Reliability of Noninvasive Oximetry in Black Subjects during Exercise and Hypoxia1-4
R. JORGE ZEBALLOS and IDELLE M. WEISMAN
Introduction
The last 20 yr have witnessed the increased use of ear oximetry for the noninvasive determination of arterial oxygen saturation in a wide spectrum of clinical situations (1-5). Although early studies demonstrated good validation of ear oximetry values with arterial O2 saturation measurements (6-11), no clear consensus has emerged about its reliability in clinical work. Indeed, several investigators have recently questioned the validity of noninvasiveoximetry as a reliable indicator of arterial oxygen saturation in studies that utilized similar instrument models during exposure to hypoxia (12), during exercise in pulmonary patients (13), and during exercise in cardiac patients (14, 15). Furthermore, technology does not appear to have provided convincing solutions to the question of the reliability of pulse oximetry determinations as similar concerns have also appeared in studies that have evaluated newer, more technologically advanced pulse oximeters (16-19).
The reliability of pulse oximetry measurements during exercise and hypoxia also continues to be controversial (10, 12). More recent studies (13, 19) haveattempted to define the role of pulse oximetry in clinical exercise testing. Of the many factors that can affect the reliability of noninvasive oximetry, the effect of skin pigmentation has not been thoroughly investigated (20). Until recently, few studies have specifically addressed the clinical application and limitations of oximetry in blacks. Earlier studies included only small numbers of blacks who, for the most part, were not darkly pigmented (6, 10); these studies provided conflicting conclusions. One study (6) suggested that the accuracy of ear oximetry wasnot affected by skin pigmentation, whereas another suggested that it was affected (10). No study, to our knowledge,has specifically evaluated the effect of skin pigmentation on the ac1240
SUMMARY The effect of skin pigmentation on the reliability of noninvasive OXimetry, especially during exercise and hypoxia, has not been thoroughly Investigated. This Is the first study, to our knowledge, that specifically addresses this question. Thirty-three young black men performed multistage, steady-state cycle ergometry, breathing gas mixtures simulating different altitudes: 33 breathed gas simulating sea level (Plo, = 146 mm Hg), 11 breathed gas simulating 2,300 m (Plo, 110 mm Hg), and 22 breathed gas simulating 4,000 m (Pto, 65 mm Hg). Co-oxlmeter Sao, determinations were performed In arterial blood samples obtained concurrently with ear oximetry that _s measured using H_lett·Packard 47201A (HP) and Blox IIA oxlmeters. The mean error or bias for the [HP - Sao,] and for [Blox IIA - Sao,] ± 95% CI _re: at simulated sea level (Sao, > 96%): -0.4 ± 0.3% and 2.1 ± 0.3%; at simulated 2,300 m (range of Sao, means, 89 to 94%): -0.8 ± 0.5% and 3.5 ± 0.9%; for simulated 4,000 m (range of Sao, means, 75 to 84%): -4.8 ± 1.6% and 9.8 ± 1.8%, respectively. A bettar coefficient correlation was observed for all the pairs between Sao, versus HP (r = 0.94, P < 0.001,n = 279) than for the Sao, versus Blox IIA (r = 0.80, P < 0.001,n = 242). In conclusion, the HP oximeter appears to estimate Sao, more accurataly than the Blox IIA oximeter. The previously described overestimation for the Blox IIA ear oxlmatar and the underestimation for the HP ear oximeter at low Sao, values In whites Is exaggeratad In blacks. Although noninvasive oximetry may be used to follow desaturatlon trends In blacks, It would be unreliable to estimate abeolute Sao,. The clinical utility of nonlnv8llva oximetry In blacks Is unacceptable at values of Sao, '" 85% for the HP and < 90% for the Blox IIA oxlmetere. The effect of skin pigmentation on the reliability of the many n_r pulse oxlmetars requires further Investigation.
=
=
AM REV RESPIR DIS 1991: 144:1240-1244
curacy of ear oximetry during combined exercise and hypoxia. Accordingly, the present study was undertaken to compare the reliability of two ear oximeters (Hewlett Packard 47201A and Biox IIA) in the determination of oxygen saturation compared with concurrent measurement of arterial O2 saturation in normal black subjects at rest and during exercise under conditions of normoxia and hypoxia. Methods Subjects The data presented in this study were obtained in 33 healthy, young, nonsmoking, black male Army recruits with normal hemoglobin (AA). All served as controls for a study on the effect of hypoxia in persons with the sickle cell trait. The majority of subjects were darkly pigmented; there was no classification concerning the degree of skin pigmentation. All volunteers gavewritten informed consent. The study was approved by our hospital's Human Use Review Board. Protocol The study was conducted in El Paso at an
altitude of 1,270 m and a mean barometric pressure of 656 mm Hg. The readings of two ear oximeters, an HP-47201A (Hewlett-Packard, Waltham, MA) and a Biox IIA (Ohmeda, Boulder, CO) were compared with measured Sao. at rest and during exerciseunder the following conditions: at simulated sea level in 33 volunteers, at simulated 2,300 m (S2300m) (Received in original form April 3, 1991 and in revised form June 17, 1991) I From the Department of Clinical Investigation, William Beaumont Army Medical Center and Texas 'Iech University Regional Academic Health Center, El Paso, Texas. 1 Supported by the U.S.Naval Medical Research and Development Command, Bethesda, and by the U.S. Army Medical Research and Development Command, Fort Detrick, Frederick, Maryland. 3 The opinions or assertions contained herein are the private views of the authors and are not to be construed as official or as reflecting the views of the Department of the Army or the Department of Defense. 4 Correspondence and requests for reprints should be addressed to Dr. R. Jorge Zeballos, Department of Clinical Investigation, Joint NavyArmy Human Performance/SCT Research Laboratory, William Beaumont Army Medical Center, El Paso, TX 79920-5001.
