Spectrophotometry

Spectrophotometry uses the measurement of light absorbance to determine the concentrations of various solutes in clear solutions. Carl Matthes used this technique to determine hemoglobin oxygen saturation as early as the 1930s.⁴ The measurement is based on the Lambert-Beer law, which relates solute concentrations to the intensity of light transmitted through a solution:

(11-7)

I_trans represents the intensity of light transmitted through solution, I_in is the intensity of incident light, d is the distance light is transmitted through the solution (optical path length), C is the concentration of the solute, and ε is the extinction coefficient of the solute.

The extinction coefficient ε determines the tendency of a given solute to absorb light at a specific wavelength or color. Equation 11-7 shows it as the natural logarithm of the ratio I_in/I_trans for a solution whose concentration C and thickness d are both unity. The extinction coefficient is a known constant for a given solute at a specified wavelength. If a solute of known extinction coefficient (ε) is in solution in a cuvette, a transparent measurement chamber of known dimensions, the solute concentration can then be calculated using Equation 11-7 and the measured intensities of incident and transmitted light. If multiple solutes are present, the exponent in Equation 11-7 is simply the sum of similar expressions for each solute; for example, if there are two solutes, it is dC₁ε₁ + dC₂ε₂. The extinction coefficients of the four most common hemoglobin species in the red to IR wavelength range are shown in Figure 11-4.

Laboratory in vitro blood oximeters, usually called CO- oximeters, use spectrophotometry to determine the concentrations of several common hemoglobin species by measuring light transmitted through a cuvette filled with a hemoglobin suspension produced from lysed red blood cells.¹ The analysis above assumes 1) that both the solvent and the cuvette are transparent at the wavelengths used, 2) the light path length is exactly known, and 3) no unknown light-absorbing substances are present in the solution. It is difficult to meet these requirements precisely in clinical devices; hence CO-oximeters theoretically based on the Lambert-Beer law require empirical corrections to improve accuracy.

Engineering Principles

Conventional pulse oximeters estimate arterial hemoglobin saturation by measuring the transmission of light at two wavelengths through a pulsatile vascular tissue bed. The pulse oximeter effectively uses the finger, ear, or other tissue as a “cuvette” containing hemoglobin. However, living tissue contains a number of light absorbers other than arterial hemoglobin; these include skin, soft tissue, bone, and venous and capillary blood. Early in vivo oximeters, such as Millikan’s, compensated for these additional tissue absorbances by compressing the soft tissues to eliminate all blood during a calibration cycle. The absorbance of the bloodless tissue was then used as a baseline. Some of these oximeters heated the tissue during measurement to render it hyperemic and thus obtain an absorbance more dependent on arterial blood.

The pulse oximeter distinguishes arterial blood from other light absorbers in the tissue in a novel way. As shown in Figure 11-5, light absorbance in tissue can be divided into a constant or direct current (DC) component and a pulsating or alternating current (AC) component. The AC component of absorbance is almost exclusively the result of arterial blood pulsations; this was Aoyagi’s original hypothesis. These pulsations are caused by the systolic volume expansion of the arteriolar bed, which produces an increase in optical path length and thereby increases the absorbance (Equation 11-7). Conventional pulse oximetry thus assumes that arterial blood is the only pulsatile absorber; any other fluctuating light absorbers will constitute sources of error. The consequences of and recent modifications to this assumption are discussed below (see Motion Artifact and Adaptive Digital Signal Filtering).

Most pulse oximeters use the same two wavelengths of light: 660 nm (red) and 940 nm (near IR). The pulse oximeter measures the AC component of the light absorbance at each wavelength and then divides it by the corresponding DC component (see Fig. 11-5), yielding the pulse-added absorbances S₆₆₀ = AC₆₆₀/DC₆₆₀ and S₉₄₀ = AC₉₄₀/DC₉₄₀. The pulse-added absorbances at the two wavelengths are independent of the intensity of the incident light. The oximeter then calculates the ratio (R) of the two pulse-added absorbances:

(11-8)

The Lambert-Beer law implies that in the absence of dyshemoglobins (e.g., COHb, MetHb), the ratio R is uniquely related to the arterial hemoglobin saturation. Although the pulse oximeter saturation estimate SpO₂ can be mathematically derived from the value of R via the Lambert-Beer law, the pulse oximeter uses an empirical calibration curve to relate SpO₂ to this ratio. A typical calibration is shown in Figure 11-6. This curve, as for all other pulse oximeter calibrations, is based on experimental data from human volunteers. The calibration data are stored in the oximeter’s microprocessor memory; pulse oximeters do not require user calibration because they assume one calibration algorithm for the entire human race. This does not imply that the instrument “calibrates itself for each patient,” as is sometimes erroneously stated.

To summarize these engineering principles, conventional two-wavelength pulse oximetry requires three implicit assumptions regarding human physiology: 1) that the two and only two light absorbers in human blood are oxyhemoglobin and reduced hemoglobin; 2) that all pulsations in light absorbance are caused by fluctuations in the local volume of arterial blood; and 3) that one empirical, experimental calibration curve—the relationship between the ratio R and SpO₂—is valid for the entire human race.

Dyshemoglobins and Intravenous Dyes: Conventional Pulse Oximetry

Because the conventional pulse oximeter measures light absorbance at two wavelengths, it can deal with unknown concentrations of only two solutes, that is, the two hemoglobin species O₂Hb and RHb. If any light-absorbing substance other than O₂Hb and RHb is present, violating the first assumption above, the pulse oximeter cannot accurately estimate saturation. As shown by the light absorbance spectra in Figure 11-4, both carboxyhemoglobin (COHb) and methemoglobin (MetHb) absorb light at one or both of the wavelengths commonly used by the pulse oximeter. Significant concentrations of either dyshemoglobin will therefore produce erroneous SpO₂ values. The fact that functional saturation (SaO₂) does not depend explicitly on dyshemoglobin concentrations (Equation 11-1) does not imply that SaO₂ can be determined by a two-wavelength oximeter. In the presence of additional hemoglobins, an oximeter cannot measure the concentrations of any hemoglobin species using only two wavelengths of light (Equation 11-7).

The effects of COHb on SpO₂ values were first determined experimentally in dogs in 1987.⁶ Figure 11-7 shows SpO₂ from conventional pulse oximetry, as well as fractional saturation O₂Hb% determined by in vitro CO-oximetry, plotted as functions of COHb%. Even when COHb% increases to levels greater than 70%, the displayed SpO₂ values remain greater than 90% at all times. COHb% levels above 50% are considered potentially lethal in humans. The pulse oximeter thus interprets COHb as though it were composed mostly of O₂Hb, a fact that can be predicted from the absorbance spectra in Figure 11-4. At the red wavelength of 660 nm, COHb has roughly the same absorbance as O₂Hb; at 940 nm, COHb is relatively transparent. This is consistent with the clinical observation that patients with carboxyhemoglobinemia exhibit a bright pink skin color.

