial infarction would experience an ∼20% increase in myocardial infarct size.332

The DETO2X-AMI trial is a further randomised trial of oxygen versus air in the management of suspected myocardial infarction and was started in Sweden and will hopefully provide more useful information concerning the management of this common medical emergency.333

In the past, most emergency calls to ambulance services because of chest pain were treated with high-concentration oxygen in accordance with the Joint Royal Colleges Ambulance Liaison Committee (JRCALC) older guidance.334 However, this guidance was updated in 2009 to reflect the advice given in the 2008 BTS Emergency Oxygen Guideline with a target saturation range of 94–98% for most patients with chest pain.12 Most such patients have a final diagnosis of undifferentiated chest pain rather than acute coronary artery syndrome and most patients with undifferentiated chest pain are normoxaemic. The clinical management of a very large number of patients was therefore changed following the introduction of the 2008 BTS guideline and subsequent publications have tended to support the concept of aiming for normoxaemia in patients with suspected myocardial infarction.

Evidence statements

▸ The optimal oxygen saturation level for patients with myocardial infarction and acute coronary syndromes has not yet been investigated in clinical trials. Expert opinion advises maintaining normal or near-normal saturation unless the patient is at risk of hypercapnic respiratory failure (evidence level 4).

▸ There is evidence from two randomised studies that the administration of oxygen to non-hypoxaemic patients in the immediate management of myocardial infarction may be associated with increased infarct size (evidence level 1+).

Recommendation

F13: In myocardial infarction and acute coronary syndromes, aim at an oxygen saturation of 94–98% or 88–92% if the patient is at risk of hypercapnic respiratory failure (grade D).

8.13.2 Stroke

In the past, it was customary to give supplemental oxygen to all patients with stroke to try to improve cerebral oxygenation. There are four published randomised trials of oxygen therapy in patients with stroke with normal oxygen saturation, none of which have shown benefits. A study of supplemental oxygen (2 or 3 L/min dependent on baseline oxygen saturation) for 24 hours (n=550) found no difference in 1-year survival for the entire cohort of patients with stroke and no difference in survival for patients with more severe strokes. However, a post hoc subgroup analysis found that for patients with minor strokes, 1-year mortality was 18% in the group given oxygen and 9% in the group given air (OR 0.45; 95% CI 0.23 to 0.90; p=0.02).335 A pilot study of oxygen supplementation at a rate of 2 or 3 L/min, dependent on baseline oxygen saturation for 72 hours in the UK (n=300) suggested better neurological recovery at 1 week, but showed no difference in survival and recovery at 6 months.336 ,337 The study did not find an excess risk for minor strokes in patients given oxygen. A much larger study (n=8003) using the same dose and duration of oxygen treatment has just closed and is only available in abstract format so far, but confirms that routine oxygen treatment does not reduce mortality or improve recovery.338

A small randomised trial of high-concentration oxygen therapy (10 L/min) in non-hypoxaemic patients (n=40) showed no clinical benefit339 and a study of very high-flow oxygen (45 L/min) was terminated in 2009 after enrolling 85 patients because of excess mortality in the hyperoxaemia group (40% vs 17%), the full report had not yet been published when this guideline was prepared.340 Finally, an observational study reported increased mortality associated with hyperoxaemia in ventilated patients with stroke on ICUs.341

Post-stroke hypoxaemia is common and associated with worse outcome in some small studies.342 ,343 The risk of hypoxaemia is particularly high at night, during interward transfers and in the CT scanner.342 ,344 Regular monitoring of oxygen saturation and treatment of all episodes of hypoxaemia has been shown to improve outcome in an observational study.343 It is therefore recommended to monitor oxygen saturation regularly (at least every 4 hours, including the night) during the acute phase of stroke and to treat every episode of hypoxaemia.

The most common causes of hypoxaemia after stroke are obstruction of airflow by secretions in the upper airway or complications of the stroke such as pneumonia, pulmonary embolism or heart failure. These must be addressed in every patient who develops hypoxaemia post-stroke.

It is recommended that patients with stroke should receive supplemental oxygen only after the airway has been cleared and if this treatment is required to achieve an oxygen saturation of 94–98% (88–92% for patients with coexisting risk of COPD or other risk of respiratory acidosis). High concentrations of oxygen should be avoided unless required to correct hypoxaemia. As patients with stroke tolerate nasal cannulae better than oxygen masks, oxygen should be given via nasal cannulae unless a strictly controlled oxygen concentration is indicated.345 ,346

There has also been some discussion concerning the optimal body position for the management of patients with stroke and potential hypoxaemia. A systematic review of clinical studies in conscious patients concluded that there was limited evidence that sitting in a chair had a beneficial effect and lying positions had a deleterious effect on oxygen saturation in patients with acute stroke with respiratory comorbidities, but patients with acute stroke without respiratory comorbidities can adopt any body position.161 The authors of this review recommended that people with acute stroke and respiratory comorbidities should be positioned as upright as possible as discussed in section 6 and recommendation A5. No trials have expressly addressed the best position in unconscious and semiconscious patients with stroke. As in other patient groups with a reduced level of consciousness the recovery position is recommended to safeguard the airway.

Evidence statements

• High concentrations of oxygen in normoxaemic patients with stroke may be associated with increased mortality (evidence level 1−).

  • Routine oxygen for non-hypoxic patients with stroke does not improve recovery or reduce disability (evidence level 1−).

• Patient positioning has not been shown to affect SaO₂ as measured by pulse oximetry in patients without cardiorespiratory comorbidities, but sitting out in a chair improves oxygenation in patients with pre-existing cardiorespiratory disease (evidence level 1−).

Recommendation

F14: High concentrations of oxygen should be avoided in patients with stroke, unless required to maintain normal oxygen saturation. Aim at an oxygen saturation of 94–98% or 88–92% if the patient is at risk of hypercapnic respiratory failure (grade D).

Good practice points regarding stroke management

• Oxygen saturation should be monitored at least every 4 hours throughout the day and night in patients with acute stroke and all episodes of hypoxaemia treated.

• Patients with hypoxaemia post-stroke require medical review to establish and treat the cause.

• Oxygen should only be given once the airway has been cleared and at the lowest concentration necessary to achieve an oxygen saturation of 94–98% or 88–92% if the patient is at risk of hypercapnic respiratory failure.

• Oxygen should be given via nasal cannulae, unless there are clear indications for a different oxygen delivery system.

• Patients with stroke and cardiorespiratory comorbidities should be positioned as upright as possible, in a chair if possible (see recommendation A5).

• Patients with a reduced level of consciousness after stroke should be nursed in the recovery position with the paralysed side lowest.

8.13.3 Anxiety and hyperventilation or dysfunctional breathing

Many patients who present to hospital with breathlessness are found to have no cardiopulmonary problems and many such patients have a specific diagnosis of hyperventilation, dysfunctional breathing, vocal cord dysfunction or panic attacks, sometimes in addition to asthma or some other underlying respiratory disorder.347 Many such patients will have an abnormally high oxygen saturation of 99% or 100% and clearly do not require supplemental oxygen therapy but some patients with vocal cord dysfunction may develop respiratory failure due to upper airway obstruction. Many other non-hypoxaemic patients will present to hospital with acute breathlessness of unknown cause, and the majority of patients with an elevated respiratory rate are likely to have an organic illness. In some cases, simple investigations will reveal a specific diagnosis such as pneumothorax or pneumonia or pulmonary embolism, but many cases remain undiagnosed. A policy of giving supplemental oxygen if the saturation falls below 94% will avoid exposing patients with undiagnosed medical illnesses to the risk of hypoxaemia while avoiding the unnecessary use of oxygen in patients with behavioural or dysfunctional breathlessness.

Studies in normal volunteers have demonstrated that compensatory desaturation may occur shortly after voluntary hyperventilation.348 The mean PaO₂ of 10 male volunteers increased from 13.7 kPa (103 mm Hg) to 18.6 kPa (140 mm Hg) during hyperventilation but fell to a nadir of 7.8 kPa (58 mm Hg) about 7 min after cessation of hyperventilation and did not normalise until after a total of 17 min of observation. It is not known whether or not this occurs after pathological hyperventilation, but this phenomenon could cause considerable confusion if it should occur in an emergency department.

A traditional treatment for hyperventilation was to ask the participant to rebreathe from a paper bag to allow the carbon dioxide level in the blood to normalise. However, it has been shown that this practice can cause hypoxaemia with potentially fatal consequences.349 The average fall in PO₂ during rebreathing was 26 mm Hg (3.5 kPa) and the maximum fall was 42 mm Hg (5.6 kPa). This guideline does not recommend rebreathing from a paper bag in cases of hyperventilation.

Good practice points regarding patients with suspected hyperventilation

• Organic illness must be excluded before making a diagnosis of hyperventilation.

• Patients with a definite diagnosis of hyperventilation should have their oxygen saturation monitored. Those with normal or high SpO₂ do not require oxygen therapy.

• Rebreathing from a paper bag can be dangerous and is NOT advised as a treatment for hyperventilation.

8.13.4 Poisoning with substances other than carbon monoxide or cyanide

Many poisons and drugs can cause respiratory or cardiac depression or direct toxic effects on the lungs. The treatment of individual toxic agents is beyond the scope of this guideline. Specific antidotes such as naloxone for opioids or flumazenil for benzodiazepines should be given if available and oxygen saturation should be monitored closely. Supplemental oxygen should be given to achieve a target saturation of 94–98% pending the results of blood gas analysis (88–92% if at risk of hypercapnic respiratory failure, including patients with overdoses of narcotic agents). All potentially serious cases of poisoning should be monitored in a level 2 or 3 environment (HDU or ICU).

Three specific types of lung injury deserve special mention. Oxygen is known to be hazardous to patients with paraquat poisoning,125 ,126 and oxygen potentiates bleomycin lung injury. Because of these risks, oxygen should be given to patients with these conditions only if the oxygen saturation falls below 85%. Some authors have suggested the use of hypoxic ventilation with 14% oxygen as a specific treatment for paraquat poisoning.350 Bleomycin lung injury can be potentiated by high-concentration oxygen therapy, even if given several years after the initial lung injury.127 It is therefore recommended that high concentrations of oxygen should be avoided in patients with possible bleomycin-induced lung injury and a lower oxygen saturation target range should be accepted (eg, 88–92%).

There is evidence from animal experiments that oxygen may potentiate lung injury from aspiration of acids.128 ,129 The effect in humans is not known so patients with acid inhalation should have the usual adult target saturation range of 94–98%, but it would appear prudent to aim in the lower half of the target range for these patients and clinical trials in humans are clearly required.

Evidence statement

• For most types of poisoning, the optimal oxygen saturation level has not yet been investigated in clinical trials. Expert opinion advises maintaining normal or near-normal saturation unless the patient is at risk of hypercapnic respiratory failure or unless they require high oxygen levels (carbon monoxide or cyanide poisoning) or have ingested a poison such as paraquat or bleomycin where supplemental oxygen may aggravate tissue damage (evidence level 4).

