Pulse Oximetry in Children With Congenital Heart Disease ... - Masimo

ORIGINAL RESEARCHPulse Oximetry in Children WithCongenital Heart Disease: Effects ofCardiopulmonary Bypass and CyanosisAdalberto Torres Jr, MD, FAAP*Kelly M. Skender, MSN, PNP†Julie D. Wohrley, MD*Jean C. Aldag, PhD*Gary W. Raff, MD*G. Kris Bysani, MD*Dale M. Geiss, MD*The objective of this prospective, observational study withconsecutive sampling was to assess the reliability, bias,and precision of Nellcor N-395 (N) and Masimo SETRadical (M) pulse oximeters in children with cyanotic congenitalheart disease and children with congenital heartdisease recovering from cardiopulmonary bypass–assistedsurgery admitted to a cardiovascular operating suite andpediatric intensive care unit at a tertiary care communityhospital. Forty-six children with congenital heart diseasewere studied in 1 of 2 groups: (1) those recovering fromcardiopulmonary bypass with a serum lactic acid > 2mmol/L, and (2) those with co-oximetry measured saturations(SaO 2) < 90% and no evidence of shock.Measurements of SaO 2of whole blood were compared tosimultaneous pulse oximetry saturations (SpO 2). Datawere analyzed to detect significant differences in SpO 2readout failures between oximeters and average SpO 2–SaO 2± 1 SD for each oximeter. A total of 122 SaO 2measurementswere recorded; the median SaO 2was 83% (57 –100%). SpO 2failures after cardiopulmonary bypass were41% (25/61) for N versus 10% (6/61) for M (P < .001).There was a significant difference in bias (ie, average SpO 2– SaO 2) and precision (± 1 SD) between oximeters (N, 1.1 ±3.3 vs M, –0.2 ± 4.1; P < .001) in the postcardiopulmonarybypass group but no significant difference in bias and precisionbetween oximeters in the cyanotic congenital heartdisease group (N, 2.9 ± 4.6 vs M, 2.8 ± 6.2; P = .848). TheNellcor N-395 pulse oximeter failed more often immediatelyafter cardiopulmonary bypass than did the MasimoFrom the *University of Illinois College of Medicine at Peoriaand the †Children’s Hospital of Illinois at OSF St. Francis MedicalCenter, Peoria, Illinois.Received Jul 25, 2003, and in revised form Dec 29, 2003.Accepted for publication Dec 30, 2003.Address correspondence to Adalberto Torres Jr, MD, Children’sHospital of Illinois at OSF St. Francis Medical Center, 530 NEGlen Oak Ave, Peoria, IL, 61637; e-mail: altorres@uic.edu.Torres A Jr, Skender KM, Wohrley JD, Aldag JC, Raff GW, BysaniGK, Geiss DM. Pulse oximetry in children with congenital heartdisease: effects of cardiopulmonary bypass and cyanosis. JIntensive Care Med. 2004;19:229-234.DOI: 10.1177/0885066604263819SET Radical pulse oximeter. SpO2 measured with bothoximeters overestimated SaO2 in the presence of persistenthypoxemia.Key words: monitoring, pediatrics, cardiac surgery, oxygensaturation, pulse oximetry, arterial blood gasesPulse oximetry is considered by many clinicians asthe fifth vital sign and is widely used in many clinicalsettings [1]. Intensivists rely heavily on this noninvasiveinstrument to continuously assess the adequacyof oxygenation and to track the need andresponse to therapy. However, it has been welldocumented that pulse oximetry is less reliable andless accurate under certain conditions, such as poorperipheral perfusion and in children with hypoxemia[2-11]. Using various hypoperfusion models(eg, tourniquet, decreased ambient temperature),multiple investigators have demonstrated in adultvolunteers that pulse oximeters increasingly failand become less accurate, that is, > 2% difference inpulse oximeter saturation (SpO 2) and co-oximetermeasured saturation (SaO 2) [2,7,8]. Others havedemonstrated that pulse oximeters become lessaccurate in patients with hypoxemia, includingboth children [4,5,9] and adults [6,10]. Clinicianscaring for patients with severe shock or hypoxemia(eg, children with cyanotic heart disease) are oftenrequired to obtain arterial blood gases to accuratelyassess the patient’s oxygenation or make a lessthan adequately informed medical decision.Manufacturers of 2 newer generation pulseoximeters, the Masimo SET Radical and the NellcorN-395, claim an accuracy of ± 3% during periods ofpoor peripheral perfusion and with arterial oxygensaturations as low as 70% [12,13]. These 2 oximetershave not been validated in children with poorCopyright © 2004 Sage Publications 229

