- Email: [email protected]
Accuracy of pulse oximetry in children with cyanotic congenital heart disease
Accuracy of Pulse Oximetry in Children With Cyanotic Congenital
Heart Disease
Hubert J. Schmitt, MD, Wolfram H. Schuetz, MD, Peter A. Proeschel, PhD, and Christel Jaklin The use of a pulse oximeter to monitor arterial oxygen saturation (SaOr) is considered accurate and reliable in the range of 90% to 100%. However, differing reports exist about the accuracy with desaturation. Thus, the suitability of pulse oximetry in desaturated patients was evaluated using a Nellcor N-100 oximeter. In 56 children with cyanotic congenital heart disease, the pulse oximeter reading was compared with the direct measurement of SaOz by a CO-oximeter OSM 3. The influence of high hematocrit values on the accuracy at low saturation was also investigated. All oxygen saturation measurements (two per child) were carried out after induction of anesthesia (ketamine, fentanyl, pancuronium) during a “steady state” before the surgical procedure. The results indicate that at low levels of saturation (SaOr below 80%). pulse oximetry is not as accurate as at higher saturations,
P
ULSE OXIMETRY is a standard part of patient monitoring during anesthesia. It is noninvasive and allows continuous beat-to-beat assessment of arterial oxygen saturation. There are several studies confirming the value of pulse oximetty during acute arterial desaturation.1-3
Nevertheless,
accuracy
of pulse
there
addition,
Lee et al7 recently
oximetry
are
differing
during
severe
demonstrated
reports
about
desaturation.4-” the inaccuracy
the In
METHODS With institutional approval and informed consent from parents, 56 children with cyanotic CHD were studied on the day of cardiac surgery. The operations were for transposition of great arteries (TGA) (n = 17), tetralogy of Fallot (n = 15), single ventricle (n = 8), tricuspid atresia (n = 6), and complex cyanotic heart disease (n = 10). Noninvasive arterial oxygen saturation (SpOz) was determined using a Nellcor N-100 pulse oximeter (Nellcor, Inc, Haywood, CA). According to the user’s manual, the appropriate sensor was applied to a finger, great toe, or the dorsum of the foot, but not using the extremities where arterial or venous catheters were located. During the measurements, no artificial light was allowed to fall on the sensor. To ensure reliability, the heart rate (HR) shown by the pulse oximeter was verified by comparison against the HR on the ECG monitor. The time of all saturation measurements (two per child) was during a steady state following the induction of anesthesia after placing all monitors including the arterial catheter and rectal and esophageal temperature probes, but prior to the surgical procedure. All children were placed on heating and cooling equipment as used routinely during cardiopulmonary bypass (CPB) to keep their temperature constant. The anesthesia consisted of ketamine, fentanyl, and pancuronium. The children were ventilated with 50% 02 in air, and inhalation agents or vasoactive substances were not used. Blood pressure, HR, and body temperature were continuously monitored.
Vascular
KEY WORDS: measurement notic heart disease
techniques, pulse oximetry, cya-
When the pulse oximeter showed a constant saturation for at least 2 minutes, the value was recorded and two blood samples were obtained anaerobically from the arterial catheter. Both blood samples were immediately analyzed using a CO-oximeter OSM-3 (Radiometer, Willich, Germany) for in vitro oxygen saturation (SaO,) and a Corning 288 blood gas analyzer (Corning, Medfield, MA) for a complete set of blood gases. SaOz measured by the CO-oximeter represents fractional saturation defined as
of
pulse oximetry in severe anemia with hematocrits below 10%. In a similar way abnormally high hematocrits could reduce the accuracy of this method. To evaluate the suitability and validity of pulse oximetry in the clinical situation of continual low saturation and high hematocrit, children with cyanotic congenital heart disease (CHD) were studied. Noninvasively measured arterial oxygenation (Sp02) was compared with direct measurement of arterial oxygen saturation (SaOZ) by a CO-oximeter.
Journalof Cardiothoracic and
and overestimates the true value. Bias and precision between saturations measured by the pulse oximeter and the CO-oximeter were 5.8 and 4.8 in the group with a saturation below 80%. and 0.5 and 2.5 in the group with a saturation over 90%. respectively. Because the margin of safety for a patient is small when arterial saturation levels are under 80%, it is advisable under this condition to check the SaOr measurements by a CO-oximeter. High hematocrit levels did not seem to be responsible for impaired accuracy of pulse oximetry at saturation values below 80%. Copyright 0 1993 by W.B. Saunders Company
100%.
OzHb/(OzHb
+ Hb + COHb
+ MetHb)
where 02Hb = oxyhemoglobin, Hb = reduced hemoglobin, COHb = carboxyhemoglobin, and MetHb = methemoglobin. The blood gas status analyzed by the Corning 288 consisted of arterial pH, oxygen partial pressure (PaO,), carbon dioxide partial pressure (PaCOz), hematocrit (Hct), and calculated arterial oxygen saturation (SaOzcalc). Both machines were calibrated daily. Data are presented as mean 2 SD. Statistical analysis of paired SpOz and SaOz values was performed using standard linear regression, bias, and precision, as well as the agreement according to Bland and Altmanns Bias represents the mean difference of paired values of in vivo saturation (SpOz) measured by the pulse oximeter and in vitro saturation (Sa02) measured by the COoximeter. Precision is the standard deviation of these differences. The differences between saturation values simultaneously obtained by two different methods were plotted against the average of the two measurements. The mean difference ? SD and the limits of agreement, described by mean difference + 2 SD, within which 95% of the differences will lie, were calculated. Correlations between specific variables (SpO2, bias, Hct, pH) were tested using standard linear regression. For comparison within a group the Wilcoxon test was used. Comparison between the groups was performed with the Kruskal-Wallis test. P < 0.05 was considered significant.