Recommendations

F15: In most poisonings, aim at an oxygen saturation of 94–98% unless the patient is at risk of hypercapnic respiratory failure (grade D).

F16: In poisoning by paraquat and poisoning by bleomycin, give oxygen only if the saturation falls below 85% and reduce or stop oxygen therapy if the saturation rises above 88% (grade D).

8.13.5 Metabolic, endocrine and renal disorders

Many metabolic and renal disorders can cause metabolic acidosis which increases respiratory drive as the body tries to correct the acidosis by increased excretion of carbon dioxide via the lungs. Although these patients have tachypnoea, they do not usually complain of breathlessness and most have high oxygen saturation (unless there is a coexisting pulmonary or cardiac problem). Supplementary oxygen is not required for such patients unless the oxygen saturation is reduced. In such cases, oxygen should be given to maintain a saturation of 94–98%.

Evidence statement

• The optimal oxygen saturation level for patients with most metabolic, endocrine and renal disorders has not yet been investigated in clinical trials. Expert opinion advises maintaining normal or near-normal saturation unless the patient is at risk of hypercapnic respiratory failure (evidence level 4).

Recommendation

F17: In most metabolic and renal disorders, aim at an oxygen saturation of 94–98% unless the patient is at risk of hypercapnic respiratory failure (grade D).

8.13.6 Acute and subacute neuromuscular disorders producing respiratory muscle weakness

Patients with acute or subacute conditions affecting the respiratory muscles (often superimposed on chronic conditions) are at risk of sudden onset of respiratory failure with hypoxaemia and hypercapnia and may require non-invasive or invasive ventilatory support. This applies especially to patients with Guillain-Barre’ syndrome for whom spirometry (forced vital capacity) should be monitored carefully as this should detect the onset of severe respiratory failure prior to the development of hypoxaemia. If the oxygen level falls below the target saturation, urgent blood gas measurements should be undertaken and the patient is likely to need ventilatory support.

Good practice point

• Patients with respiratory failure due to neurological disorders or muscle disease are at high risk of dying and require urgent assessment to determine if they are likely to require non-invasive or invasive ventilator support rather than oxygen therapy. Monitor these patients with blood gases and regular spirometry (forced vital capacity). Patient's wishes regarding this form of treatment should be established as early as possible in the course of the illness, ideally before an acute episode has developed.

8.13.7 Cluster headache

Relief from cluster headache has been reported in 56–85% of cases.351–354 Benefit has recently been reported for other types of headache.355 Although this could be considered as a form of emergency oxygen therapy, these patients are not breathless or hypoxaemic. Cluster headache is one of the most severe pain disorders known to humans. Patients can attend emergency departments in an acute cluster attack but because of the relatively short duration and frequency of the episodes, home treatment of the acute cluster attack is most often undertaken. For patients with cluster headache arrangements for home and ambulatory oxygen is recommended for treatment of the acute cluster headache bout. Oxygen at a flow rate of at least 12 L/min from a reservoir mask is recommended.353 Patients who have episodic cluster headache may not need oxygen on a longer term basis and once they have been cluster headache free for over 4 weeks the oxygen supply can be uplifted. However, because of the recurrent nature of the condition, emergency supply of oxygen should be made available for cluster headache sufferers. In a study comparing the use of oxygen with air, 78% patients given oxygen at 12 L/min via reservoir mask for 15 min had a reduction in pain at 15 min as compared with 20% for those breathing air at 12 L/min (p<0.001).354 In a further study, there was a complete or substantial reduction in pain at 30 min in 56% of patients (nine participants) in 80% or more of their cluster headaches when given oxygen at 6 L/min for up to 15 min from a reservoir mask compared with 7% on air.352

Evidence statement

• Oxygen therapy can provide relief from cluster headache (evidence level 1−).

Recommendation

F18: For patients with cluster headaches, oxygen should be administered using a flow of at least 12 L/min from a reservoir mask and home oxygen should be provided (grade D).

8.15.1 Oxygen use, pulse oximetry and the incidence of hypoxaemia in the postoperative period

Postoperative breathlessness and hypoxaemia can develop due to a variety of complications including atelectasis, pneumonia and pulmonary embolism. The use of opioid analgesia commonly used for moderate-to-severe pain can lead to and exacerbate respiratory failure, which can potentially be life-threatening. Historically, it has been common practice for postoperative patients to receive routine supplemental oxygen for a period ranging from hours to days after surgery to prevent hypoxaemia, particularly those receiving opioid analgesia. High-risk patients undergoing major surgery are best managed in an ICU and HDU settings which facilitate the continuous monitoring of SpO₂ and usually end tidal CO₂ (this is standard in ventilated patients). Regular ABG analysis using arterial lines is also available in ICU and HDU settings. This allows oxygen therapy to be administered in a dose sufficient to relieve hypoxaemia without causing hyperoxia.

However, the majority of postoperative patients are managed on general surgical wards where continuous monitoring is not possible. The question of whether all postoperative patients should receive supplemental oxygen routinely is debatable and there is lack of robust evidence from RCTs to either support or refute its use.

A number of observational studies have documented a high incidence of postoperative hypoxaemia.182 ,362–364 A Cochrane review of 22 992 patients reported on the incidence of hypoxaemia in the perioperative and postoperative period.181 Unsurprisingly, it is difficult to draw any conclusions on the overall incidence of hypoxaemia due to the variation in population groups, operative procedure and anaesthetic modality.

This same Cochrane review included data from five trials in which patients were randomised to pulse oximetry monitoring or no pulse oximetry during and after surgery.181 The review showed that pulse oximetry leads to a 1.5-fold to 3-fold reduction in perioperative hypoxaemia. However, while pulse oximetry led to the detection of hypoxaemia and interventions to correct it, there was no difference in complication rates between the two groups. Additionally postoperative cognitive function, length of hospital stay and mortality was the same in both groups. The authors suggested that correcting modest hypoxaemia, by increasing the blood oxygen saturation from marginal to satisfactory may confer no benefit to the patient. However, only small numbers of studies were included and further work is required for clarification.

The incidence of postoperative hypoxaemia demonstrated in randomised trials is again, variable. Gift et al366 found no clinically significant difference in SpO₂ in nearly 300 postoperative patients randomised to four different treatments including oxygen 4 L/min via nasal cannulae, 40% oxygen via mask, nurse-led hyperinflation and no treatment.365 Canet et al365 found that the incidence of hypoxaemia was higher after breathing air compared with 35% oxygen and the incidence increased if the patients were older or if they received general anaesthesia.366

Given the potential to develop hypoxaemia and the inability to continuously monitor oxygenation in most postoperative patients, is it safer to routinely administer supplemental oxygen to all patients to prevent hypoxaemia? The major disadvantage to this approach is that while supplemental oxygen will correct hypoxaemia it can mask the ability to detect hypoventilation.228 Additionally, recently, Niesters et al107 have shown that hyperoxia has an additive effect on opioid-induced respiratory depression. Furthermore, there is a large body of evidence showing hyperoxia is associated with haemodynamic alterations which may increase myocardial ischaemia and impair cardiac performance (see section 8.13.1). A recent randomised trial of 340 patients postcardiac surgery administering nasal high-flow oxygen compared with usual care (oxygen to keep SpO₂>93%) showed no improvement in postoperative oxygenation.367 However, patients in the high-flow group were less likely to require escalation of respiratory support at any time during the study (47 patients in the high-flow group (27.8%) compared with 77 patients (45%) in the standard care group (OR 0.47, 95% CI 0.29 to 0.7, p<0.001). The authors hypothesised that the low-level airway pressure support provided by the nasal high-flow system may have influenced this outcome.

In a recent editorial, Martin and Grocott368 highlight the need to re-evaluate the unrestricted use of oxygen by anaesthetists. They discuss the relationship between oxygen therapy in perioperative care, critical care and resuscitation, highlighting the lack of clinical benefit from hyperoxaemia and the potential to lead to poor outcomes. They propose a strategy of ‘precise control of arterial oxygenation’ in which oxygen is administered to a defined target, for example, PaO₂ of 8–10 kPa or SaO₂ 88–92% thus avoiding potential harm from hypoxaemia and hyperoxaemia and improving clinical outcomes. These recommendations are in keeping with this present guideline.

The ‘routine’ use of supplemental oxygen postoperatively, therefore remains controversial.76 ,181 ,369–372 More evidence from RCTs is required. Regular, but not necessarily, continuous oxygen saturation monitoring is mandatory. In keeping with the 2004 SIGN guideline on postoperative care which recommends maintaining an oxygen saturation above 92% for postoperative patients, this present guideline recommends administering oxygen to maintain a target saturation of 94–98% for most patients, and a target saturation of 88–92% in those patients at risk of hypercapnic respiratory failure.370

Evidence statement

• Pulse oximetry in the perioperative period leads to a significant reduction in perioperative hypoxaemia. However, identification and correction of mild or moderate hypoxaemia with pulse oximetry does not lead to a reduction in complication rates, length of stay or mortality (evidence level 1+).

Good practice points

• A target saturation of 94–98% is recommended for most surgical patients except those at risk of hypercapnic respiratory failure when a range of 88–92% should be achieved.

• Pulse oximetry monitoring is recommended for postoperative patients despite the lack of evidence from randomised studies.

8.15.2 Patient-controlled analgesia

One of the commonest indications for anaesthetists to prescribe supplemental postoperative oxygen therapy is to help prevent hypoxaemia when PCA is used. In two reviews, the incidence of postoperative hypoxaemia in patients receiving PCA was found to be 11.5% and 15.2%.373 ,374

However, Cashman and Dolin's373 systematic review comparing the effects of acute postoperative pain management on respiratory function found PCA did not lead to more respiratory events when compared with intramuscular analgesia and epidural anaesthesia. They reviewed 165 studies used a variety of indicators for respiratory depression including hypoventilation, hypercarbia, oxygen desaturation and requirement for naloxone. When using respiratory rate as a measure of respiratory depression, there was no difference between all three modalities of analgesia. However, when oxygen desaturation was used as an indicator for respiratory depression the incidence of desaturation in the intravenous PCA group was significantly lower (mean 11.5% (95% CI 5.6% to 22%)) compared with epidural analgesia (mean 15.1% (95% CI 5.6% to 22%)) and intramuscular analgesia (mean 37% (95% CI 22.6% to 45.9%)). This is perhaps surprising when PCA use is the commonest indication for supplemental oxygen.

Importantly, while administering supplemental oxygen corrects hypoxaemia, it has the ability to mask respiratory depression by delaying the onset of desaturation.102 ,107 A study using transcutaneous carbon dioxide monitoring via ear lobe probe (TOSCA) compared postoperative patients using epidural analgesia and patients using morphine via PCA infusion pump.375 All patients received oxygen 4 L/min in the postoperative period. Significant desaturations were not observed in either group. However, the PCA group had a higher median PtCO₂, lower respiratory rate and longer hypercarbia time as detected by the TOSCA. This study demonstrates that transcutaneous carbon dioxide monitoring is useful in detecting opioid-induced respiratory depression. It also shows that hypoventilation is common when PCA analgesia is administered. However, outside the operating theatre and intensive care where only pulse oximetry is available, the use of supplemental oxygen may mask hypoventilation. For this reason, oxygen should be administered to correct hypoxaemia rather than prevent it. A target saturation of 94–98% is recommended in most patients except those at risk of hypercapnic respiratory failure when a range of 88–92% should be achieved.

