Validity of pulse oximetry during exercise in elite endurance athletes

Validity of pulse oximetry during exercise in elite endurance athletes DANIEL MARTIN, SCOTT POWERS, MICHAEL CICALE, NANCY COLLOP, DAVID HUANG, AND DAVE CRISWELL Departments of Physical Therapy, Exercise and Sport Sciences, and Physiology, Center for Exercise Science, and Division of Pulmonary Medicine, University of Florida, Gainesville, Florida 3261 O-01 54 MARTIN,DANIEL,SC~ POWERS,MICHAEL CICALE,NANY jects. Therefore there is a clear need for pulse oximeter DAVID HUANG,ANDDAVECRISWELL.Validityofpulse validation data during high-intensity exercise in enduroximetry during exercise in elite endurance athletes. J. Appl. ance athletes. The purpose of this study was to evaluate Physiol. 72(2): 455-458,1992.-Eleven highly trained male cy- the accuracy of pulse oximetry to estimate %HbO, in clists [maximal aerobic power (VO, ,& = 70.6 t 4.2 highly trained endurance athletes during exercise. ml kg-’ ..min-‘1 performed both high intensity constant load (go-95% Vo, -) and incremental cycle exercisetests with arterial blood samplingto evaluate the accuracy of pulse oximeter METHODS estimates (%Sp, ) of arterial oxyhemoglobin fraction of total Subjects were recruited from the university commuhemoglobin (%HbO,). Three subjectsalsoperformed an incremental exercise test in hypoxic conditions (inspired partial nity and informed consent was obtained according to InReview Board guidelines. All subjects (n = 11) pressureof O2= 89,93, or 100 Torr). Arterial %HbOz was de- stitutional termined via CO-oximetry and ranged from 72 to 99%. Three were Caucasian nonsmoking highly trained male competOhmeda 3740pulse oximeters were used to estimate %HbOz, itive cyclists. one on each ear lobe and a finger probe. The finger probe Subjects initially performed pulmonary spirometry tended to provide the best estimate of %HbO, during exercise: and CO diffusion capacity tests [forced vital capacity the mean %Sp,, - %HbOz difference for 232 exercise observa- (FVC), forced expired volume in 1 s, forced expiratory tions was 0.52 t 1.36% (SD). Finger probe %Spo, and %HbO, flow at 25-75% of FVC, maximum voluntary ventilation, were highly correlated [r = 0.98, standard error of the estimate and single-breath pulmonary diffusing capacity for CO] (SEE) = 1.32%,P < O.OOOl].The accuracy of pulse oximeters and maximal cycle ergometer tests (Monark 818) to enhas beenquestionedduring high-intensity exercise.When aerobic power was ~81% of 00, - (n = 75), the finger probe’s sure that they had normal pulmonary function (>80% of mean error was -0.01 t 1.40%. Finger probe %Spo, and the expected value) and that their \jolmar was supe%Hb02 were highly correlated (r = 0.97, SEE = 1.32%, P < rior (>65 ml. kg-l min-l). The standards of Morris et al. 0.0001). These results indicate that this pulse oximeter is a (7) and Miller et al. (6) were used to evaluate spirometry valid predictor of %HbOz in elite athletes during cycle exercise. and diffusion tests, respectively. Approximately 1 wk COLLOP,

l

l

maximal aerobic power; incremental exercise; hypoxia; oxyhemoglobin

THERE IS growing interest

(5,9-11,14) in the use of pulse oximeters to estimate the arterial oxyhemoglobin fraction of total hemoglobin (%HbO,) during exercise because of the inconvenience and hazards associated with indwelling arterial catheters in exercising humans. Several investigators have shown that pulse oximeter estimates (%SpoJ are reasonable predictors of %Hb02 in patients during exercise (1, 2, 12), whereas others have questioned the validity of the technique (3). Various pulse oximeters have been validated in healthy untrained subjects with average maximal aerobic power (vosm,) during exercise (4, 8, 13). Although there are published reports (8,9,11) of use of pulse oximetry during exercise in endurance athletes, pulse oximeters have not been validated in this population. Furthermore, some authors (3) have specifically questioned the validity of pulse oximetry during intensive exercise in highly trained sub-

