Show Summary Details
Page of

Pulse oximetry and capnography in the ICU 

Pulse oximetry and capnography in the ICU
Pulse oximetry and capnography in the ICU

Richard Lee

Page of

PRINTED FROM OXFORD MEDICINE ONLINE ( © Oxford University Press, 2022. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in Oxford Medicine Online for personal use (for details see Privacy Policy and Legal Notice).

date: 05 July 2022

Key points

  • Both pulse oximetry and capnography are essential monitors in the intensive care unit (ICU), particularly during intubation, ventilation, and transport.

  • Equivalent continuous information is not otherwise available.

  • It is important to understand the principles of measurement and limitations, for safe use and error detection.

  • End tidal carbon dioxide (PETCO2) and oxygen saturation should be regularly checked against PaCO2 and co-oximeter SO2 obtained from the blood gas machine.

  • The PECO2 trace informs endotracheal tube placement, ventilation, and blood flow to the lungs.


Pulse oximetry and capnography provide continuous, rapidly obtained, non-invasive, reasonably accurate information. In 1991 the Society of Critical Care Medicine listed both as necessary monitoring for patients on respiratory support [1]‌.

Pulse oximetry

Pulse oximeters are used ubiquitously, in emergency departments, intensive care units (ICUs), operating rooms, during patient transport, in outpatient departments, and in increasingly in general practice. As for many monitors there is no evidence of patient outcome benefit from their use [2]‌, but there is for detecting hypoxaemia and influencing management.

General principles

Oximetry has been performed in vitro since the 1930s by spectrophotometry. The absorption of known wavelengths of light, based on Beer’s Law (absorbance is proportional to the chromophore concentration), and Lambert’s Law (absorbance is proportional to the thickness of the absorbing layer), and the differing absorption of light by the Hb species are used to measure their concentrations.

Co-Oximetry in modern blood gas machines uses multiple wavelengths of light, passed through a transparent cuvette containing the blood specimen, haemolysed by ultrasound to make a suspension. The light path distances and cuvette wall thickness are known. Non-invasive clinical oximetry had been inhibited by quantifying the variable light absorption and thickness of skin, subcutaneous tissue and venous blood until the 1970s when an eight-wavelength ear oximeter for clinical use was available, but it was expensive, bulky, and required heating of the skin. In 1974, Aoyagi described the principle of pulse oximetry [3]‌. He noticed a pulsatile variation in the tissue optical density caused by arterial pulsation and realized that if the optical density of the pulsatile component was measured at two appropriate wavelengths a ratio could be obtained, which related closely to arterial SaO2.The development of compact light emitting diodes (LEDs) and microprocessors led to easy application, and rapid machine response. The first commercial pulse oximeter was marketed in 1983.

The pulse oximeter probe is applied to a part of the body, where pulsatile arterial blood flow is sensed. Probes may be disposable or reusable, and are designed to be applied to finger, forehead, nose, ear, or toe. They are calibrated for transmitted or reflected light, and should not be dismantled or used other than in accordance with the manufacturer’s instructions.

The probe consists of two LEDs—one emits 660nm (red, R) and the other 940 nm (infrared, IR) light, chosen for maximum separation in absorbance characteristics of deoxyhaemoglobin (deoxyHb) and oxyhaemoglobin (oxyHb)—and a photodiode, which detects the intensity of the light—reflected or transmitted. The LEDs strobe at 720 Hz, sequentially (R, IR, and both OFF to provide a reference), cycling at 30 cycles/sec. The transmitted or reflected signal is digitalized, the difference between systolic and diastolic measured to give a pulsatile component (ac) and non-pulsatile component (dc), therefore allowing for quantification of absorption due to tissue, venous blood, nail, and polish. A ratio of ratios, R, is then obtained [(ac R/ac IR)/(dc IR/dc R)], simplified to R = ac 660.dc 660/ac 940.dc 940. R is empirically calibrated against look up tables of R versus known SaO2 [4]‌. The tables are obtained using co-oximetry and having normal young volunteers breathe hypoxic mixtures to produce saturations in the range 70–100% SaO2.

Pulse oximeters usually also calculate heart rate and provide visual measures of signal quality.

The monitor may calculate SpO2 25–30 times per second, but averages SpO2 over 3–12 seconds to allow artefact rejection, and update the screen every 1–2 seconds, depending on fast or slow mode, and the manufacturer.

