Correspondence: A.H. Kendrick, Dept of Respiratory Medicine, Bristol Royal Infirmary, Bristol, BS2 8HW, UK.

Few laboratory measurements have had as great an impact on clinical practice as the determination of arterial blood gas levels. The arterial carbon dioxide tension (Pa,CO2) indicates whether acid/base disorders are of respiratory or metabolic origin, and is a sensitive and specific indicator of the adequacy of alveolar ventilation. An elevated Pa,CO2 indicates that alveolar ventilation is insufficient, whereas a reduced Pa,CO2 indicates increased alveolar ventilation.
Measurements of arterial oxygen tension (Pa,O2) indicate the adequacy of pulmonary oxygen exchange. A reduced Pa,O2 indicates a low inspired oxygen tension (PO2), abnormal lung function or the presence of a shunt. In addition, measurements of pH and bicarbonate permit classification of the acid/base disturbance.
Arterial blood gas levels are the gold standard. However, there are many situations in which occasional or continuous measurements of carbon dioxide tension (PCO2) and PO2 are useful, including bronchoscopy, exercise testing, sleep studies and manipulation of ventilators. Often, occasional or continuous measurements are not carried out because of the risks, discomfort and delay involved in intra-arterial blood sampling. The use, therefore, of noninvasive techniques provides the opportunity for obtaining measurements in almost any clinical situation.

The recognition of the presence of hypoxaemia by the human eye is rather poor. In their classic study, Comroe and Bothello [1] observed that 47% of observers could not detect hypoxaemia until oxygen saturation (SO2) fell below 80%. Even when the SO2 had fallen to 71-75%, 25% of observers did not detect hypoxaemia.
The detrimental effects of hypoxaemia have been recognized for a long time, and were summarized succinctly by J.S. Haldane: "anoxaemia not only stops the machine, but wrecks the machinery". The long-term O2 therapy trials [2, 3] have clearly shown that hypoxaemia has an adverse effect on longevity, but that this can be reversed to some extent by administration of O2.
Pulse oximeters now proliferate in hospitals and in primary care practice. Despite this, little data is available regarding what happens to patients in terms of SO2 during admission to hospital. It is known that episodic hypoxaemia is more common than suspected and that, where hypoxaemia (<90%) lasts for „5 min within the first 24 h after admission, the long-term prognosis is 3.3 times worse that when there are no hypoxaemic events within the first 24 h (fig. 1) [4].

Fig. 1. - Patient survival from the day of admission to hospital to 4 months after admission. Pulse oximetry was monitored for the first 24 h after admission ( &emdash;&emdash; : Sp,O2 <90%; - - - : Sp,O2 >90%). (From [4].)

The first pulse oximeter was commercially manufactured in 1975 and provided the first noninvasive assessment of SO2 (oxygen saturation obtained from pulse oximeter (Sp,O2)). It was "arguably the most significant technological advance ever made in monitoring the well-being and safety of patients during anaesthesia, recovery and critical care" [5]. In many countries, pulse oximetry is now an essential component of the "standard of care" and is used as a tool to alert the clinician to the presence of hypoxaemia and guide the more rapid treatment of serious hypoxaemia [6].

The principles of oximetry are based on the spectrophotometric absorption of specific wavelengths of light by blood. The first such device was used in 1935 by K. Mathes, but the sensor was so cumbersome that its use, even as a trend monitor, was impractical [7, 8].
Important advances were made during World War II, especially with pilots who needed to be monitored at high altitude in unpressurized cockpits. Milikan devised a lightweight sensor and referred to this as an "oximeter" [9]. Despite these developments, the oximeter remained a difficult practical instrument to use until the 1960s, when a self-calibrating eight-wavelength ear oximeter was developed and marketed by Hewlett-Packard. The device was accurate to within ±4% within the range 65-100%, but was rendered inaccurate in the presence of jaundice, carboxyhaemoglobin or skin pigments. Despite these limitations, this device quickly became the standard clinical and laboratory tool within pulmonary medicine, despite its cost.
The real breakthrough in "pulse oximetry" came in the mid 1970s, when T. Aoyagi observed that the variation in tissue arterial blood volume with each pulse could be used to obtain a signal dependent only on the characteristics of arterial blood [10]. This resulted in much simpler sensors being developed, and, with further improvements in microprocessor technology, has led to small hand-held pulse oximeters with an acceptable degree of accuracy [11].

Pulse oximetry uses readily available light-emitting diodes (LEDs) at two wavelengths: 660 nm (red) and 940 nm (near infrared). These wavelengths are convenient, as they are the points of greatest separation between the haemoglobin and oxyhaemoglobin absorption spectra (fig. 2). As the emitted light passes through the finger or earlobe, some of the energy is absorbed by arterial and venous blood, tissue and the variable pulsations of arterial blood. Using electronic circuitry, the signals in the infrared and red wavelengths are equalized and the ratio of red:infrared light is calculated, which is directly related to Sp,O2 by pulse oximetry (fig. 3). Each second, approximately 600 individual measurements are made and fed into an algorithm contained within the microprocessor, compared with stored values and then weighted using formulae that are specific to each manufacturer. The displayed value is averaged over the previous 3-6 s and updated every 0.5-1 s. This averaging tends to reduce the effects of artefacts and erroneous signals.

Fig. 2. - Absorption spectra of oxygenated (&emdash;&emdash; ) and deoxygenated (- - -) haemoglobin, showing the two wavelengths most commonly used for measurements (660 and 940 nm). (From [12].)

Fig. 3. - a) Components of the absorbance signal; b) equalization of direct current (DC) levels to give scaled signals; and c) a schema of the electronic circuitry of a pulse oximeter. b) The upper absorbance correspond to "raw" signals (DC unequal) and the lower ones to "scaled" signals (DC equal). AC: alternating current; RAM: random-access memory; EPROM: erasable programmable read-only memory; Sp,O2: oxygen saturation obtained from pulse oximeter.

Most pulse oximeters display a plethysmographic waveform, in the form of either a waveform or a signal strength indicator. These waveforms can be used to assess the effects of artefacts or poor signal quality (fig. 4).