later, subjects reported to the laboratory for a series of four or five cycle ergometer tests with arterial blood sampling. The first test was a maximal incremental test; the subjects, breathing room air began at 70 W and increased 35 W every 90 s until volitional exhaustion. Three constant load tests followed the incremental test and consisted of 3 min of warm-up at -50% maximal power output, followed by a 3-min bout at -90-95% of maximal power output. To provide a wide range of hemoglobin saturation during exercise, the inspired partial pressure of 0, for the constant load tests were 164, 149, and 121 Torr. Gases were administered in a random single-blind fashion. The hyperoxic and hypoxic gases were supplied from high-pressure cylinders and were warmed and humidified. The normoxic gas sample was provided from a Tissot spirometer open to room air. Subjects breathed the selected gas mixture for 10 min before the resting blood samples were obtained. All exercise tests were separated by l-1.5 h to allow blood gases and pH to return to normal. Finally, to evaluate the accuracy of pulse oximetry during exercise with low hemoglobin saturation, three

0161-7567/92 $2.00Copyright 0 1992the American Physiological Society

455

456 TABLE

VALIDITY

1. %HbO, bias, variability,

correlation,

OF PI LSE OXIMETRY

and

confidence limits for entire data set Probe

n

%Hb02 Bias

Left ear Right ear Finger

273 273 273

0.72% -0.60%* 0.54%

%HbOz Bias 95% Confidence Limits

Standard Deviation

2.52 2.31 1.35

0.94 0.95 0.98

0.42 to 1.02 -0.87 to -0.32 098 to 0.70

%HbOz, oxyhemoglobin fraction of total hemoglobin; %Sp,,, pulse oximetry measurements of total hemoglobin; %SO,, available hemoglobin saturation. %HbO, bias = %Sp,, - %HbO,. * Left ear and finger probe biases were significantly different from right ear probe bias (P < 0.05). Mean t SD %HbO, and %SO, were 92.1t7.1% and 93.6t7.3% (SD), respectively.

subjects performed an additional maximal incremental test in which the fraction of inspired 0, was 89,93, or 100 Torr. Barometric pressure for all tests ranged from 760 to 764 Torr. 0, uptake was measured every minute during exercise with a Sensormedics metabolic measurement cart. The 0, and CO, analyzers were calibrated with appropriate precision gases immediately before every test. The volume- measuring turbine and temperatu .re tran .sducer were calibrated on a daily basis. An Allen’s test was used to ensure adequate ulnar collateral circulation before catheterization of the radial artery of the subjects’ nondominant extremity. The insertion site was anesthetized with lidocaine, and the catheter was inserted and taped into place. A 3-ml blood sample was anaerobically withdrawn into a heparinized syringe at rest and during the last 30 s of each exercise stager After collection, blood samples were capped, placed in ice, and analyzed within 60 min on an Instrumentation Laboratory model 282 CO-oximeter for methemoglobin (MetHb), carboxyhemoglobin (HbCO), + HbCO + MetHb + %HbO, [ %HbO, = (HbO,/HbO, Hb) X 1001, and “available” hemoglobin saturation + Hb) X 1001. The CO-oxi[%So,; %So, = HbO,/(HbO, meter was calibrated immediately before each subject’s samples were analyzed and checked periodically thereafter. The CO-oximeter operator was not apprised of the pulse oximeter results. The reliability of the CO-oximeter was assessed by performing duplicate analysis on 50 randomly selected samples. The difference in %HbO, between the two samples was 0.02 t 0.26% (SD), which was not significantly different from zero (P = 0.67). The mean difference between 50 duplicate %SO, samples was

0.11 t 0.29% (P = 0.02). Three identical pulse oximeters were used (Ohemda 3740, software revision E) in the fast mode (3-s average) of operation. Oximeter probes were placed on the pinna of both ears and on the index finger of the noncatheterized limb. Before the probes were placed, the sites were vigorously cleaned with alcohol and gauze pads, and the probe cables were taped to the subject to minimize motion artifact. The %Sp,, values were manually recorded while arterial blood samples were drawn. The oximeters’ poor signal alarm and pulse waveform were closely monitored by the recording investigator to ensure that spurious data were not included. When a pulse oximeter’s loss of signal warning was present, saturation data for that narticular machine for that time neriod were not used.

