CHAPTER 11

Pulse Oximetry
Steven J. Barker*

CHAPTER OUTLINE

OVERVIEW Multiwavelength Pulse Oximetry

Wavelength Variability
HEMOGLOBIN SATURATION AND OXYGEN
TRANSPORT Signal/Noise Ratio
Motion Artifact and Adaptive Digital
HISTORY OF PULSE OXIMETRY Signal Filtering
PHYSICS AND ENGINEERING OF PULSE CLINICAL APPLICATIONS: ACCURACY, RESPONSE,
OXIMETRY AND LIMITATIONS
Spectrophotometry Accuracy Studies
Engineering Principles Plethysmograph Variability Applications
SOURCES OF ERROR Limitations of Pulse Oximetry
Dyshemoglobins and Intravenous Dyes:
Conventional Pulse Oximetry

OVERVIEW HEMOGLOBIN SATURATION

AND OXYGEN TRANSPORT
No monitor of oxygenation has had as much impact on the
practice of anesthesiology as the pulse oximeter. Unknown The pulse oximeter provides a noninvasive estimate of
in the operating room (OR) before the 1980s, the pulse arterial hemoglobin saturation, a variable directly pro-
oximeter is now a minimum standard of care for all patients portional to the oxygen content of arterial blood. Two
who receive anesthetics, whether general, regional, or definitions of hemoglobin saturation are in current use.
local. Its operation requires no special skill or training, and The older definition, called functional saturation, or SaO2,
its use is noninvasive and therefore almost risk free. The is related to the concentrations of oxyhemoglobin (O2Hb)
pulse oximeter gives continuous, real-time estimates of and deoxygenated, or “reduced,” hemoglobin (RHb) as
arterial hemoglobin saturation, which can warn of hypox- follows:
emia from many causes, including loss of airway patency,
loss of oxygen supply, and increases in venous admixture.
SaO2 = [(O2 Hb)/(O2 Hb + RHb)] × 100 %  (11-1)
As illustrated in Figure 11-1, pulse oximetry provides a
monitor of the second of four stages of the oxygen trans-
port process. This important advance goes beyond the first Additional species of hemoglobin are often present in
stage of inspired gas monitoring, but it does not ensure the adult blood, including carboxyhemoglobin (COHb) and
third stage of adequate oxygen delivery to vital organs. methemoglobin (MetHb). This leads to the definition of
This chapter reviews the historic development of pulse fractional hemoglobin saturation, or O2Hb%, which is the
oximetry, its underlying physical and engineering princi- ratio of oxyhemoglobin to the total concentration of all
ples, and recent improvements such as artifact reduction hemoglobin species:
and multiwavelength processing. An understanding of
these principles will enable the reader to predict sources
of measurement error. Pulse oximeter accuracy, response, O2 Hb % = [O2 Hb/(O2 Hb + RHb + COHb + MetHb)] × 100 %
clinical applications, limitations, and future potential are  = (O2 Hb/Hb) × 100 % (11-2)
also discussed.

In this formula, Hb is the total hemoglobin, the sum of

*Dr. Barker is a member of the Scientific Advisory Board of Masimo all species present. Fractional saturation is sometimes
Corporation (Irvine, CA) and a community member of their Board of
Directors. He is not an employee or a paid consultant of Masimo or
called oxyhemoglobin fraction or oxyhemoglobin percent.1
any other company involved in pulse oximetry. Fractional arterial hemoglobin saturation is related to

256
 11  Pulse Oximetry 257

Noninvasive Invasive

Lungs FiO2 meter (fuel cell,

(inspired/expired gas) polarographic, paramagnetic)
Mass spectrometer
Raman spectrometer
Magnetoacoustic

Arterial blood Pulse oximeter Intraarterial optode

or Clarke electrode

Transcutaneous PO2 Organ surface PO2

Tissue
Niroscope (e.g., liver, brain)
(cells, mitochondria
cytochromes) Cerebral oximeter Intramuscular fiberoptic PO2
Skin color

Fiberoptic pulmonary artery

Venous blood
oximeter (SvO2)
FIGURE 11-1  n  The four stages of the oxygen transport system, showing that the pulse oximeter monitors oxygen at the level of the
arterial blood. Respired gas monitors can confirm only that oxygen is being delivered to the lungs, but the pulse oximeter also moni-
tors the function of the lungs in transporting this oxygen to the arterial blood. Pulse oximetry does not guarantee that oxygen is being
delivered to or used by the tissues; this can be determined only by monitors that function further down the oxygen transport chain.
FiO2, fractional concentration of inspired oxygen; PO2, oxygen tension; S v O2, mixed venous oxygen saturation.

the arterial oxygen content, CaO2, by the following 100

formula:
90

CaO2 = (1.37 × Hb [O2 Hb %/100 ])+ (0.003 × PaO2 ) (11-3) 80

O2 saturation of hemoglobin (%)

70
Here PaO2 is the arterial oxygen tension in millimeters
of mercury. The first expression in brackets in Equa- 60
tion 11-3 represents the oxygen bound to hemoglobin,
which under normal conditions (Hb = 15 g/dL and 50
O2Hb% = 98) equals approximately 20 mL oxygen per 40
100 mL blood. The second expression represents PO2 (mm Hg) O2 Sat (%)
�oxygen dissolved in plasma, which equals 0.3 mL per 30
100 mL for a PaO2 of 100 mm Hg. Plasma-dissolved 60 90
40 75
oxygen usually does not play a significant role in oxygen 20
27 50
transport. Equation 11-3 shows that arterial oxygen
10
content is directly proportional to both total hemoglo-
bin (Hb) and fractional saturation. O2Hb% and PaO2 0
are related by the oxyhemoglobin dissociation curve, 0 10 20 30 40 50 60 70 80 90 100 110
shown in Figure 11-2. Under normal conditions, this PO2 (mm Hg)
relationship predicts a hemoglobin saturation for adults FIGURE 11-2  n  The oxyhemoglobin dissociation curve. Hemoglo-
of 50% at a PaO2 of 27 mm Hg, 75% at a PaO2 of bin saturation is plotted as a function of arterial oxygen tension
40 mm Hg (the typical venous blood value), and 90% at (PaO2) in millimeters of mercury. Under normal conditions for
adults, a PaO2 of 27 mm Hg yields a saturation of 50% (P50). The
a PaO2 of 60 mm Hg. This normal dissociation curve is curve is shifted to the right by acidosis, hypercarbia, increases
shifted to the right by acidosis, hypercarbia, hyperther- in 2,3-diphosphoglycerate (2,3-DPG), and hyperthermia.
mia, and increases in 2,3-diphosphoglycerate (2,3-DPG)
concentration. Note that for PaO2 values greater than
90 mm Hg, O2Hb% is nearly independent of PaO2. The factor of 10 appears because CaO2 is measured in mil-
This saturation property of hemoglobin has important liliters per deciliter, whereas cardiac output is measured in
implications in the clinical interpretation of pulse oxim- liters per minute. The quantity of oxygen consumed per
eter data, discussed below. minute is then the difference between the arterial oxygen
The amount of oxygen delivered to the tissues by the delivery (O2del) and the venous oxygen return (O2ret):
arterial blood (O2del) is simply the product of the arterial
oxygen content (CaO2) and the cardiac output, or
VO2 = O2del − O2ret = (CaO2 − CvO2 ) (11-5)

O2del = CaO2 × Cardiac output × 10  (11-4) × Cardiac output × 10
258 PART III  Patient Monitors

100

95

90

85 Intubating
O2 saturation (%)

blowing
Airway
Even
Scope respiration
80

A Abdominal
R.L., 4 yrs Coughing - respiration
75 breath holding
Tonsillectomy
Open-drop technique
FIGURE 11-3  n  Ear oximeter hemoglobin satura-
Vinethene ether
tion plotted as a function of time. Graph for a
70 4-year-old child undergoing general anesthesia
for tonsillectomy under open-drop ether with no
supplemental oxygen. Note the significant
desaturation associated with breath holding
65 during induction of anesthesia. Saturation does
not return to its preinduction baseline value at
any time during the record. (From Steven RC,
A Breath holding Slater HM, Johnson AL, et€al: The oximeter: a tech-
60 nical aid for the anesthesiologist. Anesthesiology
11:40 AM 11:45 11:50 11:55 12:00 12:05 12:10 1951;12:548.)

Arterial oxygen content and mixed venous oxygen con- Glenn Millikan created the first lightweight ear oxime-
tent (CvO2) in this form of the Fick equation can be ter in the early 1940s for aviation research. Millikan coined
replaced by corresponding expressions in terms of hemo- the term oximeter to describe his device, which measured
globin saturation from Equation 11-3: hemoglobin saturation in pilots flying at high altitudes.
Similar devices developed in the 1940s were used by Wood
VO2 = 13.7 × Cardiac output × (O2 Hb % a − O2 Hb % v )/100 (11-6) and others in the OR. The first perioperative application
of an in€vivo oximeter to appear in the anesthesiology lit-
Subscripts a and v denote arterial and mixed venous frac- erature was published in 1951.2 Figure 11-3 shows a record
tional hemoglobin saturations, respectively. The small of the hemoglobin saturation obtained from an ear oxim-
contributions of plasma-dissolved oxygen have been eter plotted against time during a tonsillectomy. Even
neglected in this form of the Fick equation. Equation 11-6 though this record shows a dramatic fall in saturation dur-
describes the relationship between oxygen consumption, ing the induction of anesthesia, curiously described as
arterial and mixed venous hemoglobin saturations, total “breath holding” in the original figure, the device drew
hemoglobin, and cardiac output. An understanding of this little attention from anesthesiologists until much later.
equation is vital to the interpretation of data from pulse The lack of acceptance of early in€vivo oximeters was
oximeters and other oxygen monitors and to understand- mostly due to their serious limitations. They were deli-
ing their relationship to other hemodynamic variables. cate instruments that required technician support for
calibration and operation. In addition, the earpiece sen-
sor was large and cumbersome, and it produced enough
HISTORY OF PULSE OXIMETRY heat to cause occasional burns. Hewlett-Packard (HP)
made a major advance in ear oximetry in the 1970s, when
Although the pulse oximeter became a monitoring stan- it marketed a self-calibrating, eight-wavelength ear oxim-
dard in the OR in the 1980s, in€vivo oximeters date to the eter. The HP oximeter was reasonably accurate for intra-
1930s. In 1935, Carl Matthes developed the first instru- operative monitoring, but it was still burdened by the size
ment that measured hemoglobin oxygen saturation by and bulky nature of the sensor as well as the expense of
transilluminating tissue. Matthes’s device used two wave- the instrument.3 Although the HP oximeter became a
lengths of light, one visible and one infrared (IR), much like standard tool in pulmonary function laboratories, it had
the modern pulse oximeter. This instrument could follow virtually no impact in the OR.
saturation trends but was difficult to calibrate. J.R. Squires The first true pulse oximeter was invented by Takuo
developed a similar instrument that calibrated itself by Aoyagi in 1975.4 While investigating a method to measure
compressing the ear to eliminate blood, a technique used intravenous dye washout curves using light transmission
later in the first commercially marketed in€vivo oximeters. through the ear, Aoyagi discovered that his data on light
 11  Pulse Oximetry 259

