clinical performance of an in-line, ex vivo point- of-care
TRANSCRIPT
Clinical Performance of an In-Line, ex Vivo Point-of-Care Monitor: A Multicenter Study
Glenn F. Billman,1 Amy B. Hughes,2 Golde G. Dudell,3 Elizabeth Waldman,1
Lisa M. Adcock,4 Dan M. Hall,5 Edmund N. Orsini, Jr.,6 Adolph J. Koska,7
Linda J. Van Marter,8 Neil N. Finer,9 Jeff C. Kulhavy,10 Ronald D. Feld,10 andJohn A. Widness2*
Background: The management of critically ill infantsand neonates includes frequent determination of arte-rial blood gas, electrolyte, and hematocrit values. Anobjective of attached point-of-care patient monitoring isto provide clinically relevant data without the adverseconsequences associated with serial phlebotomy.Methods: We prospectively determined the mean dif-ference (and SD of the difference) from laboratorymethods of an in-line, ex vivo monitor, the VIA LVMBlood Gas and Chemistry Monitoring System® (VIALVM Monitor; Metracor Technologies, Inc.), in 100 crit-ically ill neonates and infants at seven children’s hos-pitals. In doing so, we examined monitor stability withcontinuous use. In vivo patient test results from labora-tory benchtop analyzers were compared with those fromthe VIA LVM Monitor on paired samples. In a separatein vitro comparison, benchtop analyzer and monitor testresults were compared on whole-blood split samples.
Results: A total of 1414 concurrent, paired-sample mea-surements were obtained. The mean differences (SD ofdifferences) from laboratory methods and r values forthe combined data for the VIA LVM Monitor from theseven sites were 0.001 (0.026) and 0.97 for pH, 0.7 (3.6)mmHg and 0.94 for PCO2, 4.2 (9.6) mmHg and 0.98 forPO2, 0.0 (2.9) mmol/L and 0.87 for sodium, 0.1 (0.2)mmol/L and 0.96 for potassium, and 0.3% (2.9%) and 0.90for hematocrit. Performance results were similar amongthe study sites with increasing time of monitor use andbetween in vivo paired-sample and in vitro split-sampletest results.Conclusion: The VIA LVM Monitor can be used toassess critically ill neonates and infants.© 2002 American Association for Clinical Chemistry
Technologic innovations in the development and fabrica-tion of biosensors and microprocessors have led to thedevelopment and patient bedside use of small, highlyaccurate point-of-care (POC)11 devices. These devicesinclude “analyzers”, which require permanent removal ofblood specimens from patients, and “monitors”, which donot. POC devices have the potential for promoting im-proved patient outcomes through their ability to shortenthe therapeutic turnaround time of diagnostic tests and toincrease accuracy by reducing or eliminating preanalyticerror and specimen preparation (1, 2). Because monitorsare designed to operate as closed systems with insignifi-cant blood loss, the need for erythrocyte transfusions toneonates and the risk of nosocomial infections may alsobe reduced. These advantages have motivated the devel-opment and introduction of continuous and near-contin-uous patient-attached POC monitors.
Despite numerous reports describing the clinical per-
1 Department of Pathology, Children’s Hospital of San Diego, San Diego,CA 92123.
2 Department of Pediatrics, The Children’s Hospital of Iowa, Iowa City, IA52242.
3 Department of Pediatrics, Division of Neonatology, Children’s Hospitalof San Diego, San Diego, CA 92123.
4 Texas Children’s Hospital, Baylor College of Medicine, Houston, TX77030.
Departments of 5 Pediatrics and 6 Pathology, The Children’s Hospital,Denver, CO 80218.
7 Department of Anesthesia, Driscoll Children’s Hospital, Corpus Christi,TX 78411.
8 Department of Pediatrics, Division of Newborn Medicine, Children’sHospital, Boston, MA 02115.
9 Department of Pediatrics, Division of Neonatology, University of Cali-fornia San Diego Medical Center, San Diego, CA 92103.
10 Department of Pathology, The University of Iowa, Iowa City, IA 52242.*Address correspondence to this author at: University of Iowa Hospitals &
Clinics, 200 Hawkins Dr., W222-1 GH, Iowa City, IA 52242-1083. Fax 319-356-4685; e-mail [email protected].
Received in revised form June 20, 2002; accepted August 21, 2002.
11 Nonstandard abbreviations: POC, point-of-care; Hct, hematocrit; andNICU, neonatal intensive care unit.
Clinical Chemistry 48:112030–2043 (2002) Pediatric Clinical
Chemistry
2030
Dow
nloaded from https://academ
ic.oup.com/clinchem
/article/48/11/2030/5642240 by guest on 15 Novem
ber 2021
formance characteristics of near-patient POC devices, fewstudies have included infants and children (3–7), andnone of the reports on infants and children have includedsufficient numbers of individuals drawn from multiplesettings to allow generalization about the bedside perfor-mance of POC monitoring in this population. In thisstudy, we prospectively evaluated the performance of arecently introduced in-line, ex vivo monitor specificallydesigned for use in critically ill neonates, infants, andother volume-restricted patients. This monitor, the VIALow Volume Mode Blood Gas and Chemistry MonitoringSystem® (VIA LVM Monitor; Metracor Technologies, Inc.,San Diego, CA), measures pH, Po2, Pco2, Na�, K�, andhematocrit (Hct) by automatically drawing blood from apatient’s arterial catheter, analyzing it, and reinfusing theblood sample back into the patient. The monitor’s fluidcircuitry design and its proprietary stopcock preventexcessive amounts of sterile calibration solution frombeing infused into the patient.
In the present study, we hypothesized that acceptableVIA LVM Monitor comparative mean differences, SD ofdifferences, stability, and trouble-free use would be ob-served in the context of a prospective in vivo multicenterclinical trial in which different makes and models ofinstruments in use at different study sites functioned ascomparison analyzers and in a single-site in vitro split-sample comparison of the VIA LVM Monitor with a singlecomparison analyzer.