The mean differences of %Spoq - %HbO, (%HbO, bias) and %Sp,, - %SO, (%SO, bias), standard deviations, and 95% confidence limits (95% CL) for bias were calculated. Differences between probes were compared with analysis of variance, followed with Scheffe’s tests when appropriate. Relationships between measured and predicted hemoglobin saturation were evaluated by correlation and regression. Data are presented as means -+ SD, and statistical significance was set at P < 0.05. RESULTS

Subjects’ mean age, weight, and height were 24.8 t 3.4 yr, 79.3 t 4.1 kg, and 174.1 t 13.9 cm, respectively. Mean or 70.6 t 4.2 ml. VO 2 max was 5.60 t 0.36 l/min kg -l. I]nin-l. All subjects had normal spirometry tests and CO diffusion capacity. Twenty-eight blood samples were not included in the analysis because one or more of the pulse oximeters indicated a loss of signal error. Approximately 90% of the missing pulse oximeter observations were due to loss of signal errors from the ear probes. A total of 273 resting and exercise arterial blood samples with three corresponding %Sp,, values were obtained. Mean values for HbCO and MetHb for all observations were 1.26 t 0.65% and 0.30 + O.lO%, respectively. Mean %HbO, bias of the three probes for the entire data set is given in Table 1. The finger and left ear bias values were significantly different from the right ear values, but on the basis of the bias and lowest random variability (standard deviation of the bias), the finger probe was judged to provide the best estimate of %HbO,. Finger probe %HbO, bias was significantly different from zero (P < 0.0001). Mean finger probe %SO, bias was -1.0 t 1.3% (95% CL = -1.1 to -0.8%), which was significantly different from zero (P < 0.0001). A total of 232 observations were obtained during exercise. There was a statistical difference between the finger and left ear %HbO, bias and that of the right ear (Table 2). Once again the finger probe was judged to be the best predictor of %HbO, on the basis of the lowest bias and variability. A scatter plot of finger probe %Sp,, and %HbO, for the exercise data subset is shown in Fig. 1. Finger probe %HbO, bias was significantly different from zero (P < 0.0001). Mean finger probe %SO, bias was -0.9 t 1.3% (95% CL = -1.1 to -0.8%), which was significantly different from zero (P < 0.0001). The correlation between % Sp,, and %SO,for the exercise subset was r = 0.99 and yielded the regression equation: %SO,=

2. %HbO, bias, variability, correlation, confidence limits for exercise data subset

TABLE

Probe

n

%HbO, Bias

Left ear Right ear Finger

232 232 232

0.87% -0.59%* 0.52%

Standard Deviation

2.60 2.40 1.36

and

r

%Hb02 Bias 95% Confidence Limits

0.93 0.94 0.98

0.54 to 1.21 -0.90 to -0.28 0.34 to 0.70

%HbOz bias = %Sp,, - %HbOz. * Left ear * and finger probe biases were significantly different from right ear probe bias (P < 0.05). Mean %HbOz and %SO, were 91.4rt7.5% and 92.9+7.6% (SD), respectively. See Table 1 for definition of abbreviations.

VALIDITY

OF PULSE

1

457

OXIMETRY

(P < 0.0001). The regression function predicting %SO, from %Sp, was %SO, = %Sp,, X 1.08 - 6.12 (SEE = 1.26, r = 0.99, P < 0.0001). DISCUSSION

80

Finger Probe Saturation (%SpO$ FIG. 1. Relationship between finger probe pulse oximetry of available hemoglobin (Spa,) and measured oxyhemoglobin of total hemoglobin (%HbO,) for exercise data subset. Solid gression function: %HbO, = 1.02 X %Sp, - 2.30 [r = 0.98; error of the estimate (SEE) = 1.32; n = 2321.

estimate fraction line, restandard

%Sp, x 1.049 - 3.53 [standard error of the estimate (SEEf = 1.21, P < O.OOOOl]. The validity of pulse oximetry has been questioned during high-intensity exercise, so bias was analyzed as a function of percent 60, max. Observations (n = 75) when aerobic power (~oJ was ~81% of vo, maxwere of particular interest, because arterial desaturation is most likely to occur during high-intensity exercise. When TO, was r81% Ofv02max, mean %HbO, (91.4 t 7.4%) and mean %Sp,, (91.3 t 6.8%) were not statistically different (P = 0.96). The %HbO, bias values of the left ear and finger probe were significantly smaller than those of the right ear (Table 3), although the bias range between the three probes was small (+0.25 to -1.1%). There were no signifrcant differences observed within the ear probes across the five VO, classes. Although statistically significant differences were present within the finger probe results, no practically significant differences emerged (Table 3), and finger probe %HbO, bias did not increase as relative 0, uptake increased. Once again, the finger probe was judged to provide the best estimate of %HbO,. A scatter plot of finger probe %Sp,, and %HbO, for the high-intensity exercise data subset is shown in Fig. 2. Mean finger probe %SO, bias was -1.2 t 1.4% (95% CL = -1.6 to -0.9%), which was significantly different from zero TABLE