Hemoglobin Extinction Curves

10 Red Infrared

Extinction coefficient
1.0 Methemoglobin

FIGURE 11-4  n  Extinction coefficient (ε) plotted versus light Oxyhemoglobin

wavelength in nanometers for the four most common
hemoglobin species: oxyhemoglobin (O2Hb), reduced
hemoglobin (RHb), carboxyhemoglobin (COHb), and 0.1 Reduced
methemoglobin (MetHb). The absorbance of COHb is hemoglobin
very similar to that of O2Hb in the visible red wave-
lengths. MetHb has a high absorbance over a broad
spectrum, giving it a characteristic brown color. (Modi- Carboxyhemoglobin
fied from Tremper KK, Barker SJ: Pulse oximetry: applica-
tions and limitations. In Advances in oxygen monitoring: .01
International Anesthesiology Clinics. Boston, 1987, Little, 600 640 680 720 760 800 840 880 920 960 1000
Brown, pp 155-175.) Wavelength  (nm)

absorbance versus time contained fluctuations caused by the in clear solutions. Carl Matthes used this technique to
arterial pulse. In dealing with this “artifact,” he discovered determine hemoglobin oxygen saturation as early as the
that the relative amplitudes of the fluctuations at the two 1930s.4 The measurement is based on the Lambert-Beer
light wavelengths varied with arterial hemoglobin satura- law, which relates solute concentrations to the intensity
tion. This fortuitous discovery led him to the creation of the of light transmitted through a solution:
first two-wavelength pulse oximeter, which was developed
and marketed by Nihon Kohden Corporation. However,  Itrans = Iin e( − dCε) (11-7)
Aoyagi’s oximeter used filtered light sources and fiberoptic
transmission cables between the instrument and the earlobe Itrans represents the intensity of light transmitted through
sensor, rendering it somewhat awkward for use in the OR. solution, Iin is the intensity of incident light, d is the dis-
The next breakthrough in technology came in the late tance light is transmitted through the solution (optical
1970s, when Scott Wilbur of the Biox Corporation (later path length), C is the concentration of the solute, and ε is
Ohmeda) developed the first ear sensor that used light- the extinction coefficient of the solute.
emitting diodes (LEDs) and solid-state photodetectors The extinction coefficient ε determines the tendency of
built into the sensor itself. The fiberoptic cables of Aoya- a given solute to absorb light at a specific wavelength or
gi’s pulse oximeters were thereby replaced by a thin elec- color. Equation 11-7 shows it as the natural logarithm of
trical cable.5 The accuracy of the pulse oximeter was also the ratio Iin/Itrans for a solution whose concentration C and
improved by the incorporation of digital microprocessors thickness d are both unity. The extinction coefficient is a
into the instrument. Further electronic improvements known constant for a given solute at a specified wavelength.
were made by Biox and Nellcor in the early 1980s, and If a solute of known extinction coefficient (ε) is in solution
the pulse oximeter was ready to take its place as a stan- in a cuvette, a transparent measurement chamber of known
dard perioperative monitor. dimensions, the solute concentration can then be calcu-
The success of Nellcor in marketing its N-100 pulse lated using Equation 11-7 and the measured intensities of
oximeter to anesthesiologists in the mid-1980s brought incident and transmitted light. If multiple solutes are pres-
these devices into the OR in large numbers. The instru- ent, the exponent in Equation 11-7 is simply the sum of
ment had now become fairly reliable and easy to use as similar expressions for each solute; for example, if there are
well as relatively inexpensive. It gained rapid acceptance two solutes, it is dC1ε1 + dC2ε2. The extinction coefficients
and became a standard of care in the OR by 1986. Today, of the four most common hemoglobin species in the red to
no anesthesiologist would feel comfortable inducing gen- IR wavelength range are shown in Figure 11-4.
eral anesthesia without a functioning pulse oximeter. An Laboratory in€vitro blood oximeters, usually called CO-
excellent review of the history of pulse oximetry and the oximeters, use spectrophotometry to determine the con-
development of blood gas analysis has been written by centrations of several common hemoglobin species by
Severinghaus and Astrup.4 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
PHYSICS AND ENGINEERING and the cuvette are transparent at the wavelengths used,
OF PULSE OXIMETRY 2) the light path length is exactly known, and 3) no
unknown light-absorbing substances are present in the
Spectrophotometry solution. It is difficult to meet these requirements pre-
cisely in clinical devices; hence CO-oximeters theoreti-
Spectrophotometry uses the measurement of light absor- cally based on the Lambert-Beer law require empirical
bance to determine the concentrations of various solutes corrections to improve accuracy.
 260 PART III  Patient Monitors

AC Absorption due to pulsatile arterial blood

Light absorption

Absorption due to nonpulsatile arterial blood

Absorption due to venous and capillary blood
FIGURE 11-5  n  Schematic of the light absorbances of
living tissue plotted versus time. The fixed direct
DC current (DC) absorbance results from solid tissues,
Absorption due to tissue venous and capillary blood, and nonpulsatile arte-
rial blood. The alternating current (AC) component
is caused by pulsations in the arterial blood volume.
(Modified from Ohmeda Pulse Oximeter Model 3700
Time service manual. Boulder, CO, 1986, Ohmeda, p 22.)

Engineering Principles 100

90
Conventional pulse oximeters estimate arterial hemoglo-
80
bin saturation by measuring the transmission of light at
two wavelengths through a pulsatile vascular tissue bed. 70
The pulse oximeter effectively uses the finger, ear, or 60

SpO2 (%)
other tissue as a “cuvette” containing hemoglobin. How- 50
ever, living tissue contains a number of light absorbers 40
other than arterial hemoglobin; these include skin, soft
30
tissue, bone, and venous and capillary blood. Early in€vivo
oximeters, such as Millikan’s, compensated for these 20
additional tissue absorbances by compressing the soft tis- 10
sues to eliminate all blood during a calibration cycle. The 0
absorbance of the bloodless tissue was then used as a 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4
baseline. Some of these oximeters heated the tissue dur- AC660/DC660
ing measurement to render it hyperemic and thus obtain R
AC940/DC940
an absorbance more dependent on arterial blood.
FIGURE 11-6  n  A typical pulse oximeter calibration algorithm, in
The pulse oximeter distinguishes arterial blood from which estimated oxygenation (SpO2) is plotted against the ratio
other light absorbers in the tissue in a novel way. As R from Equation 11-8. The value of R varies from roughly 0.4 at
shown in Figure 11-5, light absorbance in tissue can be 100% saturation to 3.4 at 0% saturation. Note that an R value of
divided into a constant or direct current (DC) component 1.0 corresponds to an SpO2 reading of 85%. This calibration
curve is a best fit of experimental data obtained on healthy vol-
and a pulsating or alternating current (AC) component. unteers. AC, alternating current; DC, direct current. (Modified
The AC component of absorbance is almost exclusively from Pologe JA: Pulse oximetry: technical aspects of machine
the result of arterial blood pulsations; this was Aoyagi’s design. In Tremper KK, Barker SJ, editors: Advances in oxygen mon-
original hypothesis. These pulsations are caused by the itoring: International Anesthesiology Clinics. Boston, 1987, Little,
systolic volume expansion of the arteriolar bed, which Brown, p 142.)
produces an increase in optical path length and thereby
increases the absorbance (Equation 11-7). Conventional Although the pulse oximeter saturation estimate SpO2
pulse oximetry thus assumes that arterial blood is the only can be mathematically derived from the value of R via
pulsatile absorber; any other fluctuating light absorbers the Lambert-Beer law, the pulse oximeter uses an empir-
will constitute sources of error. The consequences of and ical calibration curve to relate SpO2 to this ratio. A typi-
recent modifications to this assumption are discussed cal calibration is shown in Figure 11-6. This curve, as for
below (see Motion Artifact and Adaptive Digital Signal all other pulse oximeter calibrations, is based on experi-
Filtering). mental data from human volunteers. The calibration
Most pulse oximeters use the same two wavelengths of data are stored in the oximeter’s microprocessor mem-
light: 660 nm (red) and 940 nm (near IR). The pulse ory; pulse oximeters do not require user calibration
oximeter measures the AC component of the light absor- because they assume one calibration algorithm for the
bance at each wavelength and then divides it by the cor- entire human race. This does not imply that the instru-
responding DC component (see Fig. 11-5), yielding the ment “calibrates itself for each patient,” as is sometimes
pulse-added absorbances S660 = AC660/DC660 and S940 = erroneously stated.
AC940/DC940. The pulse-added absorbances at the two To summarize these engineering principles, conven-
wavelengths are independent of the intensity of the inci- tional two-wavelength pulse oximetry requires three
dent light. The oximeter then calculates the ratio (R) of implicit assumptions regarding human physiology: 1)
the two pulse-added absorbances: that the two and only two light absorbers in human blood
are oxyhemoglobin and reduced hemoglobin; 2) that all
 R = [AC660 /DC660 ]/[AC940 /DC940 ] (11-8) pulsations in light absorbance are caused by fluctuations
in the local volume of arterial blood; and 3) that one
The Lambert-Beer law implies that in the absence of empirical, experimental calibration curve—the relation-
dyshemoglobins (e.g., COHb, MetHb), the ratio R is ship between the ratio R and SpO2—is valid for the entire
uniquely related to the arterial hemoglobin saturation. human race.
 11  Pulse Oximetry 261

SOURCES OF ERROR 100%

SpO2
Given the physics and engineering principles of pulse 80
oximetry, as well as the three assumptions noted above,

SpO2 and O2Hb%

the major sources of error in SpO2 readings are predict-
able. This section examines the common sources of error 60
as well as some design approaches and new developments
aimed at reducing errors. The wise user of pulse oximetry 40
must be aware of these problems and be able to anticipate
erroneous data.
20 FiO2  1.0 O2Hb%

Dyshemoglobins and Intravenous Dyes:

0
Conventional Pulse Oximetry 0 20 40 60 80 100%
Because the conventional pulse oximeter measures light COHb%
absorbance at two wavelengths, it can deal with unknown FIGURE 11-7  n  Estimated oxygen saturation (SpO2) and fractional
saturation (O2Hb%) plotted against carboxyhemoglobin level
concentrations of only two solutes, that is, the two (COHb%) for dogs inhaling carbon monoxide at 200 ppm. SpO2
hemoglobin species O2Hb and RHb. If any light-absorb- seriously overestimates arterial fractional hemoglobin oxygen
ing substance other than O2Hb and RHb is present, vio- saturation in the presence of COHb and remains greater than
lating the first assumption above, the pulse oximeter 90% even for a COHb% of 70. The pulse oximeter “sees” COHb
cannot accurately estimate saturation. As shown by the as though it were mostly O2Hb. (From Barker S, Tremper K: The
effect of carbon monoxide inhalation on pulse oximetry and transcu-
light absorbance spectra in Figure 11-4, both carboxyhe- taneous PO2. Anesthesiology 1987;66:677-679.)
moglobin (COHb) and methemoglobin (MetHb) absorb
light at one or both of the wavelengths commonly used
by the pulse oximeter. Significant concentrations of 100
either dyshemoglobin will therefore produce erroneous
SpO2 values. The fact that functional saturation (SaO2)
90
does not depend explicitly on dyshemoglobin concentra- SpO2
tions (Equation 11-1) does not imply that SaO2 can be
determined by a two-wavelength oximeter. In the pres-
80
ence of additional hemoglobins, an oximeter cannot
measure the concentrations of any hemoglobin species
SpO2 and HbO2%

using only two wavelengths of light (Equation 11-7).

