Maximizing the Laboratory Setting for Testing Devices and Understanding Statistical Output in Pulse Oximetry : Anesthesia & Analgesia

Secondary Logo

Journal Logo

Review Article: Review Article

Maximizing the Laboratory Setting for Testing Devices and Understanding Statistical Output in Pulse Oximetry

Batchelder, Paul B. RRT; Raley, Dena M. BSBE

Author Information
doi: 10.1213/01.ane.0000268495.35207.ab
  • Free


Testing medical monitoring devices in a human performance laboratory is usually conducted when there is no other, less complicated, reference that can be used to “stand in” for the patient.

Using pulse oximetry as the example, we will give a brief description of comprehensive developmental testing in a laboratory setting, then limit the scope of the discussion to one facet of testing, that of baseline Spo2 accuracy determination in best-case conditions. We will go into some detail regarding design of study for this type of testing, including methods that are currently used to control variables introduced by the reference device, device under test, and the physiology of the human test subject. Finally, using examples from a physiological monitoring reference laboratory1, we will describe the statistical output of laboratory testing currently required by regulatory organizations along with practical implications to clinical performance.


The less direct the measurement and the more interfering variables involved, the more likely empirical human testing will be required. Development and validation of some devices does not require empirical laboratory testing in humans. For instance, some devices use traceable reference gases that stand in for the human patient and provide a comparison standard that also allows calibration by the clinician after the device has left the factory (1).

There is no standard reference material for pulse oximeters, and they are not suitable for calibration by the user. To paraphrase the current International Organization for Standardization (ISO) Pulse Oximetry Standard “There is today no accepted method of verifying the correct calibration of a pulse oximeter probe/pulse oximeter monitor combination other than testing on human beings. This is due to the complexity of the optical intricacies of the interaction of light and human tissue upon which pulse oximetry depends” (2); While some aspects of functionality can be appraised with the use of simulators (2), pulse oximetry involves a host of physio-optical interactions and, as such, is in the category of devices that require empirical human accuracy tests to develop, calibrate, and validate (3,4). Table 1 provides a brief outline of a comprehensive set of pulse oximetry tests, most of which involve empirical testing in human subjects.

Table 1:
Human Laboratory Tests Commonly Used During Pulse Oximetry Development—Clinimark Laboratories

A complete description of the pulse oximetry development process is outside the scope of this paper. Instead, we will confine our discussion to that of Basic Accuracy Testing, which is the one test that is required to validate basic SpO2 accuracy by regulatory agencies (2). The Basic Accuracy Test consists of a Human Arterial Desaturation Study (induced hypoxia), which provides a “best case” estimate of the “trueness” of the pulse oximeter reading in nominal conditions. Nominal conditions are defined here as best-case circumstances with no interfering conditions (5). The ISO 9919 standard refers to these circumstances as, “… clinical laboratory conditions under which … many of the known sources of error in pulse oximetry are virtually eliminated. Examples of such sources of error are low perfusion, EMI, motion, nail polish, pulse oximeter probe mispositioning and ambient light.”


Maximization of the laboratory setting requires a study design with methodology that allows control of variables that can introduce additional error to the determination of saturation trueness in nominal conditions. While a comprehensive analysis of the effects of different study designs is outside the scope of this paper, we will inventory several factors mentioned in the literature that have been shown to increase error in the determination of pulse oximeter accuracy. Taken together, the methods listed here can control many of the known issues that can obscure the final results.

We have categorized these study design factors into four distinct areas. They are Human Physiology Management, Device Under Test Issues, Preanalytical— handling the blood sample, and Analytical— management of the Co-oximeter.

Human Physiology Management

The human body continuously attempts to maintain homeostasis with an arterial saturation in the range of 90%–100% (6). This is in opposition to the goal of a pulse oximeter validation study, which is to cause the oxygen saturation to be less than that which the body considers suitable. During a desaturation study, not only are low saturation levels required but stable plateaus must also be induced. In effect, the body is constantly at odds with the goals of a desaturation study.

Some of the more important physiological test conditions that can be manipulated to limit these potential sources of error include:

  • Delay time in the systemic arterial blood supply
  • Creation of saturation plateaus
  • Reduction of interfering substances
  • Physiological conditions that increase the noise to signal ratio

Delay Time

In the presence of a changing saturation, since the pulse oximetry sensor is at a slightly different location than the arterial sample site, there is a difference between when a change in saturation reaches the co-oximeter sample site and when that same change in saturation reaches the pulse oximeter sensor site (7).