Fig. 4. - Common pulsatile waveforms from a Biox 3740 pulse oximeter (Datex-Ohmeda Division Instrumentarium Corp., Helsinki, Finland). a) Normal plethysmographic waveform showing peak pressure and notch (arrow); b) low signal indicator suggesting poor site perfusion or that the securing tape over the probe is too tight; c) electronic interference observed during electrosurgery, magnetic resonance imaging or other electrical/electronic devices; and d) motion artefact or poor probe positioning. The numbers above a and b represent oxygen saturation (96 and 94%) pulse (77 and 64 beats·min-1). (From [13].)

In addition to SO2, the LED cycles between successive pulsatile signals can be used to calculate the cardiac frequency. When the cardiac frequency, as displayed on the pulse oximeter, differs significantly from that obtained from electrocardiography, the Sp,O2 may also be erroneous, and needs to be interpreted with caution.

Oximeters are not easy to calibrate and it is generally accepted that in vivo calibration need only take place during the design and development of the instrument. In cases in which calibration is required, blood samples need to be taken and analysed using a co-oximeter. The only method of performing an in vivo calibration is to use normal volunteers, and an indwelling radial artery catheter, to obtain estimates of both arterial oxygen saturation (Sa,O2) and Sp,O2 while the subjects breathe gas mixtures containing combinations of O2 and nitrogen. Since volunteers should not be taken below 80% SO2, calibration at levels below this is by extrapolation.
As an alternative to in vivo calibration, Nonin Medical (Plymouth, USA) produces the Finger Phantom. This device simulates the light absorption and arterial blood flow of the human finger. The accuracy of the oximeter system can be assessed at 97, 90% and 80% SO2. Although these devices have been developed specifically for Nonin pulse oximeters, they can be used with other makes of pulse oximeter and transmittance sensor. Variations in displayed value, however, may occur with pulse oximeters from other manufacturers.

The accuracy of commercial pulse oximeters is generally 2-3% in the SO2 range 70-100%. Below this, accuracy is obtained by extrapolation, and is therefore poorer [14]. In addition, the accuracy between different commercial brands of oximeters may be very different despite using similar hardware. The differences are most likely to be due to the algorithms used in processing the light intensity signals from which Sp,O2 is obtained [15].
Numerous studies have assessed the accuracy of pulse oximeters, usually comparing them to a co-oximeter. In cases in which SO2 is normal, pulse oximeters have an accuracy of ~±2% [16-18], even in critically ill patients [19] or in patients undergoing respiratory diagnostic screening [20]. At SO2 <80%, the accuracy can deteriorate to as little as ±10% and there is a tendency to under-read [16, 19, 21, 22].
In patients with chronic obstructive pulmonary disease who were hypoxaemic [23] and in critically ill patients, in whom Sa,O2 was as low as 63% [24], there was poor agreement between pulse oximeter and co-oximeter readings, the oximeter under-reading at SO2 of <80% [24]. This may, however, be dependent on the oximeter used for comparison [25].
In a recent meta-analysis of the measurement of Sa,O2 by pulse oximetry, Jensen et al. [26] concluded that, from the 74 studies included in the analysis, pulse oximeters were accurate to within 2% in the range 70-100% Sa,O2. They also concluded that finger probes were more accurate and pulse oximeters failed to accurately record the true Sa,O2 during severe or rapid desaturation, hypotension, hypothermia, dyshaemoglobinaemia and low perfusion states.

Most pulse oximeters are purchased with a standard reusable finger-clip probe, although disposable finger probes are available. Less often used are ear-lobe, toe, nose and forehead probes. Probes are also available for neonates, children and adults. The accuracy of these probes varies with type and site of location of probe [27]. In general, finger probes appear to be more accurate than other probes (fig. 5).

Fig. 5. - Accuracy of pulse oximeter probes (1: Radiometer (Radiometer America, Inc., Westlake, OH, USA); 2: Sensormedics Oxyshuttle; 3: Criticare CSI 5O4; 4: Invivo 4500; 5: Datex Satlite (Datex-Ohmeda Division Instrumentarium Corp., Helsinki, Finland); 6: Ohmeda Biox 3740; 7: Ohmeda 3700; 8: Novametrix 5O5 (Nonin Medical Inc., Plymouth, MN, USA) placed on the finger (white box), ear (linear box ///), nose (linear box \\\) or forehead (hatched box) compared to co-oximeter measurements under conditions of poor perfusion. The bias of the finger probes ranged 0.2-1.7 and that of the other probes 0.1-8.1. (From [27].)

The dynamic response of the oximeter system also affects accuracy, especially when following dynamic or transient changes. There have been few studies investigating the dynamic response of oximeters [28-32]. In all of these studies, the response time was faster for ear probes as compared to finger probes (fig. 6), the difference being up to 24 s.

Fig. 6. - Response characteristics of ear (- - -) and finger pulse oximeter probes (- - - -) in a subject experiencing severe oxygen desaturation. Compared with oxygen saturation calculated from expiratory oxygen tension (&emdash;&emdash;), to ear probe shows a lag of ~50 s. Sp,O2: oxygen saturation obtained from pulse oximeters. (From [33].)

Most pulse oximeters average the output over a fixed number of seconds. Some, such as the Ohmeda Biox 3740 (Datex-Ohmeda Division Instrumentarium Corp., Helsinki, Finland), permit selection of the averaging time in the range 3-12 s, depending on the application, with the default being either 6 or 12 s, depending on the software version. Care needs to be taken to ensure that the appropriate averaging time is selected. Evidence from overnight sleep recordings has indicated that Sp,O2 may be significantly underestimated with longer averaging times [34, 35]. The number of 4% desaturations per hour was significantly lower at an averaging time of 12 s (mean 1.9; range 0-8.6) compared to 6 s (mean 5.8; range 0-40) in simultaneous recordings made from a finger probe using an Ohmeda Biox 3740 [34].
Many pulse oximeters have a preset fixed averaging time, but that of some, such as the Ohmeda Biox range, can be set at the commencement of each application. A fast averaging time is required during exercise studies and the use of 3 s is appropriate. In cases in which gradual changes or following a trend is required, a longer signal averaging time, such as 12 s; may be appropriate. Currently, in sleep studies, an averaging time of 5-6 s is used. Of note, however, is that, the shorter the averaging time, the greater the degree of artefact encountered in the recorded signal. The choice, therefore, of signal averaging time is a balance between the accuracy of the signal recorded in reflecting Sp,O2 changes and the degree of noise in the resulting recorded signal.