NONINVASIVE OXIMETRY IN BLACK SUBJECTS DURING EXERCISE ANO HYPOXIA
1241
TABLE 1 in 11, and at simulated 4,000 m (S4000m) in the other 22 subjects. SPECTROPHOTOMETRIC ARTERIAL OXIMETRY (IL 282 CO-OXIMETER), HP AND BIOX IIA EAR The experimental conditions were achieved OXIMETRY, AND vo, DURING STEADY·STATE EXERCISE (SSE) AT SIMULATED SEA LEVEL IN 33 SUBJECTS' by having the subjects breathe gas mixtures with different fractions of inspired oxygen Baseline SSE I SSE II SSE III Recovery (Flo,) with the balance N" through a respira96.9 ± 0.8 96.6 ± 0.9 96.9 ± 0.6 97.0 ± 0.8 96.6 ± 1.1 Sao,. % tory CPAP face mask (Vital Signs) connectCOHb, % 1 :I: 0.7 0.8:1: 0.7 0.6:1: 0.6 0.6:1: 0.6 0.5 ± 0.4 ed to a 120L reservoir bag for 2 h before and 0.6:1: 0.2 0.7:1: 0.3 0.7:1: 0.5 0.7 :I: 0.3 0.6:1: 0.2 during the exercise tests. Simulated sea level MetHb, % 98.4 ± 0.6 98.4:1: 0.9 97.7 ± 1.0t 98.2 :I: 0.7 97.9:1: 0.8* Funct. Sao" % was obtained using a gas mixture with Flo, 96.9:1: 1.8 96.6:1: 1.6 HP oximetry, °Al 96.3:1: 1.4* 95.8:1: 1.6* 96.5 :I: 1.3 = 24010, PI(), = 146 mm Hg, S2300m was Biox oximetry, 0Al 99.0:1: 1.4t 99.3:1: 1.2t 98.6:1: 2.0t 98.2:1: 2.5t 99.1 ± 1.1t obtained with Flo, = 18%, Plo, = 110 mm 0.03 ± 1.9 -0.3 ± 1.9 -0.3 ± 1.6 -0.7:1:1.4 -0.5:1: 1.4 [HP - Sao,), % Hg, and S4000m was obtained with Flo, = (:I: 0.7) (:I: 0.6) (:I: 0.5) (:I: 0.7) (:I: 0.6) 14%, PI(), = 85 mm Hg. 2.1 :I: 1.3 2.5:1: 1.6 2.1 :I: 2.4 1.7:1: 2.7 2.2 :I: 1.5 [Biox - Sao,]. % (± 0.6) Subjects exercised on an electronically (:I: 0.5) (:I: 0.9) (:I: 1.1) (:I: 0.7) vo, Umin 0.40:1: 0.10 2.08 ± 0.26 2.67 ± 0.32 3.05 ± 0.36 1.35 ± 0.53 braked cycle ergometer (KEM 3, Mijnhardt, Holland) using a progressive multistage Definition of abbreviations: COHb = carboxyhemoglobin; MetHb = methemoglobin; Funcl. Sao, = corrected arterial 0, satusteady-state protocol of 5 min for each stage. ration. The power for each stage was chosen based • Values are mean ± SO. Values in parentheses are 95% CI for the difference between the two means. t p < 0.05 from Sao,. on the maximal power achieved during an inp < 0.05 from baseline for Sao, and HP and Biox oximetry. cremental exercisetest performed on the previous day. During the exercise test, the subjects men studied was 19 ± 1.7yr. The results there was no significant difference bebreathed through a mouthpiece connected to of the spectrophotometric oximetry in ar- tween HP oximetry values and measured a two-way respiratory valve (Hans Rudolph, Kansas City, MO) with the inspiratory port terial blood, the HP and Biox IIA mea- Sao., whereas Biox IIA oximetry values surements, and VOl under all conditions, were significantly higher than measured connected to the gas reservoir. Oxygen consumption (Vo,) was measured using a breath- at simulated sea level, S2300m, and Sao. under all conditions except recovby-breath computerized system (System 2001; S4000m, respectively, are shown in tables ery. The mean error or bias ± SD for Medical Graphics Corp., St. Paul, MN). A 1 to 3. Ear oximetry measures function- [HP - Sao.] for all the data collected 20-gauge radial arterial catheter was inserted al arterial O. saturation, which express- at S2300m was -0.8 ± 1.7, with a 95% prior to the exercise test. Arterial blood sam- es O. saturation in relation to the avail- CI of ± 0.5, and for [Biox IIA - Sao.] ples were obtained at rest, during the last minable Hb. The spectrophotometric ox- it was 3.5 ± 2.8, with a 95% CI of ± 0.9. ute of each steady-state exercise, and 2 min At S4000m (table 3) HP oximetry after exercise. ABG samples were analyzed imeter measures fractional O. saturation, which is the ratio of O.Hb to total Hb. values were significantly lower than meain an automated blood gas analyzer (System sured Sao. under all conditions except 1303; Instrumentation Laboratories, Lexing- The values of functional and fractional ton, MA). Oxyhemoglobin (O,Hb), carboxy- . O. saturation are very close when COHb recovery, whereas Biox IIA oximetry hemoglobin (COHb), and methemoglobin + MetHb are less than 3070 (21,22). Un- values were significantly higher than (MetHb) were measured in a spectrophoto- der all the conditions studied, COHb + measured Sao. under all conditions. For metric oximeter (IL282 CO-Oximeter; In- MetHb were less than 2%. Because frac- the pooled data collected at S4000m, the strumentation Laboratories). All ABG sam- tional saturation corrected to functional mean error or bias ± SD for [HP ples were tonometrically corroborated. saturation was consistently only 10J0 highSao.I was -4.8 ± 6.9, with a 9511J'o CI Ear oximetry determinations were perof ± 1.6, and for [Biox IIA - Sao.] it of the analyses and comparisons er, all formed concurrently with arterial blood samwere done using absolute (uncorrected) was 9.8 ± 6.9, with a 95% CI of ± 1.8. pling. The earpiece of the HP oximeter was In figure 1,measured Sao. is compared values for Sao, and ear oximetry. securely placed on the ear using the plastic At simulated sea level (table 1), no with HP and Biox IIA oximetry values support piece and headband to hold the ear cover. The earclip of the Biox IIA was placed statistical differences were observed be- under all conditions for each of the simuon the other ear using the ear probe retainer. tween HP oximetry values and measured lated altitudes studied. These graphs Both ears were rubbed with isopropyl alcoSao. at rest, during exercise, and at recov- demonstrate that the Biox IIA signifihol prior to the placement of the ear probes. ery. Biox IIA oximetry values wereslight- cantlyoverestimated Sao. for all condiBoth ear oximeters were used in accordance ly but significantly higher than measured tions except recovery at S2300m. In conwith the manufacturer's instructions. In ortrast, HP oximetry values were similar Sao. under all conditions. The 95% CI der to minimize ambient light interference, the lights in the laboratory were turned off of the mean error or bias for [Biox IIA to measured Sao. at simulated sea level and at S2300m. However,at S4OOOm, HP - Sao.] was larger than for the [HP during testing. Sao.] under all conditions. When all the oximetry significantly underestimated The SPSS/PC + (Chicago, Illinois) statistidata collected at rest, during exercise, and Sao•. The magnitude of the differences cal software program was used for the statistirecovery at simulated sea level were between ear oximetry and measured Sao, cal analysis. Values are reported as mean ± SD and 95% confidence intervals (CI). 1\vopooled, the mean error or bias ± SD for became greater with simulated altitude. way analysis of variance with repeated meaAlthough Sao. did not change with ex[HP - Sao.] was -0.4 ± 1.7, with a sures was used to detect significant differ95% CI of ± 0.3, and for [Biox IIA ercise at simulated sea level (table 1), ences. When required, Student's t test was emSao.] it was 2.1 ± 2.0, with a 95% CI simulated altitude significantly potentiatployed. Relationships were determined using of ± 0.3. Exercise itself did not induce ed the reduction in Sao. with exercise(taPearson's correlation coefficient. Significance significant changes in measured Sao. or bles 2 and 3). In general, Biox IIA and was chosen at the 5% level (p < 0.05).
*
Results
The age (mean ± SD) of the 33 black
HP and Biox IIA oximetry values at simulated sea level. It can be seen in table 2 that at S2300m
HP ear oximetry followed the trend in changes in Sao. during exercise(figure 1). When all the data collectedat rest, dur-
1242
ZEBALLDS AND WEISMAN
o SoD,
TABLE 2 SPECTROPHOTOMETRIC ARTERIAL OXIMETRY (IL 282 CO-OXIMETER), HP AND BIOX IlA EAR OXIMETRY, AND VO, DURING STEADY-STATE EXERCISE (SSE) AT SIMULATED 2,300 METERS IN 11 SUBJECTS' Baseline
SSE I
SSE II
Recovery
£'. HP so,
0 aiox so,
·p
Fig. 1. Comparison of noninvasiveoximetry (measured with Hewlen-Packard and Biox IIA oximeters) with Sao, (Co-oximeter) at three simulated altitudes, at rest and during exercise and recovery. The results are presented as mean ± SEM. Asterisk indicates a significant difference from Sao, at p < 0.05. Open circles ; Sao,; open triangles ; HP So,; open squares ; BIOX So,.
75.4 ± 5.7t 0.8::1: 0.5 0.7::1: 0.2 76.4::1: 5.5:\: 68.5::1: 12.1t:1: 85.0 ± 6.3tt -7.1 ± 9.0 (::I: 4.6) 10.4 ± 6.8 (± 3.8) 2.54 ± 0.29
For definition of abbreviations, see table 1. • Values are mean :I: SO. Values in parentheses are 95% CI for the difference between the two means. t p < 0.05 from Sao,. p < 0.05 from baseline for Sao, and HP and Siox oximetry.
Discussion
90
z
0
79.2 ± 5.4t 0.7 ± 0.4 0.7::1: 0.2 80.2 ± 4.8t 77.3 ± 10.5 88.9 ± 6.7t -2.6 ± 8.2 (± 4.5) 10.1 ::I: 7.8 (± 4.6) 1.22 ± 0.45
77.8 ± 5.3t 0.8::1: 0.4 0.7 ± 0.2 79.0 ::I: 5.2; 71.8 ± 8.5t t 87.7 ± 6.3t t -6.0 ± 5.7 (± 2.6) 10.6 ± 7.3 (::I: 3.7) 2.06 ± 0.23
ing exercise, and recovery for all three simulated altitudes were pooled together (figure 2), good correlations were observed for HP versus Sao) (r = 0.94) and for Biox IIA versus Sao) (r = 0.80). The relationships for HP oximetry versus Sao, and Biox IIA oximetry versus Sao2 for each subject with more than four pairs of measurements are reported in table 4.
g
0
r-.;
70
1
65 Bcsellne
Exl
Ex II
Recovery
NONINVASIVE OXIMETRY IN BLACK SUBJECTS DURING EXERCISE AND HYPOXIA
1243
with values obtained in other studies in white subjects (11) and during exercise y = 0.69x + 29.01 (10, 12). At lower levelsof Sa0 2 « 85%), SEE = 2.74 90 the HP significantly underestimated Sao, whereas the Biox IIA continued to sig~ 80 nificantly overestimate Sa0 2 • A similar ~ pattern for two wavelength pulse oxieters 70 o has been reported by Smyth and colo VI 60 leagues (12) in white subjects, and more recently by Cahan and coworkers (23). 50 In the latter study (23), five newer 2 Fig. 2. Relationship of Hewlett-Packard wavelength pulse oximeters were comand Biox IIA oximetry determinations 40~-~~-~--~--~---:,:----:-: 90 100 50 60 70 80 with Sao. (Co-oximeter)for all the mea40 pared with the HP in six black and six surements performed at rest and durEAR OXIt.AETER SATURATION (r.J white subjects. The investigatorsconcluding exercise and recovery at simulated ed that pulse oximeters give higher values sea level, simulated 2,300m, and simuBIOX (POOLED DATA) than does the HP oximeter, a tendency lated 4,000 m. The solid line represents 100 the regression line. The dashed line that was more pronounced in blacks than y = 1.12x -15.34 represents the line of identity. in whites. On the basis of data from the 90 SEE = 4.99 present study and from that of Cahan and coworkers (23), it would appear that 80 the HP, which utilizes eight wavelengths ~ 70 and has, through internal calibration 0 0 techniques, been designed to account for VI 60 skin pigmentation (24) is more accurate in blacks than is pulse oximetry. The 50 differences in the results of Ries and coworkers (10) and those of the present 40 40 50 60 70 80 90 100 study may reflect the greater number of EAR OXIt.AETER SATURATION (r.) low Sa0 2 values « 85%) in the present study as well as the dark pigmentation of our subjects. The data from the present study in conditions (rest, exercise, recovery) at ment. The overestimation of Sao, by new- blacks would therefore suggest an exagsimulated sea level,S2300m,and S4OOOm, er pulse oximeter models has also been geration of those patterns of responses are pooled, a better correlation for the recently reported (19, 23). In turn, the un- for the Biox IIA and HP ear oximeters HP (r = 0.94)than for the Biox IIA (r = derestimation of Sao, by the HP oximeter . that has been previously reported in 0.80) with measured Sao, is again demon- during profound hypoxia in the present whites (12, 21). Caution would appear strated (figure 2). study has also been previously described necessary, therefore, when using noninThe overestimation of Sao, by the Biox when Sao, was less than 65% (21). vasive oximetry in black subjects. IIA has been previously described in In a study by Ries and coworkers (10), This message remains timely despite whites during hypoxia (Sao 2 < 75%) (8); skin pigmentation was noted to reduce the newer generation of pulse oximeters. the overestimation, furthermore, was not- the accuracy of ear oximetry measure- At low levels of O saturation, all of the 2 ed to be progressive as Sa0 2 fell (11). ments for the HP, but not for the Biox pulse oximeters appear to become inacSmyth and colleagues (12),in a study that IIA. Results from the present study, how- curate (16, 17, 23). Furthermore, accuracompared the HP 47201A with the Biox ever, would suggest that skin pigmenta- cy of the various oximeters differs sigIIA during combined hypoxia and exercise tion does affect the accuracy of the Biox nificantly (16, 17). Severinghaus and in whites, also found that below an Sa0 2 IIA but not the HP for Sa0 2 measure- Naifeh (16) recently examined the accuraof 83070, the Biox IIA overestimated true ments >85% (figure 1).The bias for [HP cy of six pulse oximeters in the presence Sao., whereas the HP estimations were - Sao 2 ] in the present study in blacks of profound hypoxia (Sa0 40 to 70%). within the error (~ 2070) of the instru- above 85% Sao, compares very favorably Bias (mean of differences) 2and precision (SD) differed widely between instruments. The same investigators also obTABLE 4 served that, in general, ear probes were SUMMARY OF THE INDIVIDUAL CORRELATIONS OF HP AND BIOX IIA more accurate than finger probes. In EAR OXIMETRY WITH Sao. FOR EACH VOLUNTEER WITH FOUR many instances, downward trends in oxOR MORE PAIRS OF MEASUREMENTS· imetry were not detected despite signifiBiox versus Sao. HP versus Sao. cant drops in measured Sa0 2 • A more recent evaluation of 14 pulse oximeters in (n subjects) (%) (n subjects) (%) r Range whites reported substantial differences in 15 52 0.90 to o.sst 29 91 bias and precision, induced by a step hyp17 2 6 5 0.80 to 0.89t oxic test, between pulse oximeters at low 3 1 0.70 to 0.79t 0 0 NS 1 3 8 28 Sa0 2 (17). Weagree with Ries and coworkers (10), • Mean = 8 pairs. t p < 0.05. and with the results in more recent studies HEWLETI-PACKARD (POOLED DATA)
N
N
1244
(13, 19),which have concluded that pulse oximetry cannot predict absolute values of Sao., In the latter study (19), which evaluated three newer pulse oximeters at rest and during exercise, it was suggested that despite technological advances, the accuracy of these latest devices has not been improved over that reported by Ries and coworkers (10) for the Biox IIA. Individual correlations between the HP ear oximeter and Sao, revealed a very good correlation (r = 0.90 to 0.99) in almost all subjects (970/0); in contrast, a significantly lesser number of subjects (720/0) demonstrated similar levels of correlation between the Biox IIA ear oximeter measurements and Sao, (table 4). Unfortunately, it is difficult to predict in which subjects oximetry will correlate with Sao z ' In conclusion, the HP 4nOlA appears to estimate Sao, more accurately than the Biox IIA in black subjects. The previously described overestimation for the Biox IIA and underestimation for the HP 47201A at low Sao, values in whites appears to be exaggerated in blacks. Although noninvasive oximetry may be helpful in establishing desaturation trends in black subjects, it would be unreliable to estimate absolute Sao, values from these measurements. The results of the present study, and those reported by other investigators utilizing newer pulse oximeters in whites (16, 17, 19) and blacks (23) warrant caution in the clinical use of pulse oximetry in black subjects during hypoxia and exercise. Our data suggest that the clinical utility of noninva-
ZEBALLOS AND WEISMAN
sive oximetry in black subjects is unacceptable at values of Sao, ~ 85 % for the HP 47201A and < 90% for Biox IIA. Additional studies are required to determine the effect of skin pigmentation on the reliability of the many newer pulse oximeters. Acknowledgment The writers wish to thank Drs. Bruce Johnson, Craig Smith, and Timothy Martin for assistance in exercisetesting during the various phases of this study; David Lopez for technical assistance; Dr. Gavin Gregory for statistical analysis; and Debbie Angerman for manuscript preparation.
References 1. Tobin MJ. Respiratory monitoring in the intensive care unit. Am Rev Respir Dis 1988; 138: 1625-42. 2. Misiano DR, Meyerhoff ME, Collison ME. Current and future directions in the technology relating to bedside testing of critically ill patients. Chest 1990; 97:204S-14S. 3. Schnapp LM, Cohen NH. Pulse oximetry, uses and abuses. Chest 1990; 98:1244-50. 4. Eichorn JH, Cooper JB, Cullen 01, Maier WR, Philip JH, Seeman RG. Standards for patient monitoring during anesthesia at the Harvard medical school. JAMA 1986; 256:1017-20. 5. Decker MJ, Hoekje PL, Strohl KP. Ambulatory monitoring of arterial oxygensaturation. Chest 1989; 95:717-22. 6. Saunders NA, Powles ACP, Rebuck AS. Ear oximetry: accuracy and practicability in the assessment of arterial oxygenation. Am Rev Respir Dis 1976; 113:745-9. 7. Poppius H, Viljanein AA. A new ear oximeter for assessment of exercise-induced arterial desaturation in patients with pulmonary disease. Scand J Respir Dis 1977; 58:279-83. 8. Chapman KR, D'Urzo A, Rebuck AS. The accuracy and response characteristics of a simplified
ear oximeter. Chest 1983; 83:860-4. 9. TweeddalePM, Douglas NJ. Evaluation of Biox IIA ear oximeter. Thorax 1985; 40:825-7. 10. Ries AL, Farrow JT, Clausen JL. Accuracy of two ear oximeters at rest and during exercise in pulmonary patients. Am Rev Respir Dis 1985; 1132:685-9. 11. Chapman KR, Liu FLW,Watson RM, Rebuck AS. Range of accuracy of two wavelength oximetry. Chest 1986; 89:540-2. 12. Smyth RJ, D'Urzo AD, SlutskyAS, Galko BM, Rebuck AS. Ear oximetry during combined hypoxia and exercise. J Appl Physiol 1986; 60:716-9. 13. HibbetAW, Ruppel GL. Sensitivity and specificity of pulse oximetry to detect oxygen desaturation during exercise (abstract). Am Rev Respir Dis 1990; 141:A225. 14. Hansen JE, Casaburi R. Validity of ear oximetry in clinical exercisetesting.Chest 1987;91:333-7. 15. Ries AL. Oximetry-Know thy limits. Chest 1987; 91:316. 16. Severinghaus JW, Naifeh KH. Accuracy of response of six pulse oximeters to profound hypoxia. Anesthesiology 1987; 67:551-8. 17. Severinghaus JW, Naifeh KH, Koh SO. Errors in 14 pulse oximeters during profound hypoxia. J Clin Monit 1989; 5:72-81. 18. Sykes MK. Pulse oximetry: A "Which" hunt? J Clin Monit 1989; 5:69-71. 19. Escourrou PJL, Delaperche MF, Visseaux A. Reliability of pulse oximetry during exercisein pulmonary patients. Chest 1990; 97:635-8. 20. Tobin MJ. Respiratory monitoring. JAMA 1990; 264:244-51. 21. Douglas NJ, Brash HM, Wraith PK, et at. Accuracy,sensitivityto carboxyhemoglobin, and speed of response of the Hewlett-Packard 4720lA ear oximeter. Am Rev Respir Dis 1979; 119:311-3. 22. Tremper KK. Pulse oximetry. Chest 1989; 95:713-5. 23. Cahan C, Decker MJ, Hoekje PL, Strohl KP. Agreement between noninvasive oximetric values for oxygen saturation. Chest 1990; 97:814-9. 24. Merrick EB, Hayes TJ. Continuous, noninvasivemeasurements of arterial blood oxygenlevels. Hewlett-Packard J 1976; 2-9.
R. JORGE ZEBALLOS and IDELLE M. WEISMAN
Introduction
The last 20 yr have witnessed the increased use of ear oximetry for the noninvasive determination of arterial oxygen saturation in a wide spectrum of clinical situations (1-5). Although early studies demonstrated good validation of ear oximetry values with arterial O2 saturation measurements (6-11), no clear consensus has emerged about its reliability in clinical work. Indeed, several investigators have recently questioned the validity of noninvasiveoximetry as a reliable indicator of arterial oxygen saturation in studies that utilized similar instrument models during exposure to hypoxia (12), during exercise in pulmonary patients (13), and during exercise in cardiac patients (14, 15). Furthermore, technology does not appear to have provided convincing solutions to the question of the reliability of pulse oximetry determinations as similar concerns have also appeared in studies that have evaluated newer, more technologically advanced pulse oximeters (16-19).
The reliability of pulse oximetry measurements during exercise and hypoxia also continues to be controversial (10, 12). More recent studies (13, 19) haveattempted to define the role of pulse oximetry in clinical exercise testing. Of the many factors that can affect the reliability of noninvasive oximetry, the effect of skin pigmentation has not been thoroughly investigated (20). Until recently, few studies have specifically addressed the clinical application and limitations of oximetry in blacks. Earlier studies included only small numbers of blacks who, for the most part, were not darkly pigmented (6, 10); these studies provided conflicting conclusions. One study (6) suggested that the accuracy of ear oximetry wasnot affected by skin pigmentation, whereas another suggested that it was affected (10). No study, to our knowledge,has specifically evaluated the effect of skin pigmentation on the ac1240
SUMMARY The effect of skin pigmentation on the reliability of noninvasive OXimetry, especially during exercise and hypoxia, has not been thoroughly Investigated. This Is the first study, to our knowledge, that specifically addresses this question. Thirty-three young black men performed multistage, steady-state cycle ergometry, breathing gas mixtures simulating different altitudes: 33 breathed gas simulating sea level (Plo, = 146 mm Hg), 11 breathed gas simulating 2,300 m (Plo, 110 mm Hg), and 22 breathed gas simulating 4,000 m (Pto, 65 mm Hg). Co-oxlmeter Sao, determinations were performed In arterial blood samples obtained concurrently with ear oximetry that _s measured using H_lett·Packard 47201A (HP) and Blox IIA oxlmeters. The mean error or bias for the [HP - Sao,] and for [Blox IIA - Sao,] ± 95% CI _re: at simulated sea level (Sao, > 96%): -0.4 ± 0.3% and 2.1 ± 0.3%; at simulated 2,300 m (range of Sao, means, 89 to 94%): -0.8 ± 0.5% and 3.5 ± 0.9%; for simulated 4,000 m (range of Sao, means, 75 to 84%): -4.8 ± 1.6% and 9.8 ± 1.8%, respectively. A bettar coefficient correlation was observed for all the pairs between Sao, versus HP (r = 0.94, P < 0.001,n = 279) than for the Sao, versus Blox IIA (r = 0.80, P < 0.001,n = 242). In conclusion, the HP oximeter appears to estimate Sao, more accurataly than the Blox IIA oximeter. The previously described overestimation for the Blox IIA ear oxlmatar and the underestimation for the HP ear oximeter at low Sao, values In whites Is exaggeratad In blacks. Although noninvasive oximetry may be used to follow desaturatlon trends In blacks, It would be unreliable to estimate abeolute Sao,. The clinical utility of nonlnv8llva oximetry In blacks Is unacceptable at values of Sao, '" 85% for the HP and < 90% for the Blox IIA oxlmetere. The effect of skin pigmentation on the reliability of the many n_r pulse oxlmetars requires further Investigation.
=
=
AM REV RESPIR DIS 1991: 144:1240-1244
curacy of ear oximetry during combined exercise and hypoxia. Accordingly, the present study was undertaken to compare the reliability of two ear oximeters (Hewlett Packard 47201A and Biox IIA) in the determination of oxygen saturation compared with concurrent measurement of arterial O2 saturation in normal black subjects at rest and during exercise under conditions of normoxia and hypoxia. Methods Subjects The data presented in this study were obtained in 33 healthy, young, nonsmoking, black male Army recruits with normal hemoglobin (AA). All served as controls for a study on the effect of hypoxia in persons with the sickle cell trait. The majority of subjects were darkly pigmented; there was no classification concerning the degree of skin pigmentation. All volunteers gavewritten informed consent. The study was approved by our hospital's Human Use Review Board. Protocol The study was conducted in El Paso at an
altitude of 1,270 m and a mean barometric pressure of 656 mm Hg. The readings of two ear oximeters, an HP-47201A (Hewlett-Packard, Waltham, MA) and a Biox IIA (Ohmeda, Boulder, CO) were compared with measured Sao. at rest and during exerciseunder the following conditions: at simulated sea level in 33 volunteers, at simulated 2,300 m (S2300m) (Received in original form April 3, 1991 and in revised form June 17, 1991) I From the Department of Clinical Investigation, William Beaumont Army Medical Center and Texas 'Iech University Regional Academic Health Center, El Paso, Texas. 1 Supported by the U.S.Naval Medical Research and Development Command, Bethesda, and by the U.S. Army Medical Research and Development Command, Fort Detrick, Frederick, Maryland. 3 The opinions or assertions contained herein are the private views of the authors and are not to be construed as official or as reflecting the views of the Department of the Army or the Department of Defense. 4 Correspondence and requests for reprints should be addressed to Dr. R. Jorge Zeballos, Department of Clinical Investigation, Joint NavyArmy Human Performance/SCT Research Laboratory, William Beaumont Army Medical Center, El Paso, TX 79920-5001.
NONINVASIVE OXIMETRY IN BLACK SUBJECTS DURING EXERCISE ANO HYPOXIA
1241
TABLE 1 in 11, and at simulated 4,000 m (S4000m) in the other 22 subjects. SPECTROPHOTOMETRIC ARTERIAL OXIMETRY (IL 282 CO-OXIMETER), HP AND BIOX IIA EAR The experimental conditions were achieved OXIMETRY, AND vo, DURING STEADY·STATE EXERCISE (SSE) AT SIMULATED SEA LEVEL IN 33 SUBJECTS' by having the subjects breathe gas mixtures with different fractions of inspired oxygen Baseline SSE I SSE II SSE III Recovery (Flo,) with the balance N" through a respira96.9 ± 0.8 96.6 ± 0.9 96.9 ± 0.6 97.0 ± 0.8 96.6 ± 1.1 Sao,. % tory CPAP face mask (Vital Signs) connectCOHb, % 1 :I: 0.7 0.8:1: 0.7 0.6:1: 0.6 0.6:1: 0.6 0.5 ± 0.4 ed to a 120L reservoir bag for 2 h before and 0.6:1: 0.2 0.7:1: 0.3 0.7:1: 0.5 0.7 :I: 0.3 0.6:1: 0.2 during the exercise tests. Simulated sea level MetHb, % 98.4 ± 0.6 98.4:1: 0.9 97.7 ± 1.0t 98.2 :I: 0.7 97.9:1: 0.8* Funct. Sao" % was obtained using a gas mixture with Flo, 96.9:1: 1.8 96.6:1: 1.6 HP oximetry, °Al 96.3:1: 1.4* 95.8:1: 1.6* 96.5 :I: 1.3 = 24010, PI(), = 146 mm Hg, S2300m was Biox oximetry, 0Al 99.0:1: 1.4t 99.3:1: 1.2t 98.6:1: 2.0t 98.2:1: 2.5t 99.1 ± 1.1t obtained with Flo, = 18%, Plo, = 110 mm 0.03 ± 1.9 -0.3 ± 1.9 -0.3 ± 1.6 -0.7:1:1.4 -0.5:1: 1.4 [HP - Sao,), % Hg, and S4000m was obtained with Flo, = (:I: 0.7) (:I: 0.6) (:I: 0.5) (:I: 0.7) (:I: 0.6) 14%, PI(), = 85 mm Hg. 2.1 :I: 1.3 2.5:1: 1.6 2.1 :I: 2.4 1.7:1: 2.7 2.2 :I: 1.5 [Biox - Sao,]. % (± 0.6) Subjects exercised on an electronically (:I: 0.5) (:I: 0.9) (:I: 1.1) (:I: 0.7) vo, Umin 0.40:1: 0.10 2.08 ± 0.26 2.67 ± 0.32 3.05 ± 0.36 1.35 ± 0.53 braked cycle ergometer (KEM 3, Mijnhardt, Holland) using a progressive multistage Definition of abbreviations: COHb = carboxyhemoglobin; MetHb = methemoglobin; Funcl. Sao, = corrected arterial 0, satusteady-state protocol of 5 min for each stage. ration. The power for each stage was chosen based • Values are mean ± SO. Values in parentheses are 95% CI for the difference between the two means. t p < 0.05 from Sao,. on the maximal power achieved during an inp < 0.05 from baseline for Sao, and HP and Biox oximetry. cremental exercisetest performed on the previous day. During the exercise test, the subjects men studied was 19 ± 1.7yr. The results there was no significant difference bebreathed through a mouthpiece connected to of the spectrophotometric oximetry in ar- tween HP oximetry values and measured a two-way respiratory valve (Hans Rudolph, Kansas City, MO) with the inspiratory port terial blood, the HP and Biox IIA mea- Sao., whereas Biox IIA oximetry values surements, and VOl under all conditions, were significantly higher than measured connected to the gas reservoir. Oxygen consumption (Vo,) was measured using a breath- at simulated sea level, S2300m, and Sao. under all conditions except recovby-breath computerized system (System 2001; S4000m, respectively, are shown in tables ery. The mean error or bias ± SD for Medical Graphics Corp., St. Paul, MN). A 1 to 3. Ear oximetry measures function- [HP - Sao.] for all the data collected 20-gauge radial arterial catheter was inserted al arterial O. saturation, which express- at S2300m was -0.8 ± 1.7, with a 95% prior to the exercise test. Arterial blood sam- es O. saturation in relation to the avail- CI of ± 0.5, and for [Biox IIA - Sao.] ples were obtained at rest, during the last minable Hb. The spectrophotometric ox- it was 3.5 ± 2.8, with a 95% CI of ± 0.9. ute of each steady-state exercise, and 2 min At S4000m (table 3) HP oximetry after exercise. ABG samples were analyzed imeter measures fractional O. saturation, which is the ratio of O.Hb to total Hb. values were significantly lower than meain an automated blood gas analyzer (System sured Sao. under all conditions except 1303; Instrumentation Laboratories, Lexing- The values of functional and fractional ton, MA). Oxyhemoglobin (O,Hb), carboxy- . O. saturation are very close when COHb recovery, whereas Biox IIA oximetry hemoglobin (COHb), and methemoglobin + MetHb are less than 3070 (21,22). Un- values were significantly higher than (MetHb) were measured in a spectrophoto- der all the conditions studied, COHb + measured Sao. under all conditions. For metric oximeter (IL282 CO-Oximeter; In- MetHb were less than 2%. Because frac- the pooled data collected at S4000m, the strumentation Laboratories). All ABG sam- tional saturation corrected to functional mean error or bias ± SD for [HP ples were tonometrically corroborated. saturation was consistently only 10J0 highSao.I was -4.8 ± 6.9, with a 9511J'o CI Ear oximetry determinations were perof ± 1.6, and for [Biox IIA - Sao.] it of the analyses and comparisons er, all formed concurrently with arterial blood samwere done using absolute (uncorrected) was 9.8 ± 6.9, with a 95% CI of ± 1.8. pling. The earpiece of the HP oximeter was In figure 1,measured Sao. is compared values for Sao, and ear oximetry. securely placed on the ear using the plastic At simulated sea level (table 1), no with HP and Biox IIA oximetry values support piece and headband to hold the ear cover. The earclip of the Biox IIA was placed statistical differences were observed be- under all conditions for each of the simuon the other ear using the ear probe retainer. tween HP oximetry values and measured lated altitudes studied. These graphs Both ears were rubbed with isopropyl alcoSao. at rest, during exercise, and at recov- demonstrate that the Biox IIA signifihol prior to the placement of the ear probes. ery. Biox IIA oximetry values wereslight- cantlyoverestimated Sao. for all condiBoth ear oximeters were used in accordance ly but significantly higher than measured tions except recovery at S2300m. In conwith the manufacturer's instructions. In ortrast, HP oximetry values were similar Sao. under all conditions. The 95% CI der to minimize ambient light interference, the lights in the laboratory were turned off of the mean error or bias for [Biox IIA to measured Sao. at simulated sea level and at S2300m. However,at S4OOOm, HP - Sao.] was larger than for the [HP during testing. Sao.] under all conditions. When all the oximetry significantly underestimated The SPSS/PC + (Chicago, Illinois) statistidata collected at rest, during exercise, and Sao•. The magnitude of the differences cal software program was used for the statistirecovery at simulated sea level were between ear oximetry and measured Sao, cal analysis. Values are reported as mean ± SD and 95% confidence intervals (CI). 1\vopooled, the mean error or bias ± SD for became greater with simulated altitude. way analysis of variance with repeated meaAlthough Sao. did not change with ex[HP - Sao.] was -0.4 ± 1.7, with a sures was used to detect significant differ95% CI of ± 0.3, and for [Biox IIA ercise at simulated sea level (table 1), ences. When required, Student's t test was emSao.] it was 2.1 ± 2.0, with a 95% CI simulated altitude significantly potentiatployed. Relationships were determined using of ± 0.3. Exercise itself did not induce ed the reduction in Sao. with exercise(taPearson's correlation coefficient. Significance significant changes in measured Sao. or bles 2 and 3). In general, Biox IIA and was chosen at the 5% level (p < 0.05).
*
Results
The age (mean ± SD) of the 33 black
HP and Biox IIA oximetry values at simulated sea level. It can be seen in table 2 that at S2300m
HP ear oximetry followed the trend in changes in Sao. during exercise(figure 1). When all the data collectedat rest, dur-
1242
ZEBALLDS AND WEISMAN
o SoD,
TABLE 2 SPECTROPHOTOMETRIC ARTERIAL OXIMETRY (IL 282 CO-OXIMETER), HP AND BIOX IlA EAR OXIMETRY, AND VO, DURING STEADY-STATE EXERCISE (SSE) AT SIMULATED 2,300 METERS IN 11 SUBJECTS' Baseline
SSE I
SSE II
Recovery
£'. HP so,
0 aiox so,
·p
Fig. 1. Comparison of noninvasiveoximetry (measured with Hewlen-Packard and Biox IIA oximeters) with Sao, (Co-oximeter) at three simulated altitudes, at rest and during exercise and recovery. The results are presented as mean ± SEM. Asterisk indicates a significant difference from Sao, at p < 0.05. Open circles ; Sao,; open triangles ; HP So,; open squares ; BIOX So,.
75.4 ± 5.7t 0.8::1: 0.5 0.7::1: 0.2 76.4::1: 5.5:\: 68.5::1: 12.1t:1: 85.0 ± 6.3tt -7.1 ± 9.0 (::I: 4.6) 10.4 ± 6.8 (± 3.8) 2.54 ± 0.29
For definition of abbreviations, see table 1. • Values are mean :I: SO. Values in parentheses are 95% CI for the difference between the two means. t p < 0.05 from Sao,. p < 0.05 from baseline for Sao, and HP and Siox oximetry.
Discussion
90
z
0
79.2 ± 5.4t 0.7 ± 0.4 0.7::1: 0.2 80.2 ± 4.8t 77.3 ± 10.5 88.9 ± 6.7t -2.6 ± 8.2 (± 4.5) 10.1 ::I: 7.8 (± 4.6) 1.22 ± 0.45
77.8 ± 5.3t 0.8::1: 0.4 0.7 ± 0.2 79.0 ::I: 5.2; 71.8 ± 8.5t t 87.7 ± 6.3t t -6.0 ± 5.7 (± 2.6) 10.6 ± 7.3 (::I: 3.7) 2.06 ± 0.23
ing exercise, and recovery for all three simulated altitudes were pooled together (figure 2), good correlations were observed for HP versus Sao) (r = 0.94) and for Biox IIA versus Sao) (r = 0.80). The relationships for HP oximetry versus Sao, and Biox IIA oximetry versus Sao2 for each subject with more than four pairs of measurements are reported in table 4.
g
0
r-.;
70
1
65 Bcsellne
Exl
Ex II
Recovery
NONINVASIVE OXIMETRY IN BLACK SUBJECTS DURING EXERCISE AND HYPOXIA
1243
with values obtained in other studies in white subjects (11) and during exercise y = 0.69x + 29.01 (10, 12). At lower levelsof Sa0 2 « 85%), SEE = 2.74 90 the HP significantly underestimated Sao, whereas the Biox IIA continued to sig~ 80 nificantly overestimate Sa0 2 • A similar ~ pattern for two wavelength pulse oxieters 70 o has been reported by Smyth and colo VI 60 leagues (12) in white subjects, and more recently by Cahan and coworkers (23). 50 In the latter study (23), five newer 2 Fig. 2. Relationship of Hewlett-Packard wavelength pulse oximeters were comand Biox IIA oximetry determinations 40~-~~-~--~--~---:,:----:-: 90 100 50 60 70 80 with Sao. (Co-oximeter)for all the mea40 pared with the HP in six black and six surements performed at rest and durEAR OXIt.AETER SATURATION (r.J white subjects. The investigatorsconcluding exercise and recovery at simulated ed that pulse oximeters give higher values sea level, simulated 2,300m, and simuBIOX (POOLED DATA) than does the HP oximeter, a tendency lated 4,000 m. The solid line represents 100 the regression line. The dashed line that was more pronounced in blacks than y = 1.12x -15.34 represents the line of identity. in whites. On the basis of data from the 90 SEE = 4.99 present study and from that of Cahan and coworkers (23), it would appear that 80 the HP, which utilizes eight wavelengths ~ 70 and has, through internal calibration 0 0 techniques, been designed to account for VI 60 skin pigmentation (24) is more accurate in blacks than is pulse oximetry. The 50 differences in the results of Ries and coworkers (10) and those of the present 40 40 50 60 70 80 90 100 study may reflect the greater number of EAR OXIt.AETER SATURATION (r.) low Sa0 2 values « 85%) in the present study as well as the dark pigmentation of our subjects. The data from the present study in conditions (rest, exercise, recovery) at ment. The overestimation of Sao, by new- blacks would therefore suggest an exagsimulated sea level,S2300m,and S4OOOm, er pulse oximeter models has also been geration of those patterns of responses are pooled, a better correlation for the recently reported (19, 23). In turn, the un- for the Biox IIA and HP ear oximeters HP (r = 0.94)than for the Biox IIA (r = derestimation of Sao, by the HP oximeter . that has been previously reported in 0.80) with measured Sao, is again demon- during profound hypoxia in the present whites (12, 21). Caution would appear strated (figure 2). study has also been previously described necessary, therefore, when using noninThe overestimation of Sao, by the Biox when Sao, was less than 65% (21). vasive oximetry in black subjects. IIA has been previously described in In a study by Ries and coworkers (10), This message remains timely despite whites during hypoxia (Sao 2 < 75%) (8); skin pigmentation was noted to reduce the newer generation of pulse oximeters. the overestimation, furthermore, was not- the accuracy of ear oximetry measure- At low levels of O saturation, all of the 2 ed to be progressive as Sa0 2 fell (11). ments for the HP, but not for the Biox pulse oximeters appear to become inacSmyth and colleagues (12),in a study that IIA. Results from the present study, how- curate (16, 17, 23). Furthermore, accuracompared the HP 47201A with the Biox ever, would suggest that skin pigmenta- cy of the various oximeters differs sigIIA during combined hypoxia and exercise tion does affect the accuracy of the Biox nificantly (16, 17). Severinghaus and in whites, also found that below an Sa0 2 IIA but not the HP for Sa0 2 measure- Naifeh (16) recently examined the accuraof 83070, the Biox IIA overestimated true ments >85% (figure 1).The bias for [HP cy of six pulse oximeters in the presence Sao., whereas the HP estimations were - Sao 2 ] in the present study in blacks of profound hypoxia (Sa0 40 to 70%). within the error (~ 2070) of the instru- above 85% Sao, compares very favorably Bias (mean of differences) 2and precision (SD) differed widely between instruments. The same investigators also obTABLE 4 served that, in general, ear probes were SUMMARY OF THE INDIVIDUAL CORRELATIONS OF HP AND BIOX IIA more accurate than finger probes. In EAR OXIMETRY WITH Sao. FOR EACH VOLUNTEER WITH FOUR many instances, downward trends in oxOR MORE PAIRS OF MEASUREMENTS· imetry were not detected despite signifiBiox versus Sao. HP versus Sao. cant drops in measured Sa0 2 • A more recent evaluation of 14 pulse oximeters in (n subjects) (%) (n subjects) (%) r Range whites reported substantial differences in 15 52 0.90 to o.sst 29 91 bias and precision, induced by a step hyp17 2 6 5 0.80 to 0.89t oxic test, between pulse oximeters at low 3 1 0.70 to 0.79t 0 0 NS 1 3 8 28 Sa0 2 (17). Weagree with Ries and coworkers (10), • Mean = 8 pairs. t p < 0.05. and with the results in more recent studies HEWLETI-PACKARD (POOLED DATA)
N
N
1244
(13, 19),which have concluded that pulse oximetry cannot predict absolute values of Sao., In the latter study (19), which evaluated three newer pulse oximeters at rest and during exercise, it was suggested that despite technological advances, the accuracy of these latest devices has not been improved over that reported by Ries and coworkers (10) for the Biox IIA. Individual correlations between the HP ear oximeter and Sao, revealed a very good correlation (r = 0.90 to 0.99) in almost all subjects (970/0); in contrast, a significantly lesser number of subjects (720/0) demonstrated similar levels of correlation between the Biox IIA ear oximeter measurements and Sao, (table 4). Unfortunately, it is difficult to predict in which subjects oximetry will correlate with Sao z ' In conclusion, the HP 4nOlA appears to estimate Sao, more accurately than the Biox IIA in black subjects. The previously described overestimation for the Biox IIA and underestimation for the HP 47201A at low Sao, values in whites appears to be exaggerated in blacks. Although noninvasive oximetry may be helpful in establishing desaturation trends in black subjects, it would be unreliable to estimate absolute Sao, values from these measurements. The results of the present study, and those reported by other investigators utilizing newer pulse oximeters in whites (16, 17, 19) and blacks (23) warrant caution in the clinical use of pulse oximetry in black subjects during hypoxia and exercise. Our data suggest that the clinical utility of noninva-
ZEBALLOS AND WEISMAN
sive oximetry in black subjects is unacceptable at values of Sao, ~ 85 % for the HP 47201A and < 90% for Biox IIA. Additional studies are required to determine the effect of skin pigmentation on the reliability of the many newer pulse oximeters. Acknowledgment The writers wish to thank Drs. Bruce Johnson, Craig Smith, and Timothy Martin for assistance in exercisetesting during the various phases of this study; David Lopez for technical assistance; Dr. Gavin Gregory for statistical analysis; and Debbie Angerman for manuscript preparation.
References 1. Tobin MJ. Respiratory monitoring in the intensive care unit. Am Rev Respir Dis 1988; 138: 1625-42. 2. Misiano DR, Meyerhoff ME, Collison ME. Current and future directions in the technology relating to bedside testing of critically ill patients. Chest 1990; 97:204S-14S. 3. Schnapp LM, Cohen NH. Pulse oximetry, uses and abuses. Chest 1990; 98:1244-50. 4. Eichorn JH, Cooper JB, Cullen 01, Maier WR, Philip JH, Seeman RG. Standards for patient monitoring during anesthesia at the Harvard medical school. JAMA 1986; 256:1017-20. 5. Decker MJ, Hoekje PL, Strohl KP. Ambulatory monitoring of arterial oxygensaturation. Chest 1989; 95:717-22. 6. Saunders NA, Powles ACP, Rebuck AS. Ear oximetry: accuracy and practicability in the assessment of arterial oxygenation. Am Rev Respir Dis 1976; 113:745-9. 7. Poppius H, Viljanein AA. A new ear oximeter for assessment of exercise-induced arterial desaturation in patients with pulmonary disease. Scand J Respir Dis 1977; 58:279-83. 8. Chapman KR, D'Urzo A, Rebuck AS. The accuracy and response characteristics of a simplified
ear oximeter. Chest 1983; 83:860-4. 9. TweeddalePM, Douglas NJ. Evaluation of Biox IIA ear oximeter. Thorax 1985; 40:825-7. 10. Ries AL, Farrow JT, Clausen JL. Accuracy of two ear oximeters at rest and during exercise in pulmonary patients. Am Rev Respir Dis 1985; 1132:685-9. 11. Chapman KR, Liu FLW,Watson RM, Rebuck AS. Range of accuracy of two wavelength oximetry. Chest 1986; 89:540-2. 12. Smyth RJ, D'Urzo AD, SlutskyAS, Galko BM, Rebuck AS. Ear oximetry during combined hypoxia and exercise. J Appl Physiol 1986; 60:716-9. 13. HibbetAW, Ruppel GL. Sensitivity and specificity of pulse oximetry to detect oxygen desaturation during exercise (abstract). Am Rev Respir Dis 1990; 141:A225. 14. Hansen JE, Casaburi R. Validity of ear oximetry in clinical exercisetesting.Chest 1987;91:333-7. 15. Ries AL. Oximetry-Know thy limits. Chest 1987; 91:316. 16. Severinghaus JW, Naifeh KH. Accuracy of response of six pulse oximeters to profound hypoxia. Anesthesiology 1987; 67:551-8. 17. Severinghaus JW, Naifeh KH, Koh SO. Errors in 14 pulse oximeters during profound hypoxia. J Clin Monit 1989; 5:72-81. 18. Sykes MK. Pulse oximetry: A "Which" hunt? J Clin Monit 1989; 5:69-71. 19. Escourrou PJL, Delaperche MF, Visseaux A. Reliability of pulse oximetry during exercisein pulmonary patients. Chest 1990; 97:635-8. 20. Tobin MJ. Respiratory monitoring. JAMA 1990; 264:244-51. 21. Douglas NJ, Brash HM, Wraith PK, et at. Accuracy,sensitivityto carboxyhemoglobin, and speed of response of the Hewlett-Packard 4720lA ear oximeter. Am Rev Respir Dis 1979; 119:311-3. 22. Tremper KK. Pulse oximetry. Chest 1989; 95:713-5. 23. Cahan C, Decker MJ, Hoekje PL, Strohl KP. Agreement between noninvasive oximetric values for oxygen saturation. Chest 1990; 97:814-9. 24. Merrick EB, Hayes TJ. Continuous, noninvasivemeasurements of arterial blood oxygenlevels. Hewlett-Packard J 1976; 2-9.