70
The effects of COHb on SpO2 values were first deter-
mined experimentally in dogs in 1987.6 Figure 11-7
shows SpO2 from conventional pulse oximetry, as well as 60 O2Hb%
fractional saturation O2Hb% determined by in€vitro CO-
oximetry, plotted as functions of COHb%. Even when
COHb% increases to levels greater than 70%, the dis- 50
played SpO2 values remain greater than 90% at all times.
COHb% levels above 50% are considered potentially
lethal in humans. The pulse oximeter thus interprets 40
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 30
has roughly the same absorbance as O2Hb; at 940 nm, 0 20 40 60
COHb is relatively transparent. This is consistent with MetHb%
the clinical observation that patients with carboxyhemo- FIGURE 11-8  n  Estimated oxygen saturation (SpO2) and fractional
globinemia exhibit a bright pink skin color. saturation (O2Hb%) vs. methemoglobin level (MetHb%) for dogs
The effects of MetHb on conventional SpO2 values with benzocaine-induced methemoglobinemia. Although SpO2
shows a downward trend with increasing MetHb%, O2Hb% is con-
were similarly evaluated in animal experiments in 1989.7 sistently overestimated, and it appears that a plateau is reached at
Figure 11-8 shows SpO2 and O2Hb% determined by an SpO2 of 85%. When fractional concentration of inspired oxy-
CO-oximetry and plotted as functions of MetHb%. As gen (FiO2) is varied during this experiment, SpO2 measures nei-
in the case of COHb, the presence of MetHb causes the ther functional nor fractional saturation. (From Barker SJ, Tremper
pulse oximeter to overestimate fractional hemoglobin KK, Hyatt J: Effects of methemoglobinemia on pulse oximetry and
mixed venous oximetry. Anesthesiology 1989;70:112-117.)
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 approxi- has high absorbance values at both wavelengths used by
mately 85%. For MetHb% values greater than roughly the pulse oximeter. This high absorbance, which tends
30%, there is little further decrease in SpO2 (see Figure to give MetHb its characteristic brown color, adds to
11-8). This fact is again consistent with the light absor- both the numerator and denominator of the ratio R in
bance spectra in Figure 11-4, which shows that MetHb Equation 11-8. Increasing both the numerator and
262 PART III  Patient Monitors

denominator of this ratio by a fixed amount tends to 1.0

drive the value of R toward 1.0. The pulse oximeter cali- Methylene blue
bration curve of Figure 11-6 shows that an R value of 1.0 Indigo carmine
0.8 Indocyanine green
corresponds to an SpO2 value of 85%. This may explain

Extinction Coefficient
why the pulse oximeter tends to read near 85% satura-
tion in the presence of high MetHb levels.7 A case report 0.6
of methemoglobinemia caused by herbal medicines illus-
trates this phenomenon.8 0.4
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 0.2
usual pulse oximeter wavelengths, 660 and 940 nm, are
similar to the corresponding values for adult hemoglobin 0.0
(HbA). This is fortunate because the percentage of HbF 200 300 400 500 600 700 800
present in neonatal blood varies with gestational age and Wavelength (nm)
is not very predictable. HbF also produces small errors in FIGURE 11-9  n  Extinction coefficient (ε) versus wavelength (λ) in
multiwavelength in€vitro CO-oximeters. The oxygenated the range 200 to 800 nm for three dyes: methylene blue, indigo
state of HbF is interpreted by some older laboratory carmine, and indocyanine green. (Modified from Scheller MS,
Unger RJ, Kelner MJ: Effects of intravenously administered dyes on
oximeters as consisting partially of COHb.9 pulse oximetry readings. Anesthesiology 1986;65[5]:550-552.)
Theoretical considerations suggest that sickle hemo-
globin (HbS) should also have little effect on pulse oxim-
eter accuracy, but this is difficult to confirm experimentally. in blood, appears to have no significant effect on SpO2 at
It would be unethical to subject sickle cell patients to concentrations seen clinically.15
intentional hypoxemia to study pulse oximeter accuracy. Nail polish has variable effects upon SpO2 values and
A few clinical case reports have involved sickle cell usually produces falsely low readings.16 Highly opaque,
patients, either under healthy conditions or during a acrylic nail coverings can prevent the pulse oximeter from
sickle cell crisis; however, these have produced conflict- detecting any pulsatile absorbance at all. This problem
ing results. One issue is that there is no clear gold stan- can usually be averted by rotating the fingertip sensor 90
dard for comparison of SpO2 values in sickle cell patients. degrees so that the coated fingernail does not fall within
One study, which concluded that SpO2 overestimated the optical path.
SaO2 with a bias of 6.9%, used a CO-oximeter as the gold
standard.10 However, laboratory CO-oximeters are gen-
erally designed to function in the presence of only four
Multiwavelength Pulse Oximetry
types of hemoglobin: 1) reduced Hb, 2) O2Hb, 3) COHb, Until 2005, only a few investigations had been made into
and 4) MetHb. Accuracy in the presence of HbS is not the possibilities of multiwavelength pulse oximetry.
specified. Some studies have used as a standard the SaO2 �Aoyagi, the inventor of pulse oximetry, performed experi-
calculated from the PaO2, which is measured by a blood ments with three-wavelength sensors in 2002.17 Although
gas electrode (Clarke electrode). This method is dubious his purpose in using an additional wavelength was to
because it must assume a “normal” oxygen-hemoglobin improve the accuracy of the SpO2 value, he noted that
dissociation curve (see Fig. 11-2), when it is well known “Dyshemoglobins (e.g., COHb and MetHb) can be mea-
that sickle cell patients have right-shifted dissociation sured with a multiwavelength system.” A four-wavelength
curves. At least two studies have measured the O2 disso- pulse oximeter that estimated total hemoglobin was
ciation curves of individual sickle cell patients and then reported by Noiri and colleagues in 2005.18 However, this
used these to calculate SaO2 from PaO2.11,12 These stud- device was never commercially produced and did not see
ies concluded that pulse oximeter accuracy is maintained widespread clinical use. Several investigators have used
in sickle cell disease as long as differences in O2 dissocia- multiwavelength IR in€ vivo oximetry to estimate hemo-
tion curves are accounted for. Other abnormal hemoglo- globin saturation in the internal jugular vein or in brain
bins, such as hemoglobin Bassett, also have produced tissue, but again, only the ratio of oxyhemoglobin to total
erroneous SpO2 readings.13 hemoglobin was measured.19,20
The ratio R, and hence the SpO2 value, can be affected In 2005, the Masimo Corporation (Irvine, CA)
by any substance present in the blood that absorbs light at announced the development of their Rainbow technology
660 or 940 nm. Dyes injected intravenously for diagnos- Rad-57 pulse oximeter. This device uses eight wavelengths
tic purposes can have significant effects on SpO2. For of light to measure SpO2, SpCO (pulse oximeter estimate
example, intravenous methylene blue produces sudden of COHb%), and SpMet (pulse oximeter estimate of
large decreases in SpO2 values in normal subjects.14 MetHb%). The handheld, battery-powered Rad-57 was
Indigo carmine yields small decreases in SpO2, and indo- later followed by a bench-top version, Radical-7. The first
cyanine green has an intermediate effect. The extinction human study of the performance of the Rad-57 was pub-
coefficients of these three dyes are plotted against light lished in 2006.21 Twenty healthy volunteers were instru-
wavelength in Figure 11-9.14 A comparison of these mented with radial artery catheters, electrocardiograms
absorbance spectra with those of hemoglobin (see Fig. (ECGs), sphygmomanometers, and Rad-57 sensors on the
11-4) predicts the relative effects of the three dyes on fingers. In the COHb arm of the study, volunteers were
SpO2 values. Bilirubin, another common pigment found exposed to inspired CO at 150 ppm until their COHb%
 11  Pulse Oximetry 263

levels reached 15%. Arterial blood samples were analyzed Carboxyhemoglobin Elimination
by CO-oximetry, and the resulting COHb% values were 50
compared with the SpCO readings. The bias—that is, the SpCO
mean difference between SpCO and COHb% by CO- 40 Venous COHb%
oximeter—was −1.22%, and the precision (standard devia-
tion [SD] of the difference) was 2.19%. These values are 30