3. Mean %HbO, bias as a function

Probe Zeft ear Right ear Finger

120%

Our results show that the pulse oximeter can accurately predict %HbO, in highly trained athletes during incremental and constant-load exercise over the range of 72-99% HbO, saturation. Furthermore, the accuracy was not adversely affected by increasing exercise intensity. Several other authors have evaluated the accuracy of predecessors of the present oximeter, utilizing patients and healthy subjects with average vo2 max. Kagle et al. (4) evaluated the Ohmeda 3700 with finger and ear probes in normoxic and hypoxic states during rest in healthy subjects. The correlation (r = 0.78) between %Sp,, and %HbO, with the finger probe was lower than the present data, whereas the ear probe results were similar (r = 0.98). The authors attributed the discrepancy between values from the probes to the time delay for blood flow in the finger to “arterialize.” Unfortunately, bias and variability results were not presented. Cecil et al. (1) reported a bias of -0.31 and a correlation of 0.83 for 311 observations collected from a diverse sample of 152 resting patients with the Ohmeda 3700. Powers et al. (8) evaluated the Ohmeda 3700 during exercise in subjects with average VO, and found that the finger probe bias was 1.0 t 1.8% when %HbO, was 92% (n = 43 observations). Pulse oximetry has not been evaluated during intensive exercise in subjects with superior Vozmar, which has led to criticism of studies reporting reductions in %Spo, in athletes. Hansen and Casaburi (3) have hypothesized that declines in ear perfusion near-maximal exercise may be responsible for the reduction in %Sp,, observed with use of ear probes during strenuous exercise, and as such these reductions may not represent a true decrease in %SO,. We have not presentedany data quantifying changes in ear lobe or finger perfusion during exercise, but even if ear and finger perfusion did substantially decrease during exercise, these reductions did not prevent the oximeters from accurately estimating %HbO,. The finger probe generally performed the best with respect to minimizing motion artifact, bias, and random variability. The ear probes may have been more variable than the finger probe, because many of our highly trained

of VOW,,

21-40%

41-60%

61-80%

81-100%

75 0.25+2.60t -1.llt2.53 -0.01+1.40t

0.04t1.93 43

1.6922.33 22

1.0222.44 80

l.lOt2.83 53

-0.53t1.74 0.59t1.30

-0.17t2.52 1.20t0.93*

-0.23t2.17 0.82t1.41*

-0.66t2.43 0.551-1.12

%HbO, bias = %Spo - %HbOz. There were no significant differences within ear probes (P > 0.05). * Significantly different from Sl-100% (P < 0.05); t significantly different from right ear (P < 0.05). The 95% confidence limits for finger probe %HbO, bias are -0.33 to 0.32. Mean %HbO, and %SO, for VO, 281-100% of %70srnax subset were 91.4t7.4% and 92.6t7.4% (SD), respectively. See Table 1 for definition of abbreviations.

458

VALIDITY

OF PULSE

OXIMETRY

during high-intensity exercise in elite endurance athletes. In conclusion, we found the Ohmeda 3740 oximeter, equipped with a finger probe, to be a valid predictor of %HbO, in elite endurance athletes over the range of 7299% hemoglobin saturation during constant-load exercise and maximal incremental exercise in normoxia and acute hypoxia. The authors thank Ian McCloud and Laura Thomas of Shands Hospital at the University of Florida for assistance with blood analysis and pulmonary function testing. This study was supported in part by a grant from the Ohmeda Corporation. Address for correspondence: D. Martin, Box 100154, JHMHC, Gainesville, FL 32610-0154. Received 14 December 1990; accepted in final form 22 August 1991. 1

90 Finger Probe Saturation (%Sp02) 80

100

FIG. 2. Relationship between finger probe Spqz and measured %HbOz when at ~81% of maximal aerobic power. Solid line, regression function: %HbOz = 1.07 X %Spo, - 6.58 (r = 0.97; SEE = 1.32; n 7 75).