To minimize the effects of delay time, choice of an arterial sample site that is close to the pulse oximeter senor site is preferable. Another factor is the patency of the arterial catheter. A well-flowing arterial catheter is required (8,9). If the blood flows too slowly, the chances of missing the “stable” saturation period are increased.

Saturation Plateau

A saturation plateau condition must be created that lasts long enough to allow the co-oximeter sample site and the pulse oximeter sensor site to experience the same oxygen saturation for a period of time that will consider the averaging time of the pulse oximeter (10), the length of time to draw the arterial sample, and time for the pulse oximeter to restabilize after the potential interruption of blood flow during sampling (2,7,11).

Factors that can affect a stable saturation plateau are:

  • Changes in the physical status of the test subject
  • Fio2 administration equipment
  • Methods to identify the quality of the plateau

Some changes in the physical status of the test subject, while not considered unhealthy, can impact the subject's physiology to a degree that the ability to maintain a stable saturation plateau is impaired. An example of this is shown in Figure 1. After strenuous exercise, while breathing an Fio2 of 0.1 for 12 min, the test subject exhibited highly unstable saturation levels that were not stable enough to use in comparison of pulse oximetry to co-oximetry.

Figure 1.:
Unstable saturation after strenuous exercise.

Because of the body's continual attempt to maintain a saturation in the 90% range, the system that is used to administer low Fio2 levels must be configured, so that attempts of the subject's physiology to maintain homeostasis above 90% SaO2 can be overcome. Two primary methods have been described. One is a system that gradually delivers varying Fio2's in such a way that the saturation is slowly manipulated (12). The other involves voluntary hyperventilation of the test subject with larger, rapid, less gradual steps in the saturation (13). To achieve a stable saturation plateau, a method of determining when the plateau has been reached is also necessary (2,3,13).

Interfering Substances

Common interfering substances are carboxyhemoglobin, methemoglobin, nail polish, and artificial nails (14–16). When testing the pulse oximeters ability to read correctly in baseline conditions, the carboxyhemoglobin and methemoglobin should be restricted to normal values, unless the device is able to distinguish carbon monoxide and methemoglobin from oxygen. Nail polish and artificial nails should be removed before testing.

Interfering Physiological Conditions

Low perfusion and motion are two interfering physiological conditions that occur frequently and must be eliminated in baseline performance testing (2,12,17–21). The hands, and sometimes the entire body, of the test subject may be actively warmed to eliminate low perfusion and to improve circulation (22–24) (Table 2).

Table 2:
Compensations for Physiological Limitations

Device Under Test Issues

Intradevice variations in both the pulse oximeter and the sensor and device-related issues that directly influence quality of the readings can be controlled so that the influence is significantly minimized.

Intradevice Variation

Intradevice variation can originate in the manufacturing process when small differences in component tolerance add up to a measurable difference in final device performance. An example of this is shown in Figure 2, which shows manufacturing variance over an 8-month period. The average difference in most of the pulse oximeters manufactured during this period varied by 0.1% saturation in human room air Precision testing. However, one manufacturing run resulted in a jump of 0.33% saturation in the average difference testing. This would have resulted in a much larger difference in the low saturation ranges of a hypoxia test. The use of multiple units in a test can diminish the influence of intradevice variability. Multiple sensors of each sensor style should also be tested simultaneously, due to similar potential intrasensor variability.

Figure 2.:
Spo2 Precision in human room air tests.

Influences in Quality of Readings

Four device-related factors that can influence quality of the reading are the method of data collection, sensor placement, optical cross-talk, and electrical interference (noise sources from other equipment in the laboratory or under test).

Hand data collection requires that the data be entered manually into the statistical analysis program. This can be a source of inaccuracy in the study that can readily be eliminated by the use of electronic data collection. Great attention to sensor placement at the start of each test is important to ensure that all systems read as close to each other as possible. Inaccuracy as much as 12% has been reported as a result of sensor malposition (25). Optical interference is light reaching the sensor detector from any source other than from the LED's of the sensor. This can be environmental or from LED's of adjacent pulse oximeter sensors, which is known as “cross-talk.” Pulse oximeter probes should be covered with opaque material to prevent light interference (3,26–28). Finally, devices and sensors must be electrically shielded to prevent introduction of added noise from nearby electrical devices (3) (Table 3).