From the Department of Anesthesiology, Friedrich-AlexanderUniversityErlangen-Ntimberg Erlangen, Germany. Presented as a poster at the Sixth Annual Meeting of the European Association of Cardiothoracic Anesthesiologists, June 4-7, 1991, Milano, Italy. Address reprint requests to Dr. H.J. Schmitt, Department of Anesthesiology, Maximiliansplatz, University Erlangen-Niimberg, D-8520 Erlangen, Germany. Copyright o I993 by W.B. Saunders Company 1053-0770/93/0701-0012$03.0010
Anesthesia, Vol7, No 1 (February), 1993: pp 61-66
61
SCHMITT ET AL
62
Table 1. Measured Clinical Parameters saoz
COHb
MetHb
Range (%I
(%)
W)
0.9 -c 0.7
0.5 -t 0.2
1.0 c 0.7
0.5 -t 0.2
7.40 -c 0.04*
0.9 +- 0.6
0.4 2 0.2
7.37 2 0.05”
>9O(n
21)
=
80to90(n
= 44)
<80 (n = 47)
PH
7.42 t 0.04*
PaOz
PaC02
Hct
BP systolic
BP diastolic
(mmHgl
(mmHg)
WI
WmHg)
ImmHg)
74 t 137
33 c 5
44 -t 7*
51 +6t
34 2 5
48 + 9”
42 -+ 5t
36 rt 5
52 z IO*
86-+ 16
495
HR
Temperature
(beatsimin)
(“C)
14
97 + 21*
36.2 -c 0.5
82-c 15
50 + 14
107 +- 22*
36.3 -c 0.7
81 k 15
49k
119 2 16*
36.5 + 0.9
11
NOTE. Mean ? SD. *P < 0.05. tP < 0.01 between the groups.
and saturation measured by the CO-oximeter (SaOa) were 0.5% and 2.5% in the group with a saturation over 90%, 1.9% and 2.7% at an SaOl between 80% and 90%, and 5.8% and 4.8% in the third group with an SaOZ below 80%, showing the pulse oximeter to overestimate arterial saturation at low saturation values (Table 2). Paired differences of SpOZ and SaOz varied significantly among the three groups (P < 0.01). Figures 2 through 4 show the agreement between SpOZ and SaOz for each group analyzed according to Bland and A1tmann.s The limits of agreement (mean difference k 2 SD) were 0.5% t 5% in the group with a saturation over 90%, 1.9% + 5.4% in the group with a saturation between 80% - 90%, and 5.8% +- 9.6% in the group with a saturation below 80%. Bias and precision between saturations calculated by the blood gas analyzer (SaO-zcalc) and SaOz measured by the CO-oximeter were 1.3 and 1.2 at a saturation over 90%, 1.7 and 2.1 at a saturation between 80% and 90%, 3.2 and 2.9 at a saturation below 80%, respectively.
RESULTS
The ages of the 56 children ranged from 3 days to 13 years (mean 24.5 months), and their weight varied between 3.3 and 40 kg (mean 10.7 kg). There were a total of 112
pulse oximetry measurements and simultaneously drawn arterial blood samples, which were statistically analyzed. The obtained data were divided into three groups: SaOl over 90%, between 80% and 90%, and below 80%. Table 1 shows the measured clinical parameters of the three groups. The mean COHb was 1% and the mean MetHb was 0.4% in all groups. Regression analysis of all simultaneously measured SpOz and SaOz values yielded the following equation: SpOZ = 0.725 SaO?, + 25% with a correlation coefficient of r = 0.91 (Fig 1). Saturation data are presented in Table 2. SaOz and SpOZ differed significantly (P < 0.01) within the second group, saturation between 80% and 90%, and within the third group, saturation below 80%. Bias and precision between saturation measured by the pulse oximeter (SpO$
SpO2 = 0.72
0
10
20
30
40
50 Sa02 (%)
60
Sa02 + 25 (%)
70
80
90
100
Fig 1. Comparison of arterial oxygen saturltions measured with the pulse oximeter (SpO,) and CO-oximeter (SaO,). The solid line is shown for reference. Linear regrearion ofthe 112simultaneous measurements yields the dashed line.
Table 2. Saturation Data SaO* W)
Group Mean k SD
(Sa02 range %) >90
Range
Bias -t Precision
(%) Range
93.4
f
88-99
0.5
84.2
‘- 2.9t
80-89.2
86 + 3.9t
79-97
1.9 + 2.7”
(-2)-8
<80
70.7
2 7.5t
49-79.9
76 + 6t
62-87
5.8 2 4.8’
(-2)-16
(n = 47)
between the groups.
tP < 0.01 within the 9roup.
2.3
Mean -t SD
80 to 90 (n = 44)
*P < 0.01
In = 21)
Paired Difference
SPO2 I%) Range 90-97
93.8
+ 3
2 2.5*
C-51-6
PULSE OXIMETRY
IN CYANOTIC
HEART DISEASE
15.
63
n=21
10. Fig 2. Range over 90%: Agreement between saturation measured by pulse oximeter (SpO,) and by CO-oximeter (SaO*). The difference between simultaneous SpO* and SaOs measurements [SpOz - SaOJ is plotted against the average of the two measurements [(SpOz +SaO,):2]. The solid line represents the mean difference (d = 0.5%). the broken line represents the limits of agreement (d + 2SD = 5.5%. d 2SD = -4.5%).
d+2SD 5_--------------____ OS-. O.~--~~&~&. d-2SD _~~~_~~~~~~-.--------_-
-10, . _ . . . , . , , , . , . , 4’5 sb 5.5 60 65 70 75 80 (Sp02 + Sa02) / 2 (%)
DISCUSSION
Pulse oximetry has become the most important patient monitoring device during many diagnostic and surgical procedures. Therefore, it is critical to evaluate the suitability and validity of this noninvasive method even at extreme clinical ranges. These results show that pulse oximetry measurements at saturation levels below 80% are not as accurate as those at SaOz values over 80%, and tend to overestimate the SaOs at low levels. During marked desaturation (SaOz below 80%), the deviation of a given pulse oximeter reading may be twice that found in the usual clinical situation with an Sa02 over 90%. There are several known factors that can cause incorrect saturation measurements by pulse oximetry. Increasing levels of MetHb lead to a decrease of SpO2 readings. At
45
.
, 50
.
, 55
a0 0
The linear regression analysis of paired differences between SpOz and SaOz revealed no influence of either age, weight, or pH. The correlation between hematocrit and bias was 0.21 at hematocrits below 40%, 0.28 at hematocrits between 40% and SO%, and 0.16 at hematocrits over 50%.
-104
8
.
, 60
.
, . , . , . , 65 :L32)Y2 80 (Spo2 (96)
.
, . , 85 90
.
, 95
.
. 100
very high levels of MetHb of 60% or more, an Sp02 of 85% is shown on the pulse oximeter independent of true arterial oxygen saturation.9J0 In the present study, the MetHb was 0.4% in all three groups and, therefore, cannot be responsible for the inaccuracy. Carboxyhemoglobin is treated by the pulse oximeter as oxyhemoglobin, so that SpOz represents the sum of COHb and O&RX In all three groups, COHb was only about 1% and, therefore, did not play a significant role in respect to accuracy. Avoidable causes for incorrect readings from pulse oximeters include improper placement of the sensor,ii ambient light shining on the sensor,r2 or infrared heat lamps.13 These potential artifacts were avoided during the measurements. Vasoconstriction either induced by cold14 or by vasoactive drugs’ may also disturb signal detection by a pulse oximeter as shown by differing HRs detected by the pulse oximeter and electrocardiogram. The same problem is seen in shock or low cardiac output syndromes.ls Most investigators agree that pulse oximetry is usually a reliable monitor over a wide range of arterial saturations,4,6J6,17but that pulse oximeters tend to measure less accurately at lower saturations4v6J7 Chapman et al6 pointed
.
, 85
.
, 90
.
, 95
.
c 100
Fig 3. Range between 80% and 90%: Agreement between saturation measured by pulse oximeter (Sp&) and by CO-oximeter (Sa02). The difference between simultaneous SpOs and Sa02 measurements (Sp02 SaOJ is plotted against the average of the two measurements [(SpOz + SaOJ:2]. The solid line represents the limits of agreement (d + 2SD = 7.3%. d -2SD = -3.5%).
SCHMITT ET AL
n=47 15._____----~_~__~~~---~
A
A
A
A
10 -
d+2SD
i A
AA
A &
A
5.
A
AA d
4
A\
A
A
A A
,” A
0. ___-------
_------
>A^ A
A&
--
-5 .-
-107 45
d-2SD
.
I 50
,
, 55
,
, 60
.
,
.
,
m ,
m ,
I
65 70 75 80 (Sp02 + Sa02) / 2 (%)
out that at an SaOz under 75%, the instrument readings were progressively higher than the SaOz measured by a CO-oximeter.6 Boxer et al4 found in their study of children with cyanotic CHD that the pulse oximeter was less precise at an Sa02 under 80%. Bias and precision were -0.56 and 2.33 at SaOz over 80%, and -1.09 and 4.46 at Sa02 below 80%. Because their bias represents CO-oximeter minus pulse oximeter values, the negativity of the reported bias numbers indicates an increasing overestimation of Sa02 by pulse oximetry at low saturation, which is similar to the present results. Lee et al7 recently studied the effects of a falling hematocrit on pulse oximetry. They demonstrated an increase in pulse oximeter failure rates at hematocrit levels below 10% and, surprisingly, a similar increase in pulse oximeter failures at hematocrits over 40%. Because there were only a few data points in the high hematocrit range, the authors did not draw conclusions from that observation. During severe anemia with low hemoglobin concentrations, the difference between reflected and transmitted light caused by oxidized and reduced hemoglobin is very small relative to the background noise. This small signal-to-noise ratio is considered responsible for less accurate measurements.7 In the present study, high hematocrit levels seem unlikely to be responsible for the increasing inaccuracy of
, 85
,
, 90
I
, 95
.
-
r 100
Fig 4. Range below 80%: Agreement between saturation measured by pulse oximeter (SpOz) and by CO-oximeter (SaOz). The difference between simultaneous SpO, and SaO, measurements (SpO, - SaOJ is plotted against the average of the two measurements [(SpO, + SaO,):Z]. The solid line represents the mean difference (d = 5.8%). the broken line represents the limit of agreement (d + 2SD = 15.4%,d - 2SD =-3.8%).
pulse oximetry at low saturations, because linear regression analysis showed no influence of hematocrit on bias. Bias and precision between SaOl measured by the CO-oximeter and SaOz calculated by blood gas analyzer showed better accordance than between SaOz and SpO?. The blood gas analyzer measures POz, pH, and PC02, and calculates SaOz using the formulas of Kelman, Severinghaus, or Siggaard-Andersen. The calculation of SaOz per se is only correct when COHb and MetHb levels are so low that they can be neglected. I* In this study, COHb was 1% and MetHb was 0.4%, offering an ideal basis for the suitability and accuracy of Sa02 calculated by the blood gas analyzer. It is concluded that pulse oximetry is a valuable monitor showing the arterial blood oxygen saturation, and that it is accurate over a wide range of saturations. During marked desaturation, however, especially in the SaO? range under 80%, the accuracy of the pulse oximeter is impaired. With desaturation, the margin of safety for the patient is small; thus, it is advisable to keep in mind that the pulse oximeter reading may differ significantly from the actual value. In the presence of desaturation, it is reasonable to check pulse oximeter readings by a CO-oximeter or, if not available, and no significant COHb or MetHb levels are suspected, by a blood gas analyzer.
REFERENCES 1. Mihm FG, Halperin BD: Noninvasive detection of profound arterial desaturations using a pulse oximetry device. Anesthesiology 62:85-87, 1985 2. Knill RL, Clement RN, Kieraszewicz NT, Dodgson BG: Assesment of two noninvasive monitors of arterial oxygenation in anesthetized man. Anesth Aualg 61:582-586,1982 3. Yelderman M, New W Jr: Evaluation of pulse oximetry. Anesthesiology 59:349-352,1983 4. Boxer RA, Gottesfeld I, Singh S, et al: Noninvasive pulse oximetry in children with cyanotic congenital heart disease. Crit Care Med 15:1062-1064,1987 5. Ridley SA: A comparison of two pulse oximeters. Anaesthesia 43:136-140, 1988
6. Chapman KR, Liu FL, Watson RM, Rebuck AS: Range of accuracy of two-wavelength oximetry. Chest 89:540-542, 1986 7. Lee S, Tremper KK, Barker S: Effects of anemia on pulse oximetry and continuous mixed venous hemoglobin saturation monitoring in dogs. Anesthesiology 75:118-122, 1991 8. Bland JM, Altmann DG: Statistical methods for assessing agreement between two methods of clinical measurement. Lancer 8:307-310, 1986 9. Rieder HU, Frei FJ, Zbinden AM, Thomson DA: Pulse oximetry in methaemoglobinaemia. Anaesthesia 44:326-327, 1989 10. Alexander CM, Teller LE, Gross JB: Principles of pulse oximetry. Anesth Analg 68:368-376,1989 11. Kelleher JF, Ruff RH: The penumbra effect: Vasomotion-
PULSE OXIMETRY
IN CYANOTIC
HEART DISEASE
dependent pulse oximeter artifact due to probe malposition. Anesthesiology 71:787-791,1989 12. Costarino AT, Davis DA, Keon TP: Falsely normal saturation reading with the pulse oximeter. Anesthesiology 67:830-831, 1987 13. Brooks TD, Paulus DA, Winkle WE: Infrared heat lamps interfere with pulse oximeters. Anesthesiology 61:630, 1984 14. Wilkins CJ, Moores M, Hanning CD: Comparison of pulse oximeters: Effects of vasoconstriction and venous engorgement. Br J Anaesth 62:439-444,1989 15. Tremper KK, Hufstedler S, Zaccari J, et al: Pulse oximetty
65
and transcutaneous PO2 during hemorrhagic and normotensive shock in dogs. Anesthesiology 61:163A, 1984 16. Nickerson BG, Sarkisian C, Tremper KK: Bias and precision of pulse oximeters and arterial oximeters. Chest 93:515-517, 1988 17. Kagle DM, Alexander CM, Berko RS, et al: Evaluation of the Ohmeda 3700 pulse oximeter: Steady-state and transient response characteristics. Anesthesiology 66:376-380,1987 18. Zander R: Vergleich der berechneten mit der in vitro gemessenen arteriellen 02-Slttigung, in Zander R, Mertzlufft FO (ed): Der Sauerstoff-Status des arteriellen Blutes, Symp. Mainz 1986, pp 120-125
Heart Disease
Hubert J. Schmitt, MD, Wolfram H. Schuetz, MD, Peter A. Proeschel, PhD, and Christel Jaklin The use of a pulse oximeter to monitor arterial oxygen saturation (SaOr) is considered accurate and reliable in the range of 90% to 100%. However, differing reports exist about the accuracy with desaturation. Thus, the suitability of pulse oximetry in desaturated patients was evaluated using a Nellcor N-100 oximeter. In 56 children with cyanotic congenital heart disease, the pulse oximeter reading was compared with the direct measurement of SaOz by a CO-oximeter OSM 3. The influence of high hematocrit values on the accuracy at low saturation was also investigated. All oxygen saturation measurements (two per child) were carried out after induction of anesthesia (ketamine, fentanyl, pancuronium) during a “steady state” before the surgical procedure. The results indicate that at low levels of saturation (SaOr below 80%). pulse oximetry is not as accurate as at higher saturations,
P
ULSE OXIMETRY is a standard part of patient monitoring during anesthesia. It is noninvasive and allows continuous beat-to-beat assessment of arterial oxygen saturation. There are several studies confirming the value of pulse oximetty during acute arterial desaturation.1-3
Nevertheless,
accuracy
of pulse
there
addition,
Lee et al7 recently
oximetry
are
differing
during
severe
demonstrated
reports
about
desaturation.4-” the inaccuracy
the In
METHODS With institutional approval and informed consent from parents, 56 children with cyanotic CHD were studied on the day of cardiac surgery. The operations were for transposition of great arteries (TGA) (n = 17), tetralogy of Fallot (n = 15), single ventricle (n = 8), tricuspid atresia (n = 6), and complex cyanotic heart disease (n = 10). Noninvasive arterial oxygen saturation (SpOz) was determined using a Nellcor N-100 pulse oximeter (Nellcor, Inc, Haywood, CA). According to the user’s manual, the appropriate sensor was applied to a finger, great toe, or the dorsum of the foot, but not using the extremities where arterial or venous catheters were located. During the measurements, no artificial light was allowed to fall on the sensor. To ensure reliability, the heart rate (HR) shown by the pulse oximeter was verified by comparison against the HR on the ECG monitor. The time of all saturation measurements (two per child) was during a steady state following the induction of anesthesia after placing all monitors including the arterial catheter and rectal and esophageal temperature probes, but prior to the surgical procedure. All children were placed on heating and cooling equipment as used routinely during cardiopulmonary bypass (CPB) to keep their temperature constant. The anesthesia consisted of ketamine, fentanyl, and pancuronium. The children were ventilated with 50% 02 in air, and inhalation agents or vasoactive substances were not used. Blood pressure, HR, and body temperature were continuously monitored.
Vascular
KEY WORDS: measurement notic heart disease
techniques, pulse oximetry, cya-
When the pulse oximeter showed a constant saturation for at least 2 minutes, the value was recorded and two blood samples were obtained anaerobically from the arterial catheter. Both blood samples were immediately analyzed using a CO-oximeter OSM-3 (Radiometer, Willich, Germany) for in vitro oxygen saturation (SaO,) and a Corning 288 blood gas analyzer (Corning, Medfield, MA) for a complete set of blood gases. SaOz measured by the CO-oximeter represents fractional saturation defined as
of
pulse oximetry in severe anemia with hematocrits below 10%. In a similar way abnormally high hematocrits could reduce the accuracy of this method. To evaluate the suitability and validity of pulse oximetry in the clinical situation of continual low saturation and high hematocrit, children with cyanotic congenital heart disease (CHD) were studied. Noninvasively measured arterial oxygenation (Sp02) was compared with direct measurement of arterial oxygen saturation (SaOZ) by a CO-oximeter.
Journalof Cardiothoracic and
and overestimates the true value. Bias and precision between saturations measured by the pulse oximeter and the CO-oximeter were 5.8 and 4.8 in the group with a saturation below 80%. and 0.5 and 2.5 in the group with a saturation over 90%. respectively. Because the margin of safety for a patient is small when arterial saturation levels are under 80%, it is advisable under this condition to check the SaOr measurements by a CO-oximeter. High hematocrit levels did not seem to be responsible for impaired accuracy of pulse oximetry at saturation values below 80%. Copyright 0 1993 by W.B. Saunders Company
100%.
OzHb/(OzHb
+ Hb + COHb
+ MetHb)
where 02Hb = oxyhemoglobin, Hb = reduced hemoglobin, COHb = carboxyhemoglobin, and MetHb = methemoglobin. The blood gas status analyzed by the Corning 288 consisted of arterial pH, oxygen partial pressure (PaO,), carbon dioxide partial pressure (PaCOz), hematocrit (Hct), and calculated arterial oxygen saturation (SaOzcalc). Both machines were calibrated daily. Data are presented as mean 2 SD. Statistical analysis of paired SpOz and SaOz values was performed using standard linear regression, bias, and precision, as well as the agreement according to Bland and Altmanns Bias represents the mean difference of paired values of in vivo saturation (SpOz) measured by the pulse oximeter and in vitro saturation (Sa02) measured by the COoximeter. Precision is the standard deviation of these differences. The differences between saturation values simultaneously obtained by two different methods were plotted against the average of the two measurements. The mean difference ? SD and the limits of agreement, described by mean difference + 2 SD, within which 95% of the differences will lie, were calculated. Correlations between specific variables (SpO2, bias, Hct, pH) were tested using standard linear regression. For comparison within a group the Wilcoxon test was used. Comparison between the groups was performed with the Kruskal-Wallis test. P < 0.05 was considered significant.
From the Department of Anesthesiology, Friedrich-AlexanderUniversityErlangen-Ntimberg Erlangen, Germany. Presented as a poster at the Sixth Annual Meeting of the European Association of Cardiothoracic Anesthesiologists, June 4-7, 1991, Milano, Italy. Address reprint requests to Dr. H.J. Schmitt, Department of Anesthesiology, Maximiliansplatz, University Erlangen-Niimberg, D-8520 Erlangen, Germany. Copyright o I993 by W.B. Saunders Company 1053-0770/93/0701-0012$03.0010
Anesthesia, Vol7, No 1 (February), 1993: pp 61-66
61
SCHMITT ET AL
62
Table 1. Measured Clinical Parameters saoz
COHb
MetHb
Range (%I
(%)
W)
0.9 -c 0.7
0.5 -t 0.2
1.0 c 0.7
0.5 -t 0.2
7.40 -c 0.04*
0.9 +- 0.6
0.4 2 0.2
7.37 2 0.05”
>9O(n
21)
=
80to90(n
= 44)
<80 (n = 47)
PH
7.42 t 0.04*
PaOz
PaC02
Hct
BP systolic
BP diastolic
(mmHgl
(mmHg)
WI
WmHg)
ImmHg)
74 t 137
33 c 5
44 -t 7*
51 +6t
34 2 5
48 + 9”
42 -+ 5t
36 rt 5
52 z IO*
86-+ 16
495
HR
Temperature
(beatsimin)
(“C)
14
97 + 21*
36.2 -c 0.5
82-c 15
50 + 14
107 +- 22*
36.3 -c 0.7
81 k 15
49k
119 2 16*
36.5 + 0.9
11
NOTE. Mean ? SD. *P < 0.05. tP < 0.01 between the groups.
and saturation measured by the CO-oximeter (SaOa) were 0.5% and 2.5% in the group with a saturation over 90%, 1.9% and 2.7% at an SaOl between 80% and 90%, and 5.8% and 4.8% in the third group with an SaOZ below 80%, showing the pulse oximeter to overestimate arterial saturation at low saturation values (Table 2). Paired differences of SpOZ and SaOz varied significantly among the three groups (P < 0.01). Figures 2 through 4 show the agreement between SpOZ and SaOz for each group analyzed according to Bland and A1tmann.s The limits of agreement (mean difference k 2 SD) were 0.5% t 5% in the group with a saturation over 90%, 1.9% + 5.4% in the group with a saturation between 80% - 90%, and 5.8% +- 9.6% in the group with a saturation below 80%. Bias and precision between saturations calculated by the blood gas analyzer (SaO-zcalc) and SaOz measured by the CO-oximeter were 1.3 and 1.2 at a saturation over 90%, 1.7 and 2.1 at a saturation between 80% and 90%, 3.2 and 2.9 at a saturation below 80%, respectively.
RESULTS
The ages of the 56 children ranged from 3 days to 13 years (mean 24.5 months), and their weight varied between 3.3 and 40 kg (mean 10.7 kg). There were a total of 112
pulse oximetry measurements and simultaneously drawn arterial blood samples, which were statistically analyzed. The obtained data were divided into three groups: SaOl over 90%, between 80% and 90%, and below 80%. Table 1 shows the measured clinical parameters of the three groups. The mean COHb was 1% and the mean MetHb was 0.4% in all groups. Regression analysis of all simultaneously measured SpOz and SaOz values yielded the following equation: SpOZ = 0.725 SaO?, + 25% with a correlation coefficient of r = 0.91 (Fig 1). Saturation data are presented in Table 2. SaOz and SpOZ differed significantly (P < 0.01) within the second group, saturation between 80% and 90%, and within the third group, saturation below 80%. Bias and precision between saturation measured by the pulse oximeter (SpO$
SpO2 = 0.72
0
10
20
30
40
50 Sa02 (%)
60
Sa02 + 25 (%)
70
80
90
100
Fig 1. Comparison of arterial oxygen saturltions measured with the pulse oximeter (SpO,) and CO-oximeter (SaO,). The solid line is shown for reference. Linear regrearion ofthe 112simultaneous measurements yields the dashed line.
Table 2. Saturation Data SaO* W)
Group Mean k SD
(Sa02 range %) >90
Range
Bias -t Precision
(%) Range
93.4
f
88-99
0.5
84.2
‘- 2.9t
80-89.2
86 + 3.9t
79-97
1.9 + 2.7”
(-2)-8
<80
70.7
2 7.5t
49-79.9
76 + 6t
62-87
5.8 2 4.8’
(-2)-16
(n = 47)
between the groups.
tP < 0.01 within the 9roup.
2.3
Mean -t SD
80 to 90 (n = 44)
*P < 0.01
In = 21)
Paired Difference
SPO2 I%) Range 90-97
93.8
+ 3
2 2.5*
C-51-6
PULSE OXIMETRY
IN CYANOTIC
HEART DISEASE
15.
63
n=21
10. Fig 2. Range over 90%: Agreement between saturation measured by pulse oximeter (SpO,) and by CO-oximeter (SaO*). The difference between simultaneous SpO* and SaOs measurements [SpOz - SaOJ is plotted against the average of the two measurements [(SpOz +SaO,):2]. The solid line represents the mean difference (d = 0.5%). the broken line represents the limits of agreement (d + 2SD = 5.5%. d 2SD = -4.5%).
d+2SD 5_--------------____ OS-. O.~--~~&~&. d-2SD _~~~_~~~~~~-.--------_-
-10, . _ . . . , . , , , . , . , 4’5 sb 5.5 60 65 70 75 80 (Sp02 + Sa02) / 2 (%)
DISCUSSION
Pulse oximetry has become the most important patient monitoring device during many diagnostic and surgical procedures. Therefore, it is critical to evaluate the suitability and validity of this noninvasive method even at extreme clinical ranges. These results show that pulse oximetry measurements at saturation levels below 80% are not as accurate as those at SaOz values over 80%, and tend to overestimate the SaOs at low levels. During marked desaturation (SaOz below 80%), the deviation of a given pulse oximeter reading may be twice that found in the usual clinical situation with an Sa02 over 90%. There are several known factors that can cause incorrect saturation measurements by pulse oximetry. Increasing levels of MetHb lead to a decrease of SpO2 readings. At
45
.
, 50
.
, 55
a0 0
The linear regression analysis of paired differences between SpOz and SaOz revealed no influence of either age, weight, or pH. The correlation between hematocrit and bias was 0.21 at hematocrits below 40%, 0.28 at hematocrits between 40% and SO%, and 0.16 at hematocrits over 50%.
-104
8
.
, 60
.
, . , . , . , 65 :L32)Y2 80 (Spo2 (96)
.
, . , 85 90
.
, 95
.
. 100
very high levels of MetHb of 60% or more, an Sp02 of 85% is shown on the pulse oximeter independent of true arterial oxygen saturation.9J0 In the present study, the MetHb was 0.4% in all three groups and, therefore, cannot be responsible for the inaccuracy. Carboxyhemoglobin is treated by the pulse oximeter as oxyhemoglobin, so that SpOz represents the sum of COHb and O&RX In all three groups, COHb was only about 1% and, therefore, did not play a significant role in respect to accuracy. Avoidable causes for incorrect readings from pulse oximeters include improper placement of the sensor,ii ambient light shining on the sensor,r2 or infrared heat lamps.13 These potential artifacts were avoided during the measurements. Vasoconstriction either induced by cold14 or by vasoactive drugs’ may also disturb signal detection by a pulse oximeter as shown by differing HRs detected by the pulse oximeter and electrocardiogram. The same problem is seen in shock or low cardiac output syndromes.ls Most investigators agree that pulse oximetry is usually a reliable monitor over a wide range of arterial saturations,4,6J6,17but that pulse oximeters tend to measure less accurately at lower saturations4v6J7 Chapman et al6 pointed
.
, 85
.
, 90
.
, 95
.
c 100
Fig 3. Range between 80% and 90%: Agreement between saturation measured by pulse oximeter (Sp&) and by CO-oximeter (Sa02). The difference between simultaneous SpOs and Sa02 measurements (Sp02 SaOJ is plotted against the average of the two measurements [(SpOz + SaOJ:2]. The solid line represents the limits of agreement (d + 2SD = 7.3%. d -2SD = -3.5%).
SCHMITT ET AL
n=47 15._____----~_~__~~~---~
A
A
A
A
10 -
d+2SD
i A
AA
A &
A
5.
A
AA d
4
A\
A
A
A A
,” A
0. ___-------
_------
>A^ A
A&
--
-5 .-
-107 45
d-2SD
.
I 50
,
, 55
,
, 60
.
,
.
,
m ,
m ,
I
65 70 75 80 (Sp02 + Sa02) / 2 (%)
out that at an SaOz under 75%, the instrument readings were progressively higher than the SaOz measured by a CO-oximeter.6 Boxer et al4 found in their study of children with cyanotic CHD that the pulse oximeter was less precise at an Sa02 under 80%. Bias and precision were -0.56 and 2.33 at SaOz over 80%, and -1.09 and 4.46 at Sa02 below 80%. Because their bias represents CO-oximeter minus pulse oximeter values, the negativity of the reported bias numbers indicates an increasing overestimation of Sa02 by pulse oximetry at low saturation, which is similar to the present results. Lee et al7 recently studied the effects of a falling hematocrit on pulse oximetry. They demonstrated an increase in pulse oximeter failure rates at hematocrit levels below 10% and, surprisingly, a similar increase in pulse oximeter failures at hematocrits over 40%. Because there were only a few data points in the high hematocrit range, the authors did not draw conclusions from that observation. During severe anemia with low hemoglobin concentrations, the difference between reflected and transmitted light caused by oxidized and reduced hemoglobin is very small relative to the background noise. This small signal-to-noise ratio is considered responsible for less accurate measurements.7 In the present study, high hematocrit levels seem unlikely to be responsible for the increasing inaccuracy of
, 85
,
, 90
I
, 95
.