A delay of around 1–2 minutes occurs after application of a finger probe before achieving a measurement, and a sudden reduction in FiO2 produces a fall in SpO2 with a lag of 10 seconds for ear probes and 50 seconds for finger probes [4]‌.


Oximetry is the measurement of oxyHb in blood as a percentage of Hb species. Because of the use of only two wavelengths of light the only species of Hb targeted in pulse oximetry are oxyHb and deoxyHb. Co-oximeters using multiple wavelengths of light (commonly 128) may report functional saturation (SO2, the ratio: oxyHb/(oxyHb + deoxyHb)) and fractional saturation (FO2Hb, the ratio oxyHb/tHb including dyshaemoglobins) and report methaemoglobin (MetHb) and carboxyhaemoglobin (COHb)). The terminology is often misused and misunderstood [5]‌.


The pulse oximeter reading generally achieves a mean bias of ±2% in normal volunteers in the SO2 range of 80–100%, although the International Standard requires ±4%. There are large variations in bias and error between machines, manufacturers, clinical situations, sexes, and saturation levels [6]‌ and it is suggested that a pulse oximeter SO2 of 94% is necessary to ensure a co-oximeter SO2 of 90% from a simultaneous arterial blood gas [7]. Bias increases as Hb and SO2 fall.


There are many potential sources of error [8]‌. Pulse oximeters require a detectable pulse and perform poorly in poor perfusion states. They are insensitive to large changes in PaO2 in the high PaO2 range and increasing bias occurs at low PaO2. They are calibrated using co-oximetry, which is also prone to errors [9].

Errors can be described as ‘safe’ (under-read or fail to read, leading to false alarms) or ‘unsafe’ errors (over-read, leading to false security) due to the following factors.

Patient factors

  • Abnormal Hb [10]:

    • Pulse oximeters read HbCO as mostly HbO2, over-read SO2 1% for every 1% of COHb in blood.

    • MetHb absorbs both wavelengths tending to an R of 1, forces SpO2 towards 85% as MetHb rises.

    • Fetal Hb absorbs similarly to adult Hb, produces no clinically significant error.

  • Dyes: methylene blue and indocyanine green absorb red light, artificially lower SpO2.

  • Other light absorbents: bilirubin does not interfere with oximetry.

  • Nail varnish: blue, black, green, beige, purple, and white polish artificially lower SpO2. Racial pigment does not significantly affect the measurement; difficulty with signal may be overcome by using the nail bed, less pigmented.

  • Pulsatile veins (tricuspid regurgitation, venous congestion) lower SpO2.

  • Non-pulsatile flow on cardiopulmonary bypass decreases signal to noise ratio, measurement difficult.

  • Anaemia increases bias at low SO2.

Equipment factors

  • Delay due to averaging and blood flow to peripheral tissues.

  • Flooding of external light may force a value of 85% or trigger alarm.

  • Penumbra or optical shunting around the tissue will produce false readings, countered by the ‘LED OFF’ period as a reference in the machine algorithm.

  • Movement artefact may trigger an alarm or variable errors. Seizures and helicopter vibration are reported to not interfere.

  • Fluorescent light falsely lowers SpO2, but is countered by the LED OFF period as reference in the machine algorithm.

  • Radio-frequency interference:

    • MRI use requires a link to the patient with fibre optic cables, may produce a falsely low SpO2.

    • Cautery artificially decreases SpO2 with older machines, interferes with heart rate estimation.

Advances have occurred in new software and hardware for movement artefact rejection, e.g. signal extraction technology (SET), Fourier artefact suppression (FAST) and Oxismart technology [11]. Multiple wavelength pulse oximeters are available to measure other haemoglobin species, but inherent limitations of the techniques confound the accuracy [12].

Capnography and end-tidal Co2 monitoring

Most national intensive care organizations recommend that capnography be used for all sedated or intubated patients, and failure to use or misinterpret the readings is regarded as a gap in care [13].


Methods available to measure CO2 in the gas phase include IR absorption, mass spectrometry, Raman scattering, and photo-acoustic spectroscopy. IR absorption is the commonly used technique in ICU because of cost and rapid response. The Beer and Lambert Laws are applied to measurement in a calibrated cuvette either positioned within the ventilator circuit with an adaptor (mainstream) or remote from the circuit with gas aspirated via a thin tube to the analyser (sidestream). As absorption is dependent on the number of particles of targeted substance in the gas mixture, it is a partial pressure detector. End tidal carbon dioxide partial pressure (PetCO2) measurement requires rapid response for clinically useful breath-by-breath analysis.

