INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sickle cell disease is often complicated, acutely or chronically, by pulmonary infiltrates and hypoxemia (1). Furthermore, because it promotes sickling and vasoocclusion, arterial hypoxemia can lead to nonpulmonary organ injury. Thus, the detection of arterial hypoxemia is of great importance in the management of sickle cell disease. A complicating problem is that sickle erythrocytes have strongly right-shifted oxyhemoglobin dissociation curves because of high concentrations of 2,3,-DPG and because of the severely reduced oxygen affinity of polymerized hemoglobin (2, 3). Consequently, many patients with sickle cell disease have abnormally low arterial blood oxygen saturation (Sa_O2), even at sea level and when gas exchange is normal.

The most appropriate means to assess the adequacy of arterial oxygenation would be noninvasive pulse oximetry (4). The accuracy of pulse oximetry in sickle cell disease has been subject to at least five previous studies, but methodologic problems have led to conflicting claims that pulse oximetric readings may be correct, too high, or too low (5).

Contributing to the problem is confusion regarding the terminology of blood oxygenation, specifically the term "saturation," abbreviated as SO₂, or as Sa_O2, Sv_O2 or Sv_O2 when referring to arterial, venous, or mixed venous saturation measured gasometrically or by co-oximetry (10). Although some investigators have used the term SO₂ to refer to the ratio of oxyhemoglobin (O₂Hb) to total hemoglobin (tHb) (11), most authorities recommend that the ratio O₂Hb/tHb should be termed FO₂Hb (Fa_O2Hb when referring to arterial blood and Fv_O2Hb or Fv_O2Hb when referring to venous or mixed venous blood), whereas SO₂ should be used to refer to the ratio of O₂Hb to the sum of O₂Hb and deoxyhemoglobin (HHb), taking carboxy- (COHb), met- (metHb), and sulf- (sulfHb) hemoglobins out of consideration (10, 12). For this report, the terms SO₂ and FO₂Hb are used as defined above, and Sp_O2 is used to describe pulse oximetry data (10).

We undertook this study to try to determine whether pulse oximetry accurately estimates Sa_O2 in sickle cell disease.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

All subjects gave informed consent for the study procedures, which had been approved by Montefiore's Institutional Review Board for the Protection of Human Subjects.

In 17 adult patients with sickle cell disease who were hospitalized for treatment of vasoocclusive crisis and who had not been transfused for at least 3 mo, we compared 22 pulse oximeter readings with data from simultaneously drawn arterial blood samples measured for Fa_O2Hb and Sa_O2, using co-oximetry. Pulse oximetry was carried out with an Ohmeda model 3700 pulse oximeter (Ohmeda, Boulder, CO). After at least 2 min of stable Sp_O2 readings and regular pulsatile photoplethysmography apparent on the oximeter screen, we noted the Sp_O2 reading and drew arterial blood in a strictly anaerobic fashion. In most cases, we also drew peripheral venous samples and performed co-oximetry and blood gas analysis. We maintained the blood samples on ice, and, within 10 min, measured pH, PCO₂, Po₂, tHb, and O₂Hb, COHb, metHb, and HHb percentages, using a Radiometer model ABL625 CO-oximeter and blood gas analyzer (Radiometer, Copenhagen, Denmark). We calculated Sa_O2 as: 100% × O₂Hb/(O₂Hb + HHb).

We plotted Sp_O2 against Fa_O2Hb and against Sa_O2. For both arterial and venous blood samples, we also plotted SO₂ against oxygen tension (PO₂), and against pH-corrected PO₂, using Severinghaus' formula (13). For samples with SO₂ between 20 and 80%, we calculated p50, using Severinghaus' formula (13). To assess accuracy of the pulse oximeter, we performed linear regression analyses of Sp_O2 against Fa_O2Hb and against Sa_O2, and, for each subject, we calculated the differences between Sp_O2 and Fa_O2Hb and between Sp_O2 and Sa_O2. We computed the mean and SD of the differences, and, to assess average error, we computed the mean and SD of the absolute value of the differences.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The co-oximetric and pulse oximetric data are shown in Table 1. All of our patients were anemic, and most had hypoxemia and minimally to moderately elevated carboxyhemoglobin levels. All except one venous sample and most of the arterial samples lay to the right of the normal Severinghaus curve, even after the Bohr shift was taken into account (Figure 1). Nine samples from eight subjects had SO₂ values between 20 and 80%, allowing p50 to be calculated, using Severinghaus' equations (13). All except one of these samples had p50 well above normal.