COHb%
nearly the same as the manufacturer-specified uncertain-
ties for conventional pulse oximeters in the measurement
20
of SpO2. This volunteer study obviously could not investi-
gate the more pathologic COHb% levels found in smoke-
inhalation injuries. 10
Subjects in the MetHb arm of the study were given
300 mg of intravenous sodium nitrite, a drug approved 0
by the Food and Drug Administration (FDA) for treat- 0 20 40 60 80 100 120 140 160
ment of cyanide toxicity; this resulted in methemoglobin Time (min)
levels of up to 13%. The comparison of SpMet with FIGURE 11-10  n  Carboxyhemoglobin (COHb%) and its pulse oxim-
simultaneous CO-oximeter values of MetHb% yielded a etry estimate (SpCO) plotted vs. time in an 81-year-old victim of
bias of 0.00% and a precision of 0.45%. These uncertain- smoke inhalation. (From Plante T, Harris D, Savitt J, et€al: Carboxy-
hemoglobin monitored by bedside continuous CO-oximetry.
ties for SpMet are approximately one fourth of those J Trauma 2007;63[5]:1187-1190.)
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 Several case reports were published of Rad-57 being
larger absorbance signal for the pulse oximeter’s process- used to detect and monitor methemoglobinemia. Many
ing algorithm. life-threatening methemoglobinemia cases reported involve
Laboratory volunteer studies cannot predict a moni- the use of topical benzocaine.32 This is one of a large num-
tor’s performance in clinical conditions, nor can they ber of commonly used drugs that induce methemoglobin-
establish its effects on patient outcome. For this we rely emia as a side effect, a problem that undoubtedly has been
on clinical studies and case reports, a number of which underdiagnosed in the past.33 Macknet and colleagues pre-
have been published in the past few years. Coulange and sented a case of severe benzocaine-induced methemoglo-
colleagues22 measured SpCO from Rad-57 and simulta- binemia during transesophageal echocardiography, in
neous values of COHb% from blood CO-oximetry in which the Rad-57 was used to make the diagnosis and to
smoke-inhalation victims. In data from 12 patients, they monitor treatment with methylene blue.34 Barker and col-
found a bias of −1.5% and a precision of 2.5%. Layne and leagues reported a similar case in the OR following topical
colleagues measured SpCO and COHb% (Avox 400 CO- benzocaine for awake intubation.35 In both of these cases,
oximeter) in 130 outpatients, both smokers and non- the multiwavelength pulse oximeter played a major role in
smokers.23 They found a bias of −0.65% and a precision diagnosis and treatment; in€vitro laboratory CO-oximeter
of 1.8% in a COHb range from 0% to 31%. Both of these values were delayed too much to guide treatment.
clinical studies found SpCO uncertainties to be essen- The current versions of the Masimo Rad-57 and
tially the same as in the Barker volunteer study.21 �Radical-7 have two limitations. First, they still use a con-
Numerous case reports have been published on the use ventional two-wavelength red-over-IR algorithm to calcu-
of Rad-57 to detect and measure COHb levels. Some late the SpO2 value. Thus, when levels of either COHb or
authors have used SpCO as a screening tool to identify MetHb are significant, the displayed SpO2 is subject to the
smokers and track smoking cessation.24,25 Another prom- same errors described above.6,7 Of course, the presence of
ising application for the handheld device is in fire depart- a displayed SpCO or SpMet value would alert the user to
ments and for emergency medical services (EMS) the likelihood of this SpO2 error. The second limitation is
responders. Hostler and colleagues26 showed that SpCO the existence of “crosstalk” between the MetHb and
could be used as a diagnostic tool for firefighters who COHb measurement channels. That is, in the presence of
respond to CO alarm activations. They found that patients significant MetHb levels, the instrument will display a
who require transport to the emergency department had falsely elevated SpCO value while also displaying a correct
mean SpCO levels of 27.8%, whereas patients not in need SpMet value. This situation is detected by the device,
of treatment had mean levels of 3.2%. A similar study of which will display an error message indicating that the
149 patients by the New York Fire Department concluded SpCO value may not be accurate.
that the Rad-57 “can be used as a screening tool to uncover In March 2008, Masimo announced another innova-
cases of CO poisoning in which the diagnosis is not sus- tion in multiwavelength pulse oximetry: the noninvasive
pected.”27 Firefighters themselves have been screened measurement of total hemoglobin. Macknet and col-
with SpCO in other investigations.28 Case reports from leagues36 published the first clinical validation study, in
emergency departments have shown the value of SpCO which 30 surgical patients and 18 healthy volunteers were
monitoring for tracking the progress of carbon monoxide monitored with the new prototype. The subjects followed
poisoning victims and in the detection of CO poisoning in a hemodilution protocol in which one unit of blood was
unsuspected cases.29-31 Figure 11-10 shows an example of withdrawn and replaced with normal saline. Arterial blood
SpCO trend monitoring and response to treatment in an was sampled periodically and analyzed for total hemoglo-
81-year-old woman who was trapped in a house fire.29 bin (Hbt) by a Radiometer ABL-735 CO-oximeter, and
264 PART III  Patient Monitors

18 The LED emits light over a narrow but finite range of

wavelengths. The center wavelength, or wavelength of
16
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 dif-
14 ferent extinction coefficient, particularly at the 660-nm
wavelength. Manufacturers use at least two approaches to
this problem. The simplest method is to measure the cen-
SpHb (g/dL)

12
ter wavelength of all LEDs and reject those that fall outside
a specified range, for example, 660 (±5) nm. This method is
10 effective but expensive because of the large number of
LEDs that must be discarded. The second method is to
8 store multiple calibration algorithms in the pulse oximeter
software, corresponding to different LED center wave-
lengths. The electrical connector on the sensor cable is
6 y  0.62  0.945 * x then pin coded so that only the appropriate algorithm can
r  0.882
be selected for a given sensor. Neither method entirely
4 eliminates the effect of center frequency variation. This
4 6 8 10 12 14 16 18 variability does not affect the pulse oximeter’s ability to fol-
Hbt (g/dL) low changes in saturation, but it does produce differences
FIGURE 11-11  n  Radical-7 (Masimo Corporation, Irvine, CA) mea- between sensors in the absolute value of SpO2.38
surement of total hemoglobin (SpHb) plotted against simultane-
ous laboratory CO-oximeter value. Included were 48 subjects
and 802 data points. Bias was 0.03 and precision was 1.12. Hbt, Signal/Noise Ratio
total hemoglobin. (From Macknet M, Norton S, Kimball-Jones P,
et€ al: Continuous non-invasive measurement of hemoglobin via The amplitude of the fluctuating (AC) component of the
pulse CO-oximetry. Anesth Analg 2007;105[6]:S-108.) light absorbance can be much less than 1% of the ampli-
tude of the DC component (see Fig. 11-5). Any influence
that decreases the AC component, increases the DC com-
the values obtained were compared with the simultane- ponent, or adds an artifactual AC component not related
ous Radical-7 measurements of hemoglobin (SpHb), as to arterial pulsations will worsen the signal/noise ratio. For
shown in Figure 11-11. In a range of Hbt values from 4.4 to example, the AC signal is decreased in low perfusion states,
15.8 g/dL, this comparison revealed a bias and precision such as shock; the DC signal is increased when ambient
of 0.03 and 1.12 g/dL, respectively. Macknet and col- room light reaches the detector; and artifactual AC signals
leagues36a published a similar study in 20 healthy volun- are caused by patient motion, such as shivering.
teers and concluded that pulse CO-oximetry–based SpHb The photodiode light detector in the pulse oximeter
measurement was accurate to within 1.0 g/dL compared sensor cannot discriminate one wavelength of light from
with laboratory CO-oximeter measurements. another—it is effectively color blind. The detector there-
A presentation at the 2008 World Congress of Anes- fore responds to ambient room light and to light from
thesiology described a case in which casual screening with either of the LED sources. In most two-wavelength pulse
the Radical-7 revealed an abnormally low SpHb (11 g/dL) oximeters, this problem is solved by activating the red and
in a 72-year-old man.37 This led to further testing, which infrared LEDs in an alternating sequence. During one
yielded the early diagnosis of an asymptomatic esophageal part of this sequence, both LEDs are turned off, and the
carcinoma. The early complete surgical excision resulted photodetector measures the ambient background light.
in the probable cure of a highly lethal disease. The nonin- This sequence is repeated many times per second (e.g.,
vasive measurement of Hbt by pulse oximetry is very new, 480 Hz) in an attempt to eliminate light interference from
and more case reports and clinical outcome studies can be rapidly changing ambient sources. Despite this ingenious
expected in the future. design, ambient light artifact can create problems with
Multiwavelength pulse oximetry is now capable of the the signal/noise ratio; however, this difficulty can be min-
continuous, noninvasive measurement of MetHb, car- imized by covering the sensor with an opaque shield of
boxyhemoglobin, and total hemoglobin. Technology has some sort, such as a surgical drape or towel.
made real progress in correcting one of the chief weak- If the peripheral pulse is weak, the AC absorbance sig-
nesses of pulse oximetry: performance in the presence of nal becomes smaller compared with the DC signal. The
abnormal hemoglobin species. Clinical studies and case pulse oximeter uses an automatic gain control that adjusts
reports have already documented the importance of these either the LED intensity or the photodetector amplifier
new developments, and it is hoped that further develop- gain to compensate for changes in AC signal amplitude.
ments will lead to the continuous measurement of other Unfortunately, this process also amplifies background
substances in the blood. noise from all sources, including ambient light. At the
highest amplifier gain setting, the pulse oximeter may
Wavelength Variability interpret background noise as a pulsatile absorbance and
may generate an SpO2 value from this artifact.4 This phe-
The LED used as a light source by the pulse oximeter is nomenon could be demonstrated in early pulse oximeters
not an ideal monochromatic (single-wavelength) radiator. by inserting a piece of paper between the photodetector
 11  Pulse Oximetry 265

and the LEDs. Some of these pulse oximeters would 8

amplify background noise and display a pulse and satura- *
tion value from the paper.
This signal/noise ratio problem is also illustrated by
the penumbra effect.39 If a finger sensor is partially dis- 6
lodged or malpositioned in such a manner that light
passes through the fingertip at a grazing incidence, the *

Failure rate (%)