Niesters et al107 have recently demonstrated in a proof-of-concept study that hyperoxaemia has an additive effect on opioid-induced respiratory depression. In a study of 20 healthy volunteers, there was steep reduction in minute ventilation and increase in end-tidal CO₂ when breathing 50% oxygen compared with breathing air while receiving a remifentanyl infusion. Additionally apnoeic episodes were significantly higher while breathing oxygen. An audit of blood gases from 4866 patients at a university hospital has shown that hypercapnia was commoner than hypoxaemia in blood gas samples from surgical wards, a surgical HDU, theatre and the ICU as well as on medical wards and many of these hypercapnic patients had respiratory acidosis.42

In the absence of large RCTs investigating the effects of PCA and the effects of hyperoxaemia, oxygen should be administered to correct hypoxaemia rather than prevent it. A target saturation of 94–98% is recommended in most patients except those at risk of hypercapnic respiratory failure when a range of 88–92% should be achieved.

Good practice points

• Patients using PCA should have two-hourly oximetry observations because of the risk of hypoxaemia. Oxygen should be administered to keep patients within the appropriate target saturation range.

• A target saturation of 94–98% is advised in most patients having PCA except those at risk of hypercapnic respiratory failure when a range of 88–92% should be achieved.

8.15.3 The role of hyperoxaemia in reducing postoperative complications

High levels of inspired oxygen (eg, 80%) in the perioperative and postoperative periods have in the past, been proposed to improve clinical outcomes including a reduction in the incidence of surgical site infection. The rationale for its use being that increased tissue PO₂ within surgical wounds might enhance neutrophil killing capacity resulting in reduced rates of infection. Earlier meta-analysis supported this hypothesis;376 ,377 however, three RCTs have been published subsequently and do not concur with these findings.378–380

Two recent meta-analysis again show conflicting results: Togioka et al381 included seven randomised trials and found no overall benefit of hyperoxia in reducing surgical site infection. However, a statistically significant benefit was demonstrated in subgroups undergoing general anaesthesia and colorectal surgery. Hovaguimian et al382 included nine randomised trials in their meta-analysis and concluded that high perioperative oxygen did reduce the risk of perioperative surgical site infection. Both these meta-analyses included the largest RCT in this area (the PROXI trial), including 1400 patients, which showed no benefit in the administration of 80% oxygen compared with 30% oxygen in the risk of developing surgical site infection after abdominal surgery.379 Long-term follow-up of patients with cancer within the PROXI study revealed a 45% increase in long-term mortality in the high FiO₂ group (HR was 1.45; 95% CI 1.10 to 1.90; p=0.009).383 Kurz et al384 reported in 2015 that supplemental oxygen did not reduce surgical site infection risk among patients having colorectal resections.

Hyperoxia has also been proposed to reduce the incidence of postoperative nausea and vomiting via a number of mechanisms including the prevention of subtle intestinal ischaemia reducing serotonin release and the reduction of dopamine release in the carotid bodies. This has been extensively investigated with conflicting evidence from RCTs. Two recent systematic reviews and meta-analyses show differing results.382 ,385 Orhan-Sungur et al385 included data from 10 studies and showed no overall benefit in hyperoxaemia reducing the risks of postoperative nausea and vomiting. Hovaguimian et al382 included data from 11 studies and showed high concentrations of oxygen provided a significant protective effect in a subgroup of patients who received inhalation anaesthetics without prophylactic antiemetics. However, overall there was no benefit in the composite end point postoperative nausea and vomiting.

Experimental data have shown hyperbaric oxygen improves anastomotic integrity in animal models.386 ,387 Recently, 72 patients who underwent elective open infraperitoneal anastomosis for rectal cancer were randomised to 30% or 80% oxygen at induction and for 6 hours postoperatively.388 The risk of anastomotic leak was 46% lower in the 80% FiO₂ group (RR 0.63; 95% CI 0.42 to 0.98) versus the 30% FiO₂ group. The same group has shown similar results in patients with oesophagojejunal anastomosis when the risk of anastomotic leak was 49% lower in the 80% FiO₂ group (RR 0.61; 95% CI 0.40 to 0.95) versus the 30% FiO₂ group.389 However, the numbers of patients in both studies in each treatment arm were small and further work is required from other investigators to validate this finding, especially in view of the concerns of Meyhoff et al383 that high-concentration oxygen may increase medium-term mortality in patients having cancer surgery.

Evidence statements

▸ The possible effects of hyperoxaemia in the perioperative and postoperative period remain controversial and may or may not reduce overall surgical site infection. However, there may be benefit in subgroups of patients undergoing general anaesthesia and colorectal surgery for non-malignant disease (evidence level 1−).

▸ Hyperoxaemia in the perioperative and postoperative period does not substantially reduce the incidence of postoperative nausea and vomiting but meta-analyses have come to conflicting conclusions (evidence level 1−).

▸ Hyperoxaemia in the perioperative and postoperative period has been shown to improve anastomotic integrity in patients undergoing gastric and colorectal surgery (evidence level 1−).

Recommendation

J1: Hyperoxaemia is not recommended routinely in the perioperative and postoperative period to reduce the incidence of postoperative nausea and vomiting (grade D).

Good practice point

• There is conflicting evidence concerning the balance of potential benefits and risks of perioperative hyperoxaemia to reduce the risk of surgical site infection in elective surgery and there is no evidence for this practice in patients having emergency surgical procedures. More trials are required for specific procedures and more information is required concerning long-term mortality risks to patients with cancer. In the meantime, oxygen should not be used for this indication outside of clinical trials.

8.19.1 Use of CPAP in perioperative care

Obesity has reached epidemic proportions with more than one billion overweight adults worldwide, of whom at least 300 million are obese (see also section 8.15). Obesity predisposes to OSA and OHS. CPAP is the mainstay of medical treatment for patients with OSA. It has been shown to reduce apnoeas and daytime sleepiness as well as reducing sequelae of untreated OSA such as impaired cognition, hypertension, coronary artery disease and cerebrovascular accident.435 The incidence of postoperative desaturation and respiratory failure is unsurprisingly high in patients with OSA.436 Residual anaesthetic drugs have the potential to weaken the muscles of the upper airway and depress the respiratory drive, thus amplifying the disease. CPAP in the postoperative period has been shown to reduce reintubation and severe respiratory complications after major abdominal surgery in patients without sleep-disordered breathing.437 An observational study of 16 patients with OSA reported a reduction in postoperative complications in patients who used CPAP preoperatively, on extubation and continuously for 24–48 hours after surgery.438 The British Obesity and Metabolic Surgery Society (BOMSS) recommend patients with diagnosed sleep-disordered breathing established on CPAP or bi-level positive airway pressure ventilation (BiPAP) who are undergoing surgery, to bring their machines with them and use them in the preoperative and postoperative period.439 Patients with OSA or OHS with evidence of desaturation <88% in the postoperative period should be started on CPAP or BiPAP rather than supplemental oxygen alone. Supplemental oxygen can be entrained if necessary to maintain the target saturation range, usually 88–92%.

Caution should be exercised when administering oxygen to all morbidly obese patients even those without a diagnosis of sleep-disordered breathing. One study of 40 morbidly obese patients undergoing laparoscopic bariatric surgery showed the incidence of postoperative hypoxaemia was as high in patients without OSA compared with those with a diagnosis of OSA and supplemental oxygen did not reduce the frequency of oxygen desaturations.440

Evidence statements

▸ OSA is associated with increased incidence of desaturation, respiratory failure, cardiac events and ICU admission (evidence level 1+).

▸ CPAP therapy used in the preoperative and postoperative period reduces complications in patients with OSA after surgery (evidence level 2+).

Recommendation

N1: Patients with diagnosed sleep-disordered breathing established on CPAP undergoing surgery should bring their machines with them and use them in the preoperative and postoperative period. If adequate saturations are not achieved despite CPAP therapy then assess for worsening ventilation with blood gases and oxygen should be entrained to achieve a saturation of 88–92% (grade D).

8.19.2 Use of CPAP in acute pulmonary oedema

Two meta-analysis studies and one large RCT (Gray et al443) have provided evidence that CPAP administered with supplemental oxygen has definite measurable physiological and short-term clinical benefits in pulmonary oedema.441 ,442 The longer term benefits are less clear, in particular any reduction in mortality. The meta-analysis and a prehospital RCT looking at use of CPAP suggest reduced mortality and/or intubation rates.441 ,442 ,444 However, this mortality benefit was not found in the largest and best constructed RCT.443 In addition, this study did not report a reduction in the number of intubations.

Evidence statements

▸ Patients with cardiogenic pulmonary oedema, treated with CPAP and O₂ show early improvement in gas exchange when compared with standard treatment (evidence level 1+).

▸ Reports that short-term mortality and intubation rates are decreased were not confirmed in the largest RCT (evidence level 1+).

▸ The use of CPAP therapy in the prehospital environment may be beneficial to patients with acute pulmonary oedema as it can potentially decrease the need for endotracheal intubation, improve vital signs during transport to hospital and improve short-term mortality (evidence level 1+).

Recommendation

N2: CPAP with entrained oxygen to maintain saturation 94–98% should be considered as an adjunctive treatment to improve gas exchange in patients with cardiogenic pulmonary oedema who are not responding to standard treatment in hospital care or in prehospital care (grade B).

10.1.1 Cylinders (compressed gas)

Cylinders contain compressed gas held under a very high pressure. They come in an array of sizes and hence capacity, ranging from small portable cylinders for individual patient use to large cylinders suitable for hospital use (table 12). These can be used for bedside administration where piped oxygen is not available or can be the supply for a piped system.

With recent changes in technology, high-pressure cylinders are now available (ie, filled to 200 bar rather than 137 bar which can contain 54% more gas for the same size cylinder). It is important for all users of oxygen to be aware that most oxygen cylinders are colour-coded (black cylinder with white shoulder) but some high-pressure oxygen cylinders are all white. Small lightweight cylinders are also available for ambulatory use (eg, some weigh 3.2 kg when full). All systems containing compressed gases in the UK are subject to the Pressure Systems Safety Regulations 2000 (SI 2000 No 128). These regulations are intended to prevent the risk of injury from pressurised systems. Oxygen supports combustion and there is a risk of fire if oxygen is used close to combustible materials and a source of ignition. There is one case report of a serious fire caused by spontaneous sparking in the outflow system of an oxygen cylinder.452 Although oxygen is not an explosive gas, it is possible for oxygen cylinders to explode if heated during a fire, and deaths have been recorded in these circumstances.453 ,454

Healthcare organisations must ensure that they have a policy in place which ensures the safety of patients, staff and contractors in the provision, storage, use and maintenance of compressed gas systems as required by the Health and Safety at Work Act 1974.