Table 3:
Device Under Test Recommendations

The Preanalytical Phase

The preanalytical phase is the entire process involved in handling the sample. This includes everything from obtaining the sample through insertion of the sample into the co-oximeter. The preanalytical phase can be the largest contributor of error to the arterial sample measurement (29).

Care should be taken during this entire phase. Specific steps that should be attended to are:

  • Syringe type and storage
  • Heparin
  • Identification of each syringe
  • Sample collection
  • Sample preparation after collection

Syringe Type and Storage

Studies indicate that plastic may be permeable to oxygen and carbon dioxide, which could introduce errors in the sample (30–34). Current standards call for the use of glass syringes when sample measurement will be delayed longer than 30 min after collection. In this case, it is also advised that the glass syringe be placed on ice (presumably to slow metabolic processes). Samples drawn in plastic syringes should be analyzed within 30 min and should not be iced (29,35).


The syringes should be preheparinized with dry heparin or, since liquid heparin can dilute the sample and alter the true value of a sample, care should be taken to expel all excess heparin from the syringe (36–39).

Sample Identification

Because of the large number of samples obtained in the relatively short period of a study, each syringe should be labeled in a manner that will allow clear identification. This can affect both sample management during analysis (error resulting in remeasurement) and statistical processing (facilitation of accurate data synchronization).

Sample Collection

Care should be taken to waste a sufficient volume of blood before collection of the sample in order to prevent contamination by flush solution (if used) or by old blood that may remain in the catheter from the previous sample. This is especially important when extension tubing is used.

Preparation After Collection

Immediately after sample collection, air bubbles must be expelled from the syringe (40–42). The sample should then be mixed thoroughly to dissolve the heparin. Failure to do so may lead to the formation of microclots which, in turn, can bias results, interfere with measurements, and lead to analyzer downtime (43) (Table 4).

Table 4:
Reduction of Preanalytical Errors

Analytical—Management of the Co-oximeter

The accuracy specification for commonly used co-oximeters is 1% (44,45). When one considers that the published accuracy specification of most pulse oximeter systems is 2%, it is apparent that measures to control additional inaccuracy introduced by the co-oximeter are of great importance.

There are three areas that can help limit uncertainty introduced by the co-oximeter. They are:

  • Quality assurance procedures
  • Manufacturer recommended procedures (maintenance, calibration, quality control)
  • Increased vigilance procedures (increased cleaning, use of multiple co-oximeters, and multiple syringe runs)

Quality Assurance Procedures

The Clinical Laboratory Information Act states that research laboratories are not required to meet the Clinical Laboratory Information Act requirements; however, quality assurance procedures that are required in laboratories reporting clinical data are recommended by regulatory standards (2,46). These quality assurance procedures include participation in a peer group survey similar to the CAP program, written protocols and procedures, and tracking documentation of quality control and calibration (29) (Peer Group Surveys, College of American Pathologists, 325 Waukegan Road, Northfield, Illinois 60093).

Manufacturer Recommended Procedures

Procedures recommended by the co-oximeter manufacturer are required by regulatory standards (2). These include routine maintenance, calibration, and quality control. It is crucial that the co-oximeter be cared for and operated in accordance with manufacturer directions.

Increased Vigilance Procedures

Increased Cleaning.

Pulse oximetry validation studies of this type usually involve approximately 25 syringes per test with each syringe being run 1–3 times within a period of an hour. This means that a co-oximeter receives approximately 100–375 samples in an 8 hour period with 250–1125 samples in a 4-day period. This is a higher sample load than normal clinical use; thus, attention to cleaning and protein removal is of much greater importance than in normal use. We have found that increasing the cleaning and protein removal frequency decreases error messages and other problems with the co-oximeter during this type of study (47).

Use of Multiple Co-oximeters.

Errors in the co-oximeter from manufacturer to manufacturer has been reported to be as much as 1.5% (48). A Clinimark study of three of the same models from two different manufacturers showed the intradevice variation of one manufacturer to be more than 0.3% and interdevice variation to be 0.5% between manufacturers (Fig. 3).