Torres et alperfusion or hypoxemia. We hypothesized thatSpO 2measured with the Nellcor N-395 and theMasimo SET Radical pulse oximeters were neitheraccurate nor precise when compared with SaO 2ofthe whole blood of children immediately recoveringfrom cardiopulmonary bypass, a transient periodof poor peripheral perfusion, or in children withpersistent hypoxemia (ie, an SaO 2< 90%).Materials and MethodsA prospective observational study with consecutivesampling was proposed to the community institutionalreview board and approved. The study wasdeemed exempt from requiring informed consentby the community institutional review board.Two groups of children with congenital heart diseasewere studied: (1) children undergoing cardiopulmonarybypass–assisted cardiac surgery and(2) children with persistent hypoxemia. Twentyfivechildren with congenital heart disease undergoingsurgical repair (n = 13) or palliation (n = 12)using cardiopulmonary bypass, arterial catheter inplace, and a serum lactic acid ≥ 2 mmol/L (ie, evidenceof insufficient tissue perfusion) were recruitedbetween February 27, 2001, and August 6, 2001.Patients were excluded if pulse oximeter readingswithin 2% of each other prior to the start of cardiopulmonarybypass were unobtainable and if theserum lactic acid was < 2 mmol/L after cardiopulmonarybypass.Twenty-one children with cyanotic congenitalheart disease, an arterial catheter in place, and ameasured SaO 2consistently less than 90% on serialblood gases were recruited between August 7, 2001,and December 18, 2002. Four children wereenrolled on 2 separate occasions following distinctoperations. To avoid recruiting patients in a state ofpoor perfusion similar to the postcardiopulmonarybypass period, patients in this group were excludedif they displayed any clinical features of shock(eg, tachycardia, cool peripheral extremities, oliguria,serum lactic acid ≥ 2 mmol/L, etc).The pulse oximeters tested were the Masimo SETRadical (version 3, Masimo Corporation, Irvine,Calif) and the Nellcor N-395 (Nellcor PuritanBennett, Inc, Pleasanton, Calif). The whole-bloodco-oximeter used to measure SaO 2in the postcardiopulmonarybypass group was the ILSynthesis 1725 (Instrumentation Laboratories,Lexington, Mass). Because of an institutionalchange in laboratory equipment, the whole-bloodco-oximeter used to measure SaO 2in the cyanoticcongenital heart disease group was the IL 682(Instrumenta-tion Laboratories, Lexington, Mass).A pulse oximeter probe attached to each oximeterwas attached to the same distal extremity: oneon a digit and the other on the palm or sole. Probeswere not attached to extremities with arterialcatheters in place to avoid potential interferencefrom diminished arterial blood flow. Palliativeshunts were positioned centrally, that is, aorta topulmonary artery confluence, when placed in certainpatients (eg, hypoplastic left heart syndrome,unbalanced atrioventricular canal), making it feasibleto obtain an SpO 2measurement for comparisonfrom an extremity different from the extremity withthe arterial catheter. The probes were opticallyshielded. SpO 2displayed on each pulse oximeter atthe time routine arterial blood SaO 2measurementswere recorded. Cardiopulmonary monitor heartrate, pulse oximeter heart rate, and failure of pulseoximeter to display SpO 2at the time