Reliability of Pulse Oximetry During Exercise in Pulmonary
Patients*
Pierre]. L. Escourrou, M.D.; Marie Franfoise Delaperche, M.D.; and Agnes V&aseaux, B.S.
To evaluate the reliability of pulse oximetry during exercise, we studied 101 patients primarily with chronic pulmonary diseases. Three devices were used on different patients. Radial arterial blood was sampled at rest and maximal exercise simultaneously to pulse oximetric determination. Measured blood oxygen saturation was significantly different from noninvasive saturation at rest and also at exercise for each device. Nevertheless, changes in pulse oximetry from rest to exercise were significantly correlated with measured saturation for all three devices. Direction of
changes in saturation from rest to exercise was correctly evaluated by transcutaneous oximetry in all but six instances where changes were less than 4 percent. Although measured and transcutaneous saturations are signi6cantly different, we conclude that pulse oximetry reliably estimates changes in arterial saturation between rest and exercise for a clinical purpose. None of the three tested devices was better compared with the others in estimating saturation changes at exercise. (Chest 1990; 97:635-38)
Measurement of respiratory gas exchange during exercise stress testing is useful in the evaluation of patients with respiratory diseases. 1•2 Changes in blood oxygenation during exercise may be used to detect lung disease, evaluate the severity and follow the response to therapy. However, direct sampling of arterial blood by arterial cannulation is potentially dangerous.a-& Single arterial puncture at peak exercise, although less invasive, is not always successful, especially in patients with radial arteries of small diameter such as women and children. Rapid changes in arterial blood gas levels preclude precise evaluation of peak exercise value during recovery more so when the sample is obtained later than 20 s after the end of exercise. 7•8 Blood also may be sampled at the ear or the finger at exercise, but the reliability of capillarized blood has been questioned and sampling also may be difficult at exercise. 9 On the other hand, transcutaneous oximetry devices are more and more widely used in resting clinical conditions. One previous study evaluated two ear oximeters (HP 47201 A and Biox II A) at rest and exercise in pulmonary patients and showed that accuracy was comparable and clinically relevant for both devices. 10 Another report during combined hypoxia and exercise indicated that the accuracy of the pulse oximeter Biox II was lower than that of eight wavelength HP 47201 A. 11 Since these studies, several new
oximeters with improved technology have been marketed. We therefore decided to assess the reliability at exercise of three recent pulse oximeters with more efficient signal processing.
*From Explorations Fonctionnelles, Respiratoires, Hopital A. Beclere, Ciamart, France. Su~~ed in part by a grant from Comite National Contre les "es Respiratoires et Ia Tuberculose. M Manuscript received April18; revision accepted August 17. Reprint requests: Dr. Escourrou, Explorations Rmctionneles Resp, Hopital Beclere, 92141 Clamart, France
PATIENTS
One hundred and one patients have been studied whose ages ranged from 17 to 81 years old; 72 were men and 29 were women. Five patients (two men and three women) were moderately pigmented but not black. Their pathologic conditions were: COPD (33 cases), emphysema (five cases), sarcoidosis (15 cases), pulmonary fibrosis (eight cases), Hodgkin's disease (seven cases), primary pulmonary hypertension (six cases), cirrhosis of the liver (12 cases), restrictive pulmonary disease (two cases), miscellaneous diseases (seven cases), dyspnea of unknown origin (six cases). METHODS
Exercise Protocol Patients were tested on a bicycle ergometer (ergometry system 380B Siemens) by a symptom-limited incremental protocol of 20to 30-W in 2 to 3-min stages. A 12-lead ECG was recorded at rest and the end of each stage, and the heart rate was continuously displayed (custom-made ratemeter with a time constant of 3 s). Blood pressure was measured at rest and at the end of each stage by a sphygmomanometer. Average maximal load was 90±38 W (range, 30 to 210 W) for 9±3.6 min (4 to 18 min) exercise durations.
Blood Gas Measurement
Radial arterial blood was sampled using a microsampler (AVL Biomedical Instruments-Graz, Austria) at rest sitting on the bicycle and at maximal exercise just before stopping. Blood was immediately analyzed for pH, Po2 , and Pco2 using a Corning 178 gas analyzer (Corning Glass, Melfied, MA). Standard calibration procedures were used, ie, one-point gas calibration every 20 to 30 min and a two-point calibration every 2 h. The analyzer was further checked once a week with tonometered blood and the standard deviations for duplicate measurements were 2.1 mm Hg for Po, and 1.1 mm Hg for Pco2 • The calculated Sa0 2 was CHEST I 97 I 3 I MARCH. 1990
635
computed from the measured Po., Pco2 and pH, assuming normal P.., and a body temperature of37"C for these usually short exercise durations. Hemoglobin concentration and measured SaO. were determined by a spectrophotometric method (OSM., Radiometer, Copenhagen, Denmark). When a difference between measured and calculated SaO, larger than 3 percent was observed, carbon monoxide concentration was measured by the reductase method. 11 The percentage of HbCO was then subtracted from measured saturation which does not differentiate between Hb02 and HbCO. No correction was made for bilirubin as none of the cirrhotic patients had a severe hepatic insufficiency with bilirubin level higher than 20 p.moi!L.
Pulse Oximetry Three oximeters were studied: the Biox 3700 (Ohmeda version B), CSI 501 + (Criticare), and the Nellcor N200 (version 3.1). The latest device is provided with the option of artefact reduction by ECG synchronization. The three devices continuously display saturation and heart rate. All of them were used in the fast mode: saturation and heart rate averaging times are 3 and 5 s for Biox 3700, 2 and 3 s for the Nellcor N200 and every eight beats for the CSI 501 +, respectively. The instructions of the manufacturers were followed for each oximeter: the Biox 3700 and Criticare were connected to an earpiece. After calibration, the ear probes were attached to the earlobe previously massaged for 30 s with an alcohol saturated swab; the transmission cable was looped within a headband and taped to the subject's neck to minimize any movement artifact. Because no practical earpiece was available with the Nellcor N200, this device was used with the digital probe and the unit was linked to the ECG output of the electrocardiograph for artefact reduction. When a low signal output was indicated at rest by the device, the ear lobe or finger was massaged again with a revulsive paste Finalgon (Boehringer lngelheim). Notwithstanding these procedures, a satisfactory pulse signal output was not obtained at rest in five patients which were not included in the studied population. No relationship to a specific device or pigmentation was noted in these patients with apparently poor peripheral perfusion. The Biox 3700 was used in 48 patients, the Nellcor N200 in 24 patients and the Criticare in 29 patients.
Statistical Analysis Individual transcutaneous Sa02 values were compared with calculated and measured values on blood samples by paired t tests. Accuracy has been compared between devices by one-way analysis of variance on differences between invasive and transcutaneous saturations.
A
100
B
REST 0 EXERCISE X
90
100
Regression equations between variables have been computed by the least squares method using the SPSS-PC statistical package. Significance has been assessed at the O.OSlevel.
REsuLTS On the 101 patients studied at rest and exercise, 201 calculated Sa02 values, 195 measured Sa02 values and 198 transcutaneous values were obtained. In a few cases, indeed, the volume of sampled blood was insufficient to perform both spectrophotometry and gas pressure determinations. In three patients the transcutaneous value was not obtained at exercise; the device indicated a poor signal strength with erratic Sa02 values (two patients with the Biox 3700, one with the Nellcor N200). The measured range of Sa02 was 68 to 100 percent. Accuracy of Absolute Measurements of Transcutaneous Oximetry
Figure 1 gives the regressions of transcutaneous saturation on measured Sa02 at rest and exercise for each studied device. Although the relationships were statistically significant for all three devices, the scatter was large and the standard errors of the estimates were between 2 and 3 percent. Furthermore, the range of differences between measured and transcutaneous saturation values was -9 to + 16 percent. The mean difference was 1.1 ± 3 percent. These differences were not related to skin pigmentation nor to the presence of a significant HbCO level (over 4 percent) in the six patients in whom the difference between calculated and directly measured saturations was higher than 3 percent. Consequently transcutaneous saturation values were significantly different from measured and also from calculated ones not only at rest but also at peak exercise for each of the three devices (p
Patients*
Pierre]. L. Escourrou, M.D.; Marie Franfoise Delaperche, M.D.; and Agnes V&aseaux, B.S.
To evaluate the reliability of pulse oximetry during exercise, we studied 101 patients primarily with chronic pulmonary diseases. Three devices were used on different patients. Radial arterial blood was sampled at rest and maximal exercise simultaneously to pulse oximetric determination. Measured blood oxygen saturation was significantly different from noninvasive saturation at rest and also at exercise for each device. Nevertheless, changes in pulse oximetry from rest to exercise were significantly correlated with measured saturation for all three devices. Direction of
changes in saturation from rest to exercise was correctly evaluated by transcutaneous oximetry in all but six instances where changes were less than 4 percent. Although measured and transcutaneous saturations are signi6cantly different, we conclude that pulse oximetry reliably estimates changes in arterial saturation between rest and exercise for a clinical purpose. None of the three tested devices was better compared with the others in estimating saturation changes at exercise. (Chest 1990; 97:635-38)
Measurement of respiratory gas exchange during exercise stress testing is useful in the evaluation of patients with respiratory diseases. 1•2 Changes in blood oxygenation during exercise may be used to detect lung disease, evaluate the severity and follow the response to therapy. However, direct sampling of arterial blood by arterial cannulation is potentially dangerous.a-& Single arterial puncture at peak exercise, although less invasive, is not always successful, especially in patients with radial arteries of small diameter such as women and children. Rapid changes in arterial blood gas levels preclude precise evaluation of peak exercise value during recovery more so when the sample is obtained later than 20 s after the end of exercise. 7•8 Blood also may be sampled at the ear or the finger at exercise, but the reliability of capillarized blood has been questioned and sampling also may be difficult at exercise. 9 On the other hand, transcutaneous oximetry devices are more and more widely used in resting clinical conditions. One previous study evaluated two ear oximeters (HP 47201 A and Biox II A) at rest and exercise in pulmonary patients and showed that accuracy was comparable and clinically relevant for both devices. 10 Another report during combined hypoxia and exercise indicated that the accuracy of the pulse oximeter Biox II was lower than that of eight wavelength HP 47201 A. 11 Since these studies, several new
oximeters with improved technology have been marketed. We therefore decided to assess the reliability at exercise of three recent pulse oximeters with more efficient signal processing.
*From Explorations Fonctionnelles, Respiratoires, Hopital A. Beclere, Ciamart, France. Su~~ed in part by a grant from Comite National Contre les "es Respiratoires et Ia Tuberculose. M Manuscript received April18; revision accepted August 17. Reprint requests: Dr. Escourrou, Explorations Rmctionneles Resp, Hopital Beclere, 92141 Clamart, France
PATIENTS
One hundred and one patients have been studied whose ages ranged from 17 to 81 years old; 72 were men and 29 were women. Five patients (two men and three women) were moderately pigmented but not black. Their pathologic conditions were: COPD (33 cases), emphysema (five cases), sarcoidosis (15 cases), pulmonary fibrosis (eight cases), Hodgkin's disease (seven cases), primary pulmonary hypertension (six cases), cirrhosis of the liver (12 cases), restrictive pulmonary disease (two cases), miscellaneous diseases (seven cases), dyspnea of unknown origin (six cases). METHODS
Exercise Protocol Patients were tested on a bicycle ergometer (ergometry system 380B Siemens) by a symptom-limited incremental protocol of 20to 30-W in 2 to 3-min stages. A 12-lead ECG was recorded at rest and the end of each stage, and the heart rate was continuously displayed (custom-made ratemeter with a time constant of 3 s). Blood pressure was measured at rest and at the end of each stage by a sphygmomanometer. Average maximal load was 90±38 W (range, 30 to 210 W) for 9±3.6 min (4 to 18 min) exercise durations.
Blood Gas Measurement
Radial arterial blood was sampled using a microsampler (AVL Biomedical Instruments-Graz, Austria) at rest sitting on the bicycle and at maximal exercise just before stopping. Blood was immediately analyzed for pH, Po2 , and Pco2 using a Corning 178 gas analyzer (Corning Glass, Melfied, MA). Standard calibration procedures were used, ie, one-point gas calibration every 20 to 30 min and a two-point calibration every 2 h. The analyzer was further checked once a week with tonometered blood and the standard deviations for duplicate measurements were 2.1 mm Hg for Po, and 1.1 mm Hg for Pco2 • The calculated Sa0 2 was CHEST I 97 I 3 I MARCH. 1990
635
computed from the measured Po., Pco2 and pH, assuming normal P.., and a body temperature of37"C for these usually short exercise durations. Hemoglobin concentration and measured SaO. were determined by a spectrophotometric method (OSM., Radiometer, Copenhagen, Denmark). When a difference between measured and calculated SaO, larger than 3 percent was observed, carbon monoxide concentration was measured by the reductase method. 11 The percentage of HbCO was then subtracted from measured saturation which does not differentiate between Hb02 and HbCO. No correction was made for bilirubin as none of the cirrhotic patients had a severe hepatic insufficiency with bilirubin level higher than 20 p.moi!L.
Pulse Oximetry Three oximeters were studied: the Biox 3700 (Ohmeda version B), CSI 501 + (Criticare), and the Nellcor N200 (version 3.1). The latest device is provided with the option of artefact reduction by ECG synchronization. The three devices continuously display saturation and heart rate. All of them were used in the fast mode: saturation and heart rate averaging times are 3 and 5 s for Biox 3700, 2 and 3 s for the Nellcor N200 and every eight beats for the CSI 501 +, respectively. The instructions of the manufacturers were followed for each oximeter: the Biox 3700 and Criticare were connected to an earpiece. After calibration, the ear probes were attached to the earlobe previously massaged for 30 s with an alcohol saturated swab; the transmission cable was looped within a headband and taped to the subject's neck to minimize any movement artifact. Because no practical earpiece was available with the Nellcor N200, this device was used with the digital probe and the unit was linked to the ECG output of the electrocardiograph for artefact reduction. When a low signal output was indicated at rest by the device, the ear lobe or finger was massaged again with a revulsive paste Finalgon (Boehringer lngelheim). Notwithstanding these procedures, a satisfactory pulse signal output was not obtained at rest in five patients which were not included in the studied population. No relationship to a specific device or pigmentation was noted in these patients with apparently poor peripheral perfusion. The Biox 3700 was used in 48 patients, the Nellcor N200 in 24 patients and the Criticare in 29 patients.
Statistical Analysis Individual transcutaneous Sa02 values were compared with calculated and measured values on blood samples by paired t tests. Accuracy has been compared between devices by one-way analysis of variance on differences between invasive and transcutaneous saturations.
A
100
B
REST 0 EXERCISE X
90
100
Regression equations between variables have been computed by the least squares method using the SPSS-PC statistical package. Significance has been assessed at the O.OSlevel.
REsuLTS On the 101 patients studied at rest and exercise, 201 calculated Sa02 values, 195 measured Sa02 values and 198 transcutaneous values were obtained. In a few cases, indeed, the volume of sampled blood was insufficient to perform both spectrophotometry and gas pressure determinations. In three patients the transcutaneous value was not obtained at exercise; the device indicated a poor signal strength with erratic Sa02 values (two patients with the Biox 3700, one with the Nellcor N200). The measured range of Sa02 was 68 to 100 percent. Accuracy of Absolute Measurements of Transcutaneous Oximetry
Figure 1 gives the regressions of transcutaneous saturation on measured Sa02 at rest and exercise for each studied device. Although the relationships were statistically significant for all three devices, the scatter was large and the standard errors of the estimates were between 2 and 3 percent. Furthermore, the range of differences between measured and transcutaneous saturation values was -9 to + 16 percent. The mean difference was 1.1 ± 3 percent. These differences were not related to skin pigmentation nor to the presence of a significant HbCO level (over 4 percent) in the six patients in whom the difference between calculated and directly measured saturations was higher than 3 percent. Consequently transcutaneous saturation values were significantly different from measured and also from calculated ones not only at rest but also at peak exercise for each of the three devices (p