Some, but not all, pulse oximeters have analogue and digital outputs to permit recording of data with multisignal recorders. Thus overnight sleep studies that require recording of SO2 coupled with chest wall and abdominal breathing patterns and oronasal airflow require an output signal from the oximeter to either a chart recorder or a computer-based system. Many of the hand-held devices do not have outputs that allow signals to be sent to recording devices.
Another use of the RS232 port on some oximeters is to allow remote transmission of memory-stored data to a central data analysis site. Linking a pulse oximeter to such a site, via a modem reduces costs and potentially provides an efficient data analysis service.

Memory. Many pulse oximeters now come with memory storage capabilities. This ranges 8-24 h depending on the pulse oximeter. In cases in which an application requires remote use of pulse oximetry, such as during overnight sleep studies, memory storage can be used to study the patient in the home environment rather than requiring the patient to stay in hospital overnight, using a pulse oximeter attached to a computer or data storage system. Some possible applications are listed in table 1.

Table 1. - Uses of memory storage capabilities of pulse oximeters
Application	Comment
Overnight sleep studies	Studies performed in the home and in hospital wards and prior to any therapeutic intervention.
Exercise tests	6 and 12-min walking and shuttle tests using wrist- or belt-mounted oximeters and in cases in which the operator cannot directly monitor patients during the exercise study.
Domiciliary oxygen studies	Studies monitoring patients in daily activity with or without supplemental or long-term oxygen.
CPAP/ventilator patients	Used in monitoring SO2 in patients on this therapy at home to assess success of treatment regimen.
Flight assessment	Used in monitoring patients with lung disease who wish to fly, with or without supplemental oxygen. Patients with significant lung disease should have a flight assessment study prior to flying.
Videofluoroscopy	Used for assessing patients with dysphagia during a variety of swallowing manoeuvres. Requires careful post-study analysis of changes in SO2 before, during and after the swallow.
CPAP: continuous positive airway pressure; SO2: oxygen saturation.

The decision about which memory storage oximeter to use for a specific application depends on the study environment. Some oximeters store data every 12 s into memory, i.e. five data points per minute, whereas others store data every 5 s [36]. In cases in which rapidly changing Sp,O2 events are being monitored, such as during sleep studies to detect apnoeic events, more rapid data storage is required. Conversely, in cases in which events are changing more slowly, such as monitoring during assessment of fitness to fly with the subject breathing at an inspiratory oxygen fraction (FI,O2) of 0.15, data can be stored less often as trend analysis is required rather than analysis of quickly changing events.

Files. In addition to the data storage rate, the number of files that can be stored within the memory, and the length of each file, may be important. Some pulse oximeters permit numerous files to be stored, the total number being determined by the size (in hours) of each file and the maximum storage capability. This allows greater flexibility and, for instance, permits three 8-h sleep studies, or 24 1-h videofluoroscopy studies to be recorded and stored before data analysis occurs. Alternatively, a single file of Sp,O2 and cardiac frequency data can be recorded in a subject to assess the usefulness of supplemental O2 therapy over a given 24-h period.

Analysis software. Many companies provide their own data analysis software. Care needs to be taken in choosing this software in order to ensure that it does exactly what is required. It must be capable of not only linking to the pulse oximeter and downloading the stored data but also indicating possible areas of artefact and providing an analysis and visual data presentation that the user can understand. Some download software, particularly for the analysis of overnight sleep studies, have been developed to be compatible with a variety of pulse oximeters, and, regardless of the pulse oximeter used, provide a standard and user-interactive data analysis system. The ability to export the raw data to presentation and statistical packages is also a useful feature.

There are a number of important limitations to consider when using pulse oximeters.

Motion artefact. This is a potential source of error [14, 37-40], and can significantly reduce the accuracy of Sp,O2 readings. Excessive motion artefact may lead to abandonment of the use of pulse oximetry [37]. Errors of up to 20% have been observed in simulated motion artefact studies [38]. Oximeters may use different algorithms to detect and eliminate motion artefact [39], even within a range of products from the same manufacturer [39, 40]. Care, therefore, needs to be taken to ensure that the motion artefact is reduced as much as possible.
Recently, an innovative approach, Masimo signal extraction technology, has been introduced to extract the true signal from artefact due to noise and low perfusion [41]. New algorithms for processing the red and infrared light signals enable the noise common to both signals to be measured and subtracted [42]. This has resulted in much lower error rates [39]. Using this advanced technology, Dumas et al. [43] noted that the frequency at which warning alarms occurred decreased from one every 13 min to one every 30 min and the number of false alarms from 87 to 59%.

Malpositioning. If the oximeter probe is incorrectly positioned over the finger or ear lobe, pulse oximeters can yield erroneously low SO2 [44]. Some pulse oximeters under-read at high SO2, whereas others under-read regardless of the SO2 as compared to direct arterial blood sampling. Overall, there appears to be a loss of sensitivity to changes in Sa,O2. Under-reading should prompt an intervention, whereas over-reading or failure to follow trends may induce a sense of false security and be potentially dangerous.