Clinicians using oxygen cylinders should check the labelling of the cylinder to ensure that it is an oxygen cylinder and checks should be made to ensure that the cylinder is not empty or near empty.

10.1.2 Liquid oxygen

Liquid oxygen is contained in pressure tanks and is obtained from atmospheric oxygen by fractional distillation. It has to be evaporated into a gas before use. Large tanks are often used by hospitals and small tanks can be used domestically. Portable liquid oxygen is also available in small portable containers which can be filled from the larger tanks.

10.1.3 Oxygen concentrators

Oxygen concentrators are largely used in the domiciliary setting for the provision of long-term oxygen therapy and are therefore not used in the acute setting so will not be covered further.

10.2.1 High-concentration reservoir mask (non-rebreathing mask)

This type of mask delivers oxygen at concentrations between 60% and 90% when used at a flow rate of 15 L/min (figure 9).455 The delivered oxygen concentration is variable and will depend on the mask fit and the patient's breathing pattern. These masks are most suitable for trauma and emergency use in patients in whom carbon dioxide retention is unlikely.

• Download figure
• Open in new tab
• Download powerpoint

Figure 9

High-concentration reservoir mask (non-rebreathing mask).

10.2.2 Simple face mask

This type of mask delivers oxygen concentrations between 40% and 60% (figure 10). It is sometimes referred to as an MC Mask, Medium Concentration Mask, Mary Catterall Mask or as a ‘Hudson Mask’, but the latter description is discouraged because the Hudson Company make many types of mask (including high-concentration reservoir masks). The guideline group favours the term ‘simple face mask’. The oxygen supplied to the patient will be of variable concentration depending on the flow of oxygen and the patient's breathing pattern. The concentration can be changed by increasing or decreasing the oxygen flows between 5 and 10 L/min. However, different brands of simple face mask can deliver a different oxygen concentration at a given flow rate. Flows of <5 L/min can cause increased resistance to breathing, and there is a possibility of a build-up of carbon dioxide within the mask and rebreathing may occur.115

• Download figure
• Open in new tab
• Download powerpoint

Figure 10

Simple face mask.

This mask is suitable for patients with respiratory failure without hypercapnia (type 1 respiratory failure) but is not suitable for patients with hypercapnic (type 2) respiratory failure. The mask may deliver a high concentration of oxygen (>50%) and is therefore not recommended for patients who require low-concentration oxygen therapy because of the risk of carbon dioxide retention. Patients using a simple face mask may have an inspiratory flow rate greater than the gas flow rate from the mask, so the simple face mask should not be used at flow rates below 5 L/min.115 Several publications have shown that patients who require medium-concentration oxygen therapy tend to prefer nasal cannulae to simple face masks and the cannulae are more likely to be left in position by the patient and less likely to fall off.456–459

10.2.3 Venturi mask

A Venturi mask will give an accurate concentration of oxygen to the patient regardless of oxygen flow rate (the minimum suggested flow rate is written on each Venturi device and the available options are shown in table 13; figure 11A,B). The oxygen concentration remains constant because of the Venturi principle. The gas flow into the mask is diluted with air which is entrained via the cage on the Venturi adaptor. The amount of air sucked into the cage is related to the flow of oxygen into the Venturi system. The higher the flow the more air is sucked in. The proportions remain the same and therefore the Venturi mask delivers the same concentration of oxygen as the flow rate is increased.

Venturi masks are available in the following concentrations: 24%, 28%, 31%, 35%, 40% and 60%. They are suitable for all patients needing a known concentration of oxygen, but 24% and 28% Venturi masks are particularly suited to those at risk of carbon dioxide retention. A further benefit of Venturi masks is that the flow rate of gas from the mask will usually exceed the inspiratory flow rate of the patient. One study has shown that patients with a respiratory rate >30 breaths/min often have an inspiratory flow rate above the minimum flow rate specified on the mask packaging.303 Therefore, for patients with a high respiratory rate, it is suggested that the flow rate for Venturi masks should be set above the minimum flow rate listed on the packaging (increasing the oxygen flow rate into a Venturi mask does not increase the concentration of oxygen which is delivered). The accuracy of oxygen delivery from a Venturi mask is greatly reduced if the mask is not accurately placed on the patient's face.460

Patients with a respiratory rate >30 breaths/min often have a flow rate which is above the minimum delivered by the Venturi system as specified by the flow rate recommended for the mask. The flow may be increased as shown in figure 11B.

Venturi masks deliver a constant percentage of oxygen but the effect on the patient will depend on the condition being treated and on the breathing pattern and baseline oxygen saturation of the patient. As might be expected from the oxygen dissociation curve, patients with an oxygen saturation that is already in the normal range will have a very small rise in oxygen saturation (although the PaO₂ is likely to rise substantially). However, patients with very low oxygen saturation will have a marked rise if given even a small concentration of oxygen. This is because the oxygen dissociation curve is actually a ‘rapid escalator’ rather than a ‘slippery slope’.102 ,106 This is illustrated in figure 12 which demonstrates that a small increase in inspired oxygen concentration can make a big difference when the saturation is below 80% but the response to oxygen is much less if the saturation is close to or above 90%.

10.2.4 Nasal cannulae

Nasal cannulae can be used to deliver low-concentration and medium-concentration oxygen concentrations (figure 13). However, there is wide variation in patients' breathing patterns so the same flow rate of nasal oxygen may have widely different effects on the blood oxygen and carbon dioxide levels of different patients. Nasal cannulae at 1–4 L/min can have effects on oxygen saturation approximately equivalent to those seen with 24–40% oxygen from Venturi masks.230 ,463 The actual concentration of oxygen delivered (FiO₂) cannot be predicted and so cannot be used if a calculation of A–a gradient is required. The oxygen concentration continues to rise up to flows above 6 L/min. Some patients may experience discomfort and nasal dryness at flows above 4 L/min, but they can be well tolerated (see section 10.2.5).230 ,464 Although one might expect mouth breathing to reduce the efficiency of nasal cannulae, the majority of studies have shown that mouth breathing results in either the same inspired oxygen concentration or a higher concentration, especially when the respiratory rate is increased.464 This is important because patients with acute breathlessness are likely to breathe quickly and via the mouth rather than the nose. As there is marked individual variation in breathing pattern, the flow rate must be adjusted based on oximetry measurements and, where necessary, blood gas measurements. A cross-over comparison of nasal cannulae versus a Venturi mask (both adjusted to give satisfactory initial oxygen saturation) showed that the oxygen saturation of patients with exacerbated COPD fell below 90% for 5.4 hours/day during treatment with nasal cannulae compared with 3.7 hours/day during treatment with a Venturi mask.465 However, that study was published in 1999 before oximetery was widely used to monitor oxygen saturation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 13

Nasal cannulae.

The upper range of oxygen delivery from nasal cannulae is a little lower than the output of a simple face mask, but the lower range goes a lot lower than a simple face mask which should not be used below a flow rate of 5 L/min (about 40% oxygen).115 The performance and variation of nasal cannulae for medium-concentration oxygen therapy is broadly similar to that of the simple face mask, both in laboratory experiments230 ,464 and in clinical practice.457–459 ,466 ,467 One study suggested that the saturation was lower with nasal cannulae than with simple face masks in a subgroup of men following abdominal surgery.467 However, the use of target ranges would resolve this issue. Three patient preference studies comparing nasal cannulae with simple face masks in postoperative care found that patient preference was strongly in favour of nasal cannulae with up to 88% of patients preferring cannulae to masks.457–459 Another advantage of cannulae over simple face masks is that they are less likely to be removed accidentally and they allow the patient to speak and eat.458 ,468 There are no comparisons of these devices in acute care, but there is no reason to believe that the results would be any different for patients requiring medium-concentration oxygen therapy. Nasal cannula (not Venturi masks) were used in the Austin et al's51 study which showed improved survival from controlled oxygen therapy in AECOPD.

Advantages of nasal cannulae compared with simple face masks for medium-concentration oxygen therapy

• Comfort (but a minority of patients dislike the flow of oxygen into the nose, especially above 4 L/min).

• Adjustable flow gives wide oxygen concentration range (flow rate of 1–6 L/min gives FiO₂ from ∼24% to ∼50%), suitable for variable oxygen therapy and concentration titration.

• Patient preference.

• No claustrophobic sensation.

• Not taken off to eat or speak and less likely to fall off.

• Less affected by movement of face.

• Less inspiratory resistance than simple face masks.

• No risk of rebreathing of carbon dioxide.

• Cheaper.

Disadvantages of nasal cannulae

• May cause nasal irritation or soreness.

• May not work if nose is severely congested or blocked.

• Actual concentration of oxygen (FiO₂) cannot be predicted

10.2.5 High-flow humidified oxygen via nasal cannulae

There is increasing use of HFNC as an alternative delivery device for adults requiring medium-concentration and high-concentration oxygen therapy (figure 14). There is currently a small body of evidence to support its use as well as growing interest from positive clinical observations. The therapeutic effects are thought to be multifactorial and include delivery of increased FiO₂, a CPAP effect and greater comfort for the patient when compared with face masks. To deliver HFNC, three elements are required: a patient interface, a gas delivery device and a humidifier. However, specific research on the effect of humidification and high-flow oxygen therapy is lacking (see section 10.6.1). A flow of at least 60 L/min can be accommodated via standard dimension prongs on adult nasal cannula. Several manufacturers provide these, specifically designed for high-flow application. Ritchie et al469 reported a positive airway pressure of about 7 cm H₂O in normal volunteers at a flow of 50 L/min. Trials to date are few and small; however, Tiruvoipati et al470 performed a prospective randomised trial of 50 patients postextubation, who required high-flow oxygen therapy. This study showed HFNC to be as effective as high-flow face mask in delivering oxygen to extubated patients as well as a significant difference in tolerance of HFNC and a trend towards improved comfort in this group. A recent study by Hernández et al471 randomised 527 mechanically ventilated patients at low risk of reintubation to either high-flow nasal oxygen or conventional oxygen for 24 hours after extubation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 14

High-flow humidified nasal cannulae, flow generator and humidifier system.

They reported that the use of HFNC oxygen compared with conventional oxygen therapy reduced the risk of reintubation within 72 hours (4.9% vs 12.2%; p=0.04). A prospective sequential comparison trial by Roca et al472 examined 20 patients with acute hypoxaemic respiratory failure and compared HFNC with conventional oxygen therapy. HFNC was better tolerated and more comfortable than face mask. HFNC was associated with better oxygenation and lower respiratory rate. Cuquemelle et al473 also reported greater patient comfort using this system. In further support a study by Parke et al,474 60 patients with mild-to-moderate hypoxemic respiratory failure were randomised to receive HFNC or high-flow face mask and they analysed the success of allocated therapy, NIV rate and oxygenation. Significantly, more HFNC patients succeeded with their allocated therapy and the rate of NIV in this group was 10%, compared with 30% in the high-flow face mask group. The HFNC patients also had significantly fewer desaturations.