Figure 3.:
Intra- and inter-co-oximeter variability. Tests conducted using manufacturer recommended calibration and quality control procedures.

Another Clinimark study evaluated the agreement of three co-oximeters. Arterial blood samples were obtained from eight healthy volunteers over the range of 70%–100% saturation with approximately 25 samples per subject. In this study, two co-oximeters were compared with a third. Manufacturer-recommended calibration, quality control, and maintenance procedures were followed, and all co-oximeters met the calibration and quality control acceptance criteria. Figure 4 illustrates the point that the simultaneous use of multiple co-oximeters can provide important information regarding the measure of truth during a study.

Figure 4.:
Agreement of three co-oximeters.

If only one co-oximeter were used in a study, there would be no clear indication of outlier readings or reading drift. The use of more than one system provides better identification of drift, sample errors and outlying readings. It is recommended that multiple co-oximeters be used simultaneously during a validation study.

This highlights an important issue that there is currently no standard reference material for oxygen saturation. Thus, other than the highly technique-dependant, complex, and time-consuming Van Slyke method (49), there is no practical method to verify absolute accuracy of the SO2 reading.

Multiple Syringe Runs.

Real-time quality assessment can be accomplished by measuring each syringe in each co-oximeter multiple times. For example, small fairly undetectable air bubbles and analytical errors can be identified. An excerpt of data from one study is shown in Table 5. The table includes data from one co-oximeter after one arterial syringe sample was measured four times. Run number 2 gave a saturation value approximately 3% higher than the other three runs, and was clearly inaccurate. Without this multiple view of the readings, many errors can go unnoticed (Table 6).

Table 5:
Sample Error in One Syringe
Table 6:
Techniques to Reduce Analytical Errors



The goal of statistical representation of data is to provide clinicians with a fairly concise number that gives a picture of the overall expected performance. When we test a device, the information collected is a group of data points that usually miss the mark to some degree. In effect, when accuracy is tested we are making a determination of how many data points miss the mark and by how much: in other words, the inaccuracy of the data. Thus, “accuracy” statistics, in reality, portray inaccuracy which is defined more correctly as the uncertainty of a device (50). During this discussion of statistical output, it will help the reader to keep the concept of uncertainty in mind.

The performance specification of a pulse oximeter refers to a collection of data points comparing the device to a reference that provides true values. The features of this collection of comparative data points (or data cloud) are described using various statistical formulas that condense the information into a single number. The statistics that we are concerned with in pulse oximetry are Precision, Bias, and Arms (accuracy root mean square). We will explain Precision and Bias, then show how these two statistics are combined in the Arms calculation.

Precision—Size of the Data Cloud

The size of the entire grouping of data points of the test device, or data cloud, represents random errors in the readings of repeated measurements taken under the same conditions. Random error is caused by various sources of noise that result in a scatter of the readings, and is referred to as Precision. In effect, this statistic is the standard deviation (sd) of the data cloud grouped around the best-fit line of the data cloud (indicated by line B in Fig. 5). Precision does not tell how far from truth (indicated by line A in Fig. 5) the data cloud is, just how big it is. The Precision of two devices are shown in Figures 5 and 6. Repeated readings from the device in Figure 5 with a Precision (Sres)of 2.1 indicate less random noise, or random error, than readings from the device in Figure 6 with an Sres of 5.2, indicating poorer Precision.

Figure 5.:
Synthesized data for illustrative purposes. A, truth; B, test data best fit line; C, Mean Bias; D, one Local Bias point.
Figure 6.:
Synthesized data for illustrative purposes.

The formula for Precision as used here is seen in Figure 7. In this calculation, Precision is represented by Sres, which is the sd of the residuals. A residual is the difference of one test data point from the best-fit line drawn through all of the test data.

Figure 7.:

Bias—Offset of the Data Cloud

Bias is the offset of the test data from truth caused by systematic error. Mean Bias and Local Bias are two important characteristics of bias. Mean Bias is one number that represents the mean distance, or average offset, of the entire data cloud from truth (represented by C in Fig. 5). The formula for Mean Bias is shown in Figure 8. Mean Bias is the same over the entire Spo2 range; in other words, it is the same at 70% as it is at 100%.