of the bloodgases measurement were also recorded for analysis.Core temperature, cardiopulmonary bypasstime, and aortic cross-clamp time were recordedfor analysis in the cardiopulmonary bypass group.None of the subjects in the cyanotic congenitalheart disease group were hypothermic (ie, coretemperature < 36°C) at the time of study measurements,and therefore, core temperature was notmeasured.Bias and precision, measures of agreementdescribed by Bland and Altman [14], were used tocompare SpO 2measurements by both oximeterswith the co-oximetry SaO 2measurements. Bias isdefined as the average of the differences betweenthe method being tested and the gold standard(SpO 2– SaO 2), and the precision is the standarddeviation of that difference. The expected range ofthat difference between the 2 measurements, thatis, the limits of agreement, are then calculated asthe bias ± 1.96 × SD. These data points were graphicallydisplayed as bias plots. The vertical axis representedthe SpO 2– SaO 2difference, and the horizontalaxis represented the average saturation[(SpO 2+ SaO 2)/ 2]. χ 2 test was used to detect a significantdifference in data readout failures betweenoximeters at the time of the SaO 2measurements inthe postcardiopulmonary bypass group. AWilcoxon signed ranks test was used to detect differencesin matched data (ie, pulse oximeter heartrate and cardiac monitor heart rate), and the Mann-Whitney U test was used to detect differences inunmatched data. Spearman rank correlation wasused to detect significant relationships between theabsolute SpO 2– SaO 2differences and patient characteristics(eg, hematocrit, core temperature, aortic230 Journal of Intensive Care Medicine 19(4); 2004

Pediatric Heart Pulse Oximetrycross-clamp time). Mean differences ± 1 standarddeviation were reported. A P value < .05 was consideredsignificant.ResultsTable 1 contains the characteristics of the 46patients enrolled. There were a total of 122 SaO 2measurements obtained, 61 measurementsobtained in each group. The median (range) SaO 2measurement was 83% (57%-100%). Seventy-sevenof the 122 (63%) SaO 2measurements obtainedwere less than 90%. The median serum lactic acidconcentration was 1.8 mmol/L (0.4-12.7). Themedian hematocrit was 40% (19%-54%).Postcardiopulmonary Bypass GroupThe Nellcor N-395 pulse oximeter failed to measurethe SpO 2during 41% (25/61) of the SaO 2measurementscompared to 10% (6/61) for the Masimo SETRadical pulse oximeter (P < .001). There was a significantdifference in bias ± precision but no significantdifference in heart rate between oximeters(see Table 2).There was a significantly negative relationshipbetween absolute SpO 2– SaO 2difference ofMasimo SET Radical and aortic cross-clamp time (r= –0.29, P = .03) and core temperature (r = –0.42,P = .001). Both pulse oximeters had significantlypositive relationships between absolute SpO 2–SaO 2difference and hematocrit (Nellcor N-395: r =0.46, P < .01; Masimo SET Radical: r = 0.35, P

Torres et alTable 2. Differences in Oximeter MeasurementsVariable (Mean Difference ± 1 SD) Nellcor Masimo P nPostcardiopulmonary bypass SpO 2– SaO 21.1 ± 3.3 –0.2 ± 4.1 < .001 33Oximeter heart rate – monitor heart rate 0.7 ± 6.9 –0.6 ± 1.9 .819 33Cyanotic congenital heart disease SpO 2– SaO 22.9 ± 4.6 2.8 ± 6.2 .848 61Oximeter heart rate – monitor heart rate –0.1 ± 2.0 –0.05 ± 1.8 .973 61202015151010SpO2 - SaO250-5-10-15*SpO2 - SaO250-5-10*-20-15-2560 65 70 75 80 85 90 95 100(SpO 2+SaO 2)/2 (%)Fig 1. Bias plot of Masimo SET Radical pulse oximeter (n= 55). Bold lines represent average SpO 2– SaO 2difference(bias, d), and the dotted lines represent precision(d ± 2 SD). *SpO 2– SaO 2= 0 at 100%, n = 14.