Pulse dependence and site perfusion. By their very nature, pulse oximeters are pulse-dependent, and require a pulse of regular rhythm and a site with adequate perfusion. Thus a low cardiac output, vasoconstriction or hypothermia might make it very difficult to distinguish the true signal from the background noise [17, 19, 45-50]. Some pulse oximeters display a message indicating a poor quality signal (fig. 4) [17, 45-49, 51].
In normal subjects, applying a tourniquet to produce vascular occlusion results in deterioration of bias and precision [17]. In cases in which vasoconstriction occurs with significant venous engorgement, the response time of pulse oximeters can increase [45]. The lower limits of mean systolic pressure at which oximeters fail range from 47.1±13.5 mmHg (Nellcor oximeter) to 38.7±14.5 mmHg (Criticare oximeter) to 36.0±3.5 mmHg (Ohmeda oximeter), and shows that different pulse oximeters have different pressure thresholds at which they fail to accurately record SO2 [47].
In critically ill patients under a wide range of haemodynamic conditions (temperature 32.8-39°C, cardiac index 1.4-8.7 L·min-1·m2), pulse oximeters have been shown to be accurate [52]. Similarly, in patients with a low cardiac index (<2.2 L·min-1·m2) or hypothermia (<28.5°C), the accuracy of pulse oximetry was not affected [53]. Despite the findings of these two studies, Clayton et al. [48] found that only two of 20 pulse oximeters gave readings within 4% of a reference co-oximeter 95% of the time.
These studies, in critically ill patients, suggest that, in low-perfusion states, the accuracy of pulse oximeters deteriorates, but the degree of decreased perfusion required to produce inaccurate readings has yet to be clearly defined.

Which probe? The choice of probe clearly depends on the application, and probes are manufactured for use on the ear, finger, nose or forehead. Compared to co-oximeter data, finger probes are accurate to within 0.2-1.7%, closer than the values obtained using ear, nose and forehead probes [27]. In children, it appears that the type of probe does not have any significant effect on accuracy [54].
For use during exercise testing on a treadmill, ear-lobe probes are generally preferred, as there is less likelihood of reduced perfusion to the hands when holding on to the sides of the treadmill. For overnight sleep studies, however, finger probes are more appropriate.

Substance interference. Various substances interfere with the absorbance of light at 660 and 940 nm. 1) Methylene blue, indigo carmine and indocyanine green cause inaccuracy as they each strongly absorb energy at 660 nm [55]. 2) Skin pigments may or may not affect accuracy [56, 57], and any effect may depend on the probe site used. Problems, such as low Sp,O2 readings, appear to occur more frequently in patients with dark skin pigmentation than in those with light pigmentation [57]. Similarly, in critically ill patients, greater inaccuracies in Sp,O2 reading occur in black patients as compared to white patients [58]. The reasons why skin pigmentation causes less accurate Sp,O2 readings is unclear. Darkly pigmented skin may interfere with the absorption wavelengths used in pulse oximetry. There is evidence that black nail polish [59] and the black ink used in fingerprinting [60] produce inaccurate oximeter readings. 3) Nail varnish should not, in theory, affect Sp,O2 readings as the absorbance of light is nonpulsatile and is omitted from saturation calculations. However, nail polish does appear to affect the accuracy of Sp,O2 readings [59, 61]. In general, slight decreases in Sp,O2 reading are observed with blue, green and black polishes, whereas red and purple appear to have no effect. Synthetic or acrylic nails do not appear to have significant effect [62]. As with skin pigmentation, the explanation for these observations is likely to be the effects of light absorbance on the wavelengths of light used in pulse oximeter probes (fig. 7).

Fig. 7. - Absorption spectra for five different nail polish colours (&emdash;&emdash;: green; &emdash; &emdash; &emdash;: red; - - - -: black; - - - - - -: purple; - - - - - -: blue). Under these conditions, the absence of nail varnish yields zero absorbance. The vertical arrows indicate the oximeter measurement wavelengths of 660 and 940 nm. (From [50].)

Ambient light. Normal ambient light does not affect Sp,O2 readings as the LEDs make adjustment for ambient light [63]. Fluorescent and xenon arc surgical lights, as well as bright sunlight, have been shown to cause falsely low Sp,O2 readings [64, 65]. This problem can be overcome by wrapping the probe with an opaque shield.

Dyshaemoglobins. Pulse oximeters are calibrated for adult haemoglobin (haemoglobin A) and assume that carboxyhaemoglobin and methaemoglobin are found only in small quantities. If carboxyhaemoglobin levels are high, a dangerous effect may be observed as Sp,O2 is the sum of Sa,O2 and carboxyhaemoglobin (as a percentage) levels. Since pulse oximeters measure carboxyhaemoglobin as fully oxygenated haemoglobin, this results in overestimation of the true Sa,O2 [66].
Methaemoglobin interferes in a similar way with the accuracy of oxygenated haemoglobin measurement, leading to overestimation of Sp,O2 [67, 68]. When treating methaemoglobinaemia, methylene blue is used. Although this reverses the clinical signs of methaemoglobinaemia, methylene blue causes falsely low Sp,O2 readings.

Hyperbilirubinaemia. Bilirubin does not cause interference [69, 70]. Although Veyckemans et al. [69] observed a bias of ~2.9% in jaundiced patients and ~1.7% in nonjaundiced patients, they noted that the jaundiced patients had higher carboxyhaemoglobin levels. When SO2 was corrected for carboxyhaemoglobin levels, the small differences in bias disappeared.

Anaemia. In nonhypoxaemic anaemic patients (haemoglobin concentration 5.2±0.3 g·dL-1), pulse oximetry was accurate, with a bias of 0.53% [71]. To date, there has been no investigation into the accuracy of pulse oximetry in anaemic hypoxic patients.
In sickle cell disease, care needs to be taken when comparing Sp,O2 and Sa,O2 by means of blood gas analysis [72-75]. Overall, underestimation of SO2 occurs in these patients (bias of up to 8%), which may be explained by the rightward shift of the oxyhaemoglobin dissociation curve observed in sickle cell anaemia [72]. Such a shift produces a lower haemoglobin saturation for a given blood O2 content. Additionally, the formation of sickle polymers during deoxygenation results in the erthyrocytes becoming rigid and obstructing pulmonary capillary blood flow in the extremities [76]. The resultant sludging in the pulmonary capillaries may result in lower Sp,O2 readings.

Complications. There are a few reports of problems with pulse oximetry when used in conjunction with other equipment [77] or medical treatments [78, 79]. Keidan et al. [77] noted low saturation readings when a peripheral nerve stimulator was used on the same arm as the pulse oximeter. Both Farber et al. [78] and Radu et al. [79] noted that, following photodynamic therapy, from which skin photosensitization resulted, burns occurred at the site of the oximeter probe.
One of the more serious complications of pulse oximetry is that it can provide a false sense of security. For instance, prolonged apnoeas or hypercapnic acidosis may occur in patients receiving supplemental O2 therapy, although the Sp,O2 readings appear normal [80]. In this situation, it is not the oximeters that are the problem; rather, it is the users who have the problem.
In 1994, Stoneham et al. [81] published the results of a survey of medical and nursing staff using pulse oximeters. Worryingly, 30% of physicians and 93% of nurses thought that oximeters measured Pa,O2. Some physicians did not understand the oxyhaemoglobin dissociation curve and did not recognize that Sp,O2 in the high 80s (%) equated to rather low Pa,O2. Furthermore, some physicians and nurses were not unduly worried when patients had Sp,O2 in the low 80s (%). This survey clearly demonstrates the need for appropriate training of all healthcare workers in the use and limitations of pulse oximetry.

Special environments. Pulse oximetry is used widely throughout hospitals and in the home, generally without major problems. However, the monitoring of critically ill patients requiring magnetic resonance imaging (MRI) or the use of MRI during fibreoptic bronchoscopy requires care in the use of pulse oximetry as radio frequency burns due to the use of conductive cables and sensors may result [82, 83]. Oximeters designed specifically for use with MRI should therefore be used to reduce any risks as far as possible.

Alarms. False alarms can be very common in the intensive care setting [84]. Pulse oximeters are generally marketed with default low Sp,O2 alarms set at ~90%. With many pulse oximeters, it is possible to reset these alarms to lower limits, and, with some oximeters, to switch them off completely whilst in use, with the default value being reset when the oximeter is switched off. With some of the hand-held devices, it is possible to permanently reset the Sp,O2 alarms to new lower default values, which is extremely useful when using pulse oximetry during domiciliary sleep studies, as it reduces the need for the patient to do anything except switch on the oximeter.

Troubleshooting. This is necessary when the oximeter appears not to be working correctly. In order to identify the problem, careful and systematic examination of the electrical system, the probe and patient-related factors are required. In cases in which electrical problems are noted, a qualified engineer should investigate the problem. The probe can be simply checked by observing whether or not the red LED is visible and then going on to check for other problems, as necessary. Patient-related problems have been dealt with above, and include skin pigmentation and nail polish. Some of the problems occurring and possible corrective action are given in table 2.

Table 2. - Troubleshooting: common problems in pulse oximetry, probable cause and appropriate corrective action
Problem	Probable cause	Corrective action
No signal	No mains power	Plug into mains and switch on. Check socket
	Battery requires recharging or replacement	Recharge battery or request engineer to change battery pack
	Defective mains cable	Check cable for visible damage; replace cable
	Internal circuit failure	Request engineer to service equipment
	Mains plug fuse blown	Use different cable. Request engineer to service equipment
	Batteries discharged	Replace batteries
Poor-quality signal	Poorly perfused site	Rub site vigorously to improve perfusion.
		Remove tight clothing, deflate blood pressure cuff or undo tight tape around probe site
	Perfusion at probe site inadequate for valid reading	Check patient and oximeter set-up. Often observed using finger probes during treadmill exercise tests. Ask patient to hold on to the side of the treadmill loosely, or use ear probe
Low light signal	Dirty probe or probe site	Clean probe and test site
	Malpositioned probe	Reposition probe or use different site
	Nail polish or skin pigmentation	Remove nail polish and use different site that is less pigmented, e.g. fingernail
	Probe failure	Try another probe
Inconsistent signal displayed	Artefact or noise	Check connections. Reduce patient movement as far as possible, secure probe to site with light taping, use different site, or use self-adhesive probes on fingers or toes
	Excessive ambient light	Shield probe from ambient light
	Dyshaemoglobins	Use of oximeter to monitor trends only. Actual values are inaccurate
No probe alert	Probe incorrectly connected	Reconnect probe. Use compatible probes for oximeter
Off patient alert	Probe off patient	Reattach probe. Use light taping on site if necessary
	Excessive light detected	Shield probe from ambient light
	Excessively thin tissue at probe site	Use an alternative site if possible
Low SO2 alarm regularly sounding	Check diagnosis of patient - ?sleep apnoea	Patients having apnoeas show regularly decreased SO2. Reset lower limit alarm
	Ensure patient is not becoming distressed and hypoxic, such as may occur after anaesthesia	Check patient for vital signs and monitor carefully
SO2: oxygen saturation.

Pulse oximetry provides a very simple easy-to-use noninvasive technique for assessing gas exchange function. Within its many limitations, a guide to the degree of blood oxygenation can be obtained rapidly and with an acceptable degree of accuracy in clinical practice.
It is essential that users understand the use and limitations of pulse oximetry, and that appropriate training is given to all healthcare workers regarding both the theoretical background to pulse oximetry and its practical application.

In conjunction with pulse oximetry, further noninvasive assessment of blood gas levels can be made using mixed venous (Pv,CO2) measurements and end-tidal carbon dioxide tension (PET,CO2) measurements. A brief outline of these two measurements is given here for completeness.

The two-stage method of measuring PCO2 is a simple means of estimating Pa,CO2 [85]. The patient rebreathes from a 2-L bag containing O2 for 90 s, during which time the PCO2 equilibrates with Pv,CO2 (fig. 8). However, the PCO2 in the bag is higher than resting Pv,CO2. Two minutes breathing room air is allowed to removed any retained CO2 before rebreathing for a second time to produce a plateau in 10-15 s. This plateau represents Pv,CO2. It has been practice to subtract 0.9 kPa from the Pv,CO2 to obtain Pa,CO2. For patients in respiratory failure, there is a much wider arteriovenous PCO2 difference, so it is better to multiply the mixed venous value by 0.8 to obtain Pa,CO2 [87].

Fig. 8. - Rebreathing method of carbon dioxide tension (PCO2) measurement. The subject rebreathes from a bag containing 100% oxygen (a) so that the carbon dioxide in the lungs and in the bag equilibrates. (b) (1 mmHg=0.133 kPa.). (Modified from [86].)

The rebreathing PCO2 can replace Pa,CO2 in the assessment of stable chronic airflow obstruction and is useful in detecting alveolar hyperventilation. Generally, the mixed venous measurement should be obtained in any patient whose forced expiratory volume in one second is <1.0 L or 30% of the predicted value, since, below this value, the PCO2 can be markedly increased [88-90]. It is useful when arterial blood gas measurements are not readily available or as a bedside monitoring technique. Combined with simultaneous pulse oximetry measurements, very useful noninvasive information on blood gas status can be obtained. Furthermore, failure of the Sp,O2 to attain 100% when breathing 100% O2 suggests the existence of a gas exchange defect that needs further investigation.

PET,CO2 measurements have been used by physiologists for many years. Following concerns about this measurement in the operating theatre, new commercially available systems have been produced and have been used by anaesthesiologists [91]. Gas analysis is either by infrared absorption or by mass spectrometry. The technique has been applied in the intensive therapy unit, where mass spectrometry using long probes has provided valuable information on the clinical status of patients [92]. The technique can also provide useful information when there are clinically significant changes in Pa,CO2 [93] or as a noninvasive substitute for arterial blood gas level measurements during the adjustment or withdrawal of patients from ventilation [94, 95]. The PET,CO2 is ~0.4 kPa below the Pa,CO2.
The technique has problems, as many physiological conditions may alter PET,CO2, and, therefore, make it unreliable as a marker of Pa,CO2. Changes in dead space:tidal volume ratio, positive end-expiratory pressure, ventilation/perfusion relationships, breathing pattern and cardiac output can result in precision and bias such that PET,CO2 is inaccurate by 15-20% compared with Pa,CO2.
It appears that the usefulness of this technique may not be as a substitute for measuring arterial carbon dioxide tension but rather as a means of following the trend of arterial carbon dioxide tension. In one study, changes in arterial carbon dioxide tension of >1.3 kPa were correctly identified by similar changes in most postoperative patients [95].