Figure 8.:
Mean Bias.

The Actual Bias of a pulse oximeter, however, often varies in the different Spo2 ranges. Local Bias reflects the variability in the different ranges. Local Bias is the offset of each point over the Spo2 range and can indicate complex pulse oximetry calibration curves that are not well represented by the single Mean Bias number.

Local Bias is essentially the difference of each point on the best-fit line of the data cloud (B in Fig. 5) from truth at that point (two different Local Bias points are represented by D1 and D2 in Fig. 5). In Figure 5, all of the Local Bias values are different, which causes the slope (or calibration curve) of the test data to be different from the slope of the reference data. Local Bias can represent linear or nonlinear calibration curves. Data from a pulse oximetry reference study, shown in Figure 9, illustrate this point. Note that the tail of the data cloud in the lower saturation range curves up from the reference co-oximeter. The Local Bias formula is shown in Figure 10.

Figure 9.:
Variable offset.
Figure 10.:
Local Bias.

Arms—Combined Size and Offset of the Data Cloud

In his 1989 paper, Errors in 14 pulse oximeters during profound hypoxia, Severinghaus et al. recognized the need to indicate the magnitude of both Precision and Bias when describing the performance of a pulse oximeter (13). As an index of error, he coined the term “Ambiguity” to describe the absolute sum of Bias and Precision. Although the formulas used in this paper were somewhat different from those currently used in pulse oximetry performance specifications, the purpose remains the same: to include both Precision and Bias in a single number.

The Arms statistic is required by regulatory agencies when overall accuracy of a device is evaluated. The Arms calculation is affected both by Precision and by Bias (including both Mean and Local Bias). The name comes from the fact that it is the square root of the mean of the squares of the values. The values being the differences of each data pair (test reading—reference reading). Another way of describing the Arms calculation is that it looks at a “moving” mean or sd of pairs where the reference value is different for each test–reference pair. The formula can be seen in Figure 11.

Figure 11.:
A rms.

Each test–reference data pair is used in the calculation of the overall Arms number and influences the final magnitude of the results. This means that a data cloud with an acceptable Arms can have a large Bias with a tight Precision or a poor Precision with a small Bias, but it cannot have much of both and still have an acceptable Arms. Figures 5 and 6 illustrate this point; the data sets in both figures have the same Arms. Although Figure 5 has a larger Bias than Figure 6, it has a Precision that is much tighter. In this case, it has large Bias but tight Precision. Conversely, Figure 6 has a small Bias but poor Precision (more scatter).

Thus, the Arms statistic provides one number with a similar property to Severinghaus et al.'s Ambiguity, which includes both Precision and Bias in a single number and can tell the clinician generally how Accurate the pulse oximeter is or, to use the recommendation of National Institute of Standards and Technology, the Uncertainty of the pulse oximetry reading.

Data sets from human studies often have more complex features than simply poor Precision or large Bias. One example can be seen in Figure 12, where data from one of the test subjects followed the Reference fairly well from 100% to 85%, at which time the test pulse oximeter stopped reporting the true saturation and stayed at a reading of about 83%. With that subject in the data pool, the Arms was unacceptable. Another example is shown in Figure 13. The saturation of the test pulse oximeter popped-up within a 1-min period from 75% to 90% then dropped back to read correctly. In this case, with the additional presence of other pop-ups of less magnitude, the Arms was again outside the acceptable range.

Figure 12.:
Readings frozen at approximately 83%.
Figure 13.:

In the cases of Figures 12 and 13, the Local Bias of the anomalous readings influenced the calculation to a degree great enough resulting in an unacceptably large Arms.

Practical Implications of Arms

Most pulse oximeters in use today specify an Arms of 2%. What does this mean to the clinician? How inaccurate can a pulse oximeter be and still meet this specification? What does this number not tell the clinician?

The Arms is used to describe general performance over the entire saturation range and does not describe any one point but is a compilation of all points over the full range tested. Usually, saturation readings in the 90%–100% range show a smaller Arms than 2%; readings in the 70%–80% range are more than 2%; while readings in the 80%–90% range are about 2%.