Frat et al475 randomised 310 patients with acute hypoxaemic respiratory failure (without hypercapnia) to treatment with HFNC, reservoir mask or NIV and they found that HFNC increased the number of ventilator-free days and reduced 90-day mortality but did not significantly reduce the intubation rate which was the primary end point of the study.

Larger and more comprehensive studies are still awaited but research to date is encouraging, indicating that HFNC may have an important role to play in the delivery of high-flow oxygen both in terms of physiological advantages and the tolerance and comfort of the individual patients who are not at risk of hypercapnia (evidence level 2− to 1−).

Good practice point

• High-flow nasal oxygen should be considered as an alternative to reservoir mask treatment in patients with acute respiratory failure without hypercapnia.

10.2.6 Tracheostomy mask

These devices are designed to allow oxygen to be given via a tracheostomy tube or to patients with previous laryngectomy (ie, neck breathing patients; figure 15). The oxygen flow rate should be adjusted to achieve saturation in accordance with tables 1–4 and chart 1 (figure 1). Oxygen given in this way for prolonged periods needs constant humidification and patients may need suction to remove mucus from the airway.

• Download figure
• Open in new tab
• Download powerpoint

Figure 15

Tracheostomy mask.

10.2.7 Non-invasive ventilation

This treatment option is beyond the scope of the present guideline. Readers are referred to the BTS guideline concerning the use of NIV in patients with exacerbations of COPD.209 ,299

10.3.1 Health and Safety Executive guidance for safe use of oxygen cylinders476,477

• All cylinders must be secured appropriately so they cannot move in transit (includes portable cylinders).

• No smoking in the vicinity of cylinders.

• Cylinders must be checked regularly for obvious signs of leakage.

• Cylinders must be kept out of direct sunlight.

• Green warning triangle ‘compressed gas’ should be displayed on the vehicle.

• Cylinders should never be lifted by the neck.

• They should only be changed by suitably trained personnel.

• Apart from portable cylinders, all cylinders should be moved using a cylinder trolley.

• Cylinders must be turned off/fully closed when not in use

• Cylinders must be in date. Staff should check the expiry date prior to use.

10.3.2 Oxygen use by UK ambulance services

Currently, within the UK, the ambulance service—whether NHS or private—has a range of vehicles and oxygen delivery systems at their disposal. There is an increasing use of cycle response units which tend to use the lightweight AZ or C sized cylinder with a capacity of 170 L. Motorcycle response units are generally equipped with the same AZ or C sized cylinders. Fast response units based on cars tend to be equipped with at least two of the lightweight CD sized cylinders which hold 460 L. The CD cylinder is also the size favoured by mountain, cave and mines rescue teams.

Front-line ambulances are usually equipped with piped oxygen fittings (Schraeder type) and supplied from two HX sized (2300 L) cylinders, as well as carrying at least two CD sized cylinders to power a portable oxygen-powered resuscitator. The piped supply has several outlet points placed in strategic positions to which are attached standard Schrader flow meters (0–15 L/min). This enables oxygen to be given throughout the patient's journey. The ambulance is also equipped with a portable supply which can be used at the site of an accident, taken into a patient's home or can be used when transferring a patient. They carry a range of patient interfaces for delivering the oxygen under the different circumstances encountered.

Portable resuscitators are always capable of supplying free-flow oxygen therapy as well as their resuscitator facility. Again, there are a variety of portable oxygen-powered resuscitators and it is beyond the scope of these guidelines to describe each one available for use in prehospital care. It is strongly suggested that those practitioners who need to work closely with the ambulance service should become familiar with the equipment used by their local ambulance service provider. With the possible exception of the cycle response units, all types of ambulance service response will have portable resuscitators, bag-valve mask devices and hand-operated suction as a minimum. Front-line emergency ambulances will also have rechargeable battery-powered suction available.

Patient transport service ambulances are equipped with an oxygen supply, normally an HX cylinder (2300 L) delivering the oxygen via a flow meter attached directly to the cylinder. Such vehicles also tend to carry basic hand-held suction devices. A range of oxygen masks are usually available for the different types of patients carried on such vehicles. Vehicles that are equipped with an oxygen supply should also carry oximeters to ensure appropriate use of oxygen (see section 9.8 for advice on which oxygen delivery devices should be carried in ambulances).

10.4.1 Oxygen carriage in private cars (Health and Safety Executive guidance)

When travelling by car, patients have the freedom to carry their own portable oxygen cylinder.476 ,477 Some GPs in rural areas also carry oxygen in their cars. However, it is advised that certain safety precautions should be followed:

• Patients are allowed to carry oxygen cylinders for their own use without putting any labels or signs on their vehicle. This includes public transport such as buses or taxis. It is an offence to display a hazard diamond if oxygen is not being transported in a vehicle so it is preferable not to use a hazard triangle on private vehicles.

• The cylinder should be secure within the car and cannot move during transport or in the event of an accident. The authors of this guideline are aware of one fatality caused by a patient smoking in a vehicle where an unsecured liquid oxygen cylinder spilt into the floor-well during sudden braking.

• Patients should inform their car insurance company if oxygen is carried in the car.

10.4.2 Medical centres and primary care practices

The majority of medical centres and practices should have a supply of oxygen for emergency use. Generally, cylinders with integral high-flow regulators should be ordered. Otherwise, the cylinder must be fitted with a high-flow regulator capable of delivering a flow of up to 15 L/min in order to deliver medium-concentration and high-concentration oxygen therapy. A recommended list of oxygen delivery devices for use in prehospital care is given in section 9.8.

• Emergency oxygen should be available in primary care medical centres, preferably using oxygen cylinders with integral high-flow regulators. Alternatively, oxygen cylinders fitted with high-flow regulators (delivering up to 15 L/min) must be used to allow use with reservoir masks (see recommendation V1).

10.4.3 Emergency use of oxygen in the patient's home

In patients' homes, oxygen is usually provided for long-term therapy with an oxygen concentrator and an ambulatory supply with lightweight cylinders (or a portable liquid oxygen system). Long-term oxygen therapy is covered in other guidelines. In some circumstances, there may be a supply of cylinders for short-term/short-burst therapy or palliative use. The existing home oxygen supply may be used by a patient or GP in an emergency situation before the arrival of an ambulance using the patient's existing interface. If a GP is attending a patient at home with oxygen, ideally, the use of oxygen should be guided by pulse oximetry. The existence of any oxygen alert card should be asked for so that the emergency services attending can be aware of the target saturation and oxygen supply titrated accordingly.

The patient/carers should be made aware of the following Health and Safety recommendations:476 ,477

• All cylinders should be stored on a cylinder trolley or suitably secured so they cannot be knocked over.

• There should be no trailing oxygen tubing.

• A green warning triangle for ‘compressed gas’ should be displayed by the front door (warns emergency services in the event of a fire).

• The minimum number of cylinders should be stored in the house.

• There should be no smoking in the vicinity of oxygen cylinders.

• Cylinders must be checked regularly for obvious signs of leakage.

• Cylinders must be kept out of direct sunlight.

• Oxygen must not be used near a naked flame or source of heat.

10.5.1 Perioperative and postoperative care

Medium-concentration masks and nasal cannulae are usually sufficient (target saturation 94–98%) except for patients with known significant COPD who should receive oxygen from a 24% or 28% Venturi mask or 1–2 L/min from nasal cannulae aiming at a saturation range of 88–92%.

10.5.2 Emergency departments

Medium-concentration or high-concentration oxygen is normally used (via nasal cannulae, simple face mask or reservoir mask), but particular attention should be given to patients who have hypercapnic respiratory failure when a 24% or 28% Venturi mask or nasal cannulae at a flow rate of 1–2 L/min would be appropriate.

10.5.3 General wards and respiratory wards

The method of oxygen delivery will depend on the following circumstances:

• Expected duration of treatment.

• Type of respiratory illness.

• Pattern of breathing (high or low respiratory rate and drive).

• Need for humidification.

• Risk of carbon dioxide retention.

• Presence of confusion and its effect on potential compliance.

Nasal cannulae, simple face masks, reservoir masks and Venturi masks should be used where appropriate (see tables 1–4 and charts 1 and 2 in the main guideline). Nasal cannulae at flow rates or 0.5–1.0 L/min are sometimes used as a substitute for Venturi masks in acute or postacute patients with COPD on respiratory wards (adjusting flow as necessary to achieve the desired ABG tensions). This practice requires the use of paediatric flow meters to ensure consistent and finely calibrated oxygen delivery and is not recommended outside specialist units.

10.5.4 Devices used in emergency oxygen therapy

Based on the advantages of each delivery system discussed above, the following recommendations are made for delivery of oxygen in medical emergencies. It is likely that additional equipment will be maintained in specialist units, but specialist treatment is outside the scope of the present guideline.

Good practice points (see tables 1–4)

• ✓ Most hospital patients can be managed with the same delivery device, as in good practice points which follow section 9.8, but a 24% Venturi mask should also be available.

  • The high-concentration reservoir mask at 15 L/min is the preferred means for delivering high-concentration oxygen to critically ill patients until reliable pulse oximetry monitoring has been established.

  • Nasal cannulae should be used rather than simple face masks in most situations requiring medium-concentration oxygen therapy. Nasal cannulae are preferred by patients for reasons of comfort and they are less likely to be removed during meals. There is also a cost-saving.

  • The flow rate from nasal cannulae for medium-concentration oxygen therapy should be adjusted between 2 and 6 L/min to achieve the desired target saturation.

  • Venturi masks are recommended for patients requiring precise control of FiO₂. Venturi masks can deliver a constant FiO₂ of 24%, 28%, 31%, 35%, 40% and 60% oxygen with a greater gas flow than a simple face mask. In those at risk of developing hypercapnic respiratory failure with oxygen therapy, the use of Venturi masks may reduce this risk. Furthermore, there is less likelihood of dilution of the oxygen stream by room air if the patient's inspiratory flow rate exceeds the flow rate delivered by the face mask.

• For many patients, 24–28% Venturi masks can be substituted with nasal cannulae at low-flow rates (1–2 L/min) to achieve the same target range.

• The flow rate from simple face masks should be adjusted between 5 and 10 L/min to achieve the desired target saturation. Flow rates below 5 L/min may cause carbon dioxide rebreathing and increased resistance to inspiration.

• Patients with COPD with a respiratory rate of >30 breaths/min should have the flow rate set to above the minimum flow rate specified for the Venturi mask and/or packaging (see figure 11B) (increasing the oxygen flow rate into a Venturi mask increases the total gas flow from the mask but does not increase the concentration of oxygen which is delivered).

• HFNC are well tolerated and may be used as an alternative in hypoxic adult patients requiring medium-concentration to high-concentration oxygen therapy and who are not at risk of hypercapnia.