1. Comroe JH, Bothello S. The unreliability of cyanosis in the recognition of arterial anoxaemia. Am J Med Sci 1947; 214: 1-9.
2. Medical Research Council Working Party. Long-term domiciliary oxygen therapy in chronic hypoxic cor pulmonale complicating chronic bronchitis and emphysema. Lancet 1981: i: 681-686.
3. Nocturnal Oxygen Therapy Trial group. Continuous or nocturnal oxygen therapy in hypoxaemic chronic obstructive lung disease. Ann Intern Med 1980; 297: 391-398.
4. Bownton DL, Scuderi PE, Haponik EF. The incidence and effect on outcome of hypoxaemia in hospitalized medical patients. Am J Med 1994; 97: 38-46.
5. Severinghaus JW, Naifeh KH. Accuracy and response of six pulse oximeters to profound hypoxia. Anesthesiology 1987; 67: 551-558.
6. Bownton DL. A critical review of pulse oximetry. Clin Pulm Med 1994; 1: 181-187.
7. Mathes K. Untersuchungen über die Sauerstoffsättingungen des menschlichen Arterienblutes. Arch Exp Pathol Pharmacol 1935; 179: 698-711.
8. Severinghaus JW, Astrup PB. History of blood gas analysis: VI. Oximetry. J Clin Monit 1986; 2: 270-288.
9. Milikan GA. An oximeter: an instrument for measuring continuously oxygen saturation of arterial blood in man. Rev Sci Instrum 1942; 13: 434-444.
10. Aoyagi T, Kishi M, Yamaguchi K, Watanabe S. Improvement of the ear piece oximeter. 13th Meeting of the Japanese Society for Medical electronics and Biological Engineering 1974; 90-91 (abstract).
11. Severinghaus JW, Honda Y. History of blood gas analysis: VII. Pulse oximetry. J Clin Monit 1987; 3: 135-138.
12. Moyle JTB. Pulse Oximetry. London, BMJ Publishing Group, 1994.
13. Datex-Ohmeda. Biox 3740. Users Manual.
14. Langton J, Hanning CD. Effect of motion artefact on pulse oximeters. Br J Anaesth 1990; 65: 564-570.
15. Ralston AC, Webb RK, Runciman WB. Potential errors in pulse oximetry. I. Pulse oximeter evaluation. Anaesthesia 1991; 46: 202-210.
16. Nickerson BG, Sarkisian C, Tremper KK. Bias and precision of pulse oximeters and arterial oximeters. Chest 1988; 93: 515-517.
17. Morris RW, Nairn M, Torda TA. A comparison of fifteen pulse oximeters. Part I: a clinical comparison; part II: a test of performance under conditions of poor perfusion. Anaesth Intensive Care 1989; 17: 62-73.
18. Choe H, Tashiro C, Fukumitsu K, Yagi M, Yoshiya I. Comparison of recorded values from six pulse oximeters. Crit Care Med 1989; 17: 678-682.
19. Webb RK, Ralston AC, Runciman WB. Potential errors in pulse oximetry. II. Effects of changes in saturation and signal quality. Anaesthesia 1991; 96: 207-212.
20. Chiappini F, Fuso L, Pistelli R. Accuracy of a pulse oximeter in the measurement of the oxyhaemoglobin saturation. Eur Respir J 1998; 11: 716-719.
21. Severinghaus JW, Naifeh KH, Koh SO. Errors in 14 pulse oximeters during profound hypoxaemia. J Clin Monit 1989; 5: 72-81.
22. Hannhart B, Haberer J-P, Saunier C, Laxenaire M-C. Accuracy and precision of fourteen pulse oximeters. Eur Respir J 1991; 4: 115-119.
23. Hannhart B, Michalski H, Delorme N, Chapparo G, Polu JM. Reliability of six pulse oximeters in chronic obstructive pulmonary disease. Chest 1991; 99: 842-846.
24. Taylor MB, Whitman JG. The accuracy of pulse oximeters: a comparative clinical evaluation of five pulse oximeters. Anaesthesia 1988; 43: 229-232.
25. Carter BG, Carlin JB, Tibballs J, Mead H, Hochmann M, Osbourne A. Accuracy of two pulse oximeters at low arterial hemoglobin-oxygen saturations. Crit Care Med 1998; 26: 1128-1133.
26. Jensen LA, Onyskiw JE, Prasad NG. Meta-analysis of arterial oxygen saturation monitoring by pulse oximetry in adults. Heart Lung 1998; 27: 387-408.
27. Clayton DG, Webb RK, Ralston AC, Duthie D, Runciman WB. Pulse oximeter probes: a comparison between finger, nose, ear, and forehead probes under conditions of poor perfusion. Anaesthesia 1991; 46: 260-265.
28. Severinghaus JW, Naifeh KH, Koh SO. Accuracy of response of six pulse oximeters to profound hypoxia. Anesthesiology 1987; 67: 551-558.
29. Kagle DM, Alexander CM, Berko RS, Giuffre M, Gross JB. Evaluation of the Ohmeda 3700 pulse oximeter: steady-state and transient response characteristics. Anesthesiology 1987; 66: 376-380.
30. West P, George CF, Kryger MH. Dynamic in vivo response characteristics of three oximeters. Hewlett Packard 47201 A, Biox III, and Nellcor N-100. Sleep 1987; 10: 263-271.
31. Young D, Jewkes C, Spittal M, Blogg C, Weismann J, Gradwell D. Response time of pulse oximeters assessed using acute decompression. Anesth Analg 1992; 74: 189-195.
32. Trivedi NS, Ghouri AF, Lai E, Shah NK, Barker SJ. Pulse oximeter performance during desaturation and resaturation: a comparison of seven models. J Clin Anesthesia 1997; 9: 184-188.
33. Tobin MJ. Respiratory monitoring. JAMA 1990; 264: 244-251.
34. Kendrick AH, Wiltshire N, Catterall JR. Effect of signal averaging time (TSA) on on-line pulse oximetry used for overnight sleep recordings. Am J Respir Crit Care Med 1996; 153: A714.
35. Farre R, Montserrat JM, Ballester E, Hernandez L, Rotger M, Navajas D. Importance of the pulse oximeter averaging time when measuring oxygen desaturation in sleep apnea. Sleep 1998; 21: 386-390.
36. Kendrick AH, Wiltshire N, Catterall JR. Comparison of the Minolta Pulsox-5 and Ohmeda Biox 3740 pulse oximeters for sleep studies. Am J Respir Crit Care Med 1996; 153: A714.
37. Moller JT, Johannessen NW, Espersen K, et al. Randomized evaluation of pulse oximetry in 20, 802 patients: II. Perioperative events and post operative complications. Anesthesiology 1993; 78: 445-453.
38. Plummer JL, Zakaria AZ, Ilsley AH, Fronsko RR, Owen H. Evaluation of the influence of movement on saturation readings from pulse oximeters. Anaesthesia 1995; 50: 423-426.
39. Barker SJ, Shah NK. The effects of motion on the performance of pulse oximeters in volunteers. Anaesthesiology 1997; 86: 101-108.
40. Plummer JL, Ilsley AH, Fronsko RR, Owen H. Identification of movement artefact by the Nellcor N-200 and N-3000 pulse oximeters. J Clin Monit 1997; 13: 109-113.
41. Pollard V, Prough DS. Signal extraction technology: a better mousetrap? Anesth Analg 1996; 83: 213-214.
42. Trivedi NS, Shah NK, Jacobson BP. New pulse oximeter: resistant to motion artefacts. Anesthesiology 1995; 80: S510.
43. Dumas C, Wahr JA, Tremper KK. Clinical evaluation of a prototype motion artefact resistance pulse oximeter in the recovery room. Anesth Analg 1996; 83: 269-272.
44. Barker SJ, Hyatt J, Shah NK, Kao YJ. The effect of sensor malpositioning on pulse oximeter accuracy during hypoxaemia. Anesthesiology 1993; 79: 248-254.
45. Wilkins CJ, Moores M, Hanning CD. Comparison of pulse oximeters: effects of vasoconstriction and venous engorgement. Br J Anaesth 1989; 62: 439-444.
46. Palve H, Vuori A. Accuracy of three pulse oximeters at low cardiac index, and peripheral temperature. Crit Care Med 1991; 19: 560-562.