How inaccurate can a pulse oximeter be and still meet a specification of 2%? To answer in clinical terms, Figure 14 is a set of data from an arterial desaturation study that included 12 test subjects. The Arms of this test set was 1.5%. Figure 15 is the same data set with the data from one subject given a 3.5% offset (data in oval). The addition of the offset in this one subject increased the Arms to 2.0%. Because of normal variations in the population, this type of offset is a common occurrence. Even though the Arms is 2.0%, some of the data in the oval are inaccurately high by as much as 3% in the 90's, and as much as 6% in the 70's. Thus, even in nominal patient conditions, a pulse oximeter can provide out of specifications readings in one patient and still be in specifications from a statistical point of view.

Figure 14.:
A rms of 1.5.
Figure 15.:
A rms of 2.0 one subject with an average bias of 3.5%.

At times, saturation pop-ups, drop-downs, frozen readings, and periods of no reading are seen in an ill-behaved pulse oximeter system. The Arms statistic is not designed to represent these dynamic or changing conditions; instead it is calculated from test–reference data pairs that are each collected at one point in time. The undesirable performance condition may not be reflected in the statistics unless the condition happens to occur at the moment the data are being collected. Additionally, nonreading conditions are not represented in the Arms statistic.

In summary, even with a well-behaved system, some readings were seen to be off by as much as 3%–6%, as shown in the example above; additionally, the Arms statistic does not reflect many dynamic performance conditions, such as no readings, that can be important in the clinical application of a pulse oximeter.


We have described components of study design that will maximize the laboratory setting for accuracy testing in nominal pulse oximetry conditions; outlined improvements that can be made to current pulse oximetry laboratory methods; and explained the statistics used in pulse oximetry.

As we continue to improve our ability to assess baseline performance, it is wise to keep in mind that while we have discussed the foundation of pulse oximetry development, possibly the most important aspects of pulse oximetry readings are those times that happen less often but in the most critical patients. In those conditions uncertainty of the saturation reading can be much greater than that seen in nominal conditions.

This paper describes one portion of pulse oximeter development, that of accuracy assessment in nominal conditions, which are required by regulatory organizations before clearance for use in the clinical arena. The information from this type of testing provides information about a “collection of specific intervals” during the test that compares the pulse oximeter to a co-oximeter, which can only measure point-in-time data.

Next steps would be for regulatory organizations to expand the scope of required testing that includes a more comprehensive assessment of pulse oximetry performance such as the ability of a pulse oximeter to provide saturation readings continuously. Thus, one might envision assessment of pop ups, dropouts, frozen readings, and periods of no reading through assessment of sensitivity/specificity and possibly a “Performance Index” similar to the approach taken by Barker (18,51–53). Finally, development of an accessible method to verify the absolute accuracy of the co-oximeter such as a reference material for oxygen saturation is also needed.


The authors thank Steven J. Barker, PhD, MD, Professor and Head Department of Anesthesiology University of Arizona College of Medicine for his thoughtful review of the manuscript. Also, the authors thank Paul D. Mannheimer, PhD, Chief Scientist, Nellcor/Tyco Healthcare and Sandy Weininger, PhD, Electrical Engineer, US FDA Center for Devices and Radiologic Health for assistance and suggestions in reviewing portions of the manuscript.