Materials and Methodsrole of sponsorThe present study was investigator-initiated. Althoughthe authors had complete responsibility for study design,data analysis, and reporting, the sponsor was providedwith the opportunity to provide comments and criticismsfor all of these. The authors were free to publish andcomment on any and all data without restriction by thesponsor.
patient selectionInstitutional Review Board approval was a prerequisitefor study site participation. Informed written consent wasobtained from one or both parents. Study enrollment waslimited to 100 neonates or infants younger than 8 monthsof age with umbilical or peripheral arterial cathetersinserted for clinical indications and an anticipated needfor at least 10 blood gas determinations during the sub-sequent 72 h. Admission into the trial was offered irre-spective of whether a patient was receiving medical orsurgical treatment. Similarly, the composition of a pa-tient’s primary arterial line fluid was not restricted, butinstead varied according to site preference. Monitors werediscontinued if the arterial catheter was no longer re-quired for patient care, persistent waveform dampeningoccurred, or the VIA LVM Monitor or its sensor failed.
instrument comparisonConsecutive, clinically ordered tests for blood gases, Na�,K�, and Hct were analyzed by both a central laboratoryinstrument and the in-line monitor. The study did notrestrict which laboratory instruments were used or theirlocation within the hospital. Similarly, the method ofspecimen handling and delivery to the reference labora-tory occurred in each facility’s usual fashion.
The VIA LVM Monitor evaluated in the present studyis a Food and Drug Administration-approved, patient-attached, closed-system monitor designed for use in crit-ically ill neonates, infants, and other volume-restrictedpatients. Operation of the VIA LVM Monitor was per-formed as recommended by the manufacturer. Like themonitor’s adult predecessor (8 ), the in-line neonatal/infant monitor uses conventional electrochemical detec-tion methods for measuring pH, Po2, Pco2, K�, Na�, andHct. Specifically, ion-selective electrodes are used to mea-sure pH, Pco2, K�, and Na�. A Clark Po2 electrode is usedto measure Po2. Hct values are determined indirectly bymeasuring the electrical conductance. The six sensors areorganized in an array that is entirely contained in afluid-filled, thermostatically controlled flow cell. All mea-surements are performed at 37 °C.
Before each patient attachment, the VIA LVM Monitorand sensor undergo an in vitro two-point calibration.Subsequently, one-point calibrations are automaticallyperformed every 30 min (or more frequently if directed bythe operator) using a sterile, isotonic, heparinized calibra-tion solution continuously bathing the sensor array. Thissolution is composed of 500 mL of Normosol R (AbbottLaboratories), 10 mL of 84 g/L NaHCO3 (Abbott Labora-tories), and 5 mL of sodium heparin (100 units/mL). Theaddition of heparin was optional, depending on eachsite’s preference. Fluid circuitry design restricts the vol-ume of fluid administered to the patient before, during,and after each measurement cycle and permits the samecatheter to be used clinically for pressure waveformmonitoring and fluid administration.
When the instrument operator triggers a sample anal-ysis, the VIA LVM Monitor aspirates �1.5 mL of arterialblood through microbore tubing into the sensor arrayover �1 min. The volume withdrawn is sufficient topresent an undiluted blood sample to the sensor array.Specimen equilibration and analysis take an additional70 s during which the blood sample is warmed to 37 °C.As soon as the analysis phase is completed, the results aredisplayed on the monitor screen. After analysis, all but�25 �L of blood is automatically returned to the patientalong with a 0.5-mL flush of calibration solution fluid (7 ).When not in sample mode, the sensor array is cleansed bycalibration fluid set at a flow rate of 5 mL/h and divertedinto a collection bag. The minimum interval betweenblood sampling for the monitor is �5 min. To ensure theaccuracy of measurements, fresh calibration fluid is pre-pared each day of use. In accordance with hospital-
Clinical Chemistry 48, No. 11, 2002 2031D
ownloaded from
https://academic.oup.com
/clinchem/article/48/11/2030/5642240 by guest on 15 N
ovember 2021
dictated infection control guidelines, sensors are replacedevery 72 h.
in vivo paired-sample testingThe VIA LVM Monitor and the comparison laboratoryinstrument analyzed consecutive, physician-orderedblood samples. The samples submitted to the laboratorywere drawn from a stopcock located between the patientand the in-line sensor array. This was done during themonitor’s 70-s analysis period. The transportation andanalysis of the paired sample for laboratory analysisoccurred according to the usual practice of each site.Duration of sensor use for individual study participantswas defined as the time interval between when the firstand the last blood samples were drawn.
in vitro split-sample testingFresh (�8 h old) fetal umbilical cord and adult bloodsamples were analyzed in duplicate at one study site by asplit-sample method using two analytic instruments(ABL625; Radiometer) and two in-line monitors. To en-compass the range of values to include those encounteredin the clinical trial, individual blood samples were mod-ified by gas tonometry, hemoconcentration, hemodilu-tion, or addition of electrolytes. Blood samples (5 mL)were tonometered for 5–10 min with an IL 237 Tonometer(Instrumentation Laboratory, Inc.), drawn into a syringe,and analyzed within 3 min in random sequence by each ofthe four measuring instruments.
data inclusion/exclusion criteriaVIA LVM Monitor data points were compared with thoseobtained from reference laboratory instruments if (a) themost recent one-point calibrations were within limits forboth the monitor and analyzer, (b) the monitor andlaboratory instruments both successfully performed therequested analyses, and (c) the laboratory specimen wasfree of air contamination or clots. In addition, the data setsfor each of the six analytes were screened for outliers byuse of NCCLS outlier exclusion criteria (9 ). With thisprocedure, outliers are identified as data points in whichthe difference between the two methods exceeds fourtimes the absolute mean difference for all of the datapoints. Up to 2.5% of the data points for each analyte maybe excluded. For split-sample, in vitro paired compari-sons, comparison instrument values falling within 3 SD ofthe paired-sample clinical data were included.
statistical analysis and data handlingDeming regression was used to assess the agreementbetween the results obtained with the comparison meth-ods and the VIA LVM Monitor for all of the analytesexcept Hct (for which quality-control calibrators are un-available). For Hct, Pearson correlation and simple linearregression were used to assess agreement between thetwo methods. Error analysis was performed by themethod of Bland and Altman (10 ). The mean difference of
the results of the two assay methods and SD of thedifferences between the two methods were calculatedafter the exclusion of NCCLS outliers. Statistical data arepresented as the mean � SD. Only two-tailed testing wasdone. P �0.05 was considered statistically significant.
Resultsin vivo paired-sample studyPatient and site characteristics. One hundred critically illnewborns and infants with umbilical or peripheral arterialcatheters already in use were enrolled at the seven clinicalsites (Table 1). Infant weight at enrollment ranged from3100 to 5500 g. The duration of VIA LVM Monitor use inindividual patients differed significantly by site (P�0.0001), with usage time ranging from 4 to 72 h with amean of 49 h. The different sites also varied with respectto the composition of the parenteral solutions infusedthrough the arterial line, the location of the referencelaboratory in relation to the intensive care unit, themethod for getting the sample to the reference laboratory(e.g., by hand or by pneumatic tube), and the preanalyticperiod of the reference sample. Four sites used the ChironModel 865 (Bayer, Inc.); one site used Chiron Models 855,860, and 865; and two used the Radiometer ModelABL625. During the course of the study, no patientcomplications related to the use of the VIA LVM Monitorwere reported, and no events necessitating monitor dis-continuation occurred.