IR gas analysers use the principle that all gases with two or more dissimilar atoms in the molecule absorb IR radiation. CO2 is absorbed in the band 4.3 µm, oxygen does not absorb IR, but anaesthetic gases (e.g. nitrous oxide in the 4.5 µm band) and water will.

Collision broadening is the phenomenon where the presence of a second gas (e.g. N2, N2O, C3H6) widens the absorption spectrum of the other gas so that absorption is increased. This is usually countered by a standard offset.

Features of mainstream sensors

  • Better capnograph.

  • More rapid response.

  • No gas aspirated from the airway.

  • Heavier, bulkier, and more cumbersome in the circuit.

  • Expensive electronic consumables, which are prone to damage.

  • Measuring window heated to 40°C and needs to be kept clean.

  • Periodic calibration.

Features of sidestream systems

  • Cheaper consumables.

  • Practical in non-intubated patients.

  • Prone to distorted capnogram.

  • Sampling rates 50–500 mL/min may interfere with ventilation in small patients.

  • Slow response distorts capnogram.

  • Fast aspiration falsely lowers PetCO2.

  • Sampling tube occlusion may occur due to airway secretions or water.

  • Need fluid trap and filter.

Information revealed

The presence of a continued CO2 trace indicates the intra-airway position of the endotracheal tube (ETT) following intubation. The trace acts as a disconnection alarm and apnoea detector, and may detect respiratory change (V/Q mismatch, bronchospasm, etc.), effectiveness and prognosis of cardiopulmonary resuscitation (CPR), and air embolism [14]. This continuous non-invasive information decreases need for arterial blood gas estimation (ABGE) as PetCO2 provides a useful surrogate of PaCO2 in patients with normal lung function and trend in patients with COPD.

Four phases are visible on the PECO2 time trace (Fig. 73.1):

  • I represents the expiration of carbon dioxide free gas from anatomical dead space, upper airway to bronchi.

  • II represents mixed gas from airways and alveoli.

  • III represents the alveolar plateau of gas from alveoli, rising slightly due to variable mixing and time constants. The peak value represents PetCO2

  • 0 sharp descent is due to the absence of carbon dioxide in inspired gas.

  • Two angles are observed (A and B), which may be increased by lung pathology (V/Q mismatch, asthma) and rebreathing (from incompetent valves, low fresh gas flow), respectively.

Fig. 73.1 Normal capnogram. Partial pressure of expired carbon dioxide (PECO2) graphed against time in seconds, from time 0 at the start of expiration (for phases and angles, see in Information Revealed).

Fig. 73.1 Normal capnogram. Partial pressure of expired carbon dioxide (PECO2) graphed against time in seconds, from time 0 at the start of expiration (for phases and angles, see in Information Revealed).

PetCO2 approximates PaCO2 and, hence, PaCO2.

P a C O 2   =   k . V C O 2   /   A V ,
[eqn 1]

where AV or alveolar ventilation = (TV – VD) × RR, k is a constant, VCO2 is CO2 production, AV is alveolar ventilation, TV is tidal volume, VD is dead space volume, and RR is respiratory rate.

Therefore, a change in an accurately recorded PetCO2 suggests a change in CO2 production, minute ventilation, or dead space ventilation.

Co2 production

  • Increased by fever, parenteral nutrition, malignant hyperthermia, thyrotoxicosis, tourniquet release, bicarbonate infusion.

  • Decreased by hypothermia, sedation.

Minute ventilation

  • Increased by hyperventilation.

  • Decreased by hypoventilation.

Dead space ventilation

  • Increased by shock, cardiac arrest, air embolism, pulmonary embolism, V/Q mismatch.

  • Decreased by effective CPR.

  • The gap between PetCO2 and PaCO2 is usually < 5 mmHg (0.67 kPa), but may be >15 mmHg (1.99 kPa) in ARDS and is dependent on ventilation mode, dead space, and cardiac output. Alveolar dead space can be estimated as 1– (PetCO2: PaCO2).