View larger version (12K): [in this window] [in a new window]   Figure 1.   Oxygen affinity of sickle RBCs in vivo. SO2 is plotted against PO2 of arterial and venous blood samples from patients with sickle cell disease in acute vasoocclusive crisis. Where the pH differed substantially from 7.40, SO2 values were corrected to pH 7.4, using Severinghaus' equations (13), and the corrections are plotted as arrowheads. Normal data from Severinghaus (13) are shown as the solid line. All except one venous sample and most of the arterial samples lie to the right of the normal Severinghaus curves.

Pulse oximetry generally overestimated Fa_O2Hb, by an average of 3.4 ± 3.4 percentage points (mean ± SD) (Figure 2 and Table 2). Pulse oximetry generally underestimated Sa_O2, on average by 1.1 ± 0.8 percentage points (Figure 3). In one case, a patient with 62% Sa_O2, the pulse oximeter was grossly inaccurate, reading 74%. There was only one other case in which the difference between Sp_O2 and Sa_O2 (absolute value) exceeded 5 percentage points.

View larger version (14K): [in this window] [in a new window]   Figure 2.   Correlation of pulse oximetry with oxyhemoglobin percentage. Pulse oximetry readings (SpO2) are plotted against oxyhemoglobin percentage (FaO2Hb), measured by co-oximetry. The solid line is the line of identity. Although the correlation was strong and highly significant, pulse oximetry consistently overestimated FaO2Hb by an average of 3.4 percentage points (p < 0.05).

View larger version (14K): [in this window] [in a new window]   Figure 3.   Correlation of pulse oximetry with hemoglobin saturation (SaO2 = 100% times oxyhemoglobin divided by oxyhemoglobin plus reduced hemoglobin). Pulse oximetry (SpO2) readings are plotted against SaO2, measured by co-oximetry. The solid line is the line of identity. The correlation was highly significant, and the difference between pulse oximetry reading and co-oximetric saturation averaged only 1 percentage point (NS).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We confirmed that RBCs from most patients with sickle cell disease have abnormally low oxygen affinity, resulting in far right-shifted oxyhemoglobin dissociation curves in vivo. Only a small fraction of the right shift could be explained by the normal Bohr shift (3). Because of the right-shifted curve, estimates of SO₂ from blood gas data, based on assumed normal p50, are without value.

A right-shifted oxyhemoglobin dissociation curve is generally considered adaptive in anemia, allowing "unloading" of relatively large volumes of oxygen to tissues at relatively high PO₂, which preserves a high driving pressure for diffusion of oxygen into poorly vascularized tissues and/or reduces the need for increased cardiac output. However, in sickle cell disease, to the extent that the well-preserved tissue PO₂ discourages increasing cardiac output as a compensation for the low oxygen-carrying capacity of the anemic blood, venous blood becomes even more severely deoxygenated than in other forms of anemia. Hemoglobin polymerization depends on red cell concentrations of deoxyhemoglobin (14), so the right-shifted oxyhemoglobin dissociation curve may indirectly encourage polymer formation, sickling, and perhaps the consequent organ damage.

We also showed in patients with sickle cell disease, as in other patients with measurable amounts of carboxyhemoglobin (4, 15), that pulse oximetry overestimates Fa_O2Hb, typically by a few percentage points. Pulse oximetry usually underestimates Sa_O2 (oxygenation of the total available hemoglobin) only slightly, typically by 1 percentage point. The one grossly erroneous pulse oximeter measurement was not unexpected, as pulse oximeters are calibrated for accuracy over the range of Sa_O2 values above 80%, and are known to be inaccurate at lower values, especially among patients with anemia (4). In our experience, as long as the pulse oximeter screen showed a good photoplethysmogram and yielded stable readings, the minor errors in pulse oximeter data never led to an incorrect diagnosis, either of hypoxemia or of normoxemia.

Previous studies have come to conflicting and confusing conclusions regarding the accuracy of pulse oximetry in sickle cell disease. In two small studies, one of seven (5) and the other of three (6) children, pulse oximetry was found to be reasonably accurate, as compared with Sa_O2 calculated from measured Pa_O2 and oxyhemoglobin dissociation curves that had been determined by spectrophotometry and blood gas analysis in vitro.