Clinical study data characteristics. From the 100 studyparticipants there were 1433 paired-sample results avail-able for comparison. Of these, 19 were excluded fortechnical reasons, including microbubbles in the sensorarray (n � 8), parenteral fluid contamination (n � 7), andimproper collection (n � 4). Among the remaining 1414paired data sets, differences were observed in the numberof individual paired analyte measurements. These differ-ences were the result of analysis of blood samples forselected analytes at some study sites rather than thecomplete panel of six analytes assessed by the monitor.
The application of NCCLS outlier exclusion criteria ledto the exclusion of a small number (and percentage) of thetotal number of paired samples analyzed (9 ). Theseincluded 20 for pH (1.4%), 3 for Pco2 (0.2%), 9 for Po2
(0.7%), 7 for Na� (0.6%), 14 for K� (1.3%), and 7 for Hct(0.7%). Only 2 of the 1414 paired-sample data sets hadmore than one analyte that was a NCCLS outlier.
The ranges of values encountered for the six laboratoryvariables were as follows: 6.94–7.74 for pH; 17–170mmHg for Pco2; 22–215 mmHg for Po2; 121–158 mmol/Lfor Na�; 2.0–8.4 mmol/L for K�; and 27.5–64.1% for Hct.For the paired Po2 data sets, values for the laboratoryinstruments that exceeded 217 mmHg (i.e., �3 SD abovethe mean; n � 25) were excluded from analyses. Theexclusion of values above this threshold was done toavoid the potential of confounding preanalytic errorcaused by gas diffusion and aspirated air. This Po2
2032 Billman et al.: Performance of an In-Line Point-of-Care MonitorD
ownloaded from
https://academic.oup.com
/clinchem/article/48/11/2030/5642240 by guest on 15 N
ovember 2021
threshold also delimits an upper boundary based on lackof clinical utility of values above this limit.
Clinical performance correlation, mean difference, and SD ofdifferences. All six study analytes showed highly signifi-cant agreement between the results obtained with thelaboratory instrument and the VIA LVM Monitor (Fig. 1).The correlation coefficients (r) for five of the six analyteswere �0.90, whereas the correlation coefficient for Na�
was 0.87 (P �0.0001 for all comparisons). The Bland–Altman plots of all six study analytes had mean differ-ences that approximated zero (Fig. 2).
The vertical scatter of the paired patient data samplesabout the mean differences in the Bland–Altman plotsshowed that the majority of data points fell within CLIAcriteria limits used for proficiency testing accreditation ofclinical laboratories (11 ).
secondary outcomesComparison of results among study sites. A key objective inconducting this multicenter trial was to assess how mon-itor performance was impacted by the differences inpatient care that distinguish one site from another. Ac-cordingly, the seven participating sites were asked toincorporate the VIA LVM Monitor into their existingpractices and processes. Standardization across the siteswas limited to the manufacturer’s training syllabus andthe data collection protocol. It was anticipated that resultsfrom each site would reflect the influence of factors andvariables present at that site.
Comparison of the monitor’s mean difference and SDof differences across the seven participating sites demon-strated close agreement (Table 2). Site F had the leastoperational experience with the monitor and recordedone-third to one-fifth as many paired samples as the othersites. Study sites E and F had too few Na�, K�, and Hctvalues for meaningful comparison and thus were ex-cluded from analysis of individual study sites. Most of theperformance variability noted among sites was evident assubtle differences in the mean difference. Site C had thegreatest mean pH difference offset, i.e., 0.024. For Pco2,sites A, B, and C had mean differences that were largerthan those of the other sites. Except for sites D and E,which showed a slightly lower mean Po2 difference, meanPo2 differences were similar among the other sites. ForNa�, study site G showed a greater mean differencerelative to the other sites. For K�, minimal variability inmean differences and the SD of differences were observedat all sites. For Hct, although site A showed a negativemean difference relative to the other sites, the Hct SDs ofthese differences were comparable among all sites.
Influence of duration of sensor use on results. Across the 72-hperiod of study, the number of patients, their medicalconditions, the instrument operators, and the arterial lineinfusates all differed. Despite these differences, variabilityin the mean difference and the SD of the differences
Tabl
e1.
Pat
ient
,ca
thet
er,
trea
tmen
t,an
dbe
ncht
opan
alyz
erin
form
atio
n.
Sit
eP
atie
nts,
n
Sex
Bir
thw
eigh
t,a
gA
geat
entr
y,a
days
Art
eria
lca
thet
erTy
peof
trea
tmen
tP
rean
alyt
icpe
riod
ofla
bora
tory
anal
yzer
,a
min
Com
posi
tion
ofpa
rent
eral
fluid
give
nby
arte
rial
cath
eter
Loca
tion
ofla
bora
tory
anal
yzer
MF
Per
iphe
ral
Um
bilic
alM
edic
alS
urgi
cal
A3
01
61
43
05
7�
80
154
�68
24
68
22
13.6
�6.2
9g/
LN
aCl
Mai
nho
spita
lla
bora
tory
B2
11
01
12
17
9�
14
34
7�
19
813
21
02.6
�2.6
9g/
LN
aCl
Sat
ellit
ein
NIC
UC
20
16
42
02
6�
13
05
2.2
�2.6
020
20
03.9
�2.8
9g/
LN
aCl
Sat
ellit
ene
xtto
NIC
UD
16
12
42
74
3�
10
46
2.8
�2.8
016
16
01.7
�2.0
TPN
,bde
xtro
se,
elec
trol
ytes
Sat
ellit
ene
xtto
NIC
UE
64
22
53
2�
86
15.7
�7.7
06
60
7.2
�5.6
9g/
LN
aCl
Mai
nho
spita
lla
bora
tory
F4
13
21
47
�5
04
2.3
�1.2
04
40
4.0
�3.2
TPN
,de
xtro
se,
elec
trol
ytes
Mai
nho
spita
lla
bora
tory
G3
21
28
38
�1
19
855
�70
30
03
3.7
�3.6
9g/
LN
aCl
Sat
ellit
ene
xtto
PIC
UAl
lsite
s1
00
61
39
25
19
�1
17
421
�46
35
65
75
25
6.7
�6.7
NA
NA
aM
ean
�S
D.
bTP
N,
tota
lpar
ente
raln
utrit
ion;
PIC
U,
pedi
atric
inte
nsiv
eca
reun
it;N
A,no
tav
aila
ble.