10.5.5 Flow meters

All oxygen delivery systems must have a method of taking the high pressure/flow of gas and reducing it so it can be administered to the patient at a specific flow depending on the individual's needs and mask being used. Piped oxygen points have Schrader flow meters and cylinders have pressure and flow regulators. Most oxygen flow meters use a floating ball to indicate the flow rate. The centre of the ball should be aligned with the appropriate flow rate marking. The example shown in figure 16 indicates the correct setting to deliver 2 L/min.

• Download figure
• Open in new tab
• Download powerpoint

Figure 16

Flow meter showing correct setting to deliver oxygen at a rate of 2 L/min.

10.5.6 Oxygen tubing and oxygen wall outlets

Oxygen tubing is needed to connect flow meters and regulators to the patient delivery device. It is important to ensure that all tubing is connected correctly at both ends. The National Patient Safety Agency has reported frequent adverse events related to oxygen use, including four reports of instances where an oxygen mask was connected in error to a compressed air outlet instead of an oxygen outlet. Compressed air outlets are often used to drive nebulisers in hospitals because they are quieter than electrical compressors. However, the flow meter looks very similar to an oxygen flow meter and is often mounted beside an oxygen flow meter so it is very important to ensure that air flow meters are clearly labelled. There is a similar risk with other piped gas outlets such as those delivering nitrous oxide in some hospitals. Air flow meters are never required in an emergency and should be removed from wall sockets or covered by a designated ‘hood’ when not in use, an example is shown in figure 17.

• Download figure
• Open in new tab
• Download powerpoint

Figure 17

Air outlet cover.

The guideline authors are also aware of some cases where twin oxygen outlets were in use and the wrong one had been turned on or off. For example, one patient tried to turn off the oxygen flow after finishing a nebulised treatment but accidentally turned off the oxygen flow to a neighbouring patient with serious consequences. It is recommended that patients should not be allowed to adjust oxygen flow, especially if there are dual outlets.

Healthcare organisations should take measures to eliminate the risk of oxygen tubing being connected to the incorrect wall oxygen outlet or to outlets that deliver compressed air or other gases instead of oxygen. Air flow meters should be removed from the wall sockets or covered with a designated air outlet cover when not in use. Special care should be taken if twin oxygen outlets are in use (see recommendation V2).

Evidence statement

• Matters related to oxygen tubing and equipment and wall outlets are based on expert opinion and manufacturers' advice (evidence level 4).

Recommendation

V2: Healthcare organisations should take measures to eliminate the risk of oxygen tubing being connected to the incorrect wall oxygen outlet or to outlets that deliver compressed air or other gases instead of oxygen. Air flow meters should be removed from the wall sockets or covered with a designated air outlet cover when not in use. Special care should be taken if twin oxygen outlets are in use (grade D).

10.6.1 Rationale for use of humidified oxygen

The upper airway normally warms, moistens and filters inspired gases. When these functions are impaired by a pathological process or when they are bypassed by an artificial airway, it is common practice to provide humidification. The main reason for using humidification, especially with high-flow oxygen, is that it may reduce the sensation of dryness in the upper airways that oxygen can cause. However, in the non-intubated population, there appears to be little scientific evidence of any benefit from humidified oxygen except that single doses of nebulised isotonic saline have been shown to assist sputum clearance and reduce breathlessness in patients with COPD.478 ,479 There is also evidence that humidification, when combined with physiotherapy, can increase sputum clearance in bronchiectasis.480 RCTs of the effects of humidified high-flow oxygen on patient comfort are required. Use of low-flow oxygen through nasal cannulae is well tolerated without the need for humidification.481 HFNC require warmed and humidified oxygen.

Evidence statement

• The use of humidified oxygen in acute care is based on expert opinion (evidence level 4).

Recommendations

Q1: Humidification is not required for the delivery of low-flow oxygen (mask or nasal cannulae) or for the short-term use of high-flow oxygen. It is not therefore required in prehospital care. Pending the results of clinical trials, it is reasonable to use humidified oxygen for patients who require high-flow oxygen systems for more than 24 hours or who report upper airway discomfort due to dryness (grade D).

Q2: In the emergency situation, humidified oxygen use can be confined to patients with tracheostomy or an artificial airway although these patients can be managed without humidification for short periods of time (eg, ambulance journeys) (grade D).

Q3: Humidification may also be of benefit to patients with viscous secretions causing difficulty with expectoration. This benefit can be achieved using nebulised normal saline (grade D).

10.6.2 Use of bubble humidification systems

Humidified oxygen is widely administered in hospitals across the UK and this is presumed to alleviate nasal and oral discomfort in the non-intubated patient. Humidification of supplemental oxygen was commonly delivered by bubbling oxygen through either cold or warm sterile water before it reached the patient. However, the effect on patient comfort is negligible.481 ,482 Bubble humidifiers may, however, represent an infection hazard and should not be used.483 There is no evidence of a clinically significant benefit from ‘bubble bottle’ systems but there is an infection risk.

Evidence statement

• The advice not to use bubble bottles is based on expert opinion (evidence level 4).

Recommendation

Q4: Bubble bottles which allow a stream of oxygen to bubble through a container of water should not be used because there is no evidence of a clinically significant benefit but there is a risk of infection (grade D).

10.6.3 Large volume nebulisation-based humidifiers

If humidification is required, it should ideally deliver the inspired gas at a temperature of 32–36° C. Cold water humidifiers are simple and inexpensive but less efficient than a warm water systems (about 50% relative humidity at ambient temperatures). The warm water option is more effective, targeting a relative humidity of 100%, but both systems are thought to be a potential infection control risk. Warm water humidifiers are expensive and mostly confined to ICUs and HDUs and thus outside the scope of this guideline.

Newer humidifying systems are really ‘giant nebulisers’ with a 1 L reservoir of saline or sterile water and an adjustable Venturi device (figure 18). These systems are attached directly to the oxygen flow meter and connected to an aerosol mask via Flex tube. They allow delivery of precise oxygen concentrations of 28%, 35%, 40% and 60% oxygen via their Venturi device. This requires a specific oxygen flow rate as well as adjusting the Venturi nozzle on the device. It is possible to deliver 24% oxygen using a special adaptor. These large volume humidifiers have a high humidification output. The main indication for use is to assist with expectoration of viscous sputum. There are no published randomised studies involving these devices, but it has been shown that single doses of nebulised saline can assist sputum production and relieve breathlessness in patients with COPD.478 ,479

• Download figure
• Open in new tab
• Download powerpoint

Figure 18

Large volume nebulisation-based humidifier.

Good practice points related to humidified oxygen therapy

• Consider use of a large volume oxygen humidifier device for patients requiring high-flow rates or longer term oxygen, especially if sputum retention is a clinical problem.

• In the absence of an artificial airway, the decision to humidify supplemental oxygen needs to be made on an individual basis but this practice is not evidence-based.

10.8.1 Nebulised bronchodilator therapy in asthma

In patients with acute severe asthma, oxygen should be used as the driving gas for the nebulised bronchodilators whenever possible at a gas flow rate of 6–8 L/min because these patients are at risk of hypoxaemia. If the available cylinders in general practice do not produce this flow rate, an air-driven nebuliser (with electrical compressor) should be used with supplemental oxygen by nasal cannulae at 2–6 L/min to maintain an appropriate oxygen saturation level. There is some evidence that bronchodilator therapy (whether delivered by nebuliser or by metered dose inhaler) may cause pulmonary vasodilation leading to increased V/Q mismatch and reduced blood oxygen levels in acutely ill patients and in patients with stable COPD and asthma.484–486 The lowest oxygen saturation occurs not during bronchodilator therapy but about 25 min later.486 One group reported a rise in oxygen saturation during bronchodilator therapy.487

• There is some evidence that bronchodilator therapy, however given, can cause increased V/Q mismatch and reduced blood oxygen levels in acutely ill patients shortly after treatment.

10.8.2 Nebulised bronchodilator therapy for patients with COPD and other risk factors for hypercapnic respiratory failure

When an oxygen-driven nebuliser is given to patients with COPD there is a risk of hypercapnia and acidosis due to the high FiO₂ which is delivered. In AECOPD the carbon dioxide level can rise substantially within 15 min of starting high-concentration oxygen therapy.48 Edwards et al488 have confirmed that oxygen-driven nebulisers cause elevation of the blood CO₂ level in patients with COPD. The Austin et al's51 study which demonstrated improved mortality with targeted oxygen therapy in patients with AECOPD used air-driven nebulisers to ensure that the target saturation was maintained. When nebulised bronchodilators are given to hypercapnic patients, they should ideally be given using an electrical compressor or ultrasonic nebuliser and, if necessary, supplementary oxygen should be given concurrently by using nasal cannulae at 1–4 L/min to maintain an oxygen saturation of 88–92%. For hypercapnic patients or those with a history of hypercapnia, oxygen saturation should be monitored continuously during nebulised treatments. If an electrical compressor or ultrasonic nebuliser is not available, compressed air can be used to drive the nebuliser but care should be taken to ensure that the nebuliser is not attached to an oxygen outlet (see section 10.8). Once the nebulised treatment is completed, targeted oxygen therapy should be reinstituted.

Compressed air is not routinely available for treatment by ambulance staff. In this situation, oxygen-driven nebulisers may be used but should be limited to 6 min for patients with known COPD because the nebuliser mask delivers a high concentration of oxygen (about 60%). Since most of the effective dose from a nebuliser chamber is delivered within 6 min, limiting oxygen-driven nebulisers to 6 min will minimise the risk of hypercapnic respiratory failure while delivering most of the nebulised drug dose. It must be recognised that patients with COPD may develop hypercapnia and respiratory acidosis if they are left on high-flow oxygen via a nebuliser mask system long after the nebulisation process has finished.

Evidence statement

• High-concentration oxygen therapy can cause hypercapnia within 15 min in acute COPD (evidence level 4).

• The use of oxygen to drive nebulisers is based on expert opinion and extrapolation from observational studies (evidence level 4).

Recommendations

R1: For patients with asthma, nebulisers should be driven by piped oxygen or from an oxygen cylinder fitted with a high-flow regulator capable of delivering a flow rate of >6 L/min. The patient should be changed back to his/her usual oxygen mask or cannulae when nebuliser therapy is complete. If the cylinder does not produce this flow rate, an air-driven nebuliser (with electrical compressor) should be used with supplemental oxygen by nasal cannulae at 2–6 L/min to maintain an appropriate oxygen saturation level (grade D).

R2: When nebulised bronchodilators are given to patients with hypercapnic acidosis, they should be given using an ultrasonic nebuliser or else a jet nebuliser driven by compressed air and, if necessary, supplementary oxygen should be given concurrently by nasal cannulae to maintain an oxygen saturation of 88–92%. The same precautions should be applied to patients who are at risk of hypercapnic respiratory failure prior to the availability of blood gas results and the oxygen saturation should be monitored continuously during treatment. Once the nebulised treatment is completed for patients at risk of hypercapnic respiratory failure, their previous targeted oxygen therapy should be reinstituted (grade D).