-2060 65 70 75 80 85 90 95 100(SpO 2 + SaO 2)/2 (%)Fig 2. Bias plot of Nellcor N-395 pulse oximeter (n = 36).Bold lines represent average SpO 2– SaO 2difference(bias, d), and the dotted lines represent precision (d ± 2SD). *SpO 2– SaO 2= 0 at 100%, n = 8.conditions. Masimo’s Signal Extraction Technology(SET) applies multiple algorithms to filter the pulsesignal and to extract the arterial from the venoussignal and report the true SpO 2and pulse rate duringmotion and low perfusion conditions [13]. Wecompared these 2 technologically advanced pulseoximeters to the gold standard, measured SaO 2, toevaluate their performance under 2 conditionsprone to inaccuracies: low perfusion and hypoxemia.Children who have undergone cardiopulmonarybypass with subsequent lactic acidosis areconsidered comparable to children with a pathologicallycaused low perfusion condition [15]. Themajor difference between the Nellcor N-395 andthe Masimo SET was the proportion of readout failuresin the postbypass group (41% vs 10%, respectively).A study comparing percentage of nonfunctionalmonitoring time after coronary artery bypasssurgery in 86 adults revealed that the Masimo SEToutperformed the Ohmeda 3740 pulse oximeter(1.2% ± 3.3% for Masimo SET vs 8.7% ± 16.4% forOhmeda 3740; P < .001) [16]. Iyer and colleagues[17] demonstrated that the drop in peripheral skintemperature that occurs during cardiac surgery withcardiopulmonary bypass caused unreliable pulseoximeter readings. However, 2 other studies ofNellcor pulse oximeters during cardiac surgerydemonstrated no relationship of peripheral skintemperature to accuracy or failure [18,19]. We diddemonstrate a negative correlation between thecore temperature and the absolute SpO 2– SaO 2differencefor the Masimo SET Radical but not for theNellcor N-395. The high failure rate for the NellcorN-395 during low perfusion conditions would likelyresult in a greater number of arterial blood gasesbeing performed to ensure adequate SaO 2.Another significant finding of our study was theincrease in bias and precision, that is, loss of accuracyand increase in scatter, for both oximeterswhen SaO 2was < 90%. Both manufacturers assertthat their pulse oximeter is “accurate” even at a lowSaO 2of 70% [12,13]. Bell and colleagues [9] reporteda significant increase in precision for 3 oximeters(Ohmeda 3700, Novametrix 520A, and theNellcor N-200) when the SaO 2was < 90% in childrenundergoing major operations including cardiacsurgery. Jubran and Tobin [10] found a significantincrease in bias and precision for a Nellcorbrand pulse oximeter and an Ohmeda 3700 pulseoximeter when the SaO 2was ≤ 90% compared toan SaO 2> 90% (5.1% ± 2.7% vs 1.7% ± 1.2%; P

Pediatric Heart Pulse Oximetryprecision with an SaO 2< 80% [5], and the otherreported no difference with an SaO 2of < 80% [20].The increase in scatter on the bias plots appears tooccur at (SpO 2+ SaO 2)/2 < 90% for the MasimoRadical SET compared to (SpO 2+ SaO 2)/2 < 80% forthe Nellcor N-395. Algorithms used by the manufacturersprobably do not contain adequate numbersof patients with hypoxemia, which mightexplain these findings. The apparent increase inbias and precision when SaO 2was < 90% in thepostcardiopulmonary group versus the persistenthypoxemia group may have been secondary to thecombination of low perfusion and hypoxemia inthe postcardiopulmonary group or the different cooximetermethodology used in each group (lightscatter [IL Synthesis 1725] vs spectrophotometry [IL682]). Published data comparing the 2 technologieswere lacking. Nevertheless, the results in bothgroups confirmed diminished pulse oximeter performancewhen SaO 2< 90%.It has been well documented that different pulseoximeters do not perform similarly when testedunder other adverse conditions [2-11]. For example,Van de Louw and colleagues [21] reported a statisticallysignificant (but not clinically significant) differencein bias for 3 pulse oximeters (Nellcor N-200, Ohmeda 3700, and the Hewlett