Clinical Chemistry 48, No. 11, 2002 2033D
ownloaded from
https://academic.oup.com
/clinchem/article/48/11/2030/5642240 by guest on 15 N
ovember 2021
among the six analytes throughout the entire study periodwere modest and thus unlikely to impact clinical decision-making (Fig. 3). Approximately one-fourth of the patientsenrolled in the study completed 72 h of continuousmonitoring with the VIA LVM Monitor. Results of paired
samples generated at startup and after each cycle ofcalibration and/or patient monitoring included the timeand date of the analysis. The monitor printouts were usedto determine the corresponding duration of in-line sensorarray use for each patient. To test for sensor drift, paired
Fig. 1. Correlation of laboratory instrument results with those of the VIA LVM Monitor for in vivo patient paired samples.The dashed line in each panel indicates the line of identity. The solid line represents the regression line for the range of values encountered. The equation for theregression line, the r value, the SD of the residuals (Sy�x), and the number of data points for each analyte are as follows: for pH (A), y � 0.917x � 0.62 (r � 0.97;Sy�x � 0.029; n � 1340); for PCO2 (B), y � 1.04x � 2.4 mmHg (r � 0.94; Sy�x � 3.62 mmHg; n � 1279); for PO2 (C), y � 0.922x � 2.72 mmHg (r � 0.98; Sy�x �8.75 mmHg; n � 1247); for Na� (D), y � 0.986x � 2.0 mmol/L (r � 0.87; Sy�x � 2.9 mmol/L; n � 1014); for K� (E), y � 0.901x � 0.44 mmol/L (r � 0.96; Sy�x� 0.22 mmol/L; n � 1067); for Hct (F), y � 0.914x � 3.55% (r � 0.90; Sy�x, not applicable; n � 1072).
2034 Billman et al.: Performance of an In-Line Point-of-Care MonitorD
ownloaded from
https://academic.oup.com
/clinchem/article/48/11/2030/5642240 by guest on 15 N
ovember 2021
data sets were sorted according to the duration of in-linesensor array use, and monitor mean difference and SD ofdifferences were determined. With the exception of Pco2,the data demonstrated no clinically significant sensorinstability throughout the 72-h period of investigation. Aslight increase in the SD of differences, but not in themean difference was noted for Pco2 determinations be-ginning at 48–60 h of sensor use.
in vitro split-sample studyIn vitro split-sample comparisons were performed atstudy site D. One hundred four blood samples wereanalyzed for each of the six study analytes by fourinstruments: the two in-line monitors and the two labo-ratory instruments.
NCCLS guidelines were applied to the 104 split-samplein vitro data sets to identify outliers. This procedure
Fig. 2. Differences between the results obtained with the VIA LVM Monitor and the comparison methods for the in vivo patient paired samples.The thick solid horizontal line indicates the mean difference; the two horizontal dashed lines indicate the CLIA performance criteria limits referenced to the meandifference. The two vertical lines indicate the critical outlier values established for laboratory results for pediatric patients (20). The mean difference and the SD of themean differences are indicated for each analyte. The x-axis tick marks are along the horizontal line emanating from point 0 on the y axis.
Clinical Chemistry 48, No. 11, 2002 2035D
ownloaded from
https://academic.oup.com
/clinchem/article/48/11/2030/5642240 by guest on 15 N
ovember 2021
identified 5 outliers for pH (n � 90), 6 for Pco2 (n � 87),2 for Po2 (n � 91), 5 for Hct (n � 87), 4 for Na� (n � 89),and 6 for K� (n � 88). After the range of values wasrestricted to � 3 SD of the mean in vivo paired-sampledata, the final numbers of in vitro samples used in themean difference and SD of differences calculations wereas follows: pH (n � 59), Pco2 (n � 61), Po2 (n � 102), Na�
(n � 69), K� (n � 44), and Hct (n � 89). The resulting invitro split-sample data ranges were as follows: 7.128–7.728 for pH, 10.8–72.7 mmHg for Pco2, 11–217 mmHg forPo2, 122–155 mmol/L for Na�, 1.60–5.69 mmol/L for K�,and 24.2–65.1% for Hct.
in vitro performance resultsAll six study analytes showed highly significant agree-ment between results obtained with the comparison in-strument and the VIA LVM Monitor (Fig. 4). Correlationcoefficients for five of the six analytes were �0.98,whereas the r value for Hct was 0.95 (P �0.0001 for allcomparisons). Bland–Altman plots of five of the six studyanalytes revealed mean differences that approximatedzero (Fig. 5). The mean difference for Pco2 was 3.46mmHg. The vertical scatter of the paired patient datasamples about the mean differences in the Bland–Altmanplots showed that the majority of data points fell withinCLIA criteria limits (11 ). No clinically significant differ-
ences emerged between adult blood samples and fetalcord blood samples for mean difference or SD of differ-ences.
comparison of in vivo and in vitro resultsFor pH, Na�, K�, and Hct, the multicenter mean differ-ence and SD of differences were consistent between theaggregate in vivo paired-sample clinical data and therange-limited in vitro split-sample results (Table 2). ForPco2, the mean difference identified by in vitro testingwas substantially larger than that of the composite in vivomean difference (3.46 vs 0.68 mmHg). Furthermore, forPco2, the mean difference for in vitro testing was largerthan the individual in vivo mean Pco2 difference at studysite D, the site where the in vitro test was performed (i.e.,3.46 vs 0.01 mmHg). The Po2 mean difference and SD ofthe Po2 differences predicted by in vitro testing at site Dwere comparable to the observed in vivo mean differenceand SD of differences data measured at this same site(�0.15 and 4.61 mmHg vs �0.20 and 6.7 mmHg, respec-tively), but were better than the aggregated Po2 meandifference and SD of difference data for the aggregate ofall seven sites (4.17 and 9.58 mmHg, respectively).
The r values derived from in vitro split-sample testingfor all six of the analytes were equal to or greater than ther values measured from in vivo paired-sample testing.
Table 2. Mean difference and SD of differences by study site for paired-sample in vivo and split-sample in vitro testing.
Analyte
In vivo paired sample by study site
In vitro splitsampleaSite A Site B Site C Site D Site E Site F Site G
All 7sites
pHMean difference �0.01 0.00 0.02 0.00 �0.01 0.01 0.00 0.00 0.00SD of differences 0.03 0.02 0.02 0.02 0.02 0.02 0.02 0.03 0.03n 417 204 255 291 75 26 72 1340 59
PCO2, mmHgMean difference 2.54 2.20 �1.75 0.01 0.17 �0.06 �0.95 0.68 3.46SD of differences 3.24 3.27 3.40 3.09 2.92 4.50 2.06 3.62 3.90n 364 203 255 297 73 25 62 1279 61
PO2, mmHgMean difference 6.32 3.65 6.18 �0.20 2.14 6.28 5.96 4.17 �0.15SD of differences 10.96 10.26 8.58 6.70 9.13 7.78 7.92 9.58 4.61n 381 166 248 286 73 24 69 1247 102
Na�, mmol/LMean difference �0.40 �0.90 �0.52 0.47 NAb NA 2.56 �0.06 0.19SD of differences 3.18 2.53 2.43 2.13 NA NA 2.65 2.89 1.59n 411 203 73 241 11 3 72 1014 69
K�, mmol/LMean difference 0.12 0.08 0.02 0.09 NA NA 0.08 0.11 �0.01SD of differences 0.21 0.18 0.12 0.19 NA NA 0.16 0.19 0.07n 476 202 56 244 18 2 69 1067 44
Hct, %Mean difference �1.23 1.29 0.47 0.98 NA NA 0.65 0.28 �0.15SD of differences 3.05 2.75 2.35 2.82 NA NA 2.52 2.93 2.91n 302 198 224 276 2 0 69 1072 89a Performed at site D.b NA, not applicable.
2036 Billman et al.: Performance of an In-Line Point-of-Care MonitorD
ownloaded from
https://academic.oup.com
/clinchem/article/48/11/2030/5642240 by guest on 15 N
ovember 2021
Specifically, for pH, r was 0.99 vs 0.97; for Pco2, it was 0.98vs 0.94; for Po2, it was 1.00 vs 0.98; for Na�, it was 0.98 vs0.87; for K�, it was 0.99 vs 0.97; and for Hct, it was 0.95 vs0.90.
DiscussionThis multicenter clinical study was conducted to assessthe analytic performance and stability of an in-line, exvivo monitor, the VIA LVM Monitor, which is capable ofnear-continuous measurement of arterial blood gases,electrolytes, and Hct. The performance of the monitor wascomparable to that of conventional benchtop blood gasanalyzers and to other continuous and near-continuousin-line in vivo and ex vivo monitoring devices.
Widness et al. (7 ) previously characterized the in vivoperformance of the VIA LVM Monitor in 16 neonatalpatients at one site. The current study extends and morebroadly characterizes the performance of the VIA LVMMonitor among critically ill infants and neonates withvarious medical and surgical disorders being cared for in
freestanding children’s and university-based hospitals inthe US. To better approximate “real-world usage” of themonitor, no attempt was made to restrict the study to aspecific comparison instrument, catheter location, arterialline infusate, or location of the comparison instrument.Instead, preexisting blood gas and electrolyte analyzersalong with established methods of specimen collection,transport, and analysis were used. During the 72-h studyperiod, parenteral fluids infused through the arterialcatheters used to obtain blood samples varied from site tosite. In addition, at each site monitors were set up andoperated by a variety of clinical personnel, i.e., nurses,respiratory therapists, laboratory technologists, and phy-sicians.
These diverse practice patterns were included to morethoroughly characterize vulnerabilities and strengths ofthe VIA LVM Monitor in a variety of clinical venues.Despite considerable differences among the seven sites,the analytic performance of the VIA LVM Monitor, i.e., itsmean difference and SD of differences, for five of the sixanalytes was similar to those of other POC monitors andanalyzers (Table 3). The analytic performance of Hct wasthe least desirable of the six analytes based on therelatively greater SD of the differences relative to thecomparison instruments. The monitor’s less than desir-able performance in the analysis of Hct is likely attribut-able to differences in the methodologies of the twomeasuring devices, i.e., the monitor measures Hct byelectrical conductance whereas the laboratory instru-ments measure hemoglobin spectrophotometrically andindirectly derive Hct values based on an empiric correc-tion factor.
Not unexpectedly, the analytic performance of both thein vivo paired and the in vitro split whole-blood sampleswas not better than that of the laboratory instrumentsusing quality-control materials (Fig. 6). This notwith-standing, the data derived from the present study docu-ment that the VIA LVM Monitor functioned in an equiv-alent manner at the seven sites when operated by aspectrum of medical personnel, in a variety of busycritical care environments, and during the entire 72-hperiod of sensor use. Although our study did not addresscost comparisons of the monitor with standard blood gasanalysis systems, a report on neonates weighing 500-1000 gat birth indicates that use of the monitor can be cost-effective (12 ).
The similarity in the analytic performance of the in-lineVIA LVM monitor used in the present study relative toreports of other POC analyzers and monitors (Table 3) isnoteworthy for several reasons. (a) The existing NCCLSmethods comparison protocol (9 ) is based on split-sampleanalysis. This manner of comparison is not possible for invivo or ex vivo devices being tested in the clinical settingwith patient samples. Sequential (nonidentical) pairedsamples must be used when conducting these types ofcomparison studies with attached patient monitors. (b)The VIA LVM Monitor was operated by a variety of
Fig. 3. Mean differences and SD of differences of the VIA LVM Monitorrelative to the comparison method vs the duration of sensor use for thein vivo patient paired samples.The large filled circles connected with thick solid lines indicate mean difference;the small filled circles connected with thin solid lines indicate 1 SD of differencerelative to the mean difference. The two horizontal dashed lines indicate the CLIAperformance criteria limits referenced to zero. Numbers in brackets indicate thenumber of paired data points analyzed at each 12-h time interval.
Clinical Chemistry 48, No. 11, 2002 2037D
ownloaded from
https://academic.oup.com
/clinchem/article/48/11/2030/5642240 by guest on 15 N
ovember 2021
nonlaboratorians at the seven sites, under different envi-ronmental conditions, and with four different benchtopanalyzers as comparison instruments. (c) The blood sam-ples analyzed were clinical specimens obtained frompatients with a broad range of medical and surgical
disorders whose physiologic conditions varied widely. (d)The clinical specimens submitted for traditional labora-tory analysis were subject to preanalytic error duringcollection, transport, and analysis. Because the monitorwithdraws blood samples into a closed system and does
Fig. 4. Comparison of results obtained with the VIA LVM Monitor and the laboratory instruments for the in vitro laboratory split samples.The diagonal dashed line in each panel indicates the line of identity. The solid line represents the best-fit regression line for the range of values included. The equationfor the regression line, the r value, the SD of the residuals (Sy�x), and the number of data points for each analyte are as follows: for pH (A), y � 0.873x � 0.94 (r �0.99; Sy�x � 0.019; n � 59); for PCO2 (B), y � 0.842x � 2.44 mmHg (r � 0.98; Sy�x � 2.7 mmHg; n � 61); for PO2 (C), y � 0.949x � 3.8 mmHg (r � 1.00; Sy�x �3.7 mmHg; n � 102); for Na� (D), y � 0.898x � 13.7 mmol/L (r � 0.98; Sy�x � 1.3 mmol/L; n � 69); for K� (E), y � 1.008x � 0.03 mmol/L (r � 0.99; Sy�x � 0.073mmol/L; n � 44); for Hct (F), y � 0.851x � 6.49% (r � 0.95; Sy�x, not applicable; n � 89).
2038 Billman et al.: Performance of an In-Line Point-of-Care MonitorD
ownloaded from
https://academic.oup.com
/clinchem/article/48/11/2030/5642240 by guest on 15 N
ovember 2021
not require specimen transport, preanalytic errors aregreatly reduced or eliminated.
The ability of the VIA LVM Monitor to reliably andreproducibly measure pH, Pco2, Po2, Na�, K�, and Hct isfurther supported by the results of the split-sample invitro study. The in vitro data provide a useful benchmarkof the potential analytic performance of the VIA LVMMonitor in a controlled laboratory environment. With the
exception of Pco2, the in vitro r values were improvedrelative to the in vivo r values. The generally closeagreement of the paired-sample in vivo and split-samplein vitro data suggests that the instrument’s analytic per-formance is effectively buffered from the less-controlledoperating conditions found in the neonatal intensive careunit (NICU) environment. The regimen of periodic self-calibration and the temperature-controlled flow cell are
Fig. 5. Differences between the VIA LVM Monitor and comparison methods for the in vitro laboratory split samples.The thick solid horizontal line indicates the mean difference; the two horizontal dashed lines indicate the CLIA performance criteria limits referenced to the meandifference. The two vertical lines indicate the critical outlier values established for laboratory results for pediatric patients (20). The mean difference and the SD of themean differences are indicated for each analyte. The x-axis tick marks are along the horizontal line emanating from point 0 on the y axis.
Clinical Chemistry 48, No. 11, 2002 2039D
ownloaded from
https://academic.oup.com
/clinchem/article/48/11/2030/5642240 by guest on 15 N
ovember 2021
two design features that improve the monitor’s reproduc-ibility.
The discrepancy observed with in vitro split-sampletesting of Pco2 was noted to disproportionately involvehigh Pco2 values. The observation was reported to themanufacturer of the monitor. Additional work completedwith the help of the manufacturer suggested that thisdiscrepancy was in part attributable to the “memory” leftbehind by samples with high Pco2 values. By report, thiseffect has been corrected in the subsequent softwarerelease now in use. Further work is indicated to indepen-dently verify this.
The value of monitoring devices rests on their ability toreliably provide timely, accurate, and precise laboratoryinformation. The evaluation of laboratory results frompatients in the clinical setting is traditionally based onsporadic or scheduled sample collection and analysis.Although suitable for the management of most patients,this manner of testing provides insufficient detail toassess or intervene in dynamic clinical situations. Forunstable, critically ill patients in the hospital setting, theavailability of frequent patient data is more likely toidentify unrecognized or evolving events with sufficientadvance time to permit correction before adverse eventsoccur or before emergent corrective action becomes nec-essary. Similarly, the frequent measurement of laboratorydata may permit medical personnel to more appropriatelyadjust treatment for maximum benefit.
The selection of groups of patients who will benefitfrom the intensive scrutiny provided by POC devicessuch as the VIA LVM Monitor remains to be defined. Forsome patients, the availability of near-instantaneous testresults offers the potential for more efficient, cost-effec-tive, and appropriate adjustment of pharmacologic agentsand ventilatory support. For example, within the NICU, itis well documented that cerebral blood flow and long-term neurodevelopmental outcomes are significantly im-pacted by marked increases and decreases in Pco2, e.g.,those associated with intraventricular hemorrhage andperiventricular leukomalacia (3 ). Thus, for preterm in-fants, maintaining Pco2 concentrations within definedboundaries and avoiding the extremes of hypoxia andhyperoxia that are also associated with adverse long-termoutcomes are desirable goals. The VIA LVM Monitoroffers the potential for addressing these problems byproviding timely and accurate laboratory informationwith little laboratory blood loss.
The focus of our study was limited to critically illneonates and infants. The instrument may benefit and beuseful in the care of other patient populations, especiallythose vulnerable to fluid overload because of cardiac,hepatic, or renal pathology. Patients with highly dynamicand unstable disorders are those most likely to benefitfrom the availability and use of patient-attached monitors.As a group, critically ill neonates and infants possessadditional vulnerabilities that may allow them to deriveother benefits from monitoring devices. Specifically, neo-
Tabl
e3.
Com
pariso
nof
esti
mat
edbi
asan
dim
prec
isio
nva
lues
amon
gP
OC
devi
ces.
Cat
egor
yof
PO
Cde
vice
pHP
CO
2,
mm
Hg
PO
2,
mm
Hg
Na�
,m
mol
/L
K�
,m
mol
/L
Hct
,%
Esti
mat
edbi
as
SD
ofbe
twee
n-m
etho
ddi
ffer
ence
sEs
tim
ated
bias
SD
ofbe
twee
n-m
etho
ddi
ffer
ence
sEs
tim
ated
bias
SD
ofbe
twee
n-m
etho
ddi
ffer
ence
sEs
tim
ated
bias
SD
ofbe
twee
n-m
etho
ddi
ffer
ence
sEs
tim
ated
bias
SD
ofbe
twee
n-m
etho
ddi
ffer
ence
sEs
tim
ated
bias
SD
ofbe
twee
n-m
etho
ddi
ffer
ence
s
Nea
r-pat
ient
anal
yzer
sa0
.00
0.0
20.1
41.8
1�
0.5
27.5
5�
0.0
81.3
10.0
20.1
01.1
31.2
9In
-line
/in
vivo
mon
itors
b�
0.0
00
.03
0.9
63.2
70.3
818.6
4N
AcN
AN
AN
AN
AN
A
In-li
ne/e
xvi
vom
onito
rsN
on-V
IAd
0.0
00
.03
�0.0
82.8
2�
0.5
07.9
7N
AN
AN
AN
AN
AN
AVI
ALV
M(7
)�
0.0
00
.02
0.3
52.8
40.3
97.3
00.5
22.3
40.1
70.1
80.6
12.8
0VI
ALV
M(p
rese
ntst
udy)
0.0
00
.03
e0.6
83.6
2e
4.1
79.5
8e
�0.0
62.8
9e
0.0
90.1
9e
0.2
82.9
3
aM
onito
rs:
i-Sta
tPo
rtab
leC
linic
alAn
alyz
er(4
,21–2
3);
Sen
Dx
10
0®
(Sen
Dx
Med
ical
Inc.
)(2
4);
OPT
I1
(AVL
)(2
5);
Sci
entif
icC
orp.
IRM
A®(D
iam
etric
sIn
c.)
(26
,2
7).
bM
onito
rs:
Neo
tren
d(D
iam
etric
s)(3
);Pa
ratr
end
(Dia
met
rics)
(6,2
8,2
9).
cN
A,no
tav
aila
ble.
dM
onito
rs:
Sen
sica
thO
ptic
alS
enso
rIn
c.(3
0–3
2);
CD
I-3M
(Dia
met
rics
Med
ical
Inc.
)(3
3);
BP
33
00
,(P
urita
n-B
enne
tt)
(34
);C
DI™
(Hea
lthca
re)
(35
).e
The
SD
sof
the
diff
eren
ces
ofth
equ
ality
-con
trol
stan
dard
sfo
rfiv
eof
the
six
anal
ytes
incl
ude
the
mea
nim
prec
isio
nof
the
com
paris
onm
etho
dfo
rth
ese
ven
part
icip
atin
gsi
tes,
i.e.,
pH,
�0
.00
4;
PCO
2,
�0
.83
;P O
2,�
2.8
4;N
a�,�
0.7
8;a
ndK
�,�
0.0
4.B
ecau
sequ
ality
-con
trol
sam
ples
forH
ctar
eno
tav
aila
ble,
this
anal
yte
was
notin
clud
ed.N
ote
the
sim
ilarit
yof
thes
eda
taw
ithth
e1
99
9C
olle
geof
Amer
ican
Path
olog
ists
data
show
nin
Fig.
6.
2040 Billman et al.: Performance of an In-Line Point-of-Care MonitorD
ownloaded from
https://academic.oup.com
/clinchem/article/48/11/2030/5642240 by guest on 15 N
ovember 2021
nates and infants have reduced blood volumes that pre-dispose them to anemia as a consequence of frequentlaboratory testing. Iatrogenic blood loss has been esti-mated to be responsible for up to 90% of the transfusionsadministered in the NICU (13–16). In a preliminaryreport, the use of the VIA LVM Monitor in extremely lowbirth-weight infants produced a dramatic decrease in redblood cell transfusions in the first 2 weeks of life (17 ).Other potential advantages of the use of a near-continu-ous blood gas and electrolyte monitor include reducedblood exposure of healthcare personnel and reduced ratesof nosocomial infections.
The safe and appropriate duration of sensor use re-mains to be defined. Although the present study waslimited to sensor use of 72 h, no significant deteriorationin sensor performance was identified during this period.The current 72-h period of use was dictated by existinghospital infection control practice standards. The highlystable sensor performance observed across the period ofstudy supports inquiry into whether this period mightsafely be increased in a more cost-efficient manner. Fi-nally, the 72-h sensor stability data support inquiry as towhether it is necessary to repeat the comparison of pairedVIA LVM Monitor and benchtop reference samples as
Fig. 6. Mean differences and the SD of the mean differences relative to the comparison methods for the multicenter patient split-sample (left) andthe in vitro laboratory split-sample (right) analyses of the six analytes.The mean difference and the SD of the differences are indicated by the filled circles and error bars, respectively. The two horizontal dashed lines indicate the CLIAperformance criteria limits referenced to a zero difference. The numbers in brackets indicate the number of data points for each analyte. The shaded areas indicatethe reported SDs of proficiency test results from the College of American Pathologists 1999 Aqueous Blood Gas Survey for all types of instruments and thus includeother sources of variation (36). Note that there are no College of American Pathologists aqueous proficiency test results for Hct.
Clinical Chemistry 48, No. 11, 2002 2041D
ownloaded from
https://academic.oup.com
/clinchem/article/48/11/2030/5642240 by guest on 15 N
ovember 2021
long as the initial paired-sample comparison has metquality-control criteria.
The results of this multicenter study are encouragingfrom the standpoint of the performance of the monitortested. Particularly noteworthy was the robust perfor-mance of the VIA LVM Monitor when used in a variety oflocations and operated by staff with diverse backgrounds.The monitor compared favorably with the conventionalblood gas analyzers used at the seven study sites. Weconclude that this in-line, ex vivo monitor can be reliablyused in the management of critically ill neonates andinfants when arterial catheters are in use. The optimaltime period of sensor use, the required frequency forpaired-sample quality-control testing, the manner andextent to which the near-continuous availability of dataprovided by POC monitoring impact patient care, and thecost-effectiveness of providing near-continuous monitor-ing are fundamental issues requiring additional study(18, 19).
We thank Michele Richardson, Sabrina Sood, and CathyButterfield for data entry assistance, and Mark A. Hart fortechnical assistance. For the in vitro study, we acknowl-edge Brian Hicks, Erin McGuire, David Buse, and theNICU laboratory staffs for technical support, and Karen J.Johnson, Gretchen A. Cress, and Laura K. Knosp forobtaining blood samples for the split-sample in vitrostudy. Statistical help with the Deming regression analy-sis was provided by M. Bridget Zimmerman, PhD. Thisstudy was supported in part by funding provided by theNIH General Clinical Research Centers Program (GrantRR00059), by the Child Health Corporation of America,and by Metracor Technologies Inc. (San Diego, CA). Wehad full access to all data in the study and take responsi-bility for the integrity of the data and the accuracy of thedata analyses. None of us serves on any of the advisory orgoverning boards of either the Child Health Corporationof America or Metracor Technologies, Inc. Child HealthCorporation of America and Dr. Billman hold equity inMetracor Technologies Inc.
References1. Kost GJ, Hague C. In vitro, ex vivo, and in vivo biosensor systems.
In: Kost GJ, Welsh J, eds. Handbook of clinical automation,robotics, and optimization. New York: John Wiley & Sons, 1996:648–753.
2. Kost GJ, Ehrmeyer SS, Chernow B, Winkelman JW, Zaloga GP,Dellinger RP, et al. The laboratory-clinical interface: point-of-caretesting. Chest 1999;115:1140–54.
3. Morgan C, Newell SJ, Ducker DA, Hodgkinson J, White DK, MorleyCJ, et al. Continuous neonatal blood gas monitoring using amultiparameter intra-arterial sensor. Arch Dis Child 1999;80:F93–8.
4. Murthy JN, Hicks JM, Soldin SJ. Evaluation of i-STAT portableclinical analyzer in a neonatal and pediatric intensive care unit.Clin Biochem 1997;30:385–9.
5. Tobias JD, Connors D, Strauser L, Johnson T. Continuous pH andPCO2 monitoring during respiratory failure in children with the
Paratrend 7 inserted into the peripheral venous system. J Pediatr2000;136:623–7.
6. Weiss IK, Fink S, Harrison R, Feldman JD, Brill JE. Clinical use ofcontinuous arterial blood gas monitoring in the pediatric intensivecare unit. Pediatrics 1999;103:440–5.
7. Widness JA, Kulhavy JC, Johnson KJ, Cress GA, Kromer IJ,Acarregui MJ, et al. Clinical performance of an in-line point-of-caremonitor in neonates. Pediatrics 2000;106:497–504.
8. Bailey PL, McJames S, Cluff ML, Wells DT, Orr JA, WenternskowDR. Evaluation in volunteers of the VIA V-ABG automated bedsideblood gas, chemistry, and hematocrit monitor. J Clin Monit Com-put 1998;14:339–46.
9. National Committee for Clinical Laboratory Standards. Methodcomparison and bias estimation using patient samples; approvedguideline NCCLS document EP09-A. Wayne, PA: NCCLS, 1995.
10. Bland JM, Altman DG. Statistical methods for assessing agree-ment between two methods of clinical measurement. Lancet1986;i:307–10.
11. Ehrmeyer SS, Laessig RH, Leinweber JE, Oryall JJ. Medicare/CLIAfinal rules for proficiency testing: minimum intralaboratory perfor-mance characteristics (CV and bias) needed to pass. Clin Chem1990;36:1736–40.
12. Alves-Dunkerson J, Hilsenrath PE, Cress GA, Widness JA. Costanalysis of a neonatal point-of-care monitor. Am J Clin Path2002;117:809–18.
13. Widness JA. Pathophysiology, diagnosis and prevention of anemiaduring the neonatal period. NeoReviews 2000;1:e61–8.
14. Blanchette V, Zipursky A. Assessment of anemia in newborninfants. Clin Perinatol 1984;11:489–516.
15. Obladen M, Sachsenweger M, Stahnke M. Blood sampling in verylow birth weight infants receiving different levels of intensive care.Eur J Pediatr 1988;147:399–404.
16. Ringer SA, Richardson DK, Sacher RA, Keszler M, Churchill WH.Variations in transfusion practice in neonatal intensive care.Pediatrics 1998;101:194–200.
17. Widness JA, Madan A, Stevenson DK. Reduction in RBC transfu-sions among preterm infants: results of randomized trial utilizingan in-line point-of-care blood gas and electrolyte monitor [Ab-stract]. Pediatr Res 2002;51:376A.
18. Kilgore ML, Steindel SJ, Smith JA. Cost analysis for decisionsupport: the case of comparing centralized versus distributedmethods for blood gas testing. J Healthc Manag 1999;44:207–15.
19. Rainey PM. Outcomes assessment for point-of-care testing. ClinChem 1998;44:1595–6.
20. Kost GJ. Critical limits for emergency clinician notification atUnited States children’s hospitals. Pediatrics 1991;88:597–603.
21. Gault MH, Harding CE. Evaluation of i-STAT portable clinicalanalyzer in a hemodialysis unit. Clin Biochem 1996;29:117–24.
22. Jacobs E, Vadasdi E, Sarkozi L, Coman N. Analytic evaluation ofi-STAT portable clinical analyzer and use by nonlaboratory health-care professionals. Clin Chem 1993;39:1069–74.
23. Shelledy DC, Smith WA. A comparison of the i-STAT system andCorning 278 for the measurement of arterial blood gas values.Respir Care 1997;42:693–7.
24. Lindemans J, Hoefkens P, VanKessel AL, Bonnay M, KulpmannWR, VanSuijlen JEE. Portable blood gas and electrolyte analyzerevaluated in a multi-institutional study. Clin Chem 1999;45:111–7.
25. MacIntyre NR, Lawlor BL, Carstens D, Yetsko D. Accuracy andprecision of a point-of-care blood gas analyzer incorporatingoptodes. Respir Care 1996;41:800–4.
26. Wahr JA, Lau W, Tremper KK, Hallock L, Smith K. Accuracy of anew, portable, hand-held blood gas analyzer, the IRMA®. J ClinMonit 1996;12:317–24.
2042 Billman et al.: Performance of an In-Line Point-of-Care MonitorD
ownloaded from
https://academic.oup.com
/clinchem/article/48/11/2030/5642240 by guest on 15 N
ovember 2021
27. Zaloga GP, Roberts PR, Black K. Hand-held blood gas analyzer isaccurate in the critical care setting. Crit Care Med 1996;24:957–62.
28. Zollinger A, Spahn DR, Singer T, Lalundardo MP, Stoehr S, WederW, et al. Accuracy and clinical performance of a continuousintra-arterial blood-gas monitoring system during thoracoscopicsurgery. Br J Anaesth 1997;79:47–52.
29. Abraham E, Gallagher TJ, Fink S. Clinical evaluation of a multipa-rameter intra-arterial blood-gas sensor. Intensive Care Med 1996;22:507–13.
30. Clifford PS, Muzi M. Validity of a patient-dedicated blood gasmonitoring system [Abstract]. Am J Resp Crit Care 1996;153:602.
31. McKinley BA, Parmley CL. Preliminary clinical trial of an ex vivoarterial blood gas monitor. J Crit Care 1997;12:214–20.
32. Myklejord DJ, Pritzker MR, Nicoloff DM, Emery AM, Emery RW.Clinical evaluation of the on-line SensicathTM blood gas monitoringsystem. Heart Surg Forum 1998;1:60–4.
33. Mahutte CK. On-line arterial blood gas analysis with optodes:current status. Clin Biochem 1998;31:119–30.
34. Haller M, Kilger E, Briegel J, Forst H, Peter K. Continuousintra-arterial blood gas and pH monitoring in critically ill patientswith severe respiratory failure: a prospective, criterion standardstudy. Crit Care Med 1993;22:580–6.
35. Shapiro BA, Mahutte CK, Cane RD, Gilmour IJ. Clinical perfor-mance of a blood gas monitor: a prospective multicenter trial. CritCare Med 1993;21:494–7.
36. College of American Pathologists. Critical Care/Aqueous BgasSurvey, Surveys Program. Northfield, IL: College of AmericanPathologists.
Clinical Chemistry 48, No. 11, 2002 2043D
ownloaded from
https://academic.oup.com
/clinchem/article/48/11/2030/5642240 by guest on 15 N
ovember 2021