Good practice points

Do not allow hypoxaemia to occur while administering nebulised treatments:

• For hypoxaemic patients, oxygen therapy should continue during nebulised treatments.

Driving gas for nebulised treatment in ambulances:

• During treatment by ambulance staff oxygen-driven nebulisers should be used for patients with asthma and may be used for patients with COPD in the absence of an air-driven compressor system. If oxygen is used for patients with known COPD, its use should be limited to 6 min. This will deliver most of the nebulised drug dose but limit the risk of hypercapnic respiratory failure (section 10.4). Ambulance services are encouraged to explore the feasibility of introducing battery-powered, air-driven nebulisers or portable ultrasonic nebulisers.

11.1.1 Legal status of medical oxygen: Does it need a prescription?

Medical oxygen is a drug. However, the legal status of oxygen in the UK is that of a medicinal product on the General Sales List (Medicines and Healthcare Products Regulatory Agency, personal communication). This means that the sale or dispensation of oxygen does not technically require a prescription because it is not a ‘prescription-only medicine’. This status was conferred for practical reasons to facilitate the use of oxygen in the domiciliary setting where the distribution system no longer involves pharmacies. However, the use of oxygen by paramedics, registered nurses, doctors, physiotherapists and others in emergency situations is similar to the use of all other medicinal products by these people. Clinical governance requires that the intentions of the clinician who initiates oxygen therapy should be communicated clearly to the person who actually administers oxygen to the patient and an accurate record must be kept of exactly what has been given to the patient. In this respect, oxygen is in the same category as paracetamol, aspirin, ibuprofen, antihistamines, antiemetics and many other medicines that do not require a prescription if purchased by a patient for his/her own use but do require accurate documentation if administered by a health professional to a patient. In healthcare settings, all of these medicines are conventionally recorded on a ‘prescription chart’ or ‘drug kardex’ alongside drugs in the ‘prescription-only’ category such as antibiotics.

A hospital ‘prescription chart’ is a document in which doctors and other prescribing clinicians make a list of all drugs and medicinal products required by a patient and where registered nurses and other clinicians make a record of all drugs that have been administered to the patient. In many hospitals, the prescription and documentation of administration are recorded directly into the Electronic Patient Record. These documents are also used by hospital pharmacies to dispense medicines but, for technical reasons, these hospital documents (or the equivalent documents in primary care settings or the Patient Report Form (PRF) or electronic PRF used by the ambulance services) do not have the same legal status as a prescription that is given to a patient to take to a community pharmacist for the purposes of dispensing a medicine. The ‘prescription’ of oxygen in acute healthcare settings is therefore not technically a prescription for two reasons: (1) oxygen is not a prescription-only medicine; and (2) a hospital drug chart is not truly a prescription. The second issue applies to all medicines used in healthcare settings, not just to oxygen. However, the words ‘prescription’ or ‘prescription chart’ are widely used to describe the documents which record the use of medicines in UK hospitals and other healthcare settings. For this reason, the present guideline will use the term ‘prescribe’ and ‘prescription chart’ for all orders for oxygen use. The important issue is that there must be a clear written record of all medicinal products, including oxygen, which are given to a patient by healthcare professionals. Ideally, this document should be prepared at the time when oxygen therapy is started. In emergencies, however, clinicians will treat the patient first and subsequently make written records of all treatments given, including oxygen therapy.

11.1.2 Reasons to prescribe oxygen therapy

Oxygen is prescribed for hypoxaemic patients to increase PO₂ and decrease the work of breathing necessary to maintain a given PaO₂ (see section 6). The concentration of oxygen required depends on the condition being treated; an inappropriate concentration may have serious or even lethal effects. Appropriate target saturation ranges for common medical emergencies are given in section 8 of this guideline and in tables 1–4. Each hospital should have an agreed policy and protocol for oxygen prescribing to allow staff to adjust oxygen delivery devices and to give oxygen in emergency situations prior to the availability of a prescription. The use of a patient group direction (PGD) may be appropriate if other solutions have not worked in the clinical setting. Please see details at: https://www.nice.org.uk/guidance/MPG2 and http://www.medicinesresources.nhs.uk/en/Communities/NHS/PGDs/.

Although a PGD is not strictly required for medical gases (because they are not a prescription-only medicine) it is important for every organisation to have a written policy to regulate oxygen use up to time when a prescription is written, for example, for patients arriving by ambulance in the emergency department with oxygen use already initiated by the ambulance service. For some organisations, a PGD may be considered as the best way to address this important issue. BTS audits up to 2015 have shown that only 57% of patients using supplemental oxygen in UK hospitals had a valid prescription or written order. This is a situation that would not be tolerated for any other drug.

11.1.3 Implementing an oxygen prescription policy

Oxygen prescriptions may include initial mode of delivery and flow rate (eg, 4 L/min via nasal cannulae). But the most important aspect of the prescription is to give a target range. The clinicians who administer oxygen (usually registered nurses, midwives or physiotherapists) should be trained and empowered to adjust the oxygen concentration upwards and downwards as necessary to maintain the patient in the target saturation range (including adjustments made during nebulised therapy driven by oxygen or nebulised therapy driven by air with simultaneous administration of oxygen by nasal cannulae). This will require all hospitals to have an agreed oxygen administration programme with universal access to educational materials about oxygen administration. The clinicians who monitor the oxygen saturation (often healthcare assistants) should be trained to inform those who have been trained to administer oxygen if the oxygen saturations fall above or below the target saturations. Those doing the monitoring should also understand the importance for the patient of keeping in the target range. Implementing this policy will require all organisations to have agreed procedures (which may include PGDs up to the time when a prescription is written) and training programmes for all clinical staff and regular training programmes in the safe use of oxygen and audit of outcomes.

Many hospitals have developed an ‘admission bundle’ for all patients admitted to hospital. This bundle may include checking for hospital-associated thrombosis risk, checking MRSA risk and the prescription of antithrombotic agents and anti-MRSA agents if appropriate. As part of this process, many hospitals now specify that an oxygen saturation target must be set for all patients on admission to hospital so that the nursing team will know the correct starting target range if oxygen therapy should suddenly become necessary (as well as making a callout to the medical team to assess the patient). Furthermore, this allows nurses making bedside observations to know if the oxygen saturation is appropriate for each patient whether or not the patient is receiving oxygen at the time when the observation is made. Some older people without respiratory problems may have oxygen saturation below 94% and the target range for such patients should be adjusted accordingly. It was reported in the 2011 BTS national emergency oxygen audit that almost half of UK hospitals now have a policy of setting a target saturation range for all patients at the time of admission to hospital. Policies of this sort will become even more important when the EWS system takes oxygen saturation (above or below target range) into account.16

11.1.4 Administration and monitoring of oxygen therapy

The appropriate device should be used to provide the prescribed oxygen and the effects should be monitored using pulse oximetry, monitoring of respiratory rate and close observation of the patient. Arterial or capillary blood gas analysis should be repeated if clinical progress is not satisfactory and in all cases of hypercapnia and acidosis.

11.1.5 Education of health professionals

The clinician or registered healthcare professional administering the oxygen therapy should be aware of the hazards of hypoxaemia and hyperoxaemia and the signs and symptoms of inadequate or excessive oxygen delivery.

11.1.6 How to prescribe oxygen effectively

In the past, oxygen was often not prescribed at all or prescribed on a standard hospital drug chart as ‘oxygen’. It was unusual for the prescription to include full details of what device to use, what flow rate(s) to administer and whether the prescription was for a fixed concentration of oxygen or to aim at a specific oxygen saturation target.4–10

It has been shown that a purpose-designed oxygen prescription sheet can improve oxygen prescribing in the short term, but experience has shown that free-standing oxygen prescription charts are often forgotten and unused.7 Recent audit studies by members of the guideline group (part of annual BTS hospital oxygen audits) have shown improved standards of prescribing with the use of a preprinted section for oxygen use in all hospital drug charts and with prompted electronic prescribing of oxygen for all hospital patients. This system was further enhanced by setting a desired saturation range for each patient. Suggested target saturations for common medical conditions are given in sections 8 and 9 of this guideline. It is important that healthcare professionals where oxygen is administered are familiar with the optimal saturation ranges for common conditions (see online appendix 3), and it is also important that those delivering the oxygen are familiar with the equipment in use and the best types of device to deliver low-concentration, medium-concentration and high-concentration oxygen therapy. Figure 19 shows a working example of a preprinted oxygen section for a hospital prescription chart, and figures 1 and 2 (charts 1 and 2) give advice to prescribers and advice to those delivering oxygen onwards.

The safe use of oxygen includes careful consideration of the appropriate delivery device (mask, cannulae, etc) together with an appropriate source of oxygen and an appropriate oxygen flow rate.

An oxygen target range should be prescribed for all hospital patients. The method and rate of oxygen delivery should be altered by registered nurses or other registered healthcare professionals in order to achieve the prescribed target saturation as per the hospital policy. For most conditions, oxygen should be prescribed to achieve a target saturation of 94–98% (or 88–92% for those at risk of hypercapnic respiratory failure). The nurse should sign this prescription chart or the electronic equivalent on every drug round. This signature on the drug chart confirms that the patient is either receiving appropriate oxygen therapy (the delivery device and flow rate should be documented alongside the SpO₂ on the bedside observations chart or electronic observations system) or else has SpO₂ within the target range on air at the most recent observation round. It is not necessary to measure oximetry during drug rounds but action may be required if the patient was documented to be outside of their target range at the most recent observations round and if no action has yet taken place.

Evidence statement

• Prescription of oxygen in acute illness is based on expert opinion (evidence level 4).

Recommendations for safe prescribing and safe administration of oxygen

S1: Every healthcare facility should have a standard oxygen prescription document or, preferably, a designated oxygen section on all drug-prescribing cards or guided prescription of oxygen in electronic prescribing systems (grade D).

S2: A prescription for oxygen should always be provided, except in sudden illness when it must be started immediately and documented retrospectively (grade D).

S3: Doctors and other prescribers should prescribe oxygen using a target saturation range (sections 8, 9 and 11) and sign the drug chart or electronic prescribing order (grade D).

S4: An oxygen target saturation range should be prescribed for all patients who are admitted to hospital. This will ensure that every patient will receive appropriate oxygen therapy if it should be required. It will also ensure that all clinicians are aware of the appropriate oxygen target range for every patient under their care (grade D).

Good practice points related to prescribing and administering oxygen therapy to patients

• Oxygen should be prescribed on the drug chart or electronic prescribing system using a target saturation range.

• Oxygen should be prescribed to a target saturation range rather than prescribing a fixed concentration of oxygen or FiO₂ (see recommendations A1, A2, A4 and A5).

• For hypoxaemic patients, oxygen therapy should continue during other treatments such as nebulised therapy. Clinicians should assess the clinical status of the patient prior to prescribing oxygen and the patient's condition should be reassessed frequently during oxygen use (see recommendations B1–B3).