The effects of MetHb on conventional SpO₂ values were similarly evaluated in animal experiments in 1989.⁷ Figure 11-8 shows SpO₂ and O₂Hb% determined by CO-oximetry and plotted as functions of MetHb%. As in the case of COHb, the presence of MetHb causes the pulse oximeter to overestimate fractional hemoglobin saturation. However, unlike the behavior of SpO₂ with COHb, here the SpO₂ values tend to decrease with increasing MetHb until they reach a plateau at approximately 85%. For MetHb% values greater than roughly 30%, there is little further decrease in SpO₂ (see Figure 11-8). This fact is again consistent with the light absorbance spectra in Figure 11-4, which shows that MetHb has high absorbance values at both wavelengths used by the pulse oximeter. This high absorbance, which tends to give MetHb its characteristic brown color, adds to both the numerator and denominator of the ratio R in Equation 11-8. Increasing both the numerator and denominator of this ratio by a fixed amount tends to drive the value of R toward 1.0. The pulse oximeter calibration curve of Figure 11-6 shows that an R value of 1.0 corresponds to an SpO₂ value of 85%. This may explain why the pulse oximeter tends to read near 85% saturation in the presence of high MetHb levels.⁷ A case report of methemoglobinemia caused by herbal medicines illustrates this phenomenon.⁸

Fetal hemoglobin (HbF) appears to have little effect on the accuracy of conventional pulse oximetry. This is because the extinction coefficients of HbF at the two usual pulse oximeter wavelengths, 660 and 940 nm, are similar to the corresponding values for adult hemoglobin (HbA). This is fortunate because the percentage of HbF present in neonatal blood varies with gestational age and is not very predictable. HbF also produces small errors in multiwavelength in vitro CO-oximeters. The oxygenated state of HbF is interpreted by some older laboratory oximeters as consisting partially of COHb.⁹

Theoretical considerations suggest that sickle hemoglobin (HbS) should also have little effect on pulse oximeter accuracy, but this is difficult to confirm experimentally. It would be unethical to subject sickle cell patients to intentional hypoxemia to study pulse oximeter accuracy. A few clinical case reports have involved sickle cell patients, either under healthy conditions or during a sickle cell crisis; however, these have produced conflicting results. One issue is that there is no clear gold standard for comparison of SpO₂ values in sickle cell patients. One study, which concluded that SpO₂ overestimated SaO₂ with a bias of 6.9%, used a CO-oximeter as the gold standard.¹⁰ However, laboratory CO-oximeters are generally designed to function in the presence of only four types of hemoglobin: 1) reduced Hb, 2) O₂Hb, 3) COHb, and 4) MetHb. Accuracy in the presence of HbS is not specified. Some studies have used as a standard the SaO₂ calculated from the PaO₂, which is measured by a blood gas electrode (Clarke electrode). This method is dubious because it must assume a “normal” oxygen-hemoglobin dissociation curve (see Fig. 11-2), when it is well known that sickle cell patients have right-shifted dissociation curves. At least two studies have measured the O₂ dissociation curves of individual sickle cell patients and then used these to calculate SaO₂ from PaO₂.^11,12 These studies concluded that pulse oximeter accuracy is maintained in sickle cell disease as long as differences in O₂ dissociation curves are accounted for. Other abnormal hemoglobins, such as hemoglobin Bassett, also have produced erroneous SpO₂ readings.¹³

The ratio R, and hence the SpO₂ value, can be affected by any substance present in the blood that absorbs light at 660 or 940 nm. Dyes injected intravenously for diagnostic purposes can have significant effects on SpO₂. For example, intravenous methylene blue produces sudden large decreases in SpO₂ values in normal subjects.¹⁴ Indigo carmine yields small decreases in SpO₂, and indocyanine green has an intermediate effect. The extinction coefficients of these three dyes are plotted against light wavelength in Figure 11-9.¹⁴ A comparison of these absorbance spectra with those of hemoglobin (see Fig. 11-4) predicts the relative effects of the three dyes on SpO₂ values. Bilirubin, another common pigment found in blood, appears to have no significant effect on SpO₂ at concentrations seen clinically.¹⁵

Nail polish has variable effects upon SpO₂ values and usually produces falsely low readings.¹⁶ Highly opaque, acrylic nail coverings can prevent the pulse oximeter from detecting any pulsatile absorbance at all. This problem can usually be averted by rotating the fingertip sensor 90 degrees so that the coated fingernail does not fall within the optical path.

Multiwavelength Pulse Oximetry

Until 2005, only a few investigations had been made into the possibilities of multiwavelength pulse oximetry. Aoyagi, the inventor of pulse oximetry, performed experiments with three-wavelength sensors in 2002.¹⁷ Although his purpose in using an additional wavelength was to improve the accuracy of the SpO₂ value, he noted that “Dyshemoglobins (e.g., COHb and MetHb) can be measured with a multiwavelength system.” A four-wavelength pulse oximeter that estimated total hemoglobin was reported by Noiri and colleagues in 2005.¹⁸ However, this device was never commercially produced and did not see widespread clinical use. Several investigators have used multiwavelength IR in vivo oximetry to estimate hemoglobin saturation in the internal jugular vein or in brain tissue, but again, only the ratio of oxyhemoglobin to total hemoglobin was measured.^19,20

In 2005, the Masimo Corporation (Irvine, CA) announced the development of their Rainbow technology Rad-57 pulse oximeter. This device uses eight wavelengths of light to measure SpO₂, SpCO (pulse oximeter estimate of COHb%), and SpMet (pulse oximeter estimate of MetHb%). The handheld, battery-powered Rad-57 was later followed by a bench-top version, Radical-7. The first human study of the performance of the Rad-57 was published in 2006.²¹ Twenty healthy volunteers were instrumented with radial artery catheters, electrocardiograms (ECGs), sphygmomanometers, and Rad-57 sensors on the fingers. In the COHb arm of the study, volunteers were exposed to inspired CO at 150 ppm until their COHb% levels reached 15%. Arterial blood samples were analyzed by CO-oximetry, and the resulting COHb% values were compared with the SpCO readings. The bias—that is, the mean difference between SpCO and COHb% by CO-oximeter—was −1.22%, and the precision (standard deviation [SD] of the difference) was 2.19%. These values are nearly the same as the manufacturer-specified uncertainties for conventional pulse oximeters in the measurement of SpO₂. This volunteer study obviously could not investigate the more pathologic COHb% levels found in smoke-inhalation injuries.

Subjects in the MetHb arm of the study were given 300 mg of intravenous sodium nitrite, a drug approved by the Food and Drug Administration (FDA) for treatment of cyanide toxicity; this resulted in methemoglobin levels of up to 13%. The comparison of SpMet with simultaneous CO-oximeter values of MetHb% yielded a bias of 0.00% and a precision of 0.45%. These uncertainties for SpMet are approximately one fourth of those measured for SpCO. This greater accuracy is a result of the fact that MetHb has a high light absorbance at all wavelengths in the range of interest; hence it provides a larger absorbance signal for the pulse oximeter’s processing algorithm.

Laboratory volunteer studies cannot predict a monitor’s performance in clinical conditions, nor can they establish its effects on patient outcome. For this we rely on clinical studies and case reports, a number of which have been published in the past few years. Coulange and colleagues²² measured SpCO from Rad-57 and simultaneous values of COHb% from blood CO-oximetry in smoke-inhalation victims. In data from 12 patients, they found a bias of −1.5% and a precision of 2.5%. Layne and colleagues measured SpCO and COHb% (Avox 400 CO- oximeter) in 130 outpatients, both smokers and nonsmokers.²³ They found a bias of −0.65% and a precision of 1.8% in a COHb range from 0% to 31%. Both of these clinical studies found SpCO uncertainties to be essentially the same as in the Barker volunteer study.²¹

Numerous case reports have been published on the use of Rad-57 to detect and measure COHb levels. Some authors have used SpCO as a screening tool to identify smokers and track smoking cessation.^24,25 Another promising application for the handheld device is in fire departments and for emergency medical services (EMS) responders. Hostler and colleagues²⁶ showed that SpCO could be used as a diagnostic tool for firefighters who respond to CO alarm activations. They found that patients who require transport to the emergency department had mean SpCO levels of 27.8%, whereas patients not in need of treatment had mean levels of 3.2%. A similar study of 149 patients by the New York Fire Department concluded that the Rad-57 “can be used as a screening tool to uncover cases of CO poisoning in which the diagnosis is not suspected.”²⁷ Firefighters themselves have been screened with SpCO in other investigations.²⁸ Case reports from emergency departments have shown the value of SpCO monitoring for tracking the progress of carbon monoxide poisoning victims and in the detection of CO poisoning in unsuspected cases.^29-31 Figure 11-10 shows an example of SpCO trend monitoring and response to treatment in an 81-year-old woman who was trapped in a house fire.²⁹

Several case reports were published of Rad-57 being used to detect and monitor methemoglobinemia. Many life-threatening methemoglobinemia cases reported involve the use of topical benzocaine.³² This is one of a large number of commonly used drugs that induce methemoglobinemia as a side effect, a problem that undoubtedly has been underdiagnosed in the past.³³ Macknet and colleagues presented a case of severe benzocaine-induced methemoglobinemia during transesophageal echocardiography, in which the Rad-57 was used to make the diagnosis and to monitor treatment with methylene blue.³⁴ Barker and colleagues reported a similar case in the OR following topical benzocaine for awake intubation.³⁵ In both of these cases, the multiwavelength pulse oximeter played a major role in diagnosis and treatment; in vitro laboratory CO-oximeter values were delayed too much to guide treatment.

The current versions of the Masimo Rad-57 and Radical-7 have two limitations. First, they still use a conventional two-wavelength red-over-IR algorithm to calculate the SpO₂ value. Thus, when levels of either COHb or MetHb are significant, the displayed SpO₂ is subject to the same errors described above.^6,7 Of course, the presence of a displayed SpCO or SpMet value would alert the user to the likelihood of this SpO₂ error. The second limitation is the existence of “crosstalk” between the MetHb and COHb measurement channels. That is, in the presence of significant MetHb levels, the instrument will display a falsely elevated SpCO value while also displaying a correct SpMet value. This situation is detected by the device, which will display an error message indicating that the SpCO value may not be accurate.

In March 2008, Masimo announced another innovation in multiwavelength pulse oximetry: the noninvasive measurement of total hemoglobin. Macknet and colleagues³⁶ published the first clinical validation study, in which 30 surgical patients and 18 healthy volunteers were monitored with the new prototype. The subjects followed a hemodilution protocol in which one unit of blood was withdrawn and replaced with normal saline. Arterial blood was sampled periodically and analyzed for total hemoglobin (Hbt) by a Radiometer ABL-735 CO-oximeter, and the values obtained were compared with the simultaneous Radical-7 measurements of hemoglobin (SpHb), as shown in Figure 11-11. In a range of Hbt values from 4.4 to 15.8 g/dL, this comparison revealed a bias and precision of 0.03 and 1.12 g/dL, respectively. Macknet and colleagues^36a published a similar study in 20 healthy volunteers and concluded that pulse CO-oximetry–based SpHb measurement was accurate to within 1.0 g/dL compared with laboratory CO-oximeter measurements.

A presentation at the 2008 World Congress of Anesthesiology described a case in which casual screening with the Radical-7 revealed an abnormally low SpHb (11 g/dL) in a 72-year-old man.³⁷ This led to further testing, which yielded the early diagnosis of an asymptomatic esophageal carcinoma. The early complete surgical excision resulted in the probable cure of a highly lethal disease. The noninvasive measurement of Hbt by pulse oximetry is very new, and more case reports and clinical outcome studies can be expected in the future.

Multiwavelength pulse oximetry is now capable of the continuous, noninvasive measurement of MetHb, carboxyhemoglobin, and total hemoglobin. Technology has made real progress in correcting one of the chief weaknesses of pulse oximetry: performance in the presence of abnormal hemoglobin species. Clinical studies and case reports have already documented the importance of these new developments, and it is hoped that further developments will lead to the continuous measurement of other substances in the blood.

Wavelength Variability

The LED used as a light source by the pulse oximeter is not an ideal monochromatic (single-wavelength) radiator. The LED emits light over a narrow but finite range of wavelengths. The center wavelength, or wavelength of peak energy radiation, varies by as much as 15 nm among diodes of the same specification. Figure 11-4 shows that such a variation in wavelength can yield a significantly different extinction coefficient, particularly at the 660-nm wavelength. Manufacturers use at least two approaches to this problem. The simplest method is to measure the center wavelength of all LEDs and reject those that fall outside a specified range, for example, 660 (±5) nm. This method is effective but expensive because of the large number of LEDs that must be discarded. The second method is to store multiple calibration algorithms in the pulse oximeter software, corresponding to different LED center wavelengths. The electrical connector on the sensor cable is then pin coded so that only the appropriate algorithm can be selected for a given sensor. Neither method entirely eliminates the effect of center frequency variation. This variability does not affect the pulse oximeter’s ability to follow changes in saturation, but it does produce differences between sensors in the absolute value of SpO₂.³⁸

Signal/Noise Ratio

The amplitude of the fluctuating (AC) component of the light absorbance can be much less than 1% of the amplitude of the DC component (see Fig. 11-5). Any influence that decreases the AC component, increases the DC component, or adds an artifactual AC component not related to arterial pulsations will worsen the signal/noise ratio. For example, the AC signal is decreased in low perfusion states, such as shock; the DC signal is increased when ambient room light reaches the detector; and artifactual AC signals are caused by patient motion, such as shivering.

The photodiode light detector in the pulse oximeter sensor cannot discriminate one wavelength of light from another—it is effectively color blind. The detector therefore responds to ambient room light and to light from either of the LED sources. In most two-wavelength pulse oximeters, this problem is solved by activating the red and infrared LEDs in an alternating sequence. During one part of this sequence, both LEDs are turned off, and the photodetector measures the ambient background light. This sequence is repeated many times per second (e.g., 480 Hz) in an attempt to eliminate light interference from rapidly changing ambient sources. Despite this ingenious design, ambient light artifact can create problems with the signal/noise ratio; however, this difficulty can be minimized by covering the sensor with an opaque shield of some sort, such as a surgical drape or towel.

If the peripheral pulse is weak, the AC absorbance signal becomes smaller compared with the DC signal. The pulse oximeter uses an automatic gain control that adjusts either the LED intensity or the photodetector amplifier gain to compensate for changes in AC signal amplitude. Unfortunately, this process also amplifies background noise from all sources, including ambient light. At the highest amplifier gain setting, the pulse oximeter may interpret background noise as a pulsatile absorbance and may generate an SpO₂ value from this artifact.⁴ This phenomenon could be demonstrated in early pulse oximeters by inserting a piece of paper between the photodetector and the LEDs. Some of these pulse oximeters would amplify background noise and display a pulse and saturation value from the paper.

This signal/noise ratio problem is also illustrated by the penumbra effect.³⁹ If a finger sensor is partially dislodged or malpositioned in such a manner that light passes through the fingertip at a grazing incidence, the pulse oximeter may display a correct heart rate but an erroneous SpO₂ value. The SpO₂ value from a malpositioned sensor is usually falsely low during normoxemia, but it may be falsely high during moderate hypoxemia.⁴⁰ This behavior may be another example of the R = 1.0 phenomenon (Equation 11-8).

Most pulse oximeters display some sort of visual indicator of pulsatile absorbance signal, usually in the form of an absorbance-versus-time plethysmogram. This displayed waveform usually represents the AC signal after amplification and therefore does not directly correspond to the actual amplitude of the absorbance pulsations. However, a few manufacturers display a waveform whose height actually does represent pulsatile absorbance amplitude before amplification. The user must therefore determine what the waveform of the pulse oximeter in use measures; in general, it is not safe to assume that waveform amplitude represents pulse amplitude. The plethysmograph is not a quantitative monitor of peripheral perfusion, and it cannot be relied upon to warn of impending ischemia, despite claims to the contrary.

The behavior of first-generation pulse oximeters during shock or low-perfusion states has been studied in both humans and animals.^41-45 During hemorrhagic shock, pulse oximeters may display no SpO₂ value at all, or they may give falsely low saturation estimates. Loss of signal is likely with hypothermia, severe anemia, low cardiac output states, and extremes in systemic vascular resistance. An early study of failure rates in the OR found that pulse oximeters (Nellcor N-100, Ohmeda 3700) failed in 1.12% of all patients.⁴⁶ Higher failure rates were associated with poor preoperative physical status, long operations, and elderly patients. A later study used computerized anesthesia records and found that 9% of all cases (n = 9203) had gaps in SpO₂ data of 10 minutes or more.⁴⁷ Higher failure rates were associated with higher ASA physical status number and with hypotension and hypothermia. The largest failure rate study was done by Moller and colleagues,⁴⁸ in which more than 20,000 surgical patients were followed. Interestingly, this controlled study found little difference in postoperative outcome between patients monitored with pulse oximetry and unmonitored patients, but it did find a very significant variation in failure rate with ASA physical status, shown in Figure 11-12. The failure rate for ASA-4 patients was more than five times that for ASA-1 patients; these higher failure rates in sicker patients have led some to describe the pulse oximeter as a “fair-weather friend.”

Several studies have determined thresholds for loss of signal during low-perfusion states. Lawson and colleagues⁴¹ produced gradual occlusion of blood flow with a blood pressure cuff while monitoring flow at the fingertip using a Doppler flow probe. They found loss of signal when blood flow had decreased to an average of 8.6% of its baseline value, which occurred at a cuff pressure of 96% of systolic pressure. On cuff deflation, the signal returned at a blood flow of only 4% of baseline. Severinghaus and colleagues⁴⁹ studied pulse oximeter behavior during various types of reduction in extremity blood flow, including via blood pressure cuff, brachial artery pressure clamp, and extremity elevation. Failure occurred at higher mean arterial pressures with the arterial clamp than with gravitational hypotension, showing the importance of blood volume pulsatility. These studies demonstrate that the pulse oximeter functions over a wide range of blood flows and blood pressures in the extremity. Because the pulse oximeter is designed to function independently of changes in flow or pressure, loss of signal cannot be used to determine the adequacy of peripheral perfusion, even though this has been attempted.⁵⁰ (See also the related discussion of the plethysmograph above.)

The electrosurgical unit (ESU) is another source of artifact. The intense, high-frequency, electromagnetic radiation from the ESU electrode fills the OR whenever the device is activated. The electrical cable from the pulse oximeter sensor acts as an antenna that receives the ESU radiation. Earlier generations of pulse oximeters were essentially shut down by ESU activation and required up to 30 seconds to recover. More recent pulse oximeters—that is, those made after the year 2000—are much improved in this respect and usually continue to display and update SpO₂ and heart rate values during ESU use. However, some pulse oximeters display their last “valid” SpO₂ value for up to 30 seconds during loss-of-signal periods. In these instruments, the user can be certain that the pulse oximeter is measuring SpO₂ during ESU activation only if a reasonable plethysmograph waveform is observed.

Motion Artifact and Adaptive Digital Signal Filtering

Artifacts caused by patient motion have plagued pulse oximetry, particularly in the recovery room, intensive care unit (ICU), and emergency settings. Although loss of signal or erroneous readings as a result of motion are infrequent in the OR (1% to 2%), they can cause a false alarm incidence that exceeds 50% in recovery room and ICU settings.⁵¹ Patient motion, especially shivering, causes a large, fluctuating absorbance artifact that is incorrectly interpreted by the pulse oximeter algorithm. Much of this artifact results from venous blood volume pulsations created by the motion, thus violating the second assumption of conventional pulse oximetry, that the light path length is exactly known.

Until fairly recently, manufacturers used two approaches to this problem: increased signal averaging time and ECG synchronization. In the first approach, the value of the ratio R is stored on a beat-to-beat basis and averaged for several seconds. This running average is less sensitive to patient motion but is also slower to respond to sudden saturation changes. The reduction in the false-alarm rate resulting from motion is thus accompanied by a slower response to true alarms, which is particularly risky in pediatric or neonatal applications. The second approach, ECG pulse rate synchronization, was developed by Nellcor and applied in their N-200 pulse oximeter.⁵² Although this “C-lock” feature seemed to be a useful innovation, it was not particularly successful and was abandoned in later models.

A recent, more elegant solution to motion artifact is to determine the “noise signal” and subtract it from the total signal, leaving a noise-free signal from which to calculate the SpO₂. Masimo has used this “adaptive digital filtering” approach to develop an algorithm called signal extraction technology (SET), which improves pulse oximeter performance in the presence of motion artifact.^53-55 The method makes two assumptions, the first being that most of the noise associated with motion artifact is produced by pulsations in venous blood volume, which produce the SpO₂ errors described above. In fact, errors from venous pulsations were documented in an early volunteer study that showed that SpO₂ may decrease by 8% when the arm is moved from a raised to a dependent position.⁵⁶ The second assumption is that the saturation values from the venous pulsations will be less than arterial values, which seems rather obvious. The Masimo SET electronically scans all possible values of the ratio R that correspond to saturations of 0% through 100%, and it calculates the signal intensity at each possible R value, as shown in Figure 11-13. In this example of a discrete saturation transform curve, two intensity peaks occur at 79% and 97% saturation: the higher peak corresponds to the arterial pulsations and is used to calculate SpO₂, and the lower peak at 50% presumably represents the venous pulsations. The entire sequence is repeated once per second on the most recent 6 seconds of raw data. The displayed SpO₂ value thus represents a 6-second running average of arterial hemoglobin saturation, updated every second.

This new technology has been evaluated in two volunteer experiments in which motion was induced in one hand while the other hand served as a stationary control.^53,54 In both of these experiments, done 5 years apart, the latest version of Masimo SET was compared with the current versions of other manufacturers’ instruments, most of which claimed to be “motion resistant.” The results of the more recent study are summarized in Table 11-1. These findings are supported by several clinical studies conducted on patients in postoperative care units.^55,57,58 Additional studies will determine whether this approach yields improved performance in clinical settings of low perfusion, such as with shock or after cardiopulmonary bypass.

Accuracy Studies

Most pulse oximeter manufacturers claim an uncertainty of ±2% (1 SD) for SpO₂ values between 70% and 100%. This uncertainty increases to ±3% for SpO₂ values between 50% and 70%; no accuracy is specified for SpO₂ values below 50%. This implies that for saturations above 70%, the SpO₂ value should be within 2% of the actual saturation 68% of the time and within 4% (2 SDs) 95% of the time. These accuracy specifications have not changed since the mid-1980s. The seemingly generous 2% uncertainty is the direct result of the third assumption for pulse oximetry as stated above: one empirical calibration curve—that is, the relationship between the ratio R and SpO₂—fits the entire human race.

In the 1980s and 1990s, a number of studies of pulse oximeter accuracy were done that included both laboratory volunteer tests and clinical trials.^60,61 Three of the studies from this era are of special interest because they evaluated not only accuracy but also response times to sudden changes in hemoglobin saturation.^62-64 These studies found some significant errors in pulse oximeter calibrations, which were subsequently revised by the manufacturers. As a result of such “aftermarket” software revisions, seemingly identical pulse oximeters may actually perform differently. Experimental study reports must therefore specify the software version and the pulse oximeter manufacturer and model used.

The volunteer studies of Severinghaus and Naifeh compared the performance of various pulse oximeters during severe, transient desaturations.^62,64 Subjects underwent brief (45 seconds) desaturations to an O₂Hb% value of 40% to 70%. Pulse oximeter response times to these sudden saturation changes were much shorter for earlobe sensors than for finger sensors (Figure 11-14). The time for a 50% response to rapid desaturation, or “resaturation,” ranged from 10 to 20 seconds for the earlobe sensor; for the finger sensors, it varied between 24 and 50 seconds. A similar result was obtained in another study that compared response times of finger sensors to those of both earlobe sensors and reflectance sensors on the forehead.⁶⁵ Both of these studies showed wide variations among subjects in response times for finger sensors, reflecting a wide range of lung-to-finger circulation times. These physiologic time delays and their interpatient variability should be considered in the selection of sensor sites in clinical settings, in which SpO₂ can change rapidly (e.g., the OR). However, clinical studies have also found that finger sensors are the most reliable in obtaining SpO₂ values during periods of hemodynamic instability.⁶⁶

The pulse oximeter’s response to sudden saturation changes is also affected by the signal-averaging time of the instrument, which is often user selectable. The displayed SpO₂ value represents a running average of data obtained over the most recent averaging time period, which may range between 1 and 15 seconds. The SpO₂ value will respond more quickly to a sudden change in SaO₂ if a short averaging time is selected. On the other hand, if the signal/noise ratio is marginal or artifacts are present, such as with patient motion, a longer averaging time will yield better accuracy. The user must determine the most appropriate averaging time for the clinical setting. The default averaging time, which is the value applied when the power is turned on, varies among manufacturers; hence, the user must know the default value of the particular instrument in use.

Clinical studies of accuracy usually combine data from multiple patients to determine the SpO₂ uncertainty. This procedure yields a more pessimistic view of accuracy than would be obtained by studying results for individual patients. If a sensor is placed on a patient, and the SpO₂ value is 95%, the probability is 68% that the patient’s true saturation lies between 93% and 97% (±1 SD). On the other hand, if the displayed SpO₂ on that same patient decreases from 95% to 93%, the amount by which the saturation has decreased is more certain than the original absolute SpO₂ value. In other words, trend accuracy is greater than absolute value accuracy. Variability among patients is a price we pay for the convenience of having the pulse oximeter precalibrated with a universal algorithm (see the third pulse oximetry assumption). The calibration algorithm is a best fit of data from a large number of healthy adult volunteers. Alternatively, manufacturers could have chosen to require user calibration on each individual patient. This would have improved absolute accuracy, but it would have lost an important advantage of pulse oximetry, namely, the absence of a need for user calibration.

Plethysmograph Variability Applications

In 1873, Kussmaul proposed that respiratory variation of the pulse intensity is an important sign of various physiologic derangements, including pericardial tamponade.⁶⁷ Nearly 100 years later, Kussmaul’s observation led to the measured association between pulsus paradoxus and intravascular volume status.⁶⁸ This concept evolved into a relationship between “arterial pulse pressure” from the arterial cannula and fluid responsiveness, addressing the basic clinical question of whether the patient would benefit from additional intravascular fluids.^69,70 Numerous studies over the years have shown that the traditional “static” parameters, including central venous pressure and pulmonary artery occlusion pressure, are of limited value in answering this question. Perhaps “dynamic parameters” such as pulse-pressure variation can be more predictive.⁷¹

Meanwhile, other investigators had already extrapolated this pulse-variation principle to the dependence of pulse oximeter plethysmograph variations on intravascular volume.⁷² These early efforts to use pulse oximetry for the estimation of volume status were interesting but found little clinical use because they were neither sensitive nor specific. However, more recent advances in pulse oximeter signal processing have made this noninvasive possibility more promising. Masimo has developed a new parameter called the plethysmograph variability index (PVI), which is defined as

(11-9)

where PI_max is the maximum value of the perfusion index, the ratio of AC to DC absorbance signal during the respiratory cycle, and PI_min is the minimum value during the cycle. This parameter can be continuously displayed by the pulse oximeter, thus quantifying the respiratory variation of the plethysmograph amplitude. Higher values of PVI are suggestive of lower intravascular volume status and a stronger probability of a positive hemodynamic response to fluid infusion.

Clinical studies of PVI have shown that it correlates strongly with arterial pulse pressure variations during positive-pressure ventilation.⁷³ More importantly, recent studies have found that PVI is predictive of response to intravascular volume infusion, with a sensitivity and specificity superior to those of either central venous pressure or wedge pressure.^74,75 Figure 11-15 shows typical plethysmograph waveforms and PVI trend plots for patients who are fluid responders and nonresponders.⁷⁵ As intravascular fluids are infused, the PVI falls dramatically in fluid responders, but it changes little in the nonresponders.

Plethysmograph variability, like pulsus paradoxus, is sensitive to a number of physiologic factors other than volume status (Box 11-1). The astute clinician must consider all possible causes for changes in PVI. Additional clinical data must be used to distinguish hypovolemia from tamponade, bronchospasm, or pneumothorax for example. Nevertheless, early studies suggest that pulse oximeter plethysmograph variability has excellent potential for monitoring hemodynamics in OR and critical care settings.

Limitations of Pulse Oximetry

Since the pulse oximeter became a minimum standard of care in the OR in 1986,⁷⁶ it has been unethical to perform randomized controlled studies of its clinical effectiveness in that setting.⁷⁷ Before that time, several such studies were performed. Coté and colleagues⁷⁸ studied 152 pediatric patients during anesthesia and surgery. In half of these patients, the SpO₂ data were unavailable to the anesthesiologist. The study found that “major events,” defined as SpO₂ below 85% for more than 30 seconds, occurred significantly more often in the patients for whom the SpO₂ values were hidden, and most of these events occurred in patients younger than 2 years. On the other hand, none of these controlled studies demonstrated a statistically significant difference in patient outcomes—that is, morbidity and mortality—and this includes the Moller study⁴⁸ of more than 20,000 patients cited above. This merely shows the difficulty of performing outcomes studies in anesthesiology. When a study cannot disprove the null hypothesis of no difference between treatments, it does not mean that the null hypothesis is true. This basic fact of statistics is too often overlooked in interpretations of clinical studies.