47. Falconer RJ, Robinson BJ. Comparison of pulse oximeters: accuracy at low arterial pressure in volunteers. Br J Anaesth 1990; 65: 552-557.
48. Clayton D, Webb RK, Ralston AC, Duthie D, Runciman WB. A comparison of the performance of 20 pulse oximeters under conditions of poor perfusion. Anaesthesia 1991; 46: 3-10.
49. Secker C, Spiers P. Accuracy of pulse oximetry in patients with low systemic vascular resistance. Anaesthesia 1997; 52: 127-130.
50. Villanueva R, Bell C, Kain ZN, Colingo KA. Effect of peripheral perfusion on accuracy of pulse oximetry in children. J Clin Anesth 1999; 11: 317-322.
51. Lawson D, Norley I, Korbon G, Loeb R, Ellis J. Blood flow limits and pulse oximeter signal detection. Anesthesiology 1987; 67: 599-603.
52. Tremper KK, Hufstedler SM, Barker SJ. Accuracy of a pulse oximeter in the critically ill adult: effect of temperature and haemodynamics. Anesthesiology 1985; 63: 175.
53. Palve H, Vuori A. Pulse oximetry during low cardiac output and hypothermia states immediately following open heart surgery. Crit Care Med 1989; 17: 66-69.
54. Bell C, Luther MA, Nicholson JJ, Fox CJ, Hirsh JL. Effect of probe design on accuracy and reliability of pulse oximetry in pediatric patients. J Clin Anesth 1999; 11: 323-327.
55. Scheller MS, Unger RJ, Kelner MJ. Effects of intravenously administered dyes on pulse oximeter readings. Anesthesiology 1986; 65: 550-552.
56. Saunders NA, Powles ACP, Rebuck AS. Ear oximetry: accuracy and predictability in the assessment of arterial oxygenation. Am Rev Respir Dis 1976; 113: 745-749.
57. Ries AL, Prewitt LM, Johnson JJ. Skin colour and ear oximetry. Chest 1989; 96: 287-290.
58. Jubran A, Tobin MJ. Reliability of pulse oximetry in titrating supplemental oxygen therapy in ventilator-dependent patients. Chest 1990; 97: 1420-1425.
59. Cote CJ, Goldstein EA, Fuchsman WH, Hoaglin DJ. The effect of nail polish on pulse oximetry. Anesth Analg 1988; 67: 683-686.
60. Battito MF. The effect of fingerprinting ink on pulse oximetry. Anesth Analg 1989; 69: 265.
61. Rubin AS. Nail polish can affect pulse oximeter saturation. Anesthesiology 1988; 68: 825.
62. Peters SM. The effect of acrylic nails on the measurement of oxygen saturation as determined by pulse oximetry. AANA Journal 1997; 65: 361-363.
63. Hanowell L, Eisele JH, Donns D. Ambient light affects pulse oximeters. Anesthesiology 1987; 67: 864-865.
64. Amar D, Neidzwski J, Wald A, Finck AD. Fluorescent light interferes with pulse oximetry. J Clin Monit 1989; 5: 135-136.
65. Levy H, Servilla M, Simpson SQ. Oximeter malfunction. Chest 1995; 105: 975.
66. Vegfors M, Lannmarken C. Carboxyhaemoglobinaemia and pulse oximetry. Br J Anaesth 1991; 66: 625-626.
67. Eisenkraft JB. Pulse oximeter desaturation due to methemoglobinaemia. Anesthesiology 1988; 68: 279-282.
68. Reider HU, Frei FJ, Zbinden AM, Thomson DA. Pulse oximetry in methaemoglobinaemia. Anaesthesia 1989; 44: 326-327.
69. Veyckemans F, Baele P, Guillaume JE, Willems E, Robert A, Clerbaux T. Hyperbilirubinemia does not interfere with haemoglobin saturation as measured by pulse oximetry. Anesthesiology 1989; 70: 118-122.
70. Chelluri L, Snyder JV, Bird JR. accuracy of a pulse oximetry in patients with hyperbilirubinemia. Respir Care 1991; 36: 1383-1386.
71. Jay GD, Hughes L, Renzi FP. Pulse oximetry is accurate in acute anemia from hemorrhage. Ann Emerg Med 1994; 24: 32-35.
72. Weston-Smith SG, Glass UH, Acharya J, Pearson TC. Pulse oximetry in sickle cell disease. Clin Lab Haematol 1989; 11: 185-188.
73. Homi J, Levee L, Higgs D, Thomas P, Serjeant G. Pulse oximetry in a cohort study of sickle cell disease. Clin Lab Haemotology 1997; 19: 17-22.
74. Kress JP, Pohlman AS, Hall JB. Determination of hemoglobin saturation in patients with acute sickle chest syndrome. A comparison of arterial blood gases and pulse oximetry. Chest 1999; 115: 1316-1320.
75. Comber JT, Lopez BL. Evaluation of pulse oximetry in sickle cell anemia patients presenting to the emergency department in acute vaso-occlusive crisis. Am J Emerg Med 1996; 14: 16-18.
76. Ballas SK, Smith ED. Red blood cell changes during the evaluation of the sickle cell crisis. Blood 1992; 15: 2154-2163.
77. Keidan I, Sidi A, Gravenstein D. False low pulse oximetry reading associated with the concomitant use of a peripheral nerve stimulator and an evoke-potential stimulator. J Clin Anesthesia 1997; 9: 591-596.
78. Farber NE, McNeely J, Rosner D. Skin burns associated with pulse oximetry during perioperative photodynamic therapy. Anesthesiology 1996; 84: 983-986.
79. Radu A, Zellweger M, Grosjean P, Monnier P. Pulse oximeter as a cause of skin burn during photodynamic therapy. Endoscopy 1999; 31: 831-833.
80. Hutton P, Clutton-Brock T. The benefits and pitfalls of pulse oximetry. BMJ 1993; 307: 457-458.
81. Stoneham MD, Saville GM, Wilson IH. Knowledge about pulse oximetry among medical and nursing staff. Lancet 1994; 344: 1339-1342.
82. Shellock FG, Slimp GL. Severe burn of the finger caused by using a pulse oximeter during MR imaging. Am J Roentgenol 1989; 153: 1105.
83. Brown TR, Goldstein B, Little J. Severe burns resulting from magnetic resonance imaging with cardiopulmonary monitoring. Risks and relevant safety precautions. Am J Physical Med Rehab 1993; 72: 166-167.
84. Tsien CL, Fackler JC. Poor prognosis for existing monitors in the intensive care unit. Crit Care Med 1997; 25: 614-619.
85. Campbell EJM, Howell JBL. Simple rapid methods of estimating arterial and mixed venous PCO2. BMJ 1960; i: 458-462.
86. Brewis RAL. Lecture Notes on Respiratory Disease. 4th Edn. Oxford, Blackwell Scientific, 1991; 64.
87. McEvoy JDS, Jones NL, Campbell EJM. Mixed venous and arterial PCO2. BMJ 1974; iv: 687-690.
88. MacFadden ER, Lyons HA. Arterial-blood gas tensions in asthma. N Engl J Med 1868; 278: 1027-1032.
89. Miyamoto DA, Mizuno K, Furuya K. Arterial blood gases in bronchial asthma. J Allergy 1970; 45: 248-254.
90. Palmer KN, Kelman GR. Pulmonary function in asthmatic patients in remission. BMJ 1975; i: 485-486.
91. Moon RE, Camporesi EM. Respiratory monitoring. In: Miller RD, ed. Anesthesia. 4th Edn. New York, Churchill Livingstone, 1994; pp. 1253-1291.
92. Gothard JWW, Busst CM, Branthwaite MA, Davies NJH, Denison DM. Applications of respiratory mass spectrometry to intensive care. Anaesthesia 1980; 35: 890-895.
93. Marini JJ. Monitoring during mechanical ventilation. Clin Chest Med 1988; 9: 73-84.
94. Healey CJ, Fedullo AJ, Swinburne AJ, Wahl GW. Comparison of non-invasive measurements of carbon dioxide tension during withdrawal from mechanical ventilation. Crit Care Med 1987; 15: 764-769.
95. Carroll GC. A continuous monitoring technique for management of acute pulmonary failure. Chest 1987; 92: 467-471.

Back to contents