subjects had very small earlobes, which resulted in poor placement of the oximeter probe. There were some statistically significant differences between the finger and ear probes, but the absolute magnitudes of these discrepancies were small and close to the resolution (1%) of the pulse oximeters. At least two possible explanations for differences between identical models are plausible: 1) differences in the individual devices, particularly variability in the light sources, and 2) motion artifact. Some of the subjects’ earlobes were barely large enough to accommodate the ear probe, which probably increased the incidence of motion artifact. Subjects with larger earlobes may have less variability between finger and ear probes. We would recommend that pulse oximeter users evaluate signal quality and strength in both ears, as well as the finger, and use the site that offers the best signal with the lowest incidence of motion artifact. Regardless of the reason for the %Sp,, variance across the three instruments, we believe that most of the statistically significant differences in %Sp,, bias between the three pulse oximeters are of little practical importance and are primarily the product of high statistical power resulting from a large number of observations. Possibly the most important question addressed by this work is whether this pulse oximeter can accurately estimate %HbO, during high-intensity exercise? Significant arterial desaturation is most likely to occur *during high-intensity exercise, and when VO, was 281% VO, max, the finger probe’s %HbO, bias was -0.01 t 1.40%. Clearlv. this instrument accuratelv estimates %HbO,

REFERENCES

CECIL, W. T., K. J. THORPE, E. E. FIBUCH, AND G. F. TUOHY. A clinical evaluation of the accuracy of the Nellcor N-100 and Ohmeda 3700 pulse oximeters. J. Clin. Monit. 4: 31-36, 1988. 2. ESCOURROU,P. J., M. F. DELAPERCHE, AND A. VISSEAIJX. Reliability of pulse oximetry in pulmonary patients. Chest 97: 635-638, 1990. 3. HANSEN, J. E., AND R. CASABURI. Validity of ear oximetry in clinical exercise testing. Chest 91: 333-337, 1987. 4. KAGLE, D. M., C. M. ALEXANDER, R. S., BERKO, M. GIIJFFRE,AND J. B. GROSS. Evaluation of the Ohmeda 3700 pulse oximeter: steady state and transient response characteristics. Anesthesiology 1.

66: 376-380,1987.

LAWLER, J., S.-K. POWERS, AND D. THOMPSON. Linear relationship between VO,,, and VO,,, decrement during exposure to acute hypoxia. J. Appl. Physiol. 64: 1486-1492, 1988. 6. MILLER, A., 3‘. C. THORTON, R. WARSHAW, H. ANDERSON, A. TEIRSTEIN, AND I. J. SELIKOFF. Single breath diffusing capacity in a representative sample of the population of Michigan, a large industrial state. Am. Rev. Respir. Dis. 127: 270-277, 1983. 7. MORRIS, J. F., A. KOSKI, AND L. C. JOHNSON. Spirometric standards for healthy nonsmoking adults. Am. Rev. Respir. Dis. 103: 5.

57-67,197l. 8.

POWERS, S. K., S. DODD, J. FREEMAN, G. D. AYERS, H. SAMSON, AND T. MCKNIGHT. Accuracy of pulse oximetry to estimate HbOa fraction of total Hb during exercise. J. Appl. Physiol. 67: 300-304,

9.

POWERS, S. K., S. DODD, J. LAWLER, G. LANDRY, M. KIRTLEY, T. MCKNIGHT, AND S. GRINTON. Incidence of exercise induced hypoxemia in elite athletes at sea level. Eur. J. AppZ. Physiol. Occup.

10.

POWERS, S. K., J. LAWLER, J. A. DEMPSEY, S. *DODD, AND G. LANDRY. Effects of incomplete gas exchange on Vozmex. J. AppZ.

1989.

Physiol.

Physiol.

58: 298-302,

66: 2491-2495,

1988.

1989.

POWERS, S. K., AND J. WILLIAMS. Exercise-induced hypoxemia in highly trained athletes. Sports Med. 4: 46-53, 1987. 12. REIS, A., J. T. FARROW, AND J. L. CLAUSEN. Accuracy of two ear oximeters at rest and during exercise in pulmonary patients. Am. 11.

Rev. Respir.

Dis.

132: 685-689,

1985.

SMYTH, R. J., A. D. D’URZO, A. S. SLUTSKY, B. M. GALKO, AND A. S. REBUCK. Ear oximetry during combined hypoxia and exercise. J. AppZ. Physiol. 60: 716-719, 1986. 14. WILLIAMS, J. H., S. K. POWERS, AND M. K. STUART. Hemoglobin desaturation in highly trained athletes during heavy exercise. Med. 13.

Sci. Sports

Exercise

18: 168-173,

1986.