PackardViridia 24C) when vasoactive agents were beingadministered to children admitted to a pediatricintensive care unit. Other studies of older generationsof pulse oximeters still widely used havereported variable performances when tested underdifferent, low perfusion conditions [15,22]. Macnaband colleagues [15] reported a bias < 2% for theBiox 3700 in 58 children recovering from deephypothermia for cardiac surgery regardless of coretemperature, while Ibáñez and colleagues [22]reported an absolute bias of > 4% for the Biox 3700pulse oximeter in 9 of 24 (37%) adult patientsreceiving vasoactive agents.The absolute bias and precision for both oximeterscorrelated best with the patients’ hematocrit inthe cardiopulmonary bypass group but not in thepersistent hypoxemia group. One study examiningthe accuracy of the Nellcor N-100 pulse oximeter inchildren with cyanotic congenital heart disease didnot demonstrate a correlation between high hematocritconcentrations and the oximeter’s bias andprecision [5]. Hematocrit of the whole blood mayhave a greater impact on the rheology of the redblood cells under low perfusion conditions andtherefore alter the reflective characteristics of capillaryblood. The lack of a correlation between pulsepressure and the absolute bias and precision aresimilar to the findings of Villanueva and colleagues[3], who found only a weak correlation betweenpulse pressure and the bias of the Nellcor N-200pulse oximeter.One limitation of our study was the lack ofperipheral skin temperature monitoring. Studieshave reported conflicting data regarding the relationshipbetween peripheral skin temperature andpulse oximeter performance as mentioned previously.Barker and colleagues demonstrated thatsensor malpositioning during pulse oximetry testingunder conditions of hypoxemia significantlyaltered accuracy [23]. Although sensor positioningcould not be checked during the study in the cardiopulmonarybypass group, the patients were allneuromuscularly blocked during the testing, makingit unlikely that patient motion caused sensormalpositioning. Probes were examined for sensorpositioning after the measurements were made,and none of the probes appeared displaced. Andfinally, probe position may have influenced theresults of our study. No data were available regardingdifferences in accuracy based on probe position(eg, toe vs sole of the foot). Since no attemptswere made to randomly place the different probesin the different positions, it is possible that probeposition may have influenced our results.ConclusionsThe Masimo Radical SET and Nellcor N-395 pulseoximeters have an acceptable accuracy but unacceptableprecision in children with congenital heartdisease. The Nellcor N-395 pulse oximeter failedsignificantly more often than the Masimo SETRadical pulse oximeter did in children with lacticacidosis recovering from cardiopulmonary bypass.Performance of both oximeters worsened whenSaO 2was < 90% compared to SaO 2of ≥ 90%.Clinicians should refrain from altering managementof the child with cyanotic congenital heart diseasebased solely on pulse oximetry readings with thepulse oximeters studied.AcknowledgmentsThis study was supported in part by theDepartment of Pediatrics, University of IllinoisCollege of Medicine at Peoria. We are grateful tothe cardiovascular surgery and pediatric intensivecare unit staff for their patience and are especiallygrateful to the perfusionists for their unselfish assistance.We are also grateful to the MasimoJournal of Intensive Care Medicine 19(4); 2004 233