1. ISO. International vocabulary of basic and general terms in metrology (VIM). Geneva, Switzerland: ISO, 1993
2. ISO 9919. Medical electrical equipment—particular requirements for the basic safety and essential performance of pulse oximeter equipment for medical use. Geneva, Switzerland: ISO, 2005. International Standards, 2nd ed. 2005-03-15
3. Schowalter JS. Calibration. In: Webster JG, ed. Design of pulse oximeters (Medical science series). London: Institute of Physics Publishing, 1997:159, 163–72, 184, 185, 194
4. Mannheimer PD. The physio-optics of pulse oximetry, numerical modeling and clinical experience. Doctoral Thesis, University of Luebeck, Luebeck, Germany, 2003
5. Batchelder PB, Clifford D, Goldman JM. Pulse oximetry: real world performance. In: Vincent JL, ed. Yearbook of intensive care and emergency medicine. Berlin: Springer-Verlag, 2004: 647–55
6. Elin RJ. Laboratory reference interval values of clinical importance. In: Wyngaarden JB, Smith LH, eds. Cecil textbook of medicine. 19th ed. Philadelphia: WB Saunders, 1992:2374
7. Ding ZN, Shibata K, Yamamoto K, Kobayashi T, Murakami S. Decreased circulation time in the upper limb reduces the lag time of the finger pulse oximeter response. Can J Anaesth 1993;40:189–90
8. Kulkarni M, Elsner C, Ouellet D, Zeldin R. Heparinized saline versus normal saline in maintaining patency of the radial artery catheter. Can J Surg 1994;37:37–42
9. Kaye J, Heald GR, Morton J, Weaver T. Patency of radial arterial catheters. Am J Crit Care 2001;10:104–11
10. Kelleher JF. Pulse oximetry. J Clin Monit 1989;5:37–62
11. McCarthy K, Decker MJ, Kingman PS, Stoller JK. Pulse oximetry—pulse oximeter accuracy and response time. In: Kacmarek RM, Hess D, Stoller JK, eds. Monitoring in respiratory care. St. Louis: WB Saunders, 1993:316–17
12. Gehring H, Hornberger C, Matz H, Konecny E, Schmucker P. The effects of motion artifact and low perfusion on the performance of a new generation of pulse oximeters in volunteers undergoing hypoxemia. Respir Care 2002;47:48–60
13. Severinghaus JW, Naifeh KH, Koh SO. Errors in 14 pulse oximeters during profound hypoxia. J Clin Monit 1989;5:72–81
14. Hinkelbein J, Genzwuerker HV, Soql R, Fiedler F. Effect of nail polish on oxygen saturation determined by pulse oximetry in critically ill patients. Resuscitation 2007;72:82–91
15. Cote C. Effect of nail polish on pulse oximetry. Anesth Analg 1988;67:683–6
16. Rubin AS. Nail polish color can affect pulse oximeter saturation. Anesthesiology 1988;68:825
17. Nishiyama T. Pulse oximeters demonstrate different responses during hypothermia and changes in perfusion. Can J Anaesth 2006;53:136–8
18. Barker SJ. “Motion-resistant” pulse oximetry: a comparison of new and old models. Anesth Analg 2002;95:967–72
19. Hummier HD, Englelmann A, Pohlandt F, Hogel J, Franz AR. Decreased accuracy of pulse oximetry measurements during low perfusion caused by sepsis: is the perfusion index of any value? Intensive Care Med 2006;32:1428–31
20. Barker SJ. Standardization of the testing of pulse oximeter performance. Anesth Analg 2002;94;S17–20
21. Barker SJ. The effects of motion on the performance of pulse oximeters in volunteers. Anesthesiology 1997;86:101–8
22. Zhang XY, Zhang YT. The effect of local mild cold exposure on pulse transit time. Physiol Meas 2006;27:649–60
23. Schramm WM, Bartunek A, Gilly H. Effect of local limb temperature on pulse oximetry and the plethysmographic pulse wave. Int J Clin Monit Comput 1997;14:17–22
24. Ding ZN, Shibata K, Yamamoto K, Kobayashi T, Murakami S. Decreased circulation time in the upper limb reduces the lag time of the finger pulse oximeter response. Can J Anaesth 1992;39:87–9
25. Kelleher JF, Ruff RH. The Penumbra effect: vasomotion-dependant pulse oximeter artifact due to probe malposition. Anesthesiology 1989;71:787–91
26. van Oostrom JH, Mahia ME, Gravenstein D. The Stealth Station Image Guidance System may interfere with pulse oximetry. Can J Anaesth 2005;52:379–82
27. Trivedi NS, Ghouri AF, Shah NK, Lai E, Barker SJ. Effects of motion, ambient light, and hypoperfusion on pulse oximeter function. J Clin Anesth 1997;9:179–83
28. Amar D, Neidzwski J, Wald A, Finck AD. Fluorescent light interferes with pulse oximetry. J Clin Monit Comput 1989;5:135–6
