- Email: [email protected]
Continuous arterial and venous blood gas monitoring during cardiopulmonary bypass
J THORAC CARDIOVASC SURG 1991;102:431-9
Continuous arterial and venous blood gas monitoring during cardiopulmonary bypass A new monitoring technique, based on optical fluorescence chemistry, aUows continuous monitoring of all blood gas variables during cardiopulmonary bypass. To evaluate the clinical performance of this monitor, we drew 220 arterial and 216 venous blood samples from 15 patients, and simultaneous blood gas values displayed by the monitor were compared with standard laboratory measurements. The continuous monitor predicted laboratory values with varying degrees of accuracy. (R2 values by linear regression: arterial oxygen tension 0.86, venous oxygen tension 0.36, arterial carbon dioxide tension 0.58, venous carbon dioxide tension 0.72, arterial pH 0.53, venous pH 0.58; p < 0.0001). Monitor values of arterial oxygen tension overestimated laboratory values (bias = + 43.5 mm Hg), but the laboratory reference method likely underestimated true arterial oxygen tension in the high range achieved on bypass. Monitoring of venous oxygen tension was imprecise (precision = ± 6.51 mm Hg), regardless of whether stable conditions existed during the sampling period. Monitoring of carbon dioxide tension and pH showed small bias (carbon dioxide tension within 2 mm Hg, pH within 0.03) and good precision (carbon dioxide tension within 3 mm Hg, pH within 0.03). With the development of unstable conditions on bypass, monitor arterial oxygen tension values showed a changing relationship to corresponding laboratory values. In conclusion, arterial and venous carbon dioxide tension and pH monitoring provide acceptably accurate alternatives to laboratory measurement of these variables during cardiopulmonary bypass. Arterial oxygen tension monitoring accurately indicates changes in oxygen tension in the arterial oxygen tension range typically produced during extracorporeal circulation. Oxygen tension monitoring in the venous oxygen tension range is too imprecise for clinical decision-making purposes.
Jonathan B. Mark, MD,a, b Daniel FitzGerald, CCP,a, b Terry Fenton, EdD,b,c Anna Mae Fosberg, RN, ccp,a, bWilliam Camann, MD,a, b Nick Maffeo, BA,a and James Winkelman, MD,a, b Boston. Mass.
Intermittent measurements ofblood pH, oxygen tension (P02), and carbon dioxide tension (PC02) are standard means for assessing physiologic status during cardiopulmonary bypass. In simplest terms, arterial blood gas valFrom Brigham and Women's Hospital,' Harvard Medical School," and Harvard School of Public Health," Boston, Mass. Presented in part at the Eleventh Annual Meeting of the Society of Cardiovascular Anesthesiologists, Seattle, Wash., April 16-19, 1989. Supported in part by a grant from Cardiovascular Devices, Inc., Irvine, Calif. Received for publication Sept. 25, 1989. Accepted for publication April 3, 1990. Address for reprints: Jonathan B. Mark, MD, Department of Anesthesia, Brigham and Women's Hospital, 75 Francis St., Boston, MA 02115.
12/1/21518
ues reflect pump-oxygenator performance, whereas venous blood gases indicate the adequacy of oxygen delivery to the body tissues.' If all these variables were monitored continuously, rather than measured intermittently in the laboratory, benefits might include earlier detection of deleterious trends in pump-oxygenator performance or a patient's condition, avoidance of phlebotomy, and reduction of risks to health care workers from exposure to phlebotomized blood. Developments in optical fluorescence chemistry have provided a suitable technology for continuous arterial and venous blood gas monitoring during cardiopulmonary bypass.s 3 Previous investigations in vitro': 5 and in vivo'" 7 have concluded that this kind of blood gas monitoring technique adequately follows blood gas trends but is not accurate enough to supplant traditional laboratory analyses. However, those studies employed an early version of
431
The Journal of Thoracic and Cardiovascular
4 32Mark et al.
Surgery
,----- HYPOTHERMIA ------, EVENTS
ON CPS
29°C
I
SAMPLE TIME
+5 +7 +9 +11 min I I I I 1 2 3 4
I
rl- - - - REWARM - - - - - - ,
1 Q ,V' 102% r r y,
I
5
Fig. 1. Blood sampling protocol. CPR, Cardiopulmonary bypass; concentration.
° +2 +4 +6 30C min I I I I 6 7 8 9
I
+20
l/
I 16
OFF CPS
I
12, pump blood flow; V. gas flow; 02%.
oxygen
PVC02 (/)
I.LJ -l
o, :E
u,
0
a::
W ID
:E ::::>
Z
70 60 50 40 30 20 10
-8
-4
0
4
8
-20 -16 -12 -8
-4
0
4
8
12 16
20
DIFFERENCE (mmHg) (Monitor - Lab) Fig. 2. Histograms showing differences (monitor-laboratory) for Pve02 and Pvo-.
the instrument system (GAS-STAT, Cardiovascular Devices, Inc., Irvine, Calif) and focused on arterial samples; improved sensors and algorithms incorporated in current devices have not been evaluated. Furthermore, neither the original nor the current monitoring system has been validated over a wide range of arterial and venous blood gas values. Finally, monitor performance has not been tested during well-defined stable and unstable clinical conditions. Our investigation was conducted with the newest generation optical fluorescence blood gas monitor (COl 300, Cardiovascular Devices, Inc. Irvine, Calif.). We posed three clinical questions: (1) What are the trending capabilities for Po 2, Pco-, and pH sensors in both arterial and venous blood, and what is the overall performance of the continuous blood gas monitor during cardiopulmonary bypass? (2) During stable bypass conditions, how accurate is the blood gas monitor in predicting simultaneous laboratory values? (3) What happens to the relationship between monitor and laboratory blood gas values during unstable, rapidly changing conditions?
Patients and methods Patients. Fifteen patients undergoing cardiac operations requiring cardiopulmonary bypass were studied. In each case, informed written consent was obtained in accordance with standards set by the Hospital's Human Research Committee. The average patient was 66 years of age, weighed 78 kg, and had a bypass time of 82 minutes and an aortic crossclamp time of 54 minutes. Anesthetic drugs included opioids (fentanyl, sufentanil, morphine), and volatile agents (enflurane, isoflurane), all of which have been shown not to interfere with fluorescence blood gas monitoring." Cardiopulmonary bypass. Cardiopulmonary bypass was conducted with a nonpulsatile roller pump, a hollow-fiber membrane oxygenator, and a 20 lim arterial filter. The bypass circuit was primed with lactated Ringer's solution and beeflung heparin. Patient temperature was monitored with a nasopharyngeal thermistor, and arterial and venous blood temperatures were monitored with thermistors incorporated in the continuous blood gas monitor sensors. In each patient a period of stable hypothermic cardiopulmonary bypass was established, followed by a period of rapid rewarming to normothermia. Pump blood flow was 1.1 L· min"! . m- 2 during hypothermia, and increased to 2.0 L . min-I . m- 2 during rewarming. Gas flow rates were 2.0 L . min-I during hypothermia and 3.0 L . min"! during rewarming. Oxygen concentration was maintained at
Volume 102 Number 3
Continuous monitoring duringCPB 433
September 1991
60% during hypothermia and increased to 100% during rewarming. Supplemental carbon dioxide was not added to the sweepgases. Pump blood flow,gas flow, and oxygen concentration were held constant during hypothermia and adjusted in a stepwisefashion commencing with rewarming. This allowed the bypassperiod to be divided into two phases: stable hypothermia and unstable rapid rewarming. Blood gas monitoring. Before each case the continuous bloodgas monitor was calibrated in vitro according to the manufacturer's directions, with standard reference sensors, buffer solution,and two standard calibrating gases (2.8% carbon dioxide, 5.5% oxygen, 91.7% nitrogen; 8.0% carbon dioxide, 30.0% oxygen, 62.0% nitrogen). All sensors, buffers, and standard gases were COl brand. All 30 arterial and venous sensors were successfully calibrated. The arterial sensor was placed in the arterial line 47 inches beyond the oxygenator, and the venous sensor was placed in the venous line 28 inches before the reservoirbag. Dedicated time to perform calibration and incorporate the monitoring system into the extracorporeal circuit averaged 6 minutes. During cardiopulmonary bypass, monitor blood gas values were recorded every 6 seconds on magnetic disk (Toshiba TlOOO, Tustin, Calif.). All blood gas values were measured at actual arterial and venous blood temperatures and corrected to 37° C by algorithms incorporated into the blood gas monitor. Blood sampling. For comparison with monitor values, arterial blood was aspirated from the arterial filter purge line (28 inches before the arterial sensor), and venous blood was withdrawn from the venous return line at its entrance to the reservoir bag (26 inches from the venous sensor). Blood sampling followed a predetermined protocol in each patient (Fig. 1). After stable hypothermic conditions had been established for 5 minutes (constant temperature, pump blood and gas flows, and oxygen concentration), four paired samples (arterial and venous) were drawn at 2-minute intervals, with the simultaneous monitor blood gas values electronically marked for later analysis. A fifth sample was drawn at the conclusion of the hypothermic period; immediately thereafter the patient was actively rewarmed, and step changes were made in pump blood flow, gas flow, and oxygen concentration. When venous blood temperature had increased 1° C from its nadir, sampling again commenced at 2-minute intervals for 20 minutes, during this unstable period. A total of 16 arterial and venous samples were drawn in each patient (five stable hypothermic, 11 unstable rewarming), except when cardiopulmonary bypass was terminated before all sampling could be completed. Laboratory blood gas analysis. Laboratory analyses were performed by the same technician, with a single instrument (Ciba-Corning 168 blood gas analyzer, Medfield, Mass.), located in the operating room. All blood samples were withdrawn anaerobically into 3 ml polypropylene syringes, immediately placed on ice, and analyzed in sampled order. The blood gas analyzer was calibrated on the P02 and PC02 channels with two certified gas cylinders (5% carbon dioxide, 12% oxygen, 83% nitrogen; 10% carbon dioxide, 90% nitrogen), and pH was calibrated with two buffer standards (pH = 7.381; pH = 6.838). All flush solutions, buffers, and standard gases were Ciba-Corning brand. Before the start of the study all routine maintenance was performed on the blood gas analyzer, including changing membranes and rotary port 0 rings. Daily two-point calibration was
Table I. Laboratory bloodgas analyzer controls Variable
N
Mean ± SD
Range
15 15 15
94.4 ± 1.2 127.8 ± 2.4 58.9 ± 1.5
91.6-96.7 125.6-133.1 56.1-60.9
PC02 Alkalosis
15 15 15
43.8 ± 0.6 62.9 ± 0.7 21.8 ± 0.6
42.8-45.0 62.0-64.2 20.6-22.6
pH Normal pH Acidosis pH Alkalosis
15 15 15
7.392 ± 0.006 7.141 ± 0.005 7.630 ± 0.007
P02 Normal P02 Acidosis P02 Alkalosis PC02 Normal PC02 Acidosis
7.380-7.401 7.131-7.149 7.614-7.641
SO, Standard deviation. Descriptive statistics summarizing daily controls run on the Ciba-Corning laboratory blood gas analyzer. Values for P02 and Pco, are in millimeters of mercury.
performed until calibration values were stable and reproducible. (On average, three two-point calibrations were required, which is normal after electrical interruption.) Subsequently three blood gas controls (normal acid-base, respiratory acidosis, and respiratory alkalosis) were run daily to verify instrument performance (Table I). During each study the blood gas analyzer had a one-point calibration performed hourly to check for electrode drift. When drift exceeded ± 0.005 pH units, 1.8 mm Hg P02, or 1.0 mm Hg PC02, calibration was repeated. All laboratory blood gas values were measured and reported at 37°C. Data analysis. Laboratory blood gas values and nasopharyngeal temperatures were recorded manually, and these data were combined with electronically stored, simultaneous blood gas monitor values for analysis with the SAS statistical package (SAS, Inc., Cary, N.C.). Monitor trending capabilities and overall performance were evaluated first by simple linear regression of monitor versus laboratory blood gas values. Next a multiple regression model analyzed the main effects of sample time and individual subjects along with monitor values as predictors of laboratory blood gas values. Subsequently the regression model examined interaction effects of monitor by sample time and monitor by individual subjects. These analyses tested whether the slope of the monitor/laboratory blood gas value relationship differed significantly across sample times or between individual subjects. Overall monitor performance was analyzed further by calculating the mean difference between monitor and laboratory values and the standard deviation of the difference. According to the methods of Bland and Altman'' and Barker and coworkers.? the mean difference between the two measurement techniques was considered to be the bias and the standard deviation of individual scores around that difference the precision of the monitoring instrument. Monitor performance during stable hypothermic conditions also was assessed by calculation of bias and precision for each monitored variable, as indicators of monitor accuracy. Effects of rapid rewarming and abrupt adjustment of blood and gas flows,on each variable, were analyzed by plotting individual and mean monitor and laboratory values across all 16 sampling times and by examining the relationship between monitor and laboratory blood gas values during both stable (samples 1 to 5) and unstable (samples 6 to 16) periods.
The Journal of Thoracic and Cardiovascular Surgery
4 34Mark et al.
Table II. Multiple regression analyses: Monitor, sample time, and individual subjects as predictors for laboratory values Variable Pao2 Pv02
PacoPvC02 pHa pHv
C: Interactions
B: Main effects
A: Monitor only R2
Monitor
0.86 0.36 0.58 0.72 0.53 0.58
0.0001 0.0001 0.0001 0.0001 0.0001 0.0001
Time 0.0001 0.0001 0.0001 0.075 0.0001 0.0006
Subjects
R2
Monitor/time
Monitor/subjects
R2
0.017 0.0001 0.0001 0.0001 0.0001 0.0001
0.91 0.79 0.91 0.92 0.89 0.91
0.003 0.68 0.93 0.99 0.78 0.62
0.0003 0.003 0.15 0.13 0.21 0.25
0.94 0.85 0.92 0.94 0.91 0.93
Multiple regression analyses with monitor blood gas values, sample time, and individual subjects as predictors of laboratory blood gas values. See text for details. All entries are p values and R2 values derived from regression analyses.
Table III. Overall relationship between monitor and laboratory blood gas values Variable Pao2 Pv°2 Pac02 PvC02 pHa pHv
N 220 203 220 203 220 203
Lab value (mean ± SD) 306.8 41.3 33.7 39.1 7.46 7.41
± ± ± ± ± ±
86.1 5.7 4.1 3.9 0.04 0.03
Bias ± Precision
r
± ± ± ± ± ±
0.93 0.60 0.76 0.85 0.73 0.76
+43.5 -1.44 +1.62 + 1.40 -0.025 -0.029
43.3 6.51 2.79 2.14 0.029 0.023
Slope 0.74 0.42 0.81 0.84 0.83 0.76
± ± ± ± ± ±
0.04 0.08 0.09 0.07 0.10 0.09
Intercept 48.7 24.48 5.22 5.02
± ± ± ±
14.7 3.14 3.25 2.97
SD, Standard deviation. Slopes and intercepts of the regression equations are given ± 95% confidence intervals. Values for P0 2 and PC02 are in millimeters of mercury. All r values (correlation coefficients derived from regression of monitor versus laboratory values) are significant (p < 0.0001). Intercepts for pH are not reported since these values lack physiologic significance.
Results Blood gas comparisons were performed on 220 arterial and 216 venous blood samples, with the use of 30 monitoring sensors in 15 patients. Twenty arterial and 24 venous samples were unavailable for analysis owing to early completion of bypass, arterial contamination of sampled venous blood, or loss of electronically stored monitor data. From a clinical standpoint all sensors appeared to function well, except for a single venous sensor. Venous Pco- (Pvco-) and venous pH (pHv) values were clearly erroneous in one patient, with Pvco- > 67 mm Hg and pHv < 7.21. This venous sensor was considered to have failed (sensor failure rate 3.3%), and data collected from this venous sensor were not included in further analyses. Trending and overall performance. Table II summarizes results of multiple regression analyses. The continuous blood gas device was capable of predicting laboratory blood gas values for each monitored variable with varying degrees of accuracy. Regression analyses revealed R 2 values of 0.36 and 0.86 for Pvo- and Pao-, respectively, with R 2 values for Paco-, Pvco-, pHa, and pHv being intermediate. All R 2 values were highly significant (p < 0.0001). When the effects of sample times and individual subjects were taken into account, the main effects of these variables significantly improved on
the accuracy of prediction of laboratory blood gas values, over and above the level of accuracy achieved by the monitor alone (Table II, B). (The single exception was Pvco-, where sample time was a marginally significant factor [p = 0.075].) A significant monitor value by sample time interaction was present for Pao, (Table II, C), indicating that the slope of the monitor/laboratory blood gas value relationship varied significantly over sample times. Furthermore, significant monitor by individual subject interactions were present for both Pao- and Pvo(Table II, C), indicating that the slopes of the monitor/ laboratory blood gas value relationships varied significantly across subjects for these two variables. Overall performance of the continuous blood gas monitor is summarized in Table III. Monitored values of Pao, overestimated laboratory values and Pvo- monitoring was imprecise. For each of the other monitored variables, bias was small and precision good. The imprecision of Pvo z monitoring is highlighted in Fig. 2, where Pvo- is compared with Pvco- monitoring. This comparison seemed appropriate, because these two venous blood gas variables were monitored within the same sensor group, and the distributions of values monitored covered similar physiologic ranges (Pvo- = 41 mm Hg mean, 30 mm Hg to 62 mm Hg range; Pvco- = 39 mm Hg mean, 33 mm Hg to 50 mm Hg range).
Volume 102 Number 3 September 1991
Continuous monitoring during CPS
435
40 u 'L W
a::
35f-
::> ~
a::
w a..
~ I.JJ
301-
~
25
~
2
3 4
5 6
7 8
9 10 11 12 13 14 15 16
SAMPLE TIME Fig. 3. Mean nasopharyngeal (0), arterial blood (.), and venous blood (~) temperatures at each sample time.
70 = 0.44 p= 012
r
n: 14
400 I
(\J
E
300
(\J
o
50
>
•
•
a..
a..
CD
CD
j
Slope: 022 ± 025 Intercept: 343± 11.5
E
E
E
oa
60
-a. I
0-
r ;0.99 p
200
•
...J
40
•
Intercept = 23.8± 14.2
200
300
400
40
MONITOR Po0 2 (mmHg)
50
60
70
MONITOR PV02 (mmHg)
Fig. 4. Scatterplot relating monitor Pao- versus laboratory Pao- at stable sample time 4. Slope and intercept of the regression equation are reported ± 95% confidence intervals. The line of identity is indicated.
Fig. 5. Scatterplot relating monitor Pvoz versus laboratory Pvo- at stable sample time 4. Slope and intercept of the regression equation are reported ± 95% confidence intervals. The line of identity is indicated.
Substantial differences between monitor Pvo- and laboratory Pvo- were common: monitor values were within 3 mm Hg of laboratory values 32.0% of the time; within 5 mm Hg 54.2% of the time; and within 10 mm Hg only 91.1% of the time. By contrast, monitor Pvco- was within 3 mm Hg of laboratory values 70.4% of the time and within 5 mm Hg 99.0% of the time. Accuracy during stable conditions. Stable hypothermic conditions were readily achieved during cardiopulmonary bypass. At sample time 4, nasopharyngeal, arterial blood; and venous blood temperatures were within
0.6 C of one another, and had drifted by less than 0,70 C over the previous 6 minutes (Fig. 3). Cardiopulmonary bypass blood flow,gas flow,and oxygen concentration had been constant during the same interval. During this clinically stable period, monitored blood gas values most closely predicted laboratory values (Table IV), but significant differences between techniques remained. Laboratory Pao- was still overestimated by the monitor. In addition, during stable conditions, Pvo- monitoring continued to be imprecise. Although mean bias of Pvo- was not statistically significant, the monitor frequently gross0
The Journal of Thoracic and Cardiovascular
4 36Mark et al.
Surgery
500 Ol
I
E E c
0
eu
E
0
400 300
t\I
0
a..
• LAB
200 01
o
2
4
6
8
10
12
MONITOR
14
16
SAMPLE TIME Fig. 6. Mean laboratory and monitor values for Pao- at each sample time in eight patients who had a single-step from 60% to 100%. increase in inspired oxygen concentration
en
lyoverestimatedor underestimated laboratory Pvo-. This pointisgraphically highlightedby examiningscatterplots of monitor versus laboratory valuesfor Pao, and Pv02at sample time 4 (Figs. 4 and 5). Monitored valuesof Paocorrelate highly with laboratory values (r = 0.99, P < 0.0001), whereas there is only a weak association (r = 0.44, P = 0.12) between monitor and laboratory Pvo- values during this same period. Performance during unstable conditions. Clinically unstable conditions were created with the onset of rapid rewarming during bypass. All monitored temperatures increased abruptly (see Fig. 3), as extracorporeal blood and gas flows were adjusted to new target values, and oxygenconcentration was increasedin each patient. During the rewarming period,valuesfor Pao, nearly doubled (Fig. 6), whereas values for other variables were altered little. When monitor and laboratory bloodgasvalueswere plotted acrossall 16sample times,there was no consistent pattern of perturbation of the monitor versus laboratory relationship for any variable except Pao-, The relationship between monitor Pao2 and laboratory Pao-, however, changed dramatically at the onset of the unstable period. In eight subjects who had a single-step change in inspired oxygen concentration from 60% to 100% between sample times 5 and 6, monitor Pao- overestimated laboratory Pao- by + I 0 mm Hg at sample time 5, underestimated by -18 mm Hg at sample time 6, and again overestimatedby +50 mm Hg at sample time 7 (see Fig. 6). The variability of the Pao- relationship was equally evident when data from all 15 subjectswere plotted. A particularly striking example of this phenomenonis illustrated by the data from one patient, where exact
monitor/laboratory agreement occurred by the end of the stable hypothermic period (Fig. 7, sample time 5). As Pao, rose rapidly during rewarming, the monitor sensor transiently but dramatically underestimated the simultaneouslaboratory valueby 74 mm Hg (Fig. 7, sampletime 6). The typical monitor/laboratory relationship returned at sample time 7 and was maintained for the duration of the cardiopulmonary bypass period. Discussion Trending and overall performance. Continuousblood gas monitoringprovides reliabletrending informationfor detecting changes in blood PC02, pH, and Pao-; changes in Pvo, values are least adequately predicted by this monitoring technique. Improved predictions of laboratory blood gas values wereobtained when individual subjectsand sampletimes were considered, over and above predictions provided by monitor blood gas values. The amount of incremental variance explained by subject and time effectswas modest for Pao-, but considerably larger for the other variables,approaching 43% for Pv02. This indicatesthat there were systematiccomponentsof variation in laboratory values associated with sample times and individual subjects, which were not reflected by monitor values. Although these analyses suggest there is room for improvement in the accuracy of the monitor, is the prediction providedby the current monitor good enough? A technique may be sufficient to supplant laboratory measurements for clinical purposes, but not for research requirements. Moreover, a monitoring device may be adequate for use as a supplement to intermittent laboratory measurements, while not entirely substituting for
Volume 102
Continuous monitoring during CPR 43 7
Number 3 September 1991
500
~
0'
I
E E
0
N
400 300
c
0• LAB
200
01
o
2
4
6
8
10
12
MONITOR
14
16
SAMPLE TIME Fig. 7. Laboratory and monitor values for Pao , at each sample time in one patient from the group shown in Fig. 6.
these measurements. These questions are best answered by examining each blood gas variable separately. Differences between monitor and laboratory values for Pco- were small, an overestimation bias of approximately 1.5 mm Hg. Furthermore, precision of Pco- monitoring was good, averaging ± 2.5 mm Hg. This degree of accuracy might be acceptable for clinical monitoring but unacceptable for research measurements. Monitoring of pHa and pHv like Pco-, might be considered a reasonable clinical substitute for laboratory measurement. Ninetyfivepercent of the time, monitor pH values were no more than 0.08 units lower or 0.03 units higher than the corresponding laboratory pH value. Unfortunately, monitoring of P02 was more problematic. Pao- and Pvo- values spanned entirely different, nonoverlapping ranges. Unique monitoring problems were discovered specific to these ranges. However, problems identified for Pvomonitoring, are clearly the same as would exist during Pao- monitoring in hypoxemic patients. Monitor values of Pao- overestimated laboratory values. This disagreement is mitigated by a number of factors. First, selection of laboratory values as accurate reference standards for Pao- in the high range may be questioned. Second, as a percentage of the laboratory value, the bias was lessthan 15%,a disagreement of minor clinical importance in this high P0 2 range. Third, Pao, bias tends to increase as Pao- rises. As illustrated in Fig. 4, monitor values agree well with laboratory values when P0 2 is less than 200 mm Hg, and show increasing overestimation as P0 2 rises. Pvo 2 trending and overall monitoring were the most unsatisfactory in this study. R2 values from all regression analyses were lowest for Pv02. While overall bias for Pvo-
Table IV. Relationship between monitor and
laboratory bloodgas values during stable conditions Variable
N
Lab value (Mean ± SD)
Pao2 Pv02 PaC02 PvC02 pHa pHv
15 14 15 14 15 14
220.0 44.1 36.9 41.8 7.43 7.39
± ± ± ± ± ±
62.0 4.3 3.8 4.4 0.04 0.04
Bias ± precision +24.3 + 1.01 +0.82 +0.85 -0.017 -0.029
± ± ± ± ± ±
16.8 7.91 2.94 2.18 0.025 0.022
SD, Standard deviation. Values for P02 and Pco- arc in millimeters of mercury.
was small, this became less important in view of the poor precision of Pvo- monitoring. In 95% of comparisons, monitor Pvo, values underestimated laboratory values by as much as 14.5 mm Hg and overestimated by as much as 11.5 mm Hg. Consequently, neither changes in Pv02 nor absolute values of this monitored variable can be relied on for clinical management purposes. Accuracy during stable conditions. By creating both stable and unstable clinical conditions during cardiopulmortary bypass, we hoped to learn why blood gas values from the monitor and laboratory disagreed. The continuous blood gas monitor sensors have an inherent delay in responding to changes in blood gas tensions.l-" Blood mixing, membrane gas and ion permeability, and chemical reactions within the fluorescent dyes all contribute to this lag time. By examining data from stable sample time 4, we attempted to eliminate any monitor inaccuracy attributable to sensor lag. Similar to the results from the overall data sample, those from the stable period show little bias and adequate precision for Pco- and pH monitoring. Again, monitor Pao- significantly overestimated
4 38Mark et al.
laboratory values, but this disagreement was reduced from 15% to II %. Even during the most stable interval, Pvo, monitoring was imprecise and a clinically inadequate predictor oflaboratory Pvo-, The same patterns of monitor performance were evident from both the stable data and the overall set. Performance during unstable conditions. Although temperature and Pao- changed rapidly during rapid rewarming, none of the other blood gas variables were consistently perturbed. Apparently alterations in perfusion and ventilation matched patient requirements, thus obviating dramatic changes in Pvoz, Pco-, and pH. In the absence of such changes, the relationship between monitor and laboratory values for these blood gas variables remained relatively constant across time. The interaction term from the multiple linear regression analysis underscores this point: only Pao- showed a significant monitor by sample time interaction. This implies that a given monitor Pao- value bears a different relationship to its corresponding laboratory value depending on the time at which the sample was obtained. Pao- sensor lag probably accounts for this significant time interaction. As Pao, rises abruptly, the sensor lag becomes evident (see Figs. 6 and 7). Rather than the typical overestimation of laboratory Pao-, the monitor sensor underestimated Pao- at sample time 6. The sensor was trying to catch up at this point and finally did so at sample time 7. Our experimental design did not create a rapid decline in Pao-; a similar lag problem would be anticipated in this instance. However, the Pao- sensor provided an excellent indicator of change, both in direction and magnitude. We emphasize that whatever sensor lag may exist, the delay must be far less than that associated with the most efficient sampling and laboratory analysis of blood gases.
Laboratory analysis as the reference method. By necessity, our clinical investigation relied on laboratory measurement of blood gas values as the accepted reference method. Whereas in vitro experiments have shown excellent performance of the continuous blood gas monitor, in vivocomparison of monitor values with those from a laboratory blood gas analyzer show less agreement between methods."? In an attempt to eliminate laboratory error as an apparent cause of poor in vivo blood gas monitor performance, we performed all laboratory measurements according to a rigorous protocol. One experienced technician performed all measurements, on the same blood gas analyzer, located in the operating room, thereby eliminating major delays in analysis. Controls showed excellent reproducibility, confirming the performance of the blood gas analyzer. Nonetheless, potential problems remain with any in vitro technique. We made efforts to assure anaerobic
The Journal of Thoracic and Cardiovascular Surgery
blood sampling and prompt laboratory blood gas measurements, yet unrecognized air bubbles contaminating samples, delays in laboratory testing, or use of polypropylene syringes may have altered blood gas values. W- 13 Each of these factors would cause a drift of high Pao, values down toward ambient atmospheric gas tensions. As a consequence, our laboratory technique may have underestimated true Pao- values in the range created on cardiopulmonary bypass in this study (mean laboratory Pao, 307 mm Hg, range up to 495 mm Hg). Monitor Pao- bias evident from our data should not be interpreted as a shortcoming of the monitoring device; rather, the disagreement between methods may reflect the inherent limitation of laboratory methods for accurate measurement of high Pao, values. Two additional factors distinguish monitor and laboratory methods for blood gas determinations. First, the calibration range for POz is more limited in the laboratory method (laboratory 0% to 12%;monitor 5.5% to 30%). The expanded calibration range used for the monitor might be better suited to accurate Pao- determination during bypass and less well suited to Pvo, monitoring. Second, monitor blood gas values were measured at actual patient temperature and corrected by algorithm to 37° C values. On the other hand, laboratory blood gas values were both measured and reported at 37° C. Errors in temperature measurement and limitations of algorithm accuracy would confound comparison of monitor and laboratory values. All of these issues underscore thedifficulty in performing a valid clinical comparison between a continuous, on-line monitoring technique and an intermittent, remote laboratory measurement reference method. In summary, our results suggest that Pco, and pH monitoring approach clinical standards set by the blood gas laboratory. Improvements in POz sensor technology, both lag time and accuracy in the low POz range, will enhance clinical acceptability of the optical fluorescence monitoring technique. * We gratefullyacknowledge the criticalsuggestions provided
by Drs. Leroy Vandam and Daniel Raemer, and the help in manuscript preparation by Ms. Judy Wyman. REFERENCES I. Stanley TH, Isern-Amaral J. Periodic analysis of mixed venous oxygen tension to monitor the adequacy of perfusion *A new version of the COl 300 (COl 400) has been developed since we performed our study. It has a new P02 sensor with performance characteristics designed to overcome the limita tions described in this investigation.
Volume 102 Number 3 September 1991
2.
3.
4.
5.
6.
7.
during and after cardiopulmonary bypass. Can Anaesth Soc J 1974;21:454-60. Gehrich JL, Lubbers DW, Opitz N, et al. Optical fluorescence and its application to an intravascular blood gas monitoring system. IEEE Trans Biomed Eng 1986;BME33:117-32. Miller WW, Gehrich JL, Hansmann DR, Yafuso M. Continuous in vivo monitoring of blood gases. Lab Med 1988;19:629-35. Pino JA, Bashein G, Kenny MA. In vitro assessment of a flow-through fluorometric blood gas monitor. J Clin Monit 1988;4:186-94. Alston RP, Trew A. An in vitro assessment of a monitor for continuous inline measurement of Paz, Pco- and pH during cardiopulmonary bypass. Perfusion 1987;2:13947. Bashein G, Pino JA, Nessly ML, et al. Clinical assessment of a flow-through fluorometric blood gas monitor. J Clin Monit 1988;4:195-203. Alston RP. A clinical evaluation of a monitor for inline
Continuous monitoring duringCPB 439
8.
9.
10.
11.
12.
13.
measurement of Paz, Pcoz and pH during cardiopulmonary bypass. Perfusion 1988;3:225-31. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307-10. Barker SJ, Tremper KK, Hyatt J, et al. Continuous fiberoptic arterial oxygen tension measurements in dogs. J Clin Monit 1987;3:48-52. Harsten A, Berg B, Inerot S, Muth L. Importance of correct handling of samples for the results of blood gas analysis. Acta Anaesth Scand 1988;32:365-8. Scott PV, Horton JM, Mapleson WW. Leakage of oxygen from blood and water samples stored in plastic and glass syringes. Br Med J 1971;3:512-6. Biswas CK, Ramos JM, Agroyannis B, Kerr DNS. Blood gas analysis: effect of air bubbles in syringe and delay in estimation. Br Med J 1982;284:923-7. Ishikawa S, Fornier A, Borst C, Segal MS. The effects of air bubbles and time delay on blood gas analysis. Ann Allergy 1974;33:72-7.
Continuous arterial and venous blood gas monitoring during cardiopulmonary bypass A new monitoring technique, based on optical fluorescence chemistry, aUows continuous monitoring of all blood gas variables during cardiopulmonary bypass. To evaluate the clinical performance of this monitor, we drew 220 arterial and 216 venous blood samples from 15 patients, and simultaneous blood gas values displayed by the monitor were compared with standard laboratory measurements. The continuous monitor predicted laboratory values with varying degrees of accuracy. (R2 values by linear regression: arterial oxygen tension 0.86, venous oxygen tension 0.36, arterial carbon dioxide tension 0.58, venous carbon dioxide tension 0.72, arterial pH 0.53, venous pH 0.58; p < 0.0001). Monitor values of arterial oxygen tension overestimated laboratory values (bias = + 43.5 mm Hg), but the laboratory reference method likely underestimated true arterial oxygen tension in the high range achieved on bypass. Monitoring of venous oxygen tension was imprecise (precision = ± 6.51 mm Hg), regardless of whether stable conditions existed during the sampling period. Monitoring of carbon dioxide tension and pH showed small bias (carbon dioxide tension within 2 mm Hg, pH within 0.03) and good precision (carbon dioxide tension within 3 mm Hg, pH within 0.03). With the development of unstable conditions on bypass, monitor arterial oxygen tension values showed a changing relationship to corresponding laboratory values. In conclusion, arterial and venous carbon dioxide tension and pH monitoring provide acceptably accurate alternatives to laboratory measurement of these variables during cardiopulmonary bypass. Arterial oxygen tension monitoring accurately indicates changes in oxygen tension in the arterial oxygen tension range typically produced during extracorporeal circulation. Oxygen tension monitoring in the venous oxygen tension range is too imprecise for clinical decision-making purposes.
Jonathan B. Mark, MD,a, b Daniel FitzGerald, CCP,a, b Terry Fenton, EdD,b,c Anna Mae Fosberg, RN, ccp,a, bWilliam Camann, MD,a, b Nick Maffeo, BA,a and James Winkelman, MD,a, b Boston. Mass.
Intermittent measurements ofblood pH, oxygen tension (P02), and carbon dioxide tension (PC02) are standard means for assessing physiologic status during cardiopulmonary bypass. In simplest terms, arterial blood gas valFrom Brigham and Women's Hospital,' Harvard Medical School," and Harvard School of Public Health," Boston, Mass. Presented in part at the Eleventh Annual Meeting of the Society of Cardiovascular Anesthesiologists, Seattle, Wash., April 16-19, 1989. Supported in part by a grant from Cardiovascular Devices, Inc., Irvine, Calif. Received for publication Sept. 25, 1989. Accepted for publication April 3, 1990. Address for reprints: Jonathan B. Mark, MD, Department of Anesthesia, Brigham and Women's Hospital, 75 Francis St., Boston, MA 02115.
12/1/21518
ues reflect pump-oxygenator performance, whereas venous blood gases indicate the adequacy of oxygen delivery to the body tissues.' If all these variables were monitored continuously, rather than measured intermittently in the laboratory, benefits might include earlier detection of deleterious trends in pump-oxygenator performance or a patient's condition, avoidance of phlebotomy, and reduction of risks to health care workers from exposure to phlebotomized blood. Developments in optical fluorescence chemistry have provided a suitable technology for continuous arterial and venous blood gas monitoring during cardiopulmonary bypass.s 3 Previous investigations in vitro': 5 and in vivo'" 7 have concluded that this kind of blood gas monitoring technique adequately follows blood gas trends but is not accurate enough to supplant traditional laboratory analyses. However, those studies employed an early version of
431
The Journal of Thoracic and Cardiovascular
4 32Mark et al.
Surgery
,----- HYPOTHERMIA ------, EVENTS
ON CPS
29°C
I
SAMPLE TIME
+5 +7 +9 +11 min I I I I 1 2 3 4
I
rl- - - - REWARM - - - - - - ,
1 Q ,V' 102% r r y,
I
5
Fig. 1. Blood sampling protocol. CPR, Cardiopulmonary bypass; concentration.
° +2 +4 +6 30C min I I I I 6 7 8 9
I
+20
l/
I 16
OFF CPS
I
12, pump blood flow; V. gas flow; 02%.
oxygen
PVC02 (/)
I.LJ -l
o, :E
u,
0
a::
W ID
:E ::::>
Z
70 60 50 40 30 20 10
-8
-4
0
4
8
-20 -16 -12 -8
-4
0
4
8
12 16
20
DIFFERENCE (mmHg) (Monitor - Lab) Fig. 2. Histograms showing differences (monitor-laboratory) for Pve02 and Pvo-.
the instrument system (GAS-STAT, Cardiovascular Devices, Inc., Irvine, Calif) and focused on arterial samples; improved sensors and algorithms incorporated in current devices have not been evaluated. Furthermore, neither the original nor the current monitoring system has been validated over a wide range of arterial and venous blood gas values. Finally, monitor performance has not been tested during well-defined stable and unstable clinical conditions. Our investigation was conducted with the newest generation optical fluorescence blood gas monitor (COl 300, Cardiovascular Devices, Inc. Irvine, Calif.). We posed three clinical questions: (1) What are the trending capabilities for Po 2, Pco-, and pH sensors in both arterial and venous blood, and what is the overall performance of the continuous blood gas monitor during cardiopulmonary bypass? (2) During stable bypass conditions, how accurate is the blood gas monitor in predicting simultaneous laboratory values? (3) What happens to the relationship between monitor and laboratory blood gas values during unstable, rapidly changing conditions?
Patients and methods Patients. Fifteen patients undergoing cardiac operations requiring cardiopulmonary bypass were studied. In each case, informed written consent was obtained in accordance with standards set by the Hospital's Human Research Committee. The average patient was 66 years of age, weighed 78 kg, and had a bypass time of 82 minutes and an aortic crossclamp time of 54 minutes. Anesthetic drugs included opioids (fentanyl, sufentanil, morphine), and volatile agents (enflurane, isoflurane), all of which have been shown not to interfere with fluorescence blood gas monitoring." Cardiopulmonary bypass. Cardiopulmonary bypass was conducted with a nonpulsatile roller pump, a hollow-fiber membrane oxygenator, and a 20 lim arterial filter. The bypass circuit was primed with lactated Ringer's solution and beeflung heparin. Patient temperature was monitored with a nasopharyngeal thermistor, and arterial and venous blood temperatures were monitored with thermistors incorporated in the continuous blood gas monitor sensors. In each patient a period of stable hypothermic cardiopulmonary bypass was established, followed by a period of rapid rewarming to normothermia. Pump blood flow was 1.1 L· min"! . m- 2 during hypothermia, and increased to 2.0 L . min-I . m- 2 during rewarming. Gas flow rates were 2.0 L . min-I during hypothermia and 3.0 L . min"! during rewarming. Oxygen concentration was maintained at
Volume 102 Number 3
Continuous monitoring duringCPB 433
September 1991
60% during hypothermia and increased to 100% during rewarming. Supplemental carbon dioxide was not added to the sweepgases. Pump blood flow,gas flow, and oxygen concentration were held constant during hypothermia and adjusted in a stepwisefashion commencing with rewarming. This allowed the bypassperiod to be divided into two phases: stable hypothermia and unstable rapid rewarming. Blood gas monitoring. Before each case the continuous bloodgas monitor was calibrated in vitro according to the manufacturer's directions, with standard reference sensors, buffer solution,and two standard calibrating gases (2.8% carbon dioxide, 5.5% oxygen, 91.7% nitrogen; 8.0% carbon dioxide, 30.0% oxygen, 62.0% nitrogen). All sensors, buffers, and standard gases were COl brand. All 30 arterial and venous sensors were successfully calibrated. The arterial sensor was placed in the arterial line 47 inches beyond the oxygenator, and the venous sensor was placed in the venous line 28 inches before the reservoirbag. Dedicated time to perform calibration and incorporate the monitoring system into the extracorporeal circuit averaged 6 minutes. During cardiopulmonary bypass, monitor blood gas values were recorded every 6 seconds on magnetic disk (Toshiba TlOOO, Tustin, Calif.). All blood gas values were measured at actual arterial and venous blood temperatures and corrected to 37° C by algorithms incorporated into the blood gas monitor. Blood sampling. For comparison with monitor values, arterial blood was aspirated from the arterial filter purge line (28 inches before the arterial sensor), and venous blood was withdrawn from the venous return line at its entrance to the reservoir bag (26 inches from the venous sensor). Blood sampling followed a predetermined protocol in each patient (Fig. 1). After stable hypothermic conditions had been established for 5 minutes (constant temperature, pump blood and gas flows, and oxygen concentration), four paired samples (arterial and venous) were drawn at 2-minute intervals, with the simultaneous monitor blood gas values electronically marked for later analysis. A fifth sample was drawn at the conclusion of the hypothermic period; immediately thereafter the patient was actively rewarmed, and step changes were made in pump blood flow, gas flow, and oxygen concentration. When venous blood temperature had increased 1° C from its nadir, sampling again commenced at 2-minute intervals for 20 minutes, during this unstable period. A total of 16 arterial and venous samples were drawn in each patient (five stable hypothermic, 11 unstable rewarming), except when cardiopulmonary bypass was terminated before all sampling could be completed. Laboratory blood gas analysis. Laboratory analyses were performed by the same technician, with a single instrument (Ciba-Corning 168 blood gas analyzer, Medfield, Mass.), located in the operating room. All blood samples were withdrawn anaerobically into 3 ml polypropylene syringes, immediately placed on ice, and analyzed in sampled order. The blood gas analyzer was calibrated on the P02 and PC02 channels with two certified gas cylinders (5% carbon dioxide, 12% oxygen, 83% nitrogen; 10% carbon dioxide, 90% nitrogen), and pH was calibrated with two buffer standards (pH = 7.381; pH = 6.838). All flush solutions, buffers, and standard gases were Ciba-Corning brand. Before the start of the study all routine maintenance was performed on the blood gas analyzer, including changing membranes and rotary port 0 rings. Daily two-point calibration was
Table I. Laboratory bloodgas analyzer controls Variable
N
Mean ± SD
Range
15 15 15
94.4 ± 1.2 127.8 ± 2.4 58.9 ± 1.5
91.6-96.7 125.6-133.1 56.1-60.9
PC02 Alkalosis
15 15 15
43.8 ± 0.6 62.9 ± 0.7 21.8 ± 0.6
42.8-45.0 62.0-64.2 20.6-22.6
pH Normal pH Acidosis pH Alkalosis
15 15 15
7.392 ± 0.006 7.141 ± 0.005 7.630 ± 0.007
P02 Normal P02 Acidosis P02 Alkalosis PC02 Normal PC02 Acidosis
7.380-7.401 7.131-7.149 7.614-7.641
SO, Standard deviation. Descriptive statistics summarizing daily controls run on the Ciba-Corning laboratory blood gas analyzer. Values for P02 and Pco, are in millimeters of mercury.
performed until calibration values were stable and reproducible. (On average, three two-point calibrations were required, which is normal after electrical interruption.) Subsequently three blood gas controls (normal acid-base, respiratory acidosis, and respiratory alkalosis) were run daily to verify instrument performance (Table I). During each study the blood gas analyzer had a one-point calibration performed hourly to check for electrode drift. When drift exceeded ± 0.005 pH units, 1.8 mm Hg P02, or 1.0 mm Hg PC02, calibration was repeated. All laboratory blood gas values were measured and reported at 37°C. Data analysis. Laboratory blood gas values and nasopharyngeal temperatures were recorded manually, and these data were combined with electronically stored, simultaneous blood gas monitor values for analysis with the SAS statistical package (SAS, Inc., Cary, N.C.). Monitor trending capabilities and overall performance were evaluated first by simple linear regression of monitor versus laboratory blood gas values. Next a multiple regression model analyzed the main effects of sample time and individual subjects along with monitor values as predictors of laboratory blood gas values. Subsequently the regression model examined interaction effects of monitor by sample time and monitor by individual subjects. These analyses tested whether the slope of the monitor/laboratory blood gas value relationship differed significantly across sample times or between individual subjects. Overall monitor performance was analyzed further by calculating the mean difference between monitor and laboratory values and the standard deviation of the difference. According to the methods of Bland and Altman'' and Barker and coworkers.? the mean difference between the two measurement techniques was considered to be the bias and the standard deviation of individual scores around that difference the precision of the monitoring instrument. Monitor performance during stable hypothermic conditions also was assessed by calculation of bias and precision for each monitored variable, as indicators of monitor accuracy. Effects of rapid rewarming and abrupt adjustment of blood and gas flows,on each variable, were analyzed by plotting individual and mean monitor and laboratory values across all 16 sampling times and by examining the relationship between monitor and laboratory blood gas values during both stable (samples 1 to 5) and unstable (samples 6 to 16) periods.
The Journal of Thoracic and Cardiovascular Surgery
4 34Mark et al.
Table II. Multiple regression analyses: Monitor, sample time, and individual subjects as predictors for laboratory values Variable Pao2 Pv02
PacoPvC02 pHa pHv
C: Interactions
B: Main effects
A: Monitor only R2
Monitor
0.86 0.36 0.58 0.72 0.53 0.58
0.0001 0.0001 0.0001 0.0001 0.0001 0.0001
Time 0.0001 0.0001 0.0001 0.075 0.0001 0.0006
Subjects
R2
Monitor/time
Monitor/subjects
R2
0.017 0.0001 0.0001 0.0001 0.0001 0.0001
0.91 0.79 0.91 0.92 0.89 0.91
0.003 0.68 0.93 0.99 0.78 0.62
0.0003 0.003 0.15 0.13 0.21 0.25
0.94 0.85 0.92 0.94 0.91 0.93
Multiple regression analyses with monitor blood gas values, sample time, and individual subjects as predictors of laboratory blood gas values. See text for details. All entries are p values and R2 values derived from regression analyses.
Table III. Overall relationship between monitor and laboratory blood gas values Variable Pao2 Pv°2 Pac02 PvC02 pHa pHv
N 220 203 220 203 220 203
Lab value (mean ± SD) 306.8 41.3 33.7 39.1 7.46 7.41
± ± ± ± ± ±
86.1 5.7 4.1 3.9 0.04 0.03
Bias ± Precision
r
± ± ± ± ± ±
0.93 0.60 0.76 0.85 0.73 0.76
+43.5 -1.44 +1.62 + 1.40 -0.025 -0.029
43.3 6.51 2.79 2.14 0.029 0.023
Slope 0.74 0.42 0.81 0.84 0.83 0.76
± ± ± ± ± ±
0.04 0.08 0.09 0.07 0.10 0.09
Intercept 48.7 24.48 5.22 5.02
± ± ± ±
14.7 3.14 3.25 2.97
SD, Standard deviation. Slopes and intercepts of the regression equations are given ± 95% confidence intervals. Values for P0 2 and PC02 are in millimeters of mercury. All r values (correlation coefficients derived from regression of monitor versus laboratory values) are significant (p < 0.0001). Intercepts for pH are not reported since these values lack physiologic significance.
Results Blood gas comparisons were performed on 220 arterial and 216 venous blood samples, with the use of 30 monitoring sensors in 15 patients. Twenty arterial and 24 venous samples were unavailable for analysis owing to early completion of bypass, arterial contamination of sampled venous blood, or loss of electronically stored monitor data. From a clinical standpoint all sensors appeared to function well, except for a single venous sensor. Venous Pco- (Pvco-) and venous pH (pHv) values were clearly erroneous in one patient, with Pvco- > 67 mm Hg and pHv < 7.21. This venous sensor was considered to have failed (sensor failure rate 3.3%), and data collected from this venous sensor were not included in further analyses. Trending and overall performance. Table II summarizes results of multiple regression analyses. The continuous blood gas device was capable of predicting laboratory blood gas values for each monitored variable with varying degrees of accuracy. Regression analyses revealed R 2 values of 0.36 and 0.86 for Pvo- and Pao-, respectively, with R 2 values for Paco-, Pvco-, pHa, and pHv being intermediate. All R 2 values were highly significant (p < 0.0001). When the effects of sample times and individual subjects were taken into account, the main effects of these variables significantly improved on
the accuracy of prediction of laboratory blood gas values, over and above the level of accuracy achieved by the monitor alone (Table II, B). (The single exception was Pvco-, where sample time was a marginally significant factor [p = 0.075].) A significant monitor value by sample time interaction was present for Pao, (Table II, C), indicating that the slope of the monitor/laboratory blood gas value relationship varied significantly over sample times. Furthermore, significant monitor by individual subject interactions were present for both Pao- and Pvo(Table II, C), indicating that the slopes of the monitor/ laboratory blood gas value relationships varied significantly across subjects for these two variables. Overall performance of the continuous blood gas monitor is summarized in Table III. Monitored values of Pao, overestimated laboratory values and Pvo- monitoring was imprecise. For each of the other monitored variables, bias was small and precision good. The imprecision of Pvo z monitoring is highlighted in Fig. 2, where Pvo- is compared with Pvco- monitoring. This comparison seemed appropriate, because these two venous blood gas variables were monitored within the same sensor group, and the distributions of values monitored covered similar physiologic ranges (Pvo- = 41 mm Hg mean, 30 mm Hg to 62 mm Hg range; Pvco- = 39 mm Hg mean, 33 mm Hg to 50 mm Hg range).
Volume 102 Number 3 September 1991
Continuous monitoring during CPS
435
40 u 'L W
a::
35f-
::> ~
a::
w a..
~ I.JJ
301-
~
25
~
2
3 4
5 6
7 8
9 10 11 12 13 14 15 16
SAMPLE TIME Fig. 3. Mean nasopharyngeal (0), arterial blood (.), and venous blood (~) temperatures at each sample time.
70 = 0.44 p= 012
r
n: 14
400 I
(\J
E
300
(\J
o
50
>
•
•
a..
a..
CD
CD
j
Slope: 022 ± 025 Intercept: 343± 11.5
E
E
E
oa
60
-a. I
0-
r ;0.99 p
200
•
...J
40
•
Intercept = 23.8± 14.2
200
300
400
40
MONITOR Po0 2 (mmHg)
50
60
70
MONITOR PV02 (mmHg)
Fig. 4. Scatterplot relating monitor Pao- versus laboratory Pao- at stable sample time 4. Slope and intercept of the regression equation are reported ± 95% confidence intervals. The line of identity is indicated.
Fig. 5. Scatterplot relating monitor Pvoz versus laboratory Pvo- at stable sample time 4. Slope and intercept of the regression equation are reported ± 95% confidence intervals. The line of identity is indicated.
Substantial differences between monitor Pvo- and laboratory Pvo- were common: monitor values were within 3 mm Hg of laboratory values 32.0% of the time; within 5 mm Hg 54.2% of the time; and within 10 mm Hg only 91.1% of the time. By contrast, monitor Pvco- was within 3 mm Hg of laboratory values 70.4% of the time and within 5 mm Hg 99.0% of the time. Accuracy during stable conditions. Stable hypothermic conditions were readily achieved during cardiopulmonary bypass. At sample time 4, nasopharyngeal, arterial blood; and venous blood temperatures were within
0.6 C of one another, and had drifted by less than 0,70 C over the previous 6 minutes (Fig. 3). Cardiopulmonary bypass blood flow,gas flow,and oxygen concentration had been constant during the same interval. During this clinically stable period, monitored blood gas values most closely predicted laboratory values (Table IV), but significant differences between techniques remained. Laboratory Pao- was still overestimated by the monitor. In addition, during stable conditions, Pvo- monitoring continued to be imprecise. Although mean bias of Pvo- was not statistically significant, the monitor frequently gross0
The Journal of Thoracic and Cardiovascular
4 36Mark et al.
Surgery
500 Ol
I
E E c
0
eu
E
0
400 300
t\I
0
a..
• LAB
200 01
o
2
4
6
8
10
12
MONITOR
14
16
SAMPLE TIME Fig. 6. Mean laboratory and monitor values for Pao- at each sample time in eight patients who had a single-step from 60% to 100%. increase in inspired oxygen concentration
en
lyoverestimatedor underestimated laboratory Pvo-. This pointisgraphically highlightedby examiningscatterplots of monitor versus laboratory valuesfor Pao, and Pv02at sample time 4 (Figs. 4 and 5). Monitored valuesof Paocorrelate highly with laboratory values (r = 0.99, P < 0.0001), whereas there is only a weak association (r = 0.44, P = 0.12) between monitor and laboratory Pvo- values during this same period. Performance during unstable conditions. Clinically unstable conditions were created with the onset of rapid rewarming during bypass. All monitored temperatures increased abruptly (see Fig. 3), as extracorporeal blood and gas flows were adjusted to new target values, and oxygenconcentration was increasedin each patient. During the rewarming period,valuesfor Pao, nearly doubled (Fig. 6), whereas values for other variables were altered little. When monitor and laboratory bloodgasvalueswere plotted acrossall 16sample times,there was no consistent pattern of perturbation of the monitor versus laboratory relationship for any variable except Pao-, The relationship between monitor Pao2 and laboratory Pao-, however, changed dramatically at the onset of the unstable period. In eight subjects who had a single-step change in inspired oxygen concentration from 60% to 100% between sample times 5 and 6, monitor Pao- overestimated laboratory Pao- by + I 0 mm Hg at sample time 5, underestimated by -18 mm Hg at sample time 6, and again overestimatedby +50 mm Hg at sample time 7 (see Fig. 6). The variability of the Pao- relationship was equally evident when data from all 15 subjectswere plotted. A particularly striking example of this phenomenonis illustrated by the data from one patient, where exact
monitor/laboratory agreement occurred by the end of the stable hypothermic period (Fig. 7, sample time 5). As Pao, rose rapidly during rewarming, the monitor sensor transiently but dramatically underestimated the simultaneouslaboratory valueby 74 mm Hg (Fig. 7, sampletime 6). The typical monitor/laboratory relationship returned at sample time 7 and was maintained for the duration of the cardiopulmonary bypass period. Discussion Trending and overall performance. Continuousblood gas monitoringprovides reliabletrending informationfor detecting changes in blood PC02, pH, and Pao-; changes in Pvo, values are least adequately predicted by this monitoring technique. Improved predictions of laboratory blood gas values wereobtained when individual subjectsand sampletimes were considered, over and above predictions provided by monitor blood gas values. The amount of incremental variance explained by subject and time effectswas modest for Pao-, but considerably larger for the other variables,approaching 43% for Pv02. This indicatesthat there were systematiccomponentsof variation in laboratory values associated with sample times and individual subjects, which were not reflected by monitor values. Although these analyses suggest there is room for improvement in the accuracy of the monitor, is the prediction providedby the current monitor good enough? A technique may be sufficient to supplant laboratory measurements for clinical purposes, but not for research requirements. Moreover, a monitoring device may be adequate for use as a supplement to intermittent laboratory measurements, while not entirely substituting for
Volume 102
Continuous monitoring during CPR 43 7
Number 3 September 1991
500
~
0'
I
E E
0
N
400 300
c
0• LAB
200
01
o
2
4
6
8
10
12
MONITOR
14
16
SAMPLE TIME Fig. 7. Laboratory and monitor values for Pao , at each sample time in one patient from the group shown in Fig. 6.
these measurements. These questions are best answered by examining each blood gas variable separately. Differences between monitor and laboratory values for Pco- were small, an overestimation bias of approximately 1.5 mm Hg. Furthermore, precision of Pco- monitoring was good, averaging ± 2.5 mm Hg. This degree of accuracy might be acceptable for clinical monitoring but unacceptable for research measurements. Monitoring of pHa and pHv like Pco-, might be considered a reasonable clinical substitute for laboratory measurement. Ninetyfivepercent of the time, monitor pH values were no more than 0.08 units lower or 0.03 units higher than the corresponding laboratory pH value. Unfortunately, monitoring of P02 was more problematic. Pao- and Pvo- values spanned entirely different, nonoverlapping ranges. Unique monitoring problems were discovered specific to these ranges. However, problems identified for Pvomonitoring, are clearly the same as would exist during Pao- monitoring in hypoxemic patients. Monitor values of Pao- overestimated laboratory values. This disagreement is mitigated by a number of factors. First, selection of laboratory values as accurate reference standards for Pao- in the high range may be questioned. Second, as a percentage of the laboratory value, the bias was lessthan 15%,a disagreement of minor clinical importance in this high P0 2 range. Third, Pao, bias tends to increase as Pao- rises. As illustrated in Fig. 4, monitor values agree well with laboratory values when P0 2 is less than 200 mm Hg, and show increasing overestimation as P0 2 rises. Pvo 2 trending and overall monitoring were the most unsatisfactory in this study. R2 values from all regression analyses were lowest for Pv02. While overall bias for Pvo-
Table IV. Relationship between monitor and
laboratory bloodgas values during stable conditions Variable
N
Lab value (Mean ± SD)
Pao2 Pv02 PaC02 PvC02 pHa pHv
15 14 15 14 15 14
220.0 44.1 36.9 41.8 7.43 7.39
± ± ± ± ± ±
62.0 4.3 3.8 4.4 0.04 0.04
Bias ± precision +24.3 + 1.01 +0.82 +0.85 -0.017 -0.029
± ± ± ± ± ±
16.8 7.91 2.94 2.18 0.025 0.022
SD, Standard deviation. Values for P02 and Pco- arc in millimeters of mercury.
was small, this became less important in view of the poor precision of Pvo- monitoring. In 95% of comparisons, monitor Pvo, values underestimated laboratory values by as much as 14.5 mm Hg and overestimated by as much as 11.5 mm Hg. Consequently, neither changes in Pv02 nor absolute values of this monitored variable can be relied on for clinical management purposes. Accuracy during stable conditions. By creating both stable and unstable clinical conditions during cardiopulmortary bypass, we hoped to learn why blood gas values from the monitor and laboratory disagreed. The continuous blood gas monitor sensors have an inherent delay in responding to changes in blood gas tensions.l-" Blood mixing, membrane gas and ion permeability, and chemical reactions within the fluorescent dyes all contribute to this lag time. By examining data from stable sample time 4, we attempted to eliminate any monitor inaccuracy attributable to sensor lag. Similar to the results from the overall data sample, those from the stable period show little bias and adequate precision for Pco- and pH monitoring. Again, monitor Pao- significantly overestimated
4 38Mark et al.
laboratory values, but this disagreement was reduced from 15% to II %. Even during the most stable interval, Pvo, monitoring was imprecise and a clinically inadequate predictor oflaboratory Pvo-, The same patterns of monitor performance were evident from both the stable data and the overall set. Performance during unstable conditions. Although temperature and Pao- changed rapidly during rapid rewarming, none of the other blood gas variables were consistently perturbed. Apparently alterations in perfusion and ventilation matched patient requirements, thus obviating dramatic changes in Pvoz, Pco-, and pH. In the absence of such changes, the relationship between monitor and laboratory values for these blood gas variables remained relatively constant across time. The interaction term from the multiple linear regression analysis underscores this point: only Pao- showed a significant monitor by sample time interaction. This implies that a given monitor Pao- value bears a different relationship to its corresponding laboratory value depending on the time at which the sample was obtained. Pao- sensor lag probably accounts for this significant time interaction. As Pao, rises abruptly, the sensor lag becomes evident (see Figs. 6 and 7). Rather than the typical overestimation of laboratory Pao-, the monitor sensor underestimated Pao- at sample time 6. The sensor was trying to catch up at this point and finally did so at sample time 7. Our experimental design did not create a rapid decline in Pao-; a similar lag problem would be anticipated in this instance. However, the Pao- sensor provided an excellent indicator of change, both in direction and magnitude. We emphasize that whatever sensor lag may exist, the delay must be far less than that associated with the most efficient sampling and laboratory analysis of blood gases.
Laboratory analysis as the reference method. By necessity, our clinical investigation relied on laboratory measurement of blood gas values as the accepted reference method. Whereas in vitro experiments have shown excellent performance of the continuous blood gas monitor, in vivocomparison of monitor values with those from a laboratory blood gas analyzer show less agreement between methods."? In an attempt to eliminate laboratory error as an apparent cause of poor in vivo blood gas monitor performance, we performed all laboratory measurements according to a rigorous protocol. One experienced technician performed all measurements, on the same blood gas analyzer, located in the operating room, thereby eliminating major delays in analysis. Controls showed excellent reproducibility, confirming the performance of the blood gas analyzer. Nonetheless, potential problems remain with any in vitro technique. We made efforts to assure anaerobic
The Journal of Thoracic and Cardiovascular Surgery
blood sampling and prompt laboratory blood gas measurements, yet unrecognized air bubbles contaminating samples, delays in laboratory testing, or use of polypropylene syringes may have altered blood gas values. W- 13 Each of these factors would cause a drift of high Pao, values down toward ambient atmospheric gas tensions. As a consequence, our laboratory technique may have underestimated true Pao- values in the range created on cardiopulmonary bypass in this study (mean laboratory Pao, 307 mm Hg, range up to 495 mm Hg). Monitor Pao- bias evident from our data should not be interpreted as a shortcoming of the monitoring device; rather, the disagreement between methods may reflect the inherent limitation of laboratory methods for accurate measurement of high Pao, values. Two additional factors distinguish monitor and laboratory methods for blood gas determinations. First, the calibration range for POz is more limited in the laboratory method (laboratory 0% to 12%;monitor 5.5% to 30%). The expanded calibration range used for the monitor might be better suited to accurate Pao- determination during bypass and less well suited to Pvo, monitoring. Second, monitor blood gas values were measured at actual patient temperature and corrected by algorithm to 37° C values. On the other hand, laboratory blood gas values were both measured and reported at 37° C. Errors in temperature measurement and limitations of algorithm accuracy would confound comparison of monitor and laboratory values. All of these issues underscore thedifficulty in performing a valid clinical comparison between a continuous, on-line monitoring technique and an intermittent, remote laboratory measurement reference method. In summary, our results suggest that Pco, and pH monitoring approach clinical standards set by the blood gas laboratory. Improvements in POz sensor technology, both lag time and accuracy in the low POz range, will enhance clinical acceptability of the optical fluorescence monitoring technique. * We gratefullyacknowledge the criticalsuggestions provided
by Drs. Leroy Vandam and Daniel Raemer, and the help in manuscript preparation by Ms. Judy Wyman. REFERENCES I. Stanley TH, Isern-Amaral J. Periodic analysis of mixed venous oxygen tension to monitor the adequacy of perfusion *A new version of the COl 300 (COl 400) has been developed since we performed our study. It has a new P02 sensor with performance characteristics designed to overcome the limita tions described in this investigation.
Volume 102 Number 3 September 1991
2.
3.
4.
5.
6.
7.
during and after cardiopulmonary bypass. Can Anaesth Soc J 1974;21:454-60. Gehrich JL, Lubbers DW, Opitz N, et al. Optical fluorescence and its application to an intravascular blood gas monitoring system. IEEE Trans Biomed Eng 1986;BME33:117-32. Miller WW, Gehrich JL, Hansmann DR, Yafuso M. Continuous in vivo monitoring of blood gases. Lab Med 1988;19:629-35. Pino JA, Bashein G, Kenny MA. In vitro assessment of a flow-through fluorometric blood gas monitor. J Clin Monit 1988;4:186-94. Alston RP, Trew A. An in vitro assessment of a monitor for continuous inline measurement of Paz, Pco- and pH during cardiopulmonary bypass. Perfusion 1987;2:13947. Bashein G, Pino JA, Nessly ML, et al. Clinical assessment of a flow-through fluorometric blood gas monitor. J Clin Monit 1988;4:195-203. Alston RP. A clinical evaluation of a monitor for inline
Continuous monitoring duringCPB 439
8.
9.
10.
11.
12.
13.
measurement of Paz, Pcoz and pH during cardiopulmonary bypass. Perfusion 1988;3:225-31. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307-10. Barker SJ, Tremper KK, Hyatt J, et al. Continuous fiberoptic arterial oxygen tension measurements in dogs. J Clin Monit 1987;3:48-52. Harsten A, Berg B, Inerot S, Muth L. Importance of correct handling of samples for the results of blood gas analysis. Acta Anaesth Scand 1988;32:365-8. Scott PV, Horton JM, Mapleson WW. Leakage of oxygen from blood and water samples stored in plastic and glass syringes. Br Med J 1971;3:512-6. Biswas CK, Ramos JM, Agroyannis B, Kerr DNS. Blood gas analysis: effect of air bubbles in syringe and delay in estimation. Br Med J 1982;284:923-7. Ishikawa S, Fornier A, Borst C, Segal MS. The effects of air bubbles and time delay on blood gas analysis. Ann Allergy 1974;33:72-7.