Since becoming an intraoperative standard of care in the 1980s, the pulse oximeter also has had major impact in other clinical settings. Two studies monitored SpO₂ during transport from the OR to the recovery room; both found a high incidence of desaturations with SpO₂ values below 90% in patients who did not receive supplemental oxygen.^79,80 These studies strongly support a uniform policy of transport to the postanesthesia care unit (PACU) with supplemental oxygen for all patients. Similar studies suggest that SpO₂ should be continuously monitored in most patients in the recovery room. One study of postoperative pediatric patients showed no correlation between SpO₂ and a traditional postanesthesia score based on motor activity, respirations, blood pressure, mental status, and color.⁸¹ The authors concluded that postoperative pediatric patients should be monitored by pulse oximetry or be given supplemental oxygen regardless of their apparent state of wakefulness. Another study found that 14% of adult PACU patients had SpO₂ values below 90% and that significant desaturations were associated with obesity, extensive surgery, advanced age, and poor physical status.⁸² In fact, most patients are more likely to have hypoxemia in the PACU than in the OR. Recovery room patients no longer have a protected airway and are not receiving mechanical ventilation, yet they have not fully recovered from the depressant effects of anesthesia and surgery.

The pulse oximeter warns of developing hypoxemia, but at the increased inspired oxygen fractions typically used in the OR, SpO₂ may not provide an early warning of decreasing arterial oxygen tension. The oxyhemoglobin dissociation curve (see Fig. 11-2) shows that PaO₂ must decrease to less than 80 mm Hg before SaO₂ falls significantly. A good example of this limitation is in the detection of inadvertent endobronchial intubation.

Figure 11-16 shows data from four different oxygenation monitors plotted as functions of time for an animal undergoing general anesthesia at an fractional concentration of inspired oxygen (FiO₂) of 0.5.⁸³ In addition to SpO₂, the plot shows PaO₂ values from sequential arterial blood samples, transcutaneous oxygen tension (P_TCO₂), and oxygen tension from an intraarterial fiberoptic “optode” blood gas sensor (OpPO₂). At 0 minutes on the time axis, the endotracheal tube was advanced from the trachea into the left mainstem bronchus. Within 3 minutes, the PaO₂ decreased from 360 mm Hg to 120 mm Hg, and the P_TCO₂ and OpPO₂ values also fell significantly. However, SpO₂ never fell below 98% during the entire experiment. In this situation, the pulse oximeter provided no indication that an endobronchial intubation had taken place, whereas the other monitoring techniques all showed significant changes. Only when the experiment was repeated for FiO₂ values of 0.3 or less did the pulse oximeter display consistent SpO₂ decreases of 6% or more. This illustrates an important physiologic limitation of saturation monitoring: when increased FiO₂ is used, the PaO₂ value must decrease far below its baseline before the pulse oximeter will alert the clinician. Metaphorically speaking, the pulse oximeter is a sentry standing on the edge of the cliff of desaturation (see Fig. 11-2), and it gives no warning as we approach the edge; it only sounds an alarm when we have fallen off.

Finally, no discussion of a clinical device is complete without a description of the risks and complications of the monitoring process. Because the pulse oximeter is noninvasive and typically does not produce heat or radiation, these risks might be expected to be nonexistent. Because of human error, however, this is unfortunately not true. In a case report, a Physio-Control (Redmond, WA) pulse oximeter sensor was mistakenly connected to an Ohmeda (now GE Healthcare) instrument.⁸⁴ The two oximeters used the same electrical connector for the sensor, but the internal pin connections were different; this resulted in severe thermal burns to both the finger and the earlobe of a newborn infant. The lesson from this case is that compatibility between sensor and instrument must be ensured before a sensor is placed on a patient. This risk has been reduced in more recent designs, but the clinician should never assume that just because the plug fits, it is acceptable to interconnect sensors and instruments of different manufacturers.

The pulse oximeter is unquestionably the most important advance in oxygen monitoring since the development of the blood gas analyzer. It is the only oxygen monitor that provides continuous, real-time, noninvasive data on arterial oxygenation without the need for user calibration. Because it is noninvasive and almost risk free, the pulse oximeter should be used in all clinical settings in which risk of arterial hypoxemia is present. It is a minimum standard of care in the OR and is rapidly becoming a standard in most other critical care settings. Hypoxia remains the most common cause of anesthesia-related preventable death.^85,86 Therefore the prudent use of pulse oximetry is a fundamental responsibility of the anesthesiologist.

The pulse oximeter is not an ideal instrument because it is subject to both measurement error and physiologic limitations.^87-95 With an understanding of the physics and engineering principles of pulse oximetry as outlined in this chapter, the astute clinician will be aware of measurement errors and when they are likely to occur. The impact of these errors also can be minimized by obtaining appropriate supportive data. For example, a smoke inhalation victim should have carboxyhemoglobin levels determined by in vitro CO oximetry or be monitored with a multiwavelength pulse oximeter capable of measuring COHb.²¹ The pulse oximeter signal/noise ratio should be optimized by careful sensor application, prevention of ambient light from reaching the sensor, and consideration of alternative sensor sites in patients who are shivering or who have diminished peripheral pulses. The avoidance of hypoxia is a fundamental goal of every anesthesiologist, and an understanding of the physics and physiology of arterial saturation monitoring by pulse oximetry will help accomplish this task.

1. Brown L.J. A new instrument for the simultaneous measurement of total hemoglobin, % oxyhemoglobin, % carboxyhemoglobin, % methemoglobin, and oxygen content in whole blood. IEEE Trans Biomed Eng. 1980;27:132–138.

2. Stephen C.R., Slater H.M., Johnson A.L., et al. The oximeter: a technical aid for the anesthesiologist. Anesthesiology. 1951;12:541–555.

3. Knill R.L., Clement J.L., Kieraszewicz H.T., et al. Assessment of two noninvasive monitors of arterial oxygenation in anesthetized man. Anesth Analg. 1982;61:582–586.

4. Severinghaus J.W., Astrup P.B. History of blood gas analysis. VI. Oximetry, J Clin Monit. 1986;2:270–288.

5. Wukitsch M.W., Tobler D., Pologe J., et al. Pulse oximetry: an analysis of theory, technology and practice. J Clin Monit. 1988;4:290–301.

6. Barker S.J., Tremper K.K. The effect of carbon monoxide inhalation on pulse oximetry and transcutaneous PO₂. Anesthesiology. 1987;66:677–679.

7. Barker S.J., Tremper K.K., Hyatt J. Effects of methemoglobinemia on pulse oximetry and mixed venous oximetry. Anesthesiology. 1989;70:112–117.

8. Chui J., Poon W., Chan K., Chan A., Buckley T. Nitrite-induced methemoglobinaemia: aetiology, diagnosis, and treatment. Anaesthesia. 2005;60:496–500.

9. Cornelissen P.J.H., van Del W.C., de Jong P.A. Correction factors for hemoglobin derivatives in fetal blood as measured with the IL282 CO oximeter. Clin Chem. 1983;29:1555–1556.

10. Craft J.A., Alessandrini E., Kenney L.B., et al. Comparison of oxygenation measurements in pediatric patients during sickle cell crises. J Pediatr. 1994;124:93–95.

11. Rackoff W.R., Kunkel N., Silber J.H., et al. Pulse oximetry and factors associated with hemoglobin oxygen desaturation in children with sickle cell disease. Blood. 1993;81:3422–3427.

12. Weston Smith S.G., Glass U.H., Acharya J. Pearson TC: Pulse oximetry in sickle cell disease. Clin Lab Haematol. 1989;11:185–188.

13. Das A., Sinha S., Hoyer J. Hemoglobin Bassett produces low pulse oximeter and co-oximeter readings. Chest. 2007;131:1242–1244.

14. Scheller M.S., Unger R.J., Kelner M.J. Effects of intravenously administered dyes on pulse oximetry readings. Anesthesiology. 1986;65:550–552.

15. Veyckemans F., Baele P., Guillaume J.E., et al. Hyperbilirubinemia does not interfere with hemoglobin saturation measured by pulse oximetry. Anesthesiology. 1989;70:118–122.

16. Coté C.J., Goldstein E.A., Fuchsman W.H., et al. The effect of nail polish on pulse oximetry. Anesth Analg. 1988;67:683–686.

17. Aoyagi T., Miyasaka K. The theory and applications of pulse spectrophotometry. Anesth Analg. 2002;94:S93–S95.

18. Noiri E., Kobayashi N., Takamura Y., et al. Pulse total hemoglobinometer provides accurate noninvasive monitoring. Crit Care Med. 33, 2005. E2831

19. Petrov Y., Petrova I., Patrikeev I., Esenaliev R., Prough D. Multiwavelength optoacoustic system for noninvasive monitoring of cerebral venous oxygenation: a pilot clinical test in the internal jugular vein. Optics Letters. 2006;31:1827–1829.

20. Nelson L., McCann J., Loepke A., Wu J. Development and validation of a multiwavelength spatial domain near-infrared oximeter to detect cerebral hypoxia-ischemia. J Biomed Opt. 11(6), 2006. 064022

21. Barker S., Curry J., Redford D., Morgan S. Measurement of carboxyhemoglobin and methemoglobin by pulse oximetry. Anesthesiology. 2006;105:892–897.

22. Coulange M., Barthelemy A., Hug F., Thierry A., DeHaro L. Reliability of new pulse CO-oximeter in victims of carbon monoxide poisoning. Undersea Hyperb Med. 2005;32(2):1–5.

23. Layne T., Snyder C., Brooks D., Enjeti S. Evaluation of a new pulse CO-oximeter: non-invasive measurement of carboxyhemoglobin in the outpatient pulmonary lab and emergency departments. December 2006. Presented at the 52^nd Annual Meeting of the American Association for Respiratory Care (AARC), Las Vegas.

24. Hampson N., Ecker E., Scott K. Use of noninvasive pulse CO-oximeter to measure blood carboxyhemoglobin levels in bingo players. Resp Care. 2006;51(7):758–760.

25. Light A., Grass C., Pursley D., Krause J. Carboxyhemoglobin levels in smokers vs. non-smokers in a smoking environment. December 2006. Presented at the 52^nd Annual Meeting of the American Association for Respiratory Care (AARC), Las Vegas.

26. Hostler D., Roth R., Kaufman R., et al. The incidence of carbon monoxide poisoning during CO alarm investigations. Pre-Hospital Emergency Care. 2008;12(1):115.

27. Ben-Eli D., Peruggia J., McFarland J., et al. Detecting CO: FDNY studies prehospital assessment of COHb. JEMS. 2007;32(10):S36–S37.

28. Dickinson E., Mecham C., Thom S., Shofer F., Band R. The non-invasive carboxyhemoglobin monitoring of firefighters engaged in fire suppression and overhaul operations. Pre-Hospital Emergency Care. 2008;12(1):96.

29. Plante T., Harris D., Savitt J., et al. Carboxyhemoglobin monitored by bedside continuous CO-oximetry. J Trauma. 2007;63(5):1187–1190.

30. Li Z., Gao C., Huang X., Wang G., Ge H. Clinical analyses of 429 cases of acute CO poisoning. Anesthesiology. 107, 2007. A1490

31. Suner S., Partridge R., Sucov A., et al. Non-invasive pulse CO-oximetry screening in the emergency department identifies occult carbon monoxide toxicity. J Emerg Med. 2008;34(4):441–450.

32. Throm M., Stevens M., Hansen C. Benzocaine-induced methemoglobinemia in two patients: interdisciplinary collaboration, management, and near misses. Pharmacotherapy. 2007;27(8):1206–1214.

33. Ash-Bernal R., Wise R., Wright S. Acquired methemoglobinemia: a retrospective series of 138 cases at two teaching hospitals. Medicine. 2004;83(5):265–273.

34. Macknet M., Kimball-Jones P., Applegate R., Martin R., Allard M. Benzocaine induced methemoglobinemia after TEE. Respir Care. 52(11), 2007.

35. Barker S.J., Annabi E.H. Life-threatening methemoglobinemia detected by pulse oximetry: a case report. Anesth Analg. 2009;108(3):898–899.

36. Macknet M., Norton S., Kimball-Jones P., et al. Continuous non-invasive measurement of hemoglobin via pulse CO-oximetry. Anesth Analg. 105(6), 2007. S-108

36a. Macknet M.R., Allard M., Applegate R.L., II., Rook J. The accuracy of noninvasive and continuous total hemoglobin measurement by pulse CO-oximetry in human subjects undergoing hemodilution. Anesth Analg. 2010;111:1424–1426.

37. Allard M., Viljoen J. Casual screening of haemoglobin noninvasively positively affects a colleague’s future? A case report http://www.wca2008.com/images/abstractWCA/monitoring.pdf 14^th Annual World Congress of Anesthesiology; March 2008, Capetown, South Africa. Accessed online at.

38. Pologe J.A. Pulse oximetry: technical aspects of machine design. Int Anesthesiol Clin. 1987;25:137–153.

39. Kelleher J.F., Ruff R.H. The penumbra effect: vasomotion-dependent pulse oximeter artifact due to probe malposition. Anesthesiology. 1989;71:787–791.

40. Barker S.J., Hyatt J., Shah N.K., et al. The effect of sensor malpositioning on pulse oximeter accuracy during hypoxemia. Anesthesiology. 1993;79:248–254.

41. Lawson D., Norley I., Korbon G., et al. Blood flow limits and pulse oximeter signal detection. Anesthesiology. 1987;67:599–603.

42. Narang V.P.S. Utility of the pulse oximeter during cardiopulmonary resuscitation. Anesthesiology. 1986;65:239–240.

43. Nowak G.S., Moorthy S.S., McNiece W.L. Use of pulse oximetry for assessment of collateral arterial flow. Anesthesiology. 1986;64:527.

44. Skeehan T.M., Hensley F.A., Jr. Axillary artery compression and the prone position. Anesth Analg. 1986;65:518–519.

45. Barrington K.J., Ryan C.A., Finer N.N. Pulse oximetry during hemorrhagic hypotension and cardiopulmonary resuscitation in the rabbit. J Crit Care. 1986;1:242–246.

46. Freund P.R., Overand P.T., Cooper J., et al. A prospective study of intraoperative pulse oximetry failure. J Clin Monit. 1991;7:253–258.

47. Reich D.L., Timcenko A., Bodian C.A., et al. Predictors of pulse oximetry data failure. Anesthesiology. 1996;84:859–864.

48. Moller J.T., et al. Randomized evaluation of pulse oximetry in 20,802 patients. Anesthesiology. 1993;78:436–453.

49. Severinghaus J.W., Spellman M.J., Jr. Pulse oximeter failure thresholds in hypotension and vasoconstriction. Anesthesiology. 1990;73:532–537.

50. Graham B., Paulus D.A., Caffee H.H. Pulse oximetry for vascular monitoring in upper extremity replantation surgery. J Hand Surg. 1986;11A:687–692.

51. Lawless S.T. Crying wolf: false alarms in a pediatric intensive care unit. Crit Care Med. 1994;22:981–985.

52. Nellcor N-200 pulse oximetry note number 6: C-Lock ECG synchronization principles of operation. Hayward, California: Nellcor; 1988.

53. Barker S.J., Shah N.K. Effects of motion on the performance of pulse oximeters in volunteers. Anesthesiology. 1997;86:101–108.

54. Barker S.J. “Motion-resistant” pulse oximetry: a comparison of new and old models. Anesth Analg. 2002;95:967–972.

55. Dumas C., Wahr J.A., Tremper K.K. Clinical evaluation of a prototype motion artifact resistant pulse oximeter in the recovery room. Anesth Analg. 1996;83(2):269–272.

56. Kim J.M., Arakawa K., Benson K.T., et al. Pulse oximetry and circulatory kinetics associated with pulse volume amplitude measured by photoelectric plethysmography. Anesth Analg. 1986;65:1333–1339.

57. Bohnhorst B., Peter C.S., Poets C.F. Pulse oximeters’ reliability in detecting hypoxemia and bradycardia: comparison between a conventional and two new-generation oximeters. Crit Care Med. 2000;28(5):1565–1568.

58. Malviya S., Reynolds P.I., Voepel-Lewis T., et al. False alarms and sensitivity of conventional pulse oximetry versus the Masimo SET technology in the pediatric postanesthesia care unit. Anesth Analg. 2000;90(6):1336–1340.

59. Bland J.M., Altman D.G. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1:307–310.

60. Tremper K.K., Barker S.J. Pulse oximetry. Anesthesiology. 1989;70:98–108.

61. Severinghaus J.W., Kelleher J.F. Recent developments in pulse oximetry. Anesthesiology. 1992;76:1018–1038.

62. Severinghaus J.W., Naifeh K.H. Accuracy of response of six pulse oximeters to profound hypoxia. Anesthesiology. 1987;67:551–558.

63. Kagle D.M., Alexander C.M., Berko R.S., et al. Evaluation of the Ohmeda 3700 pulse oximeter: steady-state and transient response characteristics. Anesthesiology. 1987;66:376–380.

64. Severinghaus J.W., Naifeh K.H., Koh S.O. Errors in 14 pulse oximeters during profound hypoxia. J Clin Monit. 1989;5:72–81.

65. Barker S.J., Hyatt J. Forehead reflectance pulse oximetry: time response to rapid saturation change. Anesthesiology. 1990;73(3A):A544.

66. Barker S.J., Le N., Hyatt J. Failure rates of transmission and reflectance pulse oximetry for various sensor sites. J Clin Monit. 1991;7:102–103.

67. Kussmaul A. Über schwielige Mediastino-Peikarditis und den paradoxen Puls. Berl Klin Wochenschr. 1873;10:433–435. 445-449, 461-464

68. Cohn J.N., Pinkerton A.L., Tristani F.E. Mechanism of pulsus paradoxus in clinical shock. J Clin Investigation. 1967;46(11):1744–1755.

69. Michard F., Boussat S., Chemla D., et al. Relation between respiratory changes in arterial pulse pressure and fluid responsiveness in septic patients with acute circulatory failure. Am J Respir Crit Care Med. 2000;162:134–138.

70. Lopes M.R., Oliveira M.A., Pereira V.O., et al. Goal-directed fluid management based on pulse pressure variation monitoring during high-risk surgery: a pilot randomized controlled trial. Crit Care. 11, 2007. R100

71. Michard F. Changes in arterial pressure during mechanical ventilation. Anesthesiology. 2005;103:419–428.

72. Partridge B.L. Use of pulse oximetry as a non-invasive indicator of intravascular volume status. J Clin Monitoring. 1987;3:263–268.

73. Canneson M., Delannoy B., Morand A., et al. Does pleth variability index indicate the respiratory-induced variation in the plethysmogram and arterial pressure waveforms? Anesth Analg. 2008;106:1189–1194.

74. Canneson M., Desebbe O., Rosamel P., et al. Pleth variability index to monitor the respiratory variations in the pulse oximeter plethysmographic waveform amplitude and predict fluid responsiveness in the operating theatre. Br J Anaesth. 2008;101(2):200–206.

75. Canneson M., Slieker J., Desebbe O., et al. The ability of a novel algorithm for automatic estimation of the respiratory variations in arterial pulse pressure to monitor fluid responsiveness in the operating room. Anesth Analg. 2008;106:1195–1200.

76. American Society of Anesthesiologists. Standards for basic intraoperative monitoring. Anesthesia Patient Safety Newsletter. 1987. March:3

77. Eichhorn J.H., Cooper J.B., Cullen D.J., et al. Standards for patient monitoring during anesthesia at Harvard Medical School. JAMA. 1986;256:1017–1020.

78. Coté C.J., Goldstein E.A., Coté M.A., et al. A single-blind study of pulse oximetry in children. Anesthesiology. 1988;68:184–188.

79. Pullerits J., Burrows F.A., Roy W.L. Arterial desaturation in healthy children during transfer to the recovery room. Can J Anaesth. 1987;34:470–473.

80. Tyler I.L., Tantisira B., Winter P.M., et al. Continuous monitoring of arterial oxygen saturation with pulse oximetry during transfer to the recovery room. Anesth Analg. 1985;64:1108–1112.

81. Soliman I.E., Patel R.I., Ehrenpreis M.B., et al. Recovery scores do not correlate with postoperative hypoxemia in children. Anesth Analg. 1988;67:53–56.

82. Morris R.W., Buschman A., Warren D.L., Philip J.H., Reamer D.B. The prevalence of hypoxemia detected by pulse oximetry during recovery from anesthesia. J Clin Monit. 1988;4:16–20.

83. Barker S.J., Tremper K.K., Hyatt J., et al. Comparison of three oxygen monitors in detecting endobronchial intubation. J Clin Monit. 1988;4:240–243.

84. Murphy K.G., Segunda J.A., Rockoff M.A. Severe burns from a pulse oximeter. Anesthesiology. 1990;73:350–352.

85. Keenan R.L., Boyan C.P. Cardiac arrest due to anesthesia: a study of incidence and causes. JAMA. 1985;253:2373–2377.

86. Taylor G., Larson C.P., Jr., Prestwich R. Unexpected cardiac arrest during anesthesia and surgery: an environmental study. JAMA. 1976;236:2758–2760.

87. Chapman K.R., Liu F.L.W., Watson R.M., et al. Range of accuracy of two-wavelength oximetry. Chest. 1986;4:540–542.

88. Nickerson B.G., Sakrison C., Tremper K.K. Bias and precision of pulse oximeters and arterial oximeters. Chest. 1988;93:515–517.

89. Mihm F.G., Halperin B.D. Noninvasive detection of profound arterial desaturations using a pulse oximetry device. Anesthesiology. 1985;62:85–87.

90. Cecil W.T., Thorpe K.J., Fibuch E.E., et al. A clinical evaluation of the accuracy of the Nellcor N-100 and the Ohmeda 3700 pulse oximeters. J Clin Monit. 1988;4:31–36.

91. Fait C.D., Wetzel R.C., Dean J.M., et al. Pulse oximetry in critically ill children. J Clin Monit. 1985;1:232–235.

92. Boxer R.A., Gottesfeld I., Singh S., et al. Non-invasive pulse oximetry in children with cyanotic congenital heart disease. Crit Care Med. 1987;15:1062–1064.

93. Mok J., Pintar M., Benson L., et al. Evaluation of noninvasive measurements of oxygenation in stable infants. Crit Care Med. 1986;14:960–963.

94. Durand M., Ramanathan R. Pulse oximetry for continuous oxygen monitoring in six newborn infants. J Pediatr. 1986;109:1052–1055.

95. Yelderman M., New W. Evaluation of pulse oximetry. Anesthesiology. 1983;59:349–352.