pulse oximeter may display a correct heart rate but an
erroneous SpO2 value. The SpO2 value from a malposi- 4
tioned 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).
2
Most pulse oximeters display some sort of visual indi-
cator of pulsatile absorbance signal, usually in the form of
an absorbance-versus-time plethysmogram. This dis-
played waveform usually represents the AC signal after
amplification and therefore does not directly correspond 0
1 2 3 4
to the actual amplitude of the absorbance pulsations.
However, a few manufacturers display a waveform whose ASA physical status
height actually does represent pulsatile absorbance ampli- FIGURE 11-12  n  Pulse oximeter failure rates as a function of Ameri-
can Society of Anesthesiologists (ASA) physical status. Data
tude before amplification. The user must therefore deter- shown are from both the operating room and postanesthesia
mine what the waveform of the pulse oximeter in use care unit in 10,312 patients. Failure rate increased from 1% to
measures; in general, it is not safe to assume that wave- more than 7% as physical status worsened. Asterisks denote
form amplitude represents pulse amplitude. The plethys- significance. (From Moller JT, et€al: Randomized evaluation of pulse
mograph is not a quantitative monitor of peripheral oximetry in 20,802 patients. Anesthesiology 1993;78:436-453.)
perfusion, and it cannot be relied upon to warn of impend-
ing ischemia, despite claims to the contrary. 8.6% of its baseline value, which occurred at a cuff pres-
The behavior of first-generation pulse oximeters dur- sure of 96% of systolic pressure. On cuff deflation, the
ing shock or low-perfusion states has been studied in signal returned at a blood flow of only 4% of baseline.
both humans and animals.41-45 During hemorrhagic Severinghaus and colleagues49 studied pulse oximeter
shock, pulse oximeters may display no SpO2 value at all, behavior during various types of reduction in extremity
or they may give falsely low saturation estimates. Loss of blood flow, including via blood pressure cuff, brachial
signal is likely with hypothermia, severe anemia, low artery pressure clamp, and extremity elevation. Failure
cardiac output states, and extremes in systemic vascular occurred at higher mean arterial pressures with the arte-
resistance. An early study of failure rates in the OR rial clamp than with gravitational hypotension, showing
found that pulse oximeters (Nellcor N-100, Ohmeda the importance of blood volume pulsatility. These stud-
3700) failed in 1.12% of all patients.46 Higher failure ies demonstrate that the pulse oximeter functions over a
rates were associated with poor preoperative physical wide range of blood flows and blood pressures in the
status, long operations, and elderly patients. A later extremity. Because the pulse oximeter is designed to
study used computerized anesthesia records and found function independently of changes in flow or pressure,
that 9% of all cases (n = 9203) had gaps in SpO2 data of loss of signal cannot be used to determine the adequacy
10 minutes or more.47 Higher failure rates were associ- of peripheral perfusion, even though this has been
ated with higher ASA physical status number and with attempted.50 (See also the related discussion of the ple-
hypotension and hypothermia. The largest failure rate thysmograph above.)
study was done by Moller and colleagues,48 in which The electrosurgical unit (ESU) is another source of
more than 20,000 surgical patients were followed. Inter- artifact. The intense, high-frequency, electromagnetic
estingly, this controlled study found little difference in radiation from the ESU electrode fills the OR whenever
postoperative outcome between patients monitored with the device is activated. The electrical cable from the
pulse oximetry and unmonitored patients, but it did find pulse oximeter sensor acts as an antenna that receives the
a very significant variation in failure rate with ASA phys- ESU radiation. Earlier generations of pulse oximeters
ical status, shown in Figure 11-12. The failure rate for were essentially shut down by ESU activation and
ASA-4 patients was more than five times that for ASA-1 required up to 30 seconds to recover. More recent pulse
patients; these higher failure rates in sicker patients have oximeters—that is, those made after the year 2000—are
led some to describe the pulse oximeter as a “fair-weather much improved in this respect and usually continue to
friend.” display and update SpO2 and heart rate values during
Several studies have determined thresholds for loss of ESU use. However, some pulse oximeters display their
signal during low-perfusion states. Lawson and col- last “valid” SpO2 value for up to 30 seconds during loss-
leagues41 produced gradual occlusion of blood flow with of-signal periods. In these instruments, the user can be
a blood pressure cuff while monitoring flow at the fin- certain that the pulse oximeter is measuring SpO2 during
gertip using a Doppler flow probe. They found loss of ESU activation only if a reasonable plethysmograph
signal when blood flow had decreased to an average of waveform is observed.
266 PART III  Patient Monitors

Discrete saturation transform

Adaptive correlation canceler
energy output

FIGURE 11-13  n  An example of a discrete satura-

tion transform during patient motion. The sig-
nal strength calculated by the algorithm is
plotted against saturation for all possible esti-
60 65 70 75 80 85 90 95 97 100 mated pulse oximetry (SpO2) values. The peak
rv ra at 97% corresponds to the saturation of arterial
blood (ra); the peak at 79% represents venous
SpO2 (%) blood (rv).

Motion Artifact and Adaptive Digital showed that SpO2 may decrease by 8% when the arm is
moved from a raised to a dependent position.56 The sec-
Signal Filtering ond assumption is that the saturation values from the
Artifacts caused by patient motion have plagued pulse venous pulsations will be less than arterial values, which
oximetry, particularly in the recovery room, intensive seems rather obvious. The Masimo SET electronically
care unit (ICU), and emergency settings. Although loss of scans all possible values of the ratio R that correspond to
signal or erroneous readings as a result of motion are saturations of 0% through 100%, and it calculates the sig-
infrequent in the OR (1% to 2%), they can cause a false nal intensity at each possible R value, as shown in Figure
alarm incidence that exceeds 50% in recovery room and 11-13. In this example of a discrete saturation transform
ICU settings.51 Patient motion, especially shivering, curve, two intensity peaks occur at 79% and 97% satura-
causes a large, fluctuating absorbance artifact that is tion: the higher peak corresponds to the arterial pulsations
incorrectly interpreted by the pulse oximeter algorithm. and is used to calculate SpO2, and the lower peak at 50%
Much of this artifact results from venous blood volume presumably represents the venous pulsations. The entire
pulsations created by the motion, thus violating the sec- sequence is repeated once per second on the most recent
ond assumption of conventional pulse oximetry, that the 6 seconds of raw data. The displayed SpO2 value thus rep-
light path length is exactly known. resents a 6-second running average of arterial hemoglobin
Until fairly recently, manufacturers used two approach� saturation, updated every second.
�es to this problem: increased signal averaging time and This new technology has been evaluated in two volun-
ECG synchronization. In the first approach, the value of teer experiments in which motion was induced in one hand
the ratio R is stored on a beat-to-beat basis and averaged while the other hand served as a stationary control.53,54 In
for several seconds. This running average is less sensitive both of these experiments, done 5 years apart, the latest
to patient motion but is also slower to respond to sudden version of Masimo SET was compared with the current
saturation changes. The reduction in the false-alarm rate versions of other manufacturers’ instruments, most of
resulting from motion is thus accompanied by a slower which claimed to be “motion resistant.” The results of the
response to true alarms, which is particularly risky in more recent study are summarized in Table 11-1. These
pediatric or neonatal applications. The second approach, findings are supported by several clinical studies conducted
ECG pulse rate synchronization, was developed by Nell- on patients in postoperative care units.55,57,58 Additional
cor and applied in their N-200 pulse oximeter.52 Although studies will determine whether this approach yields
this “C-lock” feature seemed to be a useful innovation, it improved performance in clinical settings of low perfu-
was not particularly successful and was abandoned in later sion, such as with shock or after cardiopulmonary bypass.
models.
A recent, more elegant solution to motion artifact is to
determine the “noise signal” and subtract it from the total CLINICAL APPLICATIONS: ACCURACY,
signal, leaving a noise-free signal from which to calculate RESPONSE, AND LIMITATIONS
the SpO2. Masimo has used this “adaptive digital filtering”
approach to develop an algorithm called signal extraction This section reviews clinical applications of pulse oxime-
technology (SET), which improves pulse oximeter perfor- try and its physiologic limitations, that is, the clinical
mance in the presence of motion artifact.53-55 The method changes that can and cannot be detected by saturation
makes two assumptions, the first being that most of the monitoring. In reviewing studies of pulse oximeter accu-
noise associated with motion artifact is produced by pulsa- racy, simple statistical tools are needed. Studies of pulse
tions in venous blood volume, which produce the SpO2 oximeter accuracy are generally methods-comparison
errors described above. In fact, errors from venous pulsa- studies, in which two independent methods are used to
tions were documented in an early volunteer study that measure the same variable simultaneously. One of the two
 11  Pulse Oximetry 267

covered by the data as well as agreement between the two

TABLE 11-1â•… P
 erformance of Several Pulse
methods. Similarly, linear regression slope and intercept
Oximeters During Mechanically
may be meaningless if the data points fall within a narrow
Controlled Hand Motion
range of values.
Performance Sensitivity Specificity
Pulse Oximeter Index (%) (%) (%)
Accuracy Studies
Masimo SET 93 99 97
Viridia/24C 84 78 90
Most pulse oximeter manufacturers claim an uncertainty
Hewlett-Packard 80 70 83
of ±2% (1 SD) for SpO2 values between 70% and 100%.
CMS-B This uncertainty increases to ±3% for SpO2 values
Nellcor N-395 73 70 73 between 50% and 70%; no accuracy is specified for SpO2
Datex-Ohmeda 3900 68 60 52 values below 50%. This implies that for saturations above
Nova Mars 58 40 42 70%, the SpO2 value should be within 2% of the actual
Nellcor N-295 55 39 53 saturation 68% of the time and within 4% (2 SDs) 95% of
the time. These accuracy specifications have not changed
Performance index is the percent of time during which estimated since the mid-1980s. The seemingly generous 2% uncer-
oxygenation by pulse oximetry on the hand in motion is within tainty is the direct result of the third assumption for pulse
7% of the simultaneous value on the stationary (control) hand.
Sensitivity and specificity for detection of hypoxemia (saturation oximetry as stated above: one empirical calibration curve—
values <90%) are also shown. All instruments except the N-295 that is, the relationship between the ratio R and SpO2—
claim to be “motion resistant.” fits the entire human race.
From Barker SJ: “Motion-Resistant” pulse oximetry: a comparison In the 1980s and 1990s, a number of studies of pulse
of new and old models. Anesth Analg 2002;95:967-972. oximeter accuracy were done that included both labora-
tory volunteer tests and clinical trials.60,61 Three of the
studies from this era are of special interest because they
methods usually is a new or unproven technique—in this evaluated not only accuracy but also response times to
case, pulse oximetry—and the other method is considered sudden changes in hemoglobin saturation.62-64 These
a gold standard. The most common gold standard for studies found some significant errors in pulse oximeter
pulse oximeter studies is a multiwavelength in€vitro CO- calibrations, which were subsequently revised by the
oximeter used to analyze arterial blood samples. These manufacturers. As a result of such “aftermarket” software
laboratory devices claim an uncertainty on the order of revisions, seemingly identical pulse oximeters may actu-
±1% (1 SD) for measurements of fractional saturation ally perform differently. Experimental study reports must
(Equation 11-2). Because the accuracy of current pulse therefore specify the software version and the pulse oxim-
oximeters is comparable to this figure (generally ±2%), it eter manufacturer and model used.
must be noted that both methods in all comparison stud- The volunteer studies of Severinghaus and Naifeh
ies are subject to uncertainty; a true gold standard of abso- compared the performance of various pulse oximeters
lute accuracy for any monitor is rare. during severe, transient desaturations.62,64 Subjects
The most commonly used statistics for evaluating underwent brief (45 seconds) desaturations to an O2Hb%
methods-comparison studies are bias and precision as value of 40% to 70%. Pulse oximeter response times to
defined by Bland and Altman.59 Bias is defined as the these sudden saturation changes were much shorter for
mean of the differences between simultaneous measure- earlobe sensors than for finger sensors (Figure 11-14).
ments by the two methods, and precision is the standard The time for a 50% response to rapid desaturation, or
deviation of this difference. (We have suggested calling “resaturation,” ranged from 10 to 20 seconds for the ear-
the latter quantity the imprecision because a larger value lobe sensor; for the finger sensors, it varied between 24
implies a less precise measurement.) This text defines the and 50 seconds. A similar result was obtained in another
difference between measurements as the pulse oximeter study that compared response times of finger sensors to
SpO2 value minus the CO-oximeter O2Hb% value; how- those of both earlobe sensors and reflectance sensors on
ever, in some of the literature, the opposite sign is used. the forehead.65 Both of these studies showed wide varia-
The bias indicates systematic error; that is, it shows the tions among subjects in response times for finger sensors,
tendency of one of the two methods to consistently over- reflecting a wide range of lung-to-finger circulation times.
estimate or underestimate one value relative to the other. These physiologic time delays and their interpatient vari-
The precision represents the variability or random error ability should be considered in the selection of sensor
between the two methods. If both the systematic and ran- sites in clinical settings, in which SpO2 can change rapidly
dom errors are within acceptable clinical limits, the (e.g., the OR). However, clinical studies have also found
methods-comparison study suggests that one method can that finger sensors are the most reliable in obtaining SpO2
replace the other. Unfortunately, many published meth- values during periods of hemodynamic instability.66
ods-comparison studies do not include bias and precision The pulse oximeter’s response to sudden saturation
values. Reported statistics often include a correlation changes is also affected by the signal-averaging time of
coefficient r and a linear regression slope and intercept of the instrument, which is often user selectable. The dis-
the scatterplot, or a graph of method A versus method B played SpO2 value represents a running average of data
values. Correlation coefficient is not a measure of agree- obtained over the most recent averaging time period,
ment between two variables; rather, it is a measure of which may range between 1 and 15 seconds. The SpO2
their association and is affected by the range of values value will respond more quickly to a sudden change in
268 PART III  Patient Monitors

A B Plethysmograph Variability Applications

100
In 1873, Kussmaul proposed that respiratory variation of
the pulse intensity is an important sign of various physi-
ologic 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
Sat (%)

50 arterial cannula and fluid responsiveness, addressing the

basic clinical question of whether the patient would ben-
C efit 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
OSM-3 HbO2%
parameters” such as pulse-pressure variation can be more
%Sat from PaO2 & pH
0 predictive.71
0 50 100 Meanwhile, other investigators had already extrapo-
Time (s) lated this pulse-variation principle to the dependence of
FIGURE 11-14  n  Tracings of pulse oximetry values plotted against pulse oximeter plethysmograph variations on intravascular
time for seven pulse oximeters during a rapid, brief desatura- volume.72 These early efforts to use pulse oximetry for the
tion in a healthy volunteer. Tracings labeled A represent three
earlobe sensors, tracings labeled B are four finger sensors, and estimation of volume status were interesting but found
tracing C is the pulmonary venous saturation calculated from little clinical use because they were neither sensitive nor
exhaled oxygen tension measured by mass spectrometry. The specific. However, more recent advances in pulse oximeter
earlobe sensors register the desaturation with a 10- to 15-second signal processing have made this noninvasive possibility
time lag, whereas the finger sensors show a 50-second time lag more promising. Masimo has developed a new parameter
in this volunteer. Hb, hemoglobin; OSM-3, CO-oximeter; PaO2,
arterial oxgyen tension; Sat, saturation. (Modified from Severing- called the plethysmograph variability index (PVI), which is
haus JW, Naifeh KH: Accuracy of response of six pulse oximeters to defined as
profound hypoxia. Anesthesiology 1987;67:553.)
 PVI = 100 × (PImax − PImin )/PImax (11-9)

SaO2 if a short averaging time is selected. On the other where PImax is the maximum value of the perfusion index,
hand, if the signal/noise ratio is marginal or artifacts are the ratio of AC to DC absorbance signal during the respi-
present, such as with patient motion, a longer averaging ratory cycle, and PImin is the minimum value during the
time will yield better accuracy. The user must determine cycle. This parameter can be continuously displayed by
the most appropriate averaging time for the clinical set- the pulse oximeter, thus quantifying the respiratory varia-
ting. The default averaging time, which is the value tion of the plethysmograph amplitude. Higher values of
applied when the power is turned on, varies among man- PVI are suggestive of lower intravascular volume status
ufacturers; hence, the user must know the default value of and a stronger probability of a positive hemodynamic
the particular instrument in use. response to fluid infusion.
Clinical studies of accuracy usually combine data from Clinical studies of PVI have shown that it correlates
multiple patients to determine the SpO2 uncertainty. This strongly with arterial pulse pressure variations during
procedure yields a more pessimistic view of accuracy than positive-pressure ventilation.73 More importantly, recent
would be obtained by studying results for individual studies have found that PVI is predictive of response to
patients. If a sensor is placed on a patient, and the SpO2 intravascular volume infusion, with a sensitivity and speci-
value is 95%, the probability is 68% that the patient’s true ficity superior to those of either central venous pressure or
saturation lies between 93% and 97% (±1 SD). On the wedge pressure.74,75 Figure 11-15 shows typical plethys-
other hand, if the displayed SpO2 on that same patient mograph waveforms and PVI trend plots for patients who
decreases from 95% to 93%, the amount by which the are fluid responders and nonresponders.75 As intravascu-
saturation has decreased is more certain than the original lar fluids are infused, the PVI falls dramatically in fluid
absolute SpO2 value. In other words, trend accuracy is responders, but it changes little in the nonresponders.
greater than absolute value accuracy. Variability among Plethysmograph variability, like pulsus paradoxus, is
patients is a price we pay for the convenience of having the sensitive to a number of physiologic factors other than
pulse oximeter precalibrated with a universal algorithm volume status (Box 11-1). The astute clinician must con-
(see the third pulse oximetry assumption). The calibration sider all possible causes for changes in PVI. Additional
algorithm is a best fit of data from a large number of healthy clinical data must be used to distinguish hypovolemia
adult volunteers. Alternatively, manufacturers could have from tamponade, bronchospasm, or pneumothorax for
chosen to require user calibration on each individual example. Nevertheless, early studies suggest that pulse
patient. This would have improved absolute accuracy, but oximeter plethysmograph variability has excellent poten-
it would have lost an important advantage of pulse oxime- tial for monitoring hemodynamics in OR and critical care
try, namely, the absence of a need for user calibration. settings.
 11  Pulse Oximetry 269

Pulse oximeter waveform before Pleth variability index (%)

volume expansion

Responder to

expansion
volume Perfusion index (%)

Pleth variability index (%)

Nonresponder

expansion
to volume

Perfusion index (%)

FIGURE 11-15  n  Plethysmograph (Pleth) waveforms and plethysmograph variability index trend plots for patients who are fluid responders and
nonresponders. (From 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 101[2]:200-206, 2008.)

hand, none of these controlled studies demonstrated a sta-

BOX 11-1 ╇ Physiologic Changes that Can Affect
tistically significant difference in patient outcomes—that
Plethysmograph Variability
is, morbidity and mortality—and this includes the Moller
Cardiac study48 of more than 20,000 patients cited above. This
Cardiogenic shock
merely shows the difficulty of performing outcomes stud-
Pericardial tamponade ies in anesthesiology. When a study cannot disprove the
Pericardial effusion null hypothesis of no difference between treatments, it
Constrictive pericarditis does not mean that the null hypothesis is true. This basic
Restrictive cardiomyopathy fact of statistics is too often overlooked in interpretations
Acute myocardial infarction of clinical studies.
Since becoming an intraoperative standard of care in
Pulmonary
the 1980s, the pulse oximeter also has had major impact
Asthma in other clinical settings. Two studies monitored SpO2
Tension pneumothorax during transport from the OR to the recovery room;
Pulmonary embolism
Bronchospasm
both found a high incidence of desaturations with SpO2
Airway obstruction values below 90% in patients who did not receive supple-
mental oxygen.79,80 These studies strongly support a uni-
Noncardiac, Nonpulmonary form policy of transport to the postanesthesia care unit
Hypovolemia (PACU) with supplemental oxygen for all patients. Simi-
Septic shock lar studies suggest that SpO2 should be continuously
Anaphylactic shock monitored in most patients in the recovery room. One
Diaphragmatic hernia study of postoperative pediatric patients showed no cor-
Superior vena cava obstruction relation between SpO2 and a traditional postanesthesia
Extreme obesity score based on motor activity, respirations, blood pres-
sure, mental status, and color.81 The authors concluded
that postoperative pediatric patients should be monitored
by pulse oximetry or be given supplemental oxygen
Limitations of Pulse Oximetry regardless of their apparent state of wakefulness. Another
Since the pulse oximeter became a minimum standard of study found that 14% of adult PACU patients had SpO2
care in the OR in 1986,76 it has been unethical to perform values below 90% and that significant desaturations were
randomized controlled studies of its clinical effectiveness associated with obesity, extensive surgery, advanced age,
in that setting.77 Before that time, several such studies and poor physical status.82 In fact, most patients are more
were performed. Coté and colleagues78 studied 152 pedi- likely to have hypoxemia in the PACU than in the OR.
atric patients during anesthesia and surgery. In half of Recovery room patients no longer have a protected air-
these patients, the SpO2 data were unavailable to the anes- way and are not receiving mechanical ventilation, yet
thesiologist. The study found that “major events,” defined they have not fully recovered from the depressant effects
as SpO2 below 85% for more than 30 seconds, occurred of anesthesia and surgery.
significantly more often in the patients for whom the The pulse oximeter warns of developing hypoxemia,
SpO2 values were hidden, and most of these events but at the increased inspired oxygen fractions typically
occurred in patients younger than 2 years. On the other used in the OR, SpO2 may not provide an early warning
270 PART III  Patient Monitors

400 100

SpO2 95
PaO2
300 OpPO2
PtcO2 90
PO2 (mm Hg)

SpO2(%)
85
200 FiO2=0.5
FIGURE 11-16  n  Four oxygenation variables
plotted against time from the onset of
endobronchial intubation in a dog at a
80 fraction of inspired oxygen (FiO2) of 0.5.
Arterial oxygen tension (PaO2), optode
intraarterial oxygen tension (OpPO2), and
100
transcutaneous oxygen tension (PTCO2) all
decreased rapidly during the first 2 min-
utes following endobronchial intubation,
whereas the pulse oximetry value (SpO2)
did not change significantly at any time.
(From Barker SJ, Tremper KK, Hyatt J, et€al:
0 5 10 15 20 Comparison of three oxygen monitors in
detecting endobronchial intubation. J Clin
Time (min) Monit 1988;4:241.)

of decreasing arterial oxygen tension. The oxyhemoglo- not true. In a case report, a Physio-Control (Redmond,
bin dissociation curve (see Fig. 11-2) shows that PaO2 WA) pulse oximeter sensor was mistakenly connected to
must decrease to less than 80 mm Hg before SaO2 falls an Ohmeda (now GE Healthcare) instrument.84 The two
significantly. A good example of this limitation is in the oximeters used the same electrical connector for the sen-
detection of inadvertent endobronchial intubation. sor, but the internal pin connections were different; this
Figure 11-16 shows data from four different oxygen- resulted in severe thermal burns to both the finger and
ation monitors plotted as functions of time for an animal the earlobe of a newborn infant. The lesson from this
undergoing general anesthesia at an fractional concentra- case is that compatibility between sensor and instrument
tion of inspired oxygen (FiO2) of 0.5.83 In addition to SpO2, must be ensured before a sensor is placed on a patient.
the plot shows PaO2 values from sequential arterial blood This risk has been reduced in more recent designs, but
samples, transcutaneous oxygen tension (PTCO2), and oxy- the clinician should never assume that just because the
gen tension from an intraarterial fiberoptic “optode” blood plug fits, it is acceptable to interconnect sensors and
gas sensor (OpPO2). At 0 minutes on the time axis, the instruments of different manufacturers.
endotracheal tube was advanced from the trachea into the The pulse oximeter is unquestionably the most impor-
left mainstem bronchus. Within 3 minutes, the PaO2 tant advance in oxygen monitoring since the develop-
decreased from 360 mm Hg to 120 mm Hg, and the PTCO2 ment of the blood gas analyzer. It is the only oxygen
and OpPO2 values also fell significantly. However, SpO2 monitor that provides continuous, real-time, noninvasive
never fell below 98% during the entire experiment. In this data on arterial oxygenation without the need for user
situation, the pulse oximeter provided no indication that an calibration. Because it is noninvasive and almost risk free,
endobronchial intubation had taken place, whereas the the pulse oximeter should be used in all clinical settings in
other monitoring techniques all showed significant which risk of arterial hypoxemia is present. It is a mini-
changes. Only when the experiment was repeated for FiO2 mum standard of care in the OR and is rapidly becoming
values of 0.3 or less did the pulse oximeter display consis- a standard in most other critical care settings. Hypoxia
tent SpO2 decreases of 6% or more. This illustrates an remains the most common cause of anesthesia-related
important physiologic limitation of saturation monitoring: preventable death.85,86 Therefore the prudent use of
when increased FiO2 is used, the PaO2 value must decrease pulse oximetry is a fundamental responsibility of the
far below its baseline before the pulse oximeter will alert anesthesiologist.
the clinician. Metaphorically speaking, the pulse oximeter The pulse oximeter is not an ideal instrument because
is a sentry standing on the edge of the cliff of desaturation it is subject to both measurement error and physiologic
(see Fig. 11-2), and it gives no warning as we approach the limitations.87-95 With an understanding of the physics
edge; it only sounds an alarm when we have fallen off. and engineering principles of pulse oximetry as outlined
Finally, no discussion of a clinical device is complete in this chapter, the astute clinician will be aware of mea-
without a description of the risks and complications of surement errors and when they are likely to occur. The
the monitoring process. Because the pulse oximeter is impact of these errors also can be minimized by obtaining
noninvasive and typically does not produce heat or radia- appropriate supportive data. For example, a smoke inha-
tion, these risks might be expected to be nonexistent. lation victim should have carboxyhemoglobin levels
Because of human error, however, this is unfortunately determined by in€ vitro CO oximetry or be monitored
 11  Pulse Oximetry 271

with a multiwavelength pulse oximeter capable of mea- 23. Layne T, Snyder C, Brooks D, Enjeti S: Evaluation of a new pulse
suring COHb.21 The pulse oximeter signal/noise ratio CO-oximeter: non-invasive measurement of carboxyhemoglobin in the
outpatient pulmonary lab and emergency departments. Presented at the
should be optimized by careful sensor application, pre- 52nd Annual Meeting of the American Association for Respiratory
vention of ambient light from reaching the sensor, and Care (AARC), Las Vegas, December 2006.
consideration of alternative sensor sites in patients who 24. Hampson N, Ecker E, Scott K: Use of noninvasive pulse CO-
are shivering or who have diminished peripheral pulses. oximeter to measure blood carboxyhemoglobin levels in bingo
players, Resp Care 51(7):758–760, 2006.
The avoidance of hypoxia is a fundamental goal of every 25. Light A, Grass C, Pursley D, Krause J: Carboxyhemoglobin levels in
anesthesiologist, and an understanding of the physics and smokers vs. non-smokers in a smoking environment. Presented at the
physiology of arterial saturation monitoring by pulse 52nd Annual Meeting of the American Association for Respiratory
oximetry will help accomplish this task. Care (AARC), Las Vegas, December 2006.
26. Hostler D, Roth R, Kaufman R, et€ al: The incidence of carbon
monoxide poisoning during CO alarm investigations, Pre-Hospital
REFERENCES Emergency Care 12(1):115, 2008.
1. Brown LJ: A new instrument for the simultaneous measurement of 27. Ben-Eli D, Peruggia J, McFarland J, et€al: Detecting CO: FDNY
total hemoglobin, % oxyhemoglobin, % carboxyhemoglobin, % studies prehospital assessment of COHb, JEMS 32(10):S36–S37,
methemoglobin, and oxygen content in whole blood, IEEE Trans 2007.
Biomed Eng 27:132–138, 1980. 28. Dickinson E, Mecham C, Thom S, Shofer F, Band R: The non-
2. Stephen CR, Slater HM, Johnson AL, et€al: The oximeter: a tech- invasive carboxyhemoglobin monitoring of firefighters engaged in
nical aid for the anesthesiologist, Anesthesiology 12:541–555, 1951. fire suppression and overhaul operations, Pre-Hospital Emergency
3. Knill RL, Clement JL, Kieraszewicz HT, et€al: Assessment of two Care 12(1):96, 2008.
noninvasive monitors of arterial oxygenation in anesthetized man, 29. Plante T, Harris D, Savitt J, et€al: Carboxyhemoglobin monitored
Anesth Analg 61:582–586, 1982. by bedside continuous CO-oximetry, J Trauma 63(5):1187–1190,
4. Severinghaus JW, Astrup PB: History of blood gas analysis. VI, 2007.
Oximetry, J Clin Monit 2:270–288, 1986. 30. Li Z, Gao C, Huang X, Wang G, Ge H: Clinical analyses of 429
5. Wukitsch MW, Tobler D, Pologe J, et€al: Pulse oximetry: an analy- cases of acute CO poisoning, Anesthesiology 107: A1490, 2007.
sis of theory, technology and practice, J Clin Monit 4:290–301, 1988. 31. Suner S, Partridge R, Sucov A, et€ al: Non-invasive pulse CO-
6. Barker SJ, Tremper KK: The effect of carbon monoxide inhalation on oximetry screening in the emergency department identifies occult
pulse oximetry and transcutaneous PO2, Anesthesiology 66:677–679, carbon monoxide toxicity, J Emerg Med 34(4):441–450, 2008.
1987. 32. Throm M, Stevens M, Hansen C: Benzocaine-induced methemo-
7. Barker SJ, Tremper KK, Hyatt J: Effects of methemoglobinemia globinemia in two patients: interdisciplinary collaboration, man-
on pulse oximetry and mixed venous oximetry, Anesthesiology agement, and near misses, Pharmacotherapy 27(8):1206–1214, 2007.
70:112–117, 1989. 33. Ash-Bernal R, Wise R, Wright S: Acquired methemoglobinemia: a
8. Chui J, Poon W, Chan K, Chan A, Buckley T: Nitrite-induced retrospective series of 138 cases at two teaching hospitals, Medicine
methemoglobinaemia: aetiology, diagnosis, and treatment, Anaes- 83(5):265–273, 2004.
thesia 60:496–500, 2005. 34. Macknet M, Kimball-Jones P, Applegate R, Martin R, Allard M:
9. Cornelissen PJH, van Del WC, de Jong PA: Correction factors for Benzocaine induced methemoglobinemia after TEE, Respir Care
hemoglobin derivatives in fetal blood as measured with the IL282 52(11), 2007.
CO oximeter, Clin Chem 29:1555–1556, 1983. 35. Barker SJ, Annabi EH: Life-threatening methemoglobinemia
10. Craft JA, Alessandrini E, Kenney LB, et€al: Comparison of oxygen- detected by pulse oximetry: a case report, Anesth Analg 108(3):898–
ation measurements in pediatric patients during sickle cell crises, 899, 2009.
J Pediatr 124:93–95, 1994. 36. Macknet M, Norton S, Kimball-Jones P, et€ al: Continuous non-
11. Rackoff WR, Kunkel N, Silber JH, et€al: Pulse oximetry and factors invasive measurement of hemoglobin via pulse CO-oximetry,
associated with hemoglobin oxygen desaturation in children with Anesth Analg 105(6): 2007, S-108.
sickle cell disease, Blood 81:3422–3427, 1993. 36a. Macknet MR, Allard M, Applegate RL II, Rook J: The accuracy of
12. Weston Smith SG, Glass UH, Acharya J: Pearson TC: Pulse oxim- noninvasive and continuous total hemoglobin measurement by
etry in sickle cell disease, Clin Lab Haematol 11:185–188, 1989. pulse CO-oximetry in human subjects undergoing hemodilution,
13. Das A, Sinha S, Hoyer J: Hemoglobin Bassett produces low pulse Anesth Analg 111:1424–1426, 2010.
oximeter and co-oximeter readings, Chest 131:1242–1244, 2007. 37. Allard M, Viljoen J: Casual screening of haemoglobin noninvasively
14. Scheller MS, Unger RJ, Kelner MJ: Effects of intravenously admin- positively affects a colleague’s future? A case report. 14th Annual World
istered dyes on pulse oximetry readings, Anesthesiology 65:550–552, Congress of Anesthesiology; March 2008, Capetown, South Africa.
1986. Accessed online at http://www.wca2008.com/images/abstractWCA/
15. Veyckemans F, Baele P, Guillaume JE, et€al: Hyperbilirubinemia monitoring.pdf.
does not interfere with hemoglobin saturation measured by pulse 38. Pologe JA: Pulse oximetry: technical aspects of machine design, Int
oximetry, Anesthesiology 70:118–122, 1989. Anesthesiol Clin 25:137–153, 1987.
16. Coté CJ, Goldstein EA, Fuchsman WH, et€al: The effect of nail 39. Kelleher JF, Ruff RH: The penumbra effect: vasomotion-Â�
polish on pulse oximetry, Anesth Analg 67:683–686, 1988. dependent pulse oximeter artifact due to probe malposition,
17. Aoyagi T, Miyasaka K: The theory and applications of pulse spec- Â�Anesthesiology 71:787–791, 1989.
trophotometry, Anesth Analg 94:S93–S95, 2002. 40. Barker SJ, Hyatt J, Shah NK, et€al: The effect of sensor malposi-
18. Noiri E, Kobayashi N, Takamura Y, et€al: Pulse total hemoglobi- tioning on pulse oximeter accuracy during hypoxemia, Anesthesiol-
nometer provides accurate noninvasive monitoring, Crit Care Med ogy 79:248–254, 1993.
33: E2831, 2005. 41. Lawson D, Norley I, Korbon G, et€al: Blood flow limits and pulse
19. Petrov Y, Petrova I, Patrikeev I, Esenaliev R, Prough D: Multi- oximeter signal detection, Anesthesiology 67:599–603, 1987.
wavelength optoacoustic system for noninvasive monitoring of 42. Narang VPS: Utility of the pulse oximeter during cardiopulmo-
cerebral venous oxygenation: a pilot clinical test in the internal nary resuscitation, Anesthesiology 65:239–240, 1986.
jugular vein, Optics Letters 31:1827–1829, 2006. 43. Nowak GS, Moorthy SS, McNiece WL: Use of pulse oximetry for
20. Nelson L, McCann J, Loepke A, Wu J: Development and valida- assessment of collateral arterial flow, Anesthesiology 64:527, 1986.
tion of a multiwavelength spatial domain near-infrared oximeter to 44. Skeehan TM, Hensley FA Jr: Axillary artery compression and the
detect cerebral hypoxia-ischemia, J Biomed Opt 11(6), 2006. prone position, Anesth Analg 65:518–519, 1986.
064022. 45. Barrington KJ, Ryan CA, Finer NN: Pulse oximetry during hem-
21. Barker S, Curry J, Redford D, Morgan S: Measurement of car- orrhagic hypotension and cardiopulmonary resuscitation in the
boxyhemoglobin and methemoglobin by pulse oximetry, Anesthesi- rabbit, J Crit Care 1:242–246, 1986.
ology 105:892–897, 2006. 46. Freund PR, Overand PT, Cooper J, et€ al: A prospective study of
22. Coulange M, Barthelemy A, Hug F, Thierry A, DeHaro L: Reli- intraoperative pulse oximetry failure, J Clin Monit 7:253–258, 1991.
ability of new pulse CO-oximeter in victims of carbon monoxide 47. Reich DL, Timcenko A, Bodian CA, et€ al: Predictors of pulse
poisoning, Undersea Hyperb Med 32(2):1–5, 2005. oximetry data failure, Anesthesiology 84:859–864, 1996.
272 PART III  Patient Monitors

48. Moller JT, et€ al: Randomized evaluation of pulse oximetry in 72. Partridge BL: Use of pulse oximetry as a non-invasive indicator of
20,802 patients, Anesthesiology 78:436–453, 1993. intravascular volume status, J Clin Monitoring 3:263–268, 1987.
49. Severinghaus JW, Spellman MJ Jr: Pulse oximeter failure thresh- 73. Canneson M, Delannoy B, Morand A, et€al: Does pleth variability
olds in hypotension and vasoconstriction, Anesthesiology 73:532– index indicate the respiratory-induced variation in the plethys-
537, 1990. mogram and arterial pressure waveforms? Anesth Analg 106:1189–
50. Graham B, Paulus DA, Caffee HH: Pulse oximetry for vascular 1194, 2008.
monitoring in upper extremity replantation surgery, J Hand Surg 74. Canneson M, Desebbe O, Rosamel P, et€al: Pleth variability index
11A:687–692, 1986. to monitor the respiratory variations in the pulse oximeter plethys-
51. Lawless ST: Crying wolf: false alarms in a pediatric intensive care mographic waveform amplitude and predict fluid responsiveness in
unit, Crit Care Med 22:981–985, 1994. the operating theatre, Br J Anaesth 101(2):200–206, 2008.
52. Nellcor N-200 pulse oximetry note number 6: C-Lock ECG synchroniza- 75. Canneson M, Slieker J, Desebbe O, et€ al: The ability of a novel
tion principles of operation, Hayward, California, 1988, Nellcor. algorithm for automatic estimation of the respiratory variations in
53. Barker SJ, Shah NK: Effects of motion on the performance of pulse arterial pulse pressure to monitor fluid responsiveness in the oper-
oximeters in volunteers, Anesthesiology 86:101–108, 1997. ating room, Anesth Analg 106:1195–1200, 2008.
54. Barker SJ: “Motion-resistant” pulse oximetry: a comparison of new 76. American Society of Anesthesiologists: Standards for basic intraop-
and old models, Anesth Analg 95:967–972, 2002. erative monitoring, Anesthesia Patient Safety Newsletter, 1987;
55. Dumas C, Wahr JA, Tremper KK: Clinical evaluation of a proto- March:3.
type motion artifact resistant pulse oximeter in the recovery room, 77. Eichhorn JH, Cooper JB, Cullen DJ, et€al: Standards for patient
Anesth Analg 83(2):269–272, 1996. monitoring during anesthesia at Harvard Medical School, JAMA
56. Kim JM, Arakawa K, Benson KT, et€al: Pulse oximetry and circula- 256:1017–1020, 1986.
tory kinetics associated with pulse volume amplitude measured by 78. Coté CJ, Goldstein EA, Coté MA, et€ al: A single-blind study of
photoelectric plethysmography, Anesth Analg 65:1333–1339, 1986. pulse oximetry in children, Anesthesiology 68:184–188, 1988.
57. Bohnhorst B, Peter CS, Poets CF: Pulse oximeters’ reliability in 79. Pullerits J, Burrows FA, Roy WL: Arterial desaturation in healthy
detecting hypoxemia and bradycardia: comparison between a con- children during transfer to the recovery room, Can J Anaesth
ventional and two new-generation oximeters, Crit Care Med 34:470–473, 1987.
28(5):1565–1568, 2000. 80. Tyler IL, Tantisira B, Winter PM, et€al: Continuous monitoring of
58. Malviya S, Reynolds PI, Voepel-Lewis T, et€al: False alarms and arterial oxygen saturation with pulse oximetry during transfer to
sensitivity of conventional pulse oximetry versus the Masimo SET the recovery room, Anesth Analg 64:1108–1112, 1985.
technology in the pediatric postanesthesia care unit, Anesth Analg 81. Soliman IE, Patel RI, Ehrenpreis MB, et€al: Recovery scores do not
90(6):1336–1340, 2000. correlate with postoperative hypoxemia in children, Anesth Analg
59. Bland JM, Altman DG: Statistical methods for assessing agreement 67:53–56, 1988.
between two methods of clinical measurement, Lancet 1:307–310, 82. Morris RW, Buschman A, Warren DL, Philip JH, Reamer DB:
1986. The prevalence of hypoxemia detected by pulse oximetry during
60. Tremper KK, Barker SJ: Pulse oximetry, Anesthesiology 70:98–108, recovery from anesthesia, J Clin Monit 4:16–20, 1988.
1989. 83. Barker SJ, Tremper KK, Hyatt J, et€al: Comparison of three oxy-
61. Severinghaus JW, Kelleher JF: Recent developments in pulse gen monitors in detecting endobronchial intubation, J Clin Monit
oximetry, Anesthesiology 76:1018–1038, 1992. 4:240–243, 1988.
62. Severinghaus JW, Naifeh KH: Accuracy of response of six pulse 84. Murphy KG, Segunda JA, Rockoff MA: Severe burns from a pulse
oximeters to profound hypoxia, Anesthesiology 67:551–558, 1987. oximeter, Anesthesiology 73:350–352, 1990.
63. Kagle DM, Alexander CM, Berko RS, et€ al: Evaluation of the 85. Keenan RL, Boyan CP: Cardiac arrest due to anesthesia: a study of
Ohmeda 3700 pulse oximeter: steady-state and transient response incidence and causes, JAMA 253:2373–2377, 1985.
characteristics, Anesthesiology 66:376–380, 1987. 86. Taylor G, Larson CP Jr, Prestwich R: Unexpected cardiac arrest
64. Severinghaus JW, Naifeh KH, Koh SO: Errors in 14 pulse oximeters during anesthesia and surgery: an environmental study, JAMA
during profound hypoxia, J Clin Monit 5:72–81, 1989. 236:2758–2760, 1976.
65. Barker SJ, Hyatt J: Forehead reflectance pulse oximetry: time 87. Chapman KR, Liu FLW, Watson RM, et€al: Range of accuracy of
response to rapid saturation change, Anesthesiology 73(3A):A544, two-wavelength oximetry, Chest 4:540–542, 1986.
1990. 88. Nickerson BG, Sakrison C, Tremper KK: Bias and precision of
66. Barker SJ, Le N, Hyatt J: Failure rates of transmission and reflec- pulse oximeters and arterial oximeters, Chest 93:515–517, 1988.
tance pulse oximetry for various sensor sites, J Clin Monit 7:102–103, 89. Mihm FG, Halperin BD: Noninvasive detection of profound arte-
1991. rial desaturations using a pulse oximetry device, Anesthesiology
67. Kussmaul A: Über schwielige Mediastino-Peikarditis und den par- 62:85–87, 1985.
adoxen Puls, Berl Klin Wochenschr 10:433–435, 445-449, 461-464, 90. Cecil WT, Thorpe KJ, Fibuch EE, et€al: A clinical evaluation of
1873. the accuracy of the Nellcor N-100 and the Ohmeda 3700 pulse
68. Cohn JN, Pinkerton AL, Tristani FE: Mechanism of pulsus para- oximeters, J Clin Monit 4:31–36, 1988.
doxus in clinical shock, J Clin Investigation 46(11):1744–1755, 91. Fait CD, Wetzel RC, Dean JM, et€al: Pulse oximetry in critically ill
1967. children, J Clin Monit 1:232–235, 1985.
69. Michard F, Boussat S, Chemla D, et€al: Relation between respira- 92. Boxer RA, Gottesfeld I, Singh S, et€al: Non-invasive pulse oximetry
tory changes in arterial pulse pressure and fluid responsiveness in in children with cyanotic congenital heart disease, Crit Care Med
septic patients with acute circulatory failure, Am J Respir Crit Care 15:1062–1064, 1987.
Med 162:134–138, 2000. 93. Mok J, Pintar M, Benson L, et€al: Evaluation of noninvasive mea-
70. Lopes MR, Oliveira MA, Pereira VO, et€ al: Goal-directed fluid surements of oxygenation in stable infants, Crit Care Med 14:960–
management based on pulse pressure variation monitoring during 963, 1986.
high-risk surgery: a pilot randomized controlled trial, Crit Care 11: 94. Durand M, Ramanathan R: Pulse oximetry for continuous oxygen
R100, 2007. monitoring in six newborn infants, J Pediatr 109:1052–1055, 1986.
71. Michard F: Changes in arterial pressure during mechanical ventila- 95. Yelderman M, New W: Evaluation of pulse oximetry, Anesthesiology
tion, Anesthesiology 103:419–428, 2005. 59:349–352, 1983.