• In most emergency situations, oxygen is given to patients immediately without a formal prescription. The lack of a prescription should never preclude oxygen being given when needed in an emergency situation. However, a subsequent written record must be made of what oxygen therapy has been given to every patient in a similar manner to the recording of all other emergency treatment.

• If a patient has an oxygen alert card, initial oxygen therapy should be based on the guidance on the card until the results of blood gases are available.

Monitoring and adjusting therapy (see sections 9–11)

T1: Pulse oximetry must be available in all locations where emergency oxygen is being used by healthcare professionals (see also the limitations of using pulse oximetry section 7.1.2) (grade D).

T2: All documents which record oximetry measurements or blood gas results should state whether the patient is breathing air or a specified oxygen delivery device and flow rate using the abbreviations shown in table 5 (grade D).

T3: In all situations where repeated blood gas measurements are required, they should be measured as soon as possible, usually within 30 min of any treatment change, to determine if the proposed target saturations are appropriate. Consider the use of an indwelling arterial catheter if multiple samples are likely to be required (grade D).

T4: Adjustments should only be made by registered staff who have been trained to administer oxygen. If the oxygen saturation falls below the prespecified range, the concentration of oxygen should be increased; if the saturation rises above this range, the oxygen concentration should be reduced. If the monitoring of oxygen saturation is performed by unregistered staff (eg, healthcare assistants), there must be a clear protocol in place which requires that they should inform staff who are trained to administer oxygen if the oxygen saturation is above or below the target saturation (grade D).

Good practice points related to administration of oxygen therapy

• For hypoxaemic patients, oxygen therapy should continue during other treatments such as nebulised therapy. Clinicians should assess the clinical status of the patient prior to prescribing oxygen and the patient's condition should be reassessed frequently during oxygen use (see recommendations B1–B3).

• The administering healthcare professional should note the oxygen saturation before stating oxygen therapy whenever possible but never discontinue or delay oxygen therapy for seriously ill patients (see recommendation B2).

• The healthcare professional should start oxygen therapy using an appropriate delivery system and flow rate as specified in sections 8– 10 of this guideline. The target oxygen saturation should be documented on the respiratory section of the observation chart.

• Whenever possible, patients should be given an oxygen information sheet (example in web appendix 6 of this guideline on the BTS website).

• Staff should check the oxygen supply and connections on a regular basis because there have been serious incidents due to disconnection or misconnection of oxygen supplies.

• Staff must ensure that adequate oxygen is provided during transfers and while patients are in diagnostic departments. Additionally, oxygen saturation should be monitored continuously for seriously ill patients who require escorted transfers. This is because there have been serious incidents involving accidental discontinuation of oxygen or cylinders running out during interward transfers or transfers to other departments such as for x-rays.

Practical aspects of oxygen dispensing, documentation and monitoring

W1: Registered nurses and others who dispense drugs in hospitals should sign the drug chart or electronic prescribing record at every drug round and check that the patient is receiving oxygen therapy. This is to check that the patient is within the target saturation and also to check whether weaning and discontinuation should be instituted (grade D).

W2: Most patients are prescribed an oxygen target range. If patients are on air at the time of the drug round, registered nurses should sign the drug chart using a code such as ‘A’ for air and the observation chart should also be filled in using the code A for air (see table 5 and figure 19; grade D).

W3: All patients should have their oxygen saturation observed for at least 5 min after starting oxygen therapy or for patients who require an increased concentration of oxygen and after oxygen therapy has been decreased or stopped (grade D).

W4: If the oxygen saturation is above the target saturation range and the patient is stable, the delivery system or oxygen flow rate should be modified to return the saturation to within the target range (grade D).

W5: Patients who have a target saturation of 88–92% should have their blood gases measured within 30–60 min. This is to ensure that the carbon dioxide level is not rising. This recommendation also applies to those who are at risk of developing hypercapnic respiratory failure but who have a normal PCO₂ on the initial blood gas measurement (grade D).

W6: Stable patients whose oxygen saturation is within their target saturation range of 94–98% do not need repeat blood gas measurements within 30–60 min if there is no risk of hypercapnic respiratory failure and acidosis and may not need any further blood gas measurements unless there should be further deterioration including symptoms or signs of possible hypercapnia (grade D).

W7: Stable patients on oxygen treatment should have SpO₂ and physiological variables (eg, NEWS) measured four times a day (grade D).

W8: In those who have signs of critical illness (eg, NEWS 7 or above), oxygen saturation should be monitored continuously and the patient may require level 2 or 3 care on a HDU or critical care unit (grade D).

W9: If the patient is clinically stable and the oxygen saturation is within the target range, treatment should be continued (or eventually lowered) depending on the clinical situation (grade D).

W10: Oxygen therapy should be increased if the saturation is below the desired range and decreased if the saturation is above the desired range (and eventually discontinued as the patient recovers) (grade D).

W11: The new saturation (and the new delivery system) and flow rate should be recorded on the patient's observation chart after 5 min of treatment at the new oxygen concentration. Each change should be recorded by the clinician trained to administer oxygen by signing the observation chart (only changes should be signed for) (grade D).

W12: Repeat blood gas measurements are not required for stable patients who require a reduced concentration of oxygen (or cessation of oxygen therapy) to maintain the desired target saturation (grade D).

W13: Patients with no risk of hypercapnic respiratory failure do not always need repeat blood gas measurements after an increase in oxygen concentration. However, the patient requires clinical review to determine why the oxygen saturation has fallen (grade D).

W14: Patients at risk of hypercapnic respiratory failure (usually those with a target range of 88–92%; see table 4) require repeat blood gas assessment 30–60 min after an increase in oxygen therapy (to ensure that the carbon dioxide level is not rising) (grade D).

W15: For patients with no risk of hypercapnic respiratory failure, monitoring by pulse oximeter is sufficient (repeated blood gases not required) provided the patient is clinically stable and the oxygen saturation remains in the desired range, usually 94–98% (grade D).

W16: If a patient's oxygen saturation is lower than the prescribed target range, first check all aspects of the oxygen delivery system and the oximeter device for faults or errors (grade D).

W17: If a patient's oxygen saturation is consistently lower than the prescribed target range, there should be a medical review and the oxygen therapy should be increased according to an agreed written protocol (grade D).

W18: If the oxygen saturation fails to rise following 5–10 min of increased oxygen therapy or if there is clinical concern following medical review, then blood gas measurements should be repeated (grade D).

Training in oxygen prescribing and use

X1: All clinicians prescribing oxygen should have appropriate training and access to written or electronic oxygen prescribing guidelines based on this national guideline (grade D).

(Training slides for doctors and nurses are available as online supplementary appendices 7 and 8 on the BTS website.)

X2: Every hospital should have a training programme to ensure that clinical staff are familiar with the hospital's oxygen administration policies. In view of the high number of adverse incidents related to oxygen therapy, it is recommended that all acute Trusts should include basic training in oxygen use in the mandatory training programmes for all clinical staff (grade D).

11.3.1 Pulse oximeters

Pulse oximetry should be available to all healthcare staff managing patients receiving oxygen therapy and they should be trained in their use (see section 7 for technical and practical information regarding oximeter use). Clinicians should be aware that pulse oximetry gives no information about the PCO₂ or pH and most pulse oximeters are unreliable when a patient's SpO₂ falls below about 85%. Pulse oximetry is dependent on pulsatile flow, and it may not be possible to achieve a satisfactory oximeter reading in patients with cold hands, especially those with severe Raynaud's phenomenon due to collagen vascular diseases (which may also cause hypoxic lung disease). The readings may also be affected by shock, skin pigmentation, nail varnish, etc (see section 7). It is essential to record the oxygen delivery system alongside the oximetry result.

All measurements of oxygen saturation should be recorded in the observation chart along with the code for the oxygen delivery system that is being used. If the patient is breathing air at the time of the observation, this should also be recorded in the chart (see recommendation B2 and figure 19).

11.3.2 Arterial or arterialised capillary blood gases

• Arterial or arterialised capillary blood gases should be measured and the oxygen device and flow rate should be noted on arrival at hospital (or at the time when oxygen therapy becomes necessary) for most patients requiring emergency oxygen therapy (see recommendations C1–C3).

• Blood gas measurements should be repeated in all critically ill patients and in many other cases according to the response to treatment (see recommendations C1–C3).

11.3.3 Physiological monitoring: ‘track and trigger’ systems

• EWS systems such as the NEWS are useful for monitoring patients.

• Tachypnoea is a sensitive indicator of deteriorating respiratory function.

• All acutely ill patients should have physiological monitoring using EWS or a similar physiological assessment system in addition to pulse oximetry (see recommendation B3).

• The original 2012 version of the NEWS allocated inappropriate points for hypoxaemia to patients with target ranges below 94–98% who were within their target range. It was necessary to correct for this when using such charts.165 This is recognised in the 2017 update of the NEWS chart which makes special provision for patients at risk of hypercapnia. Patients at risk of hypercapnia should score no NEWS points for saturation if within their target range and they should score points if the oxygen saturation falls below the target range or if the saturation rises above the target range while breathing oxygen.

11.3.4 Monitoring during the first hour of oxygen therapy

Evidence statement

• Monitoring of oxygen therapy is based on expert opinion.

• All patients should have their oxygen saturation observed for at least 5 min after starting oxygen therapy or for patients who require an increased concentration of oxygen and after oxygen therapy has been decreased or stopped (see recommendation W3) (evidence level 4).

11.3.5 Subsequent monitoring

The exact requirements for monitoring will depend on the clinical condition of each patient. Saturations are usually measured after 1 hour of oxygen therapy and then four-hourly. Stable patients should be monitored four times a day. However, critically ill patients will require continuous monitoring of oxygen saturation and other physiological measurements.

11.3.6 When to increase oxygen therapy

In most instances, failure to achieve the desired oxygen saturation is due to the severity of the patient's illness, but it is important to check that the oximeter is correctly placed and functioning normally as well as checking that the oxygen delivery device and the oxygen flow rate are correct. If the oxygen is being delivered from a cylinder, clinicians should check the labelling of the cylinder to ensure that it is an oxygen cylinder and checks should be made to ensure that the cylinder is not empty or near empty.

11.3.7 When to lower oxygen therapy

Most conditions which require supplemental oxygen therapy will improve with treatment and it will then be possible to reduce the amount of oxygen administered to the patient. Improvement will usually be confirmed by observing an improving oxygen saturation and a reduction in the physiological score on the NEWS observation chart as discussed in section 7.

Evidence statement

• Adjustment of oxygen therapy is based on expert opinion (evidence level 4).

Recommendations

U1: Lower the oxygen concentration if the patient is clinically stable and the oxygen saturation is above the target range or if it has been in the upper zone of the target range for some time (usually 4–8 hours) (grade D).

U2: If the target saturation is maintained, the new delivery system and flow should be continued. Repeat blood gas measurements are not required. If the patient is stable, the process can be repeated and the patient can eventually be weaned off oxygen (see section 12) (grade D).