-
r 100
Fig 4. Range below 80%: Agreement between saturation measured by pulse oximeter (SpOz) and by CO-oximeter (SaOz). The difference between simultaneous SpO, and SaO, measurements (SpO, - SaOJ is plotted against the average of the two measurements [(SpO, + SaO,):Z]. The solid line represents the mean difference (d = 5.8%). the broken line represents the limit of agreement (d + 2SD = 15.4%,d - 2SD =-3.8%).
pulse oximetry at low saturations, because linear regression analysis showed no influence of hematocrit on bias. Bias and precision between SaOl measured by the CO-oximeter and SaOz calculated by blood gas analyzer showed better accordance than between SaOz and SpO?. The blood gas analyzer measures POz, pH, and PC02, and calculates SaOz using the formulas of Kelman, Severinghaus, or Siggaard-Andersen. The calculation of SaOz per se is only correct when COHb and MetHb levels are so low that they can be neglected. I* In this study, COHb was 1% and MetHb was 0.4%, offering an ideal basis for the suitability and accuracy of Sa02 calculated by the blood gas analyzer. It is concluded that pulse oximetry is a valuable monitor showing the arterial blood oxygen saturation, and that it is accurate over a wide range of saturations. During marked desaturation, however, especially in the SaO? range under 80%, the accuracy of the pulse oximeter is impaired. With desaturation, the margin of safety for the patient is small; thus, it is advisable to keep in mind that the pulse oximeter reading may differ significantly from the actual value. In the presence of desaturation, it is reasonable to check pulse oximeter readings by a CO-oximeter or, if not available, and no significant COHb or MetHb levels are suspected, by a blood gas analyzer.
REFERENCES 1. Mihm FG, Halperin BD: Noninvasive detection of profound arterial desaturations using a pulse oximetry device. Anesthesiology 62:85-87, 1985 2. Knill RL, Clement RN, Kieraszewicz NT, Dodgson BG: Assesment of two noninvasive monitors of arterial oxygenation in anesthetized man. Anesth Aualg 61:582-586,1982 3. Yelderman M, New W Jr: Evaluation of pulse oximetry. Anesthesiology 59:349-352,1983 4. Boxer RA, Gottesfeld I, Singh S, et al: Noninvasive pulse oximetry in children with cyanotic congenital heart disease. Crit Care Med 15:1062-1064,1987 5. Ridley SA: A comparison of two pulse oximeters. Anaesthesia 43:136-140, 1988
6. Chapman KR, Liu FL, Watson RM, Rebuck AS: Range of accuracy of two-wavelength oximetry. Chest 89:540-542, 1986 7. Lee S, Tremper KK, Barker S: Effects of anemia on pulse oximetry and continuous mixed venous hemoglobin saturation monitoring in dogs. Anesthesiology 75:118-122, 1991 8. Bland JM, Altmann DG: Statistical methods for assessing agreement between two methods of clinical measurement. Lancer 8:307-310, 1986 9. Rieder HU, Frei FJ, Zbinden AM, Thomson DA: Pulse oximetry in methaemoglobinaemia. Anaesthesia 44:326-327, 1989 10. Alexander CM, Teller LE, Gross JB: Principles of pulse oximetry. Anesth Analg 68:368-376,1989 11. Kelleher JF, Ruff RH: The penumbra effect: Vasomotion-
PULSE OXIMETRY
IN CYANOTIC
HEART DISEASE
dependent pulse oximeter artifact due to probe malposition. Anesthesiology 71:787-791,1989 12. Costarino AT, Davis DA, Keon TP: Falsely normal saturation reading with the pulse oximeter. Anesthesiology 67:830-831, 1987 13. Brooks TD, Paulus DA, Winkle WE: Infrared heat lamps interfere with pulse oximeters. Anesthesiology 61:630, 1984 14. Wilkins CJ, Moores M, Hanning CD: Comparison of pulse oximeters: Effects of vasoconstriction and venous engorgement. Br J Anaesth 62:439-444,1989 15. Tremper KK, Hufstedler S, Zaccari J, et al: Pulse oximetty
65
and transcutaneous PO2 during hemorrhagic and normotensive shock in dogs. Anesthesiology 61:163A, 1984 16. Nickerson BG, Sarkisian C, Tremper KK: Bias and precision of pulse oximeters and arterial oximeters. Chest 93:515-517, 1988 17. Kagle DM, Alexander CM, Berko RS, et al: Evaluation of the Ohmeda 3700 pulse oximeter: Steady-state and transient response characteristics. Anesthesiology 66:376-380,1987 18. Zander R: Vergleich der berechneten mit der in vitro gemessenen arteriellen 02-Slttigung, in Zander R, Mertzlufft FO (ed): Der Sauerstoff-Status des arteriellen Blutes, Symp. Mainz 1986, pp 120-125