29. NCCLS. Blood gas and pH analysis and related measurements; approved guideline. NCCLS document C4-A [ISBN 1-56238-444-9]. Wayne, PA: NCCLS, 2001
30. Knowles TP, Mullin RA, Hunter JA, Douce FH. Effects of syringe material, sample storage time, and temperature on blood gases and oxygen saturation in arterialized human blood samples. Respir Care 2006;51:732–6
31. Beaulieu M, Lapointe Y, Vinet B. Stability of PO2, PCO2, and pH in fresh blood samples stored in a plastic syringe with low heparin in relation to various blood-gas and hematological parameters. Clin Biochem 1999;32:101–7
32. Smeenk F, Janssen J, Arends B, Harff G, van den Bosch J, Schonberger J, Postmus PE. Effects of four different methods of sampling arterial blood and storage time on gas tensions and shunt calculation in the 100% oxygen test. Eur Respir J 1997;10:910–13
33. Liss HP, Payne C Jr. Stability of blood gases in ice and at room temperature. Chest 1993;103:1120–2
34. Schmidt C, Muller-Plathe O. Stability of PO2, PCO2, and pH in heparinized whole blood samples: influence of storage temperature with regard to leukocyte count and syringe material. Eur J Clin Chem Clin Biochem 1992;30:767–73
35. AARC. Clinical practice guideline: blood gas analysis and hemoximetry. Respir Care 2001;46:498–505
36. Avoid errors in arterial blood sampling. (Code 919-467). Copenhagen: Radiometer A/S, 1992
37. Hopper K, Rezende ML, Haskins SC. Assessment of the effect of dilution of blood samples with sodium heparin on blood gas, electrolyte, and lactate measurements in dogs. Am J Vet Res 2005;66:656–60
38. Gayed AM, Marino ME, Dolanski EA. Comparison of the effects of dry and liquid heparin on neonatal arterial blood gases. Am J Perinatol 1992;9:159–61
39. Ordog GJ, Wasserberger J, Balasubramaniam S. Effect of heparin on arterial blood gases. Ann Emerg Med 1985;14:233–8
40. Lu JY, Kao JT, Chien TI, Lee TF, Tsai KS. Effects of air bubbles and tube transportation on blood oxygen tension in arterial blood gas analysis. J Formos Med Assoc 2003;102:246–9
41. Wooley A, Hickling K. Errors in measuring blood gases in the intensive care unit: effect of delay in estimation. J Crit Care 2003;18:31–7
42. Wuillemin WA, Gerber AU. Sources of error in the pre-analytical phase of blood gas analysis. Schweiz Rundsch Med Prax 1995;84:200–3
43. Ancic P, Munoz F. Importance of pre-analytical variables in the quality of arterial blood gas determination. Rev Med Chil 1997;125:1045–8
44. OSM3 Hemoximeter, operators manual. C ed. Copenhagen: Radiometer Medical A/S, 1998. Code number 989–148
45. IL682 Co-oximeter system, operators manual. Rev. 3.0. Lexington, MA: Instrumentation Laboratory, 1995
46. Department of Health and Human Services, Centers for Disease Control and Prevention. Clinical laboratory improvements amendments (CLIA); 57 FR 7139. Atlanta, GA: Department of Health and Human Services, Centers for Disease Control and Prevention, 1992
47. Wu C, Kenny MA, Huang MC, Afromowitz MA, Yager P. Feasibility study of the spectroscopic measurement of oxyhemoglobin using whole blood with out pre-treatment. Analyst 1998;123:477–81
48. Aoyagi T. Is the co-oximeter a gold standard? Anesth Analg 2002;94(1S):A1
49. Gregory IC. Assessment of Van Slyke manometric measurements of oxygen content. J Appl Physiol 1973;34:715–17
50. Guidelines for evaluating and expressing the uncertainty of NIST measurement results, NIST technical note 1297. Gaithersburg, MD: NIST, 1994
51. Workie FA, Rais-Bahrami K, Short BL. Clinical use of new-generation pulse oximeters in the neonatal intensive care unit. Am J Perinatol 2005;22:357–60
52. Richards NM, Giuliano KK, Jones PG. A prospective comparison of 3 new-generation pulse oximetry devices during ambulation after open heart surgery. Respir Care 2006;51:29–35
53. Prielipp RC, Scuderi PE, Hines MH, Atlee JL, Butterworth JF. Comparison of a prototype esophageal oximetry probe with two conventional digital pulse oximetry monitors in aortocoronary bypass patients. J Clin Monit Comput 2000;16:201–9

1Clinimark Laboratories; before testing all studies received IRB approval from the Avista Adventist Hospital IRB Committee— subjects were consented healthy volunteers.
Cited Here

International Anesthesia Research Society