In a study of adult patients with sickle cell disease, Comber and Lopez (7) found that Sp_O2 values were much lower than Sa_O2 values calculated from arterial blood gas measurements (assuming normal p50). They concluded that pulse oximetry seriously underestimates oxygenation of patients with sickle cell disease. On the other hand, Pianosi and associates (8) found a generally good correlation between Sp_O2 and Sa_O2 calculated from blood gas measurements in arterialized capillary blood from children with sickle cell disease (8). There was substantial individual variation, which depended, not surprisingly, on p50. In the studies of both Comber and Lopez and Pianosi and associates, since p50 was not likely to have been near normal, the calculated saturation measurements must be assumed not to have been accurate. Thus, neither of these studies compared pulse oximetry to a gold standard, and, in fact, in both studies, the pulse oximetry data are more likely to have been accurate than the calculated saturations.

In another pediatric study, Craft and associates (9) compared pulse oximeter to co-oximeter data in children with sickle cell disease. Although it is not clear from their report, Craft and associates apparently compared Sp_O2 readings with measured Fa_O2Hb, which they called "saturation," rather than with conventionally defined Sa_O2 (oxyhemoglobin as a percentage of available hemoglobin rather than as a percentage of total hemoglobin) (10, 12). Craft and associates found, as we did, that pulse oximetry overestimates Fa_O2Hb (9); not surprisingly, they also found that the severity of overestimation correlated with carboxyhemoglobin percentage (Fa_COHb).

For two related reasons, it is to be expected that pulse oximeters better reflect Sa_O2 than Fa_O2Hb. First, because of the wavelengths utilized by pulse oximeters and the shapes of the absorption spectra of oxyhemoglobin and carboxyhemoglobin, pulse oximeters read oxyhemoglobin and carboxyhemoglobins similarly; in essence, a pulse oximeter reading usually reflects the sum of oxyhemoglobin and carboxyhemoglobin percentages (16). When the carboxyhemoglobin level is less than 1%, Fa_O2Hb and Sa_O2 are virtually identical. When COHb levels are even slightly higher, however (as occurs in most normal subjects), the sum of Fa_O2Hb and Fa_COHb is actually a better reflection of Sa_O2 than is Fa_O2Hb alone. Therefore, given that they can only discriminate two species of hemoglobin, the fact that pulse oximeters tend to read COHb as O₂Hb improves their ability to estimate Sa_O2. Second, normal subjects have small, but measurable and variable levels of carboxyhemoglobin and met-hemoglobin, as did the patients with sickle cell disease we studied, making their Sa_O2 somewhat higher than their Fa_O2Hb. Thus, since pulse oximeters are calibrated by comparison with co-oximetry measurements of Sa_O2 rather than Fa_O2Hb in normal subjects breathing normoxic and hypoxic gases, it comes as no surprise that pulse oximetry provides a more accurate estimate of Sa_O2 than Fa_O2Hb.

Whether Fa_O2Hb or Sa_O2 is the better index of blood oxygenation depends upon the clinical circumstances. When it is important to know blood oxygen content, Fa_O2Hb might yield a closer estimate than would Sa_O2. For example, when high percentages of carboxyhemoglobin are present, as in victims of smoke inhalation, heavy smokers, and patients with hemolytic anemia (including sickle cell disease [16]), Fa_O2Hb would reflect the state of oxygenation better than would Sa_O2. However, even if pulse oximeters accurately measured Fa_O2Hb, in order to allow accurate assessment of blood oxygen content, it would still be necessary to collect blood for measurements of tHb and Pa_O2, measurements that cannot yet be made noninvasively. For most clinical purposes in which noninvasive monitoring of oxygenation is appropriate, e.g., screening for hypoxemia or titration of supplemental oxygen delivery, a measure of Sa_O2 would be as revealing or more revealing than a measure of Fa_O2Hb because when even small amounts of either COHb or metHb are present, even thoroughly oxygenated blood will not approach 100% Fa_O2Hb. Under such circumstances, a low Fa_O2Hb might prompt a clinician to use unnecessarily high FI_O2. In sickle cell disease in particular, saturation monitoring by pulse oximetry provides an appropriate guide to the adequacy of supplemental oxygen therapy, but it must be borne in mind that, because of the right-shift of the oxyhemoglobin dissociation curve, Sp_O2 in the 92 to 95% range (at sea level) may reflect normal gas exchange.

Our study confirms in adults that pulse oximetry has acceptable accuracy for reliable clinical diagnosis of serious gas exchange abnormalities in sickle cell disease. Despite our patients' severe anemia, which has been shown to impair the accuracy of pulse oximetry (4), as long as strong and regular photoplethysmographic waves were present, we found that pulse oximeters could be relied upon not to misdiagnose either hypoxemia or normoxemia in such patients. As in all other patients, additional tests for the severity of anemia, the adequacy of cardiac output, and sometimes the carboxyhemoglobin (and met-hemoglobin) levels are required for full evaluation of oxygen transport.