Torres et alCorporation for the unreimbursed supply ofMasimo SET pulse oximeter probes. This study waspresented in part at the Society of Critical CareMedicine’s 32nd Society of Critical Care MedicineCongress, San Diego, California, January 28, 2002,and at the 31st Society of Critical Care MedicineCongress, San Antonio, Texas, January 30, 2003.References1. Tremper KK. Pulse oximetry’s final frontier. Crit Care Med.2000;28:1684-1685.2. Trivedi NS, Ghouri AF, Shah NK, Lai E, Barker SJ. Effects ofmotion, ambient light, and hypoperfusion on pulse oximeterfunction. J Clin Anesth. 1997;9:179-183.3. Villanueva R, Bell C, Kain ZN, Colingo KA. Effect ofperipheral perfusion on accuracy of pulse oximetry in children.J Clin Anesth. 1999;11:317-322.4. Carter BG, Carlin JB, Tibballs J, Mead H, Hochmann M.Accuracy of two pulse oximeters at low arterial hemoglobinoxygensaturation. Crit Care Med. 1998;26:1128-1133.5. Schmitt HJ, Schuetz WH, Proeschel, Jaklin C. Accuracy ofpulse oximetry in children with cyanotic congenital heartdisease. J Cardiothorac Vasc Anesth. 1993;7:61-65.6. Thrush D, Hodges MR. Accuracy of pulse oximetry duringhypoxemia. South Med J. 1994;87:518-521.7. Severinghaus JW, Spellman MJ. Pulse oximeter failurethresholds in hypotension and vasoconstriction.Anesthesiology. 1990;73:532-537.8. Morris RW, Nairn M, Torda TA. A comparison of fifteenpulse oximeters. Part I: a clinical comparison; part II: a testof performance under conditions of poor perfusion.Anaesth Intensive Care. 1989;17:62-82.9. Bell C, Luther MA, Nicholson JJ, Fox CJ, Hirsch JL. Effect ofprobe design on accuracy and reliability of pulse oximetryin pediatric patients. J Clin Anesth. 1999;11:323-327.10. Jubran A, Tobin MJ. Reliability of pulse oximetry in titratingsupplemental oxygen therapy in ventilator-dependentpatients. Chest. 1990;97:1420-1425.11. Bohnhorst B, Peter CS, Poets CF. Pulse oximeters’ reliabilityin detecting hypoxemia and bradycardia: comparisonbetween a conventional and two new generation oximeters.Crit Care Med. 2000;28:1565-1568.12. Nellcor Puritan Bennet, Inc. Nellcor N-395 pulse oximeterproduct description. Available at: http://www.nellcor.com/products/index.asp. Accessed January 15, 2003.13. Masimo Corporation. Radical signal extraction pulseoximeter sales sheet. Available at: http://www.masimo.com/pusleox/radical.htm. Accessed January 15, 2003.14. Bland JM, Altman DG. Statistical methods for assessingagreement between two methods of clinical measurement.Lancet. 1986;1(8476):307-310.15. Macnab AJ, Baker-Brown G, Anderson EE. Oximetry inchildren recovering from deep hypothermia for cardiacsurgery. Crit Care Med. 1990;18:1066-1069.16. Durbin CJ, Rostow SK. More reliable oximetry reduces thefrequency of arterial blood gas analyses and hastens oxygenweaning after cardiac surgery: a prospective randomizedtrial of the clinical impact of a new technology. CritCare Med. 2002;30:1735-1740.17. Iyer P, McDougall P, Loughnan P, et al. Accuracy of pulseoximetry in hypothermic neonates and infants undergoingcardiac surgery. Crit Care Med. 1996;24:507-511.18. Pälve H, Vuori A. Minimum pulse pressure and peripheraltemperature needed for pulse oximetry during cardiac surgerywith cardiopulmonary bypass. J Cardiothorac VascAnesth. 1991;5:327-330.19. Carter BG, Wiwczaruk D, Hochmann M, Henning R.Performance of transcutaneous PCO 2and pulse oximetrymonitors in newborns and infants after cardiac surgery.Anaesth Intensive Care. 2001;29:260-265.20. Boxer RA, Gottesfeld I, Singh S, et al. Noninvasive pulseoximetry in children with cyanotic congenital heart disease.Crit Care Med. 1987;15:1062-1064.21. Van de Louw A, Cracco C, Cerf C, et al. Accuracy of pulseoximetry in the intensive care unit. Intensive Care Med.2001;27:1606-1613.22. Ibañez J, Velasco J, Raurich JM. The accuracy of the Biox3700 pulse oximeter in patients receiving vasoactive therapy.Intensive Care Med. 1991;17:484-486.23. Barker SJ, Hyatt J, Shah NK, Kao J. The effect of sensormalpositioning on pulse oximeter accuracy during hypoxemia.Anesthesiology. 1993;79:248-254.234 Journal of Intensive Care Medicine 19(4); 2004