Most errors relate to technical problems of calibration, blocked sensor window or blocked sampling tube, or delay in achieving a trace. High sampling flow rate with a sidestream sensor will produce a falsely low PetCO2. Capnography provides information, which requires pattern recognition, interpretation, comparison with normal, and clinical decision-making, which are all major clinical sources of error. An absent trace after intubation in cardiac arrest may be falsely attributed to ineffective CPR and low cardiac output, but a flattened trace should at least be seen. In event of oesophageal intubation after ingestion of carbonated drinks may give an initial CO2 trace, falling after several breaths.

Common abnormal patterns

  • Absent PCO2 trace: oesophageal intubation, disconnection, apnoea, total airway obstruction, cardiac arrest without CPR.

  • Elevated Phase 1: rebreathing, incompetent circuit valves, exhausted CO2 absorber during anaesthesia, low fresh gas flow, slow sidestream sampling rate.

  • Decreased Phase II slope: slow sidestream sampling rate, kinked or blocked ETT, expiratory obstruction (e.g. asthma or chronic obstructive pulmonary disease (COPD)).

  • Absent Phase III plateau: in COPD, bronchospasm, acute respiratory distress syndrome (ARDS).

  • Elevated PetCO2: hypoventilation, malignant hyperthermia (MH), CO2 insufflated at laparoscopy, bicarbonate administration, release tourniquet.

  • Decreased PetCO2: hyperventilation, hypothermia, low cardiac output (CO), pulmonary embolus (PE) [15].


1. Society of Critical Care Medicine (1991). Guidelines For Standards of Care for Patients with Acute Respiratory Failure on Mechanical Ventilatory Support. Critical Care Medicine, 19(2), 275–8.Find this resource:

2. Pedersen T, Hovhannisyan K, and Moller AM (2010). Pulse oximetry for perioperative monitoring (Review). The Cochrane Library, Issue 12.Find this resource:

3. Severinghaus JW. (2007). Takuo Aoyagi: discovery of pulse oximetry. Anesthesia and Analgesia, 105, S1–4.Find this resource:

4. Tremper KK and Barker SJ (1989). Pulse oximetry. Anaesthesiology, 70, 98–108.Find this resource:

5. Toffaletti J and Zijlstra WG. (2007). Misconceptions in reporting oxygen saturation. Anesthesia and Analgesia, 105, S5–9.Find this resource:

6. Feiner J, Severinghaus JW, and Bickler PE (2007). Dark skin decreases the accuracy of pulse oximeters at low oxygen saturation: the effects of oximeter probe type and gender. Anesthesia and Analgesia, 105, S18–23.Find this resource:

7. Van de Louw A, Cracco C, Cerf C, et al. (2001). Accuracy of pulse oximetry in the intensive care unit. Intensive Care Medicine, 27, 1606–13.Find this resource:

8. Milner QJW and Matthews GR. (2012). An assessment of the accuracy of pulse oximeters. Anaesthesia, 67, 396–401.Find this resource:

9. Gehring H, Duembgen L, Peterlein M, HagelBerg S, and Dibbelt L (2007). Hemoximetry as the ‘gold standard’? Error assessment based on differences among identical blood gas analyser devices of five manufacturers. Anesthesia and Analgesia, 105, S24–30.Find this resource:

11. Giuliano K and Higgins TL. (2005). New-generation pulse oximetry in the care of critically ill patients. American Journal of Critical Care, 14(1), 26–37.Find this resource:

10. Barker SJ and Badal JJ. (2008). The measurement of dyshaemoglobins and total haemoglobin by pulse oximetry. Current Opinion in Anesthesiology, 21, 805–10.Find this resource:

12. Shamir M Y, Avramovich A, and Smaka T (2012). The current status of continuous non-invasive measurement of total, oxy, and methaemoglobin concentration. Anesthesia and Analgesia, 114, 972–8.Find this resource:

13. Editorial (2009). Airway incidents in critical care, the NPSA, medical training and capnography. Anaesthesia, 64, 351–7.Find this resource:

14. Soubani AO. (2001). Non-invasive monitoring of oxygen and carbon dioxide. American Journal of Emergency Medicine, 19(2), 141–6.Find this resource:

15. Sivarajan VB and Bohn D. (2011). Monitoring of standard hemodynamic parameters: Heart rate, systemic blood pressure, atrial pressure, pulse oximetry, and end-tidal CO2. Pediatric Critical Care Medicine, 12(